diff options
Diffstat (limited to 'drivers/infiniband/core')
72 files changed, 56835 insertions, 0 deletions
diff --git a/drivers/infiniband/core/Makefile b/drivers/infiniband/core/Makefile new file mode 100644 index 000000000..ccf2670ef --- /dev/null +++ b/drivers/infiniband/core/Makefile @@ -0,0 +1,44 @@ +# SPDX-License-Identifier: GPL-2.0 +infiniband-$(CONFIG_INFINIBAND_ADDR_TRANS) := rdma_cm.o +user_access-$(CONFIG_INFINIBAND_ADDR_TRANS) := rdma_ucm.o + +obj-$(CONFIG_INFINIBAND) += ib_core.o ib_cm.o iw_cm.o \ + $(infiniband-y) +obj-$(CONFIG_INFINIBAND_USER_MAD) += ib_umad.o +obj-$(CONFIG_INFINIBAND_USER_ACCESS) += ib_uverbs.o $(user_access-y) + +ib_core-y := packer.o ud_header.o verbs.o cq.o rw.o sysfs.o \ + device.o cache.o netlink.o \ + roce_gid_mgmt.o mr_pool.o addr.o sa_query.o \ + multicast.o mad.o smi.o agent.o mad_rmpp.o \ + nldev.o restrack.o counters.o ib_core_uverbs.o \ + trace.o lag.o + +ib_core-$(CONFIG_SECURITY_INFINIBAND) += security.o +ib_core-$(CONFIG_CGROUP_RDMA) += cgroup.o + +ib_cm-y := cm.o cm_trace.o + +iw_cm-y := iwcm.o iwpm_util.o iwpm_msg.o + +CFLAGS_cma_trace.o += -I$(src) +rdma_cm-y := cma.o cma_trace.o + +rdma_cm-$(CONFIG_INFINIBAND_ADDR_TRANS_CONFIGFS) += cma_configfs.o + +rdma_ucm-y := ucma.o + +ib_umad-y := user_mad.o + +ib_uverbs-y := uverbs_main.o uverbs_cmd.o uverbs_marshall.o \ + rdma_core.o uverbs_std_types.o uverbs_ioctl.o \ + uverbs_std_types_cq.o \ + uverbs_std_types_flow_action.o uverbs_std_types_dm.o \ + uverbs_std_types_mr.o uverbs_std_types_counters.o \ + uverbs_uapi.o uverbs_std_types_device.o \ + uverbs_std_types_async_fd.o \ + uverbs_std_types_srq.o \ + uverbs_std_types_wq.o \ + uverbs_std_types_qp.o +ib_uverbs-$(CONFIG_INFINIBAND_USER_MEM) += umem.o +ib_uverbs-$(CONFIG_INFINIBAND_ON_DEMAND_PAGING) += umem_odp.o diff --git a/drivers/infiniband/core/addr.c b/drivers/infiniband/core/addr.c new file mode 100644 index 000000000..65e3e7df8 --- /dev/null +++ b/drivers/infiniband/core/addr.c @@ -0,0 +1,890 @@ +/* + * Copyright (c) 2005 Voltaire Inc. All rights reserved. + * Copyright (c) 2002-2005, Network Appliance, Inc. All rights reserved. + * Copyright (c) 1999-2005, Mellanox Technologies, Inc. All rights reserved. + * Copyright (c) 2005 Intel Corporation. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include <linux/mutex.h> +#include <linux/inetdevice.h> +#include <linux/slab.h> +#include <linux/workqueue.h> +#include <linux/module.h> +#include <net/arp.h> +#include <net/neighbour.h> +#include <net/route.h> +#include <net/netevent.h> +#include <net/ipv6_stubs.h> +#include <net/ip6_route.h> +#include <rdma/ib_addr.h> +#include <rdma/ib_cache.h> +#include <rdma/ib_sa.h> +#include <rdma/ib.h> +#include <rdma/rdma_netlink.h> +#include <net/netlink.h> + +#include "core_priv.h" + +struct addr_req { + struct list_head list; + struct sockaddr_storage src_addr; + struct sockaddr_storage dst_addr; + struct rdma_dev_addr *addr; + void *context; + void (*callback)(int status, struct sockaddr *src_addr, + struct rdma_dev_addr *addr, void *context); + unsigned long timeout; + struct delayed_work work; + bool resolve_by_gid_attr; /* Consider gid attr in resolve phase */ + int status; + u32 seq; +}; + +static atomic_t ib_nl_addr_request_seq = ATOMIC_INIT(0); + +static DEFINE_SPINLOCK(lock); +static LIST_HEAD(req_list); +static struct workqueue_struct *addr_wq; + +static const struct nla_policy ib_nl_addr_policy[LS_NLA_TYPE_MAX] = { + [LS_NLA_TYPE_DGID] = {.type = NLA_BINARY, + .len = sizeof(struct rdma_nla_ls_gid), + .validation_type = NLA_VALIDATE_MIN, + .min = sizeof(struct rdma_nla_ls_gid)}, +}; + +static inline bool ib_nl_is_good_ip_resp(const struct nlmsghdr *nlh) +{ + struct nlattr *tb[LS_NLA_TYPE_MAX] = {}; + int ret; + + if (nlh->nlmsg_flags & RDMA_NL_LS_F_ERR) + return false; + + ret = nla_parse_deprecated(tb, LS_NLA_TYPE_MAX - 1, nlmsg_data(nlh), + nlmsg_len(nlh), ib_nl_addr_policy, NULL); + if (ret) + return false; + + return true; +} + +static void ib_nl_process_good_ip_rsep(const struct nlmsghdr *nlh) +{ + const struct nlattr *head, *curr; + union ib_gid gid; + struct addr_req *req; + int len, rem; + int found = 0; + + head = (const struct nlattr *)nlmsg_data(nlh); + len = nlmsg_len(nlh); + + nla_for_each_attr(curr, head, len, rem) { + if (curr->nla_type == LS_NLA_TYPE_DGID) + memcpy(&gid, nla_data(curr), nla_len(curr)); + } + + spin_lock_bh(&lock); + list_for_each_entry(req, &req_list, list) { + if (nlh->nlmsg_seq != req->seq) + continue; + /* We set the DGID part, the rest was set earlier */ + rdma_addr_set_dgid(req->addr, &gid); + req->status = 0; + found = 1; + break; + } + spin_unlock_bh(&lock); + + if (!found) + pr_info("Couldn't find request waiting for DGID: %pI6\n", + &gid); +} + +int ib_nl_handle_ip_res_resp(struct sk_buff *skb, + struct nlmsghdr *nlh, + struct netlink_ext_ack *extack) +{ + if ((nlh->nlmsg_flags & NLM_F_REQUEST) || + !(NETLINK_CB(skb).sk)) + return -EPERM; + + if (ib_nl_is_good_ip_resp(nlh)) + ib_nl_process_good_ip_rsep(nlh); + + return 0; +} + +static int ib_nl_ip_send_msg(struct rdma_dev_addr *dev_addr, + const void *daddr, + u32 seq, u16 family) +{ + struct sk_buff *skb = NULL; + struct nlmsghdr *nlh; + struct rdma_ls_ip_resolve_header *header; + void *data; + size_t size; + int attrtype; + int len; + + if (family == AF_INET) { + size = sizeof(struct in_addr); + attrtype = RDMA_NLA_F_MANDATORY | LS_NLA_TYPE_IPV4; + } else { + size = sizeof(struct in6_addr); + attrtype = RDMA_NLA_F_MANDATORY | LS_NLA_TYPE_IPV6; + } + + len = nla_total_size(sizeof(size)); + len += NLMSG_ALIGN(sizeof(*header)); + + skb = nlmsg_new(len, GFP_KERNEL); + if (!skb) + return -ENOMEM; + + data = ibnl_put_msg(skb, &nlh, seq, 0, RDMA_NL_LS, + RDMA_NL_LS_OP_IP_RESOLVE, NLM_F_REQUEST); + if (!data) { + nlmsg_free(skb); + return -ENODATA; + } + + /* Construct the family header first */ + header = skb_put(skb, NLMSG_ALIGN(sizeof(*header))); + header->ifindex = dev_addr->bound_dev_if; + nla_put(skb, attrtype, size, daddr); + + /* Repair the nlmsg header length */ + nlmsg_end(skb, nlh); + rdma_nl_multicast(&init_net, skb, RDMA_NL_GROUP_LS, GFP_KERNEL); + + /* Make the request retry, so when we get the response from userspace + * we will have something. + */ + return -ENODATA; +} + +int rdma_addr_size(const struct sockaddr *addr) +{ + switch (addr->sa_family) { + case AF_INET: + return sizeof(struct sockaddr_in); + case AF_INET6: + return sizeof(struct sockaddr_in6); + case AF_IB: + return sizeof(struct sockaddr_ib); + default: + return 0; + } +} +EXPORT_SYMBOL(rdma_addr_size); + +int rdma_addr_size_in6(struct sockaddr_in6 *addr) +{ + int ret = rdma_addr_size((struct sockaddr *) addr); + + return ret <= sizeof(*addr) ? ret : 0; +} +EXPORT_SYMBOL(rdma_addr_size_in6); + +int rdma_addr_size_kss(struct __kernel_sockaddr_storage *addr) +{ + int ret = rdma_addr_size((struct sockaddr *) addr); + + return ret <= sizeof(*addr) ? ret : 0; +} +EXPORT_SYMBOL(rdma_addr_size_kss); + +/** + * rdma_copy_src_l2_addr - Copy netdevice source addresses + * @dev_addr: Destination address pointer where to copy the addresses + * @dev: Netdevice whose source addresses to copy + * + * rdma_copy_src_l2_addr() copies source addresses from the specified netdevice. + * This includes unicast address, broadcast address, device type and + * interface index. + */ +void rdma_copy_src_l2_addr(struct rdma_dev_addr *dev_addr, + const struct net_device *dev) +{ + dev_addr->dev_type = dev->type; + memcpy(dev_addr->src_dev_addr, dev->dev_addr, MAX_ADDR_LEN); + memcpy(dev_addr->broadcast, dev->broadcast, MAX_ADDR_LEN); + dev_addr->bound_dev_if = dev->ifindex; +} +EXPORT_SYMBOL(rdma_copy_src_l2_addr); + +static struct net_device * +rdma_find_ndev_for_src_ip_rcu(struct net *net, const struct sockaddr *src_in) +{ + struct net_device *dev = NULL; + int ret = -EADDRNOTAVAIL; + + switch (src_in->sa_family) { + case AF_INET: + dev = __ip_dev_find(net, + ((const struct sockaddr_in *)src_in)->sin_addr.s_addr, + false); + if (dev) + ret = 0; + break; +#if IS_ENABLED(CONFIG_IPV6) + case AF_INET6: + for_each_netdev_rcu(net, dev) { + if (ipv6_chk_addr(net, + &((const struct sockaddr_in6 *)src_in)->sin6_addr, + dev, 1)) { + ret = 0; + break; + } + } + break; +#endif + } + return ret ? ERR_PTR(ret) : dev; +} + +int rdma_translate_ip(const struct sockaddr *addr, + struct rdma_dev_addr *dev_addr) +{ + struct net_device *dev; + + if (dev_addr->bound_dev_if) { + dev = dev_get_by_index(dev_addr->net, dev_addr->bound_dev_if); + if (!dev) + return -ENODEV; + rdma_copy_src_l2_addr(dev_addr, dev); + dev_put(dev); + return 0; + } + + rcu_read_lock(); + dev = rdma_find_ndev_for_src_ip_rcu(dev_addr->net, addr); + if (!IS_ERR(dev)) + rdma_copy_src_l2_addr(dev_addr, dev); + rcu_read_unlock(); + return PTR_ERR_OR_ZERO(dev); +} +EXPORT_SYMBOL(rdma_translate_ip); + +static void set_timeout(struct addr_req *req, unsigned long time) +{ + unsigned long delay; + + delay = time - jiffies; + if ((long)delay < 0) + delay = 0; + + mod_delayed_work(addr_wq, &req->work, delay); +} + +static void queue_req(struct addr_req *req) +{ + spin_lock_bh(&lock); + list_add_tail(&req->list, &req_list); + set_timeout(req, req->timeout); + spin_unlock_bh(&lock); +} + +static int ib_nl_fetch_ha(struct rdma_dev_addr *dev_addr, + const void *daddr, u32 seq, u16 family) +{ + if (!rdma_nl_chk_listeners(RDMA_NL_GROUP_LS)) + return -EADDRNOTAVAIL; + + return ib_nl_ip_send_msg(dev_addr, daddr, seq, family); +} + +static int dst_fetch_ha(const struct dst_entry *dst, + struct rdma_dev_addr *dev_addr, + const void *daddr) +{ + struct neighbour *n; + int ret = 0; + + n = dst_neigh_lookup(dst, daddr); + if (!n) + return -ENODATA; + + if (!(n->nud_state & NUD_VALID)) { + neigh_event_send(n, NULL); + ret = -ENODATA; + } else { + neigh_ha_snapshot(dev_addr->dst_dev_addr, n, dst->dev); + } + + neigh_release(n); + + return ret; +} + +static bool has_gateway(const struct dst_entry *dst, sa_family_t family) +{ + struct rtable *rt; + struct rt6_info *rt6; + + if (family == AF_INET) { + rt = container_of(dst, struct rtable, dst); + return rt->rt_uses_gateway; + } + + rt6 = container_of(dst, struct rt6_info, dst); + return rt6->rt6i_flags & RTF_GATEWAY; +} + +static int fetch_ha(const struct dst_entry *dst, struct rdma_dev_addr *dev_addr, + const struct sockaddr *dst_in, u32 seq) +{ + const struct sockaddr_in *dst_in4 = + (const struct sockaddr_in *)dst_in; + const struct sockaddr_in6 *dst_in6 = + (const struct sockaddr_in6 *)dst_in; + const void *daddr = (dst_in->sa_family == AF_INET) ? + (const void *)&dst_in4->sin_addr.s_addr : + (const void *)&dst_in6->sin6_addr; + sa_family_t family = dst_in->sa_family; + + might_sleep(); + + /* If we have a gateway in IB mode then it must be an IB network */ + if (has_gateway(dst, family) && dev_addr->network == RDMA_NETWORK_IB) + return ib_nl_fetch_ha(dev_addr, daddr, seq, family); + else + return dst_fetch_ha(dst, dev_addr, daddr); +} + +static int addr4_resolve(struct sockaddr *src_sock, + const struct sockaddr *dst_sock, + struct rdma_dev_addr *addr, + struct rtable **prt) +{ + struct sockaddr_in *src_in = (struct sockaddr_in *)src_sock; + const struct sockaddr_in *dst_in = + (const struct sockaddr_in *)dst_sock; + + __be32 src_ip = src_in->sin_addr.s_addr; + __be32 dst_ip = dst_in->sin_addr.s_addr; + struct rtable *rt; + struct flowi4 fl4; + int ret; + + memset(&fl4, 0, sizeof(fl4)); + fl4.daddr = dst_ip; + fl4.saddr = src_ip; + fl4.flowi4_oif = addr->bound_dev_if; + rt = ip_route_output_key(addr->net, &fl4); + ret = PTR_ERR_OR_ZERO(rt); + if (ret) + return ret; + + src_in->sin_addr.s_addr = fl4.saddr; + + addr->hoplimit = ip4_dst_hoplimit(&rt->dst); + + *prt = rt; + return 0; +} + +#if IS_ENABLED(CONFIG_IPV6) +static int addr6_resolve(struct sockaddr *src_sock, + const struct sockaddr *dst_sock, + struct rdma_dev_addr *addr, + struct dst_entry **pdst) +{ + struct sockaddr_in6 *src_in = (struct sockaddr_in6 *)src_sock; + const struct sockaddr_in6 *dst_in = + (const struct sockaddr_in6 *)dst_sock; + struct flowi6 fl6; + struct dst_entry *dst; + + memset(&fl6, 0, sizeof fl6); + fl6.daddr = dst_in->sin6_addr; + fl6.saddr = src_in->sin6_addr; + fl6.flowi6_oif = addr->bound_dev_if; + + dst = ipv6_stub->ipv6_dst_lookup_flow(addr->net, NULL, &fl6, NULL); + if (IS_ERR(dst)) + return PTR_ERR(dst); + + if (ipv6_addr_any(&src_in->sin6_addr)) + src_in->sin6_addr = fl6.saddr; + + addr->hoplimit = ip6_dst_hoplimit(dst); + + *pdst = dst; + return 0; +} +#else +static int addr6_resolve(struct sockaddr *src_sock, + const struct sockaddr *dst_sock, + struct rdma_dev_addr *addr, + struct dst_entry **pdst) +{ + return -EADDRNOTAVAIL; +} +#endif + +static int addr_resolve_neigh(const struct dst_entry *dst, + const struct sockaddr *dst_in, + struct rdma_dev_addr *addr, + unsigned int ndev_flags, + u32 seq) +{ + int ret = 0; + + if (ndev_flags & IFF_LOOPBACK) { + memcpy(addr->dst_dev_addr, addr->src_dev_addr, MAX_ADDR_LEN); + } else { + if (!(ndev_flags & IFF_NOARP)) { + /* If the device doesn't do ARP internally */ + ret = fetch_ha(dst, addr, dst_in, seq); + } + } + return ret; +} + +static int copy_src_l2_addr(struct rdma_dev_addr *dev_addr, + const struct sockaddr *dst_in, + const struct dst_entry *dst, + const struct net_device *ndev) +{ + int ret = 0; + + if (dst->dev->flags & IFF_LOOPBACK) + ret = rdma_translate_ip(dst_in, dev_addr); + else + rdma_copy_src_l2_addr(dev_addr, dst->dev); + + /* + * If there's a gateway and type of device not ARPHRD_INFINIBAND, + * we're definitely in RoCE v2 (as RoCE v1 isn't routable) set the + * network type accordingly. + */ + if (has_gateway(dst, dst_in->sa_family) && + ndev->type != ARPHRD_INFINIBAND) + dev_addr->network = dst_in->sa_family == AF_INET ? + RDMA_NETWORK_IPV4 : + RDMA_NETWORK_IPV6; + else + dev_addr->network = RDMA_NETWORK_IB; + + return ret; +} + +static int rdma_set_src_addr_rcu(struct rdma_dev_addr *dev_addr, + unsigned int *ndev_flags, + const struct sockaddr *dst_in, + const struct dst_entry *dst) +{ + struct net_device *ndev = READ_ONCE(dst->dev); + + *ndev_flags = ndev->flags; + /* A physical device must be the RDMA device to use */ + if (ndev->flags & IFF_LOOPBACK) { + /* + * RDMA (IB/RoCE, iWarp) doesn't run on lo interface or + * loopback IP address. So if route is resolved to loopback + * interface, translate that to a real ndev based on non + * loopback IP address. + */ + ndev = rdma_find_ndev_for_src_ip_rcu(dev_net(ndev), dst_in); + if (IS_ERR(ndev)) + return -ENODEV; + } + + return copy_src_l2_addr(dev_addr, dst_in, dst, ndev); +} + +static int set_addr_netns_by_gid_rcu(struct rdma_dev_addr *addr) +{ + struct net_device *ndev; + + ndev = rdma_read_gid_attr_ndev_rcu(addr->sgid_attr); + if (IS_ERR(ndev)) + return PTR_ERR(ndev); + + /* + * Since we are holding the rcu, reading net and ifindex + * are safe without any additional reference; because + * change_net_namespace() in net/core/dev.c does rcu sync + * after it changes the state to IFF_DOWN and before + * updating netdev fields {net, ifindex}. + */ + addr->net = dev_net(ndev); + addr->bound_dev_if = ndev->ifindex; + return 0; +} + +static void rdma_addr_set_net_defaults(struct rdma_dev_addr *addr) +{ + addr->net = &init_net; + addr->bound_dev_if = 0; +} + +static int addr_resolve(struct sockaddr *src_in, + const struct sockaddr *dst_in, + struct rdma_dev_addr *addr, + bool resolve_neigh, + bool resolve_by_gid_attr, + u32 seq) +{ + struct dst_entry *dst = NULL; + unsigned int ndev_flags = 0; + struct rtable *rt = NULL; + int ret; + + if (!addr->net) { + pr_warn_ratelimited("%s: missing namespace\n", __func__); + return -EINVAL; + } + + rcu_read_lock(); + if (resolve_by_gid_attr) { + if (!addr->sgid_attr) { + rcu_read_unlock(); + pr_warn_ratelimited("%s: missing gid_attr\n", __func__); + return -EINVAL; + } + /* + * If the request is for a specific gid attribute of the + * rdma_dev_addr, derive net from the netdevice of the + * GID attribute. + */ + ret = set_addr_netns_by_gid_rcu(addr); + if (ret) { + rcu_read_unlock(); + return ret; + } + } + if (src_in->sa_family == AF_INET) { + ret = addr4_resolve(src_in, dst_in, addr, &rt); + dst = &rt->dst; + } else { + ret = addr6_resolve(src_in, dst_in, addr, &dst); + } + if (ret) { + rcu_read_unlock(); + goto done; + } + ret = rdma_set_src_addr_rcu(addr, &ndev_flags, dst_in, dst); + rcu_read_unlock(); + + /* + * Resolve neighbor destination address if requested and + * only if src addr translation didn't fail. + */ + if (!ret && resolve_neigh) + ret = addr_resolve_neigh(dst, dst_in, addr, ndev_flags, seq); + + if (src_in->sa_family == AF_INET) + ip_rt_put(rt); + else + dst_release(dst); +done: + /* + * Clear the addr net to go back to its original state, only if it was + * derived from GID attribute in this context. + */ + if (resolve_by_gid_attr) + rdma_addr_set_net_defaults(addr); + return ret; +} + +static void process_one_req(struct work_struct *_work) +{ + struct addr_req *req; + struct sockaddr *src_in, *dst_in; + + req = container_of(_work, struct addr_req, work.work); + + if (req->status == -ENODATA) { + src_in = (struct sockaddr *)&req->src_addr; + dst_in = (struct sockaddr *)&req->dst_addr; + req->status = addr_resolve(src_in, dst_in, req->addr, + true, req->resolve_by_gid_attr, + req->seq); + if (req->status && time_after_eq(jiffies, req->timeout)) { + req->status = -ETIMEDOUT; + } else if (req->status == -ENODATA) { + /* requeue the work for retrying again */ + spin_lock_bh(&lock); + if (!list_empty(&req->list)) + set_timeout(req, req->timeout); + spin_unlock_bh(&lock); + return; + } + } + + req->callback(req->status, (struct sockaddr *)&req->src_addr, + req->addr, req->context); + req->callback = NULL; + + spin_lock_bh(&lock); + /* + * Although the work will normally have been canceled by the workqueue, + * it can still be requeued as long as it is on the req_list. + */ + cancel_delayed_work(&req->work); + if (!list_empty(&req->list)) { + list_del_init(&req->list); + kfree(req); + } + spin_unlock_bh(&lock); +} + +int rdma_resolve_ip(struct sockaddr *src_addr, const struct sockaddr *dst_addr, + struct rdma_dev_addr *addr, unsigned long timeout_ms, + void (*callback)(int status, struct sockaddr *src_addr, + struct rdma_dev_addr *addr, void *context), + bool resolve_by_gid_attr, void *context) +{ + struct sockaddr *src_in, *dst_in; + struct addr_req *req; + int ret = 0; + + req = kzalloc(sizeof *req, GFP_KERNEL); + if (!req) + return -ENOMEM; + + src_in = (struct sockaddr *) &req->src_addr; + dst_in = (struct sockaddr *) &req->dst_addr; + + if (src_addr) { + if (src_addr->sa_family != dst_addr->sa_family) { + ret = -EINVAL; + goto err; + } + + memcpy(src_in, src_addr, rdma_addr_size(src_addr)); + } else { + src_in->sa_family = dst_addr->sa_family; + } + + memcpy(dst_in, dst_addr, rdma_addr_size(dst_addr)); + req->addr = addr; + req->callback = callback; + req->context = context; + req->resolve_by_gid_attr = resolve_by_gid_attr; + INIT_DELAYED_WORK(&req->work, process_one_req); + req->seq = (u32)atomic_inc_return(&ib_nl_addr_request_seq); + + req->status = addr_resolve(src_in, dst_in, addr, true, + req->resolve_by_gid_attr, req->seq); + switch (req->status) { + case 0: + req->timeout = jiffies; + queue_req(req); + break; + case -ENODATA: + req->timeout = msecs_to_jiffies(timeout_ms) + jiffies; + queue_req(req); + break; + default: + ret = req->status; + goto err; + } + return ret; +err: + kfree(req); + return ret; +} +EXPORT_SYMBOL(rdma_resolve_ip); + +int roce_resolve_route_from_path(struct sa_path_rec *rec, + const struct ib_gid_attr *attr) +{ + union { + struct sockaddr _sockaddr; + struct sockaddr_in _sockaddr_in; + struct sockaddr_in6 _sockaddr_in6; + } sgid, dgid; + struct rdma_dev_addr dev_addr = {}; + int ret; + + might_sleep(); + + if (rec->roce.route_resolved) + return 0; + + rdma_gid2ip((struct sockaddr *)&sgid, &rec->sgid); + rdma_gid2ip((struct sockaddr *)&dgid, &rec->dgid); + + if (sgid._sockaddr.sa_family != dgid._sockaddr.sa_family) + return -EINVAL; + + if (!attr || !attr->ndev) + return -EINVAL; + + dev_addr.net = &init_net; + dev_addr.sgid_attr = attr; + + ret = addr_resolve((struct sockaddr *)&sgid, (struct sockaddr *)&dgid, + &dev_addr, false, true, 0); + if (ret) + return ret; + + if ((dev_addr.network == RDMA_NETWORK_IPV4 || + dev_addr.network == RDMA_NETWORK_IPV6) && + rec->rec_type != SA_PATH_REC_TYPE_ROCE_V2) + return -EINVAL; + + rec->roce.route_resolved = true; + return 0; +} + +/** + * rdma_addr_cancel - Cancel resolve ip request + * @addr: Pointer to address structure given previously + * during rdma_resolve_ip(). + * rdma_addr_cancel() is synchronous function which cancels any pending + * request if there is any. + */ +void rdma_addr_cancel(struct rdma_dev_addr *addr) +{ + struct addr_req *req, *temp_req; + struct addr_req *found = NULL; + + spin_lock_bh(&lock); + list_for_each_entry_safe(req, temp_req, &req_list, list) { + if (req->addr == addr) { + /* + * Removing from the list means we take ownership of + * the req + */ + list_del_init(&req->list); + found = req; + break; + } + } + spin_unlock_bh(&lock); + + if (!found) + return; + + /* + * sync canceling the work after removing it from the req_list + * guarentees no work is running and none will be started. + */ + cancel_delayed_work_sync(&found->work); + kfree(found); +} +EXPORT_SYMBOL(rdma_addr_cancel); + +struct resolve_cb_context { + struct completion comp; + int status; +}; + +static void resolve_cb(int status, struct sockaddr *src_addr, + struct rdma_dev_addr *addr, void *context) +{ + ((struct resolve_cb_context *)context)->status = status; + complete(&((struct resolve_cb_context *)context)->comp); +} + +int rdma_addr_find_l2_eth_by_grh(const union ib_gid *sgid, + const union ib_gid *dgid, + u8 *dmac, const struct ib_gid_attr *sgid_attr, + int *hoplimit) +{ + struct rdma_dev_addr dev_addr; + struct resolve_cb_context ctx; + union { + struct sockaddr_in _sockaddr_in; + struct sockaddr_in6 _sockaddr_in6; + } sgid_addr, dgid_addr; + int ret; + + rdma_gid2ip((struct sockaddr *)&sgid_addr, sgid); + rdma_gid2ip((struct sockaddr *)&dgid_addr, dgid); + + memset(&dev_addr, 0, sizeof(dev_addr)); + dev_addr.net = &init_net; + dev_addr.sgid_attr = sgid_attr; + + init_completion(&ctx.comp); + ret = rdma_resolve_ip((struct sockaddr *)&sgid_addr, + (struct sockaddr *)&dgid_addr, &dev_addr, 1000, + resolve_cb, true, &ctx); + if (ret) + return ret; + + wait_for_completion(&ctx.comp); + + ret = ctx.status; + if (ret) + return ret; + + memcpy(dmac, dev_addr.dst_dev_addr, ETH_ALEN); + *hoplimit = dev_addr.hoplimit; + return 0; +} + +static int netevent_callback(struct notifier_block *self, unsigned long event, + void *ctx) +{ + struct addr_req *req; + + if (event == NETEVENT_NEIGH_UPDATE) { + struct neighbour *neigh = ctx; + + if (neigh->nud_state & NUD_VALID) { + spin_lock_bh(&lock); + list_for_each_entry(req, &req_list, list) + set_timeout(req, jiffies); + spin_unlock_bh(&lock); + } + } + return 0; +} + +static struct notifier_block nb = { + .notifier_call = netevent_callback +}; + +int addr_init(void) +{ + addr_wq = alloc_ordered_workqueue("ib_addr", 0); + if (!addr_wq) + return -ENOMEM; + + register_netevent_notifier(&nb); + + return 0; +} + +void addr_cleanup(void) +{ + unregister_netevent_notifier(&nb); + destroy_workqueue(addr_wq); + WARN_ON(!list_empty(&req_list)); +} diff --git a/drivers/infiniband/core/agent.c b/drivers/infiniband/core/agent.c new file mode 100644 index 000000000..f82b4260d --- /dev/null +++ b/drivers/infiniband/core/agent.c @@ -0,0 +1,221 @@ +/* + * Copyright (c) 2004, 2005 Mellanox Technologies Ltd. All rights reserved. + * Copyright (c) 2004, 2005 Infinicon Corporation. All rights reserved. + * Copyright (c) 2004, 2005 Intel Corporation. All rights reserved. + * Copyright (c) 2004, 2005 Topspin Corporation. All rights reserved. + * Copyright (c) 2004-2007 Voltaire Corporation. All rights reserved. + * Copyright (c) 2005 Sun Microsystems, Inc. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + */ + +#include <linux/slab.h> +#include <linux/string.h> + +#include "agent.h" +#include "smi.h" +#include "mad_priv.h" + +#define SPFX "ib_agent: " + +struct ib_agent_port_private { + struct list_head port_list; + struct ib_mad_agent *agent[2]; +}; + +static DEFINE_SPINLOCK(ib_agent_port_list_lock); +static LIST_HEAD(ib_agent_port_list); + +static struct ib_agent_port_private * +__ib_get_agent_port(const struct ib_device *device, int port_num) +{ + struct ib_agent_port_private *entry; + + list_for_each_entry(entry, &ib_agent_port_list, port_list) { + if (entry->agent[1]->device == device && + entry->agent[1]->port_num == port_num) + return entry; + } + return NULL; +} + +static struct ib_agent_port_private * +ib_get_agent_port(const struct ib_device *device, int port_num) +{ + struct ib_agent_port_private *entry; + unsigned long flags; + + spin_lock_irqsave(&ib_agent_port_list_lock, flags); + entry = __ib_get_agent_port(device, port_num); + spin_unlock_irqrestore(&ib_agent_port_list_lock, flags); + return entry; +} + +void agent_send_response(const struct ib_mad_hdr *mad_hdr, const struct ib_grh *grh, + const struct ib_wc *wc, const struct ib_device *device, + int port_num, int qpn, size_t resp_mad_len, bool opa) +{ + struct ib_agent_port_private *port_priv; + struct ib_mad_agent *agent; + struct ib_mad_send_buf *send_buf; + struct ib_ah *ah; + struct ib_mad_send_wr_private *mad_send_wr; + + if (rdma_cap_ib_switch(device)) + port_priv = ib_get_agent_port(device, 0); + else + port_priv = ib_get_agent_port(device, port_num); + + if (!port_priv) { + dev_err(&device->dev, "Unable to find port agent\n"); + return; + } + + agent = port_priv->agent[qpn]; + ah = ib_create_ah_from_wc(agent->qp->pd, wc, grh, port_num); + if (IS_ERR(ah)) { + dev_err(&device->dev, "ib_create_ah_from_wc error %ld\n", + PTR_ERR(ah)); + return; + } + + if (opa && mad_hdr->base_version != OPA_MGMT_BASE_VERSION) + resp_mad_len = IB_MGMT_MAD_SIZE; + + send_buf = ib_create_send_mad(agent, wc->src_qp, wc->pkey_index, 0, + IB_MGMT_MAD_HDR, + resp_mad_len - IB_MGMT_MAD_HDR, + GFP_KERNEL, + mad_hdr->base_version); + if (IS_ERR(send_buf)) { + dev_err(&device->dev, "ib_create_send_mad error\n"); + goto err1; + } + + memcpy(send_buf->mad, mad_hdr, resp_mad_len); + send_buf->ah = ah; + + if (rdma_cap_ib_switch(device)) { + mad_send_wr = container_of(send_buf, + struct ib_mad_send_wr_private, + send_buf); + mad_send_wr->send_wr.port_num = port_num; + } + + if (ib_post_send_mad(send_buf, NULL)) { + dev_err(&device->dev, "ib_post_send_mad error\n"); + goto err2; + } + return; +err2: + ib_free_send_mad(send_buf); +err1: + rdma_destroy_ah(ah, RDMA_DESTROY_AH_SLEEPABLE); +} + +static void agent_send_handler(struct ib_mad_agent *mad_agent, + struct ib_mad_send_wc *mad_send_wc) +{ + rdma_destroy_ah(mad_send_wc->send_buf->ah, RDMA_DESTROY_AH_SLEEPABLE); + ib_free_send_mad(mad_send_wc->send_buf); +} + +int ib_agent_port_open(struct ib_device *device, int port_num) +{ + struct ib_agent_port_private *port_priv; + unsigned long flags; + int ret; + + /* Create new device info */ + port_priv = kzalloc(sizeof *port_priv, GFP_KERNEL); + if (!port_priv) { + ret = -ENOMEM; + goto error1; + } + + if (rdma_cap_ib_smi(device, port_num)) { + /* Obtain send only MAD agent for SMI QP */ + port_priv->agent[0] = ib_register_mad_agent(device, port_num, + IB_QPT_SMI, NULL, 0, + &agent_send_handler, + NULL, NULL, 0); + if (IS_ERR(port_priv->agent[0])) { + ret = PTR_ERR(port_priv->agent[0]); + goto error2; + } + } + + /* Obtain send only MAD agent for GSI QP */ + port_priv->agent[1] = ib_register_mad_agent(device, port_num, + IB_QPT_GSI, NULL, 0, + &agent_send_handler, + NULL, NULL, 0); + if (IS_ERR(port_priv->agent[1])) { + ret = PTR_ERR(port_priv->agent[1]); + goto error3; + } + + spin_lock_irqsave(&ib_agent_port_list_lock, flags); + list_add_tail(&port_priv->port_list, &ib_agent_port_list); + spin_unlock_irqrestore(&ib_agent_port_list_lock, flags); + + return 0; + +error3: + if (port_priv->agent[0]) + ib_unregister_mad_agent(port_priv->agent[0]); +error2: + kfree(port_priv); +error1: + return ret; +} + +int ib_agent_port_close(struct ib_device *device, int port_num) +{ + struct ib_agent_port_private *port_priv; + unsigned long flags; + + spin_lock_irqsave(&ib_agent_port_list_lock, flags); + port_priv = __ib_get_agent_port(device, port_num); + if (port_priv == NULL) { + spin_unlock_irqrestore(&ib_agent_port_list_lock, flags); + dev_err(&device->dev, "Port %d not found\n", port_num); + return -ENODEV; + } + list_del(&port_priv->port_list); + spin_unlock_irqrestore(&ib_agent_port_list_lock, flags); + + ib_unregister_mad_agent(port_priv->agent[1]); + if (port_priv->agent[0]) + ib_unregister_mad_agent(port_priv->agent[0]); + + kfree(port_priv); + return 0; +} diff --git a/drivers/infiniband/core/agent.h b/drivers/infiniband/core/agent.h new file mode 100644 index 000000000..65f92beda --- /dev/null +++ b/drivers/infiniband/core/agent.h @@ -0,0 +1,51 @@ +/* + * Copyright (c) 2004 Mellanox Technologies Ltd. All rights reserved. + * Copyright (c) 2004 Infinicon Corporation. All rights reserved. + * Copyright (c) 2004 Intel Corporation. All rights reserved. + * Copyright (c) 2004 Topspin Corporation. All rights reserved. + * Copyright (c) 2004 Voltaire Corporation. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#ifndef __AGENT_H_ +#define __AGENT_H_ + +#include <linux/err.h> +#include <rdma/ib_mad.h> + +extern int ib_agent_port_open(struct ib_device *device, int port_num); + +extern int ib_agent_port_close(struct ib_device *device, int port_num); + +extern void agent_send_response(const struct ib_mad_hdr *mad_hdr, const struct ib_grh *grh, + const struct ib_wc *wc, const struct ib_device *device, + int port_num, int qpn, size_t resp_mad_len, bool opa); + +#endif /* __AGENT_H_ */ diff --git a/drivers/infiniband/core/cache.c b/drivers/infiniband/core/cache.c new file mode 100644 index 000000000..7989b7e1d --- /dev/null +++ b/drivers/infiniband/core/cache.c @@ -0,0 +1,1674 @@ +/* + * Copyright (c) 2004 Topspin Communications. All rights reserved. + * Copyright (c) 2005 Intel Corporation. All rights reserved. + * Copyright (c) 2005 Sun Microsystems, Inc. All rights reserved. + * Copyright (c) 2005 Voltaire, Inc. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include <linux/module.h> +#include <linux/errno.h> +#include <linux/slab.h> +#include <linux/workqueue.h> +#include <linux/netdevice.h> +#include <net/addrconf.h> + +#include <rdma/ib_cache.h> + +#include "core_priv.h" + +struct ib_pkey_cache { + int table_len; + u16 table[]; +}; + +struct ib_update_work { + struct work_struct work; + struct ib_event event; + bool enforce_security; +}; + +union ib_gid zgid; +EXPORT_SYMBOL(zgid); + +enum gid_attr_find_mask { + GID_ATTR_FIND_MASK_GID = 1UL << 0, + GID_ATTR_FIND_MASK_NETDEV = 1UL << 1, + GID_ATTR_FIND_MASK_DEFAULT = 1UL << 2, + GID_ATTR_FIND_MASK_GID_TYPE = 1UL << 3, +}; + +enum gid_table_entry_state { + GID_TABLE_ENTRY_INVALID = 1, + GID_TABLE_ENTRY_VALID = 2, + /* + * Indicates that entry is pending to be removed, there may + * be active users of this GID entry. + * When last user of the GID entry releases reference to it, + * GID entry is detached from the table. + */ + GID_TABLE_ENTRY_PENDING_DEL = 3, +}; + +struct roce_gid_ndev_storage { + struct rcu_head rcu_head; + struct net_device *ndev; +}; + +struct ib_gid_table_entry { + struct kref kref; + struct work_struct del_work; + struct ib_gid_attr attr; + void *context; + /* Store the ndev pointer to release reference later on in + * call_rcu context because by that time gid_table_entry + * and attr might be already freed. So keep a copy of it. + * ndev_storage is freed by rcu callback. + */ + struct roce_gid_ndev_storage *ndev_storage; + enum gid_table_entry_state state; +}; + +struct ib_gid_table { + int sz; + /* In RoCE, adding a GID to the table requires: + * (a) Find if this GID is already exists. + * (b) Find a free space. + * (c) Write the new GID + * + * Delete requires different set of operations: + * (a) Find the GID + * (b) Delete it. + * + **/ + /* Any writer to data_vec must hold this lock and the write side of + * rwlock. Readers must hold only rwlock. All writers must be in a + * sleepable context. + */ + struct mutex lock; + /* rwlock protects data_vec[ix]->state and entry pointer. + */ + rwlock_t rwlock; + struct ib_gid_table_entry **data_vec; + /* bit field, each bit indicates the index of default GID */ + u32 default_gid_indices; +}; + +static void dispatch_gid_change_event(struct ib_device *ib_dev, u8 port) +{ + struct ib_event event; + + event.device = ib_dev; + event.element.port_num = port; + event.event = IB_EVENT_GID_CHANGE; + + ib_dispatch_event_clients(&event); +} + +static const char * const gid_type_str[] = { + /* IB/RoCE v1 value is set for IB_GID_TYPE_IB and IB_GID_TYPE_ROCE for + * user space compatibility reasons. + */ + [IB_GID_TYPE_IB] = "IB/RoCE v1", + [IB_GID_TYPE_ROCE] = "IB/RoCE v1", + [IB_GID_TYPE_ROCE_UDP_ENCAP] = "RoCE v2", +}; + +const char *ib_cache_gid_type_str(enum ib_gid_type gid_type) +{ + if (gid_type < ARRAY_SIZE(gid_type_str) && gid_type_str[gid_type]) + return gid_type_str[gid_type]; + + return "Invalid GID type"; +} +EXPORT_SYMBOL(ib_cache_gid_type_str); + +/** rdma_is_zero_gid - Check if given GID is zero or not. + * @gid: GID to check + * Returns true if given GID is zero, returns false otherwise. + */ +bool rdma_is_zero_gid(const union ib_gid *gid) +{ + return !memcmp(gid, &zgid, sizeof(*gid)); +} +EXPORT_SYMBOL(rdma_is_zero_gid); + +/** is_gid_index_default - Check if a given index belongs to + * reserved default GIDs or not. + * @table: GID table pointer + * @index: Index to check in GID table + * Returns true if index is one of the reserved default GID index otherwise + * returns false. + */ +static bool is_gid_index_default(const struct ib_gid_table *table, + unsigned int index) +{ + return index < 32 && (BIT(index) & table->default_gid_indices); +} + +int ib_cache_gid_parse_type_str(const char *buf) +{ + unsigned int i; + size_t len; + int err = -EINVAL; + + len = strlen(buf); + if (len == 0) + return -EINVAL; + + if (buf[len - 1] == '\n') + len--; + + for (i = 0; i < ARRAY_SIZE(gid_type_str); ++i) + if (gid_type_str[i] && !strncmp(buf, gid_type_str[i], len) && + len == strlen(gid_type_str[i])) { + err = i; + break; + } + + return err; +} +EXPORT_SYMBOL(ib_cache_gid_parse_type_str); + +static struct ib_gid_table *rdma_gid_table(struct ib_device *device, u8 port) +{ + return device->port_data[port].cache.gid; +} + +static bool is_gid_entry_free(const struct ib_gid_table_entry *entry) +{ + return !entry; +} + +static bool is_gid_entry_valid(const struct ib_gid_table_entry *entry) +{ + return entry && entry->state == GID_TABLE_ENTRY_VALID; +} + +static void schedule_free_gid(struct kref *kref) +{ + struct ib_gid_table_entry *entry = + container_of(kref, struct ib_gid_table_entry, kref); + + queue_work(ib_wq, &entry->del_work); +} + +static void put_gid_ndev(struct rcu_head *head) +{ + struct roce_gid_ndev_storage *storage = + container_of(head, struct roce_gid_ndev_storage, rcu_head); + + WARN_ON(!storage->ndev); + /* At this point its safe to release netdev reference, + * as all callers working on gid_attr->ndev are done + * using this netdev. + */ + dev_put(storage->ndev); + kfree(storage); +} + +static void free_gid_entry_locked(struct ib_gid_table_entry *entry) +{ + struct ib_device *device = entry->attr.device; + u8 port_num = entry->attr.port_num; + struct ib_gid_table *table = rdma_gid_table(device, port_num); + + dev_dbg(&device->dev, "%s port=%d index=%d gid %pI6\n", __func__, + port_num, entry->attr.index, entry->attr.gid.raw); + + write_lock_irq(&table->rwlock); + + /* + * The only way to avoid overwriting NULL in table is + * by comparing if it is same entry in table or not! + * If new entry in table is added by the time we free here, + * don't overwrite the table entry. + */ + if (entry == table->data_vec[entry->attr.index]) + table->data_vec[entry->attr.index] = NULL; + /* Now this index is ready to be allocated */ + write_unlock_irq(&table->rwlock); + + if (entry->ndev_storage) + call_rcu(&entry->ndev_storage->rcu_head, put_gid_ndev); + kfree(entry); +} + +static void free_gid_entry(struct kref *kref) +{ + struct ib_gid_table_entry *entry = + container_of(kref, struct ib_gid_table_entry, kref); + + free_gid_entry_locked(entry); +} + +/** + * free_gid_work - Release reference to the GID entry + * @work: Work structure to refer to GID entry which needs to be + * deleted. + * + * free_gid_work() frees the entry from the HCA's hardware table + * if provider supports it. It releases reference to netdevice. + */ +static void free_gid_work(struct work_struct *work) +{ + struct ib_gid_table_entry *entry = + container_of(work, struct ib_gid_table_entry, del_work); + struct ib_device *device = entry->attr.device; + u8 port_num = entry->attr.port_num; + struct ib_gid_table *table = rdma_gid_table(device, port_num); + + mutex_lock(&table->lock); + free_gid_entry_locked(entry); + mutex_unlock(&table->lock); +} + +static struct ib_gid_table_entry * +alloc_gid_entry(const struct ib_gid_attr *attr) +{ + struct ib_gid_table_entry *entry; + struct net_device *ndev; + + entry = kzalloc(sizeof(*entry), GFP_KERNEL); + if (!entry) + return NULL; + + ndev = rcu_dereference_protected(attr->ndev, 1); + if (ndev) { + entry->ndev_storage = kzalloc(sizeof(*entry->ndev_storage), + GFP_KERNEL); + if (!entry->ndev_storage) { + kfree(entry); + return NULL; + } + dev_hold(ndev); + entry->ndev_storage->ndev = ndev; + } + kref_init(&entry->kref); + memcpy(&entry->attr, attr, sizeof(*attr)); + INIT_WORK(&entry->del_work, free_gid_work); + entry->state = GID_TABLE_ENTRY_INVALID; + return entry; +} + +static void store_gid_entry(struct ib_gid_table *table, + struct ib_gid_table_entry *entry) +{ + entry->state = GID_TABLE_ENTRY_VALID; + + dev_dbg(&entry->attr.device->dev, "%s port=%d index=%d gid %pI6\n", + __func__, entry->attr.port_num, entry->attr.index, + entry->attr.gid.raw); + + lockdep_assert_held(&table->lock); + write_lock_irq(&table->rwlock); + table->data_vec[entry->attr.index] = entry; + write_unlock_irq(&table->rwlock); +} + +static void get_gid_entry(struct ib_gid_table_entry *entry) +{ + kref_get(&entry->kref); +} + +static void put_gid_entry(struct ib_gid_table_entry *entry) +{ + kref_put(&entry->kref, schedule_free_gid); +} + +static void put_gid_entry_locked(struct ib_gid_table_entry *entry) +{ + kref_put(&entry->kref, free_gid_entry); +} + +static int add_roce_gid(struct ib_gid_table_entry *entry) +{ + const struct ib_gid_attr *attr = &entry->attr; + int ret; + + if (!attr->ndev) { + dev_err(&attr->device->dev, "%s NULL netdev port=%d index=%d\n", + __func__, attr->port_num, attr->index); + return -EINVAL; + } + if (rdma_cap_roce_gid_table(attr->device, attr->port_num)) { + ret = attr->device->ops.add_gid(attr, &entry->context); + if (ret) { + dev_err(&attr->device->dev, + "%s GID add failed port=%d index=%d\n", + __func__, attr->port_num, attr->index); + return ret; + } + } + return 0; +} + +/** + * del_gid - Delete GID table entry + * + * @ib_dev: IB device whose GID entry to be deleted + * @port: Port number of the IB device + * @table: GID table of the IB device for a port + * @ix: GID entry index to delete + * + */ +static void del_gid(struct ib_device *ib_dev, u8 port, + struct ib_gid_table *table, int ix) +{ + struct roce_gid_ndev_storage *ndev_storage; + struct ib_gid_table_entry *entry; + + lockdep_assert_held(&table->lock); + + dev_dbg(&ib_dev->dev, "%s port=%d index=%d gid %pI6\n", __func__, port, + ix, table->data_vec[ix]->attr.gid.raw); + + write_lock_irq(&table->rwlock); + entry = table->data_vec[ix]; + entry->state = GID_TABLE_ENTRY_PENDING_DEL; + /* + * For non RoCE protocol, GID entry slot is ready to use. + */ + if (!rdma_protocol_roce(ib_dev, port)) + table->data_vec[ix] = NULL; + write_unlock_irq(&table->rwlock); + + ndev_storage = entry->ndev_storage; + if (ndev_storage) { + entry->ndev_storage = NULL; + rcu_assign_pointer(entry->attr.ndev, NULL); + call_rcu(&ndev_storage->rcu_head, put_gid_ndev); + } + + if (rdma_cap_roce_gid_table(ib_dev, port)) + ib_dev->ops.del_gid(&entry->attr, &entry->context); + + put_gid_entry_locked(entry); +} + +/** + * add_modify_gid - Add or modify GID table entry + * + * @table: GID table in which GID to be added or modified + * @attr: Attributes of the GID + * + * Returns 0 on success or appropriate error code. It accepts zero + * GID addition for non RoCE ports for HCA's who report them as valid + * GID. However such zero GIDs are not added to the cache. + */ +static int add_modify_gid(struct ib_gid_table *table, + const struct ib_gid_attr *attr) +{ + struct ib_gid_table_entry *entry; + int ret = 0; + + /* + * Invalidate any old entry in the table to make it safe to write to + * this index. + */ + if (is_gid_entry_valid(table->data_vec[attr->index])) + del_gid(attr->device, attr->port_num, table, attr->index); + + /* + * Some HCA's report multiple GID entries with only one valid GID, and + * leave other unused entries as the zero GID. Convert zero GIDs to + * empty table entries instead of storing them. + */ + if (rdma_is_zero_gid(&attr->gid)) + return 0; + + entry = alloc_gid_entry(attr); + if (!entry) + return -ENOMEM; + + if (rdma_protocol_roce(attr->device, attr->port_num)) { + ret = add_roce_gid(entry); + if (ret) + goto done; + } + + store_gid_entry(table, entry); + return 0; + +done: + put_gid_entry(entry); + return ret; +} + +/* rwlock should be read locked, or lock should be held */ +static int find_gid(struct ib_gid_table *table, const union ib_gid *gid, + const struct ib_gid_attr *val, bool default_gid, + unsigned long mask, int *pempty) +{ + int i = 0; + int found = -1; + int empty = pempty ? -1 : 0; + + while (i < table->sz && (found < 0 || empty < 0)) { + struct ib_gid_table_entry *data = table->data_vec[i]; + struct ib_gid_attr *attr; + int curr_index = i; + + i++; + + /* find_gid() is used during GID addition where it is expected + * to return a free entry slot which is not duplicate. + * Free entry slot is requested and returned if pempty is set, + * so lookup free slot only if requested. + */ + if (pempty && empty < 0) { + if (is_gid_entry_free(data) && + default_gid == + is_gid_index_default(table, curr_index)) { + /* + * Found an invalid (free) entry; allocate it. + * If default GID is requested, then our + * found slot must be one of the DEFAULT + * reserved slots or we fail. + * This ensures that only DEFAULT reserved + * slots are used for default property GIDs. + */ + empty = curr_index; + } + } + + /* + * Additionally find_gid() is used to find valid entry during + * lookup operation; so ignore the entries which are marked as + * pending for removal and the entries which are marked as + * invalid. + */ + if (!is_gid_entry_valid(data)) + continue; + + if (found >= 0) + continue; + + attr = &data->attr; + if (mask & GID_ATTR_FIND_MASK_GID_TYPE && + attr->gid_type != val->gid_type) + continue; + + if (mask & GID_ATTR_FIND_MASK_GID && + memcmp(gid, &data->attr.gid, sizeof(*gid))) + continue; + + if (mask & GID_ATTR_FIND_MASK_NETDEV && + attr->ndev != val->ndev) + continue; + + if (mask & GID_ATTR_FIND_MASK_DEFAULT && + is_gid_index_default(table, curr_index) != default_gid) + continue; + + found = curr_index; + } + + if (pempty) + *pempty = empty; + + return found; +} + +static void make_default_gid(struct net_device *dev, union ib_gid *gid) +{ + gid->global.subnet_prefix = cpu_to_be64(0xfe80000000000000LL); + addrconf_ifid_eui48(&gid->raw[8], dev); +} + +static int __ib_cache_gid_add(struct ib_device *ib_dev, u8 port, + union ib_gid *gid, struct ib_gid_attr *attr, + unsigned long mask, bool default_gid) +{ + struct ib_gid_table *table; + int ret = 0; + int empty; + int ix; + + /* Do not allow adding zero GID in support of + * IB spec version 1.3 section 4.1.1 point (6) and + * section 12.7.10 and section 12.7.20 + */ + if (rdma_is_zero_gid(gid)) + return -EINVAL; + + table = rdma_gid_table(ib_dev, port); + + mutex_lock(&table->lock); + + ix = find_gid(table, gid, attr, default_gid, mask, &empty); + if (ix >= 0) + goto out_unlock; + + if (empty < 0) { + ret = -ENOSPC; + goto out_unlock; + } + attr->device = ib_dev; + attr->index = empty; + attr->port_num = port; + attr->gid = *gid; + ret = add_modify_gid(table, attr); + if (!ret) + dispatch_gid_change_event(ib_dev, port); + +out_unlock: + mutex_unlock(&table->lock); + if (ret) + pr_warn("%s: unable to add gid %pI6 error=%d\n", + __func__, gid->raw, ret); + return ret; +} + +int ib_cache_gid_add(struct ib_device *ib_dev, u8 port, + union ib_gid *gid, struct ib_gid_attr *attr) +{ + unsigned long mask = GID_ATTR_FIND_MASK_GID | + GID_ATTR_FIND_MASK_GID_TYPE | + GID_ATTR_FIND_MASK_NETDEV; + + return __ib_cache_gid_add(ib_dev, port, gid, attr, mask, false); +} + +static int +_ib_cache_gid_del(struct ib_device *ib_dev, u8 port, + union ib_gid *gid, struct ib_gid_attr *attr, + unsigned long mask, bool default_gid) +{ + struct ib_gid_table *table; + int ret = 0; + int ix; + + table = rdma_gid_table(ib_dev, port); + + mutex_lock(&table->lock); + + ix = find_gid(table, gid, attr, default_gid, mask, NULL); + if (ix < 0) { + ret = -EINVAL; + goto out_unlock; + } + + del_gid(ib_dev, port, table, ix); + dispatch_gid_change_event(ib_dev, port); + +out_unlock: + mutex_unlock(&table->lock); + if (ret) + pr_debug("%s: can't delete gid %pI6 error=%d\n", + __func__, gid->raw, ret); + return ret; +} + +int ib_cache_gid_del(struct ib_device *ib_dev, u8 port, + union ib_gid *gid, struct ib_gid_attr *attr) +{ + unsigned long mask = GID_ATTR_FIND_MASK_GID | + GID_ATTR_FIND_MASK_GID_TYPE | + GID_ATTR_FIND_MASK_DEFAULT | + GID_ATTR_FIND_MASK_NETDEV; + + return _ib_cache_gid_del(ib_dev, port, gid, attr, mask, false); +} + +int ib_cache_gid_del_all_netdev_gids(struct ib_device *ib_dev, u8 port, + struct net_device *ndev) +{ + struct ib_gid_table *table; + int ix; + bool deleted = false; + + table = rdma_gid_table(ib_dev, port); + + mutex_lock(&table->lock); + + for (ix = 0; ix < table->sz; ix++) { + if (is_gid_entry_valid(table->data_vec[ix]) && + table->data_vec[ix]->attr.ndev == ndev) { + del_gid(ib_dev, port, table, ix); + deleted = true; + } + } + + mutex_unlock(&table->lock); + + if (deleted) + dispatch_gid_change_event(ib_dev, port); + + return 0; +} + +/** + * rdma_find_gid_by_port - Returns the GID entry attributes when it finds + * a valid GID entry for given search parameters. It searches for the specified + * GID value in the local software cache. + * @device: The device to query. + * @gid: The GID value to search for. + * @gid_type: The GID type to search for. + * @port_num: The port number of the device where the GID value should be + * searched. + * @ndev: In RoCE, the net device of the device. NULL means ignore. + * + * Returns sgid attributes if the GID is found with valid reference or + * returns ERR_PTR for the error. + * The caller must invoke rdma_put_gid_attr() to release the reference. + */ +const struct ib_gid_attr * +rdma_find_gid_by_port(struct ib_device *ib_dev, + const union ib_gid *gid, + enum ib_gid_type gid_type, + u8 port, struct net_device *ndev) +{ + int local_index; + struct ib_gid_table *table; + unsigned long mask = GID_ATTR_FIND_MASK_GID | + GID_ATTR_FIND_MASK_GID_TYPE; + struct ib_gid_attr val = {.ndev = ndev, .gid_type = gid_type}; + const struct ib_gid_attr *attr; + unsigned long flags; + + if (!rdma_is_port_valid(ib_dev, port)) + return ERR_PTR(-ENOENT); + + table = rdma_gid_table(ib_dev, port); + + if (ndev) + mask |= GID_ATTR_FIND_MASK_NETDEV; + + read_lock_irqsave(&table->rwlock, flags); + local_index = find_gid(table, gid, &val, false, mask, NULL); + if (local_index >= 0) { + get_gid_entry(table->data_vec[local_index]); + attr = &table->data_vec[local_index]->attr; + read_unlock_irqrestore(&table->rwlock, flags); + return attr; + } + + read_unlock_irqrestore(&table->rwlock, flags); + return ERR_PTR(-ENOENT); +} +EXPORT_SYMBOL(rdma_find_gid_by_port); + +/** + * rdma_find_gid_by_filter - Returns the GID table attribute where a + * specified GID value occurs + * @device: The device to query. + * @gid: The GID value to search for. + * @port: The port number of the device where the GID value could be + * searched. + * @filter: The filter function is executed on any matching GID in the table. + * If the filter function returns true, the corresponding index is returned, + * otherwise, we continue searching the GID table. It's guaranteed that + * while filter is executed, ndev field is valid and the structure won't + * change. filter is executed in an atomic context. filter must not be NULL. + * + * rdma_find_gid_by_filter() searches for the specified GID value + * of which the filter function returns true in the port's GID table. + * + */ +const struct ib_gid_attr *rdma_find_gid_by_filter( + struct ib_device *ib_dev, const union ib_gid *gid, u8 port, + bool (*filter)(const union ib_gid *gid, const struct ib_gid_attr *, + void *), + void *context) +{ + const struct ib_gid_attr *res = ERR_PTR(-ENOENT); + struct ib_gid_table *table; + unsigned long flags; + unsigned int i; + + if (!rdma_is_port_valid(ib_dev, port)) + return ERR_PTR(-EINVAL); + + table = rdma_gid_table(ib_dev, port); + + read_lock_irqsave(&table->rwlock, flags); + for (i = 0; i < table->sz; i++) { + struct ib_gid_table_entry *entry = table->data_vec[i]; + + if (!is_gid_entry_valid(entry)) + continue; + + if (memcmp(gid, &entry->attr.gid, sizeof(*gid))) + continue; + + if (filter(gid, &entry->attr, context)) { + get_gid_entry(entry); + res = &entry->attr; + break; + } + } + read_unlock_irqrestore(&table->rwlock, flags); + return res; +} + +static struct ib_gid_table *alloc_gid_table(int sz) +{ + struct ib_gid_table *table = kzalloc(sizeof(*table), GFP_KERNEL); + + if (!table) + return NULL; + + table->data_vec = kcalloc(sz, sizeof(*table->data_vec), GFP_KERNEL); + if (!table->data_vec) + goto err_free_table; + + mutex_init(&table->lock); + + table->sz = sz; + rwlock_init(&table->rwlock); + return table; + +err_free_table: + kfree(table); + return NULL; +} + +static void release_gid_table(struct ib_device *device, + struct ib_gid_table *table) +{ + bool leak = false; + int i; + + if (!table) + return; + + for (i = 0; i < table->sz; i++) { + if (is_gid_entry_free(table->data_vec[i])) + continue; + if (kref_read(&table->data_vec[i]->kref) > 1) { + dev_err(&device->dev, + "GID entry ref leak for index %d ref=%d\n", i, + kref_read(&table->data_vec[i]->kref)); + leak = true; + } + } + if (leak) + return; + + mutex_destroy(&table->lock); + kfree(table->data_vec); + kfree(table); +} + +static void cleanup_gid_table_port(struct ib_device *ib_dev, u8 port, + struct ib_gid_table *table) +{ + int i; + + if (!table) + return; + + mutex_lock(&table->lock); + for (i = 0; i < table->sz; ++i) { + if (is_gid_entry_valid(table->data_vec[i])) + del_gid(ib_dev, port, table, i); + } + mutex_unlock(&table->lock); +} + +void ib_cache_gid_set_default_gid(struct ib_device *ib_dev, u8 port, + struct net_device *ndev, + unsigned long gid_type_mask, + enum ib_cache_gid_default_mode mode) +{ + union ib_gid gid = { }; + struct ib_gid_attr gid_attr; + unsigned int gid_type; + unsigned long mask; + + mask = GID_ATTR_FIND_MASK_GID_TYPE | + GID_ATTR_FIND_MASK_DEFAULT | + GID_ATTR_FIND_MASK_NETDEV; + memset(&gid_attr, 0, sizeof(gid_attr)); + gid_attr.ndev = ndev; + + for (gid_type = 0; gid_type < IB_GID_TYPE_SIZE; ++gid_type) { + if (1UL << gid_type & ~gid_type_mask) + continue; + + gid_attr.gid_type = gid_type; + + if (mode == IB_CACHE_GID_DEFAULT_MODE_SET) { + make_default_gid(ndev, &gid); + __ib_cache_gid_add(ib_dev, port, &gid, + &gid_attr, mask, true); + } else if (mode == IB_CACHE_GID_DEFAULT_MODE_DELETE) { + _ib_cache_gid_del(ib_dev, port, &gid, + &gid_attr, mask, true); + } + } +} + +static void gid_table_reserve_default(struct ib_device *ib_dev, u8 port, + struct ib_gid_table *table) +{ + unsigned int i; + unsigned long roce_gid_type_mask; + unsigned int num_default_gids; + + roce_gid_type_mask = roce_gid_type_mask_support(ib_dev, port); + num_default_gids = hweight_long(roce_gid_type_mask); + /* Reserve starting indices for default GIDs */ + for (i = 0; i < num_default_gids && i < table->sz; i++) + table->default_gid_indices |= BIT(i); +} + + +static void gid_table_release_one(struct ib_device *ib_dev) +{ + unsigned int p; + + rdma_for_each_port (ib_dev, p) { + release_gid_table(ib_dev, ib_dev->port_data[p].cache.gid); + ib_dev->port_data[p].cache.gid = NULL; + } +} + +static int _gid_table_setup_one(struct ib_device *ib_dev) +{ + struct ib_gid_table *table; + unsigned int rdma_port; + + rdma_for_each_port (ib_dev, rdma_port) { + table = alloc_gid_table( + ib_dev->port_data[rdma_port].immutable.gid_tbl_len); + if (!table) + goto rollback_table_setup; + + gid_table_reserve_default(ib_dev, rdma_port, table); + ib_dev->port_data[rdma_port].cache.gid = table; + } + return 0; + +rollback_table_setup: + gid_table_release_one(ib_dev); + return -ENOMEM; +} + +static void gid_table_cleanup_one(struct ib_device *ib_dev) +{ + unsigned int p; + + rdma_for_each_port (ib_dev, p) + cleanup_gid_table_port(ib_dev, p, + ib_dev->port_data[p].cache.gid); +} + +static int gid_table_setup_one(struct ib_device *ib_dev) +{ + int err; + + err = _gid_table_setup_one(ib_dev); + + if (err) + return err; + + rdma_roce_rescan_device(ib_dev); + + return err; +} + +/** + * rdma_query_gid - Read the GID content from the GID software cache + * @device: Device to query the GID + * @port_num: Port number of the device + * @index: Index of the GID table entry to read + * @gid: Pointer to GID where to store the entry's GID + * + * rdma_query_gid() only reads the GID entry content for requested device, + * port and index. It reads for IB, RoCE and iWarp link layers. It doesn't + * hold any reference to the GID table entry in the HCA or software cache. + * + * Returns 0 on success or appropriate error code. + * + */ +int rdma_query_gid(struct ib_device *device, u8 port_num, + int index, union ib_gid *gid) +{ + struct ib_gid_table *table; + unsigned long flags; + int res = -EINVAL; + + if (!rdma_is_port_valid(device, port_num)) + return -EINVAL; + + table = rdma_gid_table(device, port_num); + read_lock_irqsave(&table->rwlock, flags); + + if (index < 0 || index >= table->sz || + !is_gid_entry_valid(table->data_vec[index])) + goto done; + + memcpy(gid, &table->data_vec[index]->attr.gid, sizeof(*gid)); + res = 0; + +done: + read_unlock_irqrestore(&table->rwlock, flags); + return res; +} +EXPORT_SYMBOL(rdma_query_gid); + +/** + * rdma_read_gid_hw_context - Read the HW GID context from GID attribute + * @attr: Potinter to the GID attribute + * + * rdma_read_gid_hw_context() reads the drivers GID HW context corresponding + * to the SGID attr. Callers are required to already be holding the reference + * to an existing GID entry. + * + * Returns the HW GID context + * + */ +void *rdma_read_gid_hw_context(const struct ib_gid_attr *attr) +{ + return container_of(attr, struct ib_gid_table_entry, attr)->context; +} +EXPORT_SYMBOL(rdma_read_gid_hw_context); + +/** + * rdma_find_gid - Returns SGID attributes if the matching GID is found. + * @device: The device to query. + * @gid: The GID value to search for. + * @gid_type: The GID type to search for. + * @ndev: In RoCE, the net device of the device. NULL means ignore. + * + * rdma_find_gid() searches for the specified GID value in the software cache. + * + * Returns GID attributes if a valid GID is found or returns ERR_PTR for the + * error. The caller must invoke rdma_put_gid_attr() to release the reference. + * + */ +const struct ib_gid_attr *rdma_find_gid(struct ib_device *device, + const union ib_gid *gid, + enum ib_gid_type gid_type, + struct net_device *ndev) +{ + unsigned long mask = GID_ATTR_FIND_MASK_GID | + GID_ATTR_FIND_MASK_GID_TYPE; + struct ib_gid_attr gid_attr_val = {.ndev = ndev, .gid_type = gid_type}; + unsigned int p; + + if (ndev) + mask |= GID_ATTR_FIND_MASK_NETDEV; + + rdma_for_each_port(device, p) { + struct ib_gid_table *table; + unsigned long flags; + int index; + + table = device->port_data[p].cache.gid; + read_lock_irqsave(&table->rwlock, flags); + index = find_gid(table, gid, &gid_attr_val, false, mask, NULL); + if (index >= 0) { + const struct ib_gid_attr *attr; + + get_gid_entry(table->data_vec[index]); + attr = &table->data_vec[index]->attr; + read_unlock_irqrestore(&table->rwlock, flags); + return attr; + } + read_unlock_irqrestore(&table->rwlock, flags); + } + + return ERR_PTR(-ENOENT); +} +EXPORT_SYMBOL(rdma_find_gid); + +int ib_get_cached_pkey(struct ib_device *device, + u8 port_num, + int index, + u16 *pkey) +{ + struct ib_pkey_cache *cache; + unsigned long flags; + int ret = 0; + + if (!rdma_is_port_valid(device, port_num)) + return -EINVAL; + + read_lock_irqsave(&device->cache_lock, flags); + + cache = device->port_data[port_num].cache.pkey; + + if (!cache || index < 0 || index >= cache->table_len) + ret = -EINVAL; + else + *pkey = cache->table[index]; + + read_unlock_irqrestore(&device->cache_lock, flags); + + return ret; +} +EXPORT_SYMBOL(ib_get_cached_pkey); + +int ib_get_cached_subnet_prefix(struct ib_device *device, + u8 port_num, + u64 *sn_pfx) +{ + unsigned long flags; + + if (!rdma_is_port_valid(device, port_num)) + return -EINVAL; + + read_lock_irqsave(&device->cache_lock, flags); + *sn_pfx = device->port_data[port_num].cache.subnet_prefix; + read_unlock_irqrestore(&device->cache_lock, flags); + + return 0; +} +EXPORT_SYMBOL(ib_get_cached_subnet_prefix); + +int ib_find_cached_pkey(struct ib_device *device, + u8 port_num, + u16 pkey, + u16 *index) +{ + struct ib_pkey_cache *cache; + unsigned long flags; + int i; + int ret = -ENOENT; + int partial_ix = -1; + + if (!rdma_is_port_valid(device, port_num)) + return -EINVAL; + + read_lock_irqsave(&device->cache_lock, flags); + + cache = device->port_data[port_num].cache.pkey; + if (!cache) { + ret = -EINVAL; + goto err; + } + + *index = -1; + + for (i = 0; i < cache->table_len; ++i) + if ((cache->table[i] & 0x7fff) == (pkey & 0x7fff)) { + if (cache->table[i] & 0x8000) { + *index = i; + ret = 0; + break; + } else + partial_ix = i; + } + + if (ret && partial_ix >= 0) { + *index = partial_ix; + ret = 0; + } + +err: + read_unlock_irqrestore(&device->cache_lock, flags); + + return ret; +} +EXPORT_SYMBOL(ib_find_cached_pkey); + +int ib_find_exact_cached_pkey(struct ib_device *device, + u8 port_num, + u16 pkey, + u16 *index) +{ + struct ib_pkey_cache *cache; + unsigned long flags; + int i; + int ret = -ENOENT; + + if (!rdma_is_port_valid(device, port_num)) + return -EINVAL; + + read_lock_irqsave(&device->cache_lock, flags); + + cache = device->port_data[port_num].cache.pkey; + if (!cache) { + ret = -EINVAL; + goto err; + } + + *index = -1; + + for (i = 0; i < cache->table_len; ++i) + if (cache->table[i] == pkey) { + *index = i; + ret = 0; + break; + } + +err: + read_unlock_irqrestore(&device->cache_lock, flags); + + return ret; +} +EXPORT_SYMBOL(ib_find_exact_cached_pkey); + +int ib_get_cached_lmc(struct ib_device *device, + u8 port_num, + u8 *lmc) +{ + unsigned long flags; + int ret = 0; + + if (!rdma_is_port_valid(device, port_num)) + return -EINVAL; + + read_lock_irqsave(&device->cache_lock, flags); + *lmc = device->port_data[port_num].cache.lmc; + read_unlock_irqrestore(&device->cache_lock, flags); + + return ret; +} +EXPORT_SYMBOL(ib_get_cached_lmc); + +int ib_get_cached_port_state(struct ib_device *device, + u8 port_num, + enum ib_port_state *port_state) +{ + unsigned long flags; + int ret = 0; + + if (!rdma_is_port_valid(device, port_num)) + return -EINVAL; + + read_lock_irqsave(&device->cache_lock, flags); + *port_state = device->port_data[port_num].cache.port_state; + read_unlock_irqrestore(&device->cache_lock, flags); + + return ret; +} +EXPORT_SYMBOL(ib_get_cached_port_state); + +/** + * rdma_get_gid_attr - Returns GID attributes for a port of a device + * at a requested gid_index, if a valid GID entry exists. + * @device: The device to query. + * @port_num: The port number on the device where the GID value + * is to be queried. + * @index: Index of the GID table entry whose attributes are to + * be queried. + * + * rdma_get_gid_attr() acquires reference count of gid attributes from the + * cached GID table. Caller must invoke rdma_put_gid_attr() to release + * reference to gid attribute regardless of link layer. + * + * Returns pointer to valid gid attribute or ERR_PTR for the appropriate error + * code. + */ +const struct ib_gid_attr * +rdma_get_gid_attr(struct ib_device *device, u8 port_num, int index) +{ + const struct ib_gid_attr *attr = ERR_PTR(-ENODATA); + struct ib_gid_table *table; + unsigned long flags; + + if (!rdma_is_port_valid(device, port_num)) + return ERR_PTR(-EINVAL); + + table = rdma_gid_table(device, port_num); + if (index < 0 || index >= table->sz) + return ERR_PTR(-EINVAL); + + read_lock_irqsave(&table->rwlock, flags); + if (!is_gid_entry_valid(table->data_vec[index])) + goto done; + + get_gid_entry(table->data_vec[index]); + attr = &table->data_vec[index]->attr; +done: + read_unlock_irqrestore(&table->rwlock, flags); + return attr; +} +EXPORT_SYMBOL(rdma_get_gid_attr); + +/** + * rdma_query_gid_table - Reads GID table entries of all the ports of a device up to max_entries. + * @device: The device to query. + * @entries: Entries where GID entries are returned. + * @max_entries: Maximum number of entries that can be returned. + * Entries array must be allocated to hold max_entries number of entries. + * @num_entries: Updated to the number of entries that were successfully read. + * + * Returns number of entries on success or appropriate error code. + */ +ssize_t rdma_query_gid_table(struct ib_device *device, + struct ib_uverbs_gid_entry *entries, + size_t max_entries) +{ + const struct ib_gid_attr *gid_attr; + ssize_t num_entries = 0, ret; + struct ib_gid_table *table; + unsigned int port_num, i; + struct net_device *ndev; + unsigned long flags; + + rdma_for_each_port(device, port_num) { + table = rdma_gid_table(device, port_num); + read_lock_irqsave(&table->rwlock, flags); + for (i = 0; i < table->sz; i++) { + if (!is_gid_entry_valid(table->data_vec[i])) + continue; + if (num_entries >= max_entries) { + ret = -EINVAL; + goto err; + } + + gid_attr = &table->data_vec[i]->attr; + + memcpy(&entries->gid, &gid_attr->gid, + sizeof(gid_attr->gid)); + entries->gid_index = gid_attr->index; + entries->port_num = gid_attr->port_num; + entries->gid_type = gid_attr->gid_type; + ndev = rcu_dereference_protected( + gid_attr->ndev, + lockdep_is_held(&table->rwlock)); + if (ndev) + entries->netdev_ifindex = ndev->ifindex; + + num_entries++; + entries++; + } + read_unlock_irqrestore(&table->rwlock, flags); + } + + return num_entries; +err: + read_unlock_irqrestore(&table->rwlock, flags); + return ret; +} +EXPORT_SYMBOL(rdma_query_gid_table); + +/** + * rdma_put_gid_attr - Release reference to the GID attribute + * @attr: Pointer to the GID attribute whose reference + * needs to be released. + * + * rdma_put_gid_attr() must be used to release reference whose + * reference is acquired using rdma_get_gid_attr() or any APIs + * which returns a pointer to the ib_gid_attr regardless of link layer + * of IB or RoCE. + * + */ +void rdma_put_gid_attr(const struct ib_gid_attr *attr) +{ + struct ib_gid_table_entry *entry = + container_of(attr, struct ib_gid_table_entry, attr); + + put_gid_entry(entry); +} +EXPORT_SYMBOL(rdma_put_gid_attr); + +/** + * rdma_hold_gid_attr - Get reference to existing GID attribute + * + * @attr: Pointer to the GID attribute whose reference + * needs to be taken. + * + * Increase the reference count to a GID attribute to keep it from being + * freed. Callers are required to already be holding a reference to attribute. + * + */ +void rdma_hold_gid_attr(const struct ib_gid_attr *attr) +{ + struct ib_gid_table_entry *entry = + container_of(attr, struct ib_gid_table_entry, attr); + + get_gid_entry(entry); +} +EXPORT_SYMBOL(rdma_hold_gid_attr); + +/** + * rdma_read_gid_attr_ndev_rcu - Read GID attribute netdevice + * which must be in UP state. + * + * @attr:Pointer to the GID attribute + * + * Returns pointer to netdevice if the netdevice was attached to GID and + * netdevice is in UP state. Caller must hold RCU lock as this API + * reads the netdev flags which can change while netdevice migrates to + * different net namespace. Returns ERR_PTR with error code otherwise. + * + */ +struct net_device *rdma_read_gid_attr_ndev_rcu(const struct ib_gid_attr *attr) +{ + struct ib_gid_table_entry *entry = + container_of(attr, struct ib_gid_table_entry, attr); + struct ib_device *device = entry->attr.device; + struct net_device *ndev = ERR_PTR(-EINVAL); + u8 port_num = entry->attr.port_num; + struct ib_gid_table *table; + unsigned long flags; + bool valid; + + table = rdma_gid_table(device, port_num); + + read_lock_irqsave(&table->rwlock, flags); + valid = is_gid_entry_valid(table->data_vec[attr->index]); + if (valid) { + ndev = rcu_dereference(attr->ndev); + if (!ndev) + ndev = ERR_PTR(-ENODEV); + } + read_unlock_irqrestore(&table->rwlock, flags); + return ndev; +} +EXPORT_SYMBOL(rdma_read_gid_attr_ndev_rcu); + +static int get_lower_dev_vlan(struct net_device *lower_dev, + struct netdev_nested_priv *priv) +{ + u16 *vlan_id = (u16 *)priv->data; + + if (is_vlan_dev(lower_dev)) + *vlan_id = vlan_dev_vlan_id(lower_dev); + + /* We are interested only in first level vlan device, so + * always return 1 to stop iterating over next level devices. + */ + return 1; +} + +/** + * rdma_read_gid_l2_fields - Read the vlan ID and source MAC address + * of a GID entry. + * + * @attr: GID attribute pointer whose L2 fields to be read + * @vlan_id: Pointer to vlan id to fill up if the GID entry has + * vlan id. It is optional. + * @smac: Pointer to smac to fill up for a GID entry. It is optional. + * + * rdma_read_gid_l2_fields() returns 0 on success and returns vlan id + * (if gid entry has vlan) and source MAC, or returns error. + */ +int rdma_read_gid_l2_fields(const struct ib_gid_attr *attr, + u16 *vlan_id, u8 *smac) +{ + struct netdev_nested_priv priv = { + .data = (void *)vlan_id, + }; + struct net_device *ndev; + + rcu_read_lock(); + ndev = rcu_dereference(attr->ndev); + if (!ndev) { + rcu_read_unlock(); + return -ENODEV; + } + if (smac) + ether_addr_copy(smac, ndev->dev_addr); + if (vlan_id) { + *vlan_id = 0xffff; + if (is_vlan_dev(ndev)) { + *vlan_id = vlan_dev_vlan_id(ndev); + } else { + /* If the netdev is upper device and if it's lower + * device is vlan device, consider vlan id of the + * the lower vlan device for this gid entry. + */ + netdev_walk_all_lower_dev_rcu(attr->ndev, + get_lower_dev_vlan, &priv); + } + } + rcu_read_unlock(); + return 0; +} +EXPORT_SYMBOL(rdma_read_gid_l2_fields); + +static int config_non_roce_gid_cache(struct ib_device *device, + u8 port, int gid_tbl_len) +{ + struct ib_gid_attr gid_attr = {}; + struct ib_gid_table *table; + int ret = 0; + int i; + + gid_attr.device = device; + gid_attr.port_num = port; + table = rdma_gid_table(device, port); + + mutex_lock(&table->lock); + for (i = 0; i < gid_tbl_len; ++i) { + if (!device->ops.query_gid) + continue; + ret = device->ops.query_gid(device, port, i, &gid_attr.gid); + if (ret) { + dev_warn(&device->dev, + "query_gid failed (%d) for index %d\n", ret, + i); + goto err; + } + gid_attr.index = i; + add_modify_gid(table, &gid_attr); + } +err: + mutex_unlock(&table->lock); + return ret; +} + +static int +ib_cache_update(struct ib_device *device, u8 port, bool enforce_security) +{ + struct ib_port_attr *tprops = NULL; + struct ib_pkey_cache *pkey_cache = NULL, *old_pkey_cache; + int i; + int ret; + + if (!rdma_is_port_valid(device, port)) + return -EINVAL; + + tprops = kmalloc(sizeof *tprops, GFP_KERNEL); + if (!tprops) + return -ENOMEM; + + ret = ib_query_port(device, port, tprops); + if (ret) { + dev_warn(&device->dev, "ib_query_port failed (%d)\n", ret); + goto err; + } + + if (!rdma_protocol_roce(device, port)) { + ret = config_non_roce_gid_cache(device, port, + tprops->gid_tbl_len); + if (ret) + goto err; + } + + if (tprops->pkey_tbl_len) { + pkey_cache = kmalloc(struct_size(pkey_cache, table, + tprops->pkey_tbl_len), + GFP_KERNEL); + if (!pkey_cache) { + ret = -ENOMEM; + goto err; + } + + pkey_cache->table_len = tprops->pkey_tbl_len; + + for (i = 0; i < pkey_cache->table_len; ++i) { + ret = ib_query_pkey(device, port, i, + pkey_cache->table + i); + if (ret) { + dev_warn(&device->dev, + "ib_query_pkey failed (%d) for index %d\n", + ret, i); + goto err; + } + } + } + + write_lock_irq(&device->cache_lock); + + old_pkey_cache = device->port_data[port].cache.pkey; + + device->port_data[port].cache.pkey = pkey_cache; + device->port_data[port].cache.lmc = tprops->lmc; + device->port_data[port].cache.port_state = tprops->state; + + device->port_data[port].cache.subnet_prefix = tprops->subnet_prefix; + write_unlock_irq(&device->cache_lock); + + if (enforce_security) + ib_security_cache_change(device, + port, + tprops->subnet_prefix); + + kfree(old_pkey_cache); + kfree(tprops); + return 0; + +err: + kfree(pkey_cache); + kfree(tprops); + return ret; +} + +static void ib_cache_event_task(struct work_struct *_work) +{ + struct ib_update_work *work = + container_of(_work, struct ib_update_work, work); + int ret; + + /* Before distributing the cache update event, first sync + * the cache. + */ + ret = ib_cache_update(work->event.device, work->event.element.port_num, + work->enforce_security); + + /* GID event is notified already for individual GID entries by + * dispatch_gid_change_event(). Hence, notifiy for rest of the + * events. + */ + if (!ret && work->event.event != IB_EVENT_GID_CHANGE) + ib_dispatch_event_clients(&work->event); + + kfree(work); +} + +static void ib_generic_event_task(struct work_struct *_work) +{ + struct ib_update_work *work = + container_of(_work, struct ib_update_work, work); + + ib_dispatch_event_clients(&work->event); + kfree(work); +} + +static bool is_cache_update_event(const struct ib_event *event) +{ + return (event->event == IB_EVENT_PORT_ERR || + event->event == IB_EVENT_PORT_ACTIVE || + event->event == IB_EVENT_LID_CHANGE || + event->event == IB_EVENT_PKEY_CHANGE || + event->event == IB_EVENT_CLIENT_REREGISTER || + event->event == IB_EVENT_GID_CHANGE); +} + +/** + * ib_dispatch_event - Dispatch an asynchronous event + * @event:Event to dispatch + * + * Low-level drivers must call ib_dispatch_event() to dispatch the + * event to all registered event handlers when an asynchronous event + * occurs. + */ +void ib_dispatch_event(const struct ib_event *event) +{ + struct ib_update_work *work; + + work = kzalloc(sizeof(*work), GFP_ATOMIC); + if (!work) + return; + + if (is_cache_update_event(event)) + INIT_WORK(&work->work, ib_cache_event_task); + else + INIT_WORK(&work->work, ib_generic_event_task); + + work->event = *event; + if (event->event == IB_EVENT_PKEY_CHANGE || + event->event == IB_EVENT_GID_CHANGE) + work->enforce_security = true; + + queue_work(ib_wq, &work->work); +} +EXPORT_SYMBOL(ib_dispatch_event); + +int ib_cache_setup_one(struct ib_device *device) +{ + unsigned int p; + int err; + + rwlock_init(&device->cache_lock); + + err = gid_table_setup_one(device); + if (err) + return err; + + rdma_for_each_port (device, p) { + err = ib_cache_update(device, p, true); + if (err) + return err; + } + + return 0; +} + +void ib_cache_release_one(struct ib_device *device) +{ + unsigned int p; + + /* + * The release function frees all the cache elements. + * This function should be called as part of freeing + * all the device's resources when the cache could no + * longer be accessed. + */ + rdma_for_each_port (device, p) + kfree(device->port_data[p].cache.pkey); + + gid_table_release_one(device); +} + +void ib_cache_cleanup_one(struct ib_device *device) +{ + /* The cleanup function waits for all in-progress workqueue + * elements and cleans up the GID cache. This function should be + * called after the device was removed from the devices list and + * all clients were removed, so the cache exists but is + * non-functional and shouldn't be updated anymore. + */ + flush_workqueue(ib_wq); + gid_table_cleanup_one(device); + + /* + * Flush the wq second time for any pending GID delete work. + */ + flush_workqueue(ib_wq); +} diff --git a/drivers/infiniband/core/cgroup.c b/drivers/infiniband/core/cgroup.c new file mode 100644 index 000000000..1f037fe01 --- /dev/null +++ b/drivers/infiniband/core/cgroup.c @@ -0,0 +1,53 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * Copyright (C) 2016 Parav Pandit <pandit.parav@gmail.com> + */ + +#include "core_priv.h" + +/** + * ib_device_register_rdmacg - register with rdma cgroup. + * @device: device to register to participate in resource + * accounting by rdma cgroup. + * + * Register with the rdma cgroup. Should be called before + * exposing rdma device to user space applications to avoid + * resource accounting leak. + */ +void ib_device_register_rdmacg(struct ib_device *device) +{ + device->cg_device.name = device->name; + rdmacg_register_device(&device->cg_device); +} + +/** + * ib_device_unregister_rdmacg - unregister with rdma cgroup. + * @device: device to unregister. + * + * Unregister with the rdma cgroup. Should be called after + * all the resources are deallocated, and after a stage when any + * other resource allocation by user application cannot be done + * for this device to avoid any leak in accounting. + */ +void ib_device_unregister_rdmacg(struct ib_device *device) +{ + rdmacg_unregister_device(&device->cg_device); +} + +int ib_rdmacg_try_charge(struct ib_rdmacg_object *cg_obj, + struct ib_device *device, + enum rdmacg_resource_type resource_index) +{ + return rdmacg_try_charge(&cg_obj->cg, &device->cg_device, + resource_index); +} +EXPORT_SYMBOL(ib_rdmacg_try_charge); + +void ib_rdmacg_uncharge(struct ib_rdmacg_object *cg_obj, + struct ib_device *device, + enum rdmacg_resource_type resource_index) +{ + rdmacg_uncharge(cg_obj->cg, &device->cg_device, + resource_index); +} +EXPORT_SYMBOL(ib_rdmacg_uncharge); diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c new file mode 100644 index 000000000..db1a25fbe --- /dev/null +++ b/drivers/infiniband/core/cm.c @@ -0,0 +1,4534 @@ +// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB +/* + * Copyright (c) 2004-2007 Intel Corporation. All rights reserved. + * Copyright (c) 2004 Topspin Corporation. All rights reserved. + * Copyright (c) 2004, 2005 Voltaire Corporation. All rights reserved. + * Copyright (c) 2005 Sun Microsystems, Inc. All rights reserved. + * Copyright (c) 2019, Mellanox Technologies inc. All rights reserved. + */ + +#include <linux/completion.h> +#include <linux/dma-mapping.h> +#include <linux/device.h> +#include <linux/module.h> +#include <linux/err.h> +#include <linux/idr.h> +#include <linux/interrupt.h> +#include <linux/random.h> +#include <linux/rbtree.h> +#include <linux/spinlock.h> +#include <linux/slab.h> +#include <linux/sysfs.h> +#include <linux/workqueue.h> +#include <linux/kdev_t.h> +#include <linux/etherdevice.h> + +#include <rdma/ib_cache.h> +#include <rdma/ib_cm.h> +#include "cm_msgs.h" +#include "core_priv.h" +#include "cm_trace.h" + +MODULE_AUTHOR("Sean Hefty"); +MODULE_DESCRIPTION("InfiniBand CM"); +MODULE_LICENSE("Dual BSD/GPL"); + +static const char * const ibcm_rej_reason_strs[] = { + [IB_CM_REJ_NO_QP] = "no QP", + [IB_CM_REJ_NO_EEC] = "no EEC", + [IB_CM_REJ_NO_RESOURCES] = "no resources", + [IB_CM_REJ_TIMEOUT] = "timeout", + [IB_CM_REJ_UNSUPPORTED] = "unsupported", + [IB_CM_REJ_INVALID_COMM_ID] = "invalid comm ID", + [IB_CM_REJ_INVALID_COMM_INSTANCE] = "invalid comm instance", + [IB_CM_REJ_INVALID_SERVICE_ID] = "invalid service ID", + [IB_CM_REJ_INVALID_TRANSPORT_TYPE] = "invalid transport type", + [IB_CM_REJ_STALE_CONN] = "stale conn", + [IB_CM_REJ_RDC_NOT_EXIST] = "RDC not exist", + [IB_CM_REJ_INVALID_GID] = "invalid GID", + [IB_CM_REJ_INVALID_LID] = "invalid LID", + [IB_CM_REJ_INVALID_SL] = "invalid SL", + [IB_CM_REJ_INVALID_TRAFFIC_CLASS] = "invalid traffic class", + [IB_CM_REJ_INVALID_HOP_LIMIT] = "invalid hop limit", + [IB_CM_REJ_INVALID_PACKET_RATE] = "invalid packet rate", + [IB_CM_REJ_INVALID_ALT_GID] = "invalid alt GID", + [IB_CM_REJ_INVALID_ALT_LID] = "invalid alt LID", + [IB_CM_REJ_INVALID_ALT_SL] = "invalid alt SL", + [IB_CM_REJ_INVALID_ALT_TRAFFIC_CLASS] = "invalid alt traffic class", + [IB_CM_REJ_INVALID_ALT_HOP_LIMIT] = "invalid alt hop limit", + [IB_CM_REJ_INVALID_ALT_PACKET_RATE] = "invalid alt packet rate", + [IB_CM_REJ_PORT_CM_REDIRECT] = "port CM redirect", + [IB_CM_REJ_PORT_REDIRECT] = "port redirect", + [IB_CM_REJ_INVALID_MTU] = "invalid MTU", + [IB_CM_REJ_INSUFFICIENT_RESP_RESOURCES] = "insufficient resp resources", + [IB_CM_REJ_CONSUMER_DEFINED] = "consumer defined", + [IB_CM_REJ_INVALID_RNR_RETRY] = "invalid RNR retry", + [IB_CM_REJ_DUPLICATE_LOCAL_COMM_ID] = "duplicate local comm ID", + [IB_CM_REJ_INVALID_CLASS_VERSION] = "invalid class version", + [IB_CM_REJ_INVALID_FLOW_LABEL] = "invalid flow label", + [IB_CM_REJ_INVALID_ALT_FLOW_LABEL] = "invalid alt flow label", + [IB_CM_REJ_VENDOR_OPTION_NOT_SUPPORTED] = + "vendor option is not supported", +}; + +const char *__attribute_const__ ibcm_reject_msg(int reason) +{ + size_t index = reason; + + if (index < ARRAY_SIZE(ibcm_rej_reason_strs) && + ibcm_rej_reason_strs[index]) + return ibcm_rej_reason_strs[index]; + else + return "unrecognized reason"; +} +EXPORT_SYMBOL(ibcm_reject_msg); + +struct cm_id_private; +struct cm_work; +static int cm_add_one(struct ib_device *device); +static void cm_remove_one(struct ib_device *device, void *client_data); +static void cm_process_work(struct cm_id_private *cm_id_priv, + struct cm_work *work); +static int cm_send_sidr_rep_locked(struct cm_id_private *cm_id_priv, + struct ib_cm_sidr_rep_param *param); +static int cm_send_dreq_locked(struct cm_id_private *cm_id_priv, + const void *private_data, u8 private_data_len); +static int cm_send_drep_locked(struct cm_id_private *cm_id_priv, + void *private_data, u8 private_data_len); +static int cm_send_rej_locked(struct cm_id_private *cm_id_priv, + enum ib_cm_rej_reason reason, void *ari, + u8 ari_length, const void *private_data, + u8 private_data_len); + +static struct ib_client cm_client = { + .name = "cm", + .add = cm_add_one, + .remove = cm_remove_one +}; + +static struct ib_cm { + spinlock_t lock; + struct list_head device_list; + rwlock_t device_lock; + struct rb_root listen_service_table; + u64 listen_service_id; + /* struct rb_root peer_service_table; todo: fix peer to peer */ + struct rb_root remote_qp_table; + struct rb_root remote_id_table; + struct rb_root remote_sidr_table; + struct xarray local_id_table; + u32 local_id_next; + __be32 random_id_operand; + struct list_head timewait_list; + struct workqueue_struct *wq; + /* Sync on cm change port state */ + spinlock_t state_lock; +} cm; + +/* Counter indexes ordered by attribute ID */ +enum { + CM_REQ_COUNTER, + CM_MRA_COUNTER, + CM_REJ_COUNTER, + CM_REP_COUNTER, + CM_RTU_COUNTER, + CM_DREQ_COUNTER, + CM_DREP_COUNTER, + CM_SIDR_REQ_COUNTER, + CM_SIDR_REP_COUNTER, + CM_LAP_COUNTER, + CM_APR_COUNTER, + CM_ATTR_COUNT, + CM_ATTR_ID_OFFSET = 0x0010, +}; + +enum { + CM_XMIT, + CM_XMIT_RETRIES, + CM_RECV, + CM_RECV_DUPLICATES, + CM_COUNTER_GROUPS +}; + +static char const counter_group_names[CM_COUNTER_GROUPS] + [sizeof("cm_rx_duplicates")] = { + "cm_tx_msgs", "cm_tx_retries", + "cm_rx_msgs", "cm_rx_duplicates" +}; + +struct cm_counter_group { + struct kobject obj; + atomic_long_t counter[CM_ATTR_COUNT]; +}; + +struct cm_counter_attribute { + struct attribute attr; + int index; +}; + +#define CM_COUNTER_ATTR(_name, _index) \ +struct cm_counter_attribute cm_##_name##_counter_attr = { \ + .attr = { .name = __stringify(_name), .mode = 0444 }, \ + .index = _index \ +} + +static CM_COUNTER_ATTR(req, CM_REQ_COUNTER); +static CM_COUNTER_ATTR(mra, CM_MRA_COUNTER); +static CM_COUNTER_ATTR(rej, CM_REJ_COUNTER); +static CM_COUNTER_ATTR(rep, CM_REP_COUNTER); +static CM_COUNTER_ATTR(rtu, CM_RTU_COUNTER); +static CM_COUNTER_ATTR(dreq, CM_DREQ_COUNTER); +static CM_COUNTER_ATTR(drep, CM_DREP_COUNTER); +static CM_COUNTER_ATTR(sidr_req, CM_SIDR_REQ_COUNTER); +static CM_COUNTER_ATTR(sidr_rep, CM_SIDR_REP_COUNTER); +static CM_COUNTER_ATTR(lap, CM_LAP_COUNTER); +static CM_COUNTER_ATTR(apr, CM_APR_COUNTER); + +static struct attribute *cm_counter_default_attrs[] = { + &cm_req_counter_attr.attr, + &cm_mra_counter_attr.attr, + &cm_rej_counter_attr.attr, + &cm_rep_counter_attr.attr, + &cm_rtu_counter_attr.attr, + &cm_dreq_counter_attr.attr, + &cm_drep_counter_attr.attr, + &cm_sidr_req_counter_attr.attr, + &cm_sidr_rep_counter_attr.attr, + &cm_lap_counter_attr.attr, + &cm_apr_counter_attr.attr, + NULL +}; + +struct cm_port { + struct cm_device *cm_dev; + struct ib_mad_agent *mad_agent; + u8 port_num; + struct list_head cm_priv_prim_list; + struct list_head cm_priv_altr_list; + struct cm_counter_group counter_group[CM_COUNTER_GROUPS]; +}; + +struct cm_device { + struct list_head list; + struct ib_device *ib_device; + u8 ack_delay; + int going_down; + struct cm_port *port[]; +}; + +struct cm_av { + struct cm_port *port; + union ib_gid dgid; + struct rdma_ah_attr ah_attr; + u16 pkey_index; + u8 timeout; +}; + +struct cm_work { + struct delayed_work work; + struct list_head list; + struct cm_port *port; + struct ib_mad_recv_wc *mad_recv_wc; /* Received MADs */ + __be32 local_id; /* Established / timewait */ + __be32 remote_id; + struct ib_cm_event cm_event; + struct sa_path_rec path[]; +}; + +struct cm_timewait_info { + struct cm_work work; + struct list_head list; + struct rb_node remote_qp_node; + struct rb_node remote_id_node; + __be64 remote_ca_guid; + __be32 remote_qpn; + u8 inserted_remote_qp; + u8 inserted_remote_id; +}; + +struct cm_id_private { + struct ib_cm_id id; + + struct rb_node service_node; + struct rb_node sidr_id_node; + spinlock_t lock; /* Do not acquire inside cm.lock */ + struct completion comp; + refcount_t refcount; + /* Number of clients sharing this ib_cm_id. Only valid for listeners. + * Protected by the cm.lock spinlock. */ + int listen_sharecount; + struct rcu_head rcu; + + struct ib_mad_send_buf *msg; + struct cm_timewait_info *timewait_info; + /* todo: use alternate port on send failure */ + struct cm_av av; + struct cm_av alt_av; + + void *private_data; + __be64 tid; + __be32 local_qpn; + __be32 remote_qpn; + enum ib_qp_type qp_type; + __be32 sq_psn; + __be32 rq_psn; + int timeout_ms; + enum ib_mtu path_mtu; + __be16 pkey; + u8 private_data_len; + u8 max_cm_retries; + u8 responder_resources; + u8 initiator_depth; + u8 retry_count; + u8 rnr_retry_count; + u8 service_timeout; + u8 target_ack_delay; + + struct list_head prim_list; + struct list_head altr_list; + /* Indicates that the send port mad is registered and av is set */ + int prim_send_port_not_ready; + int altr_send_port_not_ready; + + struct list_head work_list; + atomic_t work_count; + + struct rdma_ucm_ece ece; +}; + +static void cm_work_handler(struct work_struct *work); + +static inline void cm_deref_id(struct cm_id_private *cm_id_priv) +{ + if (refcount_dec_and_test(&cm_id_priv->refcount)) + complete(&cm_id_priv->comp); +} + +static int cm_alloc_msg(struct cm_id_private *cm_id_priv, + struct ib_mad_send_buf **msg) +{ + struct ib_mad_agent *mad_agent; + struct ib_mad_send_buf *m; + struct ib_ah *ah; + struct cm_av *av; + unsigned long flags, flags2; + int ret = 0; + + /* don't let the port to be released till the agent is down */ + spin_lock_irqsave(&cm.state_lock, flags2); + spin_lock_irqsave(&cm.lock, flags); + if (!cm_id_priv->prim_send_port_not_ready) + av = &cm_id_priv->av; + else if (!cm_id_priv->altr_send_port_not_ready && + (cm_id_priv->alt_av.port)) + av = &cm_id_priv->alt_av; + else { + pr_info("%s: not valid CM id\n", __func__); + ret = -ENODEV; + spin_unlock_irqrestore(&cm.lock, flags); + goto out; + } + spin_unlock_irqrestore(&cm.lock, flags); + /* Make sure the port haven't released the mad yet */ + mad_agent = cm_id_priv->av.port->mad_agent; + if (!mad_agent) { + pr_info("%s: not a valid MAD agent\n", __func__); + ret = -ENODEV; + goto out; + } + ah = rdma_create_ah(mad_agent->qp->pd, &av->ah_attr, 0); + if (IS_ERR(ah)) { + ret = PTR_ERR(ah); + goto out; + } + + m = ib_create_send_mad(mad_agent, cm_id_priv->id.remote_cm_qpn, + av->pkey_index, + 0, IB_MGMT_MAD_HDR, IB_MGMT_MAD_DATA, + GFP_ATOMIC, + IB_MGMT_BASE_VERSION); + if (IS_ERR(m)) { + rdma_destroy_ah(ah, 0); + ret = PTR_ERR(m); + goto out; + } + + /* Timeout set by caller if response is expected. */ + m->ah = ah; + m->retries = cm_id_priv->max_cm_retries; + + refcount_inc(&cm_id_priv->refcount); + m->context[0] = cm_id_priv; + *msg = m; + +out: + spin_unlock_irqrestore(&cm.state_lock, flags2); + return ret; +} + +static struct ib_mad_send_buf *cm_alloc_response_msg_no_ah(struct cm_port *port, + struct ib_mad_recv_wc *mad_recv_wc) +{ + return ib_create_send_mad(port->mad_agent, 1, mad_recv_wc->wc->pkey_index, + 0, IB_MGMT_MAD_HDR, IB_MGMT_MAD_DATA, + GFP_ATOMIC, + IB_MGMT_BASE_VERSION); +} + +static int cm_create_response_msg_ah(struct cm_port *port, + struct ib_mad_recv_wc *mad_recv_wc, + struct ib_mad_send_buf *msg) +{ + struct ib_ah *ah; + + ah = ib_create_ah_from_wc(port->mad_agent->qp->pd, mad_recv_wc->wc, + mad_recv_wc->recv_buf.grh, port->port_num); + if (IS_ERR(ah)) + return PTR_ERR(ah); + + msg->ah = ah; + return 0; +} + +static void cm_free_msg(struct ib_mad_send_buf *msg) +{ + if (msg->ah) + rdma_destroy_ah(msg->ah, 0); + if (msg->context[0]) + cm_deref_id(msg->context[0]); + ib_free_send_mad(msg); +} + +static int cm_alloc_response_msg(struct cm_port *port, + struct ib_mad_recv_wc *mad_recv_wc, + struct ib_mad_send_buf **msg) +{ + struct ib_mad_send_buf *m; + int ret; + + m = cm_alloc_response_msg_no_ah(port, mad_recv_wc); + if (IS_ERR(m)) + return PTR_ERR(m); + + ret = cm_create_response_msg_ah(port, mad_recv_wc, m); + if (ret) { + cm_free_msg(m); + return ret; + } + + *msg = m; + return 0; +} + +static void * cm_copy_private_data(const void *private_data, + u8 private_data_len) +{ + void *data; + + if (!private_data || !private_data_len) + return NULL; + + data = kmemdup(private_data, private_data_len, GFP_KERNEL); + if (!data) + return ERR_PTR(-ENOMEM); + + return data; +} + +static void cm_set_private_data(struct cm_id_private *cm_id_priv, + void *private_data, u8 private_data_len) +{ + if (cm_id_priv->private_data && cm_id_priv->private_data_len) + kfree(cm_id_priv->private_data); + + cm_id_priv->private_data = private_data; + cm_id_priv->private_data_len = private_data_len; +} + +static int cm_init_av_for_lap(struct cm_port *port, struct ib_wc *wc, + struct ib_grh *grh, struct cm_av *av) +{ + struct rdma_ah_attr new_ah_attr; + int ret; + + av->port = port; + av->pkey_index = wc->pkey_index; + + /* + * av->ah_attr might be initialized based on past wc during incoming + * connect request or while sending out connect request. So initialize + * a new ah_attr on stack. If initialization fails, old ah_attr is + * used for sending any responses. If initialization is successful, + * than new ah_attr is used by overwriting old one. + */ + ret = ib_init_ah_attr_from_wc(port->cm_dev->ib_device, + port->port_num, wc, + grh, &new_ah_attr); + if (ret) + return ret; + + rdma_move_ah_attr(&av->ah_attr, &new_ah_attr); + return 0; +} + +static int cm_init_av_for_response(struct cm_port *port, struct ib_wc *wc, + struct ib_grh *grh, struct cm_av *av) +{ + av->port = port; + av->pkey_index = wc->pkey_index; + return ib_init_ah_attr_from_wc(port->cm_dev->ib_device, + port->port_num, wc, + grh, &av->ah_attr); +} + +static void add_cm_id_to_port_list(struct cm_id_private *cm_id_priv, + struct cm_av *av, struct cm_port *port) +{ + unsigned long flags; + + spin_lock_irqsave(&cm.lock, flags); + if (&cm_id_priv->av == av) + list_add_tail(&cm_id_priv->prim_list, &port->cm_priv_prim_list); + else if (&cm_id_priv->alt_av == av) + list_add_tail(&cm_id_priv->altr_list, &port->cm_priv_altr_list); + else + WARN_ON(true); + spin_unlock_irqrestore(&cm.lock, flags); +} + +static struct cm_port * +get_cm_port_from_path(struct sa_path_rec *path, const struct ib_gid_attr *attr) +{ + struct cm_device *cm_dev; + struct cm_port *port = NULL; + unsigned long flags; + + if (attr) { + read_lock_irqsave(&cm.device_lock, flags); + list_for_each_entry(cm_dev, &cm.device_list, list) { + if (cm_dev->ib_device == attr->device) { + port = cm_dev->port[attr->port_num - 1]; + break; + } + } + read_unlock_irqrestore(&cm.device_lock, flags); + } else { + /* SGID attribute can be NULL in following + * conditions. + * (a) Alternative path + * (b) IB link layer without GRH + * (c) LAP send messages + */ + read_lock_irqsave(&cm.device_lock, flags); + list_for_each_entry(cm_dev, &cm.device_list, list) { + attr = rdma_find_gid(cm_dev->ib_device, + &path->sgid, + sa_conv_pathrec_to_gid_type(path), + NULL); + if (!IS_ERR(attr)) { + port = cm_dev->port[attr->port_num - 1]; + break; + } + } + read_unlock_irqrestore(&cm.device_lock, flags); + if (port) + rdma_put_gid_attr(attr); + } + return port; +} + +static int cm_init_av_by_path(struct sa_path_rec *path, + const struct ib_gid_attr *sgid_attr, + struct cm_av *av, + struct cm_id_private *cm_id_priv) +{ + struct rdma_ah_attr new_ah_attr; + struct cm_device *cm_dev; + struct cm_port *port; + int ret; + + port = get_cm_port_from_path(path, sgid_attr); + if (!port) + return -EINVAL; + cm_dev = port->cm_dev; + + ret = ib_find_cached_pkey(cm_dev->ib_device, port->port_num, + be16_to_cpu(path->pkey), &av->pkey_index); + if (ret) + return ret; + + av->port = port; + + /* + * av->ah_attr might be initialized based on wc or during + * request processing time which might have reference to sgid_attr. + * So initialize a new ah_attr on stack. + * If initialization fails, old ah_attr is used for sending any + * responses. If initialization is successful, than new ah_attr + * is used by overwriting the old one. So that right ah_attr + * can be used to return an error response. + */ + ret = ib_init_ah_attr_from_path(cm_dev->ib_device, port->port_num, path, + &new_ah_attr, sgid_attr); + if (ret) + return ret; + + av->timeout = path->packet_life_time + 1; + add_cm_id_to_port_list(cm_id_priv, av, port); + rdma_move_ah_attr(&av->ah_attr, &new_ah_attr); + return 0; +} + +static u32 cm_local_id(__be32 local_id) +{ + return (__force u32) (local_id ^ cm.random_id_operand); +} + +static struct cm_id_private *cm_acquire_id(__be32 local_id, __be32 remote_id) +{ + struct cm_id_private *cm_id_priv; + + rcu_read_lock(); + cm_id_priv = xa_load(&cm.local_id_table, cm_local_id(local_id)); + if (!cm_id_priv || cm_id_priv->id.remote_id != remote_id || + !refcount_inc_not_zero(&cm_id_priv->refcount)) + cm_id_priv = NULL; + rcu_read_unlock(); + + return cm_id_priv; +} + +/* + * Trivial helpers to strip endian annotation and compare; the + * endianness doesn't actually matter since we just need a stable + * order for the RB tree. + */ +static int be32_lt(__be32 a, __be32 b) +{ + return (__force u32) a < (__force u32) b; +} + +static int be32_gt(__be32 a, __be32 b) +{ + return (__force u32) a > (__force u32) b; +} + +static int be64_lt(__be64 a, __be64 b) +{ + return (__force u64) a < (__force u64) b; +} + +static int be64_gt(__be64 a, __be64 b) +{ + return (__force u64) a > (__force u64) b; +} + +/* + * Inserts a new cm_id_priv into the listen_service_table. Returns cm_id_priv + * if the new ID was inserted, NULL if it could not be inserted due to a + * collision, or the existing cm_id_priv ready for shared usage. + */ +static struct cm_id_private *cm_insert_listen(struct cm_id_private *cm_id_priv, + ib_cm_handler shared_handler) +{ + struct rb_node **link = &cm.listen_service_table.rb_node; + struct rb_node *parent = NULL; + struct cm_id_private *cur_cm_id_priv; + __be64 service_id = cm_id_priv->id.service_id; + __be64 service_mask = cm_id_priv->id.service_mask; + unsigned long flags; + + spin_lock_irqsave(&cm.lock, flags); + while (*link) { + parent = *link; + cur_cm_id_priv = rb_entry(parent, struct cm_id_private, + service_node); + if ((cur_cm_id_priv->id.service_mask & service_id) == + (service_mask & cur_cm_id_priv->id.service_id) && + (cm_id_priv->id.device == cur_cm_id_priv->id.device)) { + /* + * Sharing an ib_cm_id with different handlers is not + * supported + */ + if (cur_cm_id_priv->id.cm_handler != shared_handler || + cur_cm_id_priv->id.context || + WARN_ON(!cur_cm_id_priv->id.cm_handler)) { + spin_unlock_irqrestore(&cm.lock, flags); + return NULL; + } + refcount_inc(&cur_cm_id_priv->refcount); + cur_cm_id_priv->listen_sharecount++; + spin_unlock_irqrestore(&cm.lock, flags); + return cur_cm_id_priv; + } + + if (cm_id_priv->id.device < cur_cm_id_priv->id.device) + link = &(*link)->rb_left; + else if (cm_id_priv->id.device > cur_cm_id_priv->id.device) + link = &(*link)->rb_right; + else if (be64_lt(service_id, cur_cm_id_priv->id.service_id)) + link = &(*link)->rb_left; + else if (be64_gt(service_id, cur_cm_id_priv->id.service_id)) + link = &(*link)->rb_right; + else + link = &(*link)->rb_right; + } + cm_id_priv->listen_sharecount++; + rb_link_node(&cm_id_priv->service_node, parent, link); + rb_insert_color(&cm_id_priv->service_node, &cm.listen_service_table); + spin_unlock_irqrestore(&cm.lock, flags); + return cm_id_priv; +} + +static struct cm_id_private * cm_find_listen(struct ib_device *device, + __be64 service_id) +{ + struct rb_node *node = cm.listen_service_table.rb_node; + struct cm_id_private *cm_id_priv; + + while (node) { + cm_id_priv = rb_entry(node, struct cm_id_private, service_node); + if ((cm_id_priv->id.service_mask & service_id) == + cm_id_priv->id.service_id && + (cm_id_priv->id.device == device)) { + refcount_inc(&cm_id_priv->refcount); + return cm_id_priv; + } + if (device < cm_id_priv->id.device) + node = node->rb_left; + else if (device > cm_id_priv->id.device) + node = node->rb_right; + else if (be64_lt(service_id, cm_id_priv->id.service_id)) + node = node->rb_left; + else if (be64_gt(service_id, cm_id_priv->id.service_id)) + node = node->rb_right; + else + node = node->rb_right; + } + return NULL; +} + +static struct cm_timewait_info * cm_insert_remote_id(struct cm_timewait_info + *timewait_info) +{ + struct rb_node **link = &cm.remote_id_table.rb_node; + struct rb_node *parent = NULL; + struct cm_timewait_info *cur_timewait_info; + __be64 remote_ca_guid = timewait_info->remote_ca_guid; + __be32 remote_id = timewait_info->work.remote_id; + + while (*link) { + parent = *link; + cur_timewait_info = rb_entry(parent, struct cm_timewait_info, + remote_id_node); + if (be32_lt(remote_id, cur_timewait_info->work.remote_id)) + link = &(*link)->rb_left; + else if (be32_gt(remote_id, cur_timewait_info->work.remote_id)) + link = &(*link)->rb_right; + else if (be64_lt(remote_ca_guid, cur_timewait_info->remote_ca_guid)) + link = &(*link)->rb_left; + else if (be64_gt(remote_ca_guid, cur_timewait_info->remote_ca_guid)) + link = &(*link)->rb_right; + else + return cur_timewait_info; + } + timewait_info->inserted_remote_id = 1; + rb_link_node(&timewait_info->remote_id_node, parent, link); + rb_insert_color(&timewait_info->remote_id_node, &cm.remote_id_table); + return NULL; +} + +static struct cm_id_private *cm_find_remote_id(__be64 remote_ca_guid, + __be32 remote_id) +{ + struct rb_node *node = cm.remote_id_table.rb_node; + struct cm_timewait_info *timewait_info; + struct cm_id_private *res = NULL; + + spin_lock_irq(&cm.lock); + while (node) { + timewait_info = rb_entry(node, struct cm_timewait_info, + remote_id_node); + if (be32_lt(remote_id, timewait_info->work.remote_id)) + node = node->rb_left; + else if (be32_gt(remote_id, timewait_info->work.remote_id)) + node = node->rb_right; + else if (be64_lt(remote_ca_guid, timewait_info->remote_ca_guid)) + node = node->rb_left; + else if (be64_gt(remote_ca_guid, timewait_info->remote_ca_guid)) + node = node->rb_right; + else { + res = cm_acquire_id(timewait_info->work.local_id, + timewait_info->work.remote_id); + break; + } + } + spin_unlock_irq(&cm.lock); + return res; +} + +static struct cm_timewait_info * cm_insert_remote_qpn(struct cm_timewait_info + *timewait_info) +{ + struct rb_node **link = &cm.remote_qp_table.rb_node; + struct rb_node *parent = NULL; + struct cm_timewait_info *cur_timewait_info; + __be64 remote_ca_guid = timewait_info->remote_ca_guid; + __be32 remote_qpn = timewait_info->remote_qpn; + + while (*link) { + parent = *link; + cur_timewait_info = rb_entry(parent, struct cm_timewait_info, + remote_qp_node); + if (be32_lt(remote_qpn, cur_timewait_info->remote_qpn)) + link = &(*link)->rb_left; + else if (be32_gt(remote_qpn, cur_timewait_info->remote_qpn)) + link = &(*link)->rb_right; + else if (be64_lt(remote_ca_guid, cur_timewait_info->remote_ca_guid)) + link = &(*link)->rb_left; + else if (be64_gt(remote_ca_guid, cur_timewait_info->remote_ca_guid)) + link = &(*link)->rb_right; + else + return cur_timewait_info; + } + timewait_info->inserted_remote_qp = 1; + rb_link_node(&timewait_info->remote_qp_node, parent, link); + rb_insert_color(&timewait_info->remote_qp_node, &cm.remote_qp_table); + return NULL; +} + +static struct cm_id_private * cm_insert_remote_sidr(struct cm_id_private + *cm_id_priv) +{ + struct rb_node **link = &cm.remote_sidr_table.rb_node; + struct rb_node *parent = NULL; + struct cm_id_private *cur_cm_id_priv; + union ib_gid *port_gid = &cm_id_priv->av.dgid; + __be32 remote_id = cm_id_priv->id.remote_id; + + while (*link) { + parent = *link; + cur_cm_id_priv = rb_entry(parent, struct cm_id_private, + sidr_id_node); + if (be32_lt(remote_id, cur_cm_id_priv->id.remote_id)) + link = &(*link)->rb_left; + else if (be32_gt(remote_id, cur_cm_id_priv->id.remote_id)) + link = &(*link)->rb_right; + else { + int cmp; + cmp = memcmp(port_gid, &cur_cm_id_priv->av.dgid, + sizeof *port_gid); + if (cmp < 0) + link = &(*link)->rb_left; + else if (cmp > 0) + link = &(*link)->rb_right; + else + return cur_cm_id_priv; + } + } + rb_link_node(&cm_id_priv->sidr_id_node, parent, link); + rb_insert_color(&cm_id_priv->sidr_id_node, &cm.remote_sidr_table); + return NULL; +} + +static struct cm_id_private *cm_alloc_id_priv(struct ib_device *device, + ib_cm_handler cm_handler, + void *context) +{ + struct cm_id_private *cm_id_priv; + u32 id; + int ret; + + cm_id_priv = kzalloc(sizeof *cm_id_priv, GFP_KERNEL); + if (!cm_id_priv) + return ERR_PTR(-ENOMEM); + + cm_id_priv->id.state = IB_CM_IDLE; + cm_id_priv->id.device = device; + cm_id_priv->id.cm_handler = cm_handler; + cm_id_priv->id.context = context; + cm_id_priv->id.remote_cm_qpn = 1; + + RB_CLEAR_NODE(&cm_id_priv->service_node); + RB_CLEAR_NODE(&cm_id_priv->sidr_id_node); + spin_lock_init(&cm_id_priv->lock); + init_completion(&cm_id_priv->comp); + INIT_LIST_HEAD(&cm_id_priv->work_list); + INIT_LIST_HEAD(&cm_id_priv->prim_list); + INIT_LIST_HEAD(&cm_id_priv->altr_list); + atomic_set(&cm_id_priv->work_count, -1); + refcount_set(&cm_id_priv->refcount, 1); + + ret = xa_alloc_cyclic(&cm.local_id_table, &id, NULL, xa_limit_32b, + &cm.local_id_next, GFP_KERNEL); + if (ret < 0) + goto error; + cm_id_priv->id.local_id = (__force __be32)id ^ cm.random_id_operand; + + return cm_id_priv; + +error: + kfree(cm_id_priv); + return ERR_PTR(ret); +} + +/* + * Make the ID visible to the MAD handlers and other threads that use the + * xarray. + */ +static void cm_finalize_id(struct cm_id_private *cm_id_priv) +{ + xa_store(&cm.local_id_table, cm_local_id(cm_id_priv->id.local_id), + cm_id_priv, GFP_ATOMIC); +} + +struct ib_cm_id *ib_create_cm_id(struct ib_device *device, + ib_cm_handler cm_handler, + void *context) +{ + struct cm_id_private *cm_id_priv; + + cm_id_priv = cm_alloc_id_priv(device, cm_handler, context); + if (IS_ERR(cm_id_priv)) + return ERR_CAST(cm_id_priv); + + cm_finalize_id(cm_id_priv); + return &cm_id_priv->id; +} +EXPORT_SYMBOL(ib_create_cm_id); + +static struct cm_work * cm_dequeue_work(struct cm_id_private *cm_id_priv) +{ + struct cm_work *work; + + if (list_empty(&cm_id_priv->work_list)) + return NULL; + + work = list_entry(cm_id_priv->work_list.next, struct cm_work, list); + list_del(&work->list); + return work; +} + +static void cm_free_work(struct cm_work *work) +{ + if (work->mad_recv_wc) + ib_free_recv_mad(work->mad_recv_wc); + kfree(work); +} + +static void cm_queue_work_unlock(struct cm_id_private *cm_id_priv, + struct cm_work *work) + __releases(&cm_id_priv->lock) +{ + bool immediate; + + /* + * To deliver the event to the user callback we have the drop the + * spinlock, however, we need to ensure that the user callback is single + * threaded and receives events in the temporal order. If there are + * already events being processed then thread new events onto a list, + * the thread currently processing will pick them up. + */ + immediate = atomic_inc_and_test(&cm_id_priv->work_count); + if (!immediate) { + list_add_tail(&work->list, &cm_id_priv->work_list); + /* + * This routine always consumes incoming reference. Once queued + * to the work_list then a reference is held by the thread + * currently running cm_process_work() and this reference is not + * needed. + */ + cm_deref_id(cm_id_priv); + } + spin_unlock_irq(&cm_id_priv->lock); + + if (immediate) + cm_process_work(cm_id_priv, work); +} + +static inline int cm_convert_to_ms(int iba_time) +{ + /* approximate conversion to ms from 4.096us x 2^iba_time */ + return 1 << max(iba_time - 8, 0); +} + +/* + * calculate: 4.096x2^ack_timeout = 4.096x2^ack_delay + 2x4.096x2^life_time + * Because of how ack_timeout is stored, adding one doubles the timeout. + * To avoid large timeouts, select the max(ack_delay, life_time + 1), and + * increment it (round up) only if the other is within 50%. + */ +static u8 cm_ack_timeout(u8 ca_ack_delay, u8 packet_life_time) +{ + int ack_timeout = packet_life_time + 1; + + if (ack_timeout >= ca_ack_delay) + ack_timeout += (ca_ack_delay >= (ack_timeout - 1)); + else + ack_timeout = ca_ack_delay + + (ack_timeout >= (ca_ack_delay - 1)); + + return min(31, ack_timeout); +} + +static void cm_remove_remote(struct cm_id_private *cm_id_priv) +{ + struct cm_timewait_info *timewait_info = cm_id_priv->timewait_info; + + if (timewait_info->inserted_remote_id) { + rb_erase(&timewait_info->remote_id_node, &cm.remote_id_table); + timewait_info->inserted_remote_id = 0; + } + + if (timewait_info->inserted_remote_qp) { + rb_erase(&timewait_info->remote_qp_node, &cm.remote_qp_table); + timewait_info->inserted_remote_qp = 0; + } +} + +static struct cm_timewait_info * cm_create_timewait_info(__be32 local_id) +{ + struct cm_timewait_info *timewait_info; + + timewait_info = kzalloc(sizeof *timewait_info, GFP_KERNEL); + if (!timewait_info) + return ERR_PTR(-ENOMEM); + + timewait_info->work.local_id = local_id; + INIT_DELAYED_WORK(&timewait_info->work.work, cm_work_handler); + timewait_info->work.cm_event.event = IB_CM_TIMEWAIT_EXIT; + return timewait_info; +} + +static void cm_enter_timewait(struct cm_id_private *cm_id_priv) +{ + int wait_time; + unsigned long flags; + struct cm_device *cm_dev; + + lockdep_assert_held(&cm_id_priv->lock); + + cm_dev = ib_get_client_data(cm_id_priv->id.device, &cm_client); + if (!cm_dev) + return; + + spin_lock_irqsave(&cm.lock, flags); + cm_remove_remote(cm_id_priv); + list_add_tail(&cm_id_priv->timewait_info->list, &cm.timewait_list); + spin_unlock_irqrestore(&cm.lock, flags); + + /* + * The cm_id could be destroyed by the user before we exit timewait. + * To protect against this, we search for the cm_id after exiting + * timewait before notifying the user that we've exited timewait. + */ + cm_id_priv->id.state = IB_CM_TIMEWAIT; + wait_time = cm_convert_to_ms(cm_id_priv->av.timeout); + + /* Check if the device started its remove_one */ + spin_lock_irqsave(&cm.lock, flags); + if (!cm_dev->going_down) + queue_delayed_work(cm.wq, &cm_id_priv->timewait_info->work.work, + msecs_to_jiffies(wait_time)); + spin_unlock_irqrestore(&cm.lock, flags); + + /* + * The timewait_info is converted into a work and gets freed during + * cm_free_work() in cm_timewait_handler(). + */ + BUILD_BUG_ON(offsetof(struct cm_timewait_info, work) != 0); + cm_id_priv->timewait_info = NULL; +} + +static void cm_reset_to_idle(struct cm_id_private *cm_id_priv) +{ + unsigned long flags; + + lockdep_assert_held(&cm_id_priv->lock); + + cm_id_priv->id.state = IB_CM_IDLE; + if (cm_id_priv->timewait_info) { + spin_lock_irqsave(&cm.lock, flags); + cm_remove_remote(cm_id_priv); + spin_unlock_irqrestore(&cm.lock, flags); + kfree(cm_id_priv->timewait_info); + cm_id_priv->timewait_info = NULL; + } +} + +static void cm_destroy_id(struct ib_cm_id *cm_id, int err) +{ + struct cm_id_private *cm_id_priv; + struct cm_work *work; + + cm_id_priv = container_of(cm_id, struct cm_id_private, id); + spin_lock_irq(&cm_id_priv->lock); +retest: + switch (cm_id->state) { + case IB_CM_LISTEN: + spin_lock(&cm.lock); + if (--cm_id_priv->listen_sharecount > 0) { + /* The id is still shared. */ + WARN_ON(refcount_read(&cm_id_priv->refcount) == 1); + spin_unlock(&cm.lock); + spin_unlock_irq(&cm_id_priv->lock); + cm_deref_id(cm_id_priv); + return; + } + cm_id->state = IB_CM_IDLE; + rb_erase(&cm_id_priv->service_node, &cm.listen_service_table); + RB_CLEAR_NODE(&cm_id_priv->service_node); + spin_unlock(&cm.lock); + break; + case IB_CM_SIDR_REQ_SENT: + cm_id->state = IB_CM_IDLE; + ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg); + break; + case IB_CM_SIDR_REQ_RCVD: + cm_send_sidr_rep_locked(cm_id_priv, + &(struct ib_cm_sidr_rep_param){ + .status = IB_SIDR_REJECT }); + /* cm_send_sidr_rep_locked will not move to IDLE if it fails */ + cm_id->state = IB_CM_IDLE; + break; + case IB_CM_REQ_SENT: + case IB_CM_MRA_REQ_RCVD: + ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg); + cm_send_rej_locked(cm_id_priv, IB_CM_REJ_TIMEOUT, + &cm_id_priv->id.device->node_guid, + sizeof(cm_id_priv->id.device->node_guid), + NULL, 0); + break; + case IB_CM_REQ_RCVD: + if (err == -ENOMEM) { + /* Do not reject to allow future retries. */ + cm_reset_to_idle(cm_id_priv); + } else { + cm_send_rej_locked(cm_id_priv, + IB_CM_REJ_CONSUMER_DEFINED, NULL, 0, + NULL, 0); + } + break; + case IB_CM_REP_SENT: + case IB_CM_MRA_REP_RCVD: + ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg); + cm_send_rej_locked(cm_id_priv, IB_CM_REJ_CONSUMER_DEFINED, NULL, + 0, NULL, 0); + goto retest; + case IB_CM_MRA_REQ_SENT: + case IB_CM_REP_RCVD: + case IB_CM_MRA_REP_SENT: + cm_send_rej_locked(cm_id_priv, IB_CM_REJ_CONSUMER_DEFINED, NULL, + 0, NULL, 0); + break; + case IB_CM_ESTABLISHED: + if (cm_id_priv->qp_type == IB_QPT_XRC_TGT) { + cm_id->state = IB_CM_IDLE; + break; + } + cm_send_dreq_locked(cm_id_priv, NULL, 0); + goto retest; + case IB_CM_DREQ_SENT: + ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg); + cm_enter_timewait(cm_id_priv); + goto retest; + case IB_CM_DREQ_RCVD: + cm_send_drep_locked(cm_id_priv, NULL, 0); + WARN_ON(cm_id->state != IB_CM_TIMEWAIT); + goto retest; + case IB_CM_TIMEWAIT: + /* + * The cm_acquire_id in cm_timewait_handler will stop working + * once we do xa_erase below, so just move to idle here for + * consistency. + */ + cm_id->state = IB_CM_IDLE; + break; + case IB_CM_IDLE: + break; + } + WARN_ON(cm_id->state != IB_CM_IDLE); + + spin_lock(&cm.lock); + /* Required for cleanup paths related cm_req_handler() */ + if (cm_id_priv->timewait_info) { + cm_remove_remote(cm_id_priv); + kfree(cm_id_priv->timewait_info); + cm_id_priv->timewait_info = NULL; + } + if (!list_empty(&cm_id_priv->altr_list) && + (!cm_id_priv->altr_send_port_not_ready)) + list_del(&cm_id_priv->altr_list); + if (!list_empty(&cm_id_priv->prim_list) && + (!cm_id_priv->prim_send_port_not_ready)) + list_del(&cm_id_priv->prim_list); + WARN_ON(cm_id_priv->listen_sharecount); + WARN_ON(!RB_EMPTY_NODE(&cm_id_priv->service_node)); + if (!RB_EMPTY_NODE(&cm_id_priv->sidr_id_node)) + rb_erase(&cm_id_priv->sidr_id_node, &cm.remote_sidr_table); + spin_unlock(&cm.lock); + spin_unlock_irq(&cm_id_priv->lock); + + xa_erase(&cm.local_id_table, cm_local_id(cm_id->local_id)); + cm_deref_id(cm_id_priv); + wait_for_completion(&cm_id_priv->comp); + while ((work = cm_dequeue_work(cm_id_priv)) != NULL) + cm_free_work(work); + + rdma_destroy_ah_attr(&cm_id_priv->av.ah_attr); + rdma_destroy_ah_attr(&cm_id_priv->alt_av.ah_attr); + kfree(cm_id_priv->private_data); + kfree_rcu(cm_id_priv, rcu); +} + +void ib_destroy_cm_id(struct ib_cm_id *cm_id) +{ + cm_destroy_id(cm_id, 0); +} +EXPORT_SYMBOL(ib_destroy_cm_id); + +static int cm_init_listen(struct cm_id_private *cm_id_priv, __be64 service_id, + __be64 service_mask) +{ + service_mask = service_mask ? service_mask : ~cpu_to_be64(0); + service_id &= service_mask; + if ((service_id & IB_SERVICE_ID_AGN_MASK) == IB_CM_ASSIGN_SERVICE_ID && + (service_id != IB_CM_ASSIGN_SERVICE_ID)) + return -EINVAL; + + if (service_id == IB_CM_ASSIGN_SERVICE_ID) { + cm_id_priv->id.service_id = cpu_to_be64(cm.listen_service_id++); + cm_id_priv->id.service_mask = ~cpu_to_be64(0); + } else { + cm_id_priv->id.service_id = service_id; + cm_id_priv->id.service_mask = service_mask; + } + return 0; +} + +/** + * ib_cm_listen - Initiates listening on the specified service ID for + * connection and service ID resolution requests. + * @cm_id: Connection identifier associated with the listen request. + * @service_id: Service identifier matched against incoming connection + * and service ID resolution requests. The service ID should be specified + * network-byte order. If set to IB_CM_ASSIGN_SERVICE_ID, the CM will + * assign a service ID to the caller. + * @service_mask: Mask applied to service ID used to listen across a + * range of service IDs. If set to 0, the service ID is matched + * exactly. This parameter is ignored if %service_id is set to + * IB_CM_ASSIGN_SERVICE_ID. + */ +int ib_cm_listen(struct ib_cm_id *cm_id, __be64 service_id, __be64 service_mask) +{ + struct cm_id_private *cm_id_priv = + container_of(cm_id, struct cm_id_private, id); + unsigned long flags; + int ret; + + spin_lock_irqsave(&cm_id_priv->lock, flags); + if (cm_id_priv->id.state != IB_CM_IDLE) { + ret = -EINVAL; + goto out; + } + + ret = cm_init_listen(cm_id_priv, service_id, service_mask); + if (ret) + goto out; + + if (!cm_insert_listen(cm_id_priv, NULL)) { + ret = -EBUSY; + goto out; + } + + cm_id_priv->id.state = IB_CM_LISTEN; + ret = 0; + +out: + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + return ret; +} +EXPORT_SYMBOL(ib_cm_listen); + +/** + * Create a new listening ib_cm_id and listen on the given service ID. + * + * If there's an existing ID listening on that same device and service ID, + * return it. + * + * @device: Device associated with the cm_id. All related communication will + * be associated with the specified device. + * @cm_handler: Callback invoked to notify the user of CM events. + * @service_id: Service identifier matched against incoming connection + * and service ID resolution requests. The service ID should be specified + * network-byte order. If set to IB_CM_ASSIGN_SERVICE_ID, the CM will + * assign a service ID to the caller. + * + * Callers should call ib_destroy_cm_id when done with the listener ID. + */ +struct ib_cm_id *ib_cm_insert_listen(struct ib_device *device, + ib_cm_handler cm_handler, + __be64 service_id) +{ + struct cm_id_private *listen_id_priv; + struct cm_id_private *cm_id_priv; + int err = 0; + + /* Create an ID in advance, since the creation may sleep */ + cm_id_priv = cm_alloc_id_priv(device, cm_handler, NULL); + if (IS_ERR(cm_id_priv)) + return ERR_CAST(cm_id_priv); + + err = cm_init_listen(cm_id_priv, service_id, 0); + if (err) { + ib_destroy_cm_id(&cm_id_priv->id); + return ERR_PTR(err); + } + + spin_lock_irq(&cm_id_priv->lock); + listen_id_priv = cm_insert_listen(cm_id_priv, cm_handler); + if (listen_id_priv != cm_id_priv) { + spin_unlock_irq(&cm_id_priv->lock); + ib_destroy_cm_id(&cm_id_priv->id); + if (!listen_id_priv) + return ERR_PTR(-EINVAL); + return &listen_id_priv->id; + } + cm_id_priv->id.state = IB_CM_LISTEN; + spin_unlock_irq(&cm_id_priv->lock); + + /* + * A listen ID does not need to be in the xarray since it does not + * receive mads, is not placed in the remote_id or remote_qpn rbtree, + * and does not enter timewait. + */ + + return &cm_id_priv->id; +} +EXPORT_SYMBOL(ib_cm_insert_listen); + +static __be64 cm_form_tid(struct cm_id_private *cm_id_priv) +{ + u64 hi_tid, low_tid; + + hi_tid = ((u64) cm_id_priv->av.port->mad_agent->hi_tid) << 32; + low_tid = (u64)cm_id_priv->id.local_id; + return cpu_to_be64(hi_tid | low_tid); +} + +static void cm_format_mad_hdr(struct ib_mad_hdr *hdr, + __be16 attr_id, __be64 tid) +{ + hdr->base_version = IB_MGMT_BASE_VERSION; + hdr->mgmt_class = IB_MGMT_CLASS_CM; + hdr->class_version = IB_CM_CLASS_VERSION; + hdr->method = IB_MGMT_METHOD_SEND; + hdr->attr_id = attr_id; + hdr->tid = tid; +} + +static void cm_format_mad_ece_hdr(struct ib_mad_hdr *hdr, __be16 attr_id, + __be64 tid, u32 attr_mod) +{ + cm_format_mad_hdr(hdr, attr_id, tid); + hdr->attr_mod = cpu_to_be32(attr_mod); +} + +static void cm_format_req(struct cm_req_msg *req_msg, + struct cm_id_private *cm_id_priv, + struct ib_cm_req_param *param) +{ + struct sa_path_rec *pri_path = param->primary_path; + struct sa_path_rec *alt_path = param->alternate_path; + bool pri_ext = false; + + if (pri_path->rec_type == SA_PATH_REC_TYPE_OPA) + pri_ext = opa_is_extended_lid(pri_path->opa.dlid, + pri_path->opa.slid); + + cm_format_mad_ece_hdr(&req_msg->hdr, CM_REQ_ATTR_ID, + cm_form_tid(cm_id_priv), param->ece.attr_mod); + + IBA_SET(CM_REQ_LOCAL_COMM_ID, req_msg, + be32_to_cpu(cm_id_priv->id.local_id)); + IBA_SET(CM_REQ_SERVICE_ID, req_msg, be64_to_cpu(param->service_id)); + IBA_SET(CM_REQ_LOCAL_CA_GUID, req_msg, + be64_to_cpu(cm_id_priv->id.device->node_guid)); + IBA_SET(CM_REQ_LOCAL_QPN, req_msg, param->qp_num); + IBA_SET(CM_REQ_INITIATOR_DEPTH, req_msg, param->initiator_depth); + IBA_SET(CM_REQ_REMOTE_CM_RESPONSE_TIMEOUT, req_msg, + param->remote_cm_response_timeout); + cm_req_set_qp_type(req_msg, param->qp_type); + IBA_SET(CM_REQ_END_TO_END_FLOW_CONTROL, req_msg, param->flow_control); + IBA_SET(CM_REQ_STARTING_PSN, req_msg, param->starting_psn); + IBA_SET(CM_REQ_LOCAL_CM_RESPONSE_TIMEOUT, req_msg, + param->local_cm_response_timeout); + IBA_SET(CM_REQ_PARTITION_KEY, req_msg, + be16_to_cpu(param->primary_path->pkey)); + IBA_SET(CM_REQ_PATH_PACKET_PAYLOAD_MTU, req_msg, + param->primary_path->mtu); + IBA_SET(CM_REQ_MAX_CM_RETRIES, req_msg, param->max_cm_retries); + + if (param->qp_type != IB_QPT_XRC_INI) { + IBA_SET(CM_REQ_RESPONDER_RESOURCES, req_msg, + param->responder_resources); + IBA_SET(CM_REQ_RETRY_COUNT, req_msg, param->retry_count); + IBA_SET(CM_REQ_RNR_RETRY_COUNT, req_msg, + param->rnr_retry_count); + IBA_SET(CM_REQ_SRQ, req_msg, param->srq); + } + + *IBA_GET_MEM_PTR(CM_REQ_PRIMARY_LOCAL_PORT_GID, req_msg) = + pri_path->sgid; + *IBA_GET_MEM_PTR(CM_REQ_PRIMARY_REMOTE_PORT_GID, req_msg) = + pri_path->dgid; + if (pri_ext) { + IBA_GET_MEM_PTR(CM_REQ_PRIMARY_LOCAL_PORT_GID, req_msg) + ->global.interface_id = + OPA_MAKE_ID(be32_to_cpu(pri_path->opa.slid)); + IBA_GET_MEM_PTR(CM_REQ_PRIMARY_REMOTE_PORT_GID, req_msg) + ->global.interface_id = + OPA_MAKE_ID(be32_to_cpu(pri_path->opa.dlid)); + } + if (pri_path->hop_limit <= 1) { + IBA_SET(CM_REQ_PRIMARY_LOCAL_PORT_LID, req_msg, + be16_to_cpu(pri_ext ? 0 : + htons(ntohl(sa_path_get_slid( + pri_path))))); + IBA_SET(CM_REQ_PRIMARY_REMOTE_PORT_LID, req_msg, + be16_to_cpu(pri_ext ? 0 : + htons(ntohl(sa_path_get_dlid( + pri_path))))); + } else { + /* Work-around until there's a way to obtain remote LID info */ + IBA_SET(CM_REQ_PRIMARY_LOCAL_PORT_LID, req_msg, + be16_to_cpu(IB_LID_PERMISSIVE)); + IBA_SET(CM_REQ_PRIMARY_REMOTE_PORT_LID, req_msg, + be16_to_cpu(IB_LID_PERMISSIVE)); + } + IBA_SET(CM_REQ_PRIMARY_FLOW_LABEL, req_msg, + be32_to_cpu(pri_path->flow_label)); + IBA_SET(CM_REQ_PRIMARY_PACKET_RATE, req_msg, pri_path->rate); + IBA_SET(CM_REQ_PRIMARY_TRAFFIC_CLASS, req_msg, pri_path->traffic_class); + IBA_SET(CM_REQ_PRIMARY_HOP_LIMIT, req_msg, pri_path->hop_limit); + IBA_SET(CM_REQ_PRIMARY_SL, req_msg, pri_path->sl); + IBA_SET(CM_REQ_PRIMARY_SUBNET_LOCAL, req_msg, + (pri_path->hop_limit <= 1)); + IBA_SET(CM_REQ_PRIMARY_LOCAL_ACK_TIMEOUT, req_msg, + cm_ack_timeout(cm_id_priv->av.port->cm_dev->ack_delay, + pri_path->packet_life_time)); + + if (alt_path) { + bool alt_ext = false; + + if (alt_path->rec_type == SA_PATH_REC_TYPE_OPA) + alt_ext = opa_is_extended_lid(alt_path->opa.dlid, + alt_path->opa.slid); + + *IBA_GET_MEM_PTR(CM_REQ_ALTERNATE_LOCAL_PORT_GID, req_msg) = + alt_path->sgid; + *IBA_GET_MEM_PTR(CM_REQ_ALTERNATE_REMOTE_PORT_GID, req_msg) = + alt_path->dgid; + if (alt_ext) { + IBA_GET_MEM_PTR(CM_REQ_ALTERNATE_LOCAL_PORT_GID, + req_msg) + ->global.interface_id = + OPA_MAKE_ID(be32_to_cpu(alt_path->opa.slid)); + IBA_GET_MEM_PTR(CM_REQ_ALTERNATE_REMOTE_PORT_GID, + req_msg) + ->global.interface_id = + OPA_MAKE_ID(be32_to_cpu(alt_path->opa.dlid)); + } + if (alt_path->hop_limit <= 1) { + IBA_SET(CM_REQ_ALTERNATE_LOCAL_PORT_LID, req_msg, + be16_to_cpu( + alt_ext ? 0 : + htons(ntohl(sa_path_get_slid( + alt_path))))); + IBA_SET(CM_REQ_ALTERNATE_REMOTE_PORT_LID, req_msg, + be16_to_cpu( + alt_ext ? 0 : + htons(ntohl(sa_path_get_dlid( + alt_path))))); + } else { + IBA_SET(CM_REQ_ALTERNATE_LOCAL_PORT_LID, req_msg, + be16_to_cpu(IB_LID_PERMISSIVE)); + IBA_SET(CM_REQ_ALTERNATE_REMOTE_PORT_LID, req_msg, + be16_to_cpu(IB_LID_PERMISSIVE)); + } + IBA_SET(CM_REQ_ALTERNATE_FLOW_LABEL, req_msg, + be32_to_cpu(alt_path->flow_label)); + IBA_SET(CM_REQ_ALTERNATE_PACKET_RATE, req_msg, alt_path->rate); + IBA_SET(CM_REQ_ALTERNATE_TRAFFIC_CLASS, req_msg, + alt_path->traffic_class); + IBA_SET(CM_REQ_ALTERNATE_HOP_LIMIT, req_msg, + alt_path->hop_limit); + IBA_SET(CM_REQ_ALTERNATE_SL, req_msg, alt_path->sl); + IBA_SET(CM_REQ_ALTERNATE_SUBNET_LOCAL, req_msg, + (alt_path->hop_limit <= 1)); + IBA_SET(CM_REQ_ALTERNATE_LOCAL_ACK_TIMEOUT, req_msg, + cm_ack_timeout(cm_id_priv->av.port->cm_dev->ack_delay, + alt_path->packet_life_time)); + } + IBA_SET(CM_REQ_VENDOR_ID, req_msg, param->ece.vendor_id); + + if (param->private_data && param->private_data_len) + IBA_SET_MEM(CM_REQ_PRIVATE_DATA, req_msg, param->private_data, + param->private_data_len); +} + +static int cm_validate_req_param(struct ib_cm_req_param *param) +{ + if (!param->primary_path) + return -EINVAL; + + if (param->qp_type != IB_QPT_RC && param->qp_type != IB_QPT_UC && + param->qp_type != IB_QPT_XRC_INI) + return -EINVAL; + + if (param->private_data && + param->private_data_len > IB_CM_REQ_PRIVATE_DATA_SIZE) + return -EINVAL; + + if (param->alternate_path && + (param->alternate_path->pkey != param->primary_path->pkey || + param->alternate_path->mtu != param->primary_path->mtu)) + return -EINVAL; + + return 0; +} + +int ib_send_cm_req(struct ib_cm_id *cm_id, + struct ib_cm_req_param *param) +{ + struct cm_id_private *cm_id_priv; + struct cm_req_msg *req_msg; + unsigned long flags; + int ret; + + ret = cm_validate_req_param(param); + if (ret) + return ret; + + /* Verify that we're not in timewait. */ + cm_id_priv = container_of(cm_id, struct cm_id_private, id); + spin_lock_irqsave(&cm_id_priv->lock, flags); + if (cm_id->state != IB_CM_IDLE || WARN_ON(cm_id_priv->timewait_info)) { + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + ret = -EINVAL; + goto out; + } + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + + cm_id_priv->timewait_info = cm_create_timewait_info(cm_id_priv-> + id.local_id); + if (IS_ERR(cm_id_priv->timewait_info)) { + ret = PTR_ERR(cm_id_priv->timewait_info); + cm_id_priv->timewait_info = NULL; + goto out; + } + + ret = cm_init_av_by_path(param->primary_path, + param->ppath_sgid_attr, &cm_id_priv->av, + cm_id_priv); + if (ret) + goto out; + if (param->alternate_path) { + ret = cm_init_av_by_path(param->alternate_path, NULL, + &cm_id_priv->alt_av, cm_id_priv); + if (ret) + goto out; + } + cm_id->service_id = param->service_id; + cm_id->service_mask = ~cpu_to_be64(0); + cm_id_priv->timeout_ms = cm_convert_to_ms( + param->primary_path->packet_life_time) * 2 + + cm_convert_to_ms( + param->remote_cm_response_timeout); + cm_id_priv->max_cm_retries = param->max_cm_retries; + cm_id_priv->initiator_depth = param->initiator_depth; + cm_id_priv->responder_resources = param->responder_resources; + cm_id_priv->retry_count = param->retry_count; + cm_id_priv->path_mtu = param->primary_path->mtu; + cm_id_priv->pkey = param->primary_path->pkey; + cm_id_priv->qp_type = param->qp_type; + + ret = cm_alloc_msg(cm_id_priv, &cm_id_priv->msg); + if (ret) + goto out; + + req_msg = (struct cm_req_msg *) cm_id_priv->msg->mad; + cm_format_req(req_msg, cm_id_priv, param); + cm_id_priv->tid = req_msg->hdr.tid; + cm_id_priv->msg->timeout_ms = cm_id_priv->timeout_ms; + cm_id_priv->msg->context[1] = (void *) (unsigned long) IB_CM_REQ_SENT; + + cm_id_priv->local_qpn = cpu_to_be32(IBA_GET(CM_REQ_LOCAL_QPN, req_msg)); + cm_id_priv->rq_psn = cpu_to_be32(IBA_GET(CM_REQ_STARTING_PSN, req_msg)); + + trace_icm_send_req(&cm_id_priv->id); + spin_lock_irqsave(&cm_id_priv->lock, flags); + ret = ib_post_send_mad(cm_id_priv->msg, NULL); + if (ret) { + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + goto error2; + } + BUG_ON(cm_id->state != IB_CM_IDLE); + cm_id->state = IB_CM_REQ_SENT; + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + return 0; + +error2: cm_free_msg(cm_id_priv->msg); +out: return ret; +} +EXPORT_SYMBOL(ib_send_cm_req); + +static int cm_issue_rej(struct cm_port *port, + struct ib_mad_recv_wc *mad_recv_wc, + enum ib_cm_rej_reason reason, + enum cm_msg_response msg_rejected, + void *ari, u8 ari_length) +{ + struct ib_mad_send_buf *msg = NULL; + struct cm_rej_msg *rej_msg, *rcv_msg; + int ret; + + ret = cm_alloc_response_msg(port, mad_recv_wc, &msg); + if (ret) + return ret; + + /* We just need common CM header information. Cast to any message. */ + rcv_msg = (struct cm_rej_msg *) mad_recv_wc->recv_buf.mad; + rej_msg = (struct cm_rej_msg *) msg->mad; + + cm_format_mad_hdr(&rej_msg->hdr, CM_REJ_ATTR_ID, rcv_msg->hdr.tid); + IBA_SET(CM_REJ_REMOTE_COMM_ID, rej_msg, + IBA_GET(CM_REJ_LOCAL_COMM_ID, rcv_msg)); + IBA_SET(CM_REJ_LOCAL_COMM_ID, rej_msg, + IBA_GET(CM_REJ_REMOTE_COMM_ID, rcv_msg)); + IBA_SET(CM_REJ_MESSAGE_REJECTED, rej_msg, msg_rejected); + IBA_SET(CM_REJ_REASON, rej_msg, reason); + + if (ari && ari_length) { + IBA_SET(CM_REJ_REJECTED_INFO_LENGTH, rej_msg, ari_length); + IBA_SET_MEM(CM_REJ_ARI, rej_msg, ari, ari_length); + } + + trace_icm_issue_rej( + IBA_GET(CM_REJ_LOCAL_COMM_ID, rcv_msg), + IBA_GET(CM_REJ_REMOTE_COMM_ID, rcv_msg)); + ret = ib_post_send_mad(msg, NULL); + if (ret) + cm_free_msg(msg); + + return ret; +} + +static bool cm_req_has_alt_path(struct cm_req_msg *req_msg) +{ + return ((cpu_to_be16( + IBA_GET(CM_REQ_ALTERNATE_LOCAL_PORT_LID, req_msg))) || + (ib_is_opa_gid(IBA_GET_MEM_PTR(CM_REQ_ALTERNATE_LOCAL_PORT_GID, + req_msg)))); +} + +static void cm_path_set_rec_type(struct ib_device *ib_device, u8 port_num, + struct sa_path_rec *path, union ib_gid *gid) +{ + if (ib_is_opa_gid(gid) && rdma_cap_opa_ah(ib_device, port_num)) + path->rec_type = SA_PATH_REC_TYPE_OPA; + else + path->rec_type = SA_PATH_REC_TYPE_IB; +} + +static void cm_format_path_lid_from_req(struct cm_req_msg *req_msg, + struct sa_path_rec *primary_path, + struct sa_path_rec *alt_path, + struct ib_wc *wc) +{ + u32 lid; + + if (primary_path->rec_type != SA_PATH_REC_TYPE_OPA) { + sa_path_set_dlid(primary_path, wc->slid); + sa_path_set_slid(primary_path, + IBA_GET(CM_REQ_PRIMARY_REMOTE_PORT_LID, + req_msg)); + } else { + lid = opa_get_lid_from_gid(IBA_GET_MEM_PTR( + CM_REQ_PRIMARY_LOCAL_PORT_GID, req_msg)); + sa_path_set_dlid(primary_path, lid); + + lid = opa_get_lid_from_gid(IBA_GET_MEM_PTR( + CM_REQ_PRIMARY_REMOTE_PORT_GID, req_msg)); + sa_path_set_slid(primary_path, lid); + } + + if (!cm_req_has_alt_path(req_msg)) + return; + + if (alt_path->rec_type != SA_PATH_REC_TYPE_OPA) { + sa_path_set_dlid(alt_path, + IBA_GET(CM_REQ_ALTERNATE_LOCAL_PORT_LID, + req_msg)); + sa_path_set_slid(alt_path, + IBA_GET(CM_REQ_ALTERNATE_REMOTE_PORT_LID, + req_msg)); + } else { + lid = opa_get_lid_from_gid(IBA_GET_MEM_PTR( + CM_REQ_ALTERNATE_LOCAL_PORT_GID, req_msg)); + sa_path_set_dlid(alt_path, lid); + + lid = opa_get_lid_from_gid(IBA_GET_MEM_PTR( + CM_REQ_ALTERNATE_REMOTE_PORT_GID, req_msg)); + sa_path_set_slid(alt_path, lid); + } +} + +static void cm_format_paths_from_req(struct cm_req_msg *req_msg, + struct sa_path_rec *primary_path, + struct sa_path_rec *alt_path, + struct ib_wc *wc) +{ + primary_path->dgid = + *IBA_GET_MEM_PTR(CM_REQ_PRIMARY_LOCAL_PORT_GID, req_msg); + primary_path->sgid = + *IBA_GET_MEM_PTR(CM_REQ_PRIMARY_REMOTE_PORT_GID, req_msg); + primary_path->flow_label = + cpu_to_be32(IBA_GET(CM_REQ_PRIMARY_FLOW_LABEL, req_msg)); + primary_path->hop_limit = IBA_GET(CM_REQ_PRIMARY_HOP_LIMIT, req_msg); + primary_path->traffic_class = + IBA_GET(CM_REQ_PRIMARY_TRAFFIC_CLASS, req_msg); + primary_path->reversible = 1; + primary_path->pkey = + cpu_to_be16(IBA_GET(CM_REQ_PARTITION_KEY, req_msg)); + primary_path->sl = IBA_GET(CM_REQ_PRIMARY_SL, req_msg); + primary_path->mtu_selector = IB_SA_EQ; + primary_path->mtu = IBA_GET(CM_REQ_PATH_PACKET_PAYLOAD_MTU, req_msg); + primary_path->rate_selector = IB_SA_EQ; + primary_path->rate = IBA_GET(CM_REQ_PRIMARY_PACKET_RATE, req_msg); + primary_path->packet_life_time_selector = IB_SA_EQ; + primary_path->packet_life_time = + IBA_GET(CM_REQ_PRIMARY_LOCAL_ACK_TIMEOUT, req_msg); + primary_path->packet_life_time -= (primary_path->packet_life_time > 0); + primary_path->service_id = + cpu_to_be64(IBA_GET(CM_REQ_SERVICE_ID, req_msg)); + if (sa_path_is_roce(primary_path)) + primary_path->roce.route_resolved = false; + + if (cm_req_has_alt_path(req_msg)) { + alt_path->dgid = *IBA_GET_MEM_PTR( + CM_REQ_ALTERNATE_LOCAL_PORT_GID, req_msg); + alt_path->sgid = *IBA_GET_MEM_PTR( + CM_REQ_ALTERNATE_REMOTE_PORT_GID, req_msg); + alt_path->flow_label = cpu_to_be32( + IBA_GET(CM_REQ_ALTERNATE_FLOW_LABEL, req_msg)); + alt_path->hop_limit = + IBA_GET(CM_REQ_ALTERNATE_HOP_LIMIT, req_msg); + alt_path->traffic_class = + IBA_GET(CM_REQ_ALTERNATE_TRAFFIC_CLASS, req_msg); + alt_path->reversible = 1; + alt_path->pkey = + cpu_to_be16(IBA_GET(CM_REQ_PARTITION_KEY, req_msg)); + alt_path->sl = IBA_GET(CM_REQ_ALTERNATE_SL, req_msg); + alt_path->mtu_selector = IB_SA_EQ; + alt_path->mtu = + IBA_GET(CM_REQ_PATH_PACKET_PAYLOAD_MTU, req_msg); + alt_path->rate_selector = IB_SA_EQ; + alt_path->rate = IBA_GET(CM_REQ_ALTERNATE_PACKET_RATE, req_msg); + alt_path->packet_life_time_selector = IB_SA_EQ; + alt_path->packet_life_time = + IBA_GET(CM_REQ_ALTERNATE_LOCAL_ACK_TIMEOUT, req_msg); + alt_path->packet_life_time -= (alt_path->packet_life_time > 0); + alt_path->service_id = + cpu_to_be64(IBA_GET(CM_REQ_SERVICE_ID, req_msg)); + + if (sa_path_is_roce(alt_path)) + alt_path->roce.route_resolved = false; + } + cm_format_path_lid_from_req(req_msg, primary_path, alt_path, wc); +} + +static u16 cm_get_bth_pkey(struct cm_work *work) +{ + struct ib_device *ib_dev = work->port->cm_dev->ib_device; + u8 port_num = work->port->port_num; + u16 pkey_index = work->mad_recv_wc->wc->pkey_index; + u16 pkey; + int ret; + + ret = ib_get_cached_pkey(ib_dev, port_num, pkey_index, &pkey); + if (ret) { + dev_warn_ratelimited(&ib_dev->dev, "ib_cm: Couldn't retrieve pkey for incoming request (port %d, pkey index %d). %d\n", + port_num, pkey_index, ret); + return 0; + } + + return pkey; +} + +/** + * Convert OPA SGID to IB SGID + * ULPs (such as IPoIB) do not understand OPA GIDs and will + * reject them as the local_gid will not match the sgid. Therefore, + * change the pathrec's SGID to an IB SGID. + * + * @work: Work completion + * @path: Path record + */ +static void cm_opa_to_ib_sgid(struct cm_work *work, + struct sa_path_rec *path) +{ + struct ib_device *dev = work->port->cm_dev->ib_device; + u8 port_num = work->port->port_num; + + if (rdma_cap_opa_ah(dev, port_num) && + (ib_is_opa_gid(&path->sgid))) { + union ib_gid sgid; + + if (rdma_query_gid(dev, port_num, 0, &sgid)) { + dev_warn(&dev->dev, + "Error updating sgid in CM request\n"); + return; + } + + path->sgid = sgid; + } +} + +static void cm_format_req_event(struct cm_work *work, + struct cm_id_private *cm_id_priv, + struct ib_cm_id *listen_id) +{ + struct cm_req_msg *req_msg; + struct ib_cm_req_event_param *param; + + req_msg = (struct cm_req_msg *)work->mad_recv_wc->recv_buf.mad; + param = &work->cm_event.param.req_rcvd; + param->listen_id = listen_id; + param->bth_pkey = cm_get_bth_pkey(work); + param->port = cm_id_priv->av.port->port_num; + param->primary_path = &work->path[0]; + cm_opa_to_ib_sgid(work, param->primary_path); + if (cm_req_has_alt_path(req_msg)) { + param->alternate_path = &work->path[1]; + cm_opa_to_ib_sgid(work, param->alternate_path); + } else { + param->alternate_path = NULL; + } + param->remote_ca_guid = + cpu_to_be64(IBA_GET(CM_REQ_LOCAL_CA_GUID, req_msg)); + param->remote_qkey = IBA_GET(CM_REQ_LOCAL_Q_KEY, req_msg); + param->remote_qpn = IBA_GET(CM_REQ_LOCAL_QPN, req_msg); + param->qp_type = cm_req_get_qp_type(req_msg); + param->starting_psn = IBA_GET(CM_REQ_STARTING_PSN, req_msg); + param->responder_resources = IBA_GET(CM_REQ_INITIATOR_DEPTH, req_msg); + param->initiator_depth = IBA_GET(CM_REQ_RESPONDER_RESOURCES, req_msg); + param->local_cm_response_timeout = + IBA_GET(CM_REQ_REMOTE_CM_RESPONSE_TIMEOUT, req_msg); + param->flow_control = IBA_GET(CM_REQ_END_TO_END_FLOW_CONTROL, req_msg); + param->remote_cm_response_timeout = + IBA_GET(CM_REQ_LOCAL_CM_RESPONSE_TIMEOUT, req_msg); + param->retry_count = IBA_GET(CM_REQ_RETRY_COUNT, req_msg); + param->rnr_retry_count = IBA_GET(CM_REQ_RNR_RETRY_COUNT, req_msg); + param->srq = IBA_GET(CM_REQ_SRQ, req_msg); + param->ppath_sgid_attr = cm_id_priv->av.ah_attr.grh.sgid_attr; + param->ece.vendor_id = IBA_GET(CM_REQ_VENDOR_ID, req_msg); + param->ece.attr_mod = be32_to_cpu(req_msg->hdr.attr_mod); + + work->cm_event.private_data = + IBA_GET_MEM_PTR(CM_REQ_PRIVATE_DATA, req_msg); +} + +static void cm_process_work(struct cm_id_private *cm_id_priv, + struct cm_work *work) +{ + int ret; + + /* We will typically only have the current event to report. */ + ret = cm_id_priv->id.cm_handler(&cm_id_priv->id, &work->cm_event); + cm_free_work(work); + + while (!ret && !atomic_add_negative(-1, &cm_id_priv->work_count)) { + spin_lock_irq(&cm_id_priv->lock); + work = cm_dequeue_work(cm_id_priv); + spin_unlock_irq(&cm_id_priv->lock); + if (!work) + return; + + ret = cm_id_priv->id.cm_handler(&cm_id_priv->id, + &work->cm_event); + cm_free_work(work); + } + cm_deref_id(cm_id_priv); + if (ret) + cm_destroy_id(&cm_id_priv->id, ret); +} + +static void cm_format_mra(struct cm_mra_msg *mra_msg, + struct cm_id_private *cm_id_priv, + enum cm_msg_response msg_mraed, u8 service_timeout, + const void *private_data, u8 private_data_len) +{ + cm_format_mad_hdr(&mra_msg->hdr, CM_MRA_ATTR_ID, cm_id_priv->tid); + IBA_SET(CM_MRA_MESSAGE_MRAED, mra_msg, msg_mraed); + IBA_SET(CM_MRA_LOCAL_COMM_ID, mra_msg, + be32_to_cpu(cm_id_priv->id.local_id)); + IBA_SET(CM_MRA_REMOTE_COMM_ID, mra_msg, + be32_to_cpu(cm_id_priv->id.remote_id)); + IBA_SET(CM_MRA_SERVICE_TIMEOUT, mra_msg, service_timeout); + + if (private_data && private_data_len) + IBA_SET_MEM(CM_MRA_PRIVATE_DATA, mra_msg, private_data, + private_data_len); +} + +static void cm_format_rej(struct cm_rej_msg *rej_msg, + struct cm_id_private *cm_id_priv, + enum ib_cm_rej_reason reason, void *ari, + u8 ari_length, const void *private_data, + u8 private_data_len, enum ib_cm_state state) +{ + lockdep_assert_held(&cm_id_priv->lock); + + cm_format_mad_hdr(&rej_msg->hdr, CM_REJ_ATTR_ID, cm_id_priv->tid); + IBA_SET(CM_REJ_REMOTE_COMM_ID, rej_msg, + be32_to_cpu(cm_id_priv->id.remote_id)); + + switch (state) { + case IB_CM_REQ_RCVD: + IBA_SET(CM_REJ_LOCAL_COMM_ID, rej_msg, be32_to_cpu(0)); + IBA_SET(CM_REJ_MESSAGE_REJECTED, rej_msg, CM_MSG_RESPONSE_REQ); + break; + case IB_CM_MRA_REQ_SENT: + IBA_SET(CM_REJ_LOCAL_COMM_ID, rej_msg, + be32_to_cpu(cm_id_priv->id.local_id)); + IBA_SET(CM_REJ_MESSAGE_REJECTED, rej_msg, CM_MSG_RESPONSE_REQ); + break; + case IB_CM_REP_RCVD: + case IB_CM_MRA_REP_SENT: + IBA_SET(CM_REJ_LOCAL_COMM_ID, rej_msg, + be32_to_cpu(cm_id_priv->id.local_id)); + IBA_SET(CM_REJ_MESSAGE_REJECTED, rej_msg, CM_MSG_RESPONSE_REP); + break; + default: + IBA_SET(CM_REJ_LOCAL_COMM_ID, rej_msg, + be32_to_cpu(cm_id_priv->id.local_id)); + IBA_SET(CM_REJ_MESSAGE_REJECTED, rej_msg, + CM_MSG_RESPONSE_OTHER); + break; + } + + IBA_SET(CM_REJ_REASON, rej_msg, reason); + if (ari && ari_length) { + IBA_SET(CM_REJ_REJECTED_INFO_LENGTH, rej_msg, ari_length); + IBA_SET_MEM(CM_REJ_ARI, rej_msg, ari, ari_length); + } + + if (private_data && private_data_len) + IBA_SET_MEM(CM_REJ_PRIVATE_DATA, rej_msg, private_data, + private_data_len); +} + +static void cm_dup_req_handler(struct cm_work *work, + struct cm_id_private *cm_id_priv) +{ + struct ib_mad_send_buf *msg = NULL; + int ret; + + atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES]. + counter[CM_REQ_COUNTER]); + + /* Quick state check to discard duplicate REQs. */ + spin_lock_irq(&cm_id_priv->lock); + if (cm_id_priv->id.state == IB_CM_REQ_RCVD) { + spin_unlock_irq(&cm_id_priv->lock); + return; + } + spin_unlock_irq(&cm_id_priv->lock); + + ret = cm_alloc_response_msg(work->port, work->mad_recv_wc, &msg); + if (ret) + return; + + spin_lock_irq(&cm_id_priv->lock); + switch (cm_id_priv->id.state) { + case IB_CM_MRA_REQ_SENT: + cm_format_mra((struct cm_mra_msg *) msg->mad, cm_id_priv, + CM_MSG_RESPONSE_REQ, cm_id_priv->service_timeout, + cm_id_priv->private_data, + cm_id_priv->private_data_len); + break; + case IB_CM_TIMEWAIT: + cm_format_rej((struct cm_rej_msg *)msg->mad, cm_id_priv, + IB_CM_REJ_STALE_CONN, NULL, 0, NULL, 0, + IB_CM_TIMEWAIT); + break; + default: + goto unlock; + } + spin_unlock_irq(&cm_id_priv->lock); + + trace_icm_send_dup_req(&cm_id_priv->id); + ret = ib_post_send_mad(msg, NULL); + if (ret) + goto free; + return; + +unlock: spin_unlock_irq(&cm_id_priv->lock); +free: cm_free_msg(msg); +} + +static struct cm_id_private * cm_match_req(struct cm_work *work, + struct cm_id_private *cm_id_priv) +{ + struct cm_id_private *listen_cm_id_priv, *cur_cm_id_priv; + struct cm_timewait_info *timewait_info; + struct cm_req_msg *req_msg; + + req_msg = (struct cm_req_msg *)work->mad_recv_wc->recv_buf.mad; + + /* Check for possible duplicate REQ. */ + spin_lock_irq(&cm.lock); + timewait_info = cm_insert_remote_id(cm_id_priv->timewait_info); + if (timewait_info) { + cur_cm_id_priv = cm_acquire_id(timewait_info->work.local_id, + timewait_info->work.remote_id); + spin_unlock_irq(&cm.lock); + if (cur_cm_id_priv) { + cm_dup_req_handler(work, cur_cm_id_priv); + cm_deref_id(cur_cm_id_priv); + } + return NULL; + } + + /* Check for stale connections. */ + timewait_info = cm_insert_remote_qpn(cm_id_priv->timewait_info); + if (timewait_info) { + cm_remove_remote(cm_id_priv); + cur_cm_id_priv = cm_acquire_id(timewait_info->work.local_id, + timewait_info->work.remote_id); + + spin_unlock_irq(&cm.lock); + cm_issue_rej(work->port, work->mad_recv_wc, + IB_CM_REJ_STALE_CONN, CM_MSG_RESPONSE_REQ, + NULL, 0); + if (cur_cm_id_priv) { + ib_send_cm_dreq(&cur_cm_id_priv->id, NULL, 0); + cm_deref_id(cur_cm_id_priv); + } + return NULL; + } + + /* Find matching listen request. */ + listen_cm_id_priv = cm_find_listen( + cm_id_priv->id.device, + cpu_to_be64(IBA_GET(CM_REQ_SERVICE_ID, req_msg))); + if (!listen_cm_id_priv) { + cm_remove_remote(cm_id_priv); + spin_unlock_irq(&cm.lock); + cm_issue_rej(work->port, work->mad_recv_wc, + IB_CM_REJ_INVALID_SERVICE_ID, CM_MSG_RESPONSE_REQ, + NULL, 0); + return NULL; + } + spin_unlock_irq(&cm.lock); + return listen_cm_id_priv; +} + +/* + * Work-around for inter-subnet connections. If the LIDs are permissive, + * we need to override the LID/SL data in the REQ with the LID information + * in the work completion. + */ +static void cm_process_routed_req(struct cm_req_msg *req_msg, struct ib_wc *wc) +{ + if (!IBA_GET(CM_REQ_PRIMARY_SUBNET_LOCAL, req_msg)) { + if (cpu_to_be16(IBA_GET(CM_REQ_PRIMARY_LOCAL_PORT_LID, + req_msg)) == IB_LID_PERMISSIVE) { + IBA_SET(CM_REQ_PRIMARY_LOCAL_PORT_LID, req_msg, + be16_to_cpu(ib_lid_be16(wc->slid))); + IBA_SET(CM_REQ_PRIMARY_SL, req_msg, wc->sl); + } + + if (cpu_to_be16(IBA_GET(CM_REQ_PRIMARY_REMOTE_PORT_LID, + req_msg)) == IB_LID_PERMISSIVE) + IBA_SET(CM_REQ_PRIMARY_REMOTE_PORT_LID, req_msg, + wc->dlid_path_bits); + } + + if (!IBA_GET(CM_REQ_ALTERNATE_SUBNET_LOCAL, req_msg)) { + if (cpu_to_be16(IBA_GET(CM_REQ_ALTERNATE_LOCAL_PORT_LID, + req_msg)) == IB_LID_PERMISSIVE) { + IBA_SET(CM_REQ_ALTERNATE_LOCAL_PORT_LID, req_msg, + be16_to_cpu(ib_lid_be16(wc->slid))); + IBA_SET(CM_REQ_ALTERNATE_SL, req_msg, wc->sl); + } + + if (cpu_to_be16(IBA_GET(CM_REQ_ALTERNATE_REMOTE_PORT_LID, + req_msg)) == IB_LID_PERMISSIVE) + IBA_SET(CM_REQ_ALTERNATE_REMOTE_PORT_LID, req_msg, + wc->dlid_path_bits); + } +} + +static int cm_req_handler(struct cm_work *work) +{ + struct cm_id_private *cm_id_priv, *listen_cm_id_priv; + struct cm_req_msg *req_msg; + const struct ib_global_route *grh; + const struct ib_gid_attr *gid_attr; + int ret; + + req_msg = (struct cm_req_msg *)work->mad_recv_wc->recv_buf.mad; + + cm_id_priv = + cm_alloc_id_priv(work->port->cm_dev->ib_device, NULL, NULL); + if (IS_ERR(cm_id_priv)) + return PTR_ERR(cm_id_priv); + + cm_id_priv->id.remote_id = + cpu_to_be32(IBA_GET(CM_REQ_LOCAL_COMM_ID, req_msg)); + cm_id_priv->id.service_id = + cpu_to_be64(IBA_GET(CM_REQ_SERVICE_ID, req_msg)); + cm_id_priv->id.service_mask = ~cpu_to_be64(0); + cm_id_priv->tid = req_msg->hdr.tid; + cm_id_priv->timeout_ms = cm_convert_to_ms( + IBA_GET(CM_REQ_LOCAL_CM_RESPONSE_TIMEOUT, req_msg)); + cm_id_priv->max_cm_retries = IBA_GET(CM_REQ_MAX_CM_RETRIES, req_msg); + cm_id_priv->remote_qpn = + cpu_to_be32(IBA_GET(CM_REQ_LOCAL_QPN, req_msg)); + cm_id_priv->initiator_depth = + IBA_GET(CM_REQ_RESPONDER_RESOURCES, req_msg); + cm_id_priv->responder_resources = + IBA_GET(CM_REQ_INITIATOR_DEPTH, req_msg); + cm_id_priv->path_mtu = IBA_GET(CM_REQ_PATH_PACKET_PAYLOAD_MTU, req_msg); + cm_id_priv->pkey = cpu_to_be16(IBA_GET(CM_REQ_PARTITION_KEY, req_msg)); + cm_id_priv->sq_psn = cpu_to_be32(IBA_GET(CM_REQ_STARTING_PSN, req_msg)); + cm_id_priv->retry_count = IBA_GET(CM_REQ_RETRY_COUNT, req_msg); + cm_id_priv->rnr_retry_count = IBA_GET(CM_REQ_RNR_RETRY_COUNT, req_msg); + cm_id_priv->qp_type = cm_req_get_qp_type(req_msg); + + ret = cm_init_av_for_response(work->port, work->mad_recv_wc->wc, + work->mad_recv_wc->recv_buf.grh, + &cm_id_priv->av); + if (ret) + goto destroy; + cm_id_priv->timewait_info = cm_create_timewait_info(cm_id_priv-> + id.local_id); + if (IS_ERR(cm_id_priv->timewait_info)) { + ret = PTR_ERR(cm_id_priv->timewait_info); + cm_id_priv->timewait_info = NULL; + goto destroy; + } + cm_id_priv->timewait_info->work.remote_id = cm_id_priv->id.remote_id; + cm_id_priv->timewait_info->remote_ca_guid = + cpu_to_be64(IBA_GET(CM_REQ_LOCAL_CA_GUID, req_msg)); + cm_id_priv->timewait_info->remote_qpn = cm_id_priv->remote_qpn; + + /* + * Note that the ID pointer is not in the xarray at this point, + * so this set is only visible to the local thread. + */ + cm_id_priv->id.state = IB_CM_REQ_RCVD; + + listen_cm_id_priv = cm_match_req(work, cm_id_priv); + if (!listen_cm_id_priv) { + trace_icm_no_listener_err(&cm_id_priv->id); + cm_id_priv->id.state = IB_CM_IDLE; + ret = -EINVAL; + goto destroy; + } + + if (cm_id_priv->av.ah_attr.type != RDMA_AH_ATTR_TYPE_ROCE) + cm_process_routed_req(req_msg, work->mad_recv_wc->wc); + + memset(&work->path[0], 0, sizeof(work->path[0])); + if (cm_req_has_alt_path(req_msg)) + memset(&work->path[1], 0, sizeof(work->path[1])); + grh = rdma_ah_read_grh(&cm_id_priv->av.ah_attr); + gid_attr = grh->sgid_attr; + + if (gid_attr && + rdma_protocol_roce(work->port->cm_dev->ib_device, + work->port->port_num)) { + work->path[0].rec_type = + sa_conv_gid_to_pathrec_type(gid_attr->gid_type); + } else { + cm_path_set_rec_type( + work->port->cm_dev->ib_device, work->port->port_num, + &work->path[0], + IBA_GET_MEM_PTR(CM_REQ_PRIMARY_LOCAL_PORT_GID, + req_msg)); + } + if (cm_req_has_alt_path(req_msg)) + work->path[1].rec_type = work->path[0].rec_type; + cm_format_paths_from_req(req_msg, &work->path[0], + &work->path[1], work->mad_recv_wc->wc); + if (cm_id_priv->av.ah_attr.type == RDMA_AH_ATTR_TYPE_ROCE) + sa_path_set_dmac(&work->path[0], + cm_id_priv->av.ah_attr.roce.dmac); + work->path[0].hop_limit = grh->hop_limit; + ret = cm_init_av_by_path(&work->path[0], gid_attr, &cm_id_priv->av, + cm_id_priv); + if (ret) { + int err; + + err = rdma_query_gid(work->port->cm_dev->ib_device, + work->port->port_num, 0, + &work->path[0].sgid); + if (err) + ib_send_cm_rej(&cm_id_priv->id, IB_CM_REJ_INVALID_GID, + NULL, 0, NULL, 0); + else + ib_send_cm_rej(&cm_id_priv->id, IB_CM_REJ_INVALID_GID, + &work->path[0].sgid, + sizeof(work->path[0].sgid), + NULL, 0); + goto rejected; + } + if (cm_req_has_alt_path(req_msg)) { + ret = cm_init_av_by_path(&work->path[1], NULL, + &cm_id_priv->alt_av, cm_id_priv); + if (ret) { + ib_send_cm_rej(&cm_id_priv->id, + IB_CM_REJ_INVALID_ALT_GID, + &work->path[0].sgid, + sizeof(work->path[0].sgid), NULL, 0); + goto rejected; + } + } + + cm_id_priv->id.cm_handler = listen_cm_id_priv->id.cm_handler; + cm_id_priv->id.context = listen_cm_id_priv->id.context; + cm_format_req_event(work, cm_id_priv, &listen_cm_id_priv->id); + + /* Now MAD handlers can see the new ID */ + spin_lock_irq(&cm_id_priv->lock); + cm_finalize_id(cm_id_priv); + + /* Refcount belongs to the event, pairs with cm_process_work() */ + refcount_inc(&cm_id_priv->refcount); + cm_queue_work_unlock(cm_id_priv, work); + /* + * Since this ID was just created and was not made visible to other MAD + * handlers until the cm_finalize_id() above we know that the + * cm_process_work() will deliver the event and the listen_cm_id + * embedded in the event can be derefed here. + */ + cm_deref_id(listen_cm_id_priv); + return 0; + +rejected: + cm_deref_id(listen_cm_id_priv); +destroy: + ib_destroy_cm_id(&cm_id_priv->id); + return ret; +} + +static void cm_format_rep(struct cm_rep_msg *rep_msg, + struct cm_id_private *cm_id_priv, + struct ib_cm_rep_param *param) +{ + cm_format_mad_ece_hdr(&rep_msg->hdr, CM_REP_ATTR_ID, cm_id_priv->tid, + param->ece.attr_mod); + IBA_SET(CM_REP_LOCAL_COMM_ID, rep_msg, + be32_to_cpu(cm_id_priv->id.local_id)); + IBA_SET(CM_REP_REMOTE_COMM_ID, rep_msg, + be32_to_cpu(cm_id_priv->id.remote_id)); + IBA_SET(CM_REP_STARTING_PSN, rep_msg, param->starting_psn); + IBA_SET(CM_REP_RESPONDER_RESOURCES, rep_msg, + param->responder_resources); + IBA_SET(CM_REP_TARGET_ACK_DELAY, rep_msg, + cm_id_priv->av.port->cm_dev->ack_delay); + IBA_SET(CM_REP_FAILOVER_ACCEPTED, rep_msg, param->failover_accepted); + IBA_SET(CM_REP_RNR_RETRY_COUNT, rep_msg, param->rnr_retry_count); + IBA_SET(CM_REP_LOCAL_CA_GUID, rep_msg, + be64_to_cpu(cm_id_priv->id.device->node_guid)); + + if (cm_id_priv->qp_type != IB_QPT_XRC_TGT) { + IBA_SET(CM_REP_INITIATOR_DEPTH, rep_msg, + param->initiator_depth); + IBA_SET(CM_REP_END_TO_END_FLOW_CONTROL, rep_msg, + param->flow_control); + IBA_SET(CM_REP_SRQ, rep_msg, param->srq); + IBA_SET(CM_REP_LOCAL_QPN, rep_msg, param->qp_num); + } else { + IBA_SET(CM_REP_SRQ, rep_msg, 1); + IBA_SET(CM_REP_LOCAL_EE_CONTEXT_NUMBER, rep_msg, param->qp_num); + } + + IBA_SET(CM_REP_VENDOR_ID_L, rep_msg, param->ece.vendor_id); + IBA_SET(CM_REP_VENDOR_ID_M, rep_msg, param->ece.vendor_id >> 8); + IBA_SET(CM_REP_VENDOR_ID_H, rep_msg, param->ece.vendor_id >> 16); + + if (param->private_data && param->private_data_len) + IBA_SET_MEM(CM_REP_PRIVATE_DATA, rep_msg, param->private_data, + param->private_data_len); +} + +int ib_send_cm_rep(struct ib_cm_id *cm_id, + struct ib_cm_rep_param *param) +{ + struct cm_id_private *cm_id_priv; + struct ib_mad_send_buf *msg; + struct cm_rep_msg *rep_msg; + unsigned long flags; + int ret; + + if (param->private_data && + param->private_data_len > IB_CM_REP_PRIVATE_DATA_SIZE) + return -EINVAL; + + cm_id_priv = container_of(cm_id, struct cm_id_private, id); + spin_lock_irqsave(&cm_id_priv->lock, flags); + if (cm_id->state != IB_CM_REQ_RCVD && + cm_id->state != IB_CM_MRA_REQ_SENT) { + trace_icm_send_rep_err(cm_id_priv->id.local_id, cm_id->state); + ret = -EINVAL; + goto out; + } + + ret = cm_alloc_msg(cm_id_priv, &msg); + if (ret) + goto out; + + rep_msg = (struct cm_rep_msg *) msg->mad; + cm_format_rep(rep_msg, cm_id_priv, param); + msg->timeout_ms = cm_id_priv->timeout_ms; + msg->context[1] = (void *) (unsigned long) IB_CM_REP_SENT; + + trace_icm_send_rep(cm_id); + ret = ib_post_send_mad(msg, NULL); + if (ret) { + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + cm_free_msg(msg); + return ret; + } + + cm_id->state = IB_CM_REP_SENT; + cm_id_priv->msg = msg; + cm_id_priv->initiator_depth = param->initiator_depth; + cm_id_priv->responder_resources = param->responder_resources; + cm_id_priv->rq_psn = cpu_to_be32(IBA_GET(CM_REP_STARTING_PSN, rep_msg)); + WARN_ONCE(param->qp_num & 0xFF000000, + "IBTA declares QPN to be 24 bits, but it is 0x%X\n", + param->qp_num); + cm_id_priv->local_qpn = cpu_to_be32(param->qp_num & 0xFFFFFF); + +out: spin_unlock_irqrestore(&cm_id_priv->lock, flags); + return ret; +} +EXPORT_SYMBOL(ib_send_cm_rep); + +static void cm_format_rtu(struct cm_rtu_msg *rtu_msg, + struct cm_id_private *cm_id_priv, + const void *private_data, + u8 private_data_len) +{ + cm_format_mad_hdr(&rtu_msg->hdr, CM_RTU_ATTR_ID, cm_id_priv->tid); + IBA_SET(CM_RTU_LOCAL_COMM_ID, rtu_msg, + be32_to_cpu(cm_id_priv->id.local_id)); + IBA_SET(CM_RTU_REMOTE_COMM_ID, rtu_msg, + be32_to_cpu(cm_id_priv->id.remote_id)); + + if (private_data && private_data_len) + IBA_SET_MEM(CM_RTU_PRIVATE_DATA, rtu_msg, private_data, + private_data_len); +} + +int ib_send_cm_rtu(struct ib_cm_id *cm_id, + const void *private_data, + u8 private_data_len) +{ + struct cm_id_private *cm_id_priv; + struct ib_mad_send_buf *msg; + unsigned long flags; + void *data; + int ret; + + if (private_data && private_data_len > IB_CM_RTU_PRIVATE_DATA_SIZE) + return -EINVAL; + + data = cm_copy_private_data(private_data, private_data_len); + if (IS_ERR(data)) + return PTR_ERR(data); + + cm_id_priv = container_of(cm_id, struct cm_id_private, id); + spin_lock_irqsave(&cm_id_priv->lock, flags); + if (cm_id->state != IB_CM_REP_RCVD && + cm_id->state != IB_CM_MRA_REP_SENT) { + trace_icm_send_cm_rtu_err(cm_id); + ret = -EINVAL; + goto error; + } + + ret = cm_alloc_msg(cm_id_priv, &msg); + if (ret) + goto error; + + cm_format_rtu((struct cm_rtu_msg *) msg->mad, cm_id_priv, + private_data, private_data_len); + + trace_icm_send_rtu(cm_id); + ret = ib_post_send_mad(msg, NULL); + if (ret) { + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + cm_free_msg(msg); + kfree(data); + return ret; + } + + cm_id->state = IB_CM_ESTABLISHED; + cm_set_private_data(cm_id_priv, data, private_data_len); + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + return 0; + +error: spin_unlock_irqrestore(&cm_id_priv->lock, flags); + kfree(data); + return ret; +} +EXPORT_SYMBOL(ib_send_cm_rtu); + +static void cm_format_rep_event(struct cm_work *work, enum ib_qp_type qp_type) +{ + struct cm_rep_msg *rep_msg; + struct ib_cm_rep_event_param *param; + + rep_msg = (struct cm_rep_msg *)work->mad_recv_wc->recv_buf.mad; + param = &work->cm_event.param.rep_rcvd; + param->remote_ca_guid = + cpu_to_be64(IBA_GET(CM_REP_LOCAL_CA_GUID, rep_msg)); + param->remote_qkey = IBA_GET(CM_REP_LOCAL_Q_KEY, rep_msg); + param->remote_qpn = be32_to_cpu(cm_rep_get_qpn(rep_msg, qp_type)); + param->starting_psn = IBA_GET(CM_REP_STARTING_PSN, rep_msg); + param->responder_resources = IBA_GET(CM_REP_INITIATOR_DEPTH, rep_msg); + param->initiator_depth = IBA_GET(CM_REP_RESPONDER_RESOURCES, rep_msg); + param->target_ack_delay = IBA_GET(CM_REP_TARGET_ACK_DELAY, rep_msg); + param->failover_accepted = IBA_GET(CM_REP_FAILOVER_ACCEPTED, rep_msg); + param->flow_control = IBA_GET(CM_REP_END_TO_END_FLOW_CONTROL, rep_msg); + param->rnr_retry_count = IBA_GET(CM_REP_RNR_RETRY_COUNT, rep_msg); + param->srq = IBA_GET(CM_REP_SRQ, rep_msg); + param->ece.vendor_id = IBA_GET(CM_REP_VENDOR_ID_H, rep_msg) << 16; + param->ece.vendor_id |= IBA_GET(CM_REP_VENDOR_ID_M, rep_msg) << 8; + param->ece.vendor_id |= IBA_GET(CM_REP_VENDOR_ID_L, rep_msg); + param->ece.attr_mod = be32_to_cpu(rep_msg->hdr.attr_mod); + + work->cm_event.private_data = + IBA_GET_MEM_PTR(CM_REP_PRIVATE_DATA, rep_msg); +} + +static void cm_dup_rep_handler(struct cm_work *work) +{ + struct cm_id_private *cm_id_priv; + struct cm_rep_msg *rep_msg; + struct ib_mad_send_buf *msg = NULL; + int ret; + + rep_msg = (struct cm_rep_msg *) work->mad_recv_wc->recv_buf.mad; + cm_id_priv = cm_acquire_id( + cpu_to_be32(IBA_GET(CM_REP_REMOTE_COMM_ID, rep_msg)), + cpu_to_be32(IBA_GET(CM_REP_LOCAL_COMM_ID, rep_msg))); + if (!cm_id_priv) + return; + + atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES]. + counter[CM_REP_COUNTER]); + ret = cm_alloc_response_msg(work->port, work->mad_recv_wc, &msg); + if (ret) + goto deref; + + spin_lock_irq(&cm_id_priv->lock); + if (cm_id_priv->id.state == IB_CM_ESTABLISHED) + cm_format_rtu((struct cm_rtu_msg *) msg->mad, cm_id_priv, + cm_id_priv->private_data, + cm_id_priv->private_data_len); + else if (cm_id_priv->id.state == IB_CM_MRA_REP_SENT) + cm_format_mra((struct cm_mra_msg *) msg->mad, cm_id_priv, + CM_MSG_RESPONSE_REP, cm_id_priv->service_timeout, + cm_id_priv->private_data, + cm_id_priv->private_data_len); + else + goto unlock; + spin_unlock_irq(&cm_id_priv->lock); + + trace_icm_send_dup_rep(&cm_id_priv->id); + ret = ib_post_send_mad(msg, NULL); + if (ret) + goto free; + goto deref; + +unlock: spin_unlock_irq(&cm_id_priv->lock); +free: cm_free_msg(msg); +deref: cm_deref_id(cm_id_priv); +} + +static int cm_rep_handler(struct cm_work *work) +{ + struct cm_id_private *cm_id_priv; + struct cm_rep_msg *rep_msg; + int ret; + struct cm_id_private *cur_cm_id_priv; + struct cm_timewait_info *timewait_info; + + rep_msg = (struct cm_rep_msg *)work->mad_recv_wc->recv_buf.mad; + cm_id_priv = cm_acquire_id( + cpu_to_be32(IBA_GET(CM_REP_REMOTE_COMM_ID, rep_msg)), 0); + if (!cm_id_priv) { + cm_dup_rep_handler(work); + trace_icm_remote_no_priv_err( + IBA_GET(CM_REP_REMOTE_COMM_ID, rep_msg)); + return -EINVAL; + } + + cm_format_rep_event(work, cm_id_priv->qp_type); + + spin_lock_irq(&cm_id_priv->lock); + switch (cm_id_priv->id.state) { + case IB_CM_REQ_SENT: + case IB_CM_MRA_REQ_RCVD: + break; + default: + ret = -EINVAL; + trace_icm_rep_unknown_err( + IBA_GET(CM_REP_LOCAL_COMM_ID, rep_msg), + IBA_GET(CM_REP_REMOTE_COMM_ID, rep_msg), + cm_id_priv->id.state); + spin_unlock_irq(&cm_id_priv->lock); + goto error; + } + + cm_id_priv->timewait_info->work.remote_id = + cpu_to_be32(IBA_GET(CM_REP_LOCAL_COMM_ID, rep_msg)); + cm_id_priv->timewait_info->remote_ca_guid = + cpu_to_be64(IBA_GET(CM_REP_LOCAL_CA_GUID, rep_msg)); + cm_id_priv->timewait_info->remote_qpn = cm_rep_get_qpn(rep_msg, cm_id_priv->qp_type); + + spin_lock(&cm.lock); + /* Check for duplicate REP. */ + if (cm_insert_remote_id(cm_id_priv->timewait_info)) { + spin_unlock(&cm.lock); + spin_unlock_irq(&cm_id_priv->lock); + ret = -EINVAL; + trace_icm_insert_failed_err( + IBA_GET(CM_REP_REMOTE_COMM_ID, rep_msg)); + goto error; + } + /* Check for a stale connection. */ + timewait_info = cm_insert_remote_qpn(cm_id_priv->timewait_info); + if (timewait_info) { + cm_remove_remote(cm_id_priv); + cur_cm_id_priv = cm_acquire_id(timewait_info->work.local_id, + timewait_info->work.remote_id); + + spin_unlock(&cm.lock); + spin_unlock_irq(&cm_id_priv->lock); + cm_issue_rej(work->port, work->mad_recv_wc, + IB_CM_REJ_STALE_CONN, CM_MSG_RESPONSE_REP, + NULL, 0); + ret = -EINVAL; + trace_icm_staleconn_err( + IBA_GET(CM_REP_LOCAL_COMM_ID, rep_msg), + IBA_GET(CM_REP_REMOTE_COMM_ID, rep_msg)); + + if (cur_cm_id_priv) { + ib_send_cm_dreq(&cur_cm_id_priv->id, NULL, 0); + cm_deref_id(cur_cm_id_priv); + } + + goto error; + } + spin_unlock(&cm.lock); + + cm_id_priv->id.state = IB_CM_REP_RCVD; + cm_id_priv->id.remote_id = + cpu_to_be32(IBA_GET(CM_REP_LOCAL_COMM_ID, rep_msg)); + cm_id_priv->remote_qpn = cm_rep_get_qpn(rep_msg, cm_id_priv->qp_type); + cm_id_priv->initiator_depth = + IBA_GET(CM_REP_RESPONDER_RESOURCES, rep_msg); + cm_id_priv->responder_resources = + IBA_GET(CM_REP_INITIATOR_DEPTH, rep_msg); + cm_id_priv->sq_psn = cpu_to_be32(IBA_GET(CM_REP_STARTING_PSN, rep_msg)); + cm_id_priv->rnr_retry_count = IBA_GET(CM_REP_RNR_RETRY_COUNT, rep_msg); + cm_id_priv->target_ack_delay = + IBA_GET(CM_REP_TARGET_ACK_DELAY, rep_msg); + cm_id_priv->av.timeout = + cm_ack_timeout(cm_id_priv->target_ack_delay, + cm_id_priv->av.timeout - 1); + cm_id_priv->alt_av.timeout = + cm_ack_timeout(cm_id_priv->target_ack_delay, + cm_id_priv->alt_av.timeout - 1); + + ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg); + cm_queue_work_unlock(cm_id_priv, work); + return 0; + +error: + cm_deref_id(cm_id_priv); + return ret; +} + +static int cm_establish_handler(struct cm_work *work) +{ + struct cm_id_private *cm_id_priv; + + /* See comment in cm_establish about lookup. */ + cm_id_priv = cm_acquire_id(work->local_id, work->remote_id); + if (!cm_id_priv) + return -EINVAL; + + spin_lock_irq(&cm_id_priv->lock); + if (cm_id_priv->id.state != IB_CM_ESTABLISHED) { + spin_unlock_irq(&cm_id_priv->lock); + goto out; + } + + ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg); + cm_queue_work_unlock(cm_id_priv, work); + return 0; +out: + cm_deref_id(cm_id_priv); + return -EINVAL; +} + +static int cm_rtu_handler(struct cm_work *work) +{ + struct cm_id_private *cm_id_priv; + struct cm_rtu_msg *rtu_msg; + + rtu_msg = (struct cm_rtu_msg *)work->mad_recv_wc->recv_buf.mad; + cm_id_priv = cm_acquire_id( + cpu_to_be32(IBA_GET(CM_RTU_REMOTE_COMM_ID, rtu_msg)), + cpu_to_be32(IBA_GET(CM_RTU_LOCAL_COMM_ID, rtu_msg))); + if (!cm_id_priv) + return -EINVAL; + + work->cm_event.private_data = + IBA_GET_MEM_PTR(CM_RTU_PRIVATE_DATA, rtu_msg); + + spin_lock_irq(&cm_id_priv->lock); + if (cm_id_priv->id.state != IB_CM_REP_SENT && + cm_id_priv->id.state != IB_CM_MRA_REP_RCVD) { + spin_unlock_irq(&cm_id_priv->lock); + atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES]. + counter[CM_RTU_COUNTER]); + goto out; + } + cm_id_priv->id.state = IB_CM_ESTABLISHED; + + ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg); + cm_queue_work_unlock(cm_id_priv, work); + return 0; +out: + cm_deref_id(cm_id_priv); + return -EINVAL; +} + +static void cm_format_dreq(struct cm_dreq_msg *dreq_msg, + struct cm_id_private *cm_id_priv, + const void *private_data, + u8 private_data_len) +{ + cm_format_mad_hdr(&dreq_msg->hdr, CM_DREQ_ATTR_ID, + cm_form_tid(cm_id_priv)); + IBA_SET(CM_DREQ_LOCAL_COMM_ID, dreq_msg, + be32_to_cpu(cm_id_priv->id.local_id)); + IBA_SET(CM_DREQ_REMOTE_COMM_ID, dreq_msg, + be32_to_cpu(cm_id_priv->id.remote_id)); + IBA_SET(CM_DREQ_REMOTE_QPN_EECN, dreq_msg, + be32_to_cpu(cm_id_priv->remote_qpn)); + + if (private_data && private_data_len) + IBA_SET_MEM(CM_DREQ_PRIVATE_DATA, dreq_msg, private_data, + private_data_len); +} + +static int cm_send_dreq_locked(struct cm_id_private *cm_id_priv, + const void *private_data, u8 private_data_len) +{ + struct ib_mad_send_buf *msg; + int ret; + + lockdep_assert_held(&cm_id_priv->lock); + + if (private_data && private_data_len > IB_CM_DREQ_PRIVATE_DATA_SIZE) + return -EINVAL; + + if (cm_id_priv->id.state != IB_CM_ESTABLISHED) { + trace_icm_dreq_skipped(&cm_id_priv->id); + return -EINVAL; + } + + if (cm_id_priv->id.lap_state == IB_CM_LAP_SENT || + cm_id_priv->id.lap_state == IB_CM_MRA_LAP_RCVD) + ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg); + + ret = cm_alloc_msg(cm_id_priv, &msg); + if (ret) { + cm_enter_timewait(cm_id_priv); + return ret; + } + + cm_format_dreq((struct cm_dreq_msg *) msg->mad, cm_id_priv, + private_data, private_data_len); + msg->timeout_ms = cm_id_priv->timeout_ms; + msg->context[1] = (void *) (unsigned long) IB_CM_DREQ_SENT; + + trace_icm_send_dreq(&cm_id_priv->id); + ret = ib_post_send_mad(msg, NULL); + if (ret) { + cm_enter_timewait(cm_id_priv); + cm_free_msg(msg); + return ret; + } + + cm_id_priv->id.state = IB_CM_DREQ_SENT; + cm_id_priv->msg = msg; + return 0; +} + +int ib_send_cm_dreq(struct ib_cm_id *cm_id, const void *private_data, + u8 private_data_len) +{ + struct cm_id_private *cm_id_priv = + container_of(cm_id, struct cm_id_private, id); + unsigned long flags; + int ret; + + spin_lock_irqsave(&cm_id_priv->lock, flags); + ret = cm_send_dreq_locked(cm_id_priv, private_data, private_data_len); + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + return ret; +} +EXPORT_SYMBOL(ib_send_cm_dreq); + +static void cm_format_drep(struct cm_drep_msg *drep_msg, + struct cm_id_private *cm_id_priv, + const void *private_data, + u8 private_data_len) +{ + cm_format_mad_hdr(&drep_msg->hdr, CM_DREP_ATTR_ID, cm_id_priv->tid); + IBA_SET(CM_DREP_LOCAL_COMM_ID, drep_msg, + be32_to_cpu(cm_id_priv->id.local_id)); + IBA_SET(CM_DREP_REMOTE_COMM_ID, drep_msg, + be32_to_cpu(cm_id_priv->id.remote_id)); + + if (private_data && private_data_len) + IBA_SET_MEM(CM_DREP_PRIVATE_DATA, drep_msg, private_data, + private_data_len); +} + +static int cm_send_drep_locked(struct cm_id_private *cm_id_priv, + void *private_data, u8 private_data_len) +{ + struct ib_mad_send_buf *msg; + int ret; + + lockdep_assert_held(&cm_id_priv->lock); + + if (private_data && private_data_len > IB_CM_DREP_PRIVATE_DATA_SIZE) + return -EINVAL; + + if (cm_id_priv->id.state != IB_CM_DREQ_RCVD) { + trace_icm_send_drep_err(&cm_id_priv->id); + kfree(private_data); + return -EINVAL; + } + + cm_set_private_data(cm_id_priv, private_data, private_data_len); + cm_enter_timewait(cm_id_priv); + + ret = cm_alloc_msg(cm_id_priv, &msg); + if (ret) + return ret; + + cm_format_drep((struct cm_drep_msg *) msg->mad, cm_id_priv, + private_data, private_data_len); + + trace_icm_send_drep(&cm_id_priv->id); + ret = ib_post_send_mad(msg, NULL); + if (ret) { + cm_free_msg(msg); + return ret; + } + return 0; +} + +int ib_send_cm_drep(struct ib_cm_id *cm_id, const void *private_data, + u8 private_data_len) +{ + struct cm_id_private *cm_id_priv = + container_of(cm_id, struct cm_id_private, id); + unsigned long flags; + void *data; + int ret; + + data = cm_copy_private_data(private_data, private_data_len); + if (IS_ERR(data)) + return PTR_ERR(data); + + spin_lock_irqsave(&cm_id_priv->lock, flags); + ret = cm_send_drep_locked(cm_id_priv, data, private_data_len); + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + return ret; +} +EXPORT_SYMBOL(ib_send_cm_drep); + +static int cm_issue_drep(struct cm_port *port, + struct ib_mad_recv_wc *mad_recv_wc) +{ + struct ib_mad_send_buf *msg = NULL; + struct cm_dreq_msg *dreq_msg; + struct cm_drep_msg *drep_msg; + int ret; + + ret = cm_alloc_response_msg(port, mad_recv_wc, &msg); + if (ret) + return ret; + + dreq_msg = (struct cm_dreq_msg *) mad_recv_wc->recv_buf.mad; + drep_msg = (struct cm_drep_msg *) msg->mad; + + cm_format_mad_hdr(&drep_msg->hdr, CM_DREP_ATTR_ID, dreq_msg->hdr.tid); + IBA_SET(CM_DREP_REMOTE_COMM_ID, drep_msg, + IBA_GET(CM_DREQ_LOCAL_COMM_ID, dreq_msg)); + IBA_SET(CM_DREP_LOCAL_COMM_ID, drep_msg, + IBA_GET(CM_DREQ_REMOTE_COMM_ID, dreq_msg)); + + trace_icm_issue_drep( + IBA_GET(CM_DREQ_LOCAL_COMM_ID, dreq_msg), + IBA_GET(CM_DREQ_REMOTE_COMM_ID, dreq_msg)); + ret = ib_post_send_mad(msg, NULL); + if (ret) + cm_free_msg(msg); + + return ret; +} + +static int cm_dreq_handler(struct cm_work *work) +{ + struct cm_id_private *cm_id_priv; + struct cm_dreq_msg *dreq_msg; + struct ib_mad_send_buf *msg = NULL; + + dreq_msg = (struct cm_dreq_msg *)work->mad_recv_wc->recv_buf.mad; + cm_id_priv = cm_acquire_id( + cpu_to_be32(IBA_GET(CM_DREQ_REMOTE_COMM_ID, dreq_msg)), + cpu_to_be32(IBA_GET(CM_DREQ_LOCAL_COMM_ID, dreq_msg))); + if (!cm_id_priv) { + atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES]. + counter[CM_DREQ_COUNTER]); + cm_issue_drep(work->port, work->mad_recv_wc); + trace_icm_no_priv_err( + IBA_GET(CM_DREQ_LOCAL_COMM_ID, dreq_msg), + IBA_GET(CM_DREQ_REMOTE_COMM_ID, dreq_msg)); + return -EINVAL; + } + + work->cm_event.private_data = + IBA_GET_MEM_PTR(CM_DREQ_PRIVATE_DATA, dreq_msg); + + spin_lock_irq(&cm_id_priv->lock); + if (cm_id_priv->local_qpn != + cpu_to_be32(IBA_GET(CM_DREQ_REMOTE_QPN_EECN, dreq_msg))) + goto unlock; + + switch (cm_id_priv->id.state) { + case IB_CM_REP_SENT: + case IB_CM_DREQ_SENT: + ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg); + break; + case IB_CM_ESTABLISHED: + if (cm_id_priv->id.lap_state == IB_CM_LAP_SENT || + cm_id_priv->id.lap_state == IB_CM_MRA_LAP_RCVD) + ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg); + break; + case IB_CM_MRA_REP_RCVD: + break; + case IB_CM_TIMEWAIT: + atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES]. + counter[CM_DREQ_COUNTER]); + msg = cm_alloc_response_msg_no_ah(work->port, work->mad_recv_wc); + if (IS_ERR(msg)) + goto unlock; + + cm_format_drep((struct cm_drep_msg *) msg->mad, cm_id_priv, + cm_id_priv->private_data, + cm_id_priv->private_data_len); + spin_unlock_irq(&cm_id_priv->lock); + + if (cm_create_response_msg_ah(work->port, work->mad_recv_wc, msg) || + ib_post_send_mad(msg, NULL)) + cm_free_msg(msg); + goto deref; + case IB_CM_DREQ_RCVD: + atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES]. + counter[CM_DREQ_COUNTER]); + goto unlock; + default: + trace_icm_dreq_unknown_err(&cm_id_priv->id); + goto unlock; + } + cm_id_priv->id.state = IB_CM_DREQ_RCVD; + cm_id_priv->tid = dreq_msg->hdr.tid; + cm_queue_work_unlock(cm_id_priv, work); + return 0; + +unlock: spin_unlock_irq(&cm_id_priv->lock); +deref: cm_deref_id(cm_id_priv); + return -EINVAL; +} + +static int cm_drep_handler(struct cm_work *work) +{ + struct cm_id_private *cm_id_priv; + struct cm_drep_msg *drep_msg; + + drep_msg = (struct cm_drep_msg *)work->mad_recv_wc->recv_buf.mad; + cm_id_priv = cm_acquire_id( + cpu_to_be32(IBA_GET(CM_DREP_REMOTE_COMM_ID, drep_msg)), + cpu_to_be32(IBA_GET(CM_DREP_LOCAL_COMM_ID, drep_msg))); + if (!cm_id_priv) + return -EINVAL; + + work->cm_event.private_data = + IBA_GET_MEM_PTR(CM_DREP_PRIVATE_DATA, drep_msg); + + spin_lock_irq(&cm_id_priv->lock); + if (cm_id_priv->id.state != IB_CM_DREQ_SENT && + cm_id_priv->id.state != IB_CM_DREQ_RCVD) { + spin_unlock_irq(&cm_id_priv->lock); + goto out; + } + cm_enter_timewait(cm_id_priv); + + ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg); + cm_queue_work_unlock(cm_id_priv, work); + return 0; +out: + cm_deref_id(cm_id_priv); + return -EINVAL; +} + +static int cm_send_rej_locked(struct cm_id_private *cm_id_priv, + enum ib_cm_rej_reason reason, void *ari, + u8 ari_length, const void *private_data, + u8 private_data_len) +{ + enum ib_cm_state state = cm_id_priv->id.state; + struct ib_mad_send_buf *msg; + int ret; + + lockdep_assert_held(&cm_id_priv->lock); + + if ((private_data && private_data_len > IB_CM_REJ_PRIVATE_DATA_SIZE) || + (ari && ari_length > IB_CM_REJ_ARI_LENGTH)) + return -EINVAL; + + trace_icm_send_rej(&cm_id_priv->id, reason); + + switch (state) { + case IB_CM_REQ_SENT: + case IB_CM_MRA_REQ_RCVD: + case IB_CM_REQ_RCVD: + case IB_CM_MRA_REQ_SENT: + case IB_CM_REP_RCVD: + case IB_CM_MRA_REP_SENT: + cm_reset_to_idle(cm_id_priv); + ret = cm_alloc_msg(cm_id_priv, &msg); + if (ret) + return ret; + cm_format_rej((struct cm_rej_msg *)msg->mad, cm_id_priv, reason, + ari, ari_length, private_data, private_data_len, + state); + break; + case IB_CM_REP_SENT: + case IB_CM_MRA_REP_RCVD: + cm_enter_timewait(cm_id_priv); + ret = cm_alloc_msg(cm_id_priv, &msg); + if (ret) + return ret; + cm_format_rej((struct cm_rej_msg *)msg->mad, cm_id_priv, reason, + ari, ari_length, private_data, private_data_len, + state); + break; + default: + trace_icm_send_unknown_rej_err(&cm_id_priv->id); + return -EINVAL; + } + + ret = ib_post_send_mad(msg, NULL); + if (ret) { + cm_free_msg(msg); + return ret; + } + + return 0; +} + +int ib_send_cm_rej(struct ib_cm_id *cm_id, enum ib_cm_rej_reason reason, + void *ari, u8 ari_length, const void *private_data, + u8 private_data_len) +{ + struct cm_id_private *cm_id_priv = + container_of(cm_id, struct cm_id_private, id); + unsigned long flags; + int ret; + + spin_lock_irqsave(&cm_id_priv->lock, flags); + ret = cm_send_rej_locked(cm_id_priv, reason, ari, ari_length, + private_data, private_data_len); + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + return ret; +} +EXPORT_SYMBOL(ib_send_cm_rej); + +static void cm_format_rej_event(struct cm_work *work) +{ + struct cm_rej_msg *rej_msg; + struct ib_cm_rej_event_param *param; + + rej_msg = (struct cm_rej_msg *)work->mad_recv_wc->recv_buf.mad; + param = &work->cm_event.param.rej_rcvd; + param->ari = IBA_GET_MEM_PTR(CM_REJ_ARI, rej_msg); + param->ari_length = IBA_GET(CM_REJ_REJECTED_INFO_LENGTH, rej_msg); + param->reason = IBA_GET(CM_REJ_REASON, rej_msg); + work->cm_event.private_data = + IBA_GET_MEM_PTR(CM_REJ_PRIVATE_DATA, rej_msg); +} + +static struct cm_id_private * cm_acquire_rejected_id(struct cm_rej_msg *rej_msg) +{ + struct cm_id_private *cm_id_priv; + __be32 remote_id; + + remote_id = cpu_to_be32(IBA_GET(CM_REJ_LOCAL_COMM_ID, rej_msg)); + + if (IBA_GET(CM_REJ_REASON, rej_msg) == IB_CM_REJ_TIMEOUT) { + cm_id_priv = cm_find_remote_id( + *((__be64 *)IBA_GET_MEM_PTR(CM_REJ_ARI, rej_msg)), + remote_id); + } else if (IBA_GET(CM_REJ_MESSAGE_REJECTED, rej_msg) == + CM_MSG_RESPONSE_REQ) + cm_id_priv = cm_acquire_id( + cpu_to_be32(IBA_GET(CM_REJ_REMOTE_COMM_ID, rej_msg)), + 0); + else + cm_id_priv = cm_acquire_id( + cpu_to_be32(IBA_GET(CM_REJ_REMOTE_COMM_ID, rej_msg)), + remote_id); + + return cm_id_priv; +} + +static int cm_rej_handler(struct cm_work *work) +{ + struct cm_id_private *cm_id_priv; + struct cm_rej_msg *rej_msg; + + rej_msg = (struct cm_rej_msg *)work->mad_recv_wc->recv_buf.mad; + cm_id_priv = cm_acquire_rejected_id(rej_msg); + if (!cm_id_priv) + return -EINVAL; + + cm_format_rej_event(work); + + spin_lock_irq(&cm_id_priv->lock); + switch (cm_id_priv->id.state) { + case IB_CM_REQ_SENT: + case IB_CM_MRA_REQ_RCVD: + case IB_CM_REP_SENT: + case IB_CM_MRA_REP_RCVD: + ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg); + fallthrough; + case IB_CM_REQ_RCVD: + case IB_CM_MRA_REQ_SENT: + if (IBA_GET(CM_REJ_REASON, rej_msg) == IB_CM_REJ_STALE_CONN) + cm_enter_timewait(cm_id_priv); + else + cm_reset_to_idle(cm_id_priv); + break; + case IB_CM_DREQ_SENT: + ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg); + fallthrough; + case IB_CM_REP_RCVD: + case IB_CM_MRA_REP_SENT: + cm_enter_timewait(cm_id_priv); + break; + case IB_CM_ESTABLISHED: + if (cm_id_priv->id.lap_state == IB_CM_LAP_UNINIT || + cm_id_priv->id.lap_state == IB_CM_LAP_SENT) { + if (cm_id_priv->id.lap_state == IB_CM_LAP_SENT) + ib_cancel_mad(cm_id_priv->av.port->mad_agent, + cm_id_priv->msg); + cm_enter_timewait(cm_id_priv); + break; + } + fallthrough; + default: + trace_icm_rej_unknown_err(&cm_id_priv->id); + spin_unlock_irq(&cm_id_priv->lock); + goto out; + } + + cm_queue_work_unlock(cm_id_priv, work); + return 0; +out: + cm_deref_id(cm_id_priv); + return -EINVAL; +} + +int ib_send_cm_mra(struct ib_cm_id *cm_id, + u8 service_timeout, + const void *private_data, + u8 private_data_len) +{ + struct cm_id_private *cm_id_priv; + struct ib_mad_send_buf *msg; + enum ib_cm_state cm_state; + enum ib_cm_lap_state lap_state; + enum cm_msg_response msg_response; + void *data; + unsigned long flags; + int ret; + + if (private_data && private_data_len > IB_CM_MRA_PRIVATE_DATA_SIZE) + return -EINVAL; + + data = cm_copy_private_data(private_data, private_data_len); + if (IS_ERR(data)) + return PTR_ERR(data); + + cm_id_priv = container_of(cm_id, struct cm_id_private, id); + + spin_lock_irqsave(&cm_id_priv->lock, flags); + switch(cm_id_priv->id.state) { + case IB_CM_REQ_RCVD: + cm_state = IB_CM_MRA_REQ_SENT; + lap_state = cm_id->lap_state; + msg_response = CM_MSG_RESPONSE_REQ; + break; + case IB_CM_REP_RCVD: + cm_state = IB_CM_MRA_REP_SENT; + lap_state = cm_id->lap_state; + msg_response = CM_MSG_RESPONSE_REP; + break; + case IB_CM_ESTABLISHED: + if (cm_id->lap_state == IB_CM_LAP_RCVD) { + cm_state = cm_id->state; + lap_state = IB_CM_MRA_LAP_SENT; + msg_response = CM_MSG_RESPONSE_OTHER; + break; + } + fallthrough; + default: + trace_icm_send_mra_unknown_err(&cm_id_priv->id); + ret = -EINVAL; + goto error1; + } + + if (!(service_timeout & IB_CM_MRA_FLAG_DELAY)) { + ret = cm_alloc_msg(cm_id_priv, &msg); + if (ret) + goto error1; + + cm_format_mra((struct cm_mra_msg *) msg->mad, cm_id_priv, + msg_response, service_timeout, + private_data, private_data_len); + trace_icm_send_mra(cm_id); + ret = ib_post_send_mad(msg, NULL); + if (ret) + goto error2; + } + + cm_id->state = cm_state; + cm_id->lap_state = lap_state; + cm_id_priv->service_timeout = service_timeout; + cm_set_private_data(cm_id_priv, data, private_data_len); + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + return 0; + +error1: spin_unlock_irqrestore(&cm_id_priv->lock, flags); + kfree(data); + return ret; + +error2: spin_unlock_irqrestore(&cm_id_priv->lock, flags); + kfree(data); + cm_free_msg(msg); + return ret; +} +EXPORT_SYMBOL(ib_send_cm_mra); + +static struct cm_id_private * cm_acquire_mraed_id(struct cm_mra_msg *mra_msg) +{ + switch (IBA_GET(CM_MRA_MESSAGE_MRAED, mra_msg)) { + case CM_MSG_RESPONSE_REQ: + return cm_acquire_id( + cpu_to_be32(IBA_GET(CM_MRA_REMOTE_COMM_ID, mra_msg)), + 0); + case CM_MSG_RESPONSE_REP: + case CM_MSG_RESPONSE_OTHER: + return cm_acquire_id( + cpu_to_be32(IBA_GET(CM_MRA_REMOTE_COMM_ID, mra_msg)), + cpu_to_be32(IBA_GET(CM_MRA_LOCAL_COMM_ID, mra_msg))); + default: + return NULL; + } +} + +static int cm_mra_handler(struct cm_work *work) +{ + struct cm_id_private *cm_id_priv; + struct cm_mra_msg *mra_msg; + int timeout; + + mra_msg = (struct cm_mra_msg *)work->mad_recv_wc->recv_buf.mad; + cm_id_priv = cm_acquire_mraed_id(mra_msg); + if (!cm_id_priv) + return -EINVAL; + + work->cm_event.private_data = + IBA_GET_MEM_PTR(CM_MRA_PRIVATE_DATA, mra_msg); + work->cm_event.param.mra_rcvd.service_timeout = + IBA_GET(CM_MRA_SERVICE_TIMEOUT, mra_msg); + timeout = cm_convert_to_ms(IBA_GET(CM_MRA_SERVICE_TIMEOUT, mra_msg)) + + cm_convert_to_ms(cm_id_priv->av.timeout); + + spin_lock_irq(&cm_id_priv->lock); + switch (cm_id_priv->id.state) { + case IB_CM_REQ_SENT: + if (IBA_GET(CM_MRA_MESSAGE_MRAED, mra_msg) != + CM_MSG_RESPONSE_REQ || + ib_modify_mad(cm_id_priv->av.port->mad_agent, + cm_id_priv->msg, timeout)) + goto out; + cm_id_priv->id.state = IB_CM_MRA_REQ_RCVD; + break; + case IB_CM_REP_SENT: + if (IBA_GET(CM_MRA_MESSAGE_MRAED, mra_msg) != + CM_MSG_RESPONSE_REP || + ib_modify_mad(cm_id_priv->av.port->mad_agent, + cm_id_priv->msg, timeout)) + goto out; + cm_id_priv->id.state = IB_CM_MRA_REP_RCVD; + break; + case IB_CM_ESTABLISHED: + if (IBA_GET(CM_MRA_MESSAGE_MRAED, mra_msg) != + CM_MSG_RESPONSE_OTHER || + cm_id_priv->id.lap_state != IB_CM_LAP_SENT || + ib_modify_mad(cm_id_priv->av.port->mad_agent, + cm_id_priv->msg, timeout)) { + if (cm_id_priv->id.lap_state == IB_CM_MRA_LAP_RCVD) + atomic_long_inc(&work->port-> + counter_group[CM_RECV_DUPLICATES]. + counter[CM_MRA_COUNTER]); + goto out; + } + cm_id_priv->id.lap_state = IB_CM_MRA_LAP_RCVD; + break; + case IB_CM_MRA_REQ_RCVD: + case IB_CM_MRA_REP_RCVD: + atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES]. + counter[CM_MRA_COUNTER]); + fallthrough; + default: + trace_icm_mra_unknown_err(&cm_id_priv->id); + goto out; + } + + cm_id_priv->msg->context[1] = (void *) (unsigned long) + cm_id_priv->id.state; + cm_queue_work_unlock(cm_id_priv, work); + return 0; +out: + spin_unlock_irq(&cm_id_priv->lock); + cm_deref_id(cm_id_priv); + return -EINVAL; +} + +static void cm_format_path_lid_from_lap(struct cm_lap_msg *lap_msg, + struct sa_path_rec *path) +{ + u32 lid; + + if (path->rec_type != SA_PATH_REC_TYPE_OPA) { + sa_path_set_dlid(path, IBA_GET(CM_LAP_ALTERNATE_LOCAL_PORT_LID, + lap_msg)); + sa_path_set_slid(path, IBA_GET(CM_LAP_ALTERNATE_REMOTE_PORT_LID, + lap_msg)); + } else { + lid = opa_get_lid_from_gid(IBA_GET_MEM_PTR( + CM_LAP_ALTERNATE_LOCAL_PORT_GID, lap_msg)); + sa_path_set_dlid(path, lid); + + lid = opa_get_lid_from_gid(IBA_GET_MEM_PTR( + CM_LAP_ALTERNATE_REMOTE_PORT_GID, lap_msg)); + sa_path_set_slid(path, lid); + } +} + +static void cm_format_path_from_lap(struct cm_id_private *cm_id_priv, + struct sa_path_rec *path, + struct cm_lap_msg *lap_msg) +{ + path->dgid = *IBA_GET_MEM_PTR(CM_LAP_ALTERNATE_LOCAL_PORT_GID, lap_msg); + path->sgid = + *IBA_GET_MEM_PTR(CM_LAP_ALTERNATE_REMOTE_PORT_GID, lap_msg); + path->flow_label = + cpu_to_be32(IBA_GET(CM_LAP_ALTERNATE_FLOW_LABEL, lap_msg)); + path->hop_limit = IBA_GET(CM_LAP_ALTERNATE_HOP_LIMIT, lap_msg); + path->traffic_class = IBA_GET(CM_LAP_ALTERNATE_TRAFFIC_CLASS, lap_msg); + path->reversible = 1; + path->pkey = cm_id_priv->pkey; + path->sl = IBA_GET(CM_LAP_ALTERNATE_SL, lap_msg); + path->mtu_selector = IB_SA_EQ; + path->mtu = cm_id_priv->path_mtu; + path->rate_selector = IB_SA_EQ; + path->rate = IBA_GET(CM_LAP_ALTERNATE_PACKET_RATE, lap_msg); + path->packet_life_time_selector = IB_SA_EQ; + path->packet_life_time = + IBA_GET(CM_LAP_ALTERNATE_LOCAL_ACK_TIMEOUT, lap_msg); + path->packet_life_time -= (path->packet_life_time > 0); + cm_format_path_lid_from_lap(lap_msg, path); +} + +static int cm_lap_handler(struct cm_work *work) +{ + struct cm_id_private *cm_id_priv; + struct cm_lap_msg *lap_msg; + struct ib_cm_lap_event_param *param; + struct ib_mad_send_buf *msg = NULL; + int ret; + + /* Currently Alternate path messages are not supported for + * RoCE link layer. + */ + if (rdma_protocol_roce(work->port->cm_dev->ib_device, + work->port->port_num)) + return -EINVAL; + + /* todo: verify LAP request and send reject APR if invalid. */ + lap_msg = (struct cm_lap_msg *)work->mad_recv_wc->recv_buf.mad; + cm_id_priv = cm_acquire_id( + cpu_to_be32(IBA_GET(CM_LAP_REMOTE_COMM_ID, lap_msg)), + cpu_to_be32(IBA_GET(CM_LAP_LOCAL_COMM_ID, lap_msg))); + if (!cm_id_priv) + return -EINVAL; + + param = &work->cm_event.param.lap_rcvd; + memset(&work->path[0], 0, sizeof(work->path[1])); + cm_path_set_rec_type(work->port->cm_dev->ib_device, + work->port->port_num, &work->path[0], + IBA_GET_MEM_PTR(CM_LAP_ALTERNATE_LOCAL_PORT_GID, + lap_msg)); + param->alternate_path = &work->path[0]; + cm_format_path_from_lap(cm_id_priv, param->alternate_path, lap_msg); + work->cm_event.private_data = + IBA_GET_MEM_PTR(CM_LAP_PRIVATE_DATA, lap_msg); + + spin_lock_irq(&cm_id_priv->lock); + if (cm_id_priv->id.state != IB_CM_ESTABLISHED) + goto unlock; + + switch (cm_id_priv->id.lap_state) { + case IB_CM_LAP_UNINIT: + case IB_CM_LAP_IDLE: + break; + case IB_CM_MRA_LAP_SENT: + atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES]. + counter[CM_LAP_COUNTER]); + msg = cm_alloc_response_msg_no_ah(work->port, work->mad_recv_wc); + if (IS_ERR(msg)) + goto unlock; + + cm_format_mra((struct cm_mra_msg *) msg->mad, cm_id_priv, + CM_MSG_RESPONSE_OTHER, + cm_id_priv->service_timeout, + cm_id_priv->private_data, + cm_id_priv->private_data_len); + spin_unlock_irq(&cm_id_priv->lock); + + if (cm_create_response_msg_ah(work->port, work->mad_recv_wc, msg) || + ib_post_send_mad(msg, NULL)) + cm_free_msg(msg); + goto deref; + case IB_CM_LAP_RCVD: + atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES]. + counter[CM_LAP_COUNTER]); + goto unlock; + default: + goto unlock; + } + + ret = cm_init_av_for_lap(work->port, work->mad_recv_wc->wc, + work->mad_recv_wc->recv_buf.grh, + &cm_id_priv->av); + if (ret) + goto unlock; + + ret = cm_init_av_by_path(param->alternate_path, NULL, + &cm_id_priv->alt_av, cm_id_priv); + if (ret) + goto unlock; + + cm_id_priv->id.lap_state = IB_CM_LAP_RCVD; + cm_id_priv->tid = lap_msg->hdr.tid; + cm_queue_work_unlock(cm_id_priv, work); + return 0; + +unlock: spin_unlock_irq(&cm_id_priv->lock); +deref: cm_deref_id(cm_id_priv); + return -EINVAL; +} + +static int cm_apr_handler(struct cm_work *work) +{ + struct cm_id_private *cm_id_priv; + struct cm_apr_msg *apr_msg; + + /* Currently Alternate path messages are not supported for + * RoCE link layer. + */ + if (rdma_protocol_roce(work->port->cm_dev->ib_device, + work->port->port_num)) + return -EINVAL; + + apr_msg = (struct cm_apr_msg *)work->mad_recv_wc->recv_buf.mad; + cm_id_priv = cm_acquire_id( + cpu_to_be32(IBA_GET(CM_APR_REMOTE_COMM_ID, apr_msg)), + cpu_to_be32(IBA_GET(CM_APR_LOCAL_COMM_ID, apr_msg))); + if (!cm_id_priv) + return -EINVAL; /* Unmatched reply. */ + + work->cm_event.param.apr_rcvd.ap_status = + IBA_GET(CM_APR_AR_STATUS, apr_msg); + work->cm_event.param.apr_rcvd.apr_info = + IBA_GET_MEM_PTR(CM_APR_ADDITIONAL_INFORMATION, apr_msg); + work->cm_event.param.apr_rcvd.info_len = + IBA_GET(CM_APR_ADDITIONAL_INFORMATION_LENGTH, apr_msg); + work->cm_event.private_data = + IBA_GET_MEM_PTR(CM_APR_PRIVATE_DATA, apr_msg); + + spin_lock_irq(&cm_id_priv->lock); + if (cm_id_priv->id.state != IB_CM_ESTABLISHED || + (cm_id_priv->id.lap_state != IB_CM_LAP_SENT && + cm_id_priv->id.lap_state != IB_CM_MRA_LAP_RCVD)) { + spin_unlock_irq(&cm_id_priv->lock); + goto out; + } + cm_id_priv->id.lap_state = IB_CM_LAP_IDLE; + ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg); + cm_id_priv->msg = NULL; + cm_queue_work_unlock(cm_id_priv, work); + return 0; +out: + cm_deref_id(cm_id_priv); + return -EINVAL; +} + +static int cm_timewait_handler(struct cm_work *work) +{ + struct cm_timewait_info *timewait_info; + struct cm_id_private *cm_id_priv; + + timewait_info = container_of(work, struct cm_timewait_info, work); + spin_lock_irq(&cm.lock); + list_del(&timewait_info->list); + spin_unlock_irq(&cm.lock); + + cm_id_priv = cm_acquire_id(timewait_info->work.local_id, + timewait_info->work.remote_id); + if (!cm_id_priv) + return -EINVAL; + + spin_lock_irq(&cm_id_priv->lock); + if (cm_id_priv->id.state != IB_CM_TIMEWAIT || + cm_id_priv->remote_qpn != timewait_info->remote_qpn) { + spin_unlock_irq(&cm_id_priv->lock); + goto out; + } + cm_id_priv->id.state = IB_CM_IDLE; + cm_queue_work_unlock(cm_id_priv, work); + return 0; +out: + cm_deref_id(cm_id_priv); + return -EINVAL; +} + +static void cm_format_sidr_req(struct cm_sidr_req_msg *sidr_req_msg, + struct cm_id_private *cm_id_priv, + struct ib_cm_sidr_req_param *param) +{ + cm_format_mad_hdr(&sidr_req_msg->hdr, CM_SIDR_REQ_ATTR_ID, + cm_form_tid(cm_id_priv)); + IBA_SET(CM_SIDR_REQ_REQUESTID, sidr_req_msg, + be32_to_cpu(cm_id_priv->id.local_id)); + IBA_SET(CM_SIDR_REQ_PARTITION_KEY, sidr_req_msg, + be16_to_cpu(param->path->pkey)); + IBA_SET(CM_SIDR_REQ_SERVICEID, sidr_req_msg, + be64_to_cpu(param->service_id)); + + if (param->private_data && param->private_data_len) + IBA_SET_MEM(CM_SIDR_REQ_PRIVATE_DATA, sidr_req_msg, + param->private_data, param->private_data_len); +} + +int ib_send_cm_sidr_req(struct ib_cm_id *cm_id, + struct ib_cm_sidr_req_param *param) +{ + struct cm_id_private *cm_id_priv; + struct ib_mad_send_buf *msg; + unsigned long flags; + int ret; + + if (!param->path || (param->private_data && + param->private_data_len > IB_CM_SIDR_REQ_PRIVATE_DATA_SIZE)) + return -EINVAL; + + cm_id_priv = container_of(cm_id, struct cm_id_private, id); + ret = cm_init_av_by_path(param->path, param->sgid_attr, + &cm_id_priv->av, + cm_id_priv); + if (ret) + goto out; + + cm_id->service_id = param->service_id; + cm_id->service_mask = ~cpu_to_be64(0); + cm_id_priv->timeout_ms = param->timeout_ms; + cm_id_priv->max_cm_retries = param->max_cm_retries; + ret = cm_alloc_msg(cm_id_priv, &msg); + if (ret) + goto out; + + cm_format_sidr_req((struct cm_sidr_req_msg *) msg->mad, cm_id_priv, + param); + msg->timeout_ms = cm_id_priv->timeout_ms; + msg->context[1] = (void *) (unsigned long) IB_CM_SIDR_REQ_SENT; + + spin_lock_irqsave(&cm_id_priv->lock, flags); + if (cm_id->state == IB_CM_IDLE) { + trace_icm_send_sidr_req(&cm_id_priv->id); + ret = ib_post_send_mad(msg, NULL); + } else { + ret = -EINVAL; + } + + if (ret) { + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + cm_free_msg(msg); + goto out; + } + cm_id->state = IB_CM_SIDR_REQ_SENT; + cm_id_priv->msg = msg; + spin_unlock_irqrestore(&cm_id_priv->lock, flags); +out: + return ret; +} +EXPORT_SYMBOL(ib_send_cm_sidr_req); + +static void cm_format_sidr_req_event(struct cm_work *work, + const struct cm_id_private *rx_cm_id, + struct ib_cm_id *listen_id) +{ + struct cm_sidr_req_msg *sidr_req_msg; + struct ib_cm_sidr_req_event_param *param; + + sidr_req_msg = (struct cm_sidr_req_msg *) + work->mad_recv_wc->recv_buf.mad; + param = &work->cm_event.param.sidr_req_rcvd; + param->pkey = IBA_GET(CM_SIDR_REQ_PARTITION_KEY, sidr_req_msg); + param->listen_id = listen_id; + param->service_id = + cpu_to_be64(IBA_GET(CM_SIDR_REQ_SERVICEID, sidr_req_msg)); + param->bth_pkey = cm_get_bth_pkey(work); + param->port = work->port->port_num; + param->sgid_attr = rx_cm_id->av.ah_attr.grh.sgid_attr; + work->cm_event.private_data = + IBA_GET_MEM_PTR(CM_SIDR_REQ_PRIVATE_DATA, sidr_req_msg); +} + +static int cm_sidr_req_handler(struct cm_work *work) +{ + struct cm_id_private *cm_id_priv, *listen_cm_id_priv; + struct cm_sidr_req_msg *sidr_req_msg; + struct ib_wc *wc; + int ret; + + cm_id_priv = + cm_alloc_id_priv(work->port->cm_dev->ib_device, NULL, NULL); + if (IS_ERR(cm_id_priv)) + return PTR_ERR(cm_id_priv); + + /* Record SGID/SLID and request ID for lookup. */ + sidr_req_msg = (struct cm_sidr_req_msg *) + work->mad_recv_wc->recv_buf.mad; + + cm_id_priv->id.remote_id = + cpu_to_be32(IBA_GET(CM_SIDR_REQ_REQUESTID, sidr_req_msg)); + cm_id_priv->id.service_id = + cpu_to_be64(IBA_GET(CM_SIDR_REQ_SERVICEID, sidr_req_msg)); + cm_id_priv->id.service_mask = ~cpu_to_be64(0); + cm_id_priv->tid = sidr_req_msg->hdr.tid; + + wc = work->mad_recv_wc->wc; + cm_id_priv->av.dgid.global.subnet_prefix = cpu_to_be64(wc->slid); + cm_id_priv->av.dgid.global.interface_id = 0; + ret = cm_init_av_for_response(work->port, work->mad_recv_wc->wc, + work->mad_recv_wc->recv_buf.grh, + &cm_id_priv->av); + if (ret) + goto out; + + spin_lock_irq(&cm.lock); + listen_cm_id_priv = cm_insert_remote_sidr(cm_id_priv); + if (listen_cm_id_priv) { + spin_unlock_irq(&cm.lock); + atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES]. + counter[CM_SIDR_REQ_COUNTER]); + goto out; /* Duplicate message. */ + } + cm_id_priv->id.state = IB_CM_SIDR_REQ_RCVD; + listen_cm_id_priv = cm_find_listen(cm_id_priv->id.device, + cm_id_priv->id.service_id); + if (!listen_cm_id_priv) { + spin_unlock_irq(&cm.lock); + ib_send_cm_sidr_rep(&cm_id_priv->id, + &(struct ib_cm_sidr_rep_param){ + .status = IB_SIDR_UNSUPPORTED }); + goto out; /* No match. */ + } + spin_unlock_irq(&cm.lock); + + cm_id_priv->id.cm_handler = listen_cm_id_priv->id.cm_handler; + cm_id_priv->id.context = listen_cm_id_priv->id.context; + + /* + * A SIDR ID does not need to be in the xarray since it does not receive + * mads, is not placed in the remote_id or remote_qpn rbtree, and does + * not enter timewait. + */ + + cm_format_sidr_req_event(work, cm_id_priv, &listen_cm_id_priv->id); + ret = cm_id_priv->id.cm_handler(&cm_id_priv->id, &work->cm_event); + cm_free_work(work); + /* + * A pointer to the listen_cm_id is held in the event, so this deref + * must be after the event is delivered above. + */ + cm_deref_id(listen_cm_id_priv); + if (ret) + cm_destroy_id(&cm_id_priv->id, ret); + return 0; +out: + ib_destroy_cm_id(&cm_id_priv->id); + return -EINVAL; +} + +static void cm_format_sidr_rep(struct cm_sidr_rep_msg *sidr_rep_msg, + struct cm_id_private *cm_id_priv, + struct ib_cm_sidr_rep_param *param) +{ + cm_format_mad_ece_hdr(&sidr_rep_msg->hdr, CM_SIDR_REP_ATTR_ID, + cm_id_priv->tid, param->ece.attr_mod); + IBA_SET(CM_SIDR_REP_REQUESTID, sidr_rep_msg, + be32_to_cpu(cm_id_priv->id.remote_id)); + IBA_SET(CM_SIDR_REP_STATUS, sidr_rep_msg, param->status); + IBA_SET(CM_SIDR_REP_QPN, sidr_rep_msg, param->qp_num); + IBA_SET(CM_SIDR_REP_SERVICEID, sidr_rep_msg, + be64_to_cpu(cm_id_priv->id.service_id)); + IBA_SET(CM_SIDR_REP_Q_KEY, sidr_rep_msg, param->qkey); + IBA_SET(CM_SIDR_REP_VENDOR_ID_L, sidr_rep_msg, + param->ece.vendor_id & 0xFF); + IBA_SET(CM_SIDR_REP_VENDOR_ID_H, sidr_rep_msg, + (param->ece.vendor_id >> 8) & 0xFF); + + if (param->info && param->info_length) + IBA_SET_MEM(CM_SIDR_REP_ADDITIONAL_INFORMATION, sidr_rep_msg, + param->info, param->info_length); + + if (param->private_data && param->private_data_len) + IBA_SET_MEM(CM_SIDR_REP_PRIVATE_DATA, sidr_rep_msg, + param->private_data, param->private_data_len); +} + +static int cm_send_sidr_rep_locked(struct cm_id_private *cm_id_priv, + struct ib_cm_sidr_rep_param *param) +{ + struct ib_mad_send_buf *msg; + unsigned long flags; + int ret; + + lockdep_assert_held(&cm_id_priv->lock); + + if ((param->info && param->info_length > IB_CM_SIDR_REP_INFO_LENGTH) || + (param->private_data && + param->private_data_len > IB_CM_SIDR_REP_PRIVATE_DATA_SIZE)) + return -EINVAL; + + if (cm_id_priv->id.state != IB_CM_SIDR_REQ_RCVD) + return -EINVAL; + + ret = cm_alloc_msg(cm_id_priv, &msg); + if (ret) + return ret; + + cm_format_sidr_rep((struct cm_sidr_rep_msg *) msg->mad, cm_id_priv, + param); + trace_icm_send_sidr_rep(&cm_id_priv->id); + ret = ib_post_send_mad(msg, NULL); + if (ret) { + cm_free_msg(msg); + return ret; + } + cm_id_priv->id.state = IB_CM_IDLE; + spin_lock_irqsave(&cm.lock, flags); + if (!RB_EMPTY_NODE(&cm_id_priv->sidr_id_node)) { + rb_erase(&cm_id_priv->sidr_id_node, &cm.remote_sidr_table); + RB_CLEAR_NODE(&cm_id_priv->sidr_id_node); + } + spin_unlock_irqrestore(&cm.lock, flags); + return 0; +} + +int ib_send_cm_sidr_rep(struct ib_cm_id *cm_id, + struct ib_cm_sidr_rep_param *param) +{ + struct cm_id_private *cm_id_priv = + container_of(cm_id, struct cm_id_private, id); + unsigned long flags; + int ret; + + spin_lock_irqsave(&cm_id_priv->lock, flags); + ret = cm_send_sidr_rep_locked(cm_id_priv, param); + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + return ret; +} +EXPORT_SYMBOL(ib_send_cm_sidr_rep); + +static void cm_format_sidr_rep_event(struct cm_work *work, + const struct cm_id_private *cm_id_priv) +{ + struct cm_sidr_rep_msg *sidr_rep_msg; + struct ib_cm_sidr_rep_event_param *param; + + sidr_rep_msg = (struct cm_sidr_rep_msg *) + work->mad_recv_wc->recv_buf.mad; + param = &work->cm_event.param.sidr_rep_rcvd; + param->status = IBA_GET(CM_SIDR_REP_STATUS, sidr_rep_msg); + param->qkey = IBA_GET(CM_SIDR_REP_Q_KEY, sidr_rep_msg); + param->qpn = IBA_GET(CM_SIDR_REP_QPN, sidr_rep_msg); + param->info = IBA_GET_MEM_PTR(CM_SIDR_REP_ADDITIONAL_INFORMATION, + sidr_rep_msg); + param->info_len = IBA_GET(CM_SIDR_REP_ADDITIONAL_INFORMATION_LENGTH, + sidr_rep_msg); + param->sgid_attr = cm_id_priv->av.ah_attr.grh.sgid_attr; + work->cm_event.private_data = + IBA_GET_MEM_PTR(CM_SIDR_REP_PRIVATE_DATA, sidr_rep_msg); +} + +static int cm_sidr_rep_handler(struct cm_work *work) +{ + struct cm_sidr_rep_msg *sidr_rep_msg; + struct cm_id_private *cm_id_priv; + + sidr_rep_msg = (struct cm_sidr_rep_msg *) + work->mad_recv_wc->recv_buf.mad; + cm_id_priv = cm_acquire_id( + cpu_to_be32(IBA_GET(CM_SIDR_REP_REQUESTID, sidr_rep_msg)), 0); + if (!cm_id_priv) + return -EINVAL; /* Unmatched reply. */ + + spin_lock_irq(&cm_id_priv->lock); + if (cm_id_priv->id.state != IB_CM_SIDR_REQ_SENT) { + spin_unlock_irq(&cm_id_priv->lock); + goto out; + } + cm_id_priv->id.state = IB_CM_IDLE; + ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg); + spin_unlock_irq(&cm_id_priv->lock); + + cm_format_sidr_rep_event(work, cm_id_priv); + cm_process_work(cm_id_priv, work); + return 0; +out: + cm_deref_id(cm_id_priv); + return -EINVAL; +} + +static void cm_process_send_error(struct ib_mad_send_buf *msg, + enum ib_wc_status wc_status) +{ + struct cm_id_private *cm_id_priv; + struct ib_cm_event cm_event; + enum ib_cm_state state; + int ret; + + memset(&cm_event, 0, sizeof cm_event); + cm_id_priv = msg->context[0]; + + /* Discard old sends or ones without a response. */ + spin_lock_irq(&cm_id_priv->lock); + state = (enum ib_cm_state) (unsigned long) msg->context[1]; + if (msg != cm_id_priv->msg || state != cm_id_priv->id.state) + goto discard; + + trace_icm_mad_send_err(state, wc_status); + switch (state) { + case IB_CM_REQ_SENT: + case IB_CM_MRA_REQ_RCVD: + cm_reset_to_idle(cm_id_priv); + cm_event.event = IB_CM_REQ_ERROR; + break; + case IB_CM_REP_SENT: + case IB_CM_MRA_REP_RCVD: + cm_reset_to_idle(cm_id_priv); + cm_event.event = IB_CM_REP_ERROR; + break; + case IB_CM_DREQ_SENT: + cm_enter_timewait(cm_id_priv); + cm_event.event = IB_CM_DREQ_ERROR; + break; + case IB_CM_SIDR_REQ_SENT: + cm_id_priv->id.state = IB_CM_IDLE; + cm_event.event = IB_CM_SIDR_REQ_ERROR; + break; + default: + goto discard; + } + spin_unlock_irq(&cm_id_priv->lock); + cm_event.param.send_status = wc_status; + + /* No other events can occur on the cm_id at this point. */ + ret = cm_id_priv->id.cm_handler(&cm_id_priv->id, &cm_event); + cm_free_msg(msg); + if (ret) + ib_destroy_cm_id(&cm_id_priv->id); + return; +discard: + spin_unlock_irq(&cm_id_priv->lock); + cm_free_msg(msg); +} + +static void cm_send_handler(struct ib_mad_agent *mad_agent, + struct ib_mad_send_wc *mad_send_wc) +{ + struct ib_mad_send_buf *msg = mad_send_wc->send_buf; + struct cm_port *port; + u16 attr_index; + + port = mad_agent->context; + attr_index = be16_to_cpu(((struct ib_mad_hdr *) + msg->mad)->attr_id) - CM_ATTR_ID_OFFSET; + + /* + * If the send was in response to a received message (context[0] is not + * set to a cm_id), and is not a REJ, then it is a send that was + * manually retried. + */ + if (!msg->context[0] && (attr_index != CM_REJ_COUNTER)) + msg->retries = 1; + + atomic_long_add(1 + msg->retries, + &port->counter_group[CM_XMIT].counter[attr_index]); + if (msg->retries) + atomic_long_add(msg->retries, + &port->counter_group[CM_XMIT_RETRIES]. + counter[attr_index]); + + switch (mad_send_wc->status) { + case IB_WC_SUCCESS: + case IB_WC_WR_FLUSH_ERR: + cm_free_msg(msg); + break; + default: + if (msg->context[0] && msg->context[1]) + cm_process_send_error(msg, mad_send_wc->status); + else + cm_free_msg(msg); + break; + } +} + +static void cm_work_handler(struct work_struct *_work) +{ + struct cm_work *work = container_of(_work, struct cm_work, work.work); + int ret; + + switch (work->cm_event.event) { + case IB_CM_REQ_RECEIVED: + ret = cm_req_handler(work); + break; + case IB_CM_MRA_RECEIVED: + ret = cm_mra_handler(work); + break; + case IB_CM_REJ_RECEIVED: + ret = cm_rej_handler(work); + break; + case IB_CM_REP_RECEIVED: + ret = cm_rep_handler(work); + break; + case IB_CM_RTU_RECEIVED: + ret = cm_rtu_handler(work); + break; + case IB_CM_USER_ESTABLISHED: + ret = cm_establish_handler(work); + break; + case IB_CM_DREQ_RECEIVED: + ret = cm_dreq_handler(work); + break; + case IB_CM_DREP_RECEIVED: + ret = cm_drep_handler(work); + break; + case IB_CM_SIDR_REQ_RECEIVED: + ret = cm_sidr_req_handler(work); + break; + case IB_CM_SIDR_REP_RECEIVED: + ret = cm_sidr_rep_handler(work); + break; + case IB_CM_LAP_RECEIVED: + ret = cm_lap_handler(work); + break; + case IB_CM_APR_RECEIVED: + ret = cm_apr_handler(work); + break; + case IB_CM_TIMEWAIT_EXIT: + ret = cm_timewait_handler(work); + break; + default: + trace_icm_handler_err(work->cm_event.event); + ret = -EINVAL; + break; + } + if (ret) + cm_free_work(work); +} + +static int cm_establish(struct ib_cm_id *cm_id) +{ + struct cm_id_private *cm_id_priv; + struct cm_work *work; + unsigned long flags; + int ret = 0; + struct cm_device *cm_dev; + + cm_dev = ib_get_client_data(cm_id->device, &cm_client); + if (!cm_dev) + return -ENODEV; + + work = kmalloc(sizeof *work, GFP_ATOMIC); + if (!work) + return -ENOMEM; + + cm_id_priv = container_of(cm_id, struct cm_id_private, id); + spin_lock_irqsave(&cm_id_priv->lock, flags); + switch (cm_id->state) + { + case IB_CM_REP_SENT: + case IB_CM_MRA_REP_RCVD: + cm_id->state = IB_CM_ESTABLISHED; + break; + case IB_CM_ESTABLISHED: + ret = -EISCONN; + break; + default: + trace_icm_establish_err(cm_id); + ret = -EINVAL; + break; + } + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + + if (ret) { + kfree(work); + goto out; + } + + /* + * The CM worker thread may try to destroy the cm_id before it + * can execute this work item. To prevent potential deadlock, + * we need to find the cm_id once we're in the context of the + * worker thread, rather than holding a reference on it. + */ + INIT_DELAYED_WORK(&work->work, cm_work_handler); + work->local_id = cm_id->local_id; + work->remote_id = cm_id->remote_id; + work->mad_recv_wc = NULL; + work->cm_event.event = IB_CM_USER_ESTABLISHED; + + /* Check if the device started its remove_one */ + spin_lock_irqsave(&cm.lock, flags); + if (!cm_dev->going_down) { + queue_delayed_work(cm.wq, &work->work, 0); + } else { + kfree(work); + ret = -ENODEV; + } + spin_unlock_irqrestore(&cm.lock, flags); + +out: + return ret; +} + +static int cm_migrate(struct ib_cm_id *cm_id) +{ + struct cm_id_private *cm_id_priv; + struct cm_av tmp_av; + unsigned long flags; + int tmp_send_port_not_ready; + int ret = 0; + + cm_id_priv = container_of(cm_id, struct cm_id_private, id); + spin_lock_irqsave(&cm_id_priv->lock, flags); + if (cm_id->state == IB_CM_ESTABLISHED && + (cm_id->lap_state == IB_CM_LAP_UNINIT || + cm_id->lap_state == IB_CM_LAP_IDLE)) { + cm_id->lap_state = IB_CM_LAP_IDLE; + /* Swap address vector */ + tmp_av = cm_id_priv->av; + cm_id_priv->av = cm_id_priv->alt_av; + cm_id_priv->alt_av = tmp_av; + /* Swap port send ready state */ + tmp_send_port_not_ready = cm_id_priv->prim_send_port_not_ready; + cm_id_priv->prim_send_port_not_ready = cm_id_priv->altr_send_port_not_ready; + cm_id_priv->altr_send_port_not_ready = tmp_send_port_not_ready; + } else + ret = -EINVAL; + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + + return ret; +} + +int ib_cm_notify(struct ib_cm_id *cm_id, enum ib_event_type event) +{ + int ret; + + switch (event) { + case IB_EVENT_COMM_EST: + ret = cm_establish(cm_id); + break; + case IB_EVENT_PATH_MIG: + ret = cm_migrate(cm_id); + break; + default: + ret = -EINVAL; + } + return ret; +} +EXPORT_SYMBOL(ib_cm_notify); + +static void cm_recv_handler(struct ib_mad_agent *mad_agent, + struct ib_mad_send_buf *send_buf, + struct ib_mad_recv_wc *mad_recv_wc) +{ + struct cm_port *port = mad_agent->context; + struct cm_work *work; + enum ib_cm_event_type event; + bool alt_path = false; + u16 attr_id; + int paths = 0; + int going_down = 0; + + switch (mad_recv_wc->recv_buf.mad->mad_hdr.attr_id) { + case CM_REQ_ATTR_ID: + alt_path = cm_req_has_alt_path((struct cm_req_msg *) + mad_recv_wc->recv_buf.mad); + paths = 1 + (alt_path != 0); + event = IB_CM_REQ_RECEIVED; + break; + case CM_MRA_ATTR_ID: + event = IB_CM_MRA_RECEIVED; + break; + case CM_REJ_ATTR_ID: + event = IB_CM_REJ_RECEIVED; + break; + case CM_REP_ATTR_ID: + event = IB_CM_REP_RECEIVED; + break; + case CM_RTU_ATTR_ID: + event = IB_CM_RTU_RECEIVED; + break; + case CM_DREQ_ATTR_ID: + event = IB_CM_DREQ_RECEIVED; + break; + case CM_DREP_ATTR_ID: + event = IB_CM_DREP_RECEIVED; + break; + case CM_SIDR_REQ_ATTR_ID: + event = IB_CM_SIDR_REQ_RECEIVED; + break; + case CM_SIDR_REP_ATTR_ID: + event = IB_CM_SIDR_REP_RECEIVED; + break; + case CM_LAP_ATTR_ID: + paths = 1; + event = IB_CM_LAP_RECEIVED; + break; + case CM_APR_ATTR_ID: + event = IB_CM_APR_RECEIVED; + break; + default: + ib_free_recv_mad(mad_recv_wc); + return; + } + + attr_id = be16_to_cpu(mad_recv_wc->recv_buf.mad->mad_hdr.attr_id); + atomic_long_inc(&port->counter_group[CM_RECV]. + counter[attr_id - CM_ATTR_ID_OFFSET]); + + work = kmalloc(struct_size(work, path, paths), GFP_KERNEL); + if (!work) { + ib_free_recv_mad(mad_recv_wc); + return; + } + + INIT_DELAYED_WORK(&work->work, cm_work_handler); + work->cm_event.event = event; + work->mad_recv_wc = mad_recv_wc; + work->port = port; + + /* Check if the device started its remove_one */ + spin_lock_irq(&cm.lock); + if (!port->cm_dev->going_down) + queue_delayed_work(cm.wq, &work->work, 0); + else + going_down = 1; + spin_unlock_irq(&cm.lock); + + if (going_down) { + kfree(work); + ib_free_recv_mad(mad_recv_wc); + } +} + +static int cm_init_qp_init_attr(struct cm_id_private *cm_id_priv, + struct ib_qp_attr *qp_attr, + int *qp_attr_mask) +{ + unsigned long flags; + int ret; + + spin_lock_irqsave(&cm_id_priv->lock, flags); + switch (cm_id_priv->id.state) { + case IB_CM_REQ_SENT: + case IB_CM_MRA_REQ_RCVD: + case IB_CM_REQ_RCVD: + case IB_CM_MRA_REQ_SENT: + case IB_CM_REP_RCVD: + case IB_CM_MRA_REP_SENT: + case IB_CM_REP_SENT: + case IB_CM_MRA_REP_RCVD: + case IB_CM_ESTABLISHED: + *qp_attr_mask = IB_QP_STATE | IB_QP_ACCESS_FLAGS | + IB_QP_PKEY_INDEX | IB_QP_PORT; + qp_attr->qp_access_flags = IB_ACCESS_REMOTE_WRITE; + if (cm_id_priv->responder_resources) + qp_attr->qp_access_flags |= IB_ACCESS_REMOTE_READ | + IB_ACCESS_REMOTE_ATOMIC; + qp_attr->pkey_index = cm_id_priv->av.pkey_index; + qp_attr->port_num = cm_id_priv->av.port->port_num; + ret = 0; + break; + default: + trace_icm_qp_init_err(&cm_id_priv->id); + ret = -EINVAL; + break; + } + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + return ret; +} + +static int cm_init_qp_rtr_attr(struct cm_id_private *cm_id_priv, + struct ib_qp_attr *qp_attr, + int *qp_attr_mask) +{ + unsigned long flags; + int ret; + + spin_lock_irqsave(&cm_id_priv->lock, flags); + switch (cm_id_priv->id.state) { + case IB_CM_REQ_RCVD: + case IB_CM_MRA_REQ_SENT: + case IB_CM_REP_RCVD: + case IB_CM_MRA_REP_SENT: + case IB_CM_REP_SENT: + case IB_CM_MRA_REP_RCVD: + case IB_CM_ESTABLISHED: + *qp_attr_mask = IB_QP_STATE | IB_QP_AV | IB_QP_PATH_MTU | + IB_QP_DEST_QPN | IB_QP_RQ_PSN; + qp_attr->ah_attr = cm_id_priv->av.ah_attr; + qp_attr->path_mtu = cm_id_priv->path_mtu; + qp_attr->dest_qp_num = be32_to_cpu(cm_id_priv->remote_qpn); + qp_attr->rq_psn = be32_to_cpu(cm_id_priv->rq_psn); + if (cm_id_priv->qp_type == IB_QPT_RC || + cm_id_priv->qp_type == IB_QPT_XRC_TGT) { + *qp_attr_mask |= IB_QP_MAX_DEST_RD_ATOMIC | + IB_QP_MIN_RNR_TIMER; + qp_attr->max_dest_rd_atomic = + cm_id_priv->responder_resources; + qp_attr->min_rnr_timer = 0; + } + if (rdma_ah_get_dlid(&cm_id_priv->alt_av.ah_attr)) { + *qp_attr_mask |= IB_QP_ALT_PATH; + qp_attr->alt_port_num = cm_id_priv->alt_av.port->port_num; + qp_attr->alt_pkey_index = cm_id_priv->alt_av.pkey_index; + qp_attr->alt_timeout = cm_id_priv->alt_av.timeout; + qp_attr->alt_ah_attr = cm_id_priv->alt_av.ah_attr; + } + ret = 0; + break; + default: + trace_icm_qp_rtr_err(&cm_id_priv->id); + ret = -EINVAL; + break; + } + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + return ret; +} + +static int cm_init_qp_rts_attr(struct cm_id_private *cm_id_priv, + struct ib_qp_attr *qp_attr, + int *qp_attr_mask) +{ + unsigned long flags; + int ret; + + spin_lock_irqsave(&cm_id_priv->lock, flags); + switch (cm_id_priv->id.state) { + /* Allow transition to RTS before sending REP */ + case IB_CM_REQ_RCVD: + case IB_CM_MRA_REQ_SENT: + + case IB_CM_REP_RCVD: + case IB_CM_MRA_REP_SENT: + case IB_CM_REP_SENT: + case IB_CM_MRA_REP_RCVD: + case IB_CM_ESTABLISHED: + if (cm_id_priv->id.lap_state == IB_CM_LAP_UNINIT) { + *qp_attr_mask = IB_QP_STATE | IB_QP_SQ_PSN; + qp_attr->sq_psn = be32_to_cpu(cm_id_priv->sq_psn); + switch (cm_id_priv->qp_type) { + case IB_QPT_RC: + case IB_QPT_XRC_INI: + *qp_attr_mask |= IB_QP_RETRY_CNT | IB_QP_RNR_RETRY | + IB_QP_MAX_QP_RD_ATOMIC; + qp_attr->retry_cnt = cm_id_priv->retry_count; + qp_attr->rnr_retry = cm_id_priv->rnr_retry_count; + qp_attr->max_rd_atomic = cm_id_priv->initiator_depth; + fallthrough; + case IB_QPT_XRC_TGT: + *qp_attr_mask |= IB_QP_TIMEOUT; + qp_attr->timeout = cm_id_priv->av.timeout; + break; + default: + break; + } + if (rdma_ah_get_dlid(&cm_id_priv->alt_av.ah_attr)) { + *qp_attr_mask |= IB_QP_PATH_MIG_STATE; + qp_attr->path_mig_state = IB_MIG_REARM; + } + } else { + *qp_attr_mask = IB_QP_ALT_PATH | IB_QP_PATH_MIG_STATE; + qp_attr->alt_port_num = cm_id_priv->alt_av.port->port_num; + qp_attr->alt_pkey_index = cm_id_priv->alt_av.pkey_index; + qp_attr->alt_timeout = cm_id_priv->alt_av.timeout; + qp_attr->alt_ah_attr = cm_id_priv->alt_av.ah_attr; + qp_attr->path_mig_state = IB_MIG_REARM; + } + ret = 0; + break; + default: + trace_icm_qp_rts_err(&cm_id_priv->id); + ret = -EINVAL; + break; + } + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + return ret; +} + +int ib_cm_init_qp_attr(struct ib_cm_id *cm_id, + struct ib_qp_attr *qp_attr, + int *qp_attr_mask) +{ + struct cm_id_private *cm_id_priv; + int ret; + + cm_id_priv = container_of(cm_id, struct cm_id_private, id); + switch (qp_attr->qp_state) { + case IB_QPS_INIT: + ret = cm_init_qp_init_attr(cm_id_priv, qp_attr, qp_attr_mask); + break; + case IB_QPS_RTR: + ret = cm_init_qp_rtr_attr(cm_id_priv, qp_attr, qp_attr_mask); + break; + case IB_QPS_RTS: + ret = cm_init_qp_rts_attr(cm_id_priv, qp_attr, qp_attr_mask); + break; + default: + ret = -EINVAL; + break; + } + return ret; +} +EXPORT_SYMBOL(ib_cm_init_qp_attr); + +static ssize_t cm_show_counter(struct kobject *obj, struct attribute *attr, + char *buf) +{ + struct cm_counter_group *group; + struct cm_counter_attribute *cm_attr; + + group = container_of(obj, struct cm_counter_group, obj); + cm_attr = container_of(attr, struct cm_counter_attribute, attr); + + return sprintf(buf, "%ld\n", + atomic_long_read(&group->counter[cm_attr->index])); +} + +static const struct sysfs_ops cm_counter_ops = { + .show = cm_show_counter +}; + +static struct kobj_type cm_counter_obj_type = { + .sysfs_ops = &cm_counter_ops, + .default_attrs = cm_counter_default_attrs +}; + +static int cm_create_port_fs(struct cm_port *port) +{ + int i, ret; + + for (i = 0; i < CM_COUNTER_GROUPS; i++) { + ret = ib_port_register_module_stat(port->cm_dev->ib_device, + port->port_num, + &port->counter_group[i].obj, + &cm_counter_obj_type, + counter_group_names[i]); + if (ret) + goto error; + } + + return 0; + +error: + while (i--) + ib_port_unregister_module_stat(&port->counter_group[i].obj); + return ret; + +} + +static void cm_remove_port_fs(struct cm_port *port) +{ + int i; + + for (i = 0; i < CM_COUNTER_GROUPS; i++) + ib_port_unregister_module_stat(&port->counter_group[i].obj); + +} + +static int cm_add_one(struct ib_device *ib_device) +{ + struct cm_device *cm_dev; + struct cm_port *port; + struct ib_mad_reg_req reg_req = { + .mgmt_class = IB_MGMT_CLASS_CM, + .mgmt_class_version = IB_CM_CLASS_VERSION, + }; + struct ib_port_modify port_modify = { + .set_port_cap_mask = IB_PORT_CM_SUP + }; + unsigned long flags; + int ret; + int count = 0; + unsigned int i; + + cm_dev = kzalloc(struct_size(cm_dev, port, ib_device->phys_port_cnt), + GFP_KERNEL); + if (!cm_dev) + return -ENOMEM; + + cm_dev->ib_device = ib_device; + cm_dev->ack_delay = ib_device->attrs.local_ca_ack_delay; + cm_dev->going_down = 0; + + set_bit(IB_MGMT_METHOD_SEND, reg_req.method_mask); + rdma_for_each_port (ib_device, i) { + if (!rdma_cap_ib_cm(ib_device, i)) + continue; + + port = kzalloc(sizeof *port, GFP_KERNEL); + if (!port) { + ret = -ENOMEM; + goto error1; + } + + cm_dev->port[i-1] = port; + port->cm_dev = cm_dev; + port->port_num = i; + + INIT_LIST_HEAD(&port->cm_priv_prim_list); + INIT_LIST_HEAD(&port->cm_priv_altr_list); + + ret = cm_create_port_fs(port); + if (ret) + goto error1; + + port->mad_agent = ib_register_mad_agent(ib_device, i, + IB_QPT_GSI, + ®_req, + 0, + cm_send_handler, + cm_recv_handler, + port, + 0); + if (IS_ERR(port->mad_agent)) { + ret = PTR_ERR(port->mad_agent); + goto error2; + } + + ret = ib_modify_port(ib_device, i, 0, &port_modify); + if (ret) + goto error3; + + count++; + } + + if (!count) { + ret = -EOPNOTSUPP; + goto free; + } + + ib_set_client_data(ib_device, &cm_client, cm_dev); + + write_lock_irqsave(&cm.device_lock, flags); + list_add_tail(&cm_dev->list, &cm.device_list); + write_unlock_irqrestore(&cm.device_lock, flags); + return 0; + +error3: + ib_unregister_mad_agent(port->mad_agent); +error2: + cm_remove_port_fs(port); +error1: + port_modify.set_port_cap_mask = 0; + port_modify.clr_port_cap_mask = IB_PORT_CM_SUP; + kfree(port); + while (--i) { + if (!rdma_cap_ib_cm(ib_device, i)) + continue; + + port = cm_dev->port[i-1]; + ib_modify_port(ib_device, port->port_num, 0, &port_modify); + ib_unregister_mad_agent(port->mad_agent); + cm_remove_port_fs(port); + kfree(port); + } +free: + kfree(cm_dev); + return ret; +} + +static void cm_remove_one(struct ib_device *ib_device, void *client_data) +{ + struct cm_device *cm_dev = client_data; + struct cm_port *port; + struct cm_id_private *cm_id_priv; + struct ib_mad_agent *cur_mad_agent; + struct ib_port_modify port_modify = { + .clr_port_cap_mask = IB_PORT_CM_SUP + }; + unsigned long flags; + unsigned int i; + + write_lock_irqsave(&cm.device_lock, flags); + list_del(&cm_dev->list); + write_unlock_irqrestore(&cm.device_lock, flags); + + spin_lock_irq(&cm.lock); + cm_dev->going_down = 1; + spin_unlock_irq(&cm.lock); + + rdma_for_each_port (ib_device, i) { + if (!rdma_cap_ib_cm(ib_device, i)) + continue; + + port = cm_dev->port[i-1]; + ib_modify_port(ib_device, port->port_num, 0, &port_modify); + /* Mark all the cm_id's as not valid */ + spin_lock_irq(&cm.lock); + list_for_each_entry(cm_id_priv, &port->cm_priv_altr_list, altr_list) + cm_id_priv->altr_send_port_not_ready = 1; + list_for_each_entry(cm_id_priv, &port->cm_priv_prim_list, prim_list) + cm_id_priv->prim_send_port_not_ready = 1; + spin_unlock_irq(&cm.lock); + /* + * We flush the queue here after the going_down set, this + * verify that no new works will be queued in the recv handler, + * after that we can call the unregister_mad_agent + */ + flush_workqueue(cm.wq); + spin_lock_irq(&cm.state_lock); + cur_mad_agent = port->mad_agent; + port->mad_agent = NULL; + spin_unlock_irq(&cm.state_lock); + ib_unregister_mad_agent(cur_mad_agent); + cm_remove_port_fs(port); + kfree(port); + } + + kfree(cm_dev); +} + +static int __init ib_cm_init(void) +{ + int ret; + + INIT_LIST_HEAD(&cm.device_list); + rwlock_init(&cm.device_lock); + spin_lock_init(&cm.lock); + spin_lock_init(&cm.state_lock); + cm.listen_service_table = RB_ROOT; + cm.listen_service_id = be64_to_cpu(IB_CM_ASSIGN_SERVICE_ID); + cm.remote_id_table = RB_ROOT; + cm.remote_qp_table = RB_ROOT; + cm.remote_sidr_table = RB_ROOT; + xa_init_flags(&cm.local_id_table, XA_FLAGS_ALLOC); + get_random_bytes(&cm.random_id_operand, sizeof cm.random_id_operand); + INIT_LIST_HEAD(&cm.timewait_list); + + cm.wq = alloc_workqueue("ib_cm", 0, 1); + if (!cm.wq) { + ret = -ENOMEM; + goto error2; + } + + ret = ib_register_client(&cm_client); + if (ret) + goto error3; + + return 0; +error3: + destroy_workqueue(cm.wq); +error2: + return ret; +} + +static void __exit ib_cm_cleanup(void) +{ + struct cm_timewait_info *timewait_info, *tmp; + + spin_lock_irq(&cm.lock); + list_for_each_entry(timewait_info, &cm.timewait_list, list) + cancel_delayed_work(&timewait_info->work.work); + spin_unlock_irq(&cm.lock); + + ib_unregister_client(&cm_client); + destroy_workqueue(cm.wq); + + list_for_each_entry_safe(timewait_info, tmp, &cm.timewait_list, list) { + list_del(&timewait_info->list); + kfree(timewait_info); + } + + WARN_ON(!xa_empty(&cm.local_id_table)); +} + +module_init(ib_cm_init); +module_exit(ib_cm_cleanup); diff --git a/drivers/infiniband/core/cm_msgs.h b/drivers/infiniband/core/cm_msgs.h new file mode 100644 index 000000000..0cc40656b --- /dev/null +++ b/drivers/infiniband/core/cm_msgs.h @@ -0,0 +1,68 @@ +/* SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB */ +/* + * Copyright (c) 2004, 2011 Intel Corporation. All rights reserved. + * Copyright (c) 2004 Topspin Corporation. All rights reserved. + * Copyright (c) 2004 Voltaire Corporation. All rights reserved. + * Copyright (c) 2019, Mellanox Technologies inc. All rights reserved. + */ +#ifndef CM_MSGS_H +#define CM_MSGS_H + +#include <rdma/ibta_vol1_c12.h> +#include <rdma/ib_mad.h> +#include <rdma/ib_cm.h> + +/* + * Parameters to routines below should be in network-byte order, and values + * are returned in network-byte order. + */ + +#define IB_CM_CLASS_VERSION 2 /* IB specification 1.2 */ + +static inline enum ib_qp_type cm_req_get_qp_type(struct cm_req_msg *req_msg) +{ + u8 transport_type = IBA_GET(CM_REQ_TRANSPORT_SERVICE_TYPE, req_msg); + switch(transport_type) { + case 0: return IB_QPT_RC; + case 1: return IB_QPT_UC; + case 3: + switch (IBA_GET(CM_REQ_EXTENDED_TRANSPORT_TYPE, req_msg)) { + case 1: return IB_QPT_XRC_TGT; + default: return 0; + } + default: return 0; + } +} + +static inline void cm_req_set_qp_type(struct cm_req_msg *req_msg, + enum ib_qp_type qp_type) +{ + switch(qp_type) { + case IB_QPT_UC: + IBA_SET(CM_REQ_TRANSPORT_SERVICE_TYPE, req_msg, 1); + break; + case IB_QPT_XRC_INI: + IBA_SET(CM_REQ_TRANSPORT_SERVICE_TYPE, req_msg, 3); + IBA_SET(CM_REQ_EXTENDED_TRANSPORT_TYPE, req_msg, 1); + break; + default: + IBA_SET(CM_REQ_TRANSPORT_SERVICE_TYPE, req_msg, 0); + } +} + +/* Message REJected or MRAed */ +enum cm_msg_response { + CM_MSG_RESPONSE_REQ = 0x0, + CM_MSG_RESPONSE_REP = 0x1, + CM_MSG_RESPONSE_OTHER = 0x2 +}; + +static inline __be32 cm_rep_get_qpn(struct cm_rep_msg *rep_msg, enum ib_qp_type qp_type) +{ + return (qp_type == IB_QPT_XRC_INI) ? + cpu_to_be32(IBA_GET(CM_REP_LOCAL_EE_CONTEXT_NUMBER, + rep_msg)) : + cpu_to_be32(IBA_GET(CM_REP_LOCAL_QPN, rep_msg)); +} + +#endif /* CM_MSGS_H */ diff --git a/drivers/infiniband/core/cm_trace.c b/drivers/infiniband/core/cm_trace.c new file mode 100644 index 000000000..8f3482f66 --- /dev/null +++ b/drivers/infiniband/core/cm_trace.c @@ -0,0 +1,15 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * Trace points for the IB Connection Manager. + * + * Author: Chuck Lever <chuck.lever@oracle.com> + * + * Copyright (c) 2020, Oracle and/or its affiliates. + */ + +#include <rdma/rdma_cm.h> +#include "cma_priv.h" + +#define CREATE_TRACE_POINTS + +#include "cm_trace.h" diff --git a/drivers/infiniband/core/cm_trace.h b/drivers/infiniband/core/cm_trace.h new file mode 100644 index 000000000..e9d282679 --- /dev/null +++ b/drivers/infiniband/core/cm_trace.h @@ -0,0 +1,414 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +/* + * Trace point definitions for the RDMA Connect Manager. + * + * Author: Chuck Lever <chuck.lever@oracle.com> + * + * Copyright (c) 2020 Oracle and/or its affiliates. + */ + +#undef TRACE_SYSTEM +#define TRACE_SYSTEM ib_cma + +#if !defined(_TRACE_IB_CMA_H) || defined(TRACE_HEADER_MULTI_READ) + +#define _TRACE_IB_CMA_H + +#include <linux/tracepoint.h> +#include <rdma/ib_cm.h> +#include <trace/events/rdma.h> + +/* + * enum ib_cm_state, from include/rdma/ib_cm.h + */ +#define IB_CM_STATE_LIST \ + ib_cm_state(IDLE) \ + ib_cm_state(LISTEN) \ + ib_cm_state(REQ_SENT) \ + ib_cm_state(REQ_RCVD) \ + ib_cm_state(MRA_REQ_SENT) \ + ib_cm_state(MRA_REQ_RCVD) \ + ib_cm_state(REP_SENT) \ + ib_cm_state(REP_RCVD) \ + ib_cm_state(MRA_REP_SENT) \ + ib_cm_state(MRA_REP_RCVD) \ + ib_cm_state(ESTABLISHED) \ + ib_cm_state(DREQ_SENT) \ + ib_cm_state(DREQ_RCVD) \ + ib_cm_state(TIMEWAIT) \ + ib_cm_state(SIDR_REQ_SENT) \ + ib_cm_state_end(SIDR_REQ_RCVD) + +#undef ib_cm_state +#undef ib_cm_state_end +#define ib_cm_state(x) TRACE_DEFINE_ENUM(IB_CM_##x); +#define ib_cm_state_end(x) TRACE_DEFINE_ENUM(IB_CM_##x); + +IB_CM_STATE_LIST + +#undef ib_cm_state +#undef ib_cm_state_end +#define ib_cm_state(x) { IB_CM_##x, #x }, +#define ib_cm_state_end(x) { IB_CM_##x, #x } + +#define show_ib_cm_state(x) \ + __print_symbolic(x, IB_CM_STATE_LIST) + +/* + * enum ib_cm_lap_state, from include/rdma/ib_cm.h + */ +#define IB_CM_LAP_STATE_LIST \ + ib_cm_lap_state(LAP_UNINIT) \ + ib_cm_lap_state(LAP_IDLE) \ + ib_cm_lap_state(LAP_SENT) \ + ib_cm_lap_state(LAP_RCVD) \ + ib_cm_lap_state(MRA_LAP_SENT) \ + ib_cm_lap_state_end(MRA_LAP_RCVD) + +#undef ib_cm_lap_state +#undef ib_cm_lap_state_end +#define ib_cm_lap_state(x) TRACE_DEFINE_ENUM(IB_CM_##x); +#define ib_cm_lap_state_end(x) TRACE_DEFINE_ENUM(IB_CM_##x); + +IB_CM_LAP_STATE_LIST + +#undef ib_cm_lap_state +#undef ib_cm_lap_state_end +#define ib_cm_lap_state(x) { IB_CM_##x, #x }, +#define ib_cm_lap_state_end(x) { IB_CM_##x, #x } + +#define show_ib_cm_lap_state(x) \ + __print_symbolic(x, IB_CM_LAP_STATE_LIST) + +/* + * enum ib_cm_rej_reason, from include/rdma/ib_cm.h + */ +#define IB_CM_REJ_REASON_LIST \ + ib_cm_rej_reason(REJ_NO_QP) \ + ib_cm_rej_reason(REJ_NO_EEC) \ + ib_cm_rej_reason(REJ_NO_RESOURCES) \ + ib_cm_rej_reason(REJ_TIMEOUT) \ + ib_cm_rej_reason(REJ_UNSUPPORTED) \ + ib_cm_rej_reason(REJ_INVALID_COMM_ID) \ + ib_cm_rej_reason(REJ_INVALID_COMM_INSTANCE) \ + ib_cm_rej_reason(REJ_INVALID_SERVICE_ID) \ + ib_cm_rej_reason(REJ_INVALID_TRANSPORT_TYPE) \ + ib_cm_rej_reason(REJ_STALE_CONN) \ + ib_cm_rej_reason(REJ_RDC_NOT_EXIST) \ + ib_cm_rej_reason(REJ_INVALID_GID) \ + ib_cm_rej_reason(REJ_INVALID_LID) \ + ib_cm_rej_reason(REJ_INVALID_SL) \ + ib_cm_rej_reason(REJ_INVALID_TRAFFIC_CLASS) \ + ib_cm_rej_reason(REJ_INVALID_HOP_LIMIT) \ + ib_cm_rej_reason(REJ_INVALID_PACKET_RATE) \ + ib_cm_rej_reason(REJ_INVALID_ALT_GID) \ + ib_cm_rej_reason(REJ_INVALID_ALT_LID) \ + ib_cm_rej_reason(REJ_INVALID_ALT_SL) \ + ib_cm_rej_reason(REJ_INVALID_ALT_TRAFFIC_CLASS) \ + ib_cm_rej_reason(REJ_INVALID_ALT_HOP_LIMIT) \ + ib_cm_rej_reason(REJ_INVALID_ALT_PACKET_RATE) \ + ib_cm_rej_reason(REJ_PORT_CM_REDIRECT) \ + ib_cm_rej_reason(REJ_PORT_REDIRECT) \ + ib_cm_rej_reason(REJ_INVALID_MTU) \ + ib_cm_rej_reason(REJ_INSUFFICIENT_RESP_RESOURCES) \ + ib_cm_rej_reason(REJ_CONSUMER_DEFINED) \ + ib_cm_rej_reason(REJ_INVALID_RNR_RETRY) \ + ib_cm_rej_reason(REJ_DUPLICATE_LOCAL_COMM_ID) \ + ib_cm_rej_reason(REJ_INVALID_CLASS_VERSION) \ + ib_cm_rej_reason(REJ_INVALID_FLOW_LABEL) \ + ib_cm_rej_reason(REJ_INVALID_ALT_FLOW_LABEL) \ + ib_cm_rej_reason_end(REJ_VENDOR_OPTION_NOT_SUPPORTED) + +#undef ib_cm_rej_reason +#undef ib_cm_rej_reason_end +#define ib_cm_rej_reason(x) TRACE_DEFINE_ENUM(IB_CM_##x); +#define ib_cm_rej_reason_end(x) TRACE_DEFINE_ENUM(IB_CM_##x); + +IB_CM_REJ_REASON_LIST + +#undef ib_cm_rej_reason +#undef ib_cm_rej_reason_end +#define ib_cm_rej_reason(x) { IB_CM_##x, #x }, +#define ib_cm_rej_reason_end(x) { IB_CM_##x, #x } + +#define show_ib_cm_rej_reason(x) \ + __print_symbolic(x, IB_CM_REJ_REASON_LIST) + +DECLARE_EVENT_CLASS(icm_id_class, + TP_PROTO( + const struct ib_cm_id *cm_id + ), + + TP_ARGS(cm_id), + + TP_STRUCT__entry( + __field(const void *, cm_id) /* for eBPF scripts */ + __field(unsigned int, local_id) + __field(unsigned int, remote_id) + __field(unsigned long, state) + __field(unsigned long, lap_state) + ), + + TP_fast_assign( + __entry->cm_id = cm_id; + __entry->local_id = be32_to_cpu(cm_id->local_id); + __entry->remote_id = be32_to_cpu(cm_id->remote_id); + __entry->state = cm_id->state; + __entry->lap_state = cm_id->lap_state; + ), + + TP_printk("local_id=%u remote_id=%u state=%s lap_state=%s", + __entry->local_id, __entry->remote_id, + show_ib_cm_state(__entry->state), + show_ib_cm_lap_state(__entry->lap_state) + ) +); + +#define DEFINE_CM_SEND_EVENT(name) \ + DEFINE_EVENT(icm_id_class, \ + icm_send_##name, \ + TP_PROTO( \ + const struct ib_cm_id *cm_id \ + ), \ + TP_ARGS(cm_id)) + +DEFINE_CM_SEND_EVENT(req); +DEFINE_CM_SEND_EVENT(rep); +DEFINE_CM_SEND_EVENT(dup_req); +DEFINE_CM_SEND_EVENT(dup_rep); +DEFINE_CM_SEND_EVENT(rtu); +DEFINE_CM_SEND_EVENT(mra); +DEFINE_CM_SEND_EVENT(sidr_req); +DEFINE_CM_SEND_EVENT(sidr_rep); +DEFINE_CM_SEND_EVENT(dreq); +DEFINE_CM_SEND_EVENT(drep); + +TRACE_EVENT(icm_send_rej, + TP_PROTO( + const struct ib_cm_id *cm_id, + enum ib_cm_rej_reason reason + ), + + TP_ARGS(cm_id, reason), + + TP_STRUCT__entry( + __field(const void *, cm_id) + __field(u32, local_id) + __field(u32, remote_id) + __field(unsigned long, state) + __field(unsigned long, reason) + ), + + TP_fast_assign( + __entry->cm_id = cm_id; + __entry->local_id = be32_to_cpu(cm_id->local_id); + __entry->remote_id = be32_to_cpu(cm_id->remote_id); + __entry->state = cm_id->state; + __entry->reason = reason; + ), + + TP_printk("local_id=%u remote_id=%u state=%s reason=%s", + __entry->local_id, __entry->remote_id, + show_ib_cm_state(__entry->state), + show_ib_cm_rej_reason(__entry->reason) + ) +); + +#define DEFINE_CM_ERR_EVENT(name) \ + DEFINE_EVENT(icm_id_class, \ + icm_##name##_err, \ + TP_PROTO( \ + const struct ib_cm_id *cm_id \ + ), \ + TP_ARGS(cm_id)) + +DEFINE_CM_ERR_EVENT(send_cm_rtu); +DEFINE_CM_ERR_EVENT(establish); +DEFINE_CM_ERR_EVENT(no_listener); +DEFINE_CM_ERR_EVENT(send_drep); +DEFINE_CM_ERR_EVENT(dreq_unknown); +DEFINE_CM_ERR_EVENT(send_unknown_rej); +DEFINE_CM_ERR_EVENT(rej_unknown); +DEFINE_CM_ERR_EVENT(send_mra_unknown); +DEFINE_CM_ERR_EVENT(mra_unknown); +DEFINE_CM_ERR_EVENT(qp_init); +DEFINE_CM_ERR_EVENT(qp_rtr); +DEFINE_CM_ERR_EVENT(qp_rts); + +DEFINE_EVENT(icm_id_class, \ + icm_dreq_skipped, \ + TP_PROTO( \ + const struct ib_cm_id *cm_id \ + ), \ + TP_ARGS(cm_id) \ +); + +DECLARE_EVENT_CLASS(icm_local_class, + TP_PROTO( + unsigned int local_id, + unsigned int remote_id + ), + + TP_ARGS(local_id, remote_id), + + TP_STRUCT__entry( + __field(unsigned int, local_id) + __field(unsigned int, remote_id) + ), + + TP_fast_assign( + __entry->local_id = local_id; + __entry->remote_id = remote_id; + ), + + TP_printk("local_id=%u remote_id=%u", + __entry->local_id, __entry->remote_id + ) +); + +#define DEFINE_CM_LOCAL_EVENT(name) \ + DEFINE_EVENT(icm_local_class, \ + icm_##name, \ + TP_PROTO( \ + unsigned int local_id, \ + unsigned int remote_id \ + ), \ + TP_ARGS(local_id, remote_id)) + +DEFINE_CM_LOCAL_EVENT(issue_rej); +DEFINE_CM_LOCAL_EVENT(issue_drep); +DEFINE_CM_LOCAL_EVENT(staleconn_err); +DEFINE_CM_LOCAL_EVENT(no_priv_err); + +DECLARE_EVENT_CLASS(icm_remote_class, + TP_PROTO( + u32 remote_id + ), + + TP_ARGS(remote_id), + + TP_STRUCT__entry( + __field(u32, remote_id) + ), + + TP_fast_assign( + __entry->remote_id = remote_id; + ), + + TP_printk("remote_id=%u", + __entry->remote_id + ) +); + +#define DEFINE_CM_REMOTE_EVENT(name) \ + DEFINE_EVENT(icm_remote_class, \ + icm_##name, \ + TP_PROTO( \ + u32 remote_id \ + ), \ + TP_ARGS(remote_id)) + +DEFINE_CM_REMOTE_EVENT(remote_no_priv_err); +DEFINE_CM_REMOTE_EVENT(insert_failed_err); + +TRACE_EVENT(icm_send_rep_err, + TP_PROTO( + __be32 local_id, + enum ib_cm_state state + ), + + TP_ARGS(local_id, state), + + TP_STRUCT__entry( + __field(unsigned int, local_id) + __field(unsigned long, state) + ), + + TP_fast_assign( + __entry->local_id = be32_to_cpu(local_id); + __entry->state = state; + ), + + TP_printk("local_id=%u state=%s", + __entry->local_id, show_ib_cm_state(__entry->state) + ) +); + +TRACE_EVENT(icm_rep_unknown_err, + TP_PROTO( + unsigned int local_id, + unsigned int remote_id, + enum ib_cm_state state + ), + + TP_ARGS(local_id, remote_id, state), + + TP_STRUCT__entry( + __field(unsigned int, local_id) + __field(unsigned int, remote_id) + __field(unsigned long, state) + ), + + TP_fast_assign( + __entry->local_id = local_id; + __entry->remote_id = remote_id; + __entry->state = state; + ), + + TP_printk("local_id=%u remote_id=%u state=%s", + __entry->local_id, __entry->remote_id, + show_ib_cm_state(__entry->state) + ) +); + +TRACE_EVENT(icm_handler_err, + TP_PROTO( + enum ib_cm_event_type event + ), + + TP_ARGS(event), + + TP_STRUCT__entry( + __field(unsigned long, event) + ), + + TP_fast_assign( + __entry->event = event; + ), + + TP_printk("unhandled event=%s", + rdma_show_ib_cm_event(__entry->event) + ) +); + +TRACE_EVENT(icm_mad_send_err, + TP_PROTO( + enum ib_cm_state state, + enum ib_wc_status wc_status + ), + + TP_ARGS(state, wc_status), + + TP_STRUCT__entry( + __field(unsigned long, state) + __field(unsigned long, wc_status) + ), + + TP_fast_assign( + __entry->state = state; + __entry->wc_status = wc_status; + ), + + TP_printk("state=%s completion status=%s", + show_ib_cm_state(__entry->state), + rdma_show_wc_status(__entry->wc_status) + ) +); + +#endif /* _TRACE_IB_CMA_H */ + +#undef TRACE_INCLUDE_PATH +#define TRACE_INCLUDE_PATH ../../drivers/infiniband/core +#define TRACE_INCLUDE_FILE cm_trace + +#include <trace/define_trace.h> diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c new file mode 100644 index 000000000..0603d069d --- /dev/null +++ b/drivers/infiniband/core/cma.c @@ -0,0 +1,5148 @@ +// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB +/* + * Copyright (c) 2005 Voltaire Inc. All rights reserved. + * Copyright (c) 2002-2005, Network Appliance, Inc. All rights reserved. + * Copyright (c) 1999-2019, Mellanox Technologies, Inc. All rights reserved. + * Copyright (c) 2005-2006 Intel Corporation. All rights reserved. + */ + +#include <linux/completion.h> +#include <linux/in.h> +#include <linux/in6.h> +#include <linux/mutex.h> +#include <linux/random.h> +#include <linux/igmp.h> +#include <linux/xarray.h> +#include <linux/inetdevice.h> +#include <linux/slab.h> +#include <linux/module.h> +#include <net/route.h> + +#include <net/net_namespace.h> +#include <net/netns/generic.h> +#include <net/tcp.h> +#include <net/ipv6.h> +#include <net/ip_fib.h> +#include <net/ip6_route.h> + +#include <rdma/rdma_cm.h> +#include <rdma/rdma_cm_ib.h> +#include <rdma/rdma_netlink.h> +#include <rdma/ib.h> +#include <rdma/ib_cache.h> +#include <rdma/ib_cm.h> +#include <rdma/ib_sa.h> +#include <rdma/iw_cm.h> + +#include "core_priv.h" +#include "cma_priv.h" +#include "cma_trace.h" + +MODULE_AUTHOR("Sean Hefty"); +MODULE_DESCRIPTION("Generic RDMA CM Agent"); +MODULE_LICENSE("Dual BSD/GPL"); + +#define CMA_CM_RESPONSE_TIMEOUT 20 +#define CMA_QUERY_CLASSPORT_INFO_TIMEOUT 3000 +#define CMA_MAX_CM_RETRIES 15 +#define CMA_CM_MRA_SETTING (IB_CM_MRA_FLAG_DELAY | 24) +#define CMA_IBOE_PACKET_LIFETIME 18 +#define CMA_PREFERRED_ROCE_GID_TYPE IB_GID_TYPE_ROCE_UDP_ENCAP + +static const char * const cma_events[] = { + [RDMA_CM_EVENT_ADDR_RESOLVED] = "address resolved", + [RDMA_CM_EVENT_ADDR_ERROR] = "address error", + [RDMA_CM_EVENT_ROUTE_RESOLVED] = "route resolved ", + [RDMA_CM_EVENT_ROUTE_ERROR] = "route error", + [RDMA_CM_EVENT_CONNECT_REQUEST] = "connect request", + [RDMA_CM_EVENT_CONNECT_RESPONSE] = "connect response", + [RDMA_CM_EVENT_CONNECT_ERROR] = "connect error", + [RDMA_CM_EVENT_UNREACHABLE] = "unreachable", + [RDMA_CM_EVENT_REJECTED] = "rejected", + [RDMA_CM_EVENT_ESTABLISHED] = "established", + [RDMA_CM_EVENT_DISCONNECTED] = "disconnected", + [RDMA_CM_EVENT_DEVICE_REMOVAL] = "device removal", + [RDMA_CM_EVENT_MULTICAST_JOIN] = "multicast join", + [RDMA_CM_EVENT_MULTICAST_ERROR] = "multicast error", + [RDMA_CM_EVENT_ADDR_CHANGE] = "address change", + [RDMA_CM_EVENT_TIMEWAIT_EXIT] = "timewait exit", +}; + +static void cma_iboe_set_mgid(struct sockaddr *addr, union ib_gid *mgid, + enum ib_gid_type gid_type); + +const char *__attribute_const__ rdma_event_msg(enum rdma_cm_event_type event) +{ + size_t index = event; + + return (index < ARRAY_SIZE(cma_events) && cma_events[index]) ? + cma_events[index] : "unrecognized event"; +} +EXPORT_SYMBOL(rdma_event_msg); + +const char *__attribute_const__ rdma_reject_msg(struct rdma_cm_id *id, + int reason) +{ + if (rdma_ib_or_roce(id->device, id->port_num)) + return ibcm_reject_msg(reason); + + if (rdma_protocol_iwarp(id->device, id->port_num)) + return iwcm_reject_msg(reason); + + WARN_ON_ONCE(1); + return "unrecognized transport"; +} +EXPORT_SYMBOL(rdma_reject_msg); + +/** + * rdma_is_consumer_reject - return true if the consumer rejected the connect + * request. + * @id: Communication identifier that received the REJECT event. + * @reason: Value returned in the REJECT event status field. + */ +static bool rdma_is_consumer_reject(struct rdma_cm_id *id, int reason) +{ + if (rdma_ib_or_roce(id->device, id->port_num)) + return reason == IB_CM_REJ_CONSUMER_DEFINED; + + if (rdma_protocol_iwarp(id->device, id->port_num)) + return reason == -ECONNREFUSED; + + WARN_ON_ONCE(1); + return false; +} + +const void *rdma_consumer_reject_data(struct rdma_cm_id *id, + struct rdma_cm_event *ev, u8 *data_len) +{ + const void *p; + + if (rdma_is_consumer_reject(id, ev->status)) { + *data_len = ev->param.conn.private_data_len; + p = ev->param.conn.private_data; + } else { + *data_len = 0; + p = NULL; + } + return p; +} +EXPORT_SYMBOL(rdma_consumer_reject_data); + +/** + * rdma_iw_cm_id() - return the iw_cm_id pointer for this cm_id. + * @id: Communication Identifier + */ +struct iw_cm_id *rdma_iw_cm_id(struct rdma_cm_id *id) +{ + struct rdma_id_private *id_priv; + + id_priv = container_of(id, struct rdma_id_private, id); + if (id->device->node_type == RDMA_NODE_RNIC) + return id_priv->cm_id.iw; + return NULL; +} +EXPORT_SYMBOL(rdma_iw_cm_id); + +/** + * rdma_res_to_id() - return the rdma_cm_id pointer for this restrack. + * @res: rdma resource tracking entry pointer + */ +struct rdma_cm_id *rdma_res_to_id(struct rdma_restrack_entry *res) +{ + struct rdma_id_private *id_priv = + container_of(res, struct rdma_id_private, res); + + return &id_priv->id; +} +EXPORT_SYMBOL(rdma_res_to_id); + +static int cma_add_one(struct ib_device *device); +static void cma_remove_one(struct ib_device *device, void *client_data); + +static struct ib_client cma_client = { + .name = "cma", + .add = cma_add_one, + .remove = cma_remove_one +}; + +static struct ib_sa_client sa_client; +static LIST_HEAD(dev_list); +static LIST_HEAD(listen_any_list); +static DEFINE_MUTEX(lock); +static struct workqueue_struct *cma_wq; +static unsigned int cma_pernet_id; + +struct cma_pernet { + struct xarray tcp_ps; + struct xarray udp_ps; + struct xarray ipoib_ps; + struct xarray ib_ps; +}; + +static struct cma_pernet *cma_pernet(struct net *net) +{ + return net_generic(net, cma_pernet_id); +} + +static +struct xarray *cma_pernet_xa(struct net *net, enum rdma_ucm_port_space ps) +{ + struct cma_pernet *pernet = cma_pernet(net); + + switch (ps) { + case RDMA_PS_TCP: + return &pernet->tcp_ps; + case RDMA_PS_UDP: + return &pernet->udp_ps; + case RDMA_PS_IPOIB: + return &pernet->ipoib_ps; + case RDMA_PS_IB: + return &pernet->ib_ps; + default: + return NULL; + } +} + +struct cma_device { + struct list_head list; + struct ib_device *device; + struct completion comp; + refcount_t refcount; + struct list_head id_list; + enum ib_gid_type *default_gid_type; + u8 *default_roce_tos; +}; + +struct rdma_bind_list { + enum rdma_ucm_port_space ps; + struct hlist_head owners; + unsigned short port; +}; + +struct class_port_info_context { + struct ib_class_port_info *class_port_info; + struct ib_device *device; + struct completion done; + struct ib_sa_query *sa_query; + u8 port_num; +}; + +static int cma_ps_alloc(struct net *net, enum rdma_ucm_port_space ps, + struct rdma_bind_list *bind_list, int snum) +{ + struct xarray *xa = cma_pernet_xa(net, ps); + + return xa_insert(xa, snum, bind_list, GFP_KERNEL); +} + +static struct rdma_bind_list *cma_ps_find(struct net *net, + enum rdma_ucm_port_space ps, int snum) +{ + struct xarray *xa = cma_pernet_xa(net, ps); + + return xa_load(xa, snum); +} + +static void cma_ps_remove(struct net *net, enum rdma_ucm_port_space ps, + int snum) +{ + struct xarray *xa = cma_pernet_xa(net, ps); + + xa_erase(xa, snum); +} + +enum { + CMA_OPTION_AFONLY, +}; + +void cma_dev_get(struct cma_device *cma_dev) +{ + refcount_inc(&cma_dev->refcount); +} + +void cma_dev_put(struct cma_device *cma_dev) +{ + if (refcount_dec_and_test(&cma_dev->refcount)) + complete(&cma_dev->comp); +} + +struct cma_device *cma_enum_devices_by_ibdev(cma_device_filter filter, + void *cookie) +{ + struct cma_device *cma_dev; + struct cma_device *found_cma_dev = NULL; + + mutex_lock(&lock); + + list_for_each_entry(cma_dev, &dev_list, list) + if (filter(cma_dev->device, cookie)) { + found_cma_dev = cma_dev; + break; + } + + if (found_cma_dev) + cma_dev_get(found_cma_dev); + mutex_unlock(&lock); + return found_cma_dev; +} + +int cma_get_default_gid_type(struct cma_device *cma_dev, + unsigned int port) +{ + if (!rdma_is_port_valid(cma_dev->device, port)) + return -EINVAL; + + return cma_dev->default_gid_type[port - rdma_start_port(cma_dev->device)]; +} + +int cma_set_default_gid_type(struct cma_device *cma_dev, + unsigned int port, + enum ib_gid_type default_gid_type) +{ + unsigned long supported_gids; + + if (!rdma_is_port_valid(cma_dev->device, port)) + return -EINVAL; + + if (default_gid_type == IB_GID_TYPE_IB && + rdma_protocol_roce_eth_encap(cma_dev->device, port)) + default_gid_type = IB_GID_TYPE_ROCE; + + supported_gids = roce_gid_type_mask_support(cma_dev->device, port); + + if (!(supported_gids & 1 << default_gid_type)) + return -EINVAL; + + cma_dev->default_gid_type[port - rdma_start_port(cma_dev->device)] = + default_gid_type; + + return 0; +} + +int cma_get_default_roce_tos(struct cma_device *cma_dev, unsigned int port) +{ + if (!rdma_is_port_valid(cma_dev->device, port)) + return -EINVAL; + + return cma_dev->default_roce_tos[port - rdma_start_port(cma_dev->device)]; +} + +int cma_set_default_roce_tos(struct cma_device *cma_dev, unsigned int port, + u8 default_roce_tos) +{ + if (!rdma_is_port_valid(cma_dev->device, port)) + return -EINVAL; + + cma_dev->default_roce_tos[port - rdma_start_port(cma_dev->device)] = + default_roce_tos; + + return 0; +} +struct ib_device *cma_get_ib_dev(struct cma_device *cma_dev) +{ + return cma_dev->device; +} + +/* + * Device removal can occur at anytime, so we need extra handling to + * serialize notifying the user of device removal with other callbacks. + * We do this by disabling removal notification while a callback is in process, + * and reporting it after the callback completes. + */ + +struct cma_multicast { + struct rdma_id_private *id_priv; + union { + struct ib_sa_multicast *sa_mc; + struct { + struct work_struct work; + struct rdma_cm_event event; + } iboe_join; + }; + struct list_head list; + void *context; + struct sockaddr_storage addr; + u8 join_state; +}; + +struct cma_work { + struct work_struct work; + struct rdma_id_private *id; + enum rdma_cm_state old_state; + enum rdma_cm_state new_state; + struct rdma_cm_event event; +}; + +union cma_ip_addr { + struct in6_addr ip6; + struct { + __be32 pad[3]; + __be32 addr; + } ip4; +}; + +struct cma_hdr { + u8 cma_version; + u8 ip_version; /* IP version: 7:4 */ + __be16 port; + union cma_ip_addr src_addr; + union cma_ip_addr dst_addr; +}; + +#define CMA_VERSION 0x00 + +struct cma_req_info { + struct sockaddr_storage listen_addr_storage; + struct sockaddr_storage src_addr_storage; + struct ib_device *device; + union ib_gid local_gid; + __be64 service_id; + int port; + bool has_gid; + u16 pkey; +}; + +static int cma_comp_exch(struct rdma_id_private *id_priv, + enum rdma_cm_state comp, enum rdma_cm_state exch) +{ + unsigned long flags; + int ret; + + /* + * The FSM uses a funny double locking where state is protected by both + * the handler_mutex and the spinlock. State is not allowed to change + * to/from a handler_mutex protected value without also holding + * handler_mutex. + */ + if (comp == RDMA_CM_CONNECT || exch == RDMA_CM_CONNECT) + lockdep_assert_held(&id_priv->handler_mutex); + + spin_lock_irqsave(&id_priv->lock, flags); + if ((ret = (id_priv->state == comp))) + id_priv->state = exch; + spin_unlock_irqrestore(&id_priv->lock, flags); + return ret; +} + +static inline u8 cma_get_ip_ver(const struct cma_hdr *hdr) +{ + return hdr->ip_version >> 4; +} + +static inline void cma_set_ip_ver(struct cma_hdr *hdr, u8 ip_ver) +{ + hdr->ip_version = (ip_ver << 4) | (hdr->ip_version & 0xF); +} + +static int cma_igmp_send(struct net_device *ndev, union ib_gid *mgid, bool join) +{ + struct in_device *in_dev = NULL; + + if (ndev) { + rtnl_lock(); + in_dev = __in_dev_get_rtnl(ndev); + if (in_dev) { + if (join) + ip_mc_inc_group(in_dev, + *(__be32 *)(mgid->raw + 12)); + else + ip_mc_dec_group(in_dev, + *(__be32 *)(mgid->raw + 12)); + } + rtnl_unlock(); + } + return (in_dev) ? 0 : -ENODEV; +} + +static void _cma_attach_to_dev(struct rdma_id_private *id_priv, + struct cma_device *cma_dev) +{ + cma_dev_get(cma_dev); + id_priv->cma_dev = cma_dev; + id_priv->id.device = cma_dev->device; + id_priv->id.route.addr.dev_addr.transport = + rdma_node_get_transport(cma_dev->device->node_type); + list_add_tail(&id_priv->list, &cma_dev->id_list); + + trace_cm_id_attach(id_priv, cma_dev->device); +} + +static void cma_attach_to_dev(struct rdma_id_private *id_priv, + struct cma_device *cma_dev) +{ + _cma_attach_to_dev(id_priv, cma_dev); + id_priv->gid_type = + cma_dev->default_gid_type[id_priv->id.port_num - + rdma_start_port(cma_dev->device)]; +} + +static void cma_release_dev(struct rdma_id_private *id_priv) +{ + mutex_lock(&lock); + list_del(&id_priv->list); + cma_dev_put(id_priv->cma_dev); + id_priv->cma_dev = NULL; + id_priv->id.device = NULL; + if (id_priv->id.route.addr.dev_addr.sgid_attr) { + rdma_put_gid_attr(id_priv->id.route.addr.dev_addr.sgid_attr); + id_priv->id.route.addr.dev_addr.sgid_attr = NULL; + } + mutex_unlock(&lock); +} + +static inline struct sockaddr *cma_src_addr(struct rdma_id_private *id_priv) +{ + return (struct sockaddr *) &id_priv->id.route.addr.src_addr; +} + +static inline struct sockaddr *cma_dst_addr(struct rdma_id_private *id_priv) +{ + return (struct sockaddr *) &id_priv->id.route.addr.dst_addr; +} + +static inline unsigned short cma_family(struct rdma_id_private *id_priv) +{ + return id_priv->id.route.addr.src_addr.ss_family; +} + +static int cma_set_default_qkey(struct rdma_id_private *id_priv) +{ + struct ib_sa_mcmember_rec rec; + int ret = 0; + + switch (id_priv->id.ps) { + case RDMA_PS_UDP: + case RDMA_PS_IB: + id_priv->qkey = RDMA_UDP_QKEY; + break; + case RDMA_PS_IPOIB: + ib_addr_get_mgid(&id_priv->id.route.addr.dev_addr, &rec.mgid); + ret = ib_sa_get_mcmember_rec(id_priv->id.device, + id_priv->id.port_num, &rec.mgid, + &rec); + if (!ret) + id_priv->qkey = be32_to_cpu(rec.qkey); + break; + default: + break; + } + return ret; +} + +static int cma_set_qkey(struct rdma_id_private *id_priv, u32 qkey) +{ + if (!qkey || + (id_priv->qkey && (id_priv->qkey != qkey))) + return -EINVAL; + + id_priv->qkey = qkey; + return 0; +} + +static void cma_translate_ib(struct sockaddr_ib *sib, struct rdma_dev_addr *dev_addr) +{ + dev_addr->dev_type = ARPHRD_INFINIBAND; + rdma_addr_set_sgid(dev_addr, (union ib_gid *) &sib->sib_addr); + ib_addr_set_pkey(dev_addr, ntohs(sib->sib_pkey)); +} + +static int cma_translate_addr(struct sockaddr *addr, struct rdma_dev_addr *dev_addr) +{ + int ret; + + if (addr->sa_family != AF_IB) { + ret = rdma_translate_ip(addr, dev_addr); + } else { + cma_translate_ib((struct sockaddr_ib *) addr, dev_addr); + ret = 0; + } + + return ret; +} + +static const struct ib_gid_attr * +cma_validate_port(struct ib_device *device, u8 port, + enum ib_gid_type gid_type, + union ib_gid *gid, + struct rdma_id_private *id_priv) +{ + struct rdma_dev_addr *dev_addr = &id_priv->id.route.addr.dev_addr; + int bound_if_index = dev_addr->bound_dev_if; + const struct ib_gid_attr *sgid_attr; + int dev_type = dev_addr->dev_type; + struct net_device *ndev = NULL; + + if (!rdma_dev_access_netns(device, id_priv->id.route.addr.dev_addr.net)) + return ERR_PTR(-ENODEV); + + if ((dev_type == ARPHRD_INFINIBAND) && !rdma_protocol_ib(device, port)) + return ERR_PTR(-ENODEV); + + if ((dev_type != ARPHRD_INFINIBAND) && rdma_protocol_ib(device, port)) + return ERR_PTR(-ENODEV); + + if (dev_type == ARPHRD_ETHER && rdma_protocol_roce(device, port)) { + ndev = dev_get_by_index(dev_addr->net, bound_if_index); + if (!ndev) + return ERR_PTR(-ENODEV); + } else { + gid_type = IB_GID_TYPE_IB; + } + + sgid_attr = rdma_find_gid_by_port(device, gid, gid_type, port, ndev); + if (ndev) + dev_put(ndev); + return sgid_attr; +} + +static void cma_bind_sgid_attr(struct rdma_id_private *id_priv, + const struct ib_gid_attr *sgid_attr) +{ + WARN_ON(id_priv->id.route.addr.dev_addr.sgid_attr); + id_priv->id.route.addr.dev_addr.sgid_attr = sgid_attr; +} + +/** + * cma_acquire_dev_by_src_ip - Acquire cma device, port, gid attribute + * based on source ip address. + * @id_priv: cm_id which should be bound to cma device + * + * cma_acquire_dev_by_src_ip() binds cm id to cma device, port and GID attribute + * based on source IP address. It returns 0 on success or error code otherwise. + * It is applicable to active and passive side cm_id. + */ +static int cma_acquire_dev_by_src_ip(struct rdma_id_private *id_priv) +{ + struct rdma_dev_addr *dev_addr = &id_priv->id.route.addr.dev_addr; + const struct ib_gid_attr *sgid_attr; + union ib_gid gid, iboe_gid, *gidp; + struct cma_device *cma_dev; + enum ib_gid_type gid_type; + int ret = -ENODEV; + unsigned int port; + + if (dev_addr->dev_type != ARPHRD_INFINIBAND && + id_priv->id.ps == RDMA_PS_IPOIB) + return -EINVAL; + + rdma_ip2gid((struct sockaddr *)&id_priv->id.route.addr.src_addr, + &iboe_gid); + + memcpy(&gid, dev_addr->src_dev_addr + + rdma_addr_gid_offset(dev_addr), sizeof(gid)); + + mutex_lock(&lock); + list_for_each_entry(cma_dev, &dev_list, list) { + rdma_for_each_port (cma_dev->device, port) { + gidp = rdma_protocol_roce(cma_dev->device, port) ? + &iboe_gid : &gid; + gid_type = cma_dev->default_gid_type[port - 1]; + sgid_attr = cma_validate_port(cma_dev->device, port, + gid_type, gidp, id_priv); + if (!IS_ERR(sgid_attr)) { + id_priv->id.port_num = port; + cma_bind_sgid_attr(id_priv, sgid_attr); + cma_attach_to_dev(id_priv, cma_dev); + ret = 0; + goto out; + } + } + } +out: + mutex_unlock(&lock); + return ret; +} + +/** + * cma_ib_acquire_dev - Acquire cma device, port and SGID attribute + * @id_priv: cm id to bind to cma device + * @listen_id_priv: listener cm id to match against + * @req: Pointer to req structure containaining incoming + * request information + * cma_ib_acquire_dev() acquires cma device, port and SGID attribute when + * rdma device matches for listen_id and incoming request. It also verifies + * that a GID table entry is present for the source address. + * Returns 0 on success, or returns error code otherwise. + */ +static int cma_ib_acquire_dev(struct rdma_id_private *id_priv, + const struct rdma_id_private *listen_id_priv, + struct cma_req_info *req) +{ + struct rdma_dev_addr *dev_addr = &id_priv->id.route.addr.dev_addr; + const struct ib_gid_attr *sgid_attr; + enum ib_gid_type gid_type; + union ib_gid gid; + + if (dev_addr->dev_type != ARPHRD_INFINIBAND && + id_priv->id.ps == RDMA_PS_IPOIB) + return -EINVAL; + + if (rdma_protocol_roce(req->device, req->port)) + rdma_ip2gid((struct sockaddr *)&id_priv->id.route.addr.src_addr, + &gid); + else + memcpy(&gid, dev_addr->src_dev_addr + + rdma_addr_gid_offset(dev_addr), sizeof(gid)); + + gid_type = listen_id_priv->cma_dev->default_gid_type[req->port - 1]; + sgid_attr = cma_validate_port(req->device, req->port, + gid_type, &gid, id_priv); + if (IS_ERR(sgid_attr)) + return PTR_ERR(sgid_attr); + + id_priv->id.port_num = req->port; + cma_bind_sgid_attr(id_priv, sgid_attr); + /* Need to acquire lock to protect against reader + * of cma_dev->id_list such as cma_netdev_callback() and + * cma_process_remove(). + */ + mutex_lock(&lock); + cma_attach_to_dev(id_priv, listen_id_priv->cma_dev); + mutex_unlock(&lock); + rdma_restrack_add(&id_priv->res); + return 0; +} + +static int cma_iw_acquire_dev(struct rdma_id_private *id_priv, + const struct rdma_id_private *listen_id_priv) +{ + struct rdma_dev_addr *dev_addr = &id_priv->id.route.addr.dev_addr; + const struct ib_gid_attr *sgid_attr; + struct cma_device *cma_dev; + enum ib_gid_type gid_type; + int ret = -ENODEV; + unsigned int port; + union ib_gid gid; + + if (dev_addr->dev_type != ARPHRD_INFINIBAND && + id_priv->id.ps == RDMA_PS_IPOIB) + return -EINVAL; + + memcpy(&gid, dev_addr->src_dev_addr + + rdma_addr_gid_offset(dev_addr), sizeof(gid)); + + mutex_lock(&lock); + + cma_dev = listen_id_priv->cma_dev; + port = listen_id_priv->id.port_num; + gid_type = listen_id_priv->gid_type; + sgid_attr = cma_validate_port(cma_dev->device, port, + gid_type, &gid, id_priv); + if (!IS_ERR(sgid_attr)) { + id_priv->id.port_num = port; + cma_bind_sgid_attr(id_priv, sgid_attr); + ret = 0; + goto out; + } + + list_for_each_entry(cma_dev, &dev_list, list) { + rdma_for_each_port (cma_dev->device, port) { + if (listen_id_priv->cma_dev == cma_dev && + listen_id_priv->id.port_num == port) + continue; + + gid_type = cma_dev->default_gid_type[port - 1]; + sgid_attr = cma_validate_port(cma_dev->device, port, + gid_type, &gid, id_priv); + if (!IS_ERR(sgid_attr)) { + id_priv->id.port_num = port; + cma_bind_sgid_attr(id_priv, sgid_attr); + ret = 0; + goto out; + } + } + } + +out: + if (!ret) { + cma_attach_to_dev(id_priv, cma_dev); + rdma_restrack_add(&id_priv->res); + } + + mutex_unlock(&lock); + return ret; +} + +/* + * Select the source IB device and address to reach the destination IB address. + */ +static int cma_resolve_ib_dev(struct rdma_id_private *id_priv) +{ + struct cma_device *cma_dev, *cur_dev; + struct sockaddr_ib *addr; + union ib_gid gid, sgid, *dgid; + unsigned int p; + u16 pkey, index; + enum ib_port_state port_state; + int ret; + int i; + + cma_dev = NULL; + addr = (struct sockaddr_ib *) cma_dst_addr(id_priv); + dgid = (union ib_gid *) &addr->sib_addr; + pkey = ntohs(addr->sib_pkey); + + mutex_lock(&lock); + list_for_each_entry(cur_dev, &dev_list, list) { + rdma_for_each_port (cur_dev->device, p) { + if (!rdma_cap_af_ib(cur_dev->device, p)) + continue; + + if (ib_find_cached_pkey(cur_dev->device, p, pkey, &index)) + continue; + + if (ib_get_cached_port_state(cur_dev->device, p, &port_state)) + continue; + + for (i = 0; i < cur_dev->device->port_data[p].immutable.gid_tbl_len; + ++i) { + ret = rdma_query_gid(cur_dev->device, p, i, + &gid); + if (ret) + continue; + + if (!memcmp(&gid, dgid, sizeof(gid))) { + cma_dev = cur_dev; + sgid = gid; + id_priv->id.port_num = p; + goto found; + } + + if (!cma_dev && (gid.global.subnet_prefix == + dgid->global.subnet_prefix) && + port_state == IB_PORT_ACTIVE) { + cma_dev = cur_dev; + sgid = gid; + id_priv->id.port_num = p; + goto found; + } + } + } + } + mutex_unlock(&lock); + return -ENODEV; + +found: + cma_attach_to_dev(id_priv, cma_dev); + rdma_restrack_add(&id_priv->res); + mutex_unlock(&lock); + addr = (struct sockaddr_ib *)cma_src_addr(id_priv); + memcpy(&addr->sib_addr, &sgid, sizeof(sgid)); + cma_translate_ib(addr, &id_priv->id.route.addr.dev_addr); + return 0; +} + +static void cma_id_get(struct rdma_id_private *id_priv) +{ + refcount_inc(&id_priv->refcount); +} + +static void cma_id_put(struct rdma_id_private *id_priv) +{ + if (refcount_dec_and_test(&id_priv->refcount)) + complete(&id_priv->comp); +} + +static struct rdma_id_private * +__rdma_create_id(struct net *net, rdma_cm_event_handler event_handler, + void *context, enum rdma_ucm_port_space ps, + enum ib_qp_type qp_type, const struct rdma_id_private *parent) +{ + struct rdma_id_private *id_priv; + + id_priv = kzalloc(sizeof *id_priv, GFP_KERNEL); + if (!id_priv) + return ERR_PTR(-ENOMEM); + + id_priv->state = RDMA_CM_IDLE; + id_priv->id.context = context; + id_priv->id.event_handler = event_handler; + id_priv->id.ps = ps; + id_priv->id.qp_type = qp_type; + id_priv->tos_set = false; + id_priv->timeout_set = false; + id_priv->gid_type = IB_GID_TYPE_IB; + spin_lock_init(&id_priv->lock); + mutex_init(&id_priv->qp_mutex); + init_completion(&id_priv->comp); + refcount_set(&id_priv->refcount, 1); + mutex_init(&id_priv->handler_mutex); + INIT_LIST_HEAD(&id_priv->listen_list); + INIT_LIST_HEAD(&id_priv->mc_list); + get_random_bytes(&id_priv->seq_num, sizeof id_priv->seq_num); + id_priv->id.route.addr.dev_addr.net = get_net(net); + id_priv->seq_num &= 0x00ffffff; + + rdma_restrack_new(&id_priv->res, RDMA_RESTRACK_CM_ID); + if (parent) + rdma_restrack_parent_name(&id_priv->res, &parent->res); + + return id_priv; +} + +struct rdma_cm_id * +__rdma_create_kernel_id(struct net *net, rdma_cm_event_handler event_handler, + void *context, enum rdma_ucm_port_space ps, + enum ib_qp_type qp_type, const char *caller) +{ + struct rdma_id_private *ret; + + ret = __rdma_create_id(net, event_handler, context, ps, qp_type, NULL); + if (IS_ERR(ret)) + return ERR_CAST(ret); + + rdma_restrack_set_name(&ret->res, caller); + return &ret->id; +} +EXPORT_SYMBOL(__rdma_create_kernel_id); + +struct rdma_cm_id *rdma_create_user_id(rdma_cm_event_handler event_handler, + void *context, + enum rdma_ucm_port_space ps, + enum ib_qp_type qp_type) +{ + struct rdma_id_private *ret; + + ret = __rdma_create_id(current->nsproxy->net_ns, event_handler, context, + ps, qp_type, NULL); + if (IS_ERR(ret)) + return ERR_CAST(ret); + + rdma_restrack_set_name(&ret->res, NULL); + return &ret->id; +} +EXPORT_SYMBOL(rdma_create_user_id); + +static int cma_init_ud_qp(struct rdma_id_private *id_priv, struct ib_qp *qp) +{ + struct ib_qp_attr qp_attr; + int qp_attr_mask, ret; + + qp_attr.qp_state = IB_QPS_INIT; + ret = rdma_init_qp_attr(&id_priv->id, &qp_attr, &qp_attr_mask); + if (ret) + return ret; + + ret = ib_modify_qp(qp, &qp_attr, qp_attr_mask); + if (ret) + return ret; + + qp_attr.qp_state = IB_QPS_RTR; + ret = ib_modify_qp(qp, &qp_attr, IB_QP_STATE); + if (ret) + return ret; + + qp_attr.qp_state = IB_QPS_RTS; + qp_attr.sq_psn = 0; + ret = ib_modify_qp(qp, &qp_attr, IB_QP_STATE | IB_QP_SQ_PSN); + + return ret; +} + +static int cma_init_conn_qp(struct rdma_id_private *id_priv, struct ib_qp *qp) +{ + struct ib_qp_attr qp_attr; + int qp_attr_mask, ret; + + qp_attr.qp_state = IB_QPS_INIT; + ret = rdma_init_qp_attr(&id_priv->id, &qp_attr, &qp_attr_mask); + if (ret) + return ret; + + return ib_modify_qp(qp, &qp_attr, qp_attr_mask); +} + +int rdma_create_qp(struct rdma_cm_id *id, struct ib_pd *pd, + struct ib_qp_init_attr *qp_init_attr) +{ + struct rdma_id_private *id_priv; + struct ib_qp *qp; + int ret; + + id_priv = container_of(id, struct rdma_id_private, id); + if (id->device != pd->device) { + ret = -EINVAL; + goto out_err; + } + + qp_init_attr->port_num = id->port_num; + qp = ib_create_qp(pd, qp_init_attr); + if (IS_ERR(qp)) { + ret = PTR_ERR(qp); + goto out_err; + } + + if (id->qp_type == IB_QPT_UD) + ret = cma_init_ud_qp(id_priv, qp); + else + ret = cma_init_conn_qp(id_priv, qp); + if (ret) + goto out_destroy; + + id->qp = qp; + id_priv->qp_num = qp->qp_num; + id_priv->srq = (qp->srq != NULL); + trace_cm_qp_create(id_priv, pd, qp_init_attr, 0); + return 0; +out_destroy: + ib_destroy_qp(qp); +out_err: + trace_cm_qp_create(id_priv, pd, qp_init_attr, ret); + return ret; +} +EXPORT_SYMBOL(rdma_create_qp); + +void rdma_destroy_qp(struct rdma_cm_id *id) +{ + struct rdma_id_private *id_priv; + + id_priv = container_of(id, struct rdma_id_private, id); + trace_cm_qp_destroy(id_priv); + mutex_lock(&id_priv->qp_mutex); + ib_destroy_qp(id_priv->id.qp); + id_priv->id.qp = NULL; + mutex_unlock(&id_priv->qp_mutex); +} +EXPORT_SYMBOL(rdma_destroy_qp); + +static int cma_modify_qp_rtr(struct rdma_id_private *id_priv, + struct rdma_conn_param *conn_param) +{ + struct ib_qp_attr qp_attr; + int qp_attr_mask, ret; + + mutex_lock(&id_priv->qp_mutex); + if (!id_priv->id.qp) { + ret = 0; + goto out; + } + + /* Need to update QP attributes from default values. */ + qp_attr.qp_state = IB_QPS_INIT; + ret = rdma_init_qp_attr(&id_priv->id, &qp_attr, &qp_attr_mask); + if (ret) + goto out; + + ret = ib_modify_qp(id_priv->id.qp, &qp_attr, qp_attr_mask); + if (ret) + goto out; + + qp_attr.qp_state = IB_QPS_RTR; + ret = rdma_init_qp_attr(&id_priv->id, &qp_attr, &qp_attr_mask); + if (ret) + goto out; + + BUG_ON(id_priv->cma_dev->device != id_priv->id.device); + + if (conn_param) + qp_attr.max_dest_rd_atomic = conn_param->responder_resources; + ret = ib_modify_qp(id_priv->id.qp, &qp_attr, qp_attr_mask); +out: + mutex_unlock(&id_priv->qp_mutex); + return ret; +} + +static int cma_modify_qp_rts(struct rdma_id_private *id_priv, + struct rdma_conn_param *conn_param) +{ + struct ib_qp_attr qp_attr; + int qp_attr_mask, ret; + + mutex_lock(&id_priv->qp_mutex); + if (!id_priv->id.qp) { + ret = 0; + goto out; + } + + qp_attr.qp_state = IB_QPS_RTS; + ret = rdma_init_qp_attr(&id_priv->id, &qp_attr, &qp_attr_mask); + if (ret) + goto out; + + if (conn_param) + qp_attr.max_rd_atomic = conn_param->initiator_depth; + ret = ib_modify_qp(id_priv->id.qp, &qp_attr, qp_attr_mask); +out: + mutex_unlock(&id_priv->qp_mutex); + return ret; +} + +static int cma_modify_qp_err(struct rdma_id_private *id_priv) +{ + struct ib_qp_attr qp_attr; + int ret; + + mutex_lock(&id_priv->qp_mutex); + if (!id_priv->id.qp) { + ret = 0; + goto out; + } + + qp_attr.qp_state = IB_QPS_ERR; + ret = ib_modify_qp(id_priv->id.qp, &qp_attr, IB_QP_STATE); +out: + mutex_unlock(&id_priv->qp_mutex); + return ret; +} + +static int cma_ib_init_qp_attr(struct rdma_id_private *id_priv, + struct ib_qp_attr *qp_attr, int *qp_attr_mask) +{ + struct rdma_dev_addr *dev_addr = &id_priv->id.route.addr.dev_addr; + int ret; + u16 pkey; + + if (rdma_cap_eth_ah(id_priv->id.device, id_priv->id.port_num)) + pkey = 0xffff; + else + pkey = ib_addr_get_pkey(dev_addr); + + ret = ib_find_cached_pkey(id_priv->id.device, id_priv->id.port_num, + pkey, &qp_attr->pkey_index); + if (ret) + return ret; + + qp_attr->port_num = id_priv->id.port_num; + *qp_attr_mask = IB_QP_STATE | IB_QP_PKEY_INDEX | IB_QP_PORT; + + if (id_priv->id.qp_type == IB_QPT_UD) { + ret = cma_set_default_qkey(id_priv); + if (ret) + return ret; + + qp_attr->qkey = id_priv->qkey; + *qp_attr_mask |= IB_QP_QKEY; + } else { + qp_attr->qp_access_flags = 0; + *qp_attr_mask |= IB_QP_ACCESS_FLAGS; + } + return 0; +} + +int rdma_init_qp_attr(struct rdma_cm_id *id, struct ib_qp_attr *qp_attr, + int *qp_attr_mask) +{ + struct rdma_id_private *id_priv; + int ret = 0; + + id_priv = container_of(id, struct rdma_id_private, id); + if (rdma_cap_ib_cm(id->device, id->port_num)) { + if (!id_priv->cm_id.ib || (id_priv->id.qp_type == IB_QPT_UD)) + ret = cma_ib_init_qp_attr(id_priv, qp_attr, qp_attr_mask); + else + ret = ib_cm_init_qp_attr(id_priv->cm_id.ib, qp_attr, + qp_attr_mask); + + if (qp_attr->qp_state == IB_QPS_RTR) + qp_attr->rq_psn = id_priv->seq_num; + } else if (rdma_cap_iw_cm(id->device, id->port_num)) { + if (!id_priv->cm_id.iw) { + qp_attr->qp_access_flags = 0; + *qp_attr_mask = IB_QP_STATE | IB_QP_ACCESS_FLAGS; + } else + ret = iw_cm_init_qp_attr(id_priv->cm_id.iw, qp_attr, + qp_attr_mask); + qp_attr->port_num = id_priv->id.port_num; + *qp_attr_mask |= IB_QP_PORT; + } else + ret = -ENOSYS; + + if ((*qp_attr_mask & IB_QP_TIMEOUT) && id_priv->timeout_set) + qp_attr->timeout = id_priv->timeout; + + return ret; +} +EXPORT_SYMBOL(rdma_init_qp_attr); + +static inline bool cma_zero_addr(const struct sockaddr *addr) +{ + switch (addr->sa_family) { + case AF_INET: + return ipv4_is_zeronet(((struct sockaddr_in *)addr)->sin_addr.s_addr); + case AF_INET6: + return ipv6_addr_any(&((struct sockaddr_in6 *)addr)->sin6_addr); + case AF_IB: + return ib_addr_any(&((struct sockaddr_ib *)addr)->sib_addr); + default: + return false; + } +} + +static inline bool cma_loopback_addr(const struct sockaddr *addr) +{ + switch (addr->sa_family) { + case AF_INET: + return ipv4_is_loopback( + ((struct sockaddr_in *)addr)->sin_addr.s_addr); + case AF_INET6: + return ipv6_addr_loopback( + &((struct sockaddr_in6 *)addr)->sin6_addr); + case AF_IB: + return ib_addr_loopback( + &((struct sockaddr_ib *)addr)->sib_addr); + default: + return false; + } +} + +static inline bool cma_any_addr(const struct sockaddr *addr) +{ + return cma_zero_addr(addr) || cma_loopback_addr(addr); +} + +static int cma_addr_cmp(const struct sockaddr *src, const struct sockaddr *dst) +{ + if (src->sa_family != dst->sa_family) + return -1; + + switch (src->sa_family) { + case AF_INET: + return ((struct sockaddr_in *)src)->sin_addr.s_addr != + ((struct sockaddr_in *)dst)->sin_addr.s_addr; + case AF_INET6: { + struct sockaddr_in6 *src_addr6 = (struct sockaddr_in6 *)src; + struct sockaddr_in6 *dst_addr6 = (struct sockaddr_in6 *)dst; + bool link_local; + + if (ipv6_addr_cmp(&src_addr6->sin6_addr, + &dst_addr6->sin6_addr)) + return 1; + link_local = ipv6_addr_type(&dst_addr6->sin6_addr) & + IPV6_ADDR_LINKLOCAL; + /* Link local must match their scope_ids */ + return link_local ? (src_addr6->sin6_scope_id != + dst_addr6->sin6_scope_id) : + 0; + } + + default: + return ib_addr_cmp(&((struct sockaddr_ib *) src)->sib_addr, + &((struct sockaddr_ib *) dst)->sib_addr); + } +} + +static __be16 cma_port(const struct sockaddr *addr) +{ + struct sockaddr_ib *sib; + + switch (addr->sa_family) { + case AF_INET: + return ((struct sockaddr_in *) addr)->sin_port; + case AF_INET6: + return ((struct sockaddr_in6 *) addr)->sin6_port; + case AF_IB: + sib = (struct sockaddr_ib *) addr; + return htons((u16) (be64_to_cpu(sib->sib_sid) & + be64_to_cpu(sib->sib_sid_mask))); + default: + return 0; + } +} + +static inline int cma_any_port(const struct sockaddr *addr) +{ + return !cma_port(addr); +} + +static void cma_save_ib_info(struct sockaddr *src_addr, + struct sockaddr *dst_addr, + const struct rdma_cm_id *listen_id, + const struct sa_path_rec *path) +{ + struct sockaddr_ib *listen_ib, *ib; + + listen_ib = (struct sockaddr_ib *) &listen_id->route.addr.src_addr; + if (src_addr) { + ib = (struct sockaddr_ib *)src_addr; + ib->sib_family = AF_IB; + if (path) { + ib->sib_pkey = path->pkey; + ib->sib_flowinfo = path->flow_label; + memcpy(&ib->sib_addr, &path->sgid, 16); + ib->sib_sid = path->service_id; + ib->sib_scope_id = 0; + } else { + ib->sib_pkey = listen_ib->sib_pkey; + ib->sib_flowinfo = listen_ib->sib_flowinfo; + ib->sib_addr = listen_ib->sib_addr; + ib->sib_sid = listen_ib->sib_sid; + ib->sib_scope_id = listen_ib->sib_scope_id; + } + ib->sib_sid_mask = cpu_to_be64(0xffffffffffffffffULL); + } + if (dst_addr) { + ib = (struct sockaddr_ib *)dst_addr; + ib->sib_family = AF_IB; + if (path) { + ib->sib_pkey = path->pkey; + ib->sib_flowinfo = path->flow_label; + memcpy(&ib->sib_addr, &path->dgid, 16); + } + } +} + +static void cma_save_ip4_info(struct sockaddr_in *src_addr, + struct sockaddr_in *dst_addr, + struct cma_hdr *hdr, + __be16 local_port) +{ + if (src_addr) { + *src_addr = (struct sockaddr_in) { + .sin_family = AF_INET, + .sin_addr.s_addr = hdr->dst_addr.ip4.addr, + .sin_port = local_port, + }; + } + + if (dst_addr) { + *dst_addr = (struct sockaddr_in) { + .sin_family = AF_INET, + .sin_addr.s_addr = hdr->src_addr.ip4.addr, + .sin_port = hdr->port, + }; + } +} + +static void cma_save_ip6_info(struct sockaddr_in6 *src_addr, + struct sockaddr_in6 *dst_addr, + struct cma_hdr *hdr, + __be16 local_port) +{ + if (src_addr) { + *src_addr = (struct sockaddr_in6) { + .sin6_family = AF_INET6, + .sin6_addr = hdr->dst_addr.ip6, + .sin6_port = local_port, + }; + } + + if (dst_addr) { + *dst_addr = (struct sockaddr_in6) { + .sin6_family = AF_INET6, + .sin6_addr = hdr->src_addr.ip6, + .sin6_port = hdr->port, + }; + } +} + +static u16 cma_port_from_service_id(__be64 service_id) +{ + return (u16)be64_to_cpu(service_id); +} + +static int cma_save_ip_info(struct sockaddr *src_addr, + struct sockaddr *dst_addr, + const struct ib_cm_event *ib_event, + __be64 service_id) +{ + struct cma_hdr *hdr; + __be16 port; + + hdr = ib_event->private_data; + if (hdr->cma_version != CMA_VERSION) + return -EINVAL; + + port = htons(cma_port_from_service_id(service_id)); + + switch (cma_get_ip_ver(hdr)) { + case 4: + cma_save_ip4_info((struct sockaddr_in *)src_addr, + (struct sockaddr_in *)dst_addr, hdr, port); + break; + case 6: + cma_save_ip6_info((struct sockaddr_in6 *)src_addr, + (struct sockaddr_in6 *)dst_addr, hdr, port); + break; + default: + return -EAFNOSUPPORT; + } + + return 0; +} + +static int cma_save_net_info(struct sockaddr *src_addr, + struct sockaddr *dst_addr, + const struct rdma_cm_id *listen_id, + const struct ib_cm_event *ib_event, + sa_family_t sa_family, __be64 service_id) +{ + if (sa_family == AF_IB) { + if (ib_event->event == IB_CM_REQ_RECEIVED) + cma_save_ib_info(src_addr, dst_addr, listen_id, + ib_event->param.req_rcvd.primary_path); + else if (ib_event->event == IB_CM_SIDR_REQ_RECEIVED) + cma_save_ib_info(src_addr, dst_addr, listen_id, NULL); + return 0; + } + + return cma_save_ip_info(src_addr, dst_addr, ib_event, service_id); +} + +static int cma_save_req_info(const struct ib_cm_event *ib_event, + struct cma_req_info *req) +{ + const struct ib_cm_req_event_param *req_param = + &ib_event->param.req_rcvd; + const struct ib_cm_sidr_req_event_param *sidr_param = + &ib_event->param.sidr_req_rcvd; + + switch (ib_event->event) { + case IB_CM_REQ_RECEIVED: + req->device = req_param->listen_id->device; + req->port = req_param->port; + memcpy(&req->local_gid, &req_param->primary_path->sgid, + sizeof(req->local_gid)); + req->has_gid = true; + req->service_id = req_param->primary_path->service_id; + req->pkey = be16_to_cpu(req_param->primary_path->pkey); + if (req->pkey != req_param->bth_pkey) + pr_warn_ratelimited("RDMA CMA: got different BTH P_Key (0x%x) and primary path P_Key (0x%x)\n" + "RDMA CMA: in the future this may cause the request to be dropped\n", + req_param->bth_pkey, req->pkey); + break; + case IB_CM_SIDR_REQ_RECEIVED: + req->device = sidr_param->listen_id->device; + req->port = sidr_param->port; + req->has_gid = false; + req->service_id = sidr_param->service_id; + req->pkey = sidr_param->pkey; + if (req->pkey != sidr_param->bth_pkey) + pr_warn_ratelimited("RDMA CMA: got different BTH P_Key (0x%x) and SIDR request payload P_Key (0x%x)\n" + "RDMA CMA: in the future this may cause the request to be dropped\n", + sidr_param->bth_pkey, req->pkey); + break; + default: + return -EINVAL; + } + + return 0; +} + +static bool validate_ipv4_net_dev(struct net_device *net_dev, + const struct sockaddr_in *dst_addr, + const struct sockaddr_in *src_addr) +{ + __be32 daddr = dst_addr->sin_addr.s_addr, + saddr = src_addr->sin_addr.s_addr; + struct fib_result res; + struct flowi4 fl4; + int err; + bool ret; + + if (ipv4_is_multicast(saddr) || ipv4_is_lbcast(saddr) || + ipv4_is_lbcast(daddr) || ipv4_is_zeronet(saddr) || + ipv4_is_zeronet(daddr) || ipv4_is_loopback(daddr) || + ipv4_is_loopback(saddr)) + return false; + + memset(&fl4, 0, sizeof(fl4)); + fl4.flowi4_oif = net_dev->ifindex; + fl4.daddr = daddr; + fl4.saddr = saddr; + + rcu_read_lock(); + err = fib_lookup(dev_net(net_dev), &fl4, &res, 0); + ret = err == 0 && FIB_RES_DEV(res) == net_dev; + rcu_read_unlock(); + + return ret; +} + +static bool validate_ipv6_net_dev(struct net_device *net_dev, + const struct sockaddr_in6 *dst_addr, + const struct sockaddr_in6 *src_addr) +{ +#if IS_ENABLED(CONFIG_IPV6) + const int strict = ipv6_addr_type(&dst_addr->sin6_addr) & + IPV6_ADDR_LINKLOCAL; + struct rt6_info *rt = rt6_lookup(dev_net(net_dev), &dst_addr->sin6_addr, + &src_addr->sin6_addr, net_dev->ifindex, + NULL, strict); + bool ret; + + if (!rt) + return false; + + ret = rt->rt6i_idev->dev == net_dev; + ip6_rt_put(rt); + + return ret; +#else + return false; +#endif +} + +static bool validate_net_dev(struct net_device *net_dev, + const struct sockaddr *daddr, + const struct sockaddr *saddr) +{ + const struct sockaddr_in *daddr4 = (const struct sockaddr_in *)daddr; + const struct sockaddr_in *saddr4 = (const struct sockaddr_in *)saddr; + const struct sockaddr_in6 *daddr6 = (const struct sockaddr_in6 *)daddr; + const struct sockaddr_in6 *saddr6 = (const struct sockaddr_in6 *)saddr; + + switch (daddr->sa_family) { + case AF_INET: + return saddr->sa_family == AF_INET && + validate_ipv4_net_dev(net_dev, daddr4, saddr4); + + case AF_INET6: + return saddr->sa_family == AF_INET6 && + validate_ipv6_net_dev(net_dev, daddr6, saddr6); + + default: + return false; + } +} + +static struct net_device * +roce_get_net_dev_by_cm_event(const struct ib_cm_event *ib_event) +{ + const struct ib_gid_attr *sgid_attr = NULL; + struct net_device *ndev; + + if (ib_event->event == IB_CM_REQ_RECEIVED) + sgid_attr = ib_event->param.req_rcvd.ppath_sgid_attr; + else if (ib_event->event == IB_CM_SIDR_REQ_RECEIVED) + sgid_attr = ib_event->param.sidr_req_rcvd.sgid_attr; + + if (!sgid_attr) + return NULL; + + rcu_read_lock(); + ndev = rdma_read_gid_attr_ndev_rcu(sgid_attr); + if (IS_ERR(ndev)) + ndev = NULL; + else + dev_hold(ndev); + rcu_read_unlock(); + return ndev; +} + +static struct net_device *cma_get_net_dev(const struct ib_cm_event *ib_event, + struct cma_req_info *req) +{ + struct sockaddr *listen_addr = + (struct sockaddr *)&req->listen_addr_storage; + struct sockaddr *src_addr = (struct sockaddr *)&req->src_addr_storage; + struct net_device *net_dev; + const union ib_gid *gid = req->has_gid ? &req->local_gid : NULL; + int err; + + err = cma_save_ip_info(listen_addr, src_addr, ib_event, + req->service_id); + if (err) + return ERR_PTR(err); + + if (rdma_protocol_roce(req->device, req->port)) + net_dev = roce_get_net_dev_by_cm_event(ib_event); + else + net_dev = ib_get_net_dev_by_params(req->device, req->port, + req->pkey, + gid, listen_addr); + if (!net_dev) + return ERR_PTR(-ENODEV); + + return net_dev; +} + +static enum rdma_ucm_port_space rdma_ps_from_service_id(__be64 service_id) +{ + return (be64_to_cpu(service_id) >> 16) & 0xffff; +} + +static bool cma_match_private_data(struct rdma_id_private *id_priv, + const struct cma_hdr *hdr) +{ + struct sockaddr *addr = cma_src_addr(id_priv); + __be32 ip4_addr; + struct in6_addr ip6_addr; + + if (cma_any_addr(addr) && !id_priv->afonly) + return true; + + switch (addr->sa_family) { + case AF_INET: + ip4_addr = ((struct sockaddr_in *)addr)->sin_addr.s_addr; + if (cma_get_ip_ver(hdr) != 4) + return false; + if (!cma_any_addr(addr) && + hdr->dst_addr.ip4.addr != ip4_addr) + return false; + break; + case AF_INET6: + ip6_addr = ((struct sockaddr_in6 *)addr)->sin6_addr; + if (cma_get_ip_ver(hdr) != 6) + return false; + if (!cma_any_addr(addr) && + memcmp(&hdr->dst_addr.ip6, &ip6_addr, sizeof(ip6_addr))) + return false; + break; + case AF_IB: + return true; + default: + return false; + } + + return true; +} + +static bool cma_protocol_roce(const struct rdma_cm_id *id) +{ + struct ib_device *device = id->device; + const int port_num = id->port_num ?: rdma_start_port(device); + + return rdma_protocol_roce(device, port_num); +} + +static bool cma_is_req_ipv6_ll(const struct cma_req_info *req) +{ + const struct sockaddr *daddr = + (const struct sockaddr *)&req->listen_addr_storage; + const struct sockaddr_in6 *daddr6 = (const struct sockaddr_in6 *)daddr; + + /* Returns true if the req is for IPv6 link local */ + return (daddr->sa_family == AF_INET6 && + (ipv6_addr_type(&daddr6->sin6_addr) & IPV6_ADDR_LINKLOCAL)); +} + +static bool cma_match_net_dev(const struct rdma_cm_id *id, + const struct net_device *net_dev, + const struct cma_req_info *req) +{ + const struct rdma_addr *addr = &id->route.addr; + + if (!net_dev) + /* This request is an AF_IB request */ + return (!id->port_num || id->port_num == req->port) && + (addr->src_addr.ss_family == AF_IB); + + /* + * If the request is not for IPv6 link local, allow matching + * request to any netdevice of the one or multiport rdma device. + */ + if (!cma_is_req_ipv6_ll(req)) + return true; + /* + * Net namespaces must match, and if the listner is listening + * on a specific netdevice than netdevice must match as well. + */ + if (net_eq(dev_net(net_dev), addr->dev_addr.net) && + (!!addr->dev_addr.bound_dev_if == + (addr->dev_addr.bound_dev_if == net_dev->ifindex))) + return true; + else + return false; +} + +static struct rdma_id_private *cma_find_listener( + const struct rdma_bind_list *bind_list, + const struct ib_cm_id *cm_id, + const struct ib_cm_event *ib_event, + const struct cma_req_info *req, + const struct net_device *net_dev) +{ + struct rdma_id_private *id_priv, *id_priv_dev; + + lockdep_assert_held(&lock); + + if (!bind_list) + return ERR_PTR(-EINVAL); + + hlist_for_each_entry(id_priv, &bind_list->owners, node) { + if (cma_match_private_data(id_priv, ib_event->private_data)) { + if (id_priv->id.device == cm_id->device && + cma_match_net_dev(&id_priv->id, net_dev, req)) + return id_priv; + list_for_each_entry(id_priv_dev, + &id_priv->listen_list, + listen_list) { + if (id_priv_dev->id.device == cm_id->device && + cma_match_net_dev(&id_priv_dev->id, + net_dev, req)) + return id_priv_dev; + } + } + } + + return ERR_PTR(-EINVAL); +} + +static struct rdma_id_private * +cma_ib_id_from_event(struct ib_cm_id *cm_id, + const struct ib_cm_event *ib_event, + struct cma_req_info *req, + struct net_device **net_dev) +{ + struct rdma_bind_list *bind_list; + struct rdma_id_private *id_priv; + int err; + + err = cma_save_req_info(ib_event, req); + if (err) + return ERR_PTR(err); + + *net_dev = cma_get_net_dev(ib_event, req); + if (IS_ERR(*net_dev)) { + if (PTR_ERR(*net_dev) == -EAFNOSUPPORT) { + /* Assuming the protocol is AF_IB */ + *net_dev = NULL; + } else { + return ERR_CAST(*net_dev); + } + } + + mutex_lock(&lock); + /* + * Net namespace might be getting deleted while route lookup, + * cm_id lookup is in progress. Therefore, perform netdevice + * validation, cm_id lookup under rcu lock. + * RCU lock along with netdevice state check, synchronizes with + * netdevice migrating to different net namespace and also avoids + * case where net namespace doesn't get deleted while lookup is in + * progress. + * If the device state is not IFF_UP, its properties such as ifindex + * and nd_net cannot be trusted to remain valid without rcu lock. + * net/core/dev.c change_net_namespace() ensures to synchronize with + * ongoing operations on net device after device is closed using + * synchronize_net(). + */ + rcu_read_lock(); + if (*net_dev) { + /* + * If netdevice is down, it is likely that it is administratively + * down or it might be migrating to different namespace. + * In that case avoid further processing, as the net namespace + * or ifindex may change. + */ + if (((*net_dev)->flags & IFF_UP) == 0) { + id_priv = ERR_PTR(-EHOSTUNREACH); + goto err; + } + + if (!validate_net_dev(*net_dev, + (struct sockaddr *)&req->src_addr_storage, + (struct sockaddr *)&req->listen_addr_storage)) { + id_priv = ERR_PTR(-EHOSTUNREACH); + goto err; + } + } + + bind_list = cma_ps_find(*net_dev ? dev_net(*net_dev) : &init_net, + rdma_ps_from_service_id(req->service_id), + cma_port_from_service_id(req->service_id)); + id_priv = cma_find_listener(bind_list, cm_id, ib_event, req, *net_dev); +err: + rcu_read_unlock(); + mutex_unlock(&lock); + if (IS_ERR(id_priv) && *net_dev) { + dev_put(*net_dev); + *net_dev = NULL; + } + return id_priv; +} + +static inline u8 cma_user_data_offset(struct rdma_id_private *id_priv) +{ + return cma_family(id_priv) == AF_IB ? 0 : sizeof(struct cma_hdr); +} + +static void cma_cancel_route(struct rdma_id_private *id_priv) +{ + if (rdma_cap_ib_sa(id_priv->id.device, id_priv->id.port_num)) { + if (id_priv->query) + ib_sa_cancel_query(id_priv->query_id, id_priv->query); + } +} + +static void _cma_cancel_listens(struct rdma_id_private *id_priv) +{ + struct rdma_id_private *dev_id_priv; + + lockdep_assert_held(&lock); + + /* + * Remove from listen_any_list to prevent added devices from spawning + * additional listen requests. + */ + list_del(&id_priv->list); + + while (!list_empty(&id_priv->listen_list)) { + dev_id_priv = list_entry(id_priv->listen_list.next, + struct rdma_id_private, listen_list); + /* sync with device removal to avoid duplicate destruction */ + list_del_init(&dev_id_priv->list); + list_del(&dev_id_priv->listen_list); + mutex_unlock(&lock); + + rdma_destroy_id(&dev_id_priv->id); + mutex_lock(&lock); + } +} + +static void cma_cancel_listens(struct rdma_id_private *id_priv) +{ + mutex_lock(&lock); + _cma_cancel_listens(id_priv); + mutex_unlock(&lock); +} + +static void cma_cancel_operation(struct rdma_id_private *id_priv, + enum rdma_cm_state state) +{ + switch (state) { + case RDMA_CM_ADDR_QUERY: + /* + * We can avoid doing the rdma_addr_cancel() based on state, + * only RDMA_CM_ADDR_QUERY has a work that could still execute. + * Notice that the addr_handler work could still be exiting + * outside this state, however due to the interaction with the + * handler_mutex the work is guaranteed not to touch id_priv + * during exit. + */ + rdma_addr_cancel(&id_priv->id.route.addr.dev_addr); + break; + case RDMA_CM_ROUTE_QUERY: + cma_cancel_route(id_priv); + break; + case RDMA_CM_LISTEN: + if (cma_any_addr(cma_src_addr(id_priv)) && !id_priv->cma_dev) + cma_cancel_listens(id_priv); + break; + default: + break; + } +} + +static void cma_release_port(struct rdma_id_private *id_priv) +{ + struct rdma_bind_list *bind_list = id_priv->bind_list; + struct net *net = id_priv->id.route.addr.dev_addr.net; + + if (!bind_list) + return; + + mutex_lock(&lock); + hlist_del(&id_priv->node); + if (hlist_empty(&bind_list->owners)) { + cma_ps_remove(net, bind_list->ps, bind_list->port); + kfree(bind_list); + } + mutex_unlock(&lock); +} + +static void destroy_mc(struct rdma_id_private *id_priv, + struct cma_multicast *mc) +{ + bool send_only = mc->join_state == BIT(SENDONLY_FULLMEMBER_JOIN); + + if (rdma_cap_ib_mcast(id_priv->id.device, id_priv->id.port_num)) + ib_sa_free_multicast(mc->sa_mc); + + if (rdma_protocol_roce(id_priv->id.device, id_priv->id.port_num)) { + struct rdma_dev_addr *dev_addr = + &id_priv->id.route.addr.dev_addr; + struct net_device *ndev = NULL; + + if (dev_addr->bound_dev_if) + ndev = dev_get_by_index(dev_addr->net, + dev_addr->bound_dev_if); + if (ndev && !send_only) { + enum ib_gid_type gid_type; + union ib_gid mgid; + + gid_type = id_priv->cma_dev->default_gid_type + [id_priv->id.port_num - + rdma_start_port( + id_priv->cma_dev->device)]; + cma_iboe_set_mgid((struct sockaddr *)&mc->addr, &mgid, + gid_type); + cma_igmp_send(ndev, &mgid, false); + } + dev_put(ndev); + + cancel_work_sync(&mc->iboe_join.work); + } + kfree(mc); +} + +static void cma_leave_mc_groups(struct rdma_id_private *id_priv) +{ + struct cma_multicast *mc; + + while (!list_empty(&id_priv->mc_list)) { + mc = list_first_entry(&id_priv->mc_list, struct cma_multicast, + list); + list_del(&mc->list); + destroy_mc(id_priv, mc); + } +} + +static void _destroy_id(struct rdma_id_private *id_priv, + enum rdma_cm_state state) +{ + cma_cancel_operation(id_priv, state); + + rdma_restrack_del(&id_priv->res); + if (id_priv->cma_dev) { + if (rdma_cap_ib_cm(id_priv->id.device, 1)) { + if (id_priv->cm_id.ib) + ib_destroy_cm_id(id_priv->cm_id.ib); + } else if (rdma_cap_iw_cm(id_priv->id.device, 1)) { + if (id_priv->cm_id.iw) + iw_destroy_cm_id(id_priv->cm_id.iw); + } + cma_leave_mc_groups(id_priv); + cma_release_dev(id_priv); + } + + cma_release_port(id_priv); + cma_id_put(id_priv); + wait_for_completion(&id_priv->comp); + + if (id_priv->internal_id) + cma_id_put(id_priv->id.context); + + kfree(id_priv->id.route.path_rec); + + put_net(id_priv->id.route.addr.dev_addr.net); + kfree(id_priv); +} + +/* + * destroy an ID from within the handler_mutex. This ensures that no other + * handlers can start running concurrently. + */ +static void destroy_id_handler_unlock(struct rdma_id_private *id_priv) + __releases(&idprv->handler_mutex) +{ + enum rdma_cm_state state; + unsigned long flags; + + trace_cm_id_destroy(id_priv); + + /* + * Setting the state to destroyed under the handler mutex provides a + * fence against calling handler callbacks. If this is invoked due to + * the failure of a handler callback then it guarentees that no future + * handlers will be called. + */ + lockdep_assert_held(&id_priv->handler_mutex); + spin_lock_irqsave(&id_priv->lock, flags); + state = id_priv->state; + id_priv->state = RDMA_CM_DESTROYING; + spin_unlock_irqrestore(&id_priv->lock, flags); + mutex_unlock(&id_priv->handler_mutex); + _destroy_id(id_priv, state); +} + +void rdma_destroy_id(struct rdma_cm_id *id) +{ + struct rdma_id_private *id_priv = + container_of(id, struct rdma_id_private, id); + + mutex_lock(&id_priv->handler_mutex); + destroy_id_handler_unlock(id_priv); +} +EXPORT_SYMBOL(rdma_destroy_id); + +static int cma_rep_recv(struct rdma_id_private *id_priv) +{ + int ret; + + ret = cma_modify_qp_rtr(id_priv, NULL); + if (ret) + goto reject; + + ret = cma_modify_qp_rts(id_priv, NULL); + if (ret) + goto reject; + + trace_cm_send_rtu(id_priv); + ret = ib_send_cm_rtu(id_priv->cm_id.ib, NULL, 0); + if (ret) + goto reject; + + return 0; +reject: + pr_debug_ratelimited("RDMA CM: CONNECT_ERROR: failed to handle reply. status %d\n", ret); + cma_modify_qp_err(id_priv); + trace_cm_send_rej(id_priv); + ib_send_cm_rej(id_priv->cm_id.ib, IB_CM_REJ_CONSUMER_DEFINED, + NULL, 0, NULL, 0); + return ret; +} + +static void cma_set_rep_event_data(struct rdma_cm_event *event, + const struct ib_cm_rep_event_param *rep_data, + void *private_data) +{ + event->param.conn.private_data = private_data; + event->param.conn.private_data_len = IB_CM_REP_PRIVATE_DATA_SIZE; + event->param.conn.responder_resources = rep_data->responder_resources; + event->param.conn.initiator_depth = rep_data->initiator_depth; + event->param.conn.flow_control = rep_data->flow_control; + event->param.conn.rnr_retry_count = rep_data->rnr_retry_count; + event->param.conn.srq = rep_data->srq; + event->param.conn.qp_num = rep_data->remote_qpn; + + event->ece.vendor_id = rep_data->ece.vendor_id; + event->ece.attr_mod = rep_data->ece.attr_mod; +} + +static int cma_cm_event_handler(struct rdma_id_private *id_priv, + struct rdma_cm_event *event) +{ + int ret; + + lockdep_assert_held(&id_priv->handler_mutex); + + trace_cm_event_handler(id_priv, event); + ret = id_priv->id.event_handler(&id_priv->id, event); + trace_cm_event_done(id_priv, event, ret); + return ret; +} + +static int cma_ib_handler(struct ib_cm_id *cm_id, + const struct ib_cm_event *ib_event) +{ + struct rdma_id_private *id_priv = cm_id->context; + struct rdma_cm_event event = {}; + enum rdma_cm_state state; + int ret; + + mutex_lock(&id_priv->handler_mutex); + state = READ_ONCE(id_priv->state); + if ((ib_event->event != IB_CM_TIMEWAIT_EXIT && + state != RDMA_CM_CONNECT) || + (ib_event->event == IB_CM_TIMEWAIT_EXIT && + state != RDMA_CM_DISCONNECT)) + goto out; + + switch (ib_event->event) { + case IB_CM_REQ_ERROR: + case IB_CM_REP_ERROR: + event.event = RDMA_CM_EVENT_UNREACHABLE; + event.status = -ETIMEDOUT; + break; + case IB_CM_REP_RECEIVED: + if (state == RDMA_CM_CONNECT && + (id_priv->id.qp_type != IB_QPT_UD)) { + trace_cm_send_mra(id_priv); + ib_send_cm_mra(cm_id, CMA_CM_MRA_SETTING, NULL, 0); + } + if (id_priv->id.qp) { + event.status = cma_rep_recv(id_priv); + event.event = event.status ? RDMA_CM_EVENT_CONNECT_ERROR : + RDMA_CM_EVENT_ESTABLISHED; + } else { + event.event = RDMA_CM_EVENT_CONNECT_RESPONSE; + } + cma_set_rep_event_data(&event, &ib_event->param.rep_rcvd, + ib_event->private_data); + break; + case IB_CM_RTU_RECEIVED: + case IB_CM_USER_ESTABLISHED: + event.event = RDMA_CM_EVENT_ESTABLISHED; + break; + case IB_CM_DREQ_ERROR: + event.status = -ETIMEDOUT; + fallthrough; + case IB_CM_DREQ_RECEIVED: + case IB_CM_DREP_RECEIVED: + if (!cma_comp_exch(id_priv, RDMA_CM_CONNECT, + RDMA_CM_DISCONNECT)) + goto out; + event.event = RDMA_CM_EVENT_DISCONNECTED; + break; + case IB_CM_TIMEWAIT_EXIT: + event.event = RDMA_CM_EVENT_TIMEWAIT_EXIT; + break; + case IB_CM_MRA_RECEIVED: + /* ignore event */ + goto out; + case IB_CM_REJ_RECEIVED: + pr_debug_ratelimited("RDMA CM: REJECTED: %s\n", rdma_reject_msg(&id_priv->id, + ib_event->param.rej_rcvd.reason)); + cma_modify_qp_err(id_priv); + event.status = ib_event->param.rej_rcvd.reason; + event.event = RDMA_CM_EVENT_REJECTED; + event.param.conn.private_data = ib_event->private_data; + event.param.conn.private_data_len = IB_CM_REJ_PRIVATE_DATA_SIZE; + break; + default: + pr_err("RDMA CMA: unexpected IB CM event: %d\n", + ib_event->event); + goto out; + } + + ret = cma_cm_event_handler(id_priv, &event); + if (ret) { + /* Destroy the CM ID by returning a non-zero value. */ + id_priv->cm_id.ib = NULL; + destroy_id_handler_unlock(id_priv); + return ret; + } +out: + mutex_unlock(&id_priv->handler_mutex); + return 0; +} + +static struct rdma_id_private * +cma_ib_new_conn_id(const struct rdma_cm_id *listen_id, + const struct ib_cm_event *ib_event, + struct net_device *net_dev) +{ + struct rdma_id_private *listen_id_priv; + struct rdma_id_private *id_priv; + struct rdma_cm_id *id; + struct rdma_route *rt; + const sa_family_t ss_family = listen_id->route.addr.src_addr.ss_family; + struct sa_path_rec *path = ib_event->param.req_rcvd.primary_path; + const __be64 service_id = + ib_event->param.req_rcvd.primary_path->service_id; + int ret; + + listen_id_priv = container_of(listen_id, struct rdma_id_private, id); + id_priv = __rdma_create_id(listen_id->route.addr.dev_addr.net, + listen_id->event_handler, listen_id->context, + listen_id->ps, + ib_event->param.req_rcvd.qp_type, + listen_id_priv); + if (IS_ERR(id_priv)) + return NULL; + + id = &id_priv->id; + if (cma_save_net_info((struct sockaddr *)&id->route.addr.src_addr, + (struct sockaddr *)&id->route.addr.dst_addr, + listen_id, ib_event, ss_family, service_id)) + goto err; + + rt = &id->route; + rt->num_paths = ib_event->param.req_rcvd.alternate_path ? 2 : 1; + rt->path_rec = kmalloc_array(rt->num_paths, sizeof(*rt->path_rec), + GFP_KERNEL); + if (!rt->path_rec) + goto err; + + rt->path_rec[0] = *path; + if (rt->num_paths == 2) + rt->path_rec[1] = *ib_event->param.req_rcvd.alternate_path; + + if (net_dev) { + rdma_copy_src_l2_addr(&rt->addr.dev_addr, net_dev); + } else { + if (!cma_protocol_roce(listen_id) && + cma_any_addr(cma_src_addr(id_priv))) { + rt->addr.dev_addr.dev_type = ARPHRD_INFINIBAND; + rdma_addr_set_sgid(&rt->addr.dev_addr, &rt->path_rec[0].sgid); + ib_addr_set_pkey(&rt->addr.dev_addr, be16_to_cpu(rt->path_rec[0].pkey)); + } else if (!cma_any_addr(cma_src_addr(id_priv))) { + ret = cma_translate_addr(cma_src_addr(id_priv), &rt->addr.dev_addr); + if (ret) + goto err; + } + } + rdma_addr_set_dgid(&rt->addr.dev_addr, &rt->path_rec[0].dgid); + + id_priv->state = RDMA_CM_CONNECT; + return id_priv; + +err: + rdma_destroy_id(id); + return NULL; +} + +static struct rdma_id_private * +cma_ib_new_udp_id(const struct rdma_cm_id *listen_id, + const struct ib_cm_event *ib_event, + struct net_device *net_dev) +{ + const struct rdma_id_private *listen_id_priv; + struct rdma_id_private *id_priv; + struct rdma_cm_id *id; + const sa_family_t ss_family = listen_id->route.addr.src_addr.ss_family; + struct net *net = listen_id->route.addr.dev_addr.net; + int ret; + + listen_id_priv = container_of(listen_id, struct rdma_id_private, id); + id_priv = __rdma_create_id(net, listen_id->event_handler, + listen_id->context, listen_id->ps, IB_QPT_UD, + listen_id_priv); + if (IS_ERR(id_priv)) + return NULL; + + id = &id_priv->id; + if (cma_save_net_info((struct sockaddr *)&id->route.addr.src_addr, + (struct sockaddr *)&id->route.addr.dst_addr, + listen_id, ib_event, ss_family, + ib_event->param.sidr_req_rcvd.service_id)) + goto err; + + if (net_dev) { + rdma_copy_src_l2_addr(&id->route.addr.dev_addr, net_dev); + } else { + if (!cma_any_addr(cma_src_addr(id_priv))) { + ret = cma_translate_addr(cma_src_addr(id_priv), + &id->route.addr.dev_addr); + if (ret) + goto err; + } + } + + id_priv->state = RDMA_CM_CONNECT; + return id_priv; +err: + rdma_destroy_id(id); + return NULL; +} + +static void cma_set_req_event_data(struct rdma_cm_event *event, + const struct ib_cm_req_event_param *req_data, + void *private_data, int offset) +{ + event->param.conn.private_data = private_data + offset; + event->param.conn.private_data_len = IB_CM_REQ_PRIVATE_DATA_SIZE - offset; + event->param.conn.responder_resources = req_data->responder_resources; + event->param.conn.initiator_depth = req_data->initiator_depth; + event->param.conn.flow_control = req_data->flow_control; + event->param.conn.retry_count = req_data->retry_count; + event->param.conn.rnr_retry_count = req_data->rnr_retry_count; + event->param.conn.srq = req_data->srq; + event->param.conn.qp_num = req_data->remote_qpn; + + event->ece.vendor_id = req_data->ece.vendor_id; + event->ece.attr_mod = req_data->ece.attr_mod; +} + +static int cma_ib_check_req_qp_type(const struct rdma_cm_id *id, + const struct ib_cm_event *ib_event) +{ + return (((ib_event->event == IB_CM_REQ_RECEIVED) && + (ib_event->param.req_rcvd.qp_type == id->qp_type)) || + ((ib_event->event == IB_CM_SIDR_REQ_RECEIVED) && + (id->qp_type == IB_QPT_UD)) || + (!id->qp_type)); +} + +static int cma_ib_req_handler(struct ib_cm_id *cm_id, + const struct ib_cm_event *ib_event) +{ + struct rdma_id_private *listen_id, *conn_id = NULL; + struct rdma_cm_event event = {}; + struct cma_req_info req = {}; + struct net_device *net_dev; + u8 offset; + int ret; + + listen_id = cma_ib_id_from_event(cm_id, ib_event, &req, &net_dev); + if (IS_ERR(listen_id)) + return PTR_ERR(listen_id); + + trace_cm_req_handler(listen_id, ib_event->event); + if (!cma_ib_check_req_qp_type(&listen_id->id, ib_event)) { + ret = -EINVAL; + goto net_dev_put; + } + + mutex_lock(&listen_id->handler_mutex); + if (READ_ONCE(listen_id->state) != RDMA_CM_LISTEN) { + ret = -ECONNABORTED; + goto err_unlock; + } + + offset = cma_user_data_offset(listen_id); + event.event = RDMA_CM_EVENT_CONNECT_REQUEST; + if (ib_event->event == IB_CM_SIDR_REQ_RECEIVED) { + conn_id = cma_ib_new_udp_id(&listen_id->id, ib_event, net_dev); + event.param.ud.private_data = ib_event->private_data + offset; + event.param.ud.private_data_len = + IB_CM_SIDR_REQ_PRIVATE_DATA_SIZE - offset; + } else { + conn_id = cma_ib_new_conn_id(&listen_id->id, ib_event, net_dev); + cma_set_req_event_data(&event, &ib_event->param.req_rcvd, + ib_event->private_data, offset); + } + if (!conn_id) { + ret = -ENOMEM; + goto err_unlock; + } + + mutex_lock_nested(&conn_id->handler_mutex, SINGLE_DEPTH_NESTING); + ret = cma_ib_acquire_dev(conn_id, listen_id, &req); + if (ret) { + destroy_id_handler_unlock(conn_id); + goto err_unlock; + } + + conn_id->cm_id.ib = cm_id; + cm_id->context = conn_id; + cm_id->cm_handler = cma_ib_handler; + + ret = cma_cm_event_handler(conn_id, &event); + if (ret) { + /* Destroy the CM ID by returning a non-zero value. */ + conn_id->cm_id.ib = NULL; + mutex_unlock(&listen_id->handler_mutex); + destroy_id_handler_unlock(conn_id); + goto net_dev_put; + } + + if (READ_ONCE(conn_id->state) == RDMA_CM_CONNECT && + conn_id->id.qp_type != IB_QPT_UD) { + trace_cm_send_mra(cm_id->context); + ib_send_cm_mra(cm_id, CMA_CM_MRA_SETTING, NULL, 0); + } + mutex_unlock(&conn_id->handler_mutex); + +err_unlock: + mutex_unlock(&listen_id->handler_mutex); + +net_dev_put: + if (net_dev) + dev_put(net_dev); + + return ret; +} + +__be64 rdma_get_service_id(struct rdma_cm_id *id, struct sockaddr *addr) +{ + if (addr->sa_family == AF_IB) + return ((struct sockaddr_ib *) addr)->sib_sid; + + return cpu_to_be64(((u64)id->ps << 16) + be16_to_cpu(cma_port(addr))); +} +EXPORT_SYMBOL(rdma_get_service_id); + +void rdma_read_gids(struct rdma_cm_id *cm_id, union ib_gid *sgid, + union ib_gid *dgid) +{ + struct rdma_addr *addr = &cm_id->route.addr; + + if (!cm_id->device) { + if (sgid) + memset(sgid, 0, sizeof(*sgid)); + if (dgid) + memset(dgid, 0, sizeof(*dgid)); + return; + } + + if (rdma_protocol_roce(cm_id->device, cm_id->port_num)) { + if (sgid) + rdma_ip2gid((struct sockaddr *)&addr->src_addr, sgid); + if (dgid) + rdma_ip2gid((struct sockaddr *)&addr->dst_addr, dgid); + } else { + if (sgid) + rdma_addr_get_sgid(&addr->dev_addr, sgid); + if (dgid) + rdma_addr_get_dgid(&addr->dev_addr, dgid); + } +} +EXPORT_SYMBOL(rdma_read_gids); + +static int cma_iw_handler(struct iw_cm_id *iw_id, struct iw_cm_event *iw_event) +{ + struct rdma_id_private *id_priv = iw_id->context; + struct rdma_cm_event event = {}; + int ret = 0; + struct sockaddr *laddr = (struct sockaddr *)&iw_event->local_addr; + struct sockaddr *raddr = (struct sockaddr *)&iw_event->remote_addr; + + mutex_lock(&id_priv->handler_mutex); + if (READ_ONCE(id_priv->state) != RDMA_CM_CONNECT) + goto out; + + switch (iw_event->event) { + case IW_CM_EVENT_CLOSE: + event.event = RDMA_CM_EVENT_DISCONNECTED; + break; + case IW_CM_EVENT_CONNECT_REPLY: + memcpy(cma_src_addr(id_priv), laddr, + rdma_addr_size(laddr)); + memcpy(cma_dst_addr(id_priv), raddr, + rdma_addr_size(raddr)); + switch (iw_event->status) { + case 0: + event.event = RDMA_CM_EVENT_ESTABLISHED; + event.param.conn.initiator_depth = iw_event->ird; + event.param.conn.responder_resources = iw_event->ord; + break; + case -ECONNRESET: + case -ECONNREFUSED: + event.event = RDMA_CM_EVENT_REJECTED; + break; + case -ETIMEDOUT: + event.event = RDMA_CM_EVENT_UNREACHABLE; + break; + default: + event.event = RDMA_CM_EVENT_CONNECT_ERROR; + break; + } + break; + case IW_CM_EVENT_ESTABLISHED: + event.event = RDMA_CM_EVENT_ESTABLISHED; + event.param.conn.initiator_depth = iw_event->ird; + event.param.conn.responder_resources = iw_event->ord; + break; + default: + goto out; + } + + event.status = iw_event->status; + event.param.conn.private_data = iw_event->private_data; + event.param.conn.private_data_len = iw_event->private_data_len; + ret = cma_cm_event_handler(id_priv, &event); + if (ret) { + /* Destroy the CM ID by returning a non-zero value. */ + id_priv->cm_id.iw = NULL; + destroy_id_handler_unlock(id_priv); + return ret; + } + +out: + mutex_unlock(&id_priv->handler_mutex); + return ret; +} + +static int iw_conn_req_handler(struct iw_cm_id *cm_id, + struct iw_cm_event *iw_event) +{ + struct rdma_id_private *listen_id, *conn_id; + struct rdma_cm_event event = {}; + int ret = -ECONNABORTED; + struct sockaddr *laddr = (struct sockaddr *)&iw_event->local_addr; + struct sockaddr *raddr = (struct sockaddr *)&iw_event->remote_addr; + + event.event = RDMA_CM_EVENT_CONNECT_REQUEST; + event.param.conn.private_data = iw_event->private_data; + event.param.conn.private_data_len = iw_event->private_data_len; + event.param.conn.initiator_depth = iw_event->ird; + event.param.conn.responder_resources = iw_event->ord; + + listen_id = cm_id->context; + + mutex_lock(&listen_id->handler_mutex); + if (READ_ONCE(listen_id->state) != RDMA_CM_LISTEN) + goto out; + + /* Create a new RDMA id for the new IW CM ID */ + conn_id = __rdma_create_id(listen_id->id.route.addr.dev_addr.net, + listen_id->id.event_handler, + listen_id->id.context, RDMA_PS_TCP, + IB_QPT_RC, listen_id); + if (IS_ERR(conn_id)) { + ret = -ENOMEM; + goto out; + } + mutex_lock_nested(&conn_id->handler_mutex, SINGLE_DEPTH_NESTING); + conn_id->state = RDMA_CM_CONNECT; + + ret = rdma_translate_ip(laddr, &conn_id->id.route.addr.dev_addr); + if (ret) { + mutex_unlock(&listen_id->handler_mutex); + destroy_id_handler_unlock(conn_id); + return ret; + } + + ret = cma_iw_acquire_dev(conn_id, listen_id); + if (ret) { + mutex_unlock(&listen_id->handler_mutex); + destroy_id_handler_unlock(conn_id); + return ret; + } + + conn_id->cm_id.iw = cm_id; + cm_id->context = conn_id; + cm_id->cm_handler = cma_iw_handler; + + memcpy(cma_src_addr(conn_id), laddr, rdma_addr_size(laddr)); + memcpy(cma_dst_addr(conn_id), raddr, rdma_addr_size(raddr)); + + ret = cma_cm_event_handler(conn_id, &event); + if (ret) { + /* User wants to destroy the CM ID */ + conn_id->cm_id.iw = NULL; + mutex_unlock(&listen_id->handler_mutex); + destroy_id_handler_unlock(conn_id); + return ret; + } + + mutex_unlock(&conn_id->handler_mutex); + +out: + mutex_unlock(&listen_id->handler_mutex); + return ret; +} + +static int cma_ib_listen(struct rdma_id_private *id_priv) +{ + struct sockaddr *addr; + struct ib_cm_id *id; + __be64 svc_id; + + addr = cma_src_addr(id_priv); + svc_id = rdma_get_service_id(&id_priv->id, addr); + id = ib_cm_insert_listen(id_priv->id.device, + cma_ib_req_handler, svc_id); + if (IS_ERR(id)) + return PTR_ERR(id); + id_priv->cm_id.ib = id; + + return 0; +} + +static int cma_iw_listen(struct rdma_id_private *id_priv, int backlog) +{ + int ret; + struct iw_cm_id *id; + + id = iw_create_cm_id(id_priv->id.device, + iw_conn_req_handler, + id_priv); + if (IS_ERR(id)) + return PTR_ERR(id); + + mutex_lock(&id_priv->qp_mutex); + id->tos = id_priv->tos; + id->tos_set = id_priv->tos_set; + mutex_unlock(&id_priv->qp_mutex); + id_priv->cm_id.iw = id; + + memcpy(&id_priv->cm_id.iw->local_addr, cma_src_addr(id_priv), + rdma_addr_size(cma_src_addr(id_priv))); + + ret = iw_cm_listen(id_priv->cm_id.iw, backlog); + + if (ret) { + iw_destroy_cm_id(id_priv->cm_id.iw); + id_priv->cm_id.iw = NULL; + } + + return ret; +} + +static int cma_listen_handler(struct rdma_cm_id *id, + struct rdma_cm_event *event) +{ + struct rdma_id_private *id_priv = id->context; + + /* Listening IDs are always destroyed on removal */ + if (event->event == RDMA_CM_EVENT_DEVICE_REMOVAL) + return -1; + + id->context = id_priv->id.context; + id->event_handler = id_priv->id.event_handler; + trace_cm_event_handler(id_priv, event); + return id_priv->id.event_handler(id, event); +} + +static int cma_listen_on_dev(struct rdma_id_private *id_priv, + struct cma_device *cma_dev, + struct rdma_id_private **to_destroy) +{ + struct rdma_id_private *dev_id_priv; + struct net *net = id_priv->id.route.addr.dev_addr.net; + int ret; + + lockdep_assert_held(&lock); + + *to_destroy = NULL; + if (cma_family(id_priv) == AF_IB && !rdma_cap_ib_cm(cma_dev->device, 1)) + return 0; + + dev_id_priv = + __rdma_create_id(net, cma_listen_handler, id_priv, + id_priv->id.ps, id_priv->id.qp_type, id_priv); + if (IS_ERR(dev_id_priv)) + return PTR_ERR(dev_id_priv); + + dev_id_priv->state = RDMA_CM_ADDR_BOUND; + memcpy(cma_src_addr(dev_id_priv), cma_src_addr(id_priv), + rdma_addr_size(cma_src_addr(id_priv))); + + _cma_attach_to_dev(dev_id_priv, cma_dev); + rdma_restrack_add(&dev_id_priv->res); + cma_id_get(id_priv); + dev_id_priv->internal_id = 1; + dev_id_priv->afonly = id_priv->afonly; + mutex_lock(&id_priv->qp_mutex); + dev_id_priv->tos_set = id_priv->tos_set; + dev_id_priv->tos = id_priv->tos; + mutex_unlock(&id_priv->qp_mutex); + + ret = rdma_listen(&dev_id_priv->id, id_priv->backlog); + if (ret) + goto err_listen; + list_add_tail(&dev_id_priv->listen_list, &id_priv->listen_list); + return 0; +err_listen: + /* Caller must destroy this after releasing lock */ + *to_destroy = dev_id_priv; + dev_warn(&cma_dev->device->dev, "RDMA CMA: %s, error %d\n", __func__, ret); + return ret; +} + +static int cma_listen_on_all(struct rdma_id_private *id_priv) +{ + struct rdma_id_private *to_destroy; + struct cma_device *cma_dev; + int ret; + + mutex_lock(&lock); + list_add_tail(&id_priv->list, &listen_any_list); + list_for_each_entry(cma_dev, &dev_list, list) { + ret = cma_listen_on_dev(id_priv, cma_dev, &to_destroy); + if (ret) { + /* Prevent racing with cma_process_remove() */ + if (to_destroy) + list_del_init(&to_destroy->list); + goto err_listen; + } + } + mutex_unlock(&lock); + return 0; + +err_listen: + _cma_cancel_listens(id_priv); + mutex_unlock(&lock); + if (to_destroy) + rdma_destroy_id(&to_destroy->id); + return ret; +} + +void rdma_set_service_type(struct rdma_cm_id *id, int tos) +{ + struct rdma_id_private *id_priv; + + id_priv = container_of(id, struct rdma_id_private, id); + mutex_lock(&id_priv->qp_mutex); + id_priv->tos = (u8) tos; + id_priv->tos_set = true; + mutex_unlock(&id_priv->qp_mutex); +} +EXPORT_SYMBOL(rdma_set_service_type); + +/** + * rdma_set_ack_timeout() - Set the ack timeout of QP associated + * with a connection identifier. + * @id: Communication identifier to associated with service type. + * @timeout: Ack timeout to set a QP, expressed as 4.096 * 2^(timeout) usec. + * + * This function should be called before rdma_connect() on active side, + * and on passive side before rdma_accept(). It is applicable to primary + * path only. The timeout will affect the local side of the QP, it is not + * negotiated with remote side and zero disables the timer. In case it is + * set before rdma_resolve_route, the value will also be used to determine + * PacketLifeTime for RoCE. + * + * Return: 0 for success + */ +int rdma_set_ack_timeout(struct rdma_cm_id *id, u8 timeout) +{ + struct rdma_id_private *id_priv; + + if (id->qp_type != IB_QPT_RC && id->qp_type != IB_QPT_XRC_INI) + return -EINVAL; + + id_priv = container_of(id, struct rdma_id_private, id); + mutex_lock(&id_priv->qp_mutex); + id_priv->timeout = timeout; + id_priv->timeout_set = true; + mutex_unlock(&id_priv->qp_mutex); + + return 0; +} +EXPORT_SYMBOL(rdma_set_ack_timeout); + +static void cma_query_handler(int status, struct sa_path_rec *path_rec, + void *context) +{ + struct cma_work *work = context; + struct rdma_route *route; + + route = &work->id->id.route; + + if (!status) { + route->num_paths = 1; + *route->path_rec = *path_rec; + } else { + work->old_state = RDMA_CM_ROUTE_QUERY; + work->new_state = RDMA_CM_ADDR_RESOLVED; + work->event.event = RDMA_CM_EVENT_ROUTE_ERROR; + work->event.status = status; + pr_debug_ratelimited("RDMA CM: ROUTE_ERROR: failed to query path. status %d\n", + status); + } + + queue_work(cma_wq, &work->work); +} + +static int cma_query_ib_route(struct rdma_id_private *id_priv, + unsigned long timeout_ms, struct cma_work *work) +{ + struct rdma_dev_addr *dev_addr = &id_priv->id.route.addr.dev_addr; + struct sa_path_rec path_rec; + ib_sa_comp_mask comp_mask; + struct sockaddr_in6 *sin6; + struct sockaddr_ib *sib; + + memset(&path_rec, 0, sizeof path_rec); + + if (rdma_cap_opa_ah(id_priv->id.device, id_priv->id.port_num)) + path_rec.rec_type = SA_PATH_REC_TYPE_OPA; + else + path_rec.rec_type = SA_PATH_REC_TYPE_IB; + rdma_addr_get_sgid(dev_addr, &path_rec.sgid); + rdma_addr_get_dgid(dev_addr, &path_rec.dgid); + path_rec.pkey = cpu_to_be16(ib_addr_get_pkey(dev_addr)); + path_rec.numb_path = 1; + path_rec.reversible = 1; + path_rec.service_id = rdma_get_service_id(&id_priv->id, + cma_dst_addr(id_priv)); + + comp_mask = IB_SA_PATH_REC_DGID | IB_SA_PATH_REC_SGID | + IB_SA_PATH_REC_PKEY | IB_SA_PATH_REC_NUMB_PATH | + IB_SA_PATH_REC_REVERSIBLE | IB_SA_PATH_REC_SERVICE_ID; + + switch (cma_family(id_priv)) { + case AF_INET: + path_rec.qos_class = cpu_to_be16((u16) id_priv->tos); + comp_mask |= IB_SA_PATH_REC_QOS_CLASS; + break; + case AF_INET6: + sin6 = (struct sockaddr_in6 *) cma_src_addr(id_priv); + path_rec.traffic_class = (u8) (be32_to_cpu(sin6->sin6_flowinfo) >> 20); + comp_mask |= IB_SA_PATH_REC_TRAFFIC_CLASS; + break; + case AF_IB: + sib = (struct sockaddr_ib *) cma_src_addr(id_priv); + path_rec.traffic_class = (u8) (be32_to_cpu(sib->sib_flowinfo) >> 20); + comp_mask |= IB_SA_PATH_REC_TRAFFIC_CLASS; + break; + } + + id_priv->query_id = ib_sa_path_rec_get(&sa_client, id_priv->id.device, + id_priv->id.port_num, &path_rec, + comp_mask, timeout_ms, + GFP_KERNEL, cma_query_handler, + work, &id_priv->query); + + return (id_priv->query_id < 0) ? id_priv->query_id : 0; +} + +static void cma_iboe_join_work_handler(struct work_struct *work) +{ + struct cma_multicast *mc = + container_of(work, struct cma_multicast, iboe_join.work); + struct rdma_cm_event *event = &mc->iboe_join.event; + struct rdma_id_private *id_priv = mc->id_priv; + int ret; + + mutex_lock(&id_priv->handler_mutex); + if (READ_ONCE(id_priv->state) == RDMA_CM_DESTROYING || + READ_ONCE(id_priv->state) == RDMA_CM_DEVICE_REMOVAL) + goto out_unlock; + + ret = cma_cm_event_handler(id_priv, event); + WARN_ON(ret); + +out_unlock: + mutex_unlock(&id_priv->handler_mutex); + if (event->event == RDMA_CM_EVENT_MULTICAST_JOIN) + rdma_destroy_ah_attr(&event->param.ud.ah_attr); +} + +static void cma_work_handler(struct work_struct *_work) +{ + struct cma_work *work = container_of(_work, struct cma_work, work); + struct rdma_id_private *id_priv = work->id; + + mutex_lock(&id_priv->handler_mutex); + if (READ_ONCE(id_priv->state) == RDMA_CM_DESTROYING || + READ_ONCE(id_priv->state) == RDMA_CM_DEVICE_REMOVAL) + goto out_unlock; + if (work->old_state != 0 || work->new_state != 0) { + if (!cma_comp_exch(id_priv, work->old_state, work->new_state)) + goto out_unlock; + } + + if (cma_cm_event_handler(id_priv, &work->event)) { + cma_id_put(id_priv); + destroy_id_handler_unlock(id_priv); + goto out_free; + } + +out_unlock: + mutex_unlock(&id_priv->handler_mutex); + cma_id_put(id_priv); +out_free: + if (work->event.event == RDMA_CM_EVENT_MULTICAST_JOIN) + rdma_destroy_ah_attr(&work->event.param.ud.ah_attr); + kfree(work); +} + +static void cma_init_resolve_route_work(struct cma_work *work, + struct rdma_id_private *id_priv) +{ + work->id = id_priv; + INIT_WORK(&work->work, cma_work_handler); + work->old_state = RDMA_CM_ROUTE_QUERY; + work->new_state = RDMA_CM_ROUTE_RESOLVED; + work->event.event = RDMA_CM_EVENT_ROUTE_RESOLVED; +} + +static void enqueue_resolve_addr_work(struct cma_work *work, + struct rdma_id_private *id_priv) +{ + /* Balances with cma_id_put() in cma_work_handler */ + cma_id_get(id_priv); + + work->id = id_priv; + INIT_WORK(&work->work, cma_work_handler); + work->old_state = RDMA_CM_ADDR_QUERY; + work->new_state = RDMA_CM_ADDR_RESOLVED; + work->event.event = RDMA_CM_EVENT_ADDR_RESOLVED; + + queue_work(cma_wq, &work->work); +} + +static int cma_resolve_ib_route(struct rdma_id_private *id_priv, + unsigned long timeout_ms) +{ + struct rdma_route *route = &id_priv->id.route; + struct cma_work *work; + int ret; + + work = kzalloc(sizeof *work, GFP_KERNEL); + if (!work) + return -ENOMEM; + + cma_init_resolve_route_work(work, id_priv); + + if (!route->path_rec) + route->path_rec = kmalloc(sizeof *route->path_rec, GFP_KERNEL); + if (!route->path_rec) { + ret = -ENOMEM; + goto err1; + } + + ret = cma_query_ib_route(id_priv, timeout_ms, work); + if (ret) + goto err2; + + return 0; +err2: + kfree(route->path_rec); + route->path_rec = NULL; +err1: + kfree(work); + return ret; +} + +static enum ib_gid_type cma_route_gid_type(enum rdma_network_type network_type, + unsigned long supported_gids, + enum ib_gid_type default_gid) +{ + if ((network_type == RDMA_NETWORK_IPV4 || + network_type == RDMA_NETWORK_IPV6) && + test_bit(IB_GID_TYPE_ROCE_UDP_ENCAP, &supported_gids)) + return IB_GID_TYPE_ROCE_UDP_ENCAP; + + return default_gid; +} + +/* + * cma_iboe_set_path_rec_l2_fields() is helper function which sets + * path record type based on GID type. + * It also sets up other L2 fields which includes destination mac address + * netdev ifindex, of the path record. + * It returns the netdev of the bound interface for this path record entry. + */ +static struct net_device * +cma_iboe_set_path_rec_l2_fields(struct rdma_id_private *id_priv) +{ + struct rdma_route *route = &id_priv->id.route; + enum ib_gid_type gid_type = IB_GID_TYPE_ROCE; + struct rdma_addr *addr = &route->addr; + unsigned long supported_gids; + struct net_device *ndev; + + if (!addr->dev_addr.bound_dev_if) + return NULL; + + ndev = dev_get_by_index(addr->dev_addr.net, + addr->dev_addr.bound_dev_if); + if (!ndev) + return NULL; + + supported_gids = roce_gid_type_mask_support(id_priv->id.device, + id_priv->id.port_num); + gid_type = cma_route_gid_type(addr->dev_addr.network, + supported_gids, + id_priv->gid_type); + /* Use the hint from IP Stack to select GID Type */ + if (gid_type < ib_network_to_gid_type(addr->dev_addr.network)) + gid_type = ib_network_to_gid_type(addr->dev_addr.network); + route->path_rec->rec_type = sa_conv_gid_to_pathrec_type(gid_type); + + route->path_rec->roce.route_resolved = true; + sa_path_set_dmac(route->path_rec, addr->dev_addr.dst_dev_addr); + return ndev; +} + +int rdma_set_ib_path(struct rdma_cm_id *id, + struct sa_path_rec *path_rec) +{ + struct rdma_id_private *id_priv; + struct net_device *ndev; + int ret; + + id_priv = container_of(id, struct rdma_id_private, id); + if (!cma_comp_exch(id_priv, RDMA_CM_ADDR_RESOLVED, + RDMA_CM_ROUTE_RESOLVED)) + return -EINVAL; + + id->route.path_rec = kmemdup(path_rec, sizeof(*path_rec), + GFP_KERNEL); + if (!id->route.path_rec) { + ret = -ENOMEM; + goto err; + } + + if (rdma_protocol_roce(id->device, id->port_num)) { + ndev = cma_iboe_set_path_rec_l2_fields(id_priv); + if (!ndev) { + ret = -ENODEV; + goto err_free; + } + dev_put(ndev); + } + + id->route.num_paths = 1; + return 0; + +err_free: + kfree(id->route.path_rec); + id->route.path_rec = NULL; +err: + cma_comp_exch(id_priv, RDMA_CM_ROUTE_RESOLVED, RDMA_CM_ADDR_RESOLVED); + return ret; +} +EXPORT_SYMBOL(rdma_set_ib_path); + +static int cma_resolve_iw_route(struct rdma_id_private *id_priv) +{ + struct cma_work *work; + + work = kzalloc(sizeof *work, GFP_KERNEL); + if (!work) + return -ENOMEM; + + cma_init_resolve_route_work(work, id_priv); + queue_work(cma_wq, &work->work); + return 0; +} + +static int get_vlan_ndev_tc(struct net_device *vlan_ndev, int prio) +{ + struct net_device *dev; + + dev = vlan_dev_real_dev(vlan_ndev); + if (dev->num_tc) + return netdev_get_prio_tc_map(dev, prio); + + return (vlan_dev_get_egress_qos_mask(vlan_ndev, prio) & + VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT; +} + +struct iboe_prio_tc_map { + int input_prio; + int output_tc; + bool found; +}; + +static int get_lower_vlan_dev_tc(struct net_device *dev, + struct netdev_nested_priv *priv) +{ + struct iboe_prio_tc_map *map = (struct iboe_prio_tc_map *)priv->data; + + if (is_vlan_dev(dev)) + map->output_tc = get_vlan_ndev_tc(dev, map->input_prio); + else if (dev->num_tc) + map->output_tc = netdev_get_prio_tc_map(dev, map->input_prio); + else + map->output_tc = 0; + /* We are interested only in first level VLAN device, so always + * return 1 to stop iterating over next level devices. + */ + map->found = true; + return 1; +} + +static int iboe_tos_to_sl(struct net_device *ndev, int tos) +{ + struct iboe_prio_tc_map prio_tc_map = {}; + int prio = rt_tos2priority(tos); + struct netdev_nested_priv priv; + + /* If VLAN device, get it directly from the VLAN netdev */ + if (is_vlan_dev(ndev)) + return get_vlan_ndev_tc(ndev, prio); + + prio_tc_map.input_prio = prio; + priv.data = (void *)&prio_tc_map; + rcu_read_lock(); + netdev_walk_all_lower_dev_rcu(ndev, + get_lower_vlan_dev_tc, + &priv); + rcu_read_unlock(); + /* If map is found from lower device, use it; Otherwise + * continue with the current netdevice to get priority to tc map. + */ + if (prio_tc_map.found) + return prio_tc_map.output_tc; + else if (ndev->num_tc) + return netdev_get_prio_tc_map(ndev, prio); + else + return 0; +} + +static __be32 cma_get_roce_udp_flow_label(struct rdma_id_private *id_priv) +{ + struct sockaddr_in6 *addr6; + u16 dport, sport; + u32 hash, fl; + + addr6 = (struct sockaddr_in6 *)cma_src_addr(id_priv); + fl = be32_to_cpu(addr6->sin6_flowinfo) & IB_GRH_FLOWLABEL_MASK; + if ((cma_family(id_priv) != AF_INET6) || !fl) { + dport = be16_to_cpu(cma_port(cma_dst_addr(id_priv))); + sport = be16_to_cpu(cma_port(cma_src_addr(id_priv))); + hash = (u32)sport * 31 + dport; + fl = hash & IB_GRH_FLOWLABEL_MASK; + } + + return cpu_to_be32(fl); +} + +static int cma_resolve_iboe_route(struct rdma_id_private *id_priv) +{ + struct rdma_route *route = &id_priv->id.route; + struct rdma_addr *addr = &route->addr; + struct cma_work *work; + int ret; + struct net_device *ndev; + + u8 default_roce_tos = id_priv->cma_dev->default_roce_tos[id_priv->id.port_num - + rdma_start_port(id_priv->cma_dev->device)]; + u8 tos; + + mutex_lock(&id_priv->qp_mutex); + tos = id_priv->tos_set ? id_priv->tos : default_roce_tos; + mutex_unlock(&id_priv->qp_mutex); + + work = kzalloc(sizeof *work, GFP_KERNEL); + if (!work) + return -ENOMEM; + + route->path_rec = kzalloc(sizeof *route->path_rec, GFP_KERNEL); + if (!route->path_rec) { + ret = -ENOMEM; + goto err1; + } + + route->num_paths = 1; + + ndev = cma_iboe_set_path_rec_l2_fields(id_priv); + if (!ndev) { + ret = -ENODEV; + goto err2; + } + + rdma_ip2gid((struct sockaddr *)&id_priv->id.route.addr.src_addr, + &route->path_rec->sgid); + rdma_ip2gid((struct sockaddr *)&id_priv->id.route.addr.dst_addr, + &route->path_rec->dgid); + + if (((struct sockaddr *)&id_priv->id.route.addr.dst_addr)->sa_family != AF_IB) + /* TODO: get the hoplimit from the inet/inet6 device */ + route->path_rec->hop_limit = addr->dev_addr.hoplimit; + else + route->path_rec->hop_limit = 1; + route->path_rec->reversible = 1; + route->path_rec->pkey = cpu_to_be16(0xffff); + route->path_rec->mtu_selector = IB_SA_EQ; + route->path_rec->sl = iboe_tos_to_sl(ndev, tos); + route->path_rec->traffic_class = tos; + route->path_rec->mtu = iboe_get_mtu(ndev->mtu); + route->path_rec->rate_selector = IB_SA_EQ; + route->path_rec->rate = IB_RATE_PORT_CURRENT; + dev_put(ndev); + route->path_rec->packet_life_time_selector = IB_SA_EQ; + /* In case ACK timeout is set, use this value to calculate + * PacketLifeTime. As per IBTA 12.7.34, + * local ACK timeout = (2 * PacketLifeTime + Local CA’s ACK delay). + * Assuming a negligible local ACK delay, we can use + * PacketLifeTime = local ACK timeout/2 + * as a reasonable approximation for RoCE networks. + */ + mutex_lock(&id_priv->qp_mutex); + if (id_priv->timeout_set && id_priv->timeout) + route->path_rec->packet_life_time = id_priv->timeout - 1; + else + route->path_rec->packet_life_time = CMA_IBOE_PACKET_LIFETIME; + mutex_unlock(&id_priv->qp_mutex); + + if (!route->path_rec->mtu) { + ret = -EINVAL; + goto err2; + } + + if (rdma_protocol_roce_udp_encap(id_priv->id.device, + id_priv->id.port_num)) + route->path_rec->flow_label = + cma_get_roce_udp_flow_label(id_priv); + + cma_init_resolve_route_work(work, id_priv); + queue_work(cma_wq, &work->work); + + return 0; + +err2: + kfree(route->path_rec); + route->path_rec = NULL; + route->num_paths = 0; +err1: + kfree(work); + return ret; +} + +int rdma_resolve_route(struct rdma_cm_id *id, unsigned long timeout_ms) +{ + struct rdma_id_private *id_priv; + int ret; + + id_priv = container_of(id, struct rdma_id_private, id); + if (!cma_comp_exch(id_priv, RDMA_CM_ADDR_RESOLVED, RDMA_CM_ROUTE_QUERY)) + return -EINVAL; + + cma_id_get(id_priv); + if (rdma_cap_ib_sa(id->device, id->port_num)) + ret = cma_resolve_ib_route(id_priv, timeout_ms); + else if (rdma_protocol_roce(id->device, id->port_num)) + ret = cma_resolve_iboe_route(id_priv); + else if (rdma_protocol_iwarp(id->device, id->port_num)) + ret = cma_resolve_iw_route(id_priv); + else + ret = -ENOSYS; + + if (ret) + goto err; + + return 0; +err: + cma_comp_exch(id_priv, RDMA_CM_ROUTE_QUERY, RDMA_CM_ADDR_RESOLVED); + cma_id_put(id_priv); + return ret; +} +EXPORT_SYMBOL(rdma_resolve_route); + +static void cma_set_loopback(struct sockaddr *addr) +{ + switch (addr->sa_family) { + case AF_INET: + ((struct sockaddr_in *) addr)->sin_addr.s_addr = htonl(INADDR_LOOPBACK); + break; + case AF_INET6: + ipv6_addr_set(&((struct sockaddr_in6 *) addr)->sin6_addr, + 0, 0, 0, htonl(1)); + break; + default: + ib_addr_set(&((struct sockaddr_ib *) addr)->sib_addr, + 0, 0, 0, htonl(1)); + break; + } +} + +static int cma_bind_loopback(struct rdma_id_private *id_priv) +{ + struct cma_device *cma_dev, *cur_dev; + union ib_gid gid; + enum ib_port_state port_state; + unsigned int p; + u16 pkey; + int ret; + + cma_dev = NULL; + mutex_lock(&lock); + list_for_each_entry(cur_dev, &dev_list, list) { + if (cma_family(id_priv) == AF_IB && + !rdma_cap_ib_cm(cur_dev->device, 1)) + continue; + + if (!cma_dev) + cma_dev = cur_dev; + + rdma_for_each_port (cur_dev->device, p) { + if (!ib_get_cached_port_state(cur_dev->device, p, &port_state) && + port_state == IB_PORT_ACTIVE) { + cma_dev = cur_dev; + goto port_found; + } + } + } + + if (!cma_dev) { + ret = -ENODEV; + goto out; + } + + p = 1; + +port_found: + ret = rdma_query_gid(cma_dev->device, p, 0, &gid); + if (ret) + goto out; + + ret = ib_get_cached_pkey(cma_dev->device, p, 0, &pkey); + if (ret) + goto out; + + id_priv->id.route.addr.dev_addr.dev_type = + (rdma_protocol_ib(cma_dev->device, p)) ? + ARPHRD_INFINIBAND : ARPHRD_ETHER; + + rdma_addr_set_sgid(&id_priv->id.route.addr.dev_addr, &gid); + ib_addr_set_pkey(&id_priv->id.route.addr.dev_addr, pkey); + id_priv->id.port_num = p; + cma_attach_to_dev(id_priv, cma_dev); + rdma_restrack_add(&id_priv->res); + cma_set_loopback(cma_src_addr(id_priv)); +out: + mutex_unlock(&lock); + return ret; +} + +static void addr_handler(int status, struct sockaddr *src_addr, + struct rdma_dev_addr *dev_addr, void *context) +{ + struct rdma_id_private *id_priv = context; + struct rdma_cm_event event = {}; + struct sockaddr *addr; + struct sockaddr_storage old_addr; + + mutex_lock(&id_priv->handler_mutex); + if (!cma_comp_exch(id_priv, RDMA_CM_ADDR_QUERY, + RDMA_CM_ADDR_RESOLVED)) + goto out; + + /* + * Store the previous src address, so that if we fail to acquire + * matching rdma device, old address can be restored back, which helps + * to cancel the cma listen operation correctly. + */ + addr = cma_src_addr(id_priv); + memcpy(&old_addr, addr, rdma_addr_size(addr)); + memcpy(addr, src_addr, rdma_addr_size(src_addr)); + if (!status && !id_priv->cma_dev) { + status = cma_acquire_dev_by_src_ip(id_priv); + if (status) + pr_debug_ratelimited("RDMA CM: ADDR_ERROR: failed to acquire device. status %d\n", + status); + rdma_restrack_add(&id_priv->res); + } else if (status) { + pr_debug_ratelimited("RDMA CM: ADDR_ERROR: failed to resolve IP. status %d\n", status); + } + + if (status) { + memcpy(addr, &old_addr, + rdma_addr_size((struct sockaddr *)&old_addr)); + if (!cma_comp_exch(id_priv, RDMA_CM_ADDR_RESOLVED, + RDMA_CM_ADDR_BOUND)) + goto out; + event.event = RDMA_CM_EVENT_ADDR_ERROR; + event.status = status; + } else + event.event = RDMA_CM_EVENT_ADDR_RESOLVED; + + if (cma_cm_event_handler(id_priv, &event)) { + destroy_id_handler_unlock(id_priv); + return; + } +out: + mutex_unlock(&id_priv->handler_mutex); +} + +static int cma_resolve_loopback(struct rdma_id_private *id_priv) +{ + struct cma_work *work; + union ib_gid gid; + int ret; + + work = kzalloc(sizeof *work, GFP_KERNEL); + if (!work) + return -ENOMEM; + + if (!id_priv->cma_dev) { + ret = cma_bind_loopback(id_priv); + if (ret) + goto err; + } + + rdma_addr_get_sgid(&id_priv->id.route.addr.dev_addr, &gid); + rdma_addr_set_dgid(&id_priv->id.route.addr.dev_addr, &gid); + + enqueue_resolve_addr_work(work, id_priv); + return 0; +err: + kfree(work); + return ret; +} + +static int cma_resolve_ib_addr(struct rdma_id_private *id_priv) +{ + struct cma_work *work; + int ret; + + work = kzalloc(sizeof *work, GFP_KERNEL); + if (!work) + return -ENOMEM; + + if (!id_priv->cma_dev) { + ret = cma_resolve_ib_dev(id_priv); + if (ret) + goto err; + } + + rdma_addr_set_dgid(&id_priv->id.route.addr.dev_addr, (union ib_gid *) + &(((struct sockaddr_ib *) &id_priv->id.route.addr.dst_addr)->sib_addr)); + + enqueue_resolve_addr_work(work, id_priv); + return 0; +err: + kfree(work); + return ret; +} + +static int cma_bind_addr(struct rdma_cm_id *id, struct sockaddr *src_addr, + const struct sockaddr *dst_addr) +{ + struct sockaddr_storage zero_sock = {}; + + if (src_addr && src_addr->sa_family) + return rdma_bind_addr(id, src_addr); + + /* + * When the src_addr is not specified, automatically supply an any addr + */ + zero_sock.ss_family = dst_addr->sa_family; + if (IS_ENABLED(CONFIG_IPV6) && dst_addr->sa_family == AF_INET6) { + struct sockaddr_in6 *src_addr6 = + (struct sockaddr_in6 *)&zero_sock; + struct sockaddr_in6 *dst_addr6 = + (struct sockaddr_in6 *)dst_addr; + + src_addr6->sin6_scope_id = dst_addr6->sin6_scope_id; + if (ipv6_addr_type(&dst_addr6->sin6_addr) & IPV6_ADDR_LINKLOCAL) + id->route.addr.dev_addr.bound_dev_if = + dst_addr6->sin6_scope_id; + } else if (dst_addr->sa_family == AF_IB) { + ((struct sockaddr_ib *)&zero_sock)->sib_pkey = + ((struct sockaddr_ib *)dst_addr)->sib_pkey; + } + return rdma_bind_addr(id, (struct sockaddr *)&zero_sock); +} + +/* + * If required, resolve the source address for bind and leave the id_priv in + * state RDMA_CM_ADDR_BOUND. This oddly uses the state to determine the prior + * calls made by ULP, a previously bound ID will not be re-bound and src_addr is + * ignored. + */ +static int resolve_prepare_src(struct rdma_id_private *id_priv, + struct sockaddr *src_addr, + const struct sockaddr *dst_addr) +{ + int ret; + + memcpy(cma_dst_addr(id_priv), dst_addr, rdma_addr_size(dst_addr)); + if (!cma_comp_exch(id_priv, RDMA_CM_ADDR_BOUND, RDMA_CM_ADDR_QUERY)) { + /* For a well behaved ULP state will be RDMA_CM_IDLE */ + ret = cma_bind_addr(&id_priv->id, src_addr, dst_addr); + if (ret) + goto err_dst; + if (WARN_ON(!cma_comp_exch(id_priv, RDMA_CM_ADDR_BOUND, + RDMA_CM_ADDR_QUERY))) { + ret = -EINVAL; + goto err_dst; + } + } + + if (cma_family(id_priv) != dst_addr->sa_family) { + ret = -EINVAL; + goto err_state; + } + return 0; + +err_state: + cma_comp_exch(id_priv, RDMA_CM_ADDR_QUERY, RDMA_CM_ADDR_BOUND); +err_dst: + memset(cma_dst_addr(id_priv), 0, rdma_addr_size(dst_addr)); + return ret; +} + +int rdma_resolve_addr(struct rdma_cm_id *id, struct sockaddr *src_addr, + const struct sockaddr *dst_addr, unsigned long timeout_ms) +{ + struct rdma_id_private *id_priv = + container_of(id, struct rdma_id_private, id); + int ret; + + ret = resolve_prepare_src(id_priv, src_addr, dst_addr); + if (ret) + return ret; + + if (cma_any_addr(dst_addr)) { + ret = cma_resolve_loopback(id_priv); + } else { + if (dst_addr->sa_family == AF_IB) { + ret = cma_resolve_ib_addr(id_priv); + } else { + /* + * The FSM can return back to RDMA_CM_ADDR_BOUND after + * rdma_resolve_ip() is called, eg through the error + * path in addr_handler(). If this happens the existing + * request must be canceled before issuing a new one. + * Since canceling a request is a bit slow and this + * oddball path is rare, keep track once a request has + * been issued. The track turns out to be a permanent + * state since this is the only cancel as it is + * immediately before rdma_resolve_ip(). + */ + if (id_priv->used_resolve_ip) + rdma_addr_cancel(&id->route.addr.dev_addr); + else + id_priv->used_resolve_ip = 1; + ret = rdma_resolve_ip(cma_src_addr(id_priv), dst_addr, + &id->route.addr.dev_addr, + timeout_ms, addr_handler, + false, id_priv); + } + } + if (ret) + goto err; + + return 0; +err: + cma_comp_exch(id_priv, RDMA_CM_ADDR_QUERY, RDMA_CM_ADDR_BOUND); + return ret; +} +EXPORT_SYMBOL(rdma_resolve_addr); + +int rdma_set_reuseaddr(struct rdma_cm_id *id, int reuse) +{ + struct rdma_id_private *id_priv; + unsigned long flags; + int ret; + + id_priv = container_of(id, struct rdma_id_private, id); + spin_lock_irqsave(&id_priv->lock, flags); + if ((reuse && id_priv->state != RDMA_CM_LISTEN) || + id_priv->state == RDMA_CM_IDLE) { + id_priv->reuseaddr = reuse; + ret = 0; + } else { + ret = -EINVAL; + } + spin_unlock_irqrestore(&id_priv->lock, flags); + return ret; +} +EXPORT_SYMBOL(rdma_set_reuseaddr); + +int rdma_set_afonly(struct rdma_cm_id *id, int afonly) +{ + struct rdma_id_private *id_priv; + unsigned long flags; + int ret; + + id_priv = container_of(id, struct rdma_id_private, id); + spin_lock_irqsave(&id_priv->lock, flags); + if (id_priv->state == RDMA_CM_IDLE || id_priv->state == RDMA_CM_ADDR_BOUND) { + id_priv->options |= (1 << CMA_OPTION_AFONLY); + id_priv->afonly = afonly; + ret = 0; + } else { + ret = -EINVAL; + } + spin_unlock_irqrestore(&id_priv->lock, flags); + return ret; +} +EXPORT_SYMBOL(rdma_set_afonly); + +static void cma_bind_port(struct rdma_bind_list *bind_list, + struct rdma_id_private *id_priv) +{ + struct sockaddr *addr; + struct sockaddr_ib *sib; + u64 sid, mask; + __be16 port; + + lockdep_assert_held(&lock); + + addr = cma_src_addr(id_priv); + port = htons(bind_list->port); + + switch (addr->sa_family) { + case AF_INET: + ((struct sockaddr_in *) addr)->sin_port = port; + break; + case AF_INET6: + ((struct sockaddr_in6 *) addr)->sin6_port = port; + break; + case AF_IB: + sib = (struct sockaddr_ib *) addr; + sid = be64_to_cpu(sib->sib_sid); + mask = be64_to_cpu(sib->sib_sid_mask); + sib->sib_sid = cpu_to_be64((sid & mask) | (u64) ntohs(port)); + sib->sib_sid_mask = cpu_to_be64(~0ULL); + break; + } + id_priv->bind_list = bind_list; + hlist_add_head(&id_priv->node, &bind_list->owners); +} + +static int cma_alloc_port(enum rdma_ucm_port_space ps, + struct rdma_id_private *id_priv, unsigned short snum) +{ + struct rdma_bind_list *bind_list; + int ret; + + lockdep_assert_held(&lock); + + bind_list = kzalloc(sizeof *bind_list, GFP_KERNEL); + if (!bind_list) + return -ENOMEM; + + ret = cma_ps_alloc(id_priv->id.route.addr.dev_addr.net, ps, bind_list, + snum); + if (ret < 0) + goto err; + + bind_list->ps = ps; + bind_list->port = snum; + cma_bind_port(bind_list, id_priv); + return 0; +err: + kfree(bind_list); + return ret == -ENOSPC ? -EADDRNOTAVAIL : ret; +} + +static int cma_port_is_unique(struct rdma_bind_list *bind_list, + struct rdma_id_private *id_priv) +{ + struct rdma_id_private *cur_id; + struct sockaddr *daddr = cma_dst_addr(id_priv); + struct sockaddr *saddr = cma_src_addr(id_priv); + __be16 dport = cma_port(daddr); + + lockdep_assert_held(&lock); + + hlist_for_each_entry(cur_id, &bind_list->owners, node) { + struct sockaddr *cur_daddr = cma_dst_addr(cur_id); + struct sockaddr *cur_saddr = cma_src_addr(cur_id); + __be16 cur_dport = cma_port(cur_daddr); + + if (id_priv == cur_id) + continue; + + /* different dest port -> unique */ + if (!cma_any_port(daddr) && + !cma_any_port(cur_daddr) && + (dport != cur_dport)) + continue; + + /* different src address -> unique */ + if (!cma_any_addr(saddr) && + !cma_any_addr(cur_saddr) && + cma_addr_cmp(saddr, cur_saddr)) + continue; + + /* different dst address -> unique */ + if (!cma_any_addr(daddr) && + !cma_any_addr(cur_daddr) && + cma_addr_cmp(daddr, cur_daddr)) + continue; + + return -EADDRNOTAVAIL; + } + return 0; +} + +static int cma_alloc_any_port(enum rdma_ucm_port_space ps, + struct rdma_id_private *id_priv) +{ + static unsigned int last_used_port; + int low, high, remaining; + unsigned int rover; + struct net *net = id_priv->id.route.addr.dev_addr.net; + + lockdep_assert_held(&lock); + + inet_get_local_port_range(net, &low, &high); + remaining = (high - low) + 1; + rover = prandom_u32() % remaining + low; +retry: + if (last_used_port != rover) { + struct rdma_bind_list *bind_list; + int ret; + + bind_list = cma_ps_find(net, ps, (unsigned short)rover); + + if (!bind_list) { + ret = cma_alloc_port(ps, id_priv, rover); + } else { + ret = cma_port_is_unique(bind_list, id_priv); + if (!ret) + cma_bind_port(bind_list, id_priv); + } + /* + * Remember previously used port number in order to avoid + * re-using same port immediately after it is closed. + */ + if (!ret) + last_used_port = rover; + if (ret != -EADDRNOTAVAIL) + return ret; + } + if (--remaining) { + rover++; + if ((rover < low) || (rover > high)) + rover = low; + goto retry; + } + return -EADDRNOTAVAIL; +} + +/* + * Check that the requested port is available. This is called when trying to + * bind to a specific port, or when trying to listen on a bound port. In + * the latter case, the provided id_priv may already be on the bind_list, but + * we still need to check that it's okay to start listening. + */ +static int cma_check_port(struct rdma_bind_list *bind_list, + struct rdma_id_private *id_priv, uint8_t reuseaddr) +{ + struct rdma_id_private *cur_id; + struct sockaddr *addr, *cur_addr; + + lockdep_assert_held(&lock); + + addr = cma_src_addr(id_priv); + hlist_for_each_entry(cur_id, &bind_list->owners, node) { + if (id_priv == cur_id) + continue; + + if (reuseaddr && cur_id->reuseaddr) + continue; + + cur_addr = cma_src_addr(cur_id); + if (id_priv->afonly && cur_id->afonly && + (addr->sa_family != cur_addr->sa_family)) + continue; + + if (cma_any_addr(addr) || cma_any_addr(cur_addr)) + return -EADDRNOTAVAIL; + + if (!cma_addr_cmp(addr, cur_addr)) + return -EADDRINUSE; + } + return 0; +} + +static int cma_use_port(enum rdma_ucm_port_space ps, + struct rdma_id_private *id_priv) +{ + struct rdma_bind_list *bind_list; + unsigned short snum; + int ret; + + lockdep_assert_held(&lock); + + snum = ntohs(cma_port(cma_src_addr(id_priv))); + if (snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE)) + return -EACCES; + + bind_list = cma_ps_find(id_priv->id.route.addr.dev_addr.net, ps, snum); + if (!bind_list) { + ret = cma_alloc_port(ps, id_priv, snum); + } else { + ret = cma_check_port(bind_list, id_priv, id_priv->reuseaddr); + if (!ret) + cma_bind_port(bind_list, id_priv); + } + return ret; +} + +static enum rdma_ucm_port_space +cma_select_inet_ps(struct rdma_id_private *id_priv) +{ + switch (id_priv->id.ps) { + case RDMA_PS_TCP: + case RDMA_PS_UDP: + case RDMA_PS_IPOIB: + case RDMA_PS_IB: + return id_priv->id.ps; + default: + + return 0; + } +} + +static enum rdma_ucm_port_space +cma_select_ib_ps(struct rdma_id_private *id_priv) +{ + enum rdma_ucm_port_space ps = 0; + struct sockaddr_ib *sib; + u64 sid_ps, mask, sid; + + sib = (struct sockaddr_ib *) cma_src_addr(id_priv); + mask = be64_to_cpu(sib->sib_sid_mask) & RDMA_IB_IP_PS_MASK; + sid = be64_to_cpu(sib->sib_sid) & mask; + + if ((id_priv->id.ps == RDMA_PS_IB) && (sid == (RDMA_IB_IP_PS_IB & mask))) { + sid_ps = RDMA_IB_IP_PS_IB; + ps = RDMA_PS_IB; + } else if (((id_priv->id.ps == RDMA_PS_IB) || (id_priv->id.ps == RDMA_PS_TCP)) && + (sid == (RDMA_IB_IP_PS_TCP & mask))) { + sid_ps = RDMA_IB_IP_PS_TCP; + ps = RDMA_PS_TCP; + } else if (((id_priv->id.ps == RDMA_PS_IB) || (id_priv->id.ps == RDMA_PS_UDP)) && + (sid == (RDMA_IB_IP_PS_UDP & mask))) { + sid_ps = RDMA_IB_IP_PS_UDP; + ps = RDMA_PS_UDP; + } + + if (ps) { + sib->sib_sid = cpu_to_be64(sid_ps | ntohs(cma_port((struct sockaddr *) sib))); + sib->sib_sid_mask = cpu_to_be64(RDMA_IB_IP_PS_MASK | + be64_to_cpu(sib->sib_sid_mask)); + } + return ps; +} + +static int cma_get_port(struct rdma_id_private *id_priv) +{ + enum rdma_ucm_port_space ps; + int ret; + + if (cma_family(id_priv) != AF_IB) + ps = cma_select_inet_ps(id_priv); + else + ps = cma_select_ib_ps(id_priv); + if (!ps) + return -EPROTONOSUPPORT; + + mutex_lock(&lock); + if (cma_any_port(cma_src_addr(id_priv))) + ret = cma_alloc_any_port(ps, id_priv); + else + ret = cma_use_port(ps, id_priv); + mutex_unlock(&lock); + + return ret; +} + +static int cma_check_linklocal(struct rdma_dev_addr *dev_addr, + struct sockaddr *addr) +{ +#if IS_ENABLED(CONFIG_IPV6) + struct sockaddr_in6 *sin6; + + if (addr->sa_family != AF_INET6) + return 0; + + sin6 = (struct sockaddr_in6 *) addr; + + if (!(ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)) + return 0; + + if (!sin6->sin6_scope_id) + return -EINVAL; + + dev_addr->bound_dev_if = sin6->sin6_scope_id; +#endif + return 0; +} + +int rdma_listen(struct rdma_cm_id *id, int backlog) +{ + struct rdma_id_private *id_priv = + container_of(id, struct rdma_id_private, id); + int ret; + + if (!cma_comp_exch(id_priv, RDMA_CM_ADDR_BOUND, RDMA_CM_LISTEN)) { + struct sockaddr_in any_in = { + .sin_family = AF_INET, + .sin_addr.s_addr = htonl(INADDR_ANY), + }; + + /* For a well behaved ULP state will be RDMA_CM_IDLE */ + ret = rdma_bind_addr(id, (struct sockaddr *)&any_in); + if (ret) + return ret; + if (WARN_ON(!cma_comp_exch(id_priv, RDMA_CM_ADDR_BOUND, + RDMA_CM_LISTEN))) + return -EINVAL; + } + + /* + * Once the ID reaches RDMA_CM_LISTEN it is not allowed to be reusable + * any more, and has to be unique in the bind list. + */ + if (id_priv->reuseaddr) { + mutex_lock(&lock); + ret = cma_check_port(id_priv->bind_list, id_priv, 0); + if (!ret) + id_priv->reuseaddr = 0; + mutex_unlock(&lock); + if (ret) + goto err; + } + + id_priv->backlog = backlog; + if (id_priv->cma_dev) { + if (rdma_cap_ib_cm(id->device, 1)) { + ret = cma_ib_listen(id_priv); + if (ret) + goto err; + } else if (rdma_cap_iw_cm(id->device, 1)) { + ret = cma_iw_listen(id_priv, backlog); + if (ret) + goto err; + } else { + ret = -ENOSYS; + goto err; + } + } else { + ret = cma_listen_on_all(id_priv); + if (ret) + goto err; + } + + return 0; +err: + id_priv->backlog = 0; + /* + * All the failure paths that lead here will not allow the req_handler's + * to have run. + */ + cma_comp_exch(id_priv, RDMA_CM_LISTEN, RDMA_CM_ADDR_BOUND); + return ret; +} +EXPORT_SYMBOL(rdma_listen); + +int rdma_bind_addr(struct rdma_cm_id *id, struct sockaddr *addr) +{ + struct rdma_id_private *id_priv; + int ret; + struct sockaddr *daddr; + + if (addr->sa_family != AF_INET && addr->sa_family != AF_INET6 && + addr->sa_family != AF_IB) + return -EAFNOSUPPORT; + + id_priv = container_of(id, struct rdma_id_private, id); + if (!cma_comp_exch(id_priv, RDMA_CM_IDLE, RDMA_CM_ADDR_BOUND)) + return -EINVAL; + + ret = cma_check_linklocal(&id->route.addr.dev_addr, addr); + if (ret) + goto err1; + + memcpy(cma_src_addr(id_priv), addr, rdma_addr_size(addr)); + if (!cma_any_addr(addr)) { + ret = cma_translate_addr(addr, &id->route.addr.dev_addr); + if (ret) + goto err1; + + ret = cma_acquire_dev_by_src_ip(id_priv); + if (ret) + goto err1; + } + + if (!(id_priv->options & (1 << CMA_OPTION_AFONLY))) { + if (addr->sa_family == AF_INET) + id_priv->afonly = 1; +#if IS_ENABLED(CONFIG_IPV6) + else if (addr->sa_family == AF_INET6) { + struct net *net = id_priv->id.route.addr.dev_addr.net; + + id_priv->afonly = net->ipv6.sysctl.bindv6only; + } +#endif + } + daddr = cma_dst_addr(id_priv); + daddr->sa_family = addr->sa_family; + + ret = cma_get_port(id_priv); + if (ret) + goto err2; + + if (!cma_any_addr(addr)) + rdma_restrack_add(&id_priv->res); + return 0; +err2: + if (id_priv->cma_dev) + cma_release_dev(id_priv); +err1: + cma_comp_exch(id_priv, RDMA_CM_ADDR_BOUND, RDMA_CM_IDLE); + return ret; +} +EXPORT_SYMBOL(rdma_bind_addr); + +static int cma_format_hdr(void *hdr, struct rdma_id_private *id_priv) +{ + struct cma_hdr *cma_hdr; + + cma_hdr = hdr; + cma_hdr->cma_version = CMA_VERSION; + if (cma_family(id_priv) == AF_INET) { + struct sockaddr_in *src4, *dst4; + + src4 = (struct sockaddr_in *) cma_src_addr(id_priv); + dst4 = (struct sockaddr_in *) cma_dst_addr(id_priv); + + cma_set_ip_ver(cma_hdr, 4); + cma_hdr->src_addr.ip4.addr = src4->sin_addr.s_addr; + cma_hdr->dst_addr.ip4.addr = dst4->sin_addr.s_addr; + cma_hdr->port = src4->sin_port; + } else if (cma_family(id_priv) == AF_INET6) { + struct sockaddr_in6 *src6, *dst6; + + src6 = (struct sockaddr_in6 *) cma_src_addr(id_priv); + dst6 = (struct sockaddr_in6 *) cma_dst_addr(id_priv); + + cma_set_ip_ver(cma_hdr, 6); + cma_hdr->src_addr.ip6 = src6->sin6_addr; + cma_hdr->dst_addr.ip6 = dst6->sin6_addr; + cma_hdr->port = src6->sin6_port; + } + return 0; +} + +static int cma_sidr_rep_handler(struct ib_cm_id *cm_id, + const struct ib_cm_event *ib_event) +{ + struct rdma_id_private *id_priv = cm_id->context; + struct rdma_cm_event event = {}; + const struct ib_cm_sidr_rep_event_param *rep = + &ib_event->param.sidr_rep_rcvd; + int ret; + + mutex_lock(&id_priv->handler_mutex); + if (READ_ONCE(id_priv->state) != RDMA_CM_CONNECT) + goto out; + + switch (ib_event->event) { + case IB_CM_SIDR_REQ_ERROR: + event.event = RDMA_CM_EVENT_UNREACHABLE; + event.status = -ETIMEDOUT; + break; + case IB_CM_SIDR_REP_RECEIVED: + event.param.ud.private_data = ib_event->private_data; + event.param.ud.private_data_len = IB_CM_SIDR_REP_PRIVATE_DATA_SIZE; + if (rep->status != IB_SIDR_SUCCESS) { + event.event = RDMA_CM_EVENT_UNREACHABLE; + event.status = ib_event->param.sidr_rep_rcvd.status; + pr_debug_ratelimited("RDMA CM: UNREACHABLE: bad SIDR reply. status %d\n", + event.status); + break; + } + ret = cma_set_qkey(id_priv, rep->qkey); + if (ret) { + pr_debug_ratelimited("RDMA CM: ADDR_ERROR: failed to set qkey. status %d\n", ret); + event.event = RDMA_CM_EVENT_ADDR_ERROR; + event.status = ret; + break; + } + ib_init_ah_attr_from_path(id_priv->id.device, + id_priv->id.port_num, + id_priv->id.route.path_rec, + &event.param.ud.ah_attr, + rep->sgid_attr); + event.param.ud.qp_num = rep->qpn; + event.param.ud.qkey = rep->qkey; + event.event = RDMA_CM_EVENT_ESTABLISHED; + event.status = 0; + break; + default: + pr_err("RDMA CMA: unexpected IB CM event: %d\n", + ib_event->event); + goto out; + } + + ret = cma_cm_event_handler(id_priv, &event); + + rdma_destroy_ah_attr(&event.param.ud.ah_attr); + if (ret) { + /* Destroy the CM ID by returning a non-zero value. */ + id_priv->cm_id.ib = NULL; + destroy_id_handler_unlock(id_priv); + return ret; + } +out: + mutex_unlock(&id_priv->handler_mutex); + return 0; +} + +static int cma_resolve_ib_udp(struct rdma_id_private *id_priv, + struct rdma_conn_param *conn_param) +{ + struct ib_cm_sidr_req_param req; + struct ib_cm_id *id; + void *private_data; + u8 offset; + int ret; + + memset(&req, 0, sizeof req); + offset = cma_user_data_offset(id_priv); + req.private_data_len = offset + conn_param->private_data_len; + if (req.private_data_len < conn_param->private_data_len) + return -EINVAL; + + if (req.private_data_len) { + private_data = kzalloc(req.private_data_len, GFP_ATOMIC); + if (!private_data) + return -ENOMEM; + } else { + private_data = NULL; + } + + if (conn_param->private_data && conn_param->private_data_len) + memcpy(private_data + offset, conn_param->private_data, + conn_param->private_data_len); + + if (private_data) { + ret = cma_format_hdr(private_data, id_priv); + if (ret) + goto out; + req.private_data = private_data; + } + + id = ib_create_cm_id(id_priv->id.device, cma_sidr_rep_handler, + id_priv); + if (IS_ERR(id)) { + ret = PTR_ERR(id); + goto out; + } + id_priv->cm_id.ib = id; + + req.path = id_priv->id.route.path_rec; + req.sgid_attr = id_priv->id.route.addr.dev_addr.sgid_attr; + req.service_id = rdma_get_service_id(&id_priv->id, cma_dst_addr(id_priv)); + req.timeout_ms = 1 << (CMA_CM_RESPONSE_TIMEOUT - 8); + req.max_cm_retries = CMA_MAX_CM_RETRIES; + + trace_cm_send_sidr_req(id_priv); + ret = ib_send_cm_sidr_req(id_priv->cm_id.ib, &req); + if (ret) { + ib_destroy_cm_id(id_priv->cm_id.ib); + id_priv->cm_id.ib = NULL; + } +out: + kfree(private_data); + return ret; +} + +static int cma_connect_ib(struct rdma_id_private *id_priv, + struct rdma_conn_param *conn_param) +{ + struct ib_cm_req_param req; + struct rdma_route *route; + void *private_data; + struct ib_cm_id *id; + u8 offset; + int ret; + + memset(&req, 0, sizeof req); + offset = cma_user_data_offset(id_priv); + req.private_data_len = offset + conn_param->private_data_len; + if (req.private_data_len < conn_param->private_data_len) + return -EINVAL; + + if (req.private_data_len) { + private_data = kzalloc(req.private_data_len, GFP_ATOMIC); + if (!private_data) + return -ENOMEM; + } else { + private_data = NULL; + } + + if (conn_param->private_data && conn_param->private_data_len) + memcpy(private_data + offset, conn_param->private_data, + conn_param->private_data_len); + + id = ib_create_cm_id(id_priv->id.device, cma_ib_handler, id_priv); + if (IS_ERR(id)) { + ret = PTR_ERR(id); + goto out; + } + id_priv->cm_id.ib = id; + + route = &id_priv->id.route; + if (private_data) { + ret = cma_format_hdr(private_data, id_priv); + if (ret) + goto out; + req.private_data = private_data; + } + + req.primary_path = &route->path_rec[0]; + if (route->num_paths == 2) + req.alternate_path = &route->path_rec[1]; + + req.ppath_sgid_attr = id_priv->id.route.addr.dev_addr.sgid_attr; + /* Alternate path SGID attribute currently unsupported */ + req.service_id = rdma_get_service_id(&id_priv->id, cma_dst_addr(id_priv)); + req.qp_num = id_priv->qp_num; + req.qp_type = id_priv->id.qp_type; + req.starting_psn = id_priv->seq_num; + req.responder_resources = conn_param->responder_resources; + req.initiator_depth = conn_param->initiator_depth; + req.flow_control = conn_param->flow_control; + req.retry_count = min_t(u8, 7, conn_param->retry_count); + req.rnr_retry_count = min_t(u8, 7, conn_param->rnr_retry_count); + req.remote_cm_response_timeout = CMA_CM_RESPONSE_TIMEOUT; + req.local_cm_response_timeout = CMA_CM_RESPONSE_TIMEOUT; + req.max_cm_retries = CMA_MAX_CM_RETRIES; + req.srq = id_priv->srq ? 1 : 0; + req.ece.vendor_id = id_priv->ece.vendor_id; + req.ece.attr_mod = id_priv->ece.attr_mod; + + trace_cm_send_req(id_priv); + ret = ib_send_cm_req(id_priv->cm_id.ib, &req); +out: + if (ret && !IS_ERR(id)) { + ib_destroy_cm_id(id); + id_priv->cm_id.ib = NULL; + } + + kfree(private_data); + return ret; +} + +static int cma_connect_iw(struct rdma_id_private *id_priv, + struct rdma_conn_param *conn_param) +{ + struct iw_cm_id *cm_id; + int ret; + struct iw_cm_conn_param iw_param; + + cm_id = iw_create_cm_id(id_priv->id.device, cma_iw_handler, id_priv); + if (IS_ERR(cm_id)) + return PTR_ERR(cm_id); + + mutex_lock(&id_priv->qp_mutex); + cm_id->tos = id_priv->tos; + cm_id->tos_set = id_priv->tos_set; + mutex_unlock(&id_priv->qp_mutex); + + id_priv->cm_id.iw = cm_id; + + memcpy(&cm_id->local_addr, cma_src_addr(id_priv), + rdma_addr_size(cma_src_addr(id_priv))); + memcpy(&cm_id->remote_addr, cma_dst_addr(id_priv), + rdma_addr_size(cma_dst_addr(id_priv))); + + ret = cma_modify_qp_rtr(id_priv, conn_param); + if (ret) + goto out; + + if (conn_param) { + iw_param.ord = conn_param->initiator_depth; + iw_param.ird = conn_param->responder_resources; + iw_param.private_data = conn_param->private_data; + iw_param.private_data_len = conn_param->private_data_len; + iw_param.qpn = id_priv->id.qp ? id_priv->qp_num : conn_param->qp_num; + } else { + memset(&iw_param, 0, sizeof iw_param); + iw_param.qpn = id_priv->qp_num; + } + ret = iw_cm_connect(cm_id, &iw_param); +out: + if (ret) { + iw_destroy_cm_id(cm_id); + id_priv->cm_id.iw = NULL; + } + return ret; +} + +/** + * rdma_connect_locked - Initiate an active connection request. + * @id: Connection identifier to connect. + * @conn_param: Connection information used for connected QPs. + * + * Same as rdma_connect() but can only be called from the + * RDMA_CM_EVENT_ROUTE_RESOLVED handler callback. + */ +int rdma_connect_locked(struct rdma_cm_id *id, + struct rdma_conn_param *conn_param) +{ + struct rdma_id_private *id_priv = + container_of(id, struct rdma_id_private, id); + int ret; + + if (!cma_comp_exch(id_priv, RDMA_CM_ROUTE_RESOLVED, RDMA_CM_CONNECT)) + return -EINVAL; + + if (!id->qp) { + id_priv->qp_num = conn_param->qp_num; + id_priv->srq = conn_param->srq; + } + + if (rdma_cap_ib_cm(id->device, id->port_num)) { + if (id->qp_type == IB_QPT_UD) + ret = cma_resolve_ib_udp(id_priv, conn_param); + else + ret = cma_connect_ib(id_priv, conn_param); + } else if (rdma_cap_iw_cm(id->device, id->port_num)) + ret = cma_connect_iw(id_priv, conn_param); + else + ret = -ENOSYS; + if (ret) + goto err_state; + return 0; +err_state: + cma_comp_exch(id_priv, RDMA_CM_CONNECT, RDMA_CM_ROUTE_RESOLVED); + return ret; +} +EXPORT_SYMBOL(rdma_connect_locked); + +/** + * rdma_connect - Initiate an active connection request. + * @id: Connection identifier to connect. + * @conn_param: Connection information used for connected QPs. + * + * Users must have resolved a route for the rdma_cm_id to connect with by having + * called rdma_resolve_route before calling this routine. + * + * This call will either connect to a remote QP or obtain remote QP information + * for unconnected rdma_cm_id's. The actual operation is based on the + * rdma_cm_id's port space. + */ +int rdma_connect(struct rdma_cm_id *id, struct rdma_conn_param *conn_param) +{ + struct rdma_id_private *id_priv = + container_of(id, struct rdma_id_private, id); + int ret; + + mutex_lock(&id_priv->handler_mutex); + ret = rdma_connect_locked(id, conn_param); + mutex_unlock(&id_priv->handler_mutex); + return ret; +} +EXPORT_SYMBOL(rdma_connect); + +/** + * rdma_connect_ece - Initiate an active connection request with ECE data. + * @id: Connection identifier to connect. + * @conn_param: Connection information used for connected QPs. + * @ece: ECE parameters + * + * See rdma_connect() explanation. + */ +int rdma_connect_ece(struct rdma_cm_id *id, struct rdma_conn_param *conn_param, + struct rdma_ucm_ece *ece) +{ + struct rdma_id_private *id_priv = + container_of(id, struct rdma_id_private, id); + + id_priv->ece.vendor_id = ece->vendor_id; + id_priv->ece.attr_mod = ece->attr_mod; + + return rdma_connect(id, conn_param); +} +EXPORT_SYMBOL(rdma_connect_ece); + +static int cma_accept_ib(struct rdma_id_private *id_priv, + struct rdma_conn_param *conn_param) +{ + struct ib_cm_rep_param rep; + int ret; + + ret = cma_modify_qp_rtr(id_priv, conn_param); + if (ret) + goto out; + + ret = cma_modify_qp_rts(id_priv, conn_param); + if (ret) + goto out; + + memset(&rep, 0, sizeof rep); + rep.qp_num = id_priv->qp_num; + rep.starting_psn = id_priv->seq_num; + rep.private_data = conn_param->private_data; + rep.private_data_len = conn_param->private_data_len; + rep.responder_resources = conn_param->responder_resources; + rep.initiator_depth = conn_param->initiator_depth; + rep.failover_accepted = 0; + rep.flow_control = conn_param->flow_control; + rep.rnr_retry_count = min_t(u8, 7, conn_param->rnr_retry_count); + rep.srq = id_priv->srq ? 1 : 0; + rep.ece.vendor_id = id_priv->ece.vendor_id; + rep.ece.attr_mod = id_priv->ece.attr_mod; + + trace_cm_send_rep(id_priv); + ret = ib_send_cm_rep(id_priv->cm_id.ib, &rep); +out: + return ret; +} + +static int cma_accept_iw(struct rdma_id_private *id_priv, + struct rdma_conn_param *conn_param) +{ + struct iw_cm_conn_param iw_param; + int ret; + + if (!conn_param) + return -EINVAL; + + ret = cma_modify_qp_rtr(id_priv, conn_param); + if (ret) + return ret; + + iw_param.ord = conn_param->initiator_depth; + iw_param.ird = conn_param->responder_resources; + iw_param.private_data = conn_param->private_data; + iw_param.private_data_len = conn_param->private_data_len; + if (id_priv->id.qp) { + iw_param.qpn = id_priv->qp_num; + } else + iw_param.qpn = conn_param->qp_num; + + return iw_cm_accept(id_priv->cm_id.iw, &iw_param); +} + +static int cma_send_sidr_rep(struct rdma_id_private *id_priv, + enum ib_cm_sidr_status status, u32 qkey, + const void *private_data, int private_data_len) +{ + struct ib_cm_sidr_rep_param rep; + int ret; + + memset(&rep, 0, sizeof rep); + rep.status = status; + if (status == IB_SIDR_SUCCESS) { + if (qkey) + ret = cma_set_qkey(id_priv, qkey); + else + ret = cma_set_default_qkey(id_priv); + if (ret) + return ret; + rep.qp_num = id_priv->qp_num; + rep.qkey = id_priv->qkey; + + rep.ece.vendor_id = id_priv->ece.vendor_id; + rep.ece.attr_mod = id_priv->ece.attr_mod; + } + + rep.private_data = private_data; + rep.private_data_len = private_data_len; + + trace_cm_send_sidr_rep(id_priv); + return ib_send_cm_sidr_rep(id_priv->cm_id.ib, &rep); +} + +/** + * rdma_accept - Called to accept a connection request or response. + * @id: Connection identifier associated with the request. + * @conn_param: Information needed to establish the connection. This must be + * provided if accepting a connection request. If accepting a connection + * response, this parameter must be NULL. + * + * Typically, this routine is only called by the listener to accept a connection + * request. It must also be called on the active side of a connection if the + * user is performing their own QP transitions. + * + * In the case of error, a reject message is sent to the remote side and the + * state of the qp associated with the id is modified to error, such that any + * previously posted receive buffers would be flushed. + * + * This function is for use by kernel ULPs and must be called from under the + * handler callback. + */ +int rdma_accept(struct rdma_cm_id *id, struct rdma_conn_param *conn_param) +{ + struct rdma_id_private *id_priv = + container_of(id, struct rdma_id_private, id); + int ret; + + lockdep_assert_held(&id_priv->handler_mutex); + + if (READ_ONCE(id_priv->state) != RDMA_CM_CONNECT) + return -EINVAL; + + if (!id->qp && conn_param) { + id_priv->qp_num = conn_param->qp_num; + id_priv->srq = conn_param->srq; + } + + if (rdma_cap_ib_cm(id->device, id->port_num)) { + if (id->qp_type == IB_QPT_UD) { + if (conn_param) + ret = cma_send_sidr_rep(id_priv, IB_SIDR_SUCCESS, + conn_param->qkey, + conn_param->private_data, + conn_param->private_data_len); + else + ret = cma_send_sidr_rep(id_priv, IB_SIDR_SUCCESS, + 0, NULL, 0); + } else { + if (conn_param) + ret = cma_accept_ib(id_priv, conn_param); + else + ret = cma_rep_recv(id_priv); + } + } else if (rdma_cap_iw_cm(id->device, id->port_num)) + ret = cma_accept_iw(id_priv, conn_param); + else + ret = -ENOSYS; + + if (ret) + goto reject; + + return 0; +reject: + cma_modify_qp_err(id_priv); + rdma_reject(id, NULL, 0, IB_CM_REJ_CONSUMER_DEFINED); + return ret; +} +EXPORT_SYMBOL(rdma_accept); + +int rdma_accept_ece(struct rdma_cm_id *id, struct rdma_conn_param *conn_param, + struct rdma_ucm_ece *ece) +{ + struct rdma_id_private *id_priv = + container_of(id, struct rdma_id_private, id); + + id_priv->ece.vendor_id = ece->vendor_id; + id_priv->ece.attr_mod = ece->attr_mod; + + return rdma_accept(id, conn_param); +} +EXPORT_SYMBOL(rdma_accept_ece); + +void rdma_lock_handler(struct rdma_cm_id *id) +{ + struct rdma_id_private *id_priv = + container_of(id, struct rdma_id_private, id); + + mutex_lock(&id_priv->handler_mutex); +} +EXPORT_SYMBOL(rdma_lock_handler); + +void rdma_unlock_handler(struct rdma_cm_id *id) +{ + struct rdma_id_private *id_priv = + container_of(id, struct rdma_id_private, id); + + mutex_unlock(&id_priv->handler_mutex); +} +EXPORT_SYMBOL(rdma_unlock_handler); + +int rdma_notify(struct rdma_cm_id *id, enum ib_event_type event) +{ + struct rdma_id_private *id_priv; + int ret; + + id_priv = container_of(id, struct rdma_id_private, id); + if (!id_priv->cm_id.ib) + return -EINVAL; + + switch (id->device->node_type) { + case RDMA_NODE_IB_CA: + ret = ib_cm_notify(id_priv->cm_id.ib, event); + break; + default: + ret = 0; + break; + } + return ret; +} +EXPORT_SYMBOL(rdma_notify); + +int rdma_reject(struct rdma_cm_id *id, const void *private_data, + u8 private_data_len, u8 reason) +{ + struct rdma_id_private *id_priv; + int ret; + + id_priv = container_of(id, struct rdma_id_private, id); + if (!id_priv->cm_id.ib) + return -EINVAL; + + if (rdma_cap_ib_cm(id->device, id->port_num)) { + if (id->qp_type == IB_QPT_UD) { + ret = cma_send_sidr_rep(id_priv, IB_SIDR_REJECT, 0, + private_data, private_data_len); + } else { + trace_cm_send_rej(id_priv); + ret = ib_send_cm_rej(id_priv->cm_id.ib, reason, NULL, 0, + private_data, private_data_len); + } + } else if (rdma_cap_iw_cm(id->device, id->port_num)) { + ret = iw_cm_reject(id_priv->cm_id.iw, + private_data, private_data_len); + } else + ret = -ENOSYS; + + return ret; +} +EXPORT_SYMBOL(rdma_reject); + +int rdma_disconnect(struct rdma_cm_id *id) +{ + struct rdma_id_private *id_priv; + int ret; + + id_priv = container_of(id, struct rdma_id_private, id); + if (!id_priv->cm_id.ib) + return -EINVAL; + + if (rdma_cap_ib_cm(id->device, id->port_num)) { + ret = cma_modify_qp_err(id_priv); + if (ret) + goto out; + /* Initiate or respond to a disconnect. */ + trace_cm_disconnect(id_priv); + if (ib_send_cm_dreq(id_priv->cm_id.ib, NULL, 0)) { + if (!ib_send_cm_drep(id_priv->cm_id.ib, NULL, 0)) + trace_cm_sent_drep(id_priv); + } else { + trace_cm_sent_dreq(id_priv); + } + } else if (rdma_cap_iw_cm(id->device, id->port_num)) { + ret = iw_cm_disconnect(id_priv->cm_id.iw, 0); + } else + ret = -EINVAL; + +out: + return ret; +} +EXPORT_SYMBOL(rdma_disconnect); + +static void cma_make_mc_event(int status, struct rdma_id_private *id_priv, + struct ib_sa_multicast *multicast, + struct rdma_cm_event *event, + struct cma_multicast *mc) +{ + struct rdma_dev_addr *dev_addr; + enum ib_gid_type gid_type; + struct net_device *ndev; + + if (status) + pr_debug_ratelimited("RDMA CM: MULTICAST_ERROR: failed to join multicast. status %d\n", + status); + + event->status = status; + event->param.ud.private_data = mc->context; + if (status) { + event->event = RDMA_CM_EVENT_MULTICAST_ERROR; + return; + } + + dev_addr = &id_priv->id.route.addr.dev_addr; + ndev = dev_get_by_index(dev_addr->net, dev_addr->bound_dev_if); + gid_type = + id_priv->cma_dev + ->default_gid_type[id_priv->id.port_num - + rdma_start_port( + id_priv->cma_dev->device)]; + + event->event = RDMA_CM_EVENT_MULTICAST_JOIN; + if (ib_init_ah_from_mcmember(id_priv->id.device, id_priv->id.port_num, + &multicast->rec, ndev, gid_type, + &event->param.ud.ah_attr)) { + event->event = RDMA_CM_EVENT_MULTICAST_ERROR; + goto out; + } + + event->param.ud.qp_num = 0xFFFFFF; + event->param.ud.qkey = id_priv->qkey; + +out: + if (ndev) + dev_put(ndev); +} + +static int cma_ib_mc_handler(int status, struct ib_sa_multicast *multicast) +{ + struct cma_multicast *mc = multicast->context; + struct rdma_id_private *id_priv = mc->id_priv; + struct rdma_cm_event event = {}; + int ret = 0; + + mutex_lock(&id_priv->handler_mutex); + if (READ_ONCE(id_priv->state) == RDMA_CM_DEVICE_REMOVAL || + READ_ONCE(id_priv->state) == RDMA_CM_DESTROYING) + goto out; + + ret = cma_set_qkey(id_priv, be32_to_cpu(multicast->rec.qkey)); + if (!ret) { + cma_make_mc_event(status, id_priv, multicast, &event, mc); + ret = cma_cm_event_handler(id_priv, &event); + } + rdma_destroy_ah_attr(&event.param.ud.ah_attr); + WARN_ON(ret); + +out: + mutex_unlock(&id_priv->handler_mutex); + return 0; +} + +static void cma_set_mgid(struct rdma_id_private *id_priv, + struct sockaddr *addr, union ib_gid *mgid) +{ + unsigned char mc_map[MAX_ADDR_LEN]; + struct rdma_dev_addr *dev_addr = &id_priv->id.route.addr.dev_addr; + struct sockaddr_in *sin = (struct sockaddr_in *) addr; + struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *) addr; + + if (cma_any_addr(addr)) { + memset(mgid, 0, sizeof *mgid); + } else if ((addr->sa_family == AF_INET6) && + ((be32_to_cpu(sin6->sin6_addr.s6_addr32[0]) & 0xFFF0FFFF) == + 0xFF10A01B)) { + /* IPv6 address is an SA assigned MGID. */ + memcpy(mgid, &sin6->sin6_addr, sizeof *mgid); + } else if (addr->sa_family == AF_IB) { + memcpy(mgid, &((struct sockaddr_ib *) addr)->sib_addr, sizeof *mgid); + } else if (addr->sa_family == AF_INET6) { + ipv6_ib_mc_map(&sin6->sin6_addr, dev_addr->broadcast, mc_map); + if (id_priv->id.ps == RDMA_PS_UDP) + mc_map[7] = 0x01; /* Use RDMA CM signature */ + *mgid = *(union ib_gid *) (mc_map + 4); + } else { + ip_ib_mc_map(sin->sin_addr.s_addr, dev_addr->broadcast, mc_map); + if (id_priv->id.ps == RDMA_PS_UDP) + mc_map[7] = 0x01; /* Use RDMA CM signature */ + *mgid = *(union ib_gid *) (mc_map + 4); + } +} + +static int cma_join_ib_multicast(struct rdma_id_private *id_priv, + struct cma_multicast *mc) +{ + struct ib_sa_mcmember_rec rec; + struct rdma_dev_addr *dev_addr = &id_priv->id.route.addr.dev_addr; + ib_sa_comp_mask comp_mask; + int ret; + + ib_addr_get_mgid(dev_addr, &rec.mgid); + ret = ib_sa_get_mcmember_rec(id_priv->id.device, id_priv->id.port_num, + &rec.mgid, &rec); + if (ret) + return ret; + + if (!id_priv->qkey) { + ret = cma_set_default_qkey(id_priv); + if (ret) + return ret; + } + + cma_set_mgid(id_priv, (struct sockaddr *) &mc->addr, &rec.mgid); + rec.qkey = cpu_to_be32(id_priv->qkey); + rdma_addr_get_sgid(dev_addr, &rec.port_gid); + rec.pkey = cpu_to_be16(ib_addr_get_pkey(dev_addr)); + rec.join_state = mc->join_state; + + if ((rec.join_state == BIT(SENDONLY_FULLMEMBER_JOIN)) && + (!ib_sa_sendonly_fullmem_support(&sa_client, + id_priv->id.device, + id_priv->id.port_num))) { + dev_warn( + &id_priv->id.device->dev, + "RDMA CM: port %u Unable to multicast join: SM doesn't support Send Only Full Member option\n", + id_priv->id.port_num); + return -EOPNOTSUPP; + } + + comp_mask = IB_SA_MCMEMBER_REC_MGID | IB_SA_MCMEMBER_REC_PORT_GID | + IB_SA_MCMEMBER_REC_PKEY | IB_SA_MCMEMBER_REC_JOIN_STATE | + IB_SA_MCMEMBER_REC_QKEY | IB_SA_MCMEMBER_REC_SL | + IB_SA_MCMEMBER_REC_FLOW_LABEL | + IB_SA_MCMEMBER_REC_TRAFFIC_CLASS; + + if (id_priv->id.ps == RDMA_PS_IPOIB) + comp_mask |= IB_SA_MCMEMBER_REC_RATE | + IB_SA_MCMEMBER_REC_RATE_SELECTOR | + IB_SA_MCMEMBER_REC_MTU_SELECTOR | + IB_SA_MCMEMBER_REC_MTU | + IB_SA_MCMEMBER_REC_HOP_LIMIT; + + mc->sa_mc = ib_sa_join_multicast(&sa_client, id_priv->id.device, + id_priv->id.port_num, &rec, comp_mask, + GFP_KERNEL, cma_ib_mc_handler, mc); + return PTR_ERR_OR_ZERO(mc->sa_mc); +} + +static void cma_iboe_set_mgid(struct sockaddr *addr, union ib_gid *mgid, + enum ib_gid_type gid_type) +{ + struct sockaddr_in *sin = (struct sockaddr_in *)addr; + struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)addr; + + if (cma_any_addr(addr)) { + memset(mgid, 0, sizeof *mgid); + } else if (addr->sa_family == AF_INET6) { + memcpy(mgid, &sin6->sin6_addr, sizeof *mgid); + } else { + mgid->raw[0] = + (gid_type == IB_GID_TYPE_ROCE_UDP_ENCAP) ? 0 : 0xff; + mgid->raw[1] = + (gid_type == IB_GID_TYPE_ROCE_UDP_ENCAP) ? 0 : 0x0e; + mgid->raw[2] = 0; + mgid->raw[3] = 0; + mgid->raw[4] = 0; + mgid->raw[5] = 0; + mgid->raw[6] = 0; + mgid->raw[7] = 0; + mgid->raw[8] = 0; + mgid->raw[9] = 0; + mgid->raw[10] = 0xff; + mgid->raw[11] = 0xff; + *(__be32 *)(&mgid->raw[12]) = sin->sin_addr.s_addr; + } +} + +static int cma_iboe_join_multicast(struct rdma_id_private *id_priv, + struct cma_multicast *mc) +{ + struct rdma_dev_addr *dev_addr = &id_priv->id.route.addr.dev_addr; + int err = 0; + struct sockaddr *addr = (struct sockaddr *)&mc->addr; + struct net_device *ndev = NULL; + struct ib_sa_multicast ib = {}; + enum ib_gid_type gid_type; + bool send_only; + + send_only = mc->join_state == BIT(SENDONLY_FULLMEMBER_JOIN); + + if (cma_zero_addr(addr)) + return -EINVAL; + + gid_type = id_priv->cma_dev->default_gid_type[id_priv->id.port_num - + rdma_start_port(id_priv->cma_dev->device)]; + cma_iboe_set_mgid(addr, &ib.rec.mgid, gid_type); + + ib.rec.pkey = cpu_to_be16(0xffff); + if (dev_addr->bound_dev_if) + ndev = dev_get_by_index(dev_addr->net, dev_addr->bound_dev_if); + if (!ndev) + return -ENODEV; + + ib.rec.rate = IB_RATE_PORT_CURRENT; + ib.rec.hop_limit = 1; + ib.rec.mtu = iboe_get_mtu(ndev->mtu); + + if (addr->sa_family == AF_INET) { + if (gid_type == IB_GID_TYPE_ROCE_UDP_ENCAP) { + ib.rec.hop_limit = IPV6_DEFAULT_HOPLIMIT; + if (!send_only) { + err = cma_igmp_send(ndev, &ib.rec.mgid, + true); + } + } + } else { + if (gid_type == IB_GID_TYPE_ROCE_UDP_ENCAP) + err = -ENOTSUPP; + } + dev_put(ndev); + if (err || !ib.rec.mtu) + return err ?: -EINVAL; + + if (!id_priv->qkey) + cma_set_default_qkey(id_priv); + + rdma_ip2gid((struct sockaddr *)&id_priv->id.route.addr.src_addr, + &ib.rec.port_gid); + INIT_WORK(&mc->iboe_join.work, cma_iboe_join_work_handler); + cma_make_mc_event(0, id_priv, &ib, &mc->iboe_join.event, mc); + queue_work(cma_wq, &mc->iboe_join.work); + return 0; +} + +int rdma_join_multicast(struct rdma_cm_id *id, struct sockaddr *addr, + u8 join_state, void *context) +{ + struct rdma_id_private *id_priv = + container_of(id, struct rdma_id_private, id); + struct cma_multicast *mc; + int ret; + + /* Not supported for kernel QPs */ + if (WARN_ON(id->qp)) + return -EINVAL; + + /* ULP is calling this wrong. */ + if (!id->device || (READ_ONCE(id_priv->state) != RDMA_CM_ADDR_BOUND && + READ_ONCE(id_priv->state) != RDMA_CM_ADDR_RESOLVED)) + return -EINVAL; + + if (id_priv->id.qp_type != IB_QPT_UD) + return -EINVAL; + + mc = kzalloc(sizeof(*mc), GFP_KERNEL); + if (!mc) + return -ENOMEM; + + memcpy(&mc->addr, addr, rdma_addr_size(addr)); + mc->context = context; + mc->id_priv = id_priv; + mc->join_state = join_state; + + if (rdma_protocol_roce(id->device, id->port_num)) { + ret = cma_iboe_join_multicast(id_priv, mc); + if (ret) + goto out_err; + } else if (rdma_cap_ib_mcast(id->device, id->port_num)) { + ret = cma_join_ib_multicast(id_priv, mc); + if (ret) + goto out_err; + } else { + ret = -ENOSYS; + goto out_err; + } + + spin_lock(&id_priv->lock); + list_add(&mc->list, &id_priv->mc_list); + spin_unlock(&id_priv->lock); + + return 0; +out_err: + kfree(mc); + return ret; +} +EXPORT_SYMBOL(rdma_join_multicast); + +void rdma_leave_multicast(struct rdma_cm_id *id, struct sockaddr *addr) +{ + struct rdma_id_private *id_priv; + struct cma_multicast *mc; + + id_priv = container_of(id, struct rdma_id_private, id); + spin_lock_irq(&id_priv->lock); + list_for_each_entry(mc, &id_priv->mc_list, list) { + if (memcmp(&mc->addr, addr, rdma_addr_size(addr)) != 0) + continue; + list_del(&mc->list); + spin_unlock_irq(&id_priv->lock); + + WARN_ON(id_priv->cma_dev->device != id->device); + destroy_mc(id_priv, mc); + return; + } + spin_unlock_irq(&id_priv->lock); +} +EXPORT_SYMBOL(rdma_leave_multicast); + +static int cma_netdev_change(struct net_device *ndev, struct rdma_id_private *id_priv) +{ + struct rdma_dev_addr *dev_addr; + struct cma_work *work; + + dev_addr = &id_priv->id.route.addr.dev_addr; + + if ((dev_addr->bound_dev_if == ndev->ifindex) && + (net_eq(dev_net(ndev), dev_addr->net)) && + memcmp(dev_addr->src_dev_addr, ndev->dev_addr, ndev->addr_len)) { + pr_info("RDMA CM addr change for ndev %s used by id %p\n", + ndev->name, &id_priv->id); + work = kzalloc(sizeof *work, GFP_KERNEL); + if (!work) + return -ENOMEM; + + INIT_WORK(&work->work, cma_work_handler); + work->id = id_priv; + work->event.event = RDMA_CM_EVENT_ADDR_CHANGE; + cma_id_get(id_priv); + queue_work(cma_wq, &work->work); + } + + return 0; +} + +static int cma_netdev_callback(struct notifier_block *self, unsigned long event, + void *ptr) +{ + struct net_device *ndev = netdev_notifier_info_to_dev(ptr); + struct cma_device *cma_dev; + struct rdma_id_private *id_priv; + int ret = NOTIFY_DONE; + + if (event != NETDEV_BONDING_FAILOVER) + return NOTIFY_DONE; + + if (!netif_is_bond_master(ndev)) + return NOTIFY_DONE; + + mutex_lock(&lock); + list_for_each_entry(cma_dev, &dev_list, list) + list_for_each_entry(id_priv, &cma_dev->id_list, list) { + ret = cma_netdev_change(ndev, id_priv); + if (ret) + goto out; + } + +out: + mutex_unlock(&lock); + return ret; +} + +static struct notifier_block cma_nb = { + .notifier_call = cma_netdev_callback +}; + +static void cma_send_device_removal_put(struct rdma_id_private *id_priv) +{ + struct rdma_cm_event event = { .event = RDMA_CM_EVENT_DEVICE_REMOVAL }; + enum rdma_cm_state state; + unsigned long flags; + + mutex_lock(&id_priv->handler_mutex); + /* Record that we want to remove the device */ + spin_lock_irqsave(&id_priv->lock, flags); + state = id_priv->state; + if (state == RDMA_CM_DESTROYING || state == RDMA_CM_DEVICE_REMOVAL) { + spin_unlock_irqrestore(&id_priv->lock, flags); + mutex_unlock(&id_priv->handler_mutex); + cma_id_put(id_priv); + return; + } + id_priv->state = RDMA_CM_DEVICE_REMOVAL; + spin_unlock_irqrestore(&id_priv->lock, flags); + + if (cma_cm_event_handler(id_priv, &event)) { + /* + * At this point the ULP promises it won't call + * rdma_destroy_id() concurrently + */ + cma_id_put(id_priv); + mutex_unlock(&id_priv->handler_mutex); + trace_cm_id_destroy(id_priv); + _destroy_id(id_priv, state); + return; + } + mutex_unlock(&id_priv->handler_mutex); + + /* + * If this races with destroy then the thread that first assigns state + * to a destroying does the cancel. + */ + cma_cancel_operation(id_priv, state); + cma_id_put(id_priv); +} + +static void cma_process_remove(struct cma_device *cma_dev) +{ + mutex_lock(&lock); + while (!list_empty(&cma_dev->id_list)) { + struct rdma_id_private *id_priv = list_first_entry( + &cma_dev->id_list, struct rdma_id_private, list); + + list_del(&id_priv->listen_list); + list_del_init(&id_priv->list); + cma_id_get(id_priv); + mutex_unlock(&lock); + + cma_send_device_removal_put(id_priv); + + mutex_lock(&lock); + } + mutex_unlock(&lock); + + cma_dev_put(cma_dev); + wait_for_completion(&cma_dev->comp); +} + +static int cma_add_one(struct ib_device *device) +{ + struct rdma_id_private *to_destroy; + struct cma_device *cma_dev; + struct rdma_id_private *id_priv; + unsigned int i; + unsigned long supported_gids = 0; + int ret; + + cma_dev = kmalloc(sizeof(*cma_dev), GFP_KERNEL); + if (!cma_dev) + return -ENOMEM; + + cma_dev->device = device; + cma_dev->default_gid_type = kcalloc(device->phys_port_cnt, + sizeof(*cma_dev->default_gid_type), + GFP_KERNEL); + if (!cma_dev->default_gid_type) { + ret = -ENOMEM; + goto free_cma_dev; + } + + cma_dev->default_roce_tos = kcalloc(device->phys_port_cnt, + sizeof(*cma_dev->default_roce_tos), + GFP_KERNEL); + if (!cma_dev->default_roce_tos) { + ret = -ENOMEM; + goto free_gid_type; + } + + rdma_for_each_port (device, i) { + supported_gids = roce_gid_type_mask_support(device, i); + WARN_ON(!supported_gids); + if (supported_gids & (1 << CMA_PREFERRED_ROCE_GID_TYPE)) + cma_dev->default_gid_type[i - rdma_start_port(device)] = + CMA_PREFERRED_ROCE_GID_TYPE; + else + cma_dev->default_gid_type[i - rdma_start_port(device)] = + find_first_bit(&supported_gids, BITS_PER_LONG); + cma_dev->default_roce_tos[i - rdma_start_port(device)] = 0; + } + + init_completion(&cma_dev->comp); + refcount_set(&cma_dev->refcount, 1); + INIT_LIST_HEAD(&cma_dev->id_list); + ib_set_client_data(device, &cma_client, cma_dev); + + mutex_lock(&lock); + list_add_tail(&cma_dev->list, &dev_list); + list_for_each_entry(id_priv, &listen_any_list, list) { + ret = cma_listen_on_dev(id_priv, cma_dev, &to_destroy); + if (ret) + goto free_listen; + } + mutex_unlock(&lock); + + trace_cm_add_one(device); + return 0; + +free_listen: + list_del(&cma_dev->list); + mutex_unlock(&lock); + + /* cma_process_remove() will delete to_destroy */ + cma_process_remove(cma_dev); + kfree(cma_dev->default_roce_tos); +free_gid_type: + kfree(cma_dev->default_gid_type); + +free_cma_dev: + kfree(cma_dev); + return ret; +} + +static void cma_remove_one(struct ib_device *device, void *client_data) +{ + struct cma_device *cma_dev = client_data; + + trace_cm_remove_one(device); + + mutex_lock(&lock); + list_del(&cma_dev->list); + mutex_unlock(&lock); + + cma_process_remove(cma_dev); + kfree(cma_dev->default_roce_tos); + kfree(cma_dev->default_gid_type); + kfree(cma_dev); +} + +static int cma_init_net(struct net *net) +{ + struct cma_pernet *pernet = cma_pernet(net); + + xa_init(&pernet->tcp_ps); + xa_init(&pernet->udp_ps); + xa_init(&pernet->ipoib_ps); + xa_init(&pernet->ib_ps); + + return 0; +} + +static void cma_exit_net(struct net *net) +{ + struct cma_pernet *pernet = cma_pernet(net); + + WARN_ON(!xa_empty(&pernet->tcp_ps)); + WARN_ON(!xa_empty(&pernet->udp_ps)); + WARN_ON(!xa_empty(&pernet->ipoib_ps)); + WARN_ON(!xa_empty(&pernet->ib_ps)); +} + +static struct pernet_operations cma_pernet_operations = { + .init = cma_init_net, + .exit = cma_exit_net, + .id = &cma_pernet_id, + .size = sizeof(struct cma_pernet), +}; + +static int __init cma_init(void) +{ + int ret; + + /* + * There is a rare lock ordering dependency in cma_netdev_callback() + * that only happens when bonding is enabled. Teach lockdep that rtnl + * must never be nested under lock so it can find these without having + * to test with bonding. + */ + if (IS_ENABLED(CONFIG_LOCKDEP)) { + rtnl_lock(); + mutex_lock(&lock); + mutex_unlock(&lock); + rtnl_unlock(); + } + + cma_wq = alloc_ordered_workqueue("rdma_cm", WQ_MEM_RECLAIM); + if (!cma_wq) + return -ENOMEM; + + ret = register_pernet_subsys(&cma_pernet_operations); + if (ret) + goto err_wq; + + ib_sa_register_client(&sa_client); + register_netdevice_notifier(&cma_nb); + + ret = ib_register_client(&cma_client); + if (ret) + goto err; + + ret = cma_configfs_init(); + if (ret) + goto err_ib; + + return 0; + +err_ib: + ib_unregister_client(&cma_client); +err: + unregister_netdevice_notifier(&cma_nb); + ib_sa_unregister_client(&sa_client); + unregister_pernet_subsys(&cma_pernet_operations); +err_wq: + destroy_workqueue(cma_wq); + return ret; +} + +static void __exit cma_cleanup(void) +{ + cma_configfs_exit(); + ib_unregister_client(&cma_client); + unregister_netdevice_notifier(&cma_nb); + ib_sa_unregister_client(&sa_client); + unregister_pernet_subsys(&cma_pernet_operations); + destroy_workqueue(cma_wq); +} + +module_init(cma_init); +module_exit(cma_cleanup); diff --git a/drivers/infiniband/core/cma_configfs.c b/drivers/infiniband/core/cma_configfs.c new file mode 100644 index 000000000..d7cc8d5dc --- /dev/null +++ b/drivers/infiniband/core/cma_configfs.c @@ -0,0 +1,375 @@ +/* + * Copyright (c) 2015, Mellanox Technologies inc. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include <linux/module.h> +#include <linux/configfs.h> +#include <rdma/ib_verbs.h> +#include <rdma/rdma_cm.h> + +#include "core_priv.h" +#include "cma_priv.h" + +struct cma_device; + +struct cma_dev_group; + +struct cma_dev_port_group { + unsigned int port_num; + struct cma_dev_group *cma_dev_group; + struct config_group group; +}; + +struct cma_dev_group { + char name[IB_DEVICE_NAME_MAX]; + struct config_group device_group; + struct config_group ports_group; + struct cma_dev_port_group *ports; +}; + +static struct cma_dev_port_group *to_dev_port_group(struct config_item *item) +{ + struct config_group *group; + + if (!item) + return NULL; + + group = container_of(item, struct config_group, cg_item); + return container_of(group, struct cma_dev_port_group, group); +} + +static bool filter_by_name(struct ib_device *ib_dev, void *cookie) +{ + return !strcmp(dev_name(&ib_dev->dev), cookie); +} + +static int cma_configfs_params_get(struct config_item *item, + struct cma_device **pcma_dev, + struct cma_dev_port_group **pgroup) +{ + struct cma_dev_port_group *group = to_dev_port_group(item); + struct cma_device *cma_dev; + + if (!group) + return -ENODEV; + + cma_dev = cma_enum_devices_by_ibdev(filter_by_name, + group->cma_dev_group->name); + if (!cma_dev) + return -ENODEV; + + *pcma_dev = cma_dev; + *pgroup = group; + + return 0; +} + +static void cma_configfs_params_put(struct cma_device *cma_dev) +{ + cma_dev_put(cma_dev); +} + +static ssize_t default_roce_mode_show(struct config_item *item, + char *buf) +{ + struct cma_device *cma_dev; + struct cma_dev_port_group *group; + int gid_type; + ssize_t ret; + + ret = cma_configfs_params_get(item, &cma_dev, &group); + if (ret) + return ret; + + gid_type = cma_get_default_gid_type(cma_dev, group->port_num); + cma_configfs_params_put(cma_dev); + + if (gid_type < 0) + return gid_type; + + return sprintf(buf, "%s\n", ib_cache_gid_type_str(gid_type)); +} + +static ssize_t default_roce_mode_store(struct config_item *item, + const char *buf, size_t count) +{ + struct cma_device *cma_dev; + struct cma_dev_port_group *group; + int gid_type; + ssize_t ret; + + ret = cma_configfs_params_get(item, &cma_dev, &group); + if (ret) + return ret; + + gid_type = ib_cache_gid_parse_type_str(buf); + if (gid_type < 0) { + cma_configfs_params_put(cma_dev); + return -EINVAL; + } + + ret = cma_set_default_gid_type(cma_dev, group->port_num, gid_type); + + cma_configfs_params_put(cma_dev); + + return !ret ? strnlen(buf, count) : ret; +} + +CONFIGFS_ATTR(, default_roce_mode); + +static ssize_t default_roce_tos_show(struct config_item *item, char *buf) +{ + struct cma_device *cma_dev; + struct cma_dev_port_group *group; + ssize_t ret; + u8 tos; + + ret = cma_configfs_params_get(item, &cma_dev, &group); + if (ret) + return ret; + + tos = cma_get_default_roce_tos(cma_dev, group->port_num); + cma_configfs_params_put(cma_dev); + + return sprintf(buf, "%u\n", tos); +} + +static ssize_t default_roce_tos_store(struct config_item *item, + const char *buf, size_t count) +{ + struct cma_device *cma_dev; + struct cma_dev_port_group *group; + ssize_t ret; + u8 tos; + + ret = kstrtou8(buf, 0, &tos); + if (ret) + return ret; + + ret = cma_configfs_params_get(item, &cma_dev, &group); + if (ret) + return ret; + + ret = cma_set_default_roce_tos(cma_dev, group->port_num, tos); + cma_configfs_params_put(cma_dev); + + return ret ? ret : strnlen(buf, count); +} + +CONFIGFS_ATTR(, default_roce_tos); + +static struct configfs_attribute *cma_configfs_attributes[] = { + &attr_default_roce_mode, + &attr_default_roce_tos, + NULL, +}; + +static const struct config_item_type cma_port_group_type = { + .ct_attrs = cma_configfs_attributes, + .ct_owner = THIS_MODULE +}; + +static int make_cma_ports(struct cma_dev_group *cma_dev_group, + struct cma_device *cma_dev) +{ + struct ib_device *ibdev; + unsigned int i; + unsigned int ports_num; + struct cma_dev_port_group *ports; + int err; + + ibdev = cma_get_ib_dev(cma_dev); + + if (!ibdev) + return -ENODEV; + + ports_num = ibdev->phys_port_cnt; + ports = kcalloc(ports_num, sizeof(*cma_dev_group->ports), + GFP_KERNEL); + + if (!ports) { + err = -ENOMEM; + goto free; + } + + for (i = 0; i < ports_num; i++) { + char port_str[11]; + + ports[i].port_num = i + 1; + snprintf(port_str, sizeof(port_str), "%u", i + 1); + ports[i].cma_dev_group = cma_dev_group; + config_group_init_type_name(&ports[i].group, + port_str, + &cma_port_group_type); + configfs_add_default_group(&ports[i].group, + &cma_dev_group->ports_group); + + } + cma_dev_group->ports = ports; + + return 0; +free: + kfree(ports); + cma_dev_group->ports = NULL; + return err; +} + +static void release_cma_dev(struct config_item *item) +{ + struct config_group *group = container_of(item, struct config_group, + cg_item); + struct cma_dev_group *cma_dev_group = container_of(group, + struct cma_dev_group, + device_group); + + kfree(cma_dev_group); +}; + +static void release_cma_ports_group(struct config_item *item) +{ + struct config_group *group = container_of(item, struct config_group, + cg_item); + struct cma_dev_group *cma_dev_group = container_of(group, + struct cma_dev_group, + ports_group); + + kfree(cma_dev_group->ports); + cma_dev_group->ports = NULL; +}; + +static struct configfs_item_operations cma_ports_item_ops = { + .release = release_cma_ports_group +}; + +static const struct config_item_type cma_ports_group_type = { + .ct_item_ops = &cma_ports_item_ops, + .ct_owner = THIS_MODULE +}; + +static struct configfs_item_operations cma_device_item_ops = { + .release = release_cma_dev +}; + +static const struct config_item_type cma_device_group_type = { + .ct_item_ops = &cma_device_item_ops, + .ct_owner = THIS_MODULE +}; + +static struct config_group *make_cma_dev(struct config_group *group, + const char *name) +{ + int err = -ENODEV; + struct cma_device *cma_dev = cma_enum_devices_by_ibdev(filter_by_name, + (void *)name); + struct cma_dev_group *cma_dev_group = NULL; + + if (!cma_dev) + goto fail; + + cma_dev_group = kzalloc(sizeof(*cma_dev_group), GFP_KERNEL); + + if (!cma_dev_group) { + err = -ENOMEM; + goto fail; + } + + strlcpy(cma_dev_group->name, name, sizeof(cma_dev_group->name)); + + config_group_init_type_name(&cma_dev_group->ports_group, "ports", + &cma_ports_group_type); + + err = make_cma_ports(cma_dev_group, cma_dev); + if (err) + goto fail; + + config_group_init_type_name(&cma_dev_group->device_group, name, + &cma_device_group_type); + configfs_add_default_group(&cma_dev_group->ports_group, + &cma_dev_group->device_group); + + cma_dev_put(cma_dev); + return &cma_dev_group->device_group; + +fail: + if (cma_dev) + cma_dev_put(cma_dev); + kfree(cma_dev_group); + return ERR_PTR(err); +} + +static void drop_cma_dev(struct config_group *cgroup, struct config_item *item) +{ + struct config_group *group = + container_of(item, struct config_group, cg_item); + struct cma_dev_group *cma_dev_group = + container_of(group, struct cma_dev_group, device_group); + + configfs_remove_default_groups(&cma_dev_group->ports_group); + configfs_remove_default_groups(&cma_dev_group->device_group); + config_item_put(item); +} + +static struct configfs_group_operations cma_subsys_group_ops = { + .make_group = make_cma_dev, + .drop_item = drop_cma_dev, +}; + +static const struct config_item_type cma_subsys_type = { + .ct_group_ops = &cma_subsys_group_ops, + .ct_owner = THIS_MODULE, +}; + +static struct configfs_subsystem cma_subsys = { + .su_group = { + .cg_item = { + .ci_namebuf = "rdma_cm", + .ci_type = &cma_subsys_type, + }, + }, +}; + +int __init cma_configfs_init(void) +{ + int ret; + + config_group_init(&cma_subsys.su_group); + mutex_init(&cma_subsys.su_mutex); + ret = configfs_register_subsystem(&cma_subsys); + if (ret) + mutex_destroy(&cma_subsys.su_mutex); + return ret; +} + +void __exit cma_configfs_exit(void) +{ + configfs_unregister_subsystem(&cma_subsys); + mutex_destroy(&cma_subsys.su_mutex); +} diff --git a/drivers/infiniband/core/cma_priv.h b/drivers/infiniband/core/cma_priv.h new file mode 100644 index 000000000..b53f4fa5e --- /dev/null +++ b/drivers/infiniband/core/cma_priv.h @@ -0,0 +1,129 @@ +/* + * Copyright (c) 2005 Voltaire Inc. All rights reserved. + * Copyright (c) 2002-2005, Network Appliance, Inc. All rights reserved. + * Copyright (c) 1999-2005, Mellanox Technologies, Inc. All rights reserved. + * Copyright (c) 2005-2006 Intel Corporation. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#ifndef _CMA_PRIV_H +#define _CMA_PRIV_H + +enum rdma_cm_state { + RDMA_CM_IDLE, + RDMA_CM_ADDR_QUERY, + RDMA_CM_ADDR_RESOLVED, + RDMA_CM_ROUTE_QUERY, + RDMA_CM_ROUTE_RESOLVED, + RDMA_CM_CONNECT, + RDMA_CM_DISCONNECT, + RDMA_CM_ADDR_BOUND, + RDMA_CM_LISTEN, + RDMA_CM_DEVICE_REMOVAL, + RDMA_CM_DESTROYING +}; + +struct rdma_id_private { + struct rdma_cm_id id; + + struct rdma_bind_list *bind_list; + struct hlist_node node; + struct list_head list; /* listen_any_list or cma_device.list */ + struct list_head listen_list; /* per device listens */ + struct cma_device *cma_dev; + struct list_head mc_list; + + int internal_id; + enum rdma_cm_state state; + spinlock_t lock; + struct mutex qp_mutex; + + struct completion comp; + refcount_t refcount; + struct mutex handler_mutex; + + int backlog; + int timeout_ms; + struct ib_sa_query *query; + int query_id; + union { + struct ib_cm_id *ib; + struct iw_cm_id *iw; + } cm_id; + + u32 seq_num; + u32 qkey; + u32 qp_num; + u32 options; + u8 srq; + u8 tos; + u8 tos_set:1; + u8 timeout_set:1; + u8 reuseaddr; + u8 afonly; + u8 timeout; + u8 used_resolve_ip; + enum ib_gid_type gid_type; + + /* + * Internal to RDMA/core, don't use in the drivers + */ + struct rdma_restrack_entry res; + struct rdma_ucm_ece ece; +}; + +#if IS_ENABLED(CONFIG_INFINIBAND_ADDR_TRANS_CONFIGFS) +int cma_configfs_init(void); +void cma_configfs_exit(void); +#else +static inline int cma_configfs_init(void) +{ + return 0; +} + +static inline void cma_configfs_exit(void) +{ +} +#endif + +void cma_dev_get(struct cma_device *dev); +void cma_dev_put(struct cma_device *dev); +typedef bool (*cma_device_filter)(struct ib_device *, void *); +struct cma_device *cma_enum_devices_by_ibdev(cma_device_filter filter, + void *cookie); +int cma_get_default_gid_type(struct cma_device *dev, unsigned int port); +int cma_set_default_gid_type(struct cma_device *dev, unsigned int port, + enum ib_gid_type default_gid_type); +int cma_get_default_roce_tos(struct cma_device *dev, unsigned int port); +int cma_set_default_roce_tos(struct cma_device *dev, unsigned int port, + u8 default_roce_tos); +struct ib_device *cma_get_ib_dev(struct cma_device *dev); + +#endif /* _CMA_PRIV_H */ diff --git a/drivers/infiniband/core/cma_trace.c b/drivers/infiniband/core/cma_trace.c new file mode 100644 index 000000000..b314a281e --- /dev/null +++ b/drivers/infiniband/core/cma_trace.c @@ -0,0 +1,16 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * Trace points for the RDMA Connection Manager. + * + * Author: Chuck Lever <chuck.lever@oracle.com> + * + * Copyright (c) 2019, Oracle and/or its affiliates. All rights reserved. + */ + +#define CREATE_TRACE_POINTS + +#include <rdma/rdma_cm.h> +#include <rdma/ib_cm.h> +#include "cma_priv.h" + +#include "cma_trace.h" diff --git a/drivers/infiniband/core/cma_trace.h b/drivers/infiniband/core/cma_trace.h new file mode 100644 index 000000000..e45264267 --- /dev/null +++ b/drivers/infiniband/core/cma_trace.h @@ -0,0 +1,361 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +/* + * Trace point definitions for the RDMA Connect Manager. + * + * Author: Chuck Lever <chuck.lever@oracle.com> + * + * Copyright (c) 2019, Oracle and/or its affiliates. All rights reserved. + */ + +#undef TRACE_SYSTEM +#define TRACE_SYSTEM rdma_cma + +#if !defined(_TRACE_RDMA_CMA_H) || defined(TRACE_HEADER_MULTI_READ) + +#define _TRACE_RDMA_CMA_H + +#include <linux/tracepoint.h> +#include <trace/events/rdma.h> + + +DECLARE_EVENT_CLASS(cma_fsm_class, + TP_PROTO( + const struct rdma_id_private *id_priv + ), + + TP_ARGS(id_priv), + + TP_STRUCT__entry( + __field(u32, cm_id) + __field(u32, tos) + __array(unsigned char, srcaddr, sizeof(struct sockaddr_in6)) + __array(unsigned char, dstaddr, sizeof(struct sockaddr_in6)) + ), + + TP_fast_assign( + __entry->cm_id = id_priv->res.id; + __entry->tos = id_priv->tos; + memcpy(__entry->srcaddr, &id_priv->id.route.addr.src_addr, + sizeof(struct sockaddr_in6)); + memcpy(__entry->dstaddr, &id_priv->id.route.addr.dst_addr, + sizeof(struct sockaddr_in6)); + ), + + TP_printk("cm.id=%u src=%pISpc dst=%pISpc tos=%u", + __entry->cm_id, __entry->srcaddr, __entry->dstaddr, __entry->tos + ) +); + +#define DEFINE_CMA_FSM_EVENT(name) \ + DEFINE_EVENT(cma_fsm_class, cm_##name, \ + TP_PROTO( \ + const struct rdma_id_private *id_priv \ + ), \ + TP_ARGS(id_priv)) + +DEFINE_CMA_FSM_EVENT(send_rtu); +DEFINE_CMA_FSM_EVENT(send_rej); +DEFINE_CMA_FSM_EVENT(send_mra); +DEFINE_CMA_FSM_EVENT(send_sidr_req); +DEFINE_CMA_FSM_EVENT(send_sidr_rep); +DEFINE_CMA_FSM_EVENT(disconnect); +DEFINE_CMA_FSM_EVENT(sent_drep); +DEFINE_CMA_FSM_EVENT(sent_dreq); +DEFINE_CMA_FSM_EVENT(id_destroy); + +TRACE_EVENT(cm_id_attach, + TP_PROTO( + const struct rdma_id_private *id_priv, + const struct ib_device *device + ), + + TP_ARGS(id_priv, device), + + TP_STRUCT__entry( + __field(u32, cm_id) + __array(unsigned char, srcaddr, sizeof(struct sockaddr_in6)) + __array(unsigned char, dstaddr, sizeof(struct sockaddr_in6)) + __string(devname, device->name) + ), + + TP_fast_assign( + __entry->cm_id = id_priv->res.id; + memcpy(__entry->srcaddr, &id_priv->id.route.addr.src_addr, + sizeof(struct sockaddr_in6)); + memcpy(__entry->dstaddr, &id_priv->id.route.addr.dst_addr, + sizeof(struct sockaddr_in6)); + __assign_str(devname, device->name); + ), + + TP_printk("cm.id=%u src=%pISpc dst=%pISpc device=%s", + __entry->cm_id, __entry->srcaddr, __entry->dstaddr, + __get_str(devname) + ) +); + +DECLARE_EVENT_CLASS(cma_qp_class, + TP_PROTO( + const struct rdma_id_private *id_priv + ), + + TP_ARGS(id_priv), + + TP_STRUCT__entry( + __field(u32, cm_id) + __field(u32, tos) + __field(u32, qp_num) + __array(unsigned char, srcaddr, sizeof(struct sockaddr_in6)) + __array(unsigned char, dstaddr, sizeof(struct sockaddr_in6)) + ), + + TP_fast_assign( + __entry->cm_id = id_priv->res.id; + __entry->tos = id_priv->tos; + __entry->qp_num = id_priv->qp_num; + memcpy(__entry->srcaddr, &id_priv->id.route.addr.src_addr, + sizeof(struct sockaddr_in6)); + memcpy(__entry->dstaddr, &id_priv->id.route.addr.dst_addr, + sizeof(struct sockaddr_in6)); + ), + + TP_printk("cm.id=%u src=%pISpc dst=%pISpc tos=%u qp_num=%u", + __entry->cm_id, __entry->srcaddr, __entry->dstaddr, __entry->tos, + __entry->qp_num + ) +); + +#define DEFINE_CMA_QP_EVENT(name) \ + DEFINE_EVENT(cma_qp_class, cm_##name, \ + TP_PROTO( \ + const struct rdma_id_private *id_priv \ + ), \ + TP_ARGS(id_priv)) + +DEFINE_CMA_QP_EVENT(send_req); +DEFINE_CMA_QP_EVENT(send_rep); +DEFINE_CMA_QP_EVENT(qp_destroy); + +/* + * enum ib_wp_type, from include/rdma/ib_verbs.h + */ +#define IB_QP_TYPE_LIST \ + ib_qp_type(SMI) \ + ib_qp_type(GSI) \ + ib_qp_type(RC) \ + ib_qp_type(UC) \ + ib_qp_type(UD) \ + ib_qp_type(RAW_IPV6) \ + ib_qp_type(RAW_ETHERTYPE) \ + ib_qp_type(RAW_PACKET) \ + ib_qp_type(XRC_INI) \ + ib_qp_type_end(XRC_TGT) + +#undef ib_qp_type +#undef ib_qp_type_end + +#define ib_qp_type(x) TRACE_DEFINE_ENUM(IB_QPT_##x); +#define ib_qp_type_end(x) TRACE_DEFINE_ENUM(IB_QPT_##x); + +IB_QP_TYPE_LIST + +#undef ib_qp_type +#undef ib_qp_type_end + +#define ib_qp_type(x) { IB_QPT_##x, #x }, +#define ib_qp_type_end(x) { IB_QPT_##x, #x } + +#define rdma_show_qp_type(x) \ + __print_symbolic(x, IB_QP_TYPE_LIST) + + +TRACE_EVENT(cm_qp_create, + TP_PROTO( + const struct rdma_id_private *id_priv, + const struct ib_pd *pd, + const struct ib_qp_init_attr *qp_init_attr, + int rc + ), + + TP_ARGS(id_priv, pd, qp_init_attr, rc), + + TP_STRUCT__entry( + __field(u32, cm_id) + __field(u32, pd_id) + __field(u32, tos) + __field(u32, qp_num) + __field(u32, send_wr) + __field(u32, recv_wr) + __field(int, rc) + __field(unsigned long, qp_type) + __array(unsigned char, srcaddr, sizeof(struct sockaddr_in6)) + __array(unsigned char, dstaddr, sizeof(struct sockaddr_in6)) + ), + + TP_fast_assign( + __entry->cm_id = id_priv->res.id; + __entry->pd_id = pd->res.id; + __entry->tos = id_priv->tos; + __entry->send_wr = qp_init_attr->cap.max_send_wr; + __entry->recv_wr = qp_init_attr->cap.max_recv_wr; + __entry->rc = rc; + if (!rc) { + __entry->qp_num = id_priv->qp_num; + __entry->qp_type = id_priv->id.qp_type; + } else { + __entry->qp_num = 0; + __entry->qp_type = 0; + } + memcpy(__entry->srcaddr, &id_priv->id.route.addr.src_addr, + sizeof(struct sockaddr_in6)); + memcpy(__entry->dstaddr, &id_priv->id.route.addr.dst_addr, + sizeof(struct sockaddr_in6)); + ), + + TP_printk("cm.id=%u src=%pISpc dst=%pISpc tos=%u pd.id=%u qp_type=%s" + " send_wr=%u recv_wr=%u qp_num=%u rc=%d", + __entry->cm_id, __entry->srcaddr, __entry->dstaddr, + __entry->tos, __entry->pd_id, + rdma_show_qp_type(__entry->qp_type), __entry->send_wr, + __entry->recv_wr, __entry->qp_num, __entry->rc + ) +); + +TRACE_EVENT(cm_req_handler, + TP_PROTO( + const struct rdma_id_private *id_priv, + int event + ), + + TP_ARGS(id_priv, event), + + TP_STRUCT__entry( + __field(u32, cm_id) + __field(u32, tos) + __field(unsigned long, event) + __array(unsigned char, srcaddr, sizeof(struct sockaddr_in6)) + __array(unsigned char, dstaddr, sizeof(struct sockaddr_in6)) + ), + + TP_fast_assign( + __entry->cm_id = id_priv->res.id; + __entry->tos = id_priv->tos; + __entry->event = event; + memcpy(__entry->srcaddr, &id_priv->id.route.addr.src_addr, + sizeof(struct sockaddr_in6)); + memcpy(__entry->dstaddr, &id_priv->id.route.addr.dst_addr, + sizeof(struct sockaddr_in6)); + ), + + TP_printk("cm.id=%u src=%pISpc dst=%pISpc tos=%u %s (%lu)", + __entry->cm_id, __entry->srcaddr, __entry->dstaddr, __entry->tos, + rdma_show_ib_cm_event(__entry->event), __entry->event + ) +); + +TRACE_EVENT(cm_event_handler, + TP_PROTO( + const struct rdma_id_private *id_priv, + const struct rdma_cm_event *event + ), + + TP_ARGS(id_priv, event), + + TP_STRUCT__entry( + __field(u32, cm_id) + __field(u32, tos) + __field(unsigned long, event) + __field(int, status) + __array(unsigned char, srcaddr, sizeof(struct sockaddr_in6)) + __array(unsigned char, dstaddr, sizeof(struct sockaddr_in6)) + ), + + TP_fast_assign( + __entry->cm_id = id_priv->res.id; + __entry->tos = id_priv->tos; + __entry->event = event->event; + __entry->status = event->status; + memcpy(__entry->srcaddr, &id_priv->id.route.addr.src_addr, + sizeof(struct sockaddr_in6)); + memcpy(__entry->dstaddr, &id_priv->id.route.addr.dst_addr, + sizeof(struct sockaddr_in6)); + ), + + TP_printk("cm.id=%u src=%pISpc dst=%pISpc tos=%u %s (%lu/%d)", + __entry->cm_id, __entry->srcaddr, __entry->dstaddr, __entry->tos, + rdma_show_cm_event(__entry->event), __entry->event, + __entry->status + ) +); + +TRACE_EVENT(cm_event_done, + TP_PROTO( + const struct rdma_id_private *id_priv, + const struct rdma_cm_event *event, + int result + ), + + TP_ARGS(id_priv, event, result), + + TP_STRUCT__entry( + __field(u32, cm_id) + __field(u32, tos) + __field(unsigned long, event) + __field(int, result) + __array(unsigned char, srcaddr, sizeof(struct sockaddr_in6)) + __array(unsigned char, dstaddr, sizeof(struct sockaddr_in6)) + ), + + TP_fast_assign( + __entry->cm_id = id_priv->res.id; + __entry->tos = id_priv->tos; + __entry->event = event->event; + __entry->result = result; + memcpy(__entry->srcaddr, &id_priv->id.route.addr.src_addr, + sizeof(struct sockaddr_in6)); + memcpy(__entry->dstaddr, &id_priv->id.route.addr.dst_addr, + sizeof(struct sockaddr_in6)); + ), + + TP_printk("cm.id=%u src=%pISpc dst=%pISpc tos=%u %s consumer returns %d", + __entry->cm_id, __entry->srcaddr, __entry->dstaddr, __entry->tos, + rdma_show_cm_event(__entry->event), __entry->result + ) +); + +DECLARE_EVENT_CLASS(cma_client_class, + TP_PROTO( + const struct ib_device *device + ), + + TP_ARGS(device), + + TP_STRUCT__entry( + __string(name, device->name) + ), + + TP_fast_assign( + __assign_str(name, device->name); + ), + + TP_printk("device name=%s", + __get_str(name) + ) +); + +#define DEFINE_CMA_CLIENT_EVENT(name) \ + DEFINE_EVENT(cma_client_class, cm_##name, \ + TP_PROTO( \ + const struct ib_device *device \ + ), \ + TP_ARGS(device)) + +DEFINE_CMA_CLIENT_EVENT(add_one); +DEFINE_CMA_CLIENT_EVENT(remove_one); + +#endif /* _TRACE_RDMA_CMA_H */ + +#undef TRACE_INCLUDE_PATH +#define TRACE_INCLUDE_PATH . +#define TRACE_INCLUDE_FILE cma_trace + +#include <trace/define_trace.h> diff --git a/drivers/infiniband/core/core_priv.h b/drivers/infiniband/core/core_priv.h new file mode 100644 index 000000000..e84b0feda --- /dev/null +++ b/drivers/infiniband/core/core_priv.h @@ -0,0 +1,417 @@ +/* + * Copyright (c) 2004 Topspin Communications. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#ifndef _CORE_PRIV_H +#define _CORE_PRIV_H + +#include <linux/list.h> +#include <linux/spinlock.h> +#include <linux/cgroup_rdma.h> +#include <net/net_namespace.h> +#include <net/netns/generic.h> + +#include <rdma/ib_verbs.h> +#include <rdma/opa_addr.h> +#include <rdma/ib_mad.h> +#include <rdma/restrack.h> +#include "mad_priv.h" +#include "restrack.h" + +/* Total number of ports combined across all struct ib_devices's */ +#define RDMA_MAX_PORTS 8192 + +struct pkey_index_qp_list { + struct list_head pkey_index_list; + u16 pkey_index; + /* Lock to hold while iterating the qp_list. */ + spinlock_t qp_list_lock; + struct list_head qp_list; +}; + +/** + * struct rdma_dev_net - rdma net namespace metadata for a net + * @nl_sock: Pointer to netlink socket + * @net: Pointer to owner net namespace + * @id: xarray id to identify the net namespace. + */ +struct rdma_dev_net { + struct sock *nl_sock; + possible_net_t net; + u32 id; +}; + +extern const struct attribute_group ib_dev_attr_group; +extern bool ib_devices_shared_netns; +extern unsigned int rdma_dev_net_id; + +static inline struct rdma_dev_net *rdma_net_to_dev_net(struct net *net) +{ + return net_generic(net, rdma_dev_net_id); +} + +int ib_device_register_sysfs(struct ib_device *device); +void ib_device_unregister_sysfs(struct ib_device *device); +int ib_device_rename(struct ib_device *ibdev, const char *name); +int ib_device_set_dim(struct ib_device *ibdev, u8 use_dim); + +typedef void (*roce_netdev_callback)(struct ib_device *device, u8 port, + struct net_device *idev, void *cookie); + +typedef bool (*roce_netdev_filter)(struct ib_device *device, u8 port, + struct net_device *idev, void *cookie); + +struct net_device *ib_device_get_netdev(struct ib_device *ib_dev, + unsigned int port); + +void ib_enum_roce_netdev(struct ib_device *ib_dev, + roce_netdev_filter filter, + void *filter_cookie, + roce_netdev_callback cb, + void *cookie); +void ib_enum_all_roce_netdevs(roce_netdev_filter filter, + void *filter_cookie, + roce_netdev_callback cb, + void *cookie); + +typedef int (*nldev_callback)(struct ib_device *device, + struct sk_buff *skb, + struct netlink_callback *cb, + unsigned int idx); + +int ib_enum_all_devs(nldev_callback nldev_cb, struct sk_buff *skb, + struct netlink_callback *cb); + +struct ib_client_nl_info { + struct sk_buff *nl_msg; + struct device *cdev; + unsigned int port; + u64 abi; +}; +int ib_get_client_nl_info(struct ib_device *ibdev, const char *client_name, + struct ib_client_nl_info *res); + +enum ib_cache_gid_default_mode { + IB_CACHE_GID_DEFAULT_MODE_SET, + IB_CACHE_GID_DEFAULT_MODE_DELETE +}; + +int ib_cache_gid_parse_type_str(const char *buf); + +const char *ib_cache_gid_type_str(enum ib_gid_type gid_type); + +void ib_cache_gid_set_default_gid(struct ib_device *ib_dev, u8 port, + struct net_device *ndev, + unsigned long gid_type_mask, + enum ib_cache_gid_default_mode mode); + +int ib_cache_gid_add(struct ib_device *ib_dev, u8 port, + union ib_gid *gid, struct ib_gid_attr *attr); + +int ib_cache_gid_del(struct ib_device *ib_dev, u8 port, + union ib_gid *gid, struct ib_gid_attr *attr); + +int ib_cache_gid_del_all_netdev_gids(struct ib_device *ib_dev, u8 port, + struct net_device *ndev); + +int roce_gid_mgmt_init(void); +void roce_gid_mgmt_cleanup(void); + +unsigned long roce_gid_type_mask_support(struct ib_device *ib_dev, u8 port); + +int ib_cache_setup_one(struct ib_device *device); +void ib_cache_cleanup_one(struct ib_device *device); +void ib_cache_release_one(struct ib_device *device); +void ib_dispatch_event_clients(struct ib_event *event); + +#ifdef CONFIG_CGROUP_RDMA +void ib_device_register_rdmacg(struct ib_device *device); +void ib_device_unregister_rdmacg(struct ib_device *device); + +int ib_rdmacg_try_charge(struct ib_rdmacg_object *cg_obj, + struct ib_device *device, + enum rdmacg_resource_type resource_index); + +void ib_rdmacg_uncharge(struct ib_rdmacg_object *cg_obj, + struct ib_device *device, + enum rdmacg_resource_type resource_index); +#else +static inline void ib_device_register_rdmacg(struct ib_device *device) +{ +} + +static inline void ib_device_unregister_rdmacg(struct ib_device *device) +{ +} + +static inline int ib_rdmacg_try_charge(struct ib_rdmacg_object *cg_obj, + struct ib_device *device, + enum rdmacg_resource_type resource_index) +{ + return 0; +} + +static inline void ib_rdmacg_uncharge(struct ib_rdmacg_object *cg_obj, + struct ib_device *device, + enum rdmacg_resource_type resource_index) +{ +} +#endif + +static inline bool rdma_is_upper_dev_rcu(struct net_device *dev, + struct net_device *upper) +{ + return netdev_has_upper_dev_all_rcu(dev, upper); +} + +int addr_init(void); +void addr_cleanup(void); + +int ib_mad_init(void); +void ib_mad_cleanup(void); + +int ib_sa_init(void); +void ib_sa_cleanup(void); + +void rdma_nl_init(void); +void rdma_nl_exit(void); + +int ib_nl_handle_resolve_resp(struct sk_buff *skb, + struct nlmsghdr *nlh, + struct netlink_ext_ack *extack); +int ib_nl_handle_set_timeout(struct sk_buff *skb, + struct nlmsghdr *nlh, + struct netlink_ext_ack *extack); +int ib_nl_handle_ip_res_resp(struct sk_buff *skb, + struct nlmsghdr *nlh, + struct netlink_ext_ack *extack); + +int ib_get_cached_subnet_prefix(struct ib_device *device, + u8 port_num, + u64 *sn_pfx); + +#ifdef CONFIG_SECURITY_INFINIBAND +void ib_security_release_port_pkey_list(struct ib_device *device); + +void ib_security_cache_change(struct ib_device *device, + u8 port_num, + u64 subnet_prefix); + +int ib_security_modify_qp(struct ib_qp *qp, + struct ib_qp_attr *qp_attr, + int qp_attr_mask, + struct ib_udata *udata); + +int ib_create_qp_security(struct ib_qp *qp, struct ib_device *dev); +void ib_destroy_qp_security_begin(struct ib_qp_security *sec); +void ib_destroy_qp_security_abort(struct ib_qp_security *sec); +void ib_destroy_qp_security_end(struct ib_qp_security *sec); +int ib_open_shared_qp_security(struct ib_qp *qp, struct ib_device *dev); +void ib_close_shared_qp_security(struct ib_qp_security *sec); +int ib_mad_agent_security_setup(struct ib_mad_agent *agent, + enum ib_qp_type qp_type); +void ib_mad_agent_security_cleanup(struct ib_mad_agent *agent); +int ib_mad_enforce_security(struct ib_mad_agent_private *map, u16 pkey_index); +void ib_mad_agent_security_change(void); +#else +static inline void ib_security_release_port_pkey_list(struct ib_device *device) +{ +} + +static inline void ib_security_cache_change(struct ib_device *device, + u8 port_num, + u64 subnet_prefix) +{ +} + +static inline int ib_security_modify_qp(struct ib_qp *qp, + struct ib_qp_attr *qp_attr, + int qp_attr_mask, + struct ib_udata *udata) +{ + return qp->device->ops.modify_qp(qp->real_qp, + qp_attr, + qp_attr_mask, + udata); +} + +static inline int ib_create_qp_security(struct ib_qp *qp, + struct ib_device *dev) +{ + return 0; +} + +static inline void ib_destroy_qp_security_begin(struct ib_qp_security *sec) +{ +} + +static inline void ib_destroy_qp_security_abort(struct ib_qp_security *sec) +{ +} + +static inline void ib_destroy_qp_security_end(struct ib_qp_security *sec) +{ +} + +static inline int ib_open_shared_qp_security(struct ib_qp *qp, + struct ib_device *dev) +{ + return 0; +} + +static inline void ib_close_shared_qp_security(struct ib_qp_security *sec) +{ +} + +static inline int ib_mad_agent_security_setup(struct ib_mad_agent *agent, + enum ib_qp_type qp_type) +{ + return 0; +} + +static inline void ib_mad_agent_security_cleanup(struct ib_mad_agent *agent) +{ +} + +static inline int ib_mad_enforce_security(struct ib_mad_agent_private *map, + u16 pkey_index) +{ + return 0; +} + +static inline void ib_mad_agent_security_change(void) +{ +} +#endif + +struct ib_device *ib_device_get_by_index(const struct net *net, u32 index); + +/* RDMA device netlink */ +void nldev_init(void); +void nldev_exit(void); + +static inline struct ib_qp *_ib_create_qp(struct ib_device *dev, + struct ib_pd *pd, + struct ib_qp_init_attr *attr, + struct ib_udata *udata, + struct ib_uqp_object *uobj) +{ + enum ib_qp_type qp_type = attr->qp_type; + struct ib_qp *qp; + bool is_xrc; + + if (!dev->ops.create_qp) + return ERR_PTR(-EOPNOTSUPP); + + qp = dev->ops.create_qp(pd, attr, udata); + if (IS_ERR(qp)) + return qp; + + qp->device = dev; + qp->pd = pd; + qp->uobject = uobj; + qp->real_qp = qp; + + qp->qp_type = attr->qp_type; + qp->rwq_ind_tbl = attr->rwq_ind_tbl; + qp->send_cq = attr->send_cq; + qp->recv_cq = attr->recv_cq; + qp->srq = attr->srq; + qp->rwq_ind_tbl = attr->rwq_ind_tbl; + qp->event_handler = attr->event_handler; + + atomic_set(&qp->usecnt, 0); + spin_lock_init(&qp->mr_lock); + INIT_LIST_HEAD(&qp->rdma_mrs); + INIT_LIST_HEAD(&qp->sig_mrs); + + rdma_restrack_new(&qp->res, RDMA_RESTRACK_QP); + /* + * We don't track XRC QPs for now, because they don't have PD + * and more importantly they are created internaly by driver, + * see mlx5 create_dev_resources() as an example. + */ + is_xrc = qp_type == IB_QPT_XRC_INI || qp_type == IB_QPT_XRC_TGT; + if ((qp_type < IB_QPT_MAX && !is_xrc) || qp_type == IB_QPT_DRIVER) { + rdma_restrack_parent_name(&qp->res, &pd->res); + rdma_restrack_add(&qp->res); + } + return qp; +} + +struct rdma_dev_addr; +int rdma_resolve_ip_route(struct sockaddr *src_addr, + const struct sockaddr *dst_addr, + struct rdma_dev_addr *addr); + +int rdma_addr_find_l2_eth_by_grh(const union ib_gid *sgid, + const union ib_gid *dgid, + u8 *dmac, const struct ib_gid_attr *sgid_attr, + int *hoplimit); +void rdma_copy_src_l2_addr(struct rdma_dev_addr *dev_addr, + const struct net_device *dev); + +struct sa_path_rec; +int roce_resolve_route_from_path(struct sa_path_rec *rec, + const struct ib_gid_attr *attr); + +struct net_device *rdma_read_gid_attr_ndev_rcu(const struct ib_gid_attr *attr); + +void ib_free_port_attrs(struct ib_core_device *coredev); +int ib_setup_port_attrs(struct ib_core_device *coredev); + +int rdma_compatdev_set(u8 enable); + +int ib_port_register_module_stat(struct ib_device *device, u8 port_num, + struct kobject *kobj, struct kobj_type *ktype, + const char *name); +void ib_port_unregister_module_stat(struct kobject *kobj); + +int ib_device_set_netns_put(struct sk_buff *skb, + struct ib_device *dev, u32 ns_fd); + +int rdma_nl_net_init(struct rdma_dev_net *rnet); +void rdma_nl_net_exit(struct rdma_dev_net *rnet); + +struct rdma_umap_priv { + struct vm_area_struct *vma; + struct list_head list; + struct rdma_user_mmap_entry *entry; +}; + +void rdma_umap_priv_init(struct rdma_umap_priv *priv, + struct vm_area_struct *vma, + struct rdma_user_mmap_entry *entry); + +void ib_cq_pool_init(struct ib_device *dev); +void ib_cq_pool_destroy(struct ib_device *dev); + +#endif /* _CORE_PRIV_H */ diff --git a/drivers/infiniband/core/counters.c b/drivers/infiniband/core/counters.c new file mode 100644 index 000000000..e4ff0d332 --- /dev/null +++ b/drivers/infiniband/core/counters.c @@ -0,0 +1,639 @@ +// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB +/* + * Copyright (c) 2019 Mellanox Technologies. All rights reserved. + */ +#include <rdma/ib_verbs.h> +#include <rdma/rdma_counter.h> + +#include "core_priv.h" +#include "restrack.h" + +#define ALL_AUTO_MODE_MASKS (RDMA_COUNTER_MASK_QP_TYPE | RDMA_COUNTER_MASK_PID) + +static int __counter_set_mode(struct rdma_counter_mode *curr, + enum rdma_nl_counter_mode new_mode, + enum rdma_nl_counter_mask new_mask) +{ + if ((new_mode == RDMA_COUNTER_MODE_AUTO) && + ((new_mask & (~ALL_AUTO_MODE_MASKS)) || + (curr->mode != RDMA_COUNTER_MODE_NONE))) + return -EINVAL; + + curr->mode = new_mode; + curr->mask = new_mask; + return 0; +} + +/** + * rdma_counter_set_auto_mode() - Turn on/off per-port auto mode + * + * When @on is true, the @mask must be set; When @on is false, it goes + * into manual mode if there's any counter, so that the user is able to + * manually access them. + */ +int rdma_counter_set_auto_mode(struct ib_device *dev, u8 port, + bool on, enum rdma_nl_counter_mask mask) +{ + struct rdma_port_counter *port_counter; + int ret; + + port_counter = &dev->port_data[port].port_counter; + if (!port_counter->hstats) + return -EOPNOTSUPP; + + mutex_lock(&port_counter->lock); + if (on) { + ret = __counter_set_mode(&port_counter->mode, + RDMA_COUNTER_MODE_AUTO, mask); + } else { + if (port_counter->mode.mode != RDMA_COUNTER_MODE_AUTO) { + ret = -EINVAL; + goto out; + } + + if (port_counter->num_counters) + ret = __counter_set_mode(&port_counter->mode, + RDMA_COUNTER_MODE_MANUAL, 0); + else + ret = __counter_set_mode(&port_counter->mode, + RDMA_COUNTER_MODE_NONE, 0); + } + +out: + mutex_unlock(&port_counter->lock); + return ret; +} + +static struct rdma_counter *rdma_counter_alloc(struct ib_device *dev, u8 port, + enum rdma_nl_counter_mode mode) +{ + struct rdma_port_counter *port_counter; + struct rdma_counter *counter; + int ret; + + if (!dev->ops.counter_dealloc || !dev->ops.counter_alloc_stats) + return NULL; + + counter = kzalloc(sizeof(*counter), GFP_KERNEL); + if (!counter) + return NULL; + + counter->device = dev; + counter->port = port; + + rdma_restrack_new(&counter->res, RDMA_RESTRACK_COUNTER); + counter->stats = dev->ops.counter_alloc_stats(counter); + if (!counter->stats) + goto err_stats; + + port_counter = &dev->port_data[port].port_counter; + mutex_lock(&port_counter->lock); + if (mode == RDMA_COUNTER_MODE_MANUAL) { + ret = __counter_set_mode(&port_counter->mode, + RDMA_COUNTER_MODE_MANUAL, 0); + if (ret) + goto err_mode; + } + + port_counter->num_counters++; + mutex_unlock(&port_counter->lock); + + counter->mode.mode = mode; + kref_init(&counter->kref); + mutex_init(&counter->lock); + + return counter; + +err_mode: + mutex_unlock(&port_counter->lock); + kfree(counter->stats); +err_stats: + rdma_restrack_put(&counter->res); + kfree(counter); + return NULL; +} + +static void rdma_counter_free(struct rdma_counter *counter) +{ + struct rdma_port_counter *port_counter; + + port_counter = &counter->device->port_data[counter->port].port_counter; + mutex_lock(&port_counter->lock); + port_counter->num_counters--; + if (!port_counter->num_counters && + (port_counter->mode.mode == RDMA_COUNTER_MODE_MANUAL)) + __counter_set_mode(&port_counter->mode, RDMA_COUNTER_MODE_NONE, + 0); + + mutex_unlock(&port_counter->lock); + + rdma_restrack_del(&counter->res); + kfree(counter->stats); + kfree(counter); +} + +static void auto_mode_init_counter(struct rdma_counter *counter, + const struct ib_qp *qp, + enum rdma_nl_counter_mask new_mask) +{ + struct auto_mode_param *param = &counter->mode.param; + + counter->mode.mode = RDMA_COUNTER_MODE_AUTO; + counter->mode.mask = new_mask; + + if (new_mask & RDMA_COUNTER_MASK_QP_TYPE) + param->qp_type = qp->qp_type; +} + +static bool auto_mode_match(struct ib_qp *qp, struct rdma_counter *counter, + enum rdma_nl_counter_mask auto_mask) +{ + struct auto_mode_param *param = &counter->mode.param; + bool match = true; + + if (auto_mask & RDMA_COUNTER_MASK_QP_TYPE) + match &= (param->qp_type == qp->qp_type); + + if (auto_mask & RDMA_COUNTER_MASK_PID) + match &= (task_pid_nr(counter->res.task) == + task_pid_nr(qp->res.task)); + + return match; +} + +static int __rdma_counter_bind_qp(struct rdma_counter *counter, + struct ib_qp *qp) +{ + int ret; + + if (qp->counter) + return -EINVAL; + + if (!qp->device->ops.counter_bind_qp) + return -EOPNOTSUPP; + + mutex_lock(&counter->lock); + ret = qp->device->ops.counter_bind_qp(counter, qp); + mutex_unlock(&counter->lock); + + return ret; +} + +static int __rdma_counter_unbind_qp(struct ib_qp *qp) +{ + struct rdma_counter *counter = qp->counter; + int ret; + + if (!qp->device->ops.counter_unbind_qp) + return -EOPNOTSUPP; + + mutex_lock(&counter->lock); + ret = qp->device->ops.counter_unbind_qp(qp); + mutex_unlock(&counter->lock); + + return ret; +} + +static void counter_history_stat_update(struct rdma_counter *counter) +{ + struct ib_device *dev = counter->device; + struct rdma_port_counter *port_counter; + int i; + + port_counter = &dev->port_data[counter->port].port_counter; + if (!port_counter->hstats) + return; + + rdma_counter_query_stats(counter); + + for (i = 0; i < counter->stats->num_counters; i++) + port_counter->hstats->value[i] += counter->stats->value[i]; +} + +/** + * rdma_get_counter_auto_mode - Find the counter that @qp should be bound + * with in auto mode + * + * Return: The counter (with ref-count increased) if found + */ +static struct rdma_counter *rdma_get_counter_auto_mode(struct ib_qp *qp, + u8 port) +{ + struct rdma_port_counter *port_counter; + struct rdma_counter *counter = NULL; + struct ib_device *dev = qp->device; + struct rdma_restrack_entry *res; + struct rdma_restrack_root *rt; + unsigned long id = 0; + + port_counter = &dev->port_data[port].port_counter; + rt = &dev->res[RDMA_RESTRACK_COUNTER]; + xa_lock(&rt->xa); + xa_for_each(&rt->xa, id, res) { + counter = container_of(res, struct rdma_counter, res); + if ((counter->device != qp->device) || (counter->port != port)) + goto next; + + if (auto_mode_match(qp, counter, port_counter->mode.mask)) + break; +next: + counter = NULL; + } + + if (counter && !kref_get_unless_zero(&counter->kref)) + counter = NULL; + + xa_unlock(&rt->xa); + return counter; +} + +static void rdma_counter_res_add(struct rdma_counter *counter, + struct ib_qp *qp) +{ + rdma_restrack_parent_name(&counter->res, &qp->res); + rdma_restrack_add(&counter->res); +} + +static void counter_release(struct kref *kref) +{ + struct rdma_counter *counter; + + counter = container_of(kref, struct rdma_counter, kref); + counter_history_stat_update(counter); + counter->device->ops.counter_dealloc(counter); + rdma_counter_free(counter); +} + +/** + * rdma_counter_bind_qp_auto - Check and bind the QP to a counter base on + * the auto-mode rule + */ +int rdma_counter_bind_qp_auto(struct ib_qp *qp, u8 port) +{ + struct rdma_port_counter *port_counter; + struct ib_device *dev = qp->device; + struct rdma_counter *counter; + int ret; + + if (!qp->res.valid || rdma_is_kernel_res(&qp->res)) + return 0; + + if (!rdma_is_port_valid(dev, port)) + return -EINVAL; + + port_counter = &dev->port_data[port].port_counter; + if (port_counter->mode.mode != RDMA_COUNTER_MODE_AUTO) + return 0; + + counter = rdma_get_counter_auto_mode(qp, port); + if (counter) { + ret = __rdma_counter_bind_qp(counter, qp); + if (ret) { + kref_put(&counter->kref, counter_release); + return ret; + } + } else { + counter = rdma_counter_alloc(dev, port, RDMA_COUNTER_MODE_AUTO); + if (!counter) + return -ENOMEM; + + auto_mode_init_counter(counter, qp, port_counter->mode.mask); + + ret = __rdma_counter_bind_qp(counter, qp); + if (ret) { + rdma_counter_free(counter); + return ret; + } + + rdma_counter_res_add(counter, qp); + } + + return 0; +} + +/** + * rdma_counter_unbind_qp - Unbind a qp from a counter + * @force: + * true - Decrease the counter ref-count anyway (e.g., qp destroy) + */ +int rdma_counter_unbind_qp(struct ib_qp *qp, bool force) +{ + struct rdma_counter *counter = qp->counter; + int ret; + + if (!counter) + return -EINVAL; + + ret = __rdma_counter_unbind_qp(qp); + if (ret && !force) + return ret; + + kref_put(&counter->kref, counter_release); + return 0; +} + +int rdma_counter_query_stats(struct rdma_counter *counter) +{ + struct ib_device *dev = counter->device; + int ret; + + if (!dev->ops.counter_update_stats) + return -EINVAL; + + mutex_lock(&counter->lock); + ret = dev->ops.counter_update_stats(counter); + mutex_unlock(&counter->lock); + + return ret; +} + +static u64 get_running_counters_hwstat_sum(struct ib_device *dev, + u8 port, u32 index) +{ + struct rdma_restrack_entry *res; + struct rdma_restrack_root *rt; + struct rdma_counter *counter; + unsigned long id = 0; + u64 sum = 0; + + rt = &dev->res[RDMA_RESTRACK_COUNTER]; + xa_lock(&rt->xa); + xa_for_each(&rt->xa, id, res) { + if (!rdma_restrack_get(res)) + continue; + + xa_unlock(&rt->xa); + + counter = container_of(res, struct rdma_counter, res); + if ((counter->device != dev) || (counter->port != port) || + rdma_counter_query_stats(counter)) + goto next; + + sum += counter->stats->value[index]; + +next: + xa_lock(&rt->xa); + rdma_restrack_put(res); + } + + xa_unlock(&rt->xa); + return sum; +} + +/** + * rdma_counter_get_hwstat_value() - Get the sum value of all counters on a + * specific port, including the running ones and history data + */ +u64 rdma_counter_get_hwstat_value(struct ib_device *dev, u8 port, u32 index) +{ + struct rdma_port_counter *port_counter; + u64 sum; + + port_counter = &dev->port_data[port].port_counter; + if (!port_counter->hstats) + return 0; + + sum = get_running_counters_hwstat_sum(dev, port, index); + sum += port_counter->hstats->value[index]; + + return sum; +} + +static struct ib_qp *rdma_counter_get_qp(struct ib_device *dev, u32 qp_num) +{ + struct rdma_restrack_entry *res = NULL; + struct ib_qp *qp = NULL; + + res = rdma_restrack_get_byid(dev, RDMA_RESTRACK_QP, qp_num); + if (IS_ERR(res)) + return NULL; + + qp = container_of(res, struct ib_qp, res); + if (qp->qp_type == IB_QPT_RAW_PACKET && !capable(CAP_NET_RAW)) + goto err; + + return qp; + +err: + rdma_restrack_put(res); + return NULL; +} + +static int rdma_counter_bind_qp_manual(struct rdma_counter *counter, + struct ib_qp *qp) +{ + if ((counter->device != qp->device) || (counter->port != qp->port)) + return -EINVAL; + + return __rdma_counter_bind_qp(counter, qp); +} + +static struct rdma_counter *rdma_get_counter_by_id(struct ib_device *dev, + u32 counter_id) +{ + struct rdma_restrack_entry *res; + struct rdma_counter *counter; + + res = rdma_restrack_get_byid(dev, RDMA_RESTRACK_COUNTER, counter_id); + if (IS_ERR(res)) + return NULL; + + counter = container_of(res, struct rdma_counter, res); + kref_get(&counter->kref); + rdma_restrack_put(res); + + return counter; +} + +/** + * rdma_counter_bind_qpn() - Bind QP @qp_num to counter @counter_id + */ +int rdma_counter_bind_qpn(struct ib_device *dev, u8 port, + u32 qp_num, u32 counter_id) +{ + struct rdma_port_counter *port_counter; + struct rdma_counter *counter; + struct ib_qp *qp; + int ret; + + port_counter = &dev->port_data[port].port_counter; + if (port_counter->mode.mode == RDMA_COUNTER_MODE_AUTO) + return -EINVAL; + + qp = rdma_counter_get_qp(dev, qp_num); + if (!qp) + return -ENOENT; + + counter = rdma_get_counter_by_id(dev, counter_id); + if (!counter) { + ret = -ENOENT; + goto err; + } + + if (rdma_is_kernel_res(&counter->res) != rdma_is_kernel_res(&qp->res)) { + ret = -EINVAL; + goto err_task; + } + + ret = rdma_counter_bind_qp_manual(counter, qp); + if (ret) + goto err_task; + + rdma_restrack_put(&qp->res); + return 0; + +err_task: + kref_put(&counter->kref, counter_release); +err: + rdma_restrack_put(&qp->res); + return ret; +} + +/** + * rdma_counter_bind_qpn_alloc() - Alloc a counter and bind QP @qp_num to it + * The id of new counter is returned in @counter_id + */ +int rdma_counter_bind_qpn_alloc(struct ib_device *dev, u8 port, + u32 qp_num, u32 *counter_id) +{ + struct rdma_port_counter *port_counter; + struct rdma_counter *counter; + struct ib_qp *qp; + int ret; + + if (!rdma_is_port_valid(dev, port)) + return -EINVAL; + + port_counter = &dev->port_data[port].port_counter; + if (!port_counter->hstats) + return -EOPNOTSUPP; + + if (port_counter->mode.mode == RDMA_COUNTER_MODE_AUTO) + return -EINVAL; + + qp = rdma_counter_get_qp(dev, qp_num); + if (!qp) + return -ENOENT; + + if (rdma_is_port_valid(dev, qp->port) && (qp->port != port)) { + ret = -EINVAL; + goto err; + } + + counter = rdma_counter_alloc(dev, port, RDMA_COUNTER_MODE_MANUAL); + if (!counter) { + ret = -ENOMEM; + goto err; + } + + ret = rdma_counter_bind_qp_manual(counter, qp); + if (ret) + goto err_bind; + + if (counter_id) + *counter_id = counter->id; + + rdma_counter_res_add(counter, qp); + + rdma_restrack_put(&qp->res); + return ret; + +err_bind: + rdma_counter_free(counter); +err: + rdma_restrack_put(&qp->res); + return ret; +} + +/** + * rdma_counter_unbind_qpn() - Unbind QP @qp_num from a counter + */ +int rdma_counter_unbind_qpn(struct ib_device *dev, u8 port, + u32 qp_num, u32 counter_id) +{ + struct rdma_port_counter *port_counter; + struct ib_qp *qp; + int ret; + + if (!rdma_is_port_valid(dev, port)) + return -EINVAL; + + qp = rdma_counter_get_qp(dev, qp_num); + if (!qp) + return -ENOENT; + + if (rdma_is_port_valid(dev, qp->port) && (qp->port != port)) { + ret = -EINVAL; + goto out; + } + + port_counter = &dev->port_data[port].port_counter; + if (!qp->counter || qp->counter->id != counter_id || + port_counter->mode.mode != RDMA_COUNTER_MODE_MANUAL) { + ret = -EINVAL; + goto out; + } + + ret = rdma_counter_unbind_qp(qp, false); + +out: + rdma_restrack_put(&qp->res); + return ret; +} + +int rdma_counter_get_mode(struct ib_device *dev, u8 port, + enum rdma_nl_counter_mode *mode, + enum rdma_nl_counter_mask *mask) +{ + struct rdma_port_counter *port_counter; + + port_counter = &dev->port_data[port].port_counter; + *mode = port_counter->mode.mode; + *mask = port_counter->mode.mask; + + return 0; +} + +void rdma_counter_init(struct ib_device *dev) +{ + struct rdma_port_counter *port_counter; + u32 port, i; + + if (!dev->port_data) + return; + + rdma_for_each_port(dev, port) { + port_counter = &dev->port_data[port].port_counter; + port_counter->mode.mode = RDMA_COUNTER_MODE_NONE; + mutex_init(&port_counter->lock); + + if (!dev->ops.alloc_hw_stats) + continue; + + port_counter->hstats = dev->ops.alloc_hw_stats(dev, port); + if (!port_counter->hstats) + goto fail; + } + + return; + +fail: + for (i = port; i >= rdma_start_port(dev); i--) { + port_counter = &dev->port_data[port].port_counter; + kfree(port_counter->hstats); + port_counter->hstats = NULL; + mutex_destroy(&port_counter->lock); + } +} + +void rdma_counter_release(struct ib_device *dev) +{ + struct rdma_port_counter *port_counter; + u32 port; + + rdma_for_each_port(dev, port) { + port_counter = &dev->port_data[port].port_counter; + kfree(port_counter->hstats); + mutex_destroy(&port_counter->lock); + } +} diff --git a/drivers/infiniband/core/cq.c b/drivers/infiniband/core/cq.c new file mode 100644 index 000000000..12ebacf52 --- /dev/null +++ b/drivers/infiniband/core/cq.c @@ -0,0 +1,516 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * Copyright (c) 2015 HGST, a Western Digital Company. + */ +#include <linux/module.h> +#include <linux/err.h> +#include <linux/slab.h> +#include <rdma/ib_verbs.h> + +#include "core_priv.h" + +#include <trace/events/rdma_core.h> +/* Max size for shared CQ, may require tuning */ +#define IB_MAX_SHARED_CQ_SZ 4096U + +/* # of WCs to poll for with a single call to ib_poll_cq */ +#define IB_POLL_BATCH 16 +#define IB_POLL_BATCH_DIRECT 8 + +/* # of WCs to iterate over before yielding */ +#define IB_POLL_BUDGET_IRQ 256 +#define IB_POLL_BUDGET_WORKQUEUE 65536 + +#define IB_POLL_FLAGS \ + (IB_CQ_NEXT_COMP | IB_CQ_REPORT_MISSED_EVENTS) + +static const struct dim_cq_moder +rdma_dim_prof[RDMA_DIM_PARAMS_NUM_PROFILES] = { + {1, 0, 1, 0}, + {1, 0, 4, 0}, + {2, 0, 4, 0}, + {2, 0, 8, 0}, + {4, 0, 8, 0}, + {16, 0, 8, 0}, + {16, 0, 16, 0}, + {32, 0, 16, 0}, + {32, 0, 32, 0}, +}; + +static void ib_cq_rdma_dim_work(struct work_struct *w) +{ + struct dim *dim = container_of(w, struct dim, work); + struct ib_cq *cq = dim->priv; + + u16 usec = rdma_dim_prof[dim->profile_ix].usec; + u16 comps = rdma_dim_prof[dim->profile_ix].comps; + + dim->state = DIM_START_MEASURE; + + trace_cq_modify(cq, comps, usec); + cq->device->ops.modify_cq(cq, comps, usec); +} + +static void rdma_dim_init(struct ib_cq *cq) +{ + struct dim *dim; + + if (!cq->device->ops.modify_cq || !cq->device->use_cq_dim || + cq->poll_ctx == IB_POLL_DIRECT) + return; + + dim = kzalloc(sizeof(struct dim), GFP_KERNEL); + if (!dim) + return; + + dim->state = DIM_START_MEASURE; + dim->tune_state = DIM_GOING_RIGHT; + dim->profile_ix = RDMA_DIM_START_PROFILE; + dim->priv = cq; + cq->dim = dim; + + INIT_WORK(&dim->work, ib_cq_rdma_dim_work); +} + +static void rdma_dim_destroy(struct ib_cq *cq) +{ + if (!cq->dim) + return; + + cancel_work_sync(&cq->dim->work); + kfree(cq->dim); +} + +static int __poll_cq(struct ib_cq *cq, int num_entries, struct ib_wc *wc) +{ + int rc; + + rc = ib_poll_cq(cq, num_entries, wc); + trace_cq_poll(cq, num_entries, rc); + return rc; +} + +static int __ib_process_cq(struct ib_cq *cq, int budget, struct ib_wc *wcs, + int batch) +{ + int i, n, completed = 0; + + trace_cq_process(cq); + + /* + * budget might be (-1) if the caller does not + * want to bound this call, thus we need unsigned + * minimum here. + */ + while ((n = __poll_cq(cq, min_t(u32, batch, + budget - completed), wcs)) > 0) { + for (i = 0; i < n; i++) { + struct ib_wc *wc = &wcs[i]; + + if (wc->wr_cqe) + wc->wr_cqe->done(cq, wc); + else + WARN_ON_ONCE(wc->status == IB_WC_SUCCESS); + } + + completed += n; + + if (n != batch || (budget != -1 && completed >= budget)) + break; + } + + return completed; +} + +/** + * ib_process_direct_cq - process a CQ in caller context + * @cq: CQ to process + * @budget: number of CQEs to poll for + * + * This function is used to process all outstanding CQ entries. + * It does not offload CQ processing to a different context and does + * not ask for completion interrupts from the HCA. + * Using direct processing on CQ with non IB_POLL_DIRECT type may trigger + * concurrent processing. + * + * Note: do not pass -1 as %budget unless it is guaranteed that the number + * of completions that will be processed is small. + */ +int ib_process_cq_direct(struct ib_cq *cq, int budget) +{ + struct ib_wc wcs[IB_POLL_BATCH_DIRECT]; + + return __ib_process_cq(cq, budget, wcs, IB_POLL_BATCH_DIRECT); +} +EXPORT_SYMBOL(ib_process_cq_direct); + +static void ib_cq_completion_direct(struct ib_cq *cq, void *private) +{ + WARN_ONCE(1, "got unsolicited completion for CQ 0x%p\n", cq); +} + +static int ib_poll_handler(struct irq_poll *iop, int budget) +{ + struct ib_cq *cq = container_of(iop, struct ib_cq, iop); + struct dim *dim = cq->dim; + int completed; + + completed = __ib_process_cq(cq, budget, cq->wc, IB_POLL_BATCH); + if (completed < budget) { + irq_poll_complete(&cq->iop); + if (ib_req_notify_cq(cq, IB_POLL_FLAGS) > 0) { + trace_cq_reschedule(cq); + irq_poll_sched(&cq->iop); + } + } + + if (dim) + rdma_dim(dim, completed); + + return completed; +} + +static void ib_cq_completion_softirq(struct ib_cq *cq, void *private) +{ + trace_cq_schedule(cq); + irq_poll_sched(&cq->iop); +} + +static void ib_cq_poll_work(struct work_struct *work) +{ + struct ib_cq *cq = container_of(work, struct ib_cq, work); + int completed; + + completed = __ib_process_cq(cq, IB_POLL_BUDGET_WORKQUEUE, cq->wc, + IB_POLL_BATCH); + if (completed >= IB_POLL_BUDGET_WORKQUEUE || + ib_req_notify_cq(cq, IB_POLL_FLAGS) > 0) + queue_work(cq->comp_wq, &cq->work); + else if (cq->dim) + rdma_dim(cq->dim, completed); +} + +static void ib_cq_completion_workqueue(struct ib_cq *cq, void *private) +{ + trace_cq_schedule(cq); + queue_work(cq->comp_wq, &cq->work); +} + +/** + * __ib_alloc_cq allocate a completion queue + * @dev: device to allocate the CQ for + * @private: driver private data, accessible from cq->cq_context + * @nr_cqe: number of CQEs to allocate + * @comp_vector: HCA completion vectors for this CQ + * @poll_ctx: context to poll the CQ from. + * @caller: module owner name. + * + * This is the proper interface to allocate a CQ for in-kernel users. A + * CQ allocated with this interface will automatically be polled from the + * specified context. The ULP must use wr->wr_cqe instead of wr->wr_id + * to use this CQ abstraction. + */ +struct ib_cq *__ib_alloc_cq(struct ib_device *dev, void *private, int nr_cqe, + int comp_vector, enum ib_poll_context poll_ctx, + const char *caller) +{ + struct ib_cq_init_attr cq_attr = { + .cqe = nr_cqe, + .comp_vector = comp_vector, + }; + struct ib_cq *cq; + int ret = -ENOMEM; + + cq = rdma_zalloc_drv_obj(dev, ib_cq); + if (!cq) + return ERR_PTR(ret); + + cq->device = dev; + cq->cq_context = private; + cq->poll_ctx = poll_ctx; + atomic_set(&cq->usecnt, 0); + cq->comp_vector = comp_vector; + + cq->wc = kmalloc_array(IB_POLL_BATCH, sizeof(*cq->wc), GFP_KERNEL); + if (!cq->wc) + goto out_free_cq; + + rdma_restrack_new(&cq->res, RDMA_RESTRACK_CQ); + rdma_restrack_set_name(&cq->res, caller); + + ret = dev->ops.create_cq(cq, &cq_attr, NULL); + if (ret) + goto out_free_wc; + + rdma_dim_init(cq); + + switch (cq->poll_ctx) { + case IB_POLL_DIRECT: + cq->comp_handler = ib_cq_completion_direct; + break; + case IB_POLL_SOFTIRQ: + cq->comp_handler = ib_cq_completion_softirq; + + irq_poll_init(&cq->iop, IB_POLL_BUDGET_IRQ, ib_poll_handler); + ib_req_notify_cq(cq, IB_CQ_NEXT_COMP); + break; + case IB_POLL_WORKQUEUE: + case IB_POLL_UNBOUND_WORKQUEUE: + cq->comp_handler = ib_cq_completion_workqueue; + INIT_WORK(&cq->work, ib_cq_poll_work); + ib_req_notify_cq(cq, IB_CQ_NEXT_COMP); + cq->comp_wq = (cq->poll_ctx == IB_POLL_WORKQUEUE) ? + ib_comp_wq : ib_comp_unbound_wq; + break; + default: + ret = -EINVAL; + goto out_destroy_cq; + } + + rdma_restrack_add(&cq->res); + trace_cq_alloc(cq, nr_cqe, comp_vector, poll_ctx); + return cq; + +out_destroy_cq: + rdma_dim_destroy(cq); + cq->device->ops.destroy_cq(cq, NULL); +out_free_wc: + rdma_restrack_put(&cq->res); + kfree(cq->wc); +out_free_cq: + kfree(cq); + trace_cq_alloc_error(nr_cqe, comp_vector, poll_ctx, ret); + return ERR_PTR(ret); +} +EXPORT_SYMBOL(__ib_alloc_cq); + +/** + * __ib_alloc_cq_any - allocate a completion queue + * @dev: device to allocate the CQ for + * @private: driver private data, accessible from cq->cq_context + * @nr_cqe: number of CQEs to allocate + * @poll_ctx: context to poll the CQ from + * @caller: module owner name + * + * Attempt to spread ULP Completion Queues over each device's interrupt + * vectors. A simple best-effort mechanism is used. + */ +struct ib_cq *__ib_alloc_cq_any(struct ib_device *dev, void *private, + int nr_cqe, enum ib_poll_context poll_ctx, + const char *caller) +{ + static atomic_t counter; + int comp_vector = 0; + + if (dev->num_comp_vectors > 1) + comp_vector = + atomic_inc_return(&counter) % + min_t(int, dev->num_comp_vectors, num_online_cpus()); + + return __ib_alloc_cq(dev, private, nr_cqe, comp_vector, poll_ctx, + caller); +} +EXPORT_SYMBOL(__ib_alloc_cq_any); + +/** + * ib_free_cq - free a completion queue + * @cq: completion queue to free. + */ +void ib_free_cq(struct ib_cq *cq) +{ + int ret; + + if (WARN_ON_ONCE(atomic_read(&cq->usecnt))) + return; + if (WARN_ON_ONCE(cq->cqe_used)) + return; + + switch (cq->poll_ctx) { + case IB_POLL_DIRECT: + break; + case IB_POLL_SOFTIRQ: + irq_poll_disable(&cq->iop); + break; + case IB_POLL_WORKQUEUE: + case IB_POLL_UNBOUND_WORKQUEUE: + cancel_work_sync(&cq->work); + break; + default: + WARN_ON_ONCE(1); + } + + rdma_dim_destroy(cq); + trace_cq_free(cq); + ret = cq->device->ops.destroy_cq(cq, NULL); + WARN_ONCE(ret, "Destroy of kernel CQ shouldn't fail"); + rdma_restrack_del(&cq->res); + kfree(cq->wc); + kfree(cq); +} +EXPORT_SYMBOL(ib_free_cq); + +void ib_cq_pool_init(struct ib_device *dev) +{ + unsigned int i; + + spin_lock_init(&dev->cq_pools_lock); + for (i = 0; i < ARRAY_SIZE(dev->cq_pools); i++) + INIT_LIST_HEAD(&dev->cq_pools[i]); +} + +void ib_cq_pool_destroy(struct ib_device *dev) +{ + struct ib_cq *cq, *n; + unsigned int i; + + for (i = 0; i < ARRAY_SIZE(dev->cq_pools); i++) { + list_for_each_entry_safe(cq, n, &dev->cq_pools[i], + pool_entry) { + WARN_ON(cq->cqe_used); + cq->shared = false; + ib_free_cq(cq); + } + } +} + +static int ib_alloc_cqs(struct ib_device *dev, unsigned int nr_cqes, + enum ib_poll_context poll_ctx) +{ + LIST_HEAD(tmp_list); + unsigned int nr_cqs, i; + struct ib_cq *cq, *n; + int ret; + + if (poll_ctx > IB_POLL_LAST_POOL_TYPE) { + WARN_ON_ONCE(poll_ctx > IB_POLL_LAST_POOL_TYPE); + return -EINVAL; + } + + /* + * Allocate at least as many CQEs as requested, and otherwise + * a reasonable batch size so that we can share CQs between + * multiple users instead of allocating a larger number of CQs. + */ + nr_cqes = min_t(unsigned int, dev->attrs.max_cqe, + max(nr_cqes, IB_MAX_SHARED_CQ_SZ)); + nr_cqs = min_t(unsigned int, dev->num_comp_vectors, num_online_cpus()); + for (i = 0; i < nr_cqs; i++) { + cq = ib_alloc_cq(dev, NULL, nr_cqes, i, poll_ctx); + if (IS_ERR(cq)) { + ret = PTR_ERR(cq); + goto out_free_cqs; + } + cq->shared = true; + list_add_tail(&cq->pool_entry, &tmp_list); + } + + spin_lock_irq(&dev->cq_pools_lock); + list_splice(&tmp_list, &dev->cq_pools[poll_ctx]); + spin_unlock_irq(&dev->cq_pools_lock); + + return 0; + +out_free_cqs: + list_for_each_entry_safe(cq, n, &tmp_list, pool_entry) { + cq->shared = false; + ib_free_cq(cq); + } + return ret; +} + +/** + * ib_cq_pool_get() - Find the least used completion queue that matches + * a given cpu hint (or least used for wild card affinity) and fits + * nr_cqe. + * @dev: rdma device + * @nr_cqe: number of needed cqe entries + * @comp_vector_hint: completion vector hint (-1) for the driver to assign + * a comp vector based on internal counter + * @poll_ctx: cq polling context + * + * Finds a cq that satisfies @comp_vector_hint and @nr_cqe requirements and + * claim entries in it for us. In case there is no available cq, allocate + * a new cq with the requirements and add it to the device pool. + * IB_POLL_DIRECT cannot be used for shared cqs so it is not a valid value + * for @poll_ctx. + */ +struct ib_cq *ib_cq_pool_get(struct ib_device *dev, unsigned int nr_cqe, + int comp_vector_hint, + enum ib_poll_context poll_ctx) +{ + static unsigned int default_comp_vector; + unsigned int vector, num_comp_vectors; + struct ib_cq *cq, *found = NULL; + int ret; + + if (poll_ctx > IB_POLL_LAST_POOL_TYPE) { + WARN_ON_ONCE(poll_ctx > IB_POLL_LAST_POOL_TYPE); + return ERR_PTR(-EINVAL); + } + + num_comp_vectors = + min_t(unsigned int, dev->num_comp_vectors, num_online_cpus()); + /* Project the affinty to the device completion vector range */ + if (comp_vector_hint < 0) { + comp_vector_hint = + (READ_ONCE(default_comp_vector) + 1) % num_comp_vectors; + WRITE_ONCE(default_comp_vector, comp_vector_hint); + } + vector = comp_vector_hint % num_comp_vectors; + + /* + * Find the least used CQ with correct affinity and + * enough free CQ entries + */ + while (!found) { + spin_lock_irq(&dev->cq_pools_lock); + list_for_each_entry(cq, &dev->cq_pools[poll_ctx], + pool_entry) { + /* + * Check to see if we have found a CQ with the + * correct completion vector + */ + if (vector != cq->comp_vector) + continue; + if (cq->cqe_used + nr_cqe > cq->cqe) + continue; + found = cq; + break; + } + + if (found) { + found->cqe_used += nr_cqe; + spin_unlock_irq(&dev->cq_pools_lock); + + return found; + } + spin_unlock_irq(&dev->cq_pools_lock); + + /* + * Didn't find a match or ran out of CQs in the device + * pool, allocate a new array of CQs. + */ + ret = ib_alloc_cqs(dev, nr_cqe, poll_ctx); + if (ret) + return ERR_PTR(ret); + } + + return found; +} +EXPORT_SYMBOL(ib_cq_pool_get); + +/** + * ib_cq_pool_put - Return a CQ taken from a shared pool. + * @cq: The CQ to return. + * @nr_cqe: The max number of cqes that the user had requested. + */ +void ib_cq_pool_put(struct ib_cq *cq, unsigned int nr_cqe) +{ + if (WARN_ON_ONCE(nr_cqe > cq->cqe_used)) + return; + + spin_lock_irq(&cq->device->cq_pools_lock); + cq->cqe_used -= nr_cqe; + spin_unlock_irq(&cq->device->cq_pools_lock); +} +EXPORT_SYMBOL(ib_cq_pool_put); diff --git a/drivers/infiniband/core/device.c b/drivers/infiniband/core/device.c new file mode 100644 index 000000000..3c29fd04b --- /dev/null +++ b/drivers/infiniband/core/device.c @@ -0,0 +1,2831 @@ +/* + * Copyright (c) 2004 Topspin Communications. All rights reserved. + * Copyright (c) 2005 Sun Microsystems, Inc. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include <linux/module.h> +#include <linux/string.h> +#include <linux/errno.h> +#include <linux/kernel.h> +#include <linux/slab.h> +#include <linux/init.h> +#include <linux/netdevice.h> +#include <net/net_namespace.h> +#include <linux/security.h> +#include <linux/notifier.h> +#include <linux/hashtable.h> +#include <rdma/rdma_netlink.h> +#include <rdma/ib_addr.h> +#include <rdma/ib_cache.h> +#include <rdma/rdma_counter.h> + +#include "core_priv.h" +#include "restrack.h" + +MODULE_AUTHOR("Roland Dreier"); +MODULE_DESCRIPTION("core kernel InfiniBand API"); +MODULE_LICENSE("Dual BSD/GPL"); + +struct workqueue_struct *ib_comp_wq; +struct workqueue_struct *ib_comp_unbound_wq; +struct workqueue_struct *ib_wq; +EXPORT_SYMBOL_GPL(ib_wq); + +/* + * Each of the three rwsem locks (devices, clients, client_data) protects the + * xarray of the same name. Specifically it allows the caller to assert that + * the MARK will/will not be changing under the lock, and for devices and + * clients, that the value in the xarray is still a valid pointer. Change of + * the MARK is linked to the object state, so holding the lock and testing the + * MARK also asserts that the contained object is in a certain state. + * + * This is used to build a two stage register/unregister flow where objects + * can continue to be in the xarray even though they are still in progress to + * register/unregister. + * + * The xarray itself provides additional locking, and restartable iteration, + * which is also relied on. + * + * Locks should not be nested, with the exception of client_data, which is + * allowed to nest under the read side of the other two locks. + * + * The devices_rwsem also protects the device name list, any change or + * assignment of device name must also hold the write side to guarantee unique + * names. + */ + +/* + * devices contains devices that have had their names assigned. The + * devices may not be registered. Users that care about the registration + * status need to call ib_device_try_get() on the device to ensure it is + * registered, and keep it registered, for the required duration. + * + */ +static DEFINE_XARRAY_FLAGS(devices, XA_FLAGS_ALLOC); +static DECLARE_RWSEM(devices_rwsem); +#define DEVICE_REGISTERED XA_MARK_1 + +static u32 highest_client_id; +#define CLIENT_REGISTERED XA_MARK_1 +static DEFINE_XARRAY_FLAGS(clients, XA_FLAGS_ALLOC); +static DECLARE_RWSEM(clients_rwsem); + +static void ib_client_put(struct ib_client *client) +{ + if (refcount_dec_and_test(&client->uses)) + complete(&client->uses_zero); +} + +/* + * If client_data is registered then the corresponding client must also still + * be registered. + */ +#define CLIENT_DATA_REGISTERED XA_MARK_1 + +unsigned int rdma_dev_net_id; + +/* + * A list of net namespaces is maintained in an xarray. This is necessary + * because we can't get the locking right using the existing net ns list. We + * would require a init_net callback after the list is updated. + */ +static DEFINE_XARRAY_FLAGS(rdma_nets, XA_FLAGS_ALLOC); +/* + * rwsem to protect accessing the rdma_nets xarray entries. + */ +static DECLARE_RWSEM(rdma_nets_rwsem); + +bool ib_devices_shared_netns = true; +module_param_named(netns_mode, ib_devices_shared_netns, bool, 0444); +MODULE_PARM_DESC(netns_mode, + "Share device among net namespaces; default=1 (shared)"); +/** + * rdma_dev_access_netns() - Return whether an rdma device can be accessed + * from a specified net namespace or not. + * @dev: Pointer to rdma device which needs to be checked + * @net: Pointer to net namesapce for which access to be checked + * + * When the rdma device is in shared mode, it ignores the net namespace. + * When the rdma device is exclusive to a net namespace, rdma device net + * namespace is checked against the specified one. + */ +bool rdma_dev_access_netns(const struct ib_device *dev, const struct net *net) +{ + return (ib_devices_shared_netns || + net_eq(read_pnet(&dev->coredev.rdma_net), net)); +} +EXPORT_SYMBOL(rdma_dev_access_netns); + +/* + * xarray has this behavior where it won't iterate over NULL values stored in + * allocated arrays. So we need our own iterator to see all values stored in + * the array. This does the same thing as xa_for_each except that it also + * returns NULL valued entries if the array is allocating. Simplified to only + * work on simple xarrays. + */ +static void *xan_find_marked(struct xarray *xa, unsigned long *indexp, + xa_mark_t filter) +{ + XA_STATE(xas, xa, *indexp); + void *entry; + + rcu_read_lock(); + do { + entry = xas_find_marked(&xas, ULONG_MAX, filter); + if (xa_is_zero(entry)) + break; + } while (xas_retry(&xas, entry)); + rcu_read_unlock(); + + if (entry) { + *indexp = xas.xa_index; + if (xa_is_zero(entry)) + return NULL; + return entry; + } + return XA_ERROR(-ENOENT); +} +#define xan_for_each_marked(xa, index, entry, filter) \ + for (index = 0, entry = xan_find_marked(xa, &(index), filter); \ + !xa_is_err(entry); \ + (index)++, entry = xan_find_marked(xa, &(index), filter)) + +/* RCU hash table mapping netdevice pointers to struct ib_port_data */ +static DEFINE_SPINLOCK(ndev_hash_lock); +static DECLARE_HASHTABLE(ndev_hash, 5); + +static void free_netdevs(struct ib_device *ib_dev); +static void ib_unregister_work(struct work_struct *work); +static void __ib_unregister_device(struct ib_device *device); +static int ib_security_change(struct notifier_block *nb, unsigned long event, + void *lsm_data); +static void ib_policy_change_task(struct work_struct *work); +static DECLARE_WORK(ib_policy_change_work, ib_policy_change_task); + +static void __ibdev_printk(const char *level, const struct ib_device *ibdev, + struct va_format *vaf) +{ + if (ibdev && ibdev->dev.parent) + dev_printk_emit(level[1] - '0', + ibdev->dev.parent, + "%s %s %s: %pV", + dev_driver_string(ibdev->dev.parent), + dev_name(ibdev->dev.parent), + dev_name(&ibdev->dev), + vaf); + else if (ibdev) + printk("%s%s: %pV", + level, dev_name(&ibdev->dev), vaf); + else + printk("%s(NULL ib_device): %pV", level, vaf); +} + +void ibdev_printk(const char *level, const struct ib_device *ibdev, + const char *format, ...) +{ + struct va_format vaf; + va_list args; + + va_start(args, format); + + vaf.fmt = format; + vaf.va = &args; + + __ibdev_printk(level, ibdev, &vaf); + + va_end(args); +} +EXPORT_SYMBOL(ibdev_printk); + +#define define_ibdev_printk_level(func, level) \ +void func(const struct ib_device *ibdev, const char *fmt, ...) \ +{ \ + struct va_format vaf; \ + va_list args; \ + \ + va_start(args, fmt); \ + \ + vaf.fmt = fmt; \ + vaf.va = &args; \ + \ + __ibdev_printk(level, ibdev, &vaf); \ + \ + va_end(args); \ +} \ +EXPORT_SYMBOL(func); + +define_ibdev_printk_level(ibdev_emerg, KERN_EMERG); +define_ibdev_printk_level(ibdev_alert, KERN_ALERT); +define_ibdev_printk_level(ibdev_crit, KERN_CRIT); +define_ibdev_printk_level(ibdev_err, KERN_ERR); +define_ibdev_printk_level(ibdev_warn, KERN_WARNING); +define_ibdev_printk_level(ibdev_notice, KERN_NOTICE); +define_ibdev_printk_level(ibdev_info, KERN_INFO); + +static struct notifier_block ibdev_lsm_nb = { + .notifier_call = ib_security_change, +}; + +static int rdma_dev_change_netns(struct ib_device *device, struct net *cur_net, + struct net *net); + +/* Pointer to the RCU head at the start of the ib_port_data array */ +struct ib_port_data_rcu { + struct rcu_head rcu_head; + struct ib_port_data pdata[]; +}; + +static void ib_device_check_mandatory(struct ib_device *device) +{ +#define IB_MANDATORY_FUNC(x) { offsetof(struct ib_device_ops, x), #x } + static const struct { + size_t offset; + char *name; + } mandatory_table[] = { + IB_MANDATORY_FUNC(query_device), + IB_MANDATORY_FUNC(query_port), + IB_MANDATORY_FUNC(alloc_pd), + IB_MANDATORY_FUNC(dealloc_pd), + IB_MANDATORY_FUNC(create_qp), + IB_MANDATORY_FUNC(modify_qp), + IB_MANDATORY_FUNC(destroy_qp), + IB_MANDATORY_FUNC(post_send), + IB_MANDATORY_FUNC(post_recv), + IB_MANDATORY_FUNC(create_cq), + IB_MANDATORY_FUNC(destroy_cq), + IB_MANDATORY_FUNC(poll_cq), + IB_MANDATORY_FUNC(req_notify_cq), + IB_MANDATORY_FUNC(get_dma_mr), + IB_MANDATORY_FUNC(dereg_mr), + IB_MANDATORY_FUNC(get_port_immutable) + }; + int i; + + device->kverbs_provider = true; + for (i = 0; i < ARRAY_SIZE(mandatory_table); ++i) { + if (!*(void **) ((void *) &device->ops + + mandatory_table[i].offset)) { + device->kverbs_provider = false; + break; + } + } +} + +/* + * Caller must perform ib_device_put() to return the device reference count + * when ib_device_get_by_index() returns valid device pointer. + */ +struct ib_device *ib_device_get_by_index(const struct net *net, u32 index) +{ + struct ib_device *device; + + down_read(&devices_rwsem); + device = xa_load(&devices, index); + if (device) { + if (!rdma_dev_access_netns(device, net)) { + device = NULL; + goto out; + } + + if (!ib_device_try_get(device)) + device = NULL; + } +out: + up_read(&devices_rwsem); + return device; +} + +/** + * ib_device_put - Release IB device reference + * @device: device whose reference to be released + * + * ib_device_put() releases reference to the IB device to allow it to be + * unregistered and eventually free. + */ +void ib_device_put(struct ib_device *device) +{ + if (refcount_dec_and_test(&device->refcount)) + complete(&device->unreg_completion); +} +EXPORT_SYMBOL(ib_device_put); + +static struct ib_device *__ib_device_get_by_name(const char *name) +{ + struct ib_device *device; + unsigned long index; + + xa_for_each (&devices, index, device) + if (!strcmp(name, dev_name(&device->dev))) + return device; + + return NULL; +} + +/** + * ib_device_get_by_name - Find an IB device by name + * @name: The name to look for + * @driver_id: The driver ID that must match (RDMA_DRIVER_UNKNOWN matches all) + * + * Find and hold an ib_device by its name. The caller must call + * ib_device_put() on the returned pointer. + */ +struct ib_device *ib_device_get_by_name(const char *name, + enum rdma_driver_id driver_id) +{ + struct ib_device *device; + + down_read(&devices_rwsem); + device = __ib_device_get_by_name(name); + if (device && driver_id != RDMA_DRIVER_UNKNOWN && + device->ops.driver_id != driver_id) + device = NULL; + + if (device) { + if (!ib_device_try_get(device)) + device = NULL; + } + up_read(&devices_rwsem); + return device; +} +EXPORT_SYMBOL(ib_device_get_by_name); + +static int rename_compat_devs(struct ib_device *device) +{ + struct ib_core_device *cdev; + unsigned long index; + int ret = 0; + + mutex_lock(&device->compat_devs_mutex); + xa_for_each (&device->compat_devs, index, cdev) { + ret = device_rename(&cdev->dev, dev_name(&device->dev)); + if (ret) { + dev_warn(&cdev->dev, + "Fail to rename compatdev to new name %s\n", + dev_name(&device->dev)); + break; + } + } + mutex_unlock(&device->compat_devs_mutex); + return ret; +} + +int ib_device_rename(struct ib_device *ibdev, const char *name) +{ + unsigned long index; + void *client_data; + int ret; + + down_write(&devices_rwsem); + if (!strcmp(name, dev_name(&ibdev->dev))) { + up_write(&devices_rwsem); + return 0; + } + + if (__ib_device_get_by_name(name)) { + up_write(&devices_rwsem); + return -EEXIST; + } + + ret = device_rename(&ibdev->dev, name); + if (ret) { + up_write(&devices_rwsem); + return ret; + } + + strlcpy(ibdev->name, name, IB_DEVICE_NAME_MAX); + ret = rename_compat_devs(ibdev); + + downgrade_write(&devices_rwsem); + down_read(&ibdev->client_data_rwsem); + xan_for_each_marked(&ibdev->client_data, index, client_data, + CLIENT_DATA_REGISTERED) { + struct ib_client *client = xa_load(&clients, index); + + if (!client || !client->rename) + continue; + + client->rename(ibdev, client_data); + } + up_read(&ibdev->client_data_rwsem); + up_read(&devices_rwsem); + return 0; +} + +int ib_device_set_dim(struct ib_device *ibdev, u8 use_dim) +{ + if (use_dim > 1) + return -EINVAL; + ibdev->use_cq_dim = use_dim; + + return 0; +} + +static int alloc_name(struct ib_device *ibdev, const char *name) +{ + struct ib_device *device; + unsigned long index; + struct ida inuse; + int rc; + int i; + + lockdep_assert_held_write(&devices_rwsem); + ida_init(&inuse); + xa_for_each (&devices, index, device) { + char buf[IB_DEVICE_NAME_MAX]; + + if (sscanf(dev_name(&device->dev), name, &i) != 1) + continue; + if (i < 0 || i >= INT_MAX) + continue; + snprintf(buf, sizeof buf, name, i); + if (strcmp(buf, dev_name(&device->dev)) != 0) + continue; + + rc = ida_alloc_range(&inuse, i, i, GFP_KERNEL); + if (rc < 0) + goto out; + } + + rc = ida_alloc(&inuse, GFP_KERNEL); + if (rc < 0) + goto out; + + rc = dev_set_name(&ibdev->dev, name, rc); +out: + ida_destroy(&inuse); + return rc; +} + +static void ib_device_release(struct device *device) +{ + struct ib_device *dev = container_of(device, struct ib_device, dev); + + free_netdevs(dev); + WARN_ON(refcount_read(&dev->refcount)); + if (dev->port_data) { + ib_cache_release_one(dev); + ib_security_release_port_pkey_list(dev); + rdma_counter_release(dev); + kfree_rcu(container_of(dev->port_data, struct ib_port_data_rcu, + pdata[0]), + rcu_head); + } + + mutex_destroy(&dev->unregistration_lock); + mutex_destroy(&dev->compat_devs_mutex); + + xa_destroy(&dev->compat_devs); + xa_destroy(&dev->client_data); + kfree_rcu(dev, rcu_head); +} + +static int ib_device_uevent(struct device *device, + struct kobj_uevent_env *env) +{ + if (add_uevent_var(env, "NAME=%s", dev_name(device))) + return -ENOMEM; + + /* + * It would be nice to pass the node GUID with the event... + */ + + return 0; +} + +static const void *net_namespace(struct device *d) +{ + struct ib_core_device *coredev = + container_of(d, struct ib_core_device, dev); + + return read_pnet(&coredev->rdma_net); +} + +static struct class ib_class = { + .name = "infiniband", + .dev_release = ib_device_release, + .dev_uevent = ib_device_uevent, + .ns_type = &net_ns_type_operations, + .namespace = net_namespace, +}; + +static void rdma_init_coredev(struct ib_core_device *coredev, + struct ib_device *dev, struct net *net) +{ + /* This BUILD_BUG_ON is intended to catch layout change + * of union of ib_core_device and device. + * dev must be the first element as ib_core and providers + * driver uses it. Adding anything in ib_core_device before + * device will break this assumption. + */ + BUILD_BUG_ON(offsetof(struct ib_device, coredev.dev) != + offsetof(struct ib_device, dev)); + + coredev->dev.class = &ib_class; + coredev->dev.groups = dev->groups; + device_initialize(&coredev->dev); + coredev->owner = dev; + INIT_LIST_HEAD(&coredev->port_list); + write_pnet(&coredev->rdma_net, net); +} + +/** + * _ib_alloc_device - allocate an IB device struct + * @size:size of structure to allocate + * + * Low-level drivers should use ib_alloc_device() to allocate &struct + * ib_device. @size is the size of the structure to be allocated, + * including any private data used by the low-level driver. + * ib_dealloc_device() must be used to free structures allocated with + * ib_alloc_device(). + */ +struct ib_device *_ib_alloc_device(size_t size) +{ + struct ib_device *device; + + if (WARN_ON(size < sizeof(struct ib_device))) + return NULL; + + device = kzalloc(size, GFP_KERNEL); + if (!device) + return NULL; + + if (rdma_restrack_init(device)) { + kfree(device); + return NULL; + } + + device->groups[0] = &ib_dev_attr_group; + rdma_init_coredev(&device->coredev, device, &init_net); + + INIT_LIST_HEAD(&device->event_handler_list); + spin_lock_init(&device->qp_open_list_lock); + init_rwsem(&device->event_handler_rwsem); + mutex_init(&device->unregistration_lock); + /* + * client_data needs to be alloc because we don't want our mark to be + * destroyed if the user stores NULL in the client data. + */ + xa_init_flags(&device->client_data, XA_FLAGS_ALLOC); + init_rwsem(&device->client_data_rwsem); + xa_init_flags(&device->compat_devs, XA_FLAGS_ALLOC); + mutex_init(&device->compat_devs_mutex); + init_completion(&device->unreg_completion); + INIT_WORK(&device->unregistration_work, ib_unregister_work); + + device->uverbs_ex_cmd_mask = + BIT_ULL(IB_USER_VERBS_EX_CMD_CREATE_FLOW) | + BIT_ULL(IB_USER_VERBS_EX_CMD_CREATE_RWQ_IND_TBL) | + BIT_ULL(IB_USER_VERBS_EX_CMD_CREATE_WQ) | + BIT_ULL(IB_USER_VERBS_EX_CMD_DESTROY_FLOW) | + BIT_ULL(IB_USER_VERBS_EX_CMD_DESTROY_RWQ_IND_TBL) | + BIT_ULL(IB_USER_VERBS_EX_CMD_DESTROY_WQ) | + BIT_ULL(IB_USER_VERBS_EX_CMD_MODIFY_CQ) | + BIT_ULL(IB_USER_VERBS_EX_CMD_MODIFY_WQ) | + BIT_ULL(IB_USER_VERBS_EX_CMD_QUERY_DEVICE); + + return device; +} +EXPORT_SYMBOL(_ib_alloc_device); + +/** + * ib_dealloc_device - free an IB device struct + * @device:structure to free + * + * Free a structure allocated with ib_alloc_device(). + */ +void ib_dealloc_device(struct ib_device *device) +{ + if (device->ops.dealloc_driver) + device->ops.dealloc_driver(device); + + /* + * ib_unregister_driver() requires all devices to remain in the xarray + * while their ops are callable. The last op we call is dealloc_driver + * above. This is needed to create a fence on op callbacks prior to + * allowing the driver module to unload. + */ + down_write(&devices_rwsem); + if (xa_load(&devices, device->index) == device) + xa_erase(&devices, device->index); + up_write(&devices_rwsem); + + /* Expedite releasing netdev references */ + free_netdevs(device); + + WARN_ON(!xa_empty(&device->compat_devs)); + WARN_ON(!xa_empty(&device->client_data)); + WARN_ON(refcount_read(&device->refcount)); + rdma_restrack_clean(device); + /* Balances with device_initialize */ + put_device(&device->dev); +} +EXPORT_SYMBOL(ib_dealloc_device); + +/* + * add_client_context() and remove_client_context() must be safe against + * parallel calls on the same device - registration/unregistration of both the + * device and client can be occurring in parallel. + * + * The routines need to be a fence, any caller must not return until the add + * or remove is fully completed. + */ +static int add_client_context(struct ib_device *device, + struct ib_client *client) +{ + int ret = 0; + + if (!device->kverbs_provider && !client->no_kverbs_req) + return 0; + + down_write(&device->client_data_rwsem); + /* + * So long as the client is registered hold both the client and device + * unregistration locks. + */ + if (!refcount_inc_not_zero(&client->uses)) + goto out_unlock; + refcount_inc(&device->refcount); + + /* + * Another caller to add_client_context got here first and has already + * completely initialized context. + */ + if (xa_get_mark(&device->client_data, client->client_id, + CLIENT_DATA_REGISTERED)) + goto out; + + ret = xa_err(xa_store(&device->client_data, client->client_id, NULL, + GFP_KERNEL)); + if (ret) + goto out; + downgrade_write(&device->client_data_rwsem); + if (client->add) { + if (client->add(device)) { + /* + * If a client fails to add then the error code is + * ignored, but we won't call any more ops on this + * client. + */ + xa_erase(&device->client_data, client->client_id); + up_read(&device->client_data_rwsem); + ib_device_put(device); + ib_client_put(client); + return 0; + } + } + + /* Readers shall not see a client until add has been completed */ + xa_set_mark(&device->client_data, client->client_id, + CLIENT_DATA_REGISTERED); + up_read(&device->client_data_rwsem); + return 0; + +out: + ib_device_put(device); + ib_client_put(client); +out_unlock: + up_write(&device->client_data_rwsem); + return ret; +} + +static void remove_client_context(struct ib_device *device, + unsigned int client_id) +{ + struct ib_client *client; + void *client_data; + + down_write(&device->client_data_rwsem); + if (!xa_get_mark(&device->client_data, client_id, + CLIENT_DATA_REGISTERED)) { + up_write(&device->client_data_rwsem); + return; + } + client_data = xa_load(&device->client_data, client_id); + xa_clear_mark(&device->client_data, client_id, CLIENT_DATA_REGISTERED); + client = xa_load(&clients, client_id); + up_write(&device->client_data_rwsem); + + /* + * Notice we cannot be holding any exclusive locks when calling the + * remove callback as the remove callback can recurse back into any + * public functions in this module and thus try for any locks those + * functions take. + * + * For this reason clients and drivers should not call the + * unregistration functions will holdling any locks. + */ + if (client->remove) + client->remove(device, client_data); + + xa_erase(&device->client_data, client_id); + ib_device_put(device); + ib_client_put(client); +} + +static int alloc_port_data(struct ib_device *device) +{ + struct ib_port_data_rcu *pdata_rcu; + unsigned int port; + + if (device->port_data) + return 0; + + /* This can only be called once the physical port range is defined */ + if (WARN_ON(!device->phys_port_cnt)) + return -EINVAL; + + /* + * device->port_data is indexed directly by the port number to make + * access to this data as efficient as possible. + * + * Therefore port_data is declared as a 1 based array with potential + * empty slots at the beginning. + */ + pdata_rcu = kzalloc(struct_size(pdata_rcu, pdata, + rdma_end_port(device) + 1), + GFP_KERNEL); + if (!pdata_rcu) + return -ENOMEM; + /* + * The rcu_head is put in front of the port data array and the stored + * pointer is adjusted since we never need to see that member until + * kfree_rcu. + */ + device->port_data = pdata_rcu->pdata; + + rdma_for_each_port (device, port) { + struct ib_port_data *pdata = &device->port_data[port]; + + pdata->ib_dev = device; + spin_lock_init(&pdata->pkey_list_lock); + INIT_LIST_HEAD(&pdata->pkey_list); + spin_lock_init(&pdata->netdev_lock); + INIT_HLIST_NODE(&pdata->ndev_hash_link); + } + return 0; +} + +static int verify_immutable(const struct ib_device *dev, u8 port) +{ + return WARN_ON(!rdma_cap_ib_mad(dev, port) && + rdma_max_mad_size(dev, port) != 0); +} + +static int setup_port_data(struct ib_device *device) +{ + unsigned int port; + int ret; + + ret = alloc_port_data(device); + if (ret) + return ret; + + rdma_for_each_port (device, port) { + struct ib_port_data *pdata = &device->port_data[port]; + + ret = device->ops.get_port_immutable(device, port, + &pdata->immutable); + if (ret) + return ret; + + if (verify_immutable(device, port)) + return -EINVAL; + } + return 0; +} + +void ib_get_device_fw_str(struct ib_device *dev, char *str) +{ + if (dev->ops.get_dev_fw_str) + dev->ops.get_dev_fw_str(dev, str); + else + str[0] = '\0'; +} +EXPORT_SYMBOL(ib_get_device_fw_str); + +static void ib_policy_change_task(struct work_struct *work) +{ + struct ib_device *dev; + unsigned long index; + + down_read(&devices_rwsem); + xa_for_each_marked (&devices, index, dev, DEVICE_REGISTERED) { + unsigned int i; + + rdma_for_each_port (dev, i) { + u64 sp; + int ret = ib_get_cached_subnet_prefix(dev, + i, + &sp); + + WARN_ONCE(ret, + "ib_get_cached_subnet_prefix err: %d, this should never happen here\n", + ret); + if (!ret) + ib_security_cache_change(dev, i, sp); + } + } + up_read(&devices_rwsem); +} + +static int ib_security_change(struct notifier_block *nb, unsigned long event, + void *lsm_data) +{ + if (event != LSM_POLICY_CHANGE) + return NOTIFY_DONE; + + schedule_work(&ib_policy_change_work); + ib_mad_agent_security_change(); + + return NOTIFY_OK; +} + +static void compatdev_release(struct device *dev) +{ + struct ib_core_device *cdev = + container_of(dev, struct ib_core_device, dev); + + kfree(cdev); +} + +static int add_one_compat_dev(struct ib_device *device, + struct rdma_dev_net *rnet) +{ + struct ib_core_device *cdev; + int ret; + + lockdep_assert_held(&rdma_nets_rwsem); + if (!ib_devices_shared_netns) + return 0; + + /* + * Create and add compat device in all namespaces other than where it + * is currently bound to. + */ + if (net_eq(read_pnet(&rnet->net), + read_pnet(&device->coredev.rdma_net))) + return 0; + + /* + * The first of init_net() or ib_register_device() to take the + * compat_devs_mutex wins and gets to add the device. Others will wait + * for completion here. + */ + mutex_lock(&device->compat_devs_mutex); + cdev = xa_load(&device->compat_devs, rnet->id); + if (cdev) { + ret = 0; + goto done; + } + ret = xa_reserve(&device->compat_devs, rnet->id, GFP_KERNEL); + if (ret) + goto done; + + cdev = kzalloc(sizeof(*cdev), GFP_KERNEL); + if (!cdev) { + ret = -ENOMEM; + goto cdev_err; + } + + cdev->dev.parent = device->dev.parent; + rdma_init_coredev(cdev, device, read_pnet(&rnet->net)); + cdev->dev.release = compatdev_release; + ret = dev_set_name(&cdev->dev, "%s", dev_name(&device->dev)); + if (ret) + goto add_err; + + ret = device_add(&cdev->dev); + if (ret) + goto add_err; + ret = ib_setup_port_attrs(cdev); + if (ret) + goto port_err; + + ret = xa_err(xa_store(&device->compat_devs, rnet->id, + cdev, GFP_KERNEL)); + if (ret) + goto insert_err; + + mutex_unlock(&device->compat_devs_mutex); + return 0; + +insert_err: + ib_free_port_attrs(cdev); +port_err: + device_del(&cdev->dev); +add_err: + put_device(&cdev->dev); +cdev_err: + xa_release(&device->compat_devs, rnet->id); +done: + mutex_unlock(&device->compat_devs_mutex); + return ret; +} + +static void remove_one_compat_dev(struct ib_device *device, u32 id) +{ + struct ib_core_device *cdev; + + mutex_lock(&device->compat_devs_mutex); + cdev = xa_erase(&device->compat_devs, id); + mutex_unlock(&device->compat_devs_mutex); + if (cdev) { + ib_free_port_attrs(cdev); + device_del(&cdev->dev); + put_device(&cdev->dev); + } +} + +static void remove_compat_devs(struct ib_device *device) +{ + struct ib_core_device *cdev; + unsigned long index; + + xa_for_each (&device->compat_devs, index, cdev) + remove_one_compat_dev(device, index); +} + +static int add_compat_devs(struct ib_device *device) +{ + struct rdma_dev_net *rnet; + unsigned long index; + int ret = 0; + + lockdep_assert_held(&devices_rwsem); + + down_read(&rdma_nets_rwsem); + xa_for_each (&rdma_nets, index, rnet) { + ret = add_one_compat_dev(device, rnet); + if (ret) + break; + } + up_read(&rdma_nets_rwsem); + return ret; +} + +static void remove_all_compat_devs(void) +{ + struct ib_compat_device *cdev; + struct ib_device *dev; + unsigned long index; + + down_read(&devices_rwsem); + xa_for_each (&devices, index, dev) { + unsigned long c_index = 0; + + /* Hold nets_rwsem so that any other thread modifying this + * system param can sync with this thread. + */ + down_read(&rdma_nets_rwsem); + xa_for_each (&dev->compat_devs, c_index, cdev) + remove_one_compat_dev(dev, c_index); + up_read(&rdma_nets_rwsem); + } + up_read(&devices_rwsem); +} + +static int add_all_compat_devs(void) +{ + struct rdma_dev_net *rnet; + struct ib_device *dev; + unsigned long index; + int ret = 0; + + down_read(&devices_rwsem); + xa_for_each_marked (&devices, index, dev, DEVICE_REGISTERED) { + unsigned long net_index = 0; + + /* Hold nets_rwsem so that any other thread modifying this + * system param can sync with this thread. + */ + down_read(&rdma_nets_rwsem); + xa_for_each (&rdma_nets, net_index, rnet) { + ret = add_one_compat_dev(dev, rnet); + if (ret) + break; + } + up_read(&rdma_nets_rwsem); + } + up_read(&devices_rwsem); + if (ret) + remove_all_compat_devs(); + return ret; +} + +int rdma_compatdev_set(u8 enable) +{ + struct rdma_dev_net *rnet; + unsigned long index; + int ret = 0; + + down_write(&rdma_nets_rwsem); + if (ib_devices_shared_netns == enable) { + up_write(&rdma_nets_rwsem); + return 0; + } + + /* enable/disable of compat devices is not supported + * when more than default init_net exists. + */ + xa_for_each (&rdma_nets, index, rnet) { + ret++; + break; + } + if (!ret) + ib_devices_shared_netns = enable; + up_write(&rdma_nets_rwsem); + if (ret) + return -EBUSY; + + if (enable) + ret = add_all_compat_devs(); + else + remove_all_compat_devs(); + return ret; +} + +static void rdma_dev_exit_net(struct net *net) +{ + struct rdma_dev_net *rnet = rdma_net_to_dev_net(net); + struct ib_device *dev; + unsigned long index; + int ret; + + down_write(&rdma_nets_rwsem); + /* + * Prevent the ID from being re-used and hide the id from xa_for_each. + */ + ret = xa_err(xa_store(&rdma_nets, rnet->id, NULL, GFP_KERNEL)); + WARN_ON(ret); + up_write(&rdma_nets_rwsem); + + down_read(&devices_rwsem); + xa_for_each (&devices, index, dev) { + get_device(&dev->dev); + /* + * Release the devices_rwsem so that pontentially blocking + * device_del, doesn't hold the devices_rwsem for too long. + */ + up_read(&devices_rwsem); + + remove_one_compat_dev(dev, rnet->id); + + /* + * If the real device is in the NS then move it back to init. + */ + rdma_dev_change_netns(dev, net, &init_net); + + put_device(&dev->dev); + down_read(&devices_rwsem); + } + up_read(&devices_rwsem); + + rdma_nl_net_exit(rnet); + xa_erase(&rdma_nets, rnet->id); +} + +static __net_init int rdma_dev_init_net(struct net *net) +{ + struct rdma_dev_net *rnet = rdma_net_to_dev_net(net); + unsigned long index; + struct ib_device *dev; + int ret; + + write_pnet(&rnet->net, net); + + ret = rdma_nl_net_init(rnet); + if (ret) + return ret; + + /* No need to create any compat devices in default init_net. */ + if (net_eq(net, &init_net)) + return 0; + + ret = xa_alloc(&rdma_nets, &rnet->id, rnet, xa_limit_32b, GFP_KERNEL); + if (ret) { + rdma_nl_net_exit(rnet); + return ret; + } + + down_read(&devices_rwsem); + xa_for_each_marked (&devices, index, dev, DEVICE_REGISTERED) { + /* Hold nets_rwsem so that netlink command cannot change + * system configuration for device sharing mode. + */ + down_read(&rdma_nets_rwsem); + ret = add_one_compat_dev(dev, rnet); + up_read(&rdma_nets_rwsem); + if (ret) + break; + } + up_read(&devices_rwsem); + + if (ret) + rdma_dev_exit_net(net); + + return ret; +} + +/* + * Assign the unique string device name and the unique device index. This is + * undone by ib_dealloc_device. + */ +static int assign_name(struct ib_device *device, const char *name) +{ + static u32 last_id; + int ret; + + down_write(&devices_rwsem); + /* Assign a unique name to the device */ + if (strchr(name, '%')) + ret = alloc_name(device, name); + else + ret = dev_set_name(&device->dev, name); + if (ret) + goto out; + + if (__ib_device_get_by_name(dev_name(&device->dev))) { + ret = -ENFILE; + goto out; + } + strlcpy(device->name, dev_name(&device->dev), IB_DEVICE_NAME_MAX); + + ret = xa_alloc_cyclic(&devices, &device->index, device, xa_limit_31b, + &last_id, GFP_KERNEL); + if (ret > 0) + ret = 0; + +out: + up_write(&devices_rwsem); + return ret; +} + +/* + * setup_device() allocates memory and sets up data that requires calling the + * device ops, this is the only reason these actions are not done during + * ib_alloc_device. It is undone by ib_dealloc_device(). + */ +static int setup_device(struct ib_device *device) +{ + struct ib_udata uhw = {.outlen = 0, .inlen = 0}; + int ret; + + ib_device_check_mandatory(device); + + ret = setup_port_data(device); + if (ret) { + dev_warn(&device->dev, "Couldn't create per-port data\n"); + return ret; + } + + memset(&device->attrs, 0, sizeof(device->attrs)); + ret = device->ops.query_device(device, &device->attrs, &uhw); + if (ret) { + dev_warn(&device->dev, + "Couldn't query the device attributes\n"); + return ret; + } + + return 0; +} + +static void disable_device(struct ib_device *device) +{ + u32 cid; + + WARN_ON(!refcount_read(&device->refcount)); + + down_write(&devices_rwsem); + xa_clear_mark(&devices, device->index, DEVICE_REGISTERED); + up_write(&devices_rwsem); + + /* + * Remove clients in LIFO order, see assign_client_id. This could be + * more efficient if xarray learns to reverse iterate. Since no new + * clients can be added to this ib_device past this point we only need + * the maximum possible client_id value here. + */ + down_read(&clients_rwsem); + cid = highest_client_id; + up_read(&clients_rwsem); + while (cid) { + cid--; + remove_client_context(device, cid); + } + + ib_cq_pool_destroy(device); + + /* Pairs with refcount_set in enable_device */ + ib_device_put(device); + wait_for_completion(&device->unreg_completion); + + /* + * compat devices must be removed after device refcount drops to zero. + * Otherwise init_net() may add more compatdevs after removing compat + * devices and before device is disabled. + */ + remove_compat_devs(device); +} + +/* + * An enabled device is visible to all clients and to all the public facing + * APIs that return a device pointer. This always returns with a new get, even + * if it fails. + */ +static int enable_device_and_get(struct ib_device *device) +{ + struct ib_client *client; + unsigned long index; + int ret = 0; + + /* + * One ref belongs to the xa and the other belongs to this + * thread. This is needed to guard against parallel unregistration. + */ + refcount_set(&device->refcount, 2); + down_write(&devices_rwsem); + xa_set_mark(&devices, device->index, DEVICE_REGISTERED); + + /* + * By using downgrade_write() we ensure that no other thread can clear + * DEVICE_REGISTERED while we are completing the client setup. + */ + downgrade_write(&devices_rwsem); + + if (device->ops.enable_driver) { + ret = device->ops.enable_driver(device); + if (ret) + goto out; + } + + ib_cq_pool_init(device); + + down_read(&clients_rwsem); + xa_for_each_marked (&clients, index, client, CLIENT_REGISTERED) { + ret = add_client_context(device, client); + if (ret) + break; + } + up_read(&clients_rwsem); + if (!ret) + ret = add_compat_devs(device); +out: + up_read(&devices_rwsem); + return ret; +} + +static void prevent_dealloc_device(struct ib_device *ib_dev) +{ +} + +/** + * ib_register_device - Register an IB device with IB core + * @device: Device to register + * @name: unique string device name. This may include a '%' which will + * cause a unique index to be added to the passed device name. + * @dma_device: pointer to a DMA-capable device. If %NULL, then the IB + * device will be used. In this case the caller should fully + * setup the ibdev for DMA. This usually means using dma_virt_ops. + * + * Low-level drivers use ib_register_device() to register their + * devices with the IB core. All registered clients will receive a + * callback for each device that is added. @device must be allocated + * with ib_alloc_device(). + * + * If the driver uses ops.dealloc_driver and calls any ib_unregister_device() + * asynchronously then the device pointer may become freed as soon as this + * function returns. + */ +int ib_register_device(struct ib_device *device, const char *name, + struct device *dma_device) +{ + int ret; + + ret = assign_name(device, name); + if (ret) + return ret; + + /* + * If the caller does not provide a DMA capable device then the IB core + * will set up ib_sge and scatterlist structures that stash the kernel + * virtual address into the address field. + */ + WARN_ON(dma_device && !dma_device->dma_parms); + device->dma_device = dma_device; + + ret = setup_device(device); + if (ret) + return ret; + + ret = ib_cache_setup_one(device); + if (ret) { + dev_warn(&device->dev, + "Couldn't set up InfiniBand P_Key/GID cache\n"); + return ret; + } + + ib_device_register_rdmacg(device); + + rdma_counter_init(device); + + /* + * Ensure that ADD uevent is not fired because it + * is too early amd device is not initialized yet. + */ + dev_set_uevent_suppress(&device->dev, true); + ret = device_add(&device->dev); + if (ret) + goto cg_cleanup; + + ret = ib_device_register_sysfs(device); + if (ret) { + dev_warn(&device->dev, + "Couldn't register device with driver model\n"); + goto dev_cleanup; + } + + ret = enable_device_and_get(device); + if (ret) { + void (*dealloc_fn)(struct ib_device *); + + /* + * If we hit this error flow then we don't want to + * automatically dealloc the device since the caller is + * expected to call ib_dealloc_device() after + * ib_register_device() fails. This is tricky due to the + * possibility for a parallel unregistration along with this + * error flow. Since we have a refcount here we know any + * parallel flow is stopped in disable_device and will see the + * special dealloc_driver pointer, causing the responsibility to + * ib_dealloc_device() to revert back to this thread. + */ + dealloc_fn = device->ops.dealloc_driver; + device->ops.dealloc_driver = prevent_dealloc_device; + ib_device_put(device); + __ib_unregister_device(device); + device->ops.dealloc_driver = dealloc_fn; + dev_set_uevent_suppress(&device->dev, false); + return ret; + } + dev_set_uevent_suppress(&device->dev, false); + /* Mark for userspace that device is ready */ + kobject_uevent(&device->dev.kobj, KOBJ_ADD); + ib_device_put(device); + + return 0; + +dev_cleanup: + device_del(&device->dev); +cg_cleanup: + dev_set_uevent_suppress(&device->dev, false); + ib_device_unregister_rdmacg(device); + ib_cache_cleanup_one(device); + return ret; +} +EXPORT_SYMBOL(ib_register_device); + +/* Callers must hold a get on the device. */ +static void __ib_unregister_device(struct ib_device *ib_dev) +{ + /* + * We have a registration lock so that all the calls to unregister are + * fully fenced, once any unregister returns the device is truely + * unregistered even if multiple callers are unregistering it at the + * same time. This also interacts with the registration flow and + * provides sane semantics if register and unregister are racing. + */ + mutex_lock(&ib_dev->unregistration_lock); + if (!refcount_read(&ib_dev->refcount)) + goto out; + + disable_device(ib_dev); + + /* Expedite removing unregistered pointers from the hash table */ + free_netdevs(ib_dev); + + ib_device_unregister_sysfs(ib_dev); + device_del(&ib_dev->dev); + ib_device_unregister_rdmacg(ib_dev); + ib_cache_cleanup_one(ib_dev); + + /* + * Drivers using the new flow may not call ib_dealloc_device except + * in error unwind prior to registration success. + */ + if (ib_dev->ops.dealloc_driver && + ib_dev->ops.dealloc_driver != prevent_dealloc_device) { + WARN_ON(kref_read(&ib_dev->dev.kobj.kref) <= 1); + ib_dealloc_device(ib_dev); + } +out: + mutex_unlock(&ib_dev->unregistration_lock); +} + +/** + * ib_unregister_device - Unregister an IB device + * @ib_dev: The device to unregister + * + * Unregister an IB device. All clients will receive a remove callback. + * + * Callers should call this routine only once, and protect against races with + * registration. Typically it should only be called as part of a remove + * callback in an implementation of driver core's struct device_driver and + * related. + * + * If ops.dealloc_driver is used then ib_dev will be freed upon return from + * this function. + */ +void ib_unregister_device(struct ib_device *ib_dev) +{ + get_device(&ib_dev->dev); + __ib_unregister_device(ib_dev); + put_device(&ib_dev->dev); +} +EXPORT_SYMBOL(ib_unregister_device); + +/** + * ib_unregister_device_and_put - Unregister a device while holding a 'get' + * @ib_dev: The device to unregister + * + * This is the same as ib_unregister_device(), except it includes an internal + * ib_device_put() that should match a 'get' obtained by the caller. + * + * It is safe to call this routine concurrently from multiple threads while + * holding the 'get'. When the function returns the device is fully + * unregistered. + * + * Drivers using this flow MUST use the driver_unregister callback to clean up + * their resources associated with the device and dealloc it. + */ +void ib_unregister_device_and_put(struct ib_device *ib_dev) +{ + WARN_ON(!ib_dev->ops.dealloc_driver); + get_device(&ib_dev->dev); + ib_device_put(ib_dev); + __ib_unregister_device(ib_dev); + put_device(&ib_dev->dev); +} +EXPORT_SYMBOL(ib_unregister_device_and_put); + +/** + * ib_unregister_driver - Unregister all IB devices for a driver + * @driver_id: The driver to unregister + * + * This implements a fence for device unregistration. It only returns once all + * devices associated with the driver_id have fully completed their + * unregistration and returned from ib_unregister_device*(). + * + * If device's are not yet unregistered it goes ahead and starts unregistering + * them. + * + * This does not block creation of new devices with the given driver_id, that + * is the responsibility of the caller. + */ +void ib_unregister_driver(enum rdma_driver_id driver_id) +{ + struct ib_device *ib_dev; + unsigned long index; + + down_read(&devices_rwsem); + xa_for_each (&devices, index, ib_dev) { + if (ib_dev->ops.driver_id != driver_id) + continue; + + get_device(&ib_dev->dev); + up_read(&devices_rwsem); + + WARN_ON(!ib_dev->ops.dealloc_driver); + __ib_unregister_device(ib_dev); + + put_device(&ib_dev->dev); + down_read(&devices_rwsem); + } + up_read(&devices_rwsem); +} +EXPORT_SYMBOL(ib_unregister_driver); + +static void ib_unregister_work(struct work_struct *work) +{ + struct ib_device *ib_dev = + container_of(work, struct ib_device, unregistration_work); + + __ib_unregister_device(ib_dev); + put_device(&ib_dev->dev); +} + +/** + * ib_unregister_device_queued - Unregister a device using a work queue + * @ib_dev: The device to unregister + * + * This schedules an asynchronous unregistration using a WQ for the device. A + * driver should use this to avoid holding locks while doing unregistration, + * such as holding the RTNL lock. + * + * Drivers using this API must use ib_unregister_driver before module unload + * to ensure that all scheduled unregistrations have completed. + */ +void ib_unregister_device_queued(struct ib_device *ib_dev) +{ + WARN_ON(!refcount_read(&ib_dev->refcount)); + WARN_ON(!ib_dev->ops.dealloc_driver); + get_device(&ib_dev->dev); + if (!queue_work(system_unbound_wq, &ib_dev->unregistration_work)) + put_device(&ib_dev->dev); +} +EXPORT_SYMBOL(ib_unregister_device_queued); + +/* + * The caller must pass in a device that has the kref held and the refcount + * released. If the device is in cur_net and still registered then it is moved + * into net. + */ +static int rdma_dev_change_netns(struct ib_device *device, struct net *cur_net, + struct net *net) +{ + int ret2 = -EINVAL; + int ret; + + mutex_lock(&device->unregistration_lock); + + /* + * If a device not under ib_device_get() or if the unregistration_lock + * is not held, the namespace can be changed, or it can be unregistered. + * Check again under the lock. + */ + if (refcount_read(&device->refcount) == 0 || + !net_eq(cur_net, read_pnet(&device->coredev.rdma_net))) { + ret = -ENODEV; + goto out; + } + + kobject_uevent(&device->dev.kobj, KOBJ_REMOVE); + disable_device(device); + + /* + * At this point no one can be using the device, so it is safe to + * change the namespace. + */ + write_pnet(&device->coredev.rdma_net, net); + + down_read(&devices_rwsem); + /* + * Currently rdma devices are system wide unique. So the device name + * is guaranteed free in the new namespace. Publish the new namespace + * at the sysfs level. + */ + ret = device_rename(&device->dev, dev_name(&device->dev)); + up_read(&devices_rwsem); + if (ret) { + dev_warn(&device->dev, + "%s: Couldn't rename device after namespace change\n", + __func__); + /* Try and put things back and re-enable the device */ + write_pnet(&device->coredev.rdma_net, cur_net); + } + + ret2 = enable_device_and_get(device); + if (ret2) { + /* + * This shouldn't really happen, but if it does, let the user + * retry at later point. So don't disable the device. + */ + dev_warn(&device->dev, + "%s: Couldn't re-enable device after namespace change\n", + __func__); + } + kobject_uevent(&device->dev.kobj, KOBJ_ADD); + + ib_device_put(device); +out: + mutex_unlock(&device->unregistration_lock); + if (ret) + return ret; + return ret2; +} + +int ib_device_set_netns_put(struct sk_buff *skb, + struct ib_device *dev, u32 ns_fd) +{ + struct net *net; + int ret; + + net = get_net_ns_by_fd(ns_fd); + if (IS_ERR(net)) { + ret = PTR_ERR(net); + goto net_err; + } + + if (!netlink_ns_capable(skb, net->user_ns, CAP_NET_ADMIN)) { + ret = -EPERM; + goto ns_err; + } + + /* + * Currently supported only for those providers which support + * disassociation and don't do port specific sysfs init. Once a + * port_cleanup infrastructure is implemented, this limitation will be + * removed. + */ + if (!dev->ops.disassociate_ucontext || dev->ops.init_port || + ib_devices_shared_netns) { + ret = -EOPNOTSUPP; + goto ns_err; + } + + get_device(&dev->dev); + ib_device_put(dev); + ret = rdma_dev_change_netns(dev, current->nsproxy->net_ns, net); + put_device(&dev->dev); + + put_net(net); + return ret; + +ns_err: + put_net(net); +net_err: + ib_device_put(dev); + return ret; +} + +static struct pernet_operations rdma_dev_net_ops = { + .init = rdma_dev_init_net, + .exit = rdma_dev_exit_net, + .id = &rdma_dev_net_id, + .size = sizeof(struct rdma_dev_net), +}; + +static int assign_client_id(struct ib_client *client) +{ + int ret; + + down_write(&clients_rwsem); + /* + * The add/remove callbacks must be called in FIFO/LIFO order. To + * achieve this we assign client_ids so they are sorted in + * registration order. + */ + client->client_id = highest_client_id; + ret = xa_insert(&clients, client->client_id, client, GFP_KERNEL); + if (ret) + goto out; + + highest_client_id++; + xa_set_mark(&clients, client->client_id, CLIENT_REGISTERED); + +out: + up_write(&clients_rwsem); + return ret; +} + +static void remove_client_id(struct ib_client *client) +{ + down_write(&clients_rwsem); + xa_erase(&clients, client->client_id); + for (; highest_client_id; highest_client_id--) + if (xa_load(&clients, highest_client_id - 1)) + break; + up_write(&clients_rwsem); +} + +/** + * ib_register_client - Register an IB client + * @client:Client to register + * + * Upper level users of the IB drivers can use ib_register_client() to + * register callbacks for IB device addition and removal. When an IB + * device is added, each registered client's add method will be called + * (in the order the clients were registered), and when a device is + * removed, each client's remove method will be called (in the reverse + * order that clients were registered). In addition, when + * ib_register_client() is called, the client will receive an add + * callback for all devices already registered. + */ +int ib_register_client(struct ib_client *client) +{ + struct ib_device *device; + unsigned long index; + int ret; + + refcount_set(&client->uses, 1); + init_completion(&client->uses_zero); + ret = assign_client_id(client); + if (ret) + return ret; + + down_read(&devices_rwsem); + xa_for_each_marked (&devices, index, device, DEVICE_REGISTERED) { + ret = add_client_context(device, client); + if (ret) { + up_read(&devices_rwsem); + ib_unregister_client(client); + return ret; + } + } + up_read(&devices_rwsem); + return 0; +} +EXPORT_SYMBOL(ib_register_client); + +/** + * ib_unregister_client - Unregister an IB client + * @client:Client to unregister + * + * Upper level users use ib_unregister_client() to remove their client + * registration. When ib_unregister_client() is called, the client + * will receive a remove callback for each IB device still registered. + * + * This is a full fence, once it returns no client callbacks will be called, + * or are running in another thread. + */ +void ib_unregister_client(struct ib_client *client) +{ + struct ib_device *device; + unsigned long index; + + down_write(&clients_rwsem); + ib_client_put(client); + xa_clear_mark(&clients, client->client_id, CLIENT_REGISTERED); + up_write(&clients_rwsem); + + /* We do not want to have locks while calling client->remove() */ + rcu_read_lock(); + xa_for_each (&devices, index, device) { + if (!ib_device_try_get(device)) + continue; + rcu_read_unlock(); + + remove_client_context(device, client->client_id); + + ib_device_put(device); + rcu_read_lock(); + } + rcu_read_unlock(); + + /* + * remove_client_context() is not a fence, it can return even though a + * removal is ongoing. Wait until all removals are completed. + */ + wait_for_completion(&client->uses_zero); + remove_client_id(client); +} +EXPORT_SYMBOL(ib_unregister_client); + +static int __ib_get_global_client_nl_info(const char *client_name, + struct ib_client_nl_info *res) +{ + struct ib_client *client; + unsigned long index; + int ret = -ENOENT; + + down_read(&clients_rwsem); + xa_for_each_marked (&clients, index, client, CLIENT_REGISTERED) { + if (strcmp(client->name, client_name) != 0) + continue; + if (!client->get_global_nl_info) { + ret = -EOPNOTSUPP; + break; + } + ret = client->get_global_nl_info(res); + if (WARN_ON(ret == -ENOENT)) + ret = -EINVAL; + if (!ret && res->cdev) + get_device(res->cdev); + break; + } + up_read(&clients_rwsem); + return ret; +} + +static int __ib_get_client_nl_info(struct ib_device *ibdev, + const char *client_name, + struct ib_client_nl_info *res) +{ + unsigned long index; + void *client_data; + int ret = -ENOENT; + + down_read(&ibdev->client_data_rwsem); + xan_for_each_marked (&ibdev->client_data, index, client_data, + CLIENT_DATA_REGISTERED) { + struct ib_client *client = xa_load(&clients, index); + + if (!client || strcmp(client->name, client_name) != 0) + continue; + if (!client->get_nl_info) { + ret = -EOPNOTSUPP; + break; + } + ret = client->get_nl_info(ibdev, client_data, res); + if (WARN_ON(ret == -ENOENT)) + ret = -EINVAL; + + /* + * The cdev is guaranteed valid as long as we are inside the + * client_data_rwsem as remove_one can't be called. Keep it + * valid for the caller. + */ + if (!ret && res->cdev) + get_device(res->cdev); + break; + } + up_read(&ibdev->client_data_rwsem); + + return ret; +} + +/** + * ib_get_client_nl_info - Fetch the nl_info from a client + * @device - IB device + * @client_name - Name of the client + * @res - Result of the query + */ +int ib_get_client_nl_info(struct ib_device *ibdev, const char *client_name, + struct ib_client_nl_info *res) +{ + int ret; + + if (ibdev) + ret = __ib_get_client_nl_info(ibdev, client_name, res); + else + ret = __ib_get_global_client_nl_info(client_name, res); +#ifdef CONFIG_MODULES + if (ret == -ENOENT) { + request_module("rdma-client-%s", client_name); + if (ibdev) + ret = __ib_get_client_nl_info(ibdev, client_name, res); + else + ret = __ib_get_global_client_nl_info(client_name, res); + } +#endif + if (ret) { + if (ret == -ENOENT) + return -EOPNOTSUPP; + return ret; + } + + if (WARN_ON(!res->cdev)) + return -EINVAL; + return 0; +} + +/** + * ib_set_client_data - Set IB client context + * @device:Device to set context for + * @client:Client to set context for + * @data:Context to set + * + * ib_set_client_data() sets client context data that can be retrieved with + * ib_get_client_data(). This can only be called while the client is + * registered to the device, once the ib_client remove() callback returns this + * cannot be called. + */ +void ib_set_client_data(struct ib_device *device, struct ib_client *client, + void *data) +{ + void *rc; + + if (WARN_ON(IS_ERR(data))) + data = NULL; + + rc = xa_store(&device->client_data, client->client_id, data, + GFP_KERNEL); + WARN_ON(xa_is_err(rc)); +} +EXPORT_SYMBOL(ib_set_client_data); + +/** + * ib_register_event_handler - Register an IB event handler + * @event_handler:Handler to register + * + * ib_register_event_handler() registers an event handler that will be + * called back when asynchronous IB events occur (as defined in + * chapter 11 of the InfiniBand Architecture Specification). This + * callback occurs in workqueue context. + */ +void ib_register_event_handler(struct ib_event_handler *event_handler) +{ + down_write(&event_handler->device->event_handler_rwsem); + list_add_tail(&event_handler->list, + &event_handler->device->event_handler_list); + up_write(&event_handler->device->event_handler_rwsem); +} +EXPORT_SYMBOL(ib_register_event_handler); + +/** + * ib_unregister_event_handler - Unregister an event handler + * @event_handler:Handler to unregister + * + * Unregister an event handler registered with + * ib_register_event_handler(). + */ +void ib_unregister_event_handler(struct ib_event_handler *event_handler) +{ + down_write(&event_handler->device->event_handler_rwsem); + list_del(&event_handler->list); + up_write(&event_handler->device->event_handler_rwsem); +} +EXPORT_SYMBOL(ib_unregister_event_handler); + +void ib_dispatch_event_clients(struct ib_event *event) +{ + struct ib_event_handler *handler; + + down_read(&event->device->event_handler_rwsem); + + list_for_each_entry(handler, &event->device->event_handler_list, list) + handler->handler(handler, event); + + up_read(&event->device->event_handler_rwsem); +} + +static int iw_query_port(struct ib_device *device, + u8 port_num, + struct ib_port_attr *port_attr) +{ + struct in_device *inetdev; + struct net_device *netdev; + + memset(port_attr, 0, sizeof(*port_attr)); + + netdev = ib_device_get_netdev(device, port_num); + if (!netdev) + return -ENODEV; + + port_attr->max_mtu = IB_MTU_4096; + port_attr->active_mtu = ib_mtu_int_to_enum(netdev->mtu); + + if (!netif_carrier_ok(netdev)) { + port_attr->state = IB_PORT_DOWN; + port_attr->phys_state = IB_PORT_PHYS_STATE_DISABLED; + } else { + rcu_read_lock(); + inetdev = __in_dev_get_rcu(netdev); + + if (inetdev && inetdev->ifa_list) { + port_attr->state = IB_PORT_ACTIVE; + port_attr->phys_state = IB_PORT_PHYS_STATE_LINK_UP; + } else { + port_attr->state = IB_PORT_INIT; + port_attr->phys_state = + IB_PORT_PHYS_STATE_PORT_CONFIGURATION_TRAINING; + } + + rcu_read_unlock(); + } + + dev_put(netdev); + return device->ops.query_port(device, port_num, port_attr); +} + +static int __ib_query_port(struct ib_device *device, + u8 port_num, + struct ib_port_attr *port_attr) +{ + union ib_gid gid = {}; + int err; + + memset(port_attr, 0, sizeof(*port_attr)); + + err = device->ops.query_port(device, port_num, port_attr); + if (err || port_attr->subnet_prefix) + return err; + + if (rdma_port_get_link_layer(device, port_num) != + IB_LINK_LAYER_INFINIBAND) + return 0; + + err = device->ops.query_gid(device, port_num, 0, &gid); + if (err) + return err; + + port_attr->subnet_prefix = be64_to_cpu(gid.global.subnet_prefix); + return 0; +} + +/** + * ib_query_port - Query IB port attributes + * @device:Device to query + * @port_num:Port number to query + * @port_attr:Port attributes + * + * ib_query_port() returns the attributes of a port through the + * @port_attr pointer. + */ +int ib_query_port(struct ib_device *device, + u8 port_num, + struct ib_port_attr *port_attr) +{ + if (!rdma_is_port_valid(device, port_num)) + return -EINVAL; + + if (rdma_protocol_iwarp(device, port_num)) + return iw_query_port(device, port_num, port_attr); + else + return __ib_query_port(device, port_num, port_attr); +} +EXPORT_SYMBOL(ib_query_port); + +static void add_ndev_hash(struct ib_port_data *pdata) +{ + unsigned long flags; + + might_sleep(); + + spin_lock_irqsave(&ndev_hash_lock, flags); + if (hash_hashed(&pdata->ndev_hash_link)) { + hash_del_rcu(&pdata->ndev_hash_link); + spin_unlock_irqrestore(&ndev_hash_lock, flags); + /* + * We cannot do hash_add_rcu after a hash_del_rcu until the + * grace period + */ + synchronize_rcu(); + spin_lock_irqsave(&ndev_hash_lock, flags); + } + if (pdata->netdev) + hash_add_rcu(ndev_hash, &pdata->ndev_hash_link, + (uintptr_t)pdata->netdev); + spin_unlock_irqrestore(&ndev_hash_lock, flags); +} + +/** + * ib_device_set_netdev - Associate the ib_dev with an underlying net_device + * @ib_dev: Device to modify + * @ndev: net_device to affiliate, may be NULL + * @port: IB port the net_device is connected to + * + * Drivers should use this to link the ib_device to a netdev so the netdev + * shows up in interfaces like ib_enum_roce_netdev. Only one netdev may be + * affiliated with any port. + * + * The caller must ensure that the given ndev is not unregistered or + * unregistering, and that either the ib_device is unregistered or + * ib_device_set_netdev() is called with NULL when the ndev sends a + * NETDEV_UNREGISTER event. + */ +int ib_device_set_netdev(struct ib_device *ib_dev, struct net_device *ndev, + unsigned int port) +{ + struct net_device *old_ndev; + struct ib_port_data *pdata; + unsigned long flags; + int ret; + + /* + * Drivers wish to call this before ib_register_driver, so we have to + * setup the port data early. + */ + ret = alloc_port_data(ib_dev); + if (ret) + return ret; + + if (!rdma_is_port_valid(ib_dev, port)) + return -EINVAL; + + pdata = &ib_dev->port_data[port]; + spin_lock_irqsave(&pdata->netdev_lock, flags); + old_ndev = rcu_dereference_protected( + pdata->netdev, lockdep_is_held(&pdata->netdev_lock)); + if (old_ndev == ndev) { + spin_unlock_irqrestore(&pdata->netdev_lock, flags); + return 0; + } + + if (ndev) + dev_hold(ndev); + rcu_assign_pointer(pdata->netdev, ndev); + spin_unlock_irqrestore(&pdata->netdev_lock, flags); + + add_ndev_hash(pdata); + if (old_ndev) + dev_put(old_ndev); + + return 0; +} +EXPORT_SYMBOL(ib_device_set_netdev); + +static void free_netdevs(struct ib_device *ib_dev) +{ + unsigned long flags; + unsigned int port; + + if (!ib_dev->port_data) + return; + + rdma_for_each_port (ib_dev, port) { + struct ib_port_data *pdata = &ib_dev->port_data[port]; + struct net_device *ndev; + + spin_lock_irqsave(&pdata->netdev_lock, flags); + ndev = rcu_dereference_protected( + pdata->netdev, lockdep_is_held(&pdata->netdev_lock)); + if (ndev) { + spin_lock(&ndev_hash_lock); + hash_del_rcu(&pdata->ndev_hash_link); + spin_unlock(&ndev_hash_lock); + + /* + * If this is the last dev_put there is still a + * synchronize_rcu before the netdev is kfreed, so we + * can continue to rely on unlocked pointer + * comparisons after the put + */ + rcu_assign_pointer(pdata->netdev, NULL); + dev_put(ndev); + } + spin_unlock_irqrestore(&pdata->netdev_lock, flags); + } +} + +struct net_device *ib_device_get_netdev(struct ib_device *ib_dev, + unsigned int port) +{ + struct ib_port_data *pdata; + struct net_device *res; + + if (!rdma_is_port_valid(ib_dev, port)) + return NULL; + + pdata = &ib_dev->port_data[port]; + + /* + * New drivers should use ib_device_set_netdev() not the legacy + * get_netdev(). + */ + if (ib_dev->ops.get_netdev) + res = ib_dev->ops.get_netdev(ib_dev, port); + else { + spin_lock(&pdata->netdev_lock); + res = rcu_dereference_protected( + pdata->netdev, lockdep_is_held(&pdata->netdev_lock)); + if (res) + dev_hold(res); + spin_unlock(&pdata->netdev_lock); + } + + /* + * If we are starting to unregister expedite things by preventing + * propagation of an unregistering netdev. + */ + if (res && res->reg_state != NETREG_REGISTERED) { + dev_put(res); + return NULL; + } + + return res; +} + +/** + * ib_device_get_by_netdev - Find an IB device associated with a netdev + * @ndev: netdev to locate + * @driver_id: The driver ID that must match (RDMA_DRIVER_UNKNOWN matches all) + * + * Find and hold an ib_device that is associated with a netdev via + * ib_device_set_netdev(). The caller must call ib_device_put() on the + * returned pointer. + */ +struct ib_device *ib_device_get_by_netdev(struct net_device *ndev, + enum rdma_driver_id driver_id) +{ + struct ib_device *res = NULL; + struct ib_port_data *cur; + + rcu_read_lock(); + hash_for_each_possible_rcu (ndev_hash, cur, ndev_hash_link, + (uintptr_t)ndev) { + if (rcu_access_pointer(cur->netdev) == ndev && + (driver_id == RDMA_DRIVER_UNKNOWN || + cur->ib_dev->ops.driver_id == driver_id) && + ib_device_try_get(cur->ib_dev)) { + res = cur->ib_dev; + break; + } + } + rcu_read_unlock(); + + return res; +} +EXPORT_SYMBOL(ib_device_get_by_netdev); + +/** + * ib_enum_roce_netdev - enumerate all RoCE ports + * @ib_dev : IB device we want to query + * @filter: Should we call the callback? + * @filter_cookie: Cookie passed to filter + * @cb: Callback to call for each found RoCE ports + * @cookie: Cookie passed back to the callback + * + * Enumerates all of the physical RoCE ports of ib_dev + * which are related to netdevice and calls callback() on each + * device for which filter() function returns non zero. + */ +void ib_enum_roce_netdev(struct ib_device *ib_dev, + roce_netdev_filter filter, + void *filter_cookie, + roce_netdev_callback cb, + void *cookie) +{ + unsigned int port; + + rdma_for_each_port (ib_dev, port) + if (rdma_protocol_roce(ib_dev, port)) { + struct net_device *idev = + ib_device_get_netdev(ib_dev, port); + + if (filter(ib_dev, port, idev, filter_cookie)) + cb(ib_dev, port, idev, cookie); + + if (idev) + dev_put(idev); + } +} + +/** + * ib_enum_all_roce_netdevs - enumerate all RoCE devices + * @filter: Should we call the callback? + * @filter_cookie: Cookie passed to filter + * @cb: Callback to call for each found RoCE ports + * @cookie: Cookie passed back to the callback + * + * Enumerates all RoCE devices' physical ports which are related + * to netdevices and calls callback() on each device for which + * filter() function returns non zero. + */ +void ib_enum_all_roce_netdevs(roce_netdev_filter filter, + void *filter_cookie, + roce_netdev_callback cb, + void *cookie) +{ + struct ib_device *dev; + unsigned long index; + + down_read(&devices_rwsem); + xa_for_each_marked (&devices, index, dev, DEVICE_REGISTERED) + ib_enum_roce_netdev(dev, filter, filter_cookie, cb, cookie); + up_read(&devices_rwsem); +} + +/** + * ib_enum_all_devs - enumerate all ib_devices + * @cb: Callback to call for each found ib_device + * + * Enumerates all ib_devices and calls callback() on each device. + */ +int ib_enum_all_devs(nldev_callback nldev_cb, struct sk_buff *skb, + struct netlink_callback *cb) +{ + unsigned long index; + struct ib_device *dev; + unsigned int idx = 0; + int ret = 0; + + down_read(&devices_rwsem); + xa_for_each_marked (&devices, index, dev, DEVICE_REGISTERED) { + if (!rdma_dev_access_netns(dev, sock_net(skb->sk))) + continue; + + ret = nldev_cb(dev, skb, cb, idx); + if (ret) + break; + idx++; + } + up_read(&devices_rwsem); + return ret; +} + +/** + * ib_query_pkey - Get P_Key table entry + * @device:Device to query + * @port_num:Port number to query + * @index:P_Key table index to query + * @pkey:Returned P_Key + * + * ib_query_pkey() fetches the specified P_Key table entry. + */ +int ib_query_pkey(struct ib_device *device, + u8 port_num, u16 index, u16 *pkey) +{ + if (!rdma_is_port_valid(device, port_num)) + return -EINVAL; + + if (!device->ops.query_pkey) + return -EOPNOTSUPP; + + return device->ops.query_pkey(device, port_num, index, pkey); +} +EXPORT_SYMBOL(ib_query_pkey); + +/** + * ib_modify_device - Change IB device attributes + * @device:Device to modify + * @device_modify_mask:Mask of attributes to change + * @device_modify:New attribute values + * + * ib_modify_device() changes a device's attributes as specified by + * the @device_modify_mask and @device_modify structure. + */ +int ib_modify_device(struct ib_device *device, + int device_modify_mask, + struct ib_device_modify *device_modify) +{ + if (!device->ops.modify_device) + return -EOPNOTSUPP; + + return device->ops.modify_device(device, device_modify_mask, + device_modify); +} +EXPORT_SYMBOL(ib_modify_device); + +/** + * ib_modify_port - Modifies the attributes for the specified port. + * @device: The device to modify. + * @port_num: The number of the port to modify. + * @port_modify_mask: Mask used to specify which attributes of the port + * to change. + * @port_modify: New attribute values for the port. + * + * ib_modify_port() changes a port's attributes as specified by the + * @port_modify_mask and @port_modify structure. + */ +int ib_modify_port(struct ib_device *device, + u8 port_num, int port_modify_mask, + struct ib_port_modify *port_modify) +{ + int rc; + + if (!rdma_is_port_valid(device, port_num)) + return -EINVAL; + + if (device->ops.modify_port) + rc = device->ops.modify_port(device, port_num, + port_modify_mask, + port_modify); + else if (rdma_protocol_roce(device, port_num) && + ((port_modify->set_port_cap_mask & ~IB_PORT_CM_SUP) == 0 || + (port_modify->clr_port_cap_mask & ~IB_PORT_CM_SUP) == 0)) + rc = 0; + else + rc = -EOPNOTSUPP; + return rc; +} +EXPORT_SYMBOL(ib_modify_port); + +/** + * ib_find_gid - Returns the port number and GID table index where + * a specified GID value occurs. Its searches only for IB link layer. + * @device: The device to query. + * @gid: The GID value to search for. + * @port_num: The port number of the device where the GID value was found. + * @index: The index into the GID table where the GID was found. This + * parameter may be NULL. + */ +int ib_find_gid(struct ib_device *device, union ib_gid *gid, + u8 *port_num, u16 *index) +{ + union ib_gid tmp_gid; + unsigned int port; + int ret, i; + + rdma_for_each_port (device, port) { + if (!rdma_protocol_ib(device, port)) + continue; + + for (i = 0; i < device->port_data[port].immutable.gid_tbl_len; + ++i) { + ret = rdma_query_gid(device, port, i, &tmp_gid); + if (ret) + continue; + + if (!memcmp(&tmp_gid, gid, sizeof *gid)) { + *port_num = port; + if (index) + *index = i; + return 0; + } + } + } + + return -ENOENT; +} +EXPORT_SYMBOL(ib_find_gid); + +/** + * ib_find_pkey - Returns the PKey table index where a specified + * PKey value occurs. + * @device: The device to query. + * @port_num: The port number of the device to search for the PKey. + * @pkey: The PKey value to search for. + * @index: The index into the PKey table where the PKey was found. + */ +int ib_find_pkey(struct ib_device *device, + u8 port_num, u16 pkey, u16 *index) +{ + int ret, i; + u16 tmp_pkey; + int partial_ix = -1; + + for (i = 0; i < device->port_data[port_num].immutable.pkey_tbl_len; + ++i) { + ret = ib_query_pkey(device, port_num, i, &tmp_pkey); + if (ret) + return ret; + if ((pkey & 0x7fff) == (tmp_pkey & 0x7fff)) { + /* if there is full-member pkey take it.*/ + if (tmp_pkey & 0x8000) { + *index = i; + return 0; + } + if (partial_ix < 0) + partial_ix = i; + } + } + + /*no full-member, if exists take the limited*/ + if (partial_ix >= 0) { + *index = partial_ix; + return 0; + } + return -ENOENT; +} +EXPORT_SYMBOL(ib_find_pkey); + +/** + * ib_get_net_dev_by_params() - Return the appropriate net_dev + * for a received CM request + * @dev: An RDMA device on which the request has been received. + * @port: Port number on the RDMA device. + * @pkey: The Pkey the request came on. + * @gid: A GID that the net_dev uses to communicate. + * @addr: Contains the IP address that the request specified as its + * destination. + * + */ +struct net_device *ib_get_net_dev_by_params(struct ib_device *dev, + u8 port, + u16 pkey, + const union ib_gid *gid, + const struct sockaddr *addr) +{ + struct net_device *net_dev = NULL; + unsigned long index; + void *client_data; + + if (!rdma_protocol_ib(dev, port)) + return NULL; + + /* + * Holding the read side guarantees that the client will not become + * unregistered while we are calling get_net_dev_by_params() + */ + down_read(&dev->client_data_rwsem); + xan_for_each_marked (&dev->client_data, index, client_data, + CLIENT_DATA_REGISTERED) { + struct ib_client *client = xa_load(&clients, index); + + if (!client || !client->get_net_dev_by_params) + continue; + + net_dev = client->get_net_dev_by_params(dev, port, pkey, gid, + addr, client_data); + if (net_dev) + break; + } + up_read(&dev->client_data_rwsem); + + return net_dev; +} +EXPORT_SYMBOL(ib_get_net_dev_by_params); + +void ib_set_device_ops(struct ib_device *dev, const struct ib_device_ops *ops) +{ + struct ib_device_ops *dev_ops = &dev->ops; +#define SET_DEVICE_OP(ptr, name) \ + do { \ + if (ops->name) \ + if (!((ptr)->name)) \ + (ptr)->name = ops->name; \ + } while (0) + +#define SET_OBJ_SIZE(ptr, name) SET_DEVICE_OP(ptr, size_##name) + + if (ops->driver_id != RDMA_DRIVER_UNKNOWN) { + WARN_ON(dev_ops->driver_id != RDMA_DRIVER_UNKNOWN && + dev_ops->driver_id != ops->driver_id); + dev_ops->driver_id = ops->driver_id; + } + if (ops->owner) { + WARN_ON(dev_ops->owner && dev_ops->owner != ops->owner); + dev_ops->owner = ops->owner; + } + if (ops->uverbs_abi_ver) + dev_ops->uverbs_abi_ver = ops->uverbs_abi_ver; + + dev_ops->uverbs_no_driver_id_binding |= + ops->uverbs_no_driver_id_binding; + + SET_DEVICE_OP(dev_ops, add_gid); + SET_DEVICE_OP(dev_ops, advise_mr); + SET_DEVICE_OP(dev_ops, alloc_dm); + SET_DEVICE_OP(dev_ops, alloc_hw_stats); + SET_DEVICE_OP(dev_ops, alloc_mr); + SET_DEVICE_OP(dev_ops, alloc_mr_integrity); + SET_DEVICE_OP(dev_ops, alloc_mw); + SET_DEVICE_OP(dev_ops, alloc_pd); + SET_DEVICE_OP(dev_ops, alloc_rdma_netdev); + SET_DEVICE_OP(dev_ops, alloc_ucontext); + SET_DEVICE_OP(dev_ops, alloc_xrcd); + SET_DEVICE_OP(dev_ops, attach_mcast); + SET_DEVICE_OP(dev_ops, check_mr_status); + SET_DEVICE_OP(dev_ops, counter_alloc_stats); + SET_DEVICE_OP(dev_ops, counter_bind_qp); + SET_DEVICE_OP(dev_ops, counter_dealloc); + SET_DEVICE_OP(dev_ops, counter_unbind_qp); + SET_DEVICE_OP(dev_ops, counter_update_stats); + SET_DEVICE_OP(dev_ops, create_ah); + SET_DEVICE_OP(dev_ops, create_counters); + SET_DEVICE_OP(dev_ops, create_cq); + SET_DEVICE_OP(dev_ops, create_flow); + SET_DEVICE_OP(dev_ops, create_flow_action_esp); + SET_DEVICE_OP(dev_ops, create_qp); + SET_DEVICE_OP(dev_ops, create_rwq_ind_table); + SET_DEVICE_OP(dev_ops, create_srq); + SET_DEVICE_OP(dev_ops, create_wq); + SET_DEVICE_OP(dev_ops, dealloc_dm); + SET_DEVICE_OP(dev_ops, dealloc_driver); + SET_DEVICE_OP(dev_ops, dealloc_mw); + SET_DEVICE_OP(dev_ops, dealloc_pd); + SET_DEVICE_OP(dev_ops, dealloc_ucontext); + SET_DEVICE_OP(dev_ops, dealloc_xrcd); + SET_DEVICE_OP(dev_ops, del_gid); + SET_DEVICE_OP(dev_ops, dereg_mr); + SET_DEVICE_OP(dev_ops, destroy_ah); + SET_DEVICE_OP(dev_ops, destroy_counters); + SET_DEVICE_OP(dev_ops, destroy_cq); + SET_DEVICE_OP(dev_ops, destroy_flow); + SET_DEVICE_OP(dev_ops, destroy_flow_action); + SET_DEVICE_OP(dev_ops, destroy_qp); + SET_DEVICE_OP(dev_ops, destroy_rwq_ind_table); + SET_DEVICE_OP(dev_ops, destroy_srq); + SET_DEVICE_OP(dev_ops, destroy_wq); + SET_DEVICE_OP(dev_ops, detach_mcast); + SET_DEVICE_OP(dev_ops, disassociate_ucontext); + SET_DEVICE_OP(dev_ops, drain_rq); + SET_DEVICE_OP(dev_ops, drain_sq); + SET_DEVICE_OP(dev_ops, enable_driver); + SET_DEVICE_OP(dev_ops, fill_res_cm_id_entry); + SET_DEVICE_OP(dev_ops, fill_res_cq_entry); + SET_DEVICE_OP(dev_ops, fill_res_cq_entry_raw); + SET_DEVICE_OP(dev_ops, fill_res_mr_entry); + SET_DEVICE_OP(dev_ops, fill_res_mr_entry_raw); + SET_DEVICE_OP(dev_ops, fill_res_qp_entry); + SET_DEVICE_OP(dev_ops, fill_res_qp_entry_raw); + SET_DEVICE_OP(dev_ops, fill_stat_mr_entry); + SET_DEVICE_OP(dev_ops, get_dev_fw_str); + SET_DEVICE_OP(dev_ops, get_dma_mr); + SET_DEVICE_OP(dev_ops, get_hw_stats); + SET_DEVICE_OP(dev_ops, get_link_layer); + SET_DEVICE_OP(dev_ops, get_netdev); + SET_DEVICE_OP(dev_ops, get_port_immutable); + SET_DEVICE_OP(dev_ops, get_vector_affinity); + SET_DEVICE_OP(dev_ops, get_vf_config); + SET_DEVICE_OP(dev_ops, get_vf_guid); + SET_DEVICE_OP(dev_ops, get_vf_stats); + SET_DEVICE_OP(dev_ops, init_port); + SET_DEVICE_OP(dev_ops, iw_accept); + SET_DEVICE_OP(dev_ops, iw_add_ref); + SET_DEVICE_OP(dev_ops, iw_connect); + SET_DEVICE_OP(dev_ops, iw_create_listen); + SET_DEVICE_OP(dev_ops, iw_destroy_listen); + SET_DEVICE_OP(dev_ops, iw_get_qp); + SET_DEVICE_OP(dev_ops, iw_reject); + SET_DEVICE_OP(dev_ops, iw_rem_ref); + SET_DEVICE_OP(dev_ops, map_mr_sg); + SET_DEVICE_OP(dev_ops, map_mr_sg_pi); + SET_DEVICE_OP(dev_ops, mmap); + SET_DEVICE_OP(dev_ops, mmap_free); + SET_DEVICE_OP(dev_ops, modify_ah); + SET_DEVICE_OP(dev_ops, modify_cq); + SET_DEVICE_OP(dev_ops, modify_device); + SET_DEVICE_OP(dev_ops, modify_flow_action_esp); + SET_DEVICE_OP(dev_ops, modify_port); + SET_DEVICE_OP(dev_ops, modify_qp); + SET_DEVICE_OP(dev_ops, modify_srq); + SET_DEVICE_OP(dev_ops, modify_wq); + SET_DEVICE_OP(dev_ops, peek_cq); + SET_DEVICE_OP(dev_ops, poll_cq); + SET_DEVICE_OP(dev_ops, post_recv); + SET_DEVICE_OP(dev_ops, post_send); + SET_DEVICE_OP(dev_ops, post_srq_recv); + SET_DEVICE_OP(dev_ops, process_mad); + SET_DEVICE_OP(dev_ops, query_ah); + SET_DEVICE_OP(dev_ops, query_device); + SET_DEVICE_OP(dev_ops, query_gid); + SET_DEVICE_OP(dev_ops, query_pkey); + SET_DEVICE_OP(dev_ops, query_port); + SET_DEVICE_OP(dev_ops, query_qp); + SET_DEVICE_OP(dev_ops, query_srq); + SET_DEVICE_OP(dev_ops, query_ucontext); + SET_DEVICE_OP(dev_ops, rdma_netdev_get_params); + SET_DEVICE_OP(dev_ops, read_counters); + SET_DEVICE_OP(dev_ops, reg_dm_mr); + SET_DEVICE_OP(dev_ops, reg_user_mr); + SET_DEVICE_OP(dev_ops, req_ncomp_notif); + SET_DEVICE_OP(dev_ops, req_notify_cq); + SET_DEVICE_OP(dev_ops, rereg_user_mr); + SET_DEVICE_OP(dev_ops, resize_cq); + SET_DEVICE_OP(dev_ops, set_vf_guid); + SET_DEVICE_OP(dev_ops, set_vf_link_state); + + SET_OBJ_SIZE(dev_ops, ib_ah); + SET_OBJ_SIZE(dev_ops, ib_counters); + SET_OBJ_SIZE(dev_ops, ib_cq); + SET_OBJ_SIZE(dev_ops, ib_mw); + SET_OBJ_SIZE(dev_ops, ib_pd); + SET_OBJ_SIZE(dev_ops, ib_rwq_ind_table); + SET_OBJ_SIZE(dev_ops, ib_srq); + SET_OBJ_SIZE(dev_ops, ib_ucontext); + SET_OBJ_SIZE(dev_ops, ib_xrcd); +} +EXPORT_SYMBOL(ib_set_device_ops); + +#ifdef CONFIG_INFINIBAND_VIRT_DMA +int ib_dma_virt_map_sg(struct ib_device *dev, struct scatterlist *sg, int nents) +{ + struct scatterlist *s; + int i; + + for_each_sg(sg, s, nents, i) { + sg_dma_address(s) = (uintptr_t)sg_virt(s); + sg_dma_len(s) = s->length; + } + return nents; +} +EXPORT_SYMBOL(ib_dma_virt_map_sg); +#endif /* CONFIG_INFINIBAND_VIRT_DMA */ + +static const struct rdma_nl_cbs ibnl_ls_cb_table[RDMA_NL_LS_NUM_OPS] = { + [RDMA_NL_LS_OP_RESOLVE] = { + .doit = ib_nl_handle_resolve_resp, + .flags = RDMA_NL_ADMIN_PERM, + }, + [RDMA_NL_LS_OP_SET_TIMEOUT] = { + .doit = ib_nl_handle_set_timeout, + .flags = RDMA_NL_ADMIN_PERM, + }, + [RDMA_NL_LS_OP_IP_RESOLVE] = { + .doit = ib_nl_handle_ip_res_resp, + .flags = RDMA_NL_ADMIN_PERM, + }, +}; + +static int __init ib_core_init(void) +{ + int ret; + + ib_wq = alloc_workqueue("infiniband", 0, 0); + if (!ib_wq) + return -ENOMEM; + + ib_comp_wq = alloc_workqueue("ib-comp-wq", + WQ_HIGHPRI | WQ_MEM_RECLAIM | WQ_SYSFS, 0); + if (!ib_comp_wq) { + ret = -ENOMEM; + goto err; + } + + ib_comp_unbound_wq = + alloc_workqueue("ib-comp-unb-wq", + WQ_UNBOUND | WQ_HIGHPRI | WQ_MEM_RECLAIM | + WQ_SYSFS, WQ_UNBOUND_MAX_ACTIVE); + if (!ib_comp_unbound_wq) { + ret = -ENOMEM; + goto err_comp; + } + + ret = class_register(&ib_class); + if (ret) { + pr_warn("Couldn't create InfiniBand device class\n"); + goto err_comp_unbound; + } + + rdma_nl_init(); + + ret = addr_init(); + if (ret) { + pr_warn("Couldn't init IB address resolution\n"); + goto err_ibnl; + } + + ret = ib_mad_init(); + if (ret) { + pr_warn("Couldn't init IB MAD\n"); + goto err_addr; + } + + ret = ib_sa_init(); + if (ret) { + pr_warn("Couldn't init SA\n"); + goto err_mad; + } + + ret = register_blocking_lsm_notifier(&ibdev_lsm_nb); + if (ret) { + pr_warn("Couldn't register LSM notifier. ret %d\n", ret); + goto err_sa; + } + + ret = register_pernet_device(&rdma_dev_net_ops); + if (ret) { + pr_warn("Couldn't init compat dev. ret %d\n", ret); + goto err_compat; + } + + nldev_init(); + rdma_nl_register(RDMA_NL_LS, ibnl_ls_cb_table); + ret = roce_gid_mgmt_init(); + if (ret) { + pr_warn("Couldn't init RoCE GID management\n"); + goto err_parent; + } + + return 0; + +err_parent: + rdma_nl_unregister(RDMA_NL_LS); + nldev_exit(); + unregister_pernet_device(&rdma_dev_net_ops); +err_compat: + unregister_blocking_lsm_notifier(&ibdev_lsm_nb); +err_sa: + ib_sa_cleanup(); +err_mad: + ib_mad_cleanup(); +err_addr: + addr_cleanup(); +err_ibnl: + class_unregister(&ib_class); +err_comp_unbound: + destroy_workqueue(ib_comp_unbound_wq); +err_comp: + destroy_workqueue(ib_comp_wq); +err: + destroy_workqueue(ib_wq); + return ret; +} + +static void __exit ib_core_cleanup(void) +{ + roce_gid_mgmt_cleanup(); + rdma_nl_unregister(RDMA_NL_LS); + nldev_exit(); + unregister_pernet_device(&rdma_dev_net_ops); + unregister_blocking_lsm_notifier(&ibdev_lsm_nb); + ib_sa_cleanup(); + ib_mad_cleanup(); + addr_cleanup(); + rdma_nl_exit(); + class_unregister(&ib_class); + destroy_workqueue(ib_comp_unbound_wq); + destroy_workqueue(ib_comp_wq); + /* Make sure that any pending umem accounting work is done. */ + destroy_workqueue(ib_wq); + flush_workqueue(system_unbound_wq); + WARN_ON(!xa_empty(&clients)); + WARN_ON(!xa_empty(&devices)); +} + +MODULE_ALIAS_RDMA_NETLINK(RDMA_NL_LS, 4); + +/* ib core relies on netdev stack to first register net_ns_type_operations + * ns kobject type before ib_core initialization. + */ +fs_initcall(ib_core_init); +module_exit(ib_core_cleanup); diff --git a/drivers/infiniband/core/ib_core_uverbs.c b/drivers/infiniband/core/ib_core_uverbs.c new file mode 100644 index 000000000..b51bd7087 --- /dev/null +++ b/drivers/infiniband/core/ib_core_uverbs.c @@ -0,0 +1,367 @@ +// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB +/* + * Copyright (c) 2005 Mellanox Technologies. All rights reserved. + * Copyright 2018-2019 Amazon.com, Inc. or its affiliates. All rights reserved. + * Copyright 2019 Marvell. All rights reserved. + */ +#include <linux/xarray.h> +#include "uverbs.h" +#include "core_priv.h" + +/** + * rdma_umap_priv_init() - Initialize the private data of a vma + * + * @priv: The already allocated private data + * @vma: The vm area struct that needs private data + * @entry: entry into the mmap_xa that needs to be linked with + * this vma + * + * Each time we map IO memory into user space this keeps track of the + * mapping. When the device is hot-unplugged we 'zap' the mmaps in user space + * to point to the zero page and allow the hot unplug to proceed. + * + * This is necessary for cases like PCI physical hot unplug as the actual BAR + * memory may vanish after this and access to it from userspace could MCE. + * + * RDMA drivers supporting disassociation must have their user space designed + * to cope in some way with their IO pages going to the zero page. + * + */ +void rdma_umap_priv_init(struct rdma_umap_priv *priv, + struct vm_area_struct *vma, + struct rdma_user_mmap_entry *entry) +{ + struct ib_uverbs_file *ufile = vma->vm_file->private_data; + + priv->vma = vma; + if (entry) { + kref_get(&entry->ref); + priv->entry = entry; + } + vma->vm_private_data = priv; + /* vm_ops is setup in ib_uverbs_mmap() to avoid module dependencies */ + + mutex_lock(&ufile->umap_lock); + list_add(&priv->list, &ufile->umaps); + mutex_unlock(&ufile->umap_lock); +} +EXPORT_SYMBOL(rdma_umap_priv_init); + +/** + * rdma_user_mmap_io() - Map IO memory into a process + * + * @ucontext: associated user context + * @vma: the vma related to the current mmap call + * @pfn: pfn to map + * @size: size to map + * @prot: pgprot to use in remap call + * @entry: mmap_entry retrieved from rdma_user_mmap_entry_get(), or NULL + * if mmap_entry is not used by the driver + * + * This is to be called by drivers as part of their mmap() functions if they + * wish to send something like PCI-E BAR memory to userspace. + * + * Return -EINVAL on wrong flags or size, -EAGAIN on failure to map. 0 on + * success. + */ +int rdma_user_mmap_io(struct ib_ucontext *ucontext, struct vm_area_struct *vma, + unsigned long pfn, unsigned long size, pgprot_t prot, + struct rdma_user_mmap_entry *entry) +{ + struct ib_uverbs_file *ufile = ucontext->ufile; + struct rdma_umap_priv *priv; + + if (!(vma->vm_flags & VM_SHARED)) + return -EINVAL; + + if (vma->vm_end - vma->vm_start != size) + return -EINVAL; + + /* Driver is using this wrong, must be called by ib_uverbs_mmap */ + if (WARN_ON(!vma->vm_file || + vma->vm_file->private_data != ufile)) + return -EINVAL; + lockdep_assert_held(&ufile->device->disassociate_srcu); + + priv = kzalloc(sizeof(*priv), GFP_KERNEL); + if (!priv) + return -ENOMEM; + + vma->vm_page_prot = prot; + if (io_remap_pfn_range(vma, vma->vm_start, pfn, size, prot)) { + kfree(priv); + return -EAGAIN; + } + + rdma_umap_priv_init(priv, vma, entry); + return 0; +} +EXPORT_SYMBOL(rdma_user_mmap_io); + +/** + * rdma_user_mmap_entry_get_pgoff() - Get an entry from the mmap_xa + * + * @ucontext: associated user context + * @pgoff: The mmap offset >> PAGE_SHIFT + * + * This function is called when a user tries to mmap with an offset (returned + * by rdma_user_mmap_get_offset()) it initially received from the driver. The + * rdma_user_mmap_entry was created by the function + * rdma_user_mmap_entry_insert(). This function increases the refcnt of the + * entry so that it won't be deleted from the xarray in the meantime. + * + * Return an reference to an entry if exists or NULL if there is no + * match. rdma_user_mmap_entry_put() must be called to put the reference. + */ +struct rdma_user_mmap_entry * +rdma_user_mmap_entry_get_pgoff(struct ib_ucontext *ucontext, + unsigned long pgoff) +{ + struct rdma_user_mmap_entry *entry; + + if (pgoff > U32_MAX) + return NULL; + + xa_lock(&ucontext->mmap_xa); + + entry = xa_load(&ucontext->mmap_xa, pgoff); + + /* + * If refcount is zero, entry is already being deleted, driver_removed + * indicates that the no further mmaps are possible and we waiting for + * the active VMAs to be closed. + */ + if (!entry || entry->start_pgoff != pgoff || entry->driver_removed || + !kref_get_unless_zero(&entry->ref)) + goto err; + + xa_unlock(&ucontext->mmap_xa); + + ibdev_dbg(ucontext->device, "mmap: pgoff[%#lx] npages[%#zx] returned\n", + pgoff, entry->npages); + + return entry; + +err: + xa_unlock(&ucontext->mmap_xa); + return NULL; +} +EXPORT_SYMBOL(rdma_user_mmap_entry_get_pgoff); + +/** + * rdma_user_mmap_entry_get() - Get an entry from the mmap_xa + * + * @ucontext: associated user context + * @vma: the vma being mmap'd into + * + * This function is like rdma_user_mmap_entry_get_pgoff() except that it also + * checks that the VMA is correct. + */ +struct rdma_user_mmap_entry * +rdma_user_mmap_entry_get(struct ib_ucontext *ucontext, + struct vm_area_struct *vma) +{ + struct rdma_user_mmap_entry *entry; + + if (!(vma->vm_flags & VM_SHARED)) + return NULL; + entry = rdma_user_mmap_entry_get_pgoff(ucontext, vma->vm_pgoff); + if (!entry) + return NULL; + if (entry->npages * PAGE_SIZE != vma->vm_end - vma->vm_start) { + rdma_user_mmap_entry_put(entry); + return NULL; + } + return entry; +} +EXPORT_SYMBOL(rdma_user_mmap_entry_get); + +static void rdma_user_mmap_entry_free(struct kref *kref) +{ + struct rdma_user_mmap_entry *entry = + container_of(kref, struct rdma_user_mmap_entry, ref); + struct ib_ucontext *ucontext = entry->ucontext; + unsigned long i; + + /* + * Erase all entries occupied by this single entry, this is deferred + * until all VMA are closed so that the mmap offsets remain unique. + */ + xa_lock(&ucontext->mmap_xa); + for (i = 0; i < entry->npages; i++) + __xa_erase(&ucontext->mmap_xa, entry->start_pgoff + i); + xa_unlock(&ucontext->mmap_xa); + + ibdev_dbg(ucontext->device, "mmap: pgoff[%#lx] npages[%#zx] removed\n", + entry->start_pgoff, entry->npages); + + if (ucontext->device->ops.mmap_free) + ucontext->device->ops.mmap_free(entry); +} + +/** + * rdma_user_mmap_entry_put() - Drop reference to the mmap entry + * + * @entry: an entry in the mmap_xa + * + * This function is called when the mapping is closed if it was + * an io mapping or when the driver is done with the entry for + * some other reason. + * Should be called after rdma_user_mmap_entry_get was called + * and entry is no longer needed. This function will erase the + * entry and free it if its refcnt reaches zero. + */ +void rdma_user_mmap_entry_put(struct rdma_user_mmap_entry *entry) +{ + kref_put(&entry->ref, rdma_user_mmap_entry_free); +} +EXPORT_SYMBOL(rdma_user_mmap_entry_put); + +/** + * rdma_user_mmap_entry_remove() - Drop reference to entry and + * mark it as unmmapable + * + * @entry: the entry to insert into the mmap_xa + * + * Drivers can call this to prevent userspace from creating more mappings for + * entry, however existing mmaps continue to exist and ops->mmap_free() will + * not be called until all user mmaps are destroyed. + */ +void rdma_user_mmap_entry_remove(struct rdma_user_mmap_entry *entry) +{ + if (!entry) + return; + + xa_lock(&entry->ucontext->mmap_xa); + entry->driver_removed = true; + xa_unlock(&entry->ucontext->mmap_xa); + kref_put(&entry->ref, rdma_user_mmap_entry_free); +} +EXPORT_SYMBOL(rdma_user_mmap_entry_remove); + +/** + * rdma_user_mmap_entry_insert_range() - Insert an entry to the mmap_xa + * in a given range. + * + * @ucontext: associated user context. + * @entry: the entry to insert into the mmap_xa + * @length: length of the address that will be mmapped + * @min_pgoff: minimum pgoff to be returned + * @max_pgoff: maximum pgoff to be returned + * + * This function should be called by drivers that use the rdma_user_mmap + * interface for implementing their mmap syscall A database of mmap offsets is + * handled in the core and helper functions are provided to insert entries + * into the database and extract entries when the user calls mmap with the + * given offset. The function allocates a unique page offset in a given range + * that should be provided to user, the user will use the offset to retrieve + * information such as address to be mapped and how. + * + * Return: 0 on success and -ENOMEM on failure + */ +int rdma_user_mmap_entry_insert_range(struct ib_ucontext *ucontext, + struct rdma_user_mmap_entry *entry, + size_t length, u32 min_pgoff, + u32 max_pgoff) +{ + struct ib_uverbs_file *ufile = ucontext->ufile; + XA_STATE(xas, &ucontext->mmap_xa, min_pgoff); + u32 xa_first, xa_last, npages; + int err; + u32 i; + + if (!entry) + return -EINVAL; + + kref_init(&entry->ref); + entry->ucontext = ucontext; + + /* + * We want the whole allocation to be done without interruption from a + * different thread. The allocation requires finding a free range and + * storing. During the xa_insert the lock could be released, possibly + * allowing another thread to choose the same range. + */ + mutex_lock(&ufile->umap_lock); + + xa_lock(&ucontext->mmap_xa); + + /* We want to find an empty range */ + npages = (u32)DIV_ROUND_UP(length, PAGE_SIZE); + entry->npages = npages; + while (true) { + /* First find an empty index */ + xas_find_marked(&xas, max_pgoff, XA_FREE_MARK); + if (xas.xa_node == XAS_RESTART) + goto err_unlock; + + xa_first = xas.xa_index; + + /* Is there enough room to have the range? */ + if (check_add_overflow(xa_first, npages, &xa_last)) + goto err_unlock; + + /* + * Now look for the next present entry. If an entry doesn't + * exist, we found an empty range and can proceed. + */ + xas_next_entry(&xas, xa_last - 1); + if (xas.xa_node == XAS_BOUNDS || xas.xa_index >= xa_last) + break; + } + + for (i = xa_first; i < xa_last; i++) { + err = __xa_insert(&ucontext->mmap_xa, i, entry, GFP_KERNEL); + if (err) + goto err_undo; + } + + /* + * Internally the kernel uses a page offset, in libc this is a byte + * offset. Drivers should not return pgoff to userspace. + */ + entry->start_pgoff = xa_first; + xa_unlock(&ucontext->mmap_xa); + mutex_unlock(&ufile->umap_lock); + + ibdev_dbg(ucontext->device, "mmap: pgoff[%#lx] npages[%#x] inserted\n", + entry->start_pgoff, npages); + + return 0; + +err_undo: + for (; i > xa_first; i--) + __xa_erase(&ucontext->mmap_xa, i - 1); + +err_unlock: + xa_unlock(&ucontext->mmap_xa); + mutex_unlock(&ufile->umap_lock); + return -ENOMEM; +} +EXPORT_SYMBOL(rdma_user_mmap_entry_insert_range); + +/** + * rdma_user_mmap_entry_insert() - Insert an entry to the mmap_xa. + * + * @ucontext: associated user context. + * @entry: the entry to insert into the mmap_xa + * @length: length of the address that will be mmapped + * + * This function should be called by drivers that use the rdma_user_mmap + * interface for handling user mmapped addresses. The database is handled in + * the core and helper functions are provided to insert entries into the + * database and extract entries when the user calls mmap with the given offset. + * The function allocates a unique page offset that should be provided to user, + * the user will use the offset to retrieve information such as address to + * be mapped and how. + * + * Return: 0 on success and -ENOMEM on failure + */ +int rdma_user_mmap_entry_insert(struct ib_ucontext *ucontext, + struct rdma_user_mmap_entry *entry, + size_t length) +{ + return rdma_user_mmap_entry_insert_range(ucontext, entry, length, 0, + U32_MAX); +} +EXPORT_SYMBOL(rdma_user_mmap_entry_insert); diff --git a/drivers/infiniband/core/iwcm.c b/drivers/infiniband/core/iwcm.c new file mode 100644 index 000000000..75b6da000 --- /dev/null +++ b/drivers/infiniband/core/iwcm.c @@ -0,0 +1,1224 @@ +/* + * Copyright (c) 2004, 2005 Intel Corporation. All rights reserved. + * Copyright (c) 2004 Topspin Corporation. All rights reserved. + * Copyright (c) 2004, 2005 Voltaire Corporation. All rights reserved. + * Copyright (c) 2005 Sun Microsystems, Inc. All rights reserved. + * Copyright (c) 2005 Open Grid Computing, Inc. All rights reserved. + * Copyright (c) 2005 Network Appliance, Inc. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + */ +#include <linux/dma-mapping.h> +#include <linux/err.h> +#include <linux/idr.h> +#include <linux/interrupt.h> +#include <linux/rbtree.h> +#include <linux/sched.h> +#include <linux/spinlock.h> +#include <linux/workqueue.h> +#include <linux/completion.h> +#include <linux/slab.h> +#include <linux/module.h> +#include <linux/sysctl.h> + +#include <rdma/iw_cm.h> +#include <rdma/ib_addr.h> +#include <rdma/iw_portmap.h> +#include <rdma/rdma_netlink.h> + +#include "iwcm.h" + +MODULE_AUTHOR("Tom Tucker"); +MODULE_DESCRIPTION("iWARP CM"); +MODULE_LICENSE("Dual BSD/GPL"); + +static const char * const iwcm_rej_reason_strs[] = { + [ECONNRESET] = "reset by remote host", + [ECONNREFUSED] = "refused by remote application", + [ETIMEDOUT] = "setup timeout", +}; + +const char *__attribute_const__ iwcm_reject_msg(int reason) +{ + size_t index; + + /* iWARP uses negative errnos */ + index = -reason; + + if (index < ARRAY_SIZE(iwcm_rej_reason_strs) && + iwcm_rej_reason_strs[index]) + return iwcm_rej_reason_strs[index]; + else + return "unrecognized reason"; +} +EXPORT_SYMBOL(iwcm_reject_msg); + +static struct rdma_nl_cbs iwcm_nl_cb_table[RDMA_NL_IWPM_NUM_OPS] = { + [RDMA_NL_IWPM_REG_PID] = {.dump = iwpm_register_pid_cb}, + [RDMA_NL_IWPM_ADD_MAPPING] = {.dump = iwpm_add_mapping_cb}, + [RDMA_NL_IWPM_QUERY_MAPPING] = {.dump = iwpm_add_and_query_mapping_cb}, + [RDMA_NL_IWPM_REMOTE_INFO] = {.dump = iwpm_remote_info_cb}, + [RDMA_NL_IWPM_HANDLE_ERR] = {.dump = iwpm_mapping_error_cb}, + [RDMA_NL_IWPM_MAPINFO] = {.dump = iwpm_mapping_info_cb}, + [RDMA_NL_IWPM_MAPINFO_NUM] = {.dump = iwpm_ack_mapping_info_cb}, + [RDMA_NL_IWPM_HELLO] = {.dump = iwpm_hello_cb} +}; + +static struct workqueue_struct *iwcm_wq; +struct iwcm_work { + struct work_struct work; + struct iwcm_id_private *cm_id; + struct list_head list; + struct iw_cm_event event; + struct list_head free_list; +}; + +static unsigned int default_backlog = 256; + +static struct ctl_table_header *iwcm_ctl_table_hdr; +static struct ctl_table iwcm_ctl_table[] = { + { + .procname = "default_backlog", + .data = &default_backlog, + .maxlen = sizeof(default_backlog), + .mode = 0644, + .proc_handler = proc_dointvec, + }, + { } +}; + +/* + * The following services provide a mechanism for pre-allocating iwcm_work + * elements. The design pre-allocates them based on the cm_id type: + * LISTENING IDS: Get enough elements preallocated to handle the + * listen backlog. + * ACTIVE IDS: 4: CONNECT_REPLY, ESTABLISHED, DISCONNECT, CLOSE + * PASSIVE IDS: 3: ESTABLISHED, DISCONNECT, CLOSE + * + * Allocating them in connect and listen avoids having to deal + * with allocation failures on the event upcall from the provider (which + * is called in the interrupt context). + * + * One exception is when creating the cm_id for incoming connection requests. + * There are two cases: + * 1) in the event upcall, cm_event_handler(), for a listening cm_id. If + * the backlog is exceeded, then no more connection request events will + * be processed. cm_event_handler() returns -ENOMEM in this case. Its up + * to the provider to reject the connection request. + * 2) in the connection request workqueue handler, cm_conn_req_handler(). + * If work elements cannot be allocated for the new connect request cm_id, + * then IWCM will call the provider reject method. This is ok since + * cm_conn_req_handler() runs in the workqueue thread context. + */ + +static struct iwcm_work *get_work(struct iwcm_id_private *cm_id_priv) +{ + struct iwcm_work *work; + + if (list_empty(&cm_id_priv->work_free_list)) + return NULL; + work = list_entry(cm_id_priv->work_free_list.next, struct iwcm_work, + free_list); + list_del_init(&work->free_list); + return work; +} + +static void put_work(struct iwcm_work *work) +{ + list_add(&work->free_list, &work->cm_id->work_free_list); +} + +static void dealloc_work_entries(struct iwcm_id_private *cm_id_priv) +{ + struct list_head *e, *tmp; + + list_for_each_safe(e, tmp, &cm_id_priv->work_free_list) { + list_del(e); + kfree(list_entry(e, struct iwcm_work, free_list)); + } +} + +static int alloc_work_entries(struct iwcm_id_private *cm_id_priv, int count) +{ + struct iwcm_work *work; + + BUG_ON(!list_empty(&cm_id_priv->work_free_list)); + while (count--) { + work = kmalloc(sizeof(struct iwcm_work), GFP_KERNEL); + if (!work) { + dealloc_work_entries(cm_id_priv); + return -ENOMEM; + } + work->cm_id = cm_id_priv; + INIT_LIST_HEAD(&work->list); + put_work(work); + } + return 0; +} + +/* + * Save private data from incoming connection requests to + * iw_cm_event, so the low level driver doesn't have to. Adjust + * the event ptr to point to the local copy. + */ +static int copy_private_data(struct iw_cm_event *event) +{ + void *p; + + p = kmemdup(event->private_data, event->private_data_len, GFP_ATOMIC); + if (!p) + return -ENOMEM; + event->private_data = p; + return 0; +} + +static void free_cm_id(struct iwcm_id_private *cm_id_priv) +{ + dealloc_work_entries(cm_id_priv); + kfree(cm_id_priv); +} + +/* + * Release a reference on cm_id. If the last reference is being + * released, free the cm_id and return 1. + */ +static int iwcm_deref_id(struct iwcm_id_private *cm_id_priv) +{ + BUG_ON(atomic_read(&cm_id_priv->refcount)==0); + if (atomic_dec_and_test(&cm_id_priv->refcount)) { + BUG_ON(!list_empty(&cm_id_priv->work_list)); + free_cm_id(cm_id_priv); + return 1; + } + + return 0; +} + +static void add_ref(struct iw_cm_id *cm_id) +{ + struct iwcm_id_private *cm_id_priv; + cm_id_priv = container_of(cm_id, struct iwcm_id_private, id); + atomic_inc(&cm_id_priv->refcount); +} + +static void rem_ref(struct iw_cm_id *cm_id) +{ + struct iwcm_id_private *cm_id_priv; + + cm_id_priv = container_of(cm_id, struct iwcm_id_private, id); + + (void)iwcm_deref_id(cm_id_priv); +} + +static int cm_event_handler(struct iw_cm_id *cm_id, struct iw_cm_event *event); + +struct iw_cm_id *iw_create_cm_id(struct ib_device *device, + iw_cm_handler cm_handler, + void *context) +{ + struct iwcm_id_private *cm_id_priv; + + cm_id_priv = kzalloc(sizeof(*cm_id_priv), GFP_KERNEL); + if (!cm_id_priv) + return ERR_PTR(-ENOMEM); + + cm_id_priv->state = IW_CM_STATE_IDLE; + cm_id_priv->id.device = device; + cm_id_priv->id.cm_handler = cm_handler; + cm_id_priv->id.context = context; + cm_id_priv->id.event_handler = cm_event_handler; + cm_id_priv->id.add_ref = add_ref; + cm_id_priv->id.rem_ref = rem_ref; + spin_lock_init(&cm_id_priv->lock); + atomic_set(&cm_id_priv->refcount, 1); + init_waitqueue_head(&cm_id_priv->connect_wait); + init_completion(&cm_id_priv->destroy_comp); + INIT_LIST_HEAD(&cm_id_priv->work_list); + INIT_LIST_HEAD(&cm_id_priv->work_free_list); + + return &cm_id_priv->id; +} +EXPORT_SYMBOL(iw_create_cm_id); + + +static int iwcm_modify_qp_err(struct ib_qp *qp) +{ + struct ib_qp_attr qp_attr; + + if (!qp) + return -EINVAL; + + qp_attr.qp_state = IB_QPS_ERR; + return ib_modify_qp(qp, &qp_attr, IB_QP_STATE); +} + +/* + * This is really the RDMAC CLOSING state. It is most similar to the + * IB SQD QP state. + */ +static int iwcm_modify_qp_sqd(struct ib_qp *qp) +{ + struct ib_qp_attr qp_attr; + + BUG_ON(qp == NULL); + qp_attr.qp_state = IB_QPS_SQD; + return ib_modify_qp(qp, &qp_attr, IB_QP_STATE); +} + +/* + * CM_ID <-- CLOSING + * + * Block if a passive or active connection is currently being processed. Then + * process the event as follows: + * - If we are ESTABLISHED, move to CLOSING and modify the QP state + * based on the abrupt flag + * - If the connection is already in the CLOSING or IDLE state, the peer is + * disconnecting concurrently with us and we've already seen the + * DISCONNECT event -- ignore the request and return 0 + * - Disconnect on a listening endpoint returns -EINVAL + */ +int iw_cm_disconnect(struct iw_cm_id *cm_id, int abrupt) +{ + struct iwcm_id_private *cm_id_priv; + unsigned long flags; + int ret = 0; + struct ib_qp *qp = NULL; + + cm_id_priv = container_of(cm_id, struct iwcm_id_private, id); + /* Wait if we're currently in a connect or accept downcall */ + wait_event(cm_id_priv->connect_wait, + !test_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags)); + + spin_lock_irqsave(&cm_id_priv->lock, flags); + switch (cm_id_priv->state) { + case IW_CM_STATE_ESTABLISHED: + cm_id_priv->state = IW_CM_STATE_CLOSING; + + /* QP could be <nul> for user-mode client */ + if (cm_id_priv->qp) + qp = cm_id_priv->qp; + else + ret = -EINVAL; + break; + case IW_CM_STATE_LISTEN: + ret = -EINVAL; + break; + case IW_CM_STATE_CLOSING: + /* remote peer closed first */ + case IW_CM_STATE_IDLE: + /* accept or connect returned !0 */ + break; + case IW_CM_STATE_CONN_RECV: + /* + * App called disconnect before/without calling accept after + * connect_request event delivered. + */ + break; + case IW_CM_STATE_CONN_SENT: + /* Can only get here if wait above fails */ + default: + BUG(); + } + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + + if (qp) { + if (abrupt) + ret = iwcm_modify_qp_err(qp); + else + ret = iwcm_modify_qp_sqd(qp); + + /* + * If both sides are disconnecting the QP could + * already be in ERR or SQD states + */ + ret = 0; + } + + return ret; +} +EXPORT_SYMBOL(iw_cm_disconnect); + +/* + * CM_ID <-- DESTROYING + * + * Clean up all resources associated with the connection and release + * the initial reference taken by iw_create_cm_id. + */ +static void destroy_cm_id(struct iw_cm_id *cm_id) +{ + struct iwcm_id_private *cm_id_priv; + struct ib_qp *qp; + unsigned long flags; + + cm_id_priv = container_of(cm_id, struct iwcm_id_private, id); + /* + * Wait if we're currently in a connect or accept downcall. A + * listening endpoint should never block here. + */ + wait_event(cm_id_priv->connect_wait, + !test_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags)); + + /* + * Since we're deleting the cm_id, drop any events that + * might arrive before the last dereference. + */ + set_bit(IWCM_F_DROP_EVENTS, &cm_id_priv->flags); + + spin_lock_irqsave(&cm_id_priv->lock, flags); + qp = cm_id_priv->qp; + cm_id_priv->qp = NULL; + + switch (cm_id_priv->state) { + case IW_CM_STATE_LISTEN: + cm_id_priv->state = IW_CM_STATE_DESTROYING; + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + /* destroy the listening endpoint */ + cm_id->device->ops.iw_destroy_listen(cm_id); + spin_lock_irqsave(&cm_id_priv->lock, flags); + break; + case IW_CM_STATE_ESTABLISHED: + cm_id_priv->state = IW_CM_STATE_DESTROYING; + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + /* Abrupt close of the connection */ + (void)iwcm_modify_qp_err(qp); + spin_lock_irqsave(&cm_id_priv->lock, flags); + break; + case IW_CM_STATE_IDLE: + case IW_CM_STATE_CLOSING: + cm_id_priv->state = IW_CM_STATE_DESTROYING; + break; + case IW_CM_STATE_CONN_RECV: + /* + * App called destroy before/without calling accept after + * receiving connection request event notification or + * returned non zero from the event callback function. + * In either case, must tell the provider to reject. + */ + cm_id_priv->state = IW_CM_STATE_DESTROYING; + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + cm_id->device->ops.iw_reject(cm_id, NULL, 0); + spin_lock_irqsave(&cm_id_priv->lock, flags); + break; + case IW_CM_STATE_CONN_SENT: + case IW_CM_STATE_DESTROYING: + default: + BUG(); + break; + } + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + if (qp) + cm_id_priv->id.device->ops.iw_rem_ref(qp); + + if (cm_id->mapped) { + iwpm_remove_mapinfo(&cm_id->local_addr, &cm_id->m_local_addr); + iwpm_remove_mapping(&cm_id->local_addr, RDMA_NL_IWCM); + } + + (void)iwcm_deref_id(cm_id_priv); +} + +/* + * This function is only called by the application thread and cannot + * be called by the event thread. The function will wait for all + * references to be released on the cm_id and then kfree the cm_id + * object. + */ +void iw_destroy_cm_id(struct iw_cm_id *cm_id) +{ + destroy_cm_id(cm_id); +} +EXPORT_SYMBOL(iw_destroy_cm_id); + +/** + * iw_cm_check_wildcard - If IP address is 0 then use original + * @pm_addr: sockaddr containing the ip to check for wildcard + * @cm_addr: sockaddr containing the actual IP address + * @cm_outaddr: sockaddr to set IP addr which leaving port + * + * Checks the pm_addr for wildcard and then sets cm_outaddr's + * IP to the actual (cm_addr). + */ +static void iw_cm_check_wildcard(struct sockaddr_storage *pm_addr, + struct sockaddr_storage *cm_addr, + struct sockaddr_storage *cm_outaddr) +{ + if (pm_addr->ss_family == AF_INET) { + struct sockaddr_in *pm4_addr = (struct sockaddr_in *)pm_addr; + + if (pm4_addr->sin_addr.s_addr == htonl(INADDR_ANY)) { + struct sockaddr_in *cm4_addr = + (struct sockaddr_in *)cm_addr; + struct sockaddr_in *cm4_outaddr = + (struct sockaddr_in *)cm_outaddr; + + cm4_outaddr->sin_addr = cm4_addr->sin_addr; + } + } else { + struct sockaddr_in6 *pm6_addr = (struct sockaddr_in6 *)pm_addr; + + if (ipv6_addr_type(&pm6_addr->sin6_addr) == IPV6_ADDR_ANY) { + struct sockaddr_in6 *cm6_addr = + (struct sockaddr_in6 *)cm_addr; + struct sockaddr_in6 *cm6_outaddr = + (struct sockaddr_in6 *)cm_outaddr; + + cm6_outaddr->sin6_addr = cm6_addr->sin6_addr; + } + } +} + +/** + * iw_cm_map - Use portmapper to map the ports + * @cm_id: connection manager pointer + * @active: Indicates the active side when true + * returns nonzero for error only if iwpm_create_mapinfo() fails + * + * Tries to add a mapping for a port using the Portmapper. If + * successful in mapping the IP/Port it will check the remote + * mapped IP address for a wildcard IP address and replace the + * zero IP address with the remote_addr. + */ +static int iw_cm_map(struct iw_cm_id *cm_id, bool active) +{ + const char *devname = dev_name(&cm_id->device->dev); + const char *ifname = cm_id->device->iw_ifname; + struct iwpm_dev_data pm_reg_msg = {}; + struct iwpm_sa_data pm_msg; + int status; + + if (strlen(devname) >= sizeof(pm_reg_msg.dev_name) || + strlen(ifname) >= sizeof(pm_reg_msg.if_name)) + return -EINVAL; + + cm_id->m_local_addr = cm_id->local_addr; + cm_id->m_remote_addr = cm_id->remote_addr; + + strcpy(pm_reg_msg.dev_name, devname); + strcpy(pm_reg_msg.if_name, ifname); + + if (iwpm_register_pid(&pm_reg_msg, RDMA_NL_IWCM) || + !iwpm_valid_pid()) + return 0; + + cm_id->mapped = true; + pm_msg.loc_addr = cm_id->local_addr; + pm_msg.rem_addr = cm_id->remote_addr; + pm_msg.flags = (cm_id->device->iw_driver_flags & IW_F_NO_PORT_MAP) ? + IWPM_FLAGS_NO_PORT_MAP : 0; + if (active) + status = iwpm_add_and_query_mapping(&pm_msg, + RDMA_NL_IWCM); + else + status = iwpm_add_mapping(&pm_msg, RDMA_NL_IWCM); + + if (!status) { + cm_id->m_local_addr = pm_msg.mapped_loc_addr; + if (active) { + cm_id->m_remote_addr = pm_msg.mapped_rem_addr; + iw_cm_check_wildcard(&pm_msg.mapped_rem_addr, + &cm_id->remote_addr, + &cm_id->m_remote_addr); + } + } + + return iwpm_create_mapinfo(&cm_id->local_addr, + &cm_id->m_local_addr, + RDMA_NL_IWCM, pm_msg.flags); +} + +/* + * CM_ID <-- LISTEN + * + * Start listening for connect requests. Generates one CONNECT_REQUEST + * event for each inbound connect request. + */ +int iw_cm_listen(struct iw_cm_id *cm_id, int backlog) +{ + struct iwcm_id_private *cm_id_priv; + unsigned long flags; + int ret; + + cm_id_priv = container_of(cm_id, struct iwcm_id_private, id); + + if (!backlog) + backlog = default_backlog; + + ret = alloc_work_entries(cm_id_priv, backlog); + if (ret) + return ret; + + spin_lock_irqsave(&cm_id_priv->lock, flags); + switch (cm_id_priv->state) { + case IW_CM_STATE_IDLE: + cm_id_priv->state = IW_CM_STATE_LISTEN; + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + ret = iw_cm_map(cm_id, false); + if (!ret) + ret = cm_id->device->ops.iw_create_listen(cm_id, + backlog); + if (ret) + cm_id_priv->state = IW_CM_STATE_IDLE; + spin_lock_irqsave(&cm_id_priv->lock, flags); + break; + default: + ret = -EINVAL; + } + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + + return ret; +} +EXPORT_SYMBOL(iw_cm_listen); + +/* + * CM_ID <-- IDLE + * + * Rejects an inbound connection request. No events are generated. + */ +int iw_cm_reject(struct iw_cm_id *cm_id, + const void *private_data, + u8 private_data_len) +{ + struct iwcm_id_private *cm_id_priv; + unsigned long flags; + int ret; + + cm_id_priv = container_of(cm_id, struct iwcm_id_private, id); + set_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags); + + spin_lock_irqsave(&cm_id_priv->lock, flags); + if (cm_id_priv->state != IW_CM_STATE_CONN_RECV) { + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + clear_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags); + wake_up_all(&cm_id_priv->connect_wait); + return -EINVAL; + } + cm_id_priv->state = IW_CM_STATE_IDLE; + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + + ret = cm_id->device->ops.iw_reject(cm_id, private_data, + private_data_len); + + clear_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags); + wake_up_all(&cm_id_priv->connect_wait); + + return ret; +} +EXPORT_SYMBOL(iw_cm_reject); + +/* + * CM_ID <-- ESTABLISHED + * + * Accepts an inbound connection request and generates an ESTABLISHED + * event. Callers of iw_cm_disconnect and iw_destroy_cm_id will block + * until the ESTABLISHED event is received from the provider. + */ +int iw_cm_accept(struct iw_cm_id *cm_id, + struct iw_cm_conn_param *iw_param) +{ + struct iwcm_id_private *cm_id_priv; + struct ib_qp *qp; + unsigned long flags; + int ret; + + cm_id_priv = container_of(cm_id, struct iwcm_id_private, id); + set_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags); + + spin_lock_irqsave(&cm_id_priv->lock, flags); + if (cm_id_priv->state != IW_CM_STATE_CONN_RECV) { + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + clear_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags); + wake_up_all(&cm_id_priv->connect_wait); + return -EINVAL; + } + /* Get the ib_qp given the QPN */ + qp = cm_id->device->ops.iw_get_qp(cm_id->device, iw_param->qpn); + if (!qp) { + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + clear_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags); + wake_up_all(&cm_id_priv->connect_wait); + return -EINVAL; + } + cm_id->device->ops.iw_add_ref(qp); + cm_id_priv->qp = qp; + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + + ret = cm_id->device->ops.iw_accept(cm_id, iw_param); + if (ret) { + /* An error on accept precludes provider events */ + BUG_ON(cm_id_priv->state != IW_CM_STATE_CONN_RECV); + cm_id_priv->state = IW_CM_STATE_IDLE; + spin_lock_irqsave(&cm_id_priv->lock, flags); + qp = cm_id_priv->qp; + cm_id_priv->qp = NULL; + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + if (qp) + cm_id->device->ops.iw_rem_ref(qp); + clear_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags); + wake_up_all(&cm_id_priv->connect_wait); + } + + return ret; +} +EXPORT_SYMBOL(iw_cm_accept); + +/* + * Active Side: CM_ID <-- CONN_SENT + * + * If successful, results in the generation of a CONNECT_REPLY + * event. iw_cm_disconnect and iw_cm_destroy will block until the + * CONNECT_REPLY event is received from the provider. + */ +int iw_cm_connect(struct iw_cm_id *cm_id, struct iw_cm_conn_param *iw_param) +{ + struct iwcm_id_private *cm_id_priv; + int ret; + unsigned long flags; + struct ib_qp *qp = NULL; + + cm_id_priv = container_of(cm_id, struct iwcm_id_private, id); + + ret = alloc_work_entries(cm_id_priv, 4); + if (ret) + return ret; + + set_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags); + spin_lock_irqsave(&cm_id_priv->lock, flags); + + if (cm_id_priv->state != IW_CM_STATE_IDLE) { + ret = -EINVAL; + goto err; + } + + /* Get the ib_qp given the QPN */ + qp = cm_id->device->ops.iw_get_qp(cm_id->device, iw_param->qpn); + if (!qp) { + ret = -EINVAL; + goto err; + } + cm_id->device->ops.iw_add_ref(qp); + cm_id_priv->qp = qp; + cm_id_priv->state = IW_CM_STATE_CONN_SENT; + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + + ret = iw_cm_map(cm_id, true); + if (!ret) + ret = cm_id->device->ops.iw_connect(cm_id, iw_param); + if (!ret) + return 0; /* success */ + + spin_lock_irqsave(&cm_id_priv->lock, flags); + qp = cm_id_priv->qp; + cm_id_priv->qp = NULL; + cm_id_priv->state = IW_CM_STATE_IDLE; +err: + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + if (qp) + cm_id->device->ops.iw_rem_ref(qp); + clear_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags); + wake_up_all(&cm_id_priv->connect_wait); + return ret; +} +EXPORT_SYMBOL(iw_cm_connect); + +/* + * Passive Side: new CM_ID <-- CONN_RECV + * + * Handles an inbound connect request. The function creates a new + * iw_cm_id to represent the new connection and inherits the client + * callback function and other attributes from the listening parent. + * + * The work item contains a pointer to the listen_cm_id and the event. The + * listen_cm_id contains the client cm_handler, context and + * device. These are copied when the device is cloned. The event + * contains the new four tuple. + * + * An error on the child should not affect the parent, so this + * function does not return a value. + */ +static void cm_conn_req_handler(struct iwcm_id_private *listen_id_priv, + struct iw_cm_event *iw_event) +{ + unsigned long flags; + struct iw_cm_id *cm_id; + struct iwcm_id_private *cm_id_priv; + int ret; + + /* + * The provider should never generate a connection request + * event with a bad status. + */ + BUG_ON(iw_event->status); + + cm_id = iw_create_cm_id(listen_id_priv->id.device, + listen_id_priv->id.cm_handler, + listen_id_priv->id.context); + /* If the cm_id could not be created, ignore the request */ + if (IS_ERR(cm_id)) + goto out; + + cm_id->provider_data = iw_event->provider_data; + cm_id->m_local_addr = iw_event->local_addr; + cm_id->m_remote_addr = iw_event->remote_addr; + cm_id->local_addr = listen_id_priv->id.local_addr; + + ret = iwpm_get_remote_info(&listen_id_priv->id.m_local_addr, + &iw_event->remote_addr, + &cm_id->remote_addr, + RDMA_NL_IWCM); + if (ret) { + cm_id->remote_addr = iw_event->remote_addr; + } else { + iw_cm_check_wildcard(&listen_id_priv->id.m_local_addr, + &iw_event->local_addr, + &cm_id->local_addr); + iw_event->local_addr = cm_id->local_addr; + iw_event->remote_addr = cm_id->remote_addr; + } + + cm_id_priv = container_of(cm_id, struct iwcm_id_private, id); + cm_id_priv->state = IW_CM_STATE_CONN_RECV; + + /* + * We could be destroying the listening id. If so, ignore this + * upcall. + */ + spin_lock_irqsave(&listen_id_priv->lock, flags); + if (listen_id_priv->state != IW_CM_STATE_LISTEN) { + spin_unlock_irqrestore(&listen_id_priv->lock, flags); + iw_cm_reject(cm_id, NULL, 0); + iw_destroy_cm_id(cm_id); + goto out; + } + spin_unlock_irqrestore(&listen_id_priv->lock, flags); + + ret = alloc_work_entries(cm_id_priv, 3); + if (ret) { + iw_cm_reject(cm_id, NULL, 0); + iw_destroy_cm_id(cm_id); + goto out; + } + + /* Call the client CM handler */ + ret = cm_id->cm_handler(cm_id, iw_event); + if (ret) { + iw_cm_reject(cm_id, NULL, 0); + iw_destroy_cm_id(cm_id); + } + +out: + if (iw_event->private_data_len) + kfree(iw_event->private_data); +} + +/* + * Passive Side: CM_ID <-- ESTABLISHED + * + * The provider generated an ESTABLISHED event which means that + * the MPA negotion has completed successfully and we are now in MPA + * FPDU mode. + * + * This event can only be received in the CONN_RECV state. If the + * remote peer closed, the ESTABLISHED event would be received followed + * by the CLOSE event. If the app closes, it will block until we wake + * it up after processing this event. + */ +static int cm_conn_est_handler(struct iwcm_id_private *cm_id_priv, + struct iw_cm_event *iw_event) +{ + unsigned long flags; + int ret; + + spin_lock_irqsave(&cm_id_priv->lock, flags); + + /* + * We clear the CONNECT_WAIT bit here to allow the callback + * function to call iw_cm_disconnect. Calling iw_destroy_cm_id + * from a callback handler is not allowed. + */ + clear_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags); + BUG_ON(cm_id_priv->state != IW_CM_STATE_CONN_RECV); + cm_id_priv->state = IW_CM_STATE_ESTABLISHED; + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + ret = cm_id_priv->id.cm_handler(&cm_id_priv->id, iw_event); + wake_up_all(&cm_id_priv->connect_wait); + + return ret; +} + +/* + * Active Side: CM_ID <-- ESTABLISHED + * + * The app has called connect and is waiting for the established event to + * post it's requests to the server. This event will wake up anyone + * blocked in iw_cm_disconnect or iw_destroy_id. + */ +static int cm_conn_rep_handler(struct iwcm_id_private *cm_id_priv, + struct iw_cm_event *iw_event) +{ + struct ib_qp *qp = NULL; + unsigned long flags; + int ret; + + spin_lock_irqsave(&cm_id_priv->lock, flags); + /* + * Clear the connect wait bit so a callback function calling + * iw_cm_disconnect will not wait and deadlock this thread + */ + clear_bit(IWCM_F_CONNECT_WAIT, &cm_id_priv->flags); + BUG_ON(cm_id_priv->state != IW_CM_STATE_CONN_SENT); + if (iw_event->status == 0) { + cm_id_priv->id.m_local_addr = iw_event->local_addr; + cm_id_priv->id.m_remote_addr = iw_event->remote_addr; + iw_event->local_addr = cm_id_priv->id.local_addr; + iw_event->remote_addr = cm_id_priv->id.remote_addr; + cm_id_priv->state = IW_CM_STATE_ESTABLISHED; + } else { + /* REJECTED or RESET */ + qp = cm_id_priv->qp; + cm_id_priv->qp = NULL; + cm_id_priv->state = IW_CM_STATE_IDLE; + } + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + if (qp) + cm_id_priv->id.device->ops.iw_rem_ref(qp); + ret = cm_id_priv->id.cm_handler(&cm_id_priv->id, iw_event); + + if (iw_event->private_data_len) + kfree(iw_event->private_data); + + /* Wake up waiters on connect complete */ + wake_up_all(&cm_id_priv->connect_wait); + + return ret; +} + +/* + * CM_ID <-- CLOSING + * + * If in the ESTABLISHED state, move to CLOSING. + */ +static void cm_disconnect_handler(struct iwcm_id_private *cm_id_priv, + struct iw_cm_event *iw_event) +{ + unsigned long flags; + + spin_lock_irqsave(&cm_id_priv->lock, flags); + if (cm_id_priv->state == IW_CM_STATE_ESTABLISHED) + cm_id_priv->state = IW_CM_STATE_CLOSING; + spin_unlock_irqrestore(&cm_id_priv->lock, flags); +} + +/* + * CM_ID <-- IDLE + * + * If in the ESTBLISHED or CLOSING states, the QP will have have been + * moved by the provider to the ERR state. Disassociate the CM_ID from + * the QP, move to IDLE, and remove the 'connected' reference. + * + * If in some other state, the cm_id was destroyed asynchronously. + * This is the last reference that will result in waking up + * the app thread blocked in iw_destroy_cm_id. + */ +static int cm_close_handler(struct iwcm_id_private *cm_id_priv, + struct iw_cm_event *iw_event) +{ + struct ib_qp *qp; + unsigned long flags; + int ret = 0, notify_event = 0; + spin_lock_irqsave(&cm_id_priv->lock, flags); + qp = cm_id_priv->qp; + cm_id_priv->qp = NULL; + + switch (cm_id_priv->state) { + case IW_CM_STATE_ESTABLISHED: + case IW_CM_STATE_CLOSING: + cm_id_priv->state = IW_CM_STATE_IDLE; + notify_event = 1; + break; + case IW_CM_STATE_DESTROYING: + break; + default: + BUG(); + } + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + + if (qp) + cm_id_priv->id.device->ops.iw_rem_ref(qp); + if (notify_event) + ret = cm_id_priv->id.cm_handler(&cm_id_priv->id, iw_event); + return ret; +} + +static int process_event(struct iwcm_id_private *cm_id_priv, + struct iw_cm_event *iw_event) +{ + int ret = 0; + + switch (iw_event->event) { + case IW_CM_EVENT_CONNECT_REQUEST: + cm_conn_req_handler(cm_id_priv, iw_event); + break; + case IW_CM_EVENT_CONNECT_REPLY: + ret = cm_conn_rep_handler(cm_id_priv, iw_event); + break; + case IW_CM_EVENT_ESTABLISHED: + ret = cm_conn_est_handler(cm_id_priv, iw_event); + break; + case IW_CM_EVENT_DISCONNECT: + cm_disconnect_handler(cm_id_priv, iw_event); + break; + case IW_CM_EVENT_CLOSE: + ret = cm_close_handler(cm_id_priv, iw_event); + break; + default: + BUG(); + } + + return ret; +} + +/* + * Process events on the work_list for the cm_id. If the callback + * function requests that the cm_id be deleted, a flag is set in the + * cm_id flags to indicate that when the last reference is + * removed, the cm_id is to be destroyed. This is necessary to + * distinguish between an object that will be destroyed by the app + * thread asleep on the destroy_comp list vs. an object destroyed + * here synchronously when the last reference is removed. + */ +static void cm_work_handler(struct work_struct *_work) +{ + struct iwcm_work *work = container_of(_work, struct iwcm_work, work); + struct iw_cm_event levent; + struct iwcm_id_private *cm_id_priv = work->cm_id; + unsigned long flags; + int empty; + int ret = 0; + + spin_lock_irqsave(&cm_id_priv->lock, flags); + empty = list_empty(&cm_id_priv->work_list); + while (!empty) { + work = list_entry(cm_id_priv->work_list.next, + struct iwcm_work, list); + list_del_init(&work->list); + empty = list_empty(&cm_id_priv->work_list); + levent = work->event; + put_work(work); + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + + if (!test_bit(IWCM_F_DROP_EVENTS, &cm_id_priv->flags)) { + ret = process_event(cm_id_priv, &levent); + if (ret) + destroy_cm_id(&cm_id_priv->id); + } else + pr_debug("dropping event %d\n", levent.event); + if (iwcm_deref_id(cm_id_priv)) + return; + if (empty) + return; + spin_lock_irqsave(&cm_id_priv->lock, flags); + } + spin_unlock_irqrestore(&cm_id_priv->lock, flags); +} + +/* + * This function is called on interrupt context. Schedule events on + * the iwcm_wq thread to allow callback functions to downcall into + * the CM and/or block. Events are queued to a per-CM_ID + * work_list. If this is the first event on the work_list, the work + * element is also queued on the iwcm_wq thread. + * + * Each event holds a reference on the cm_id. Until the last posted + * event has been delivered and processed, the cm_id cannot be + * deleted. + * + * Returns: + * 0 - the event was handled. + * -ENOMEM - the event was not handled due to lack of resources. + */ +static int cm_event_handler(struct iw_cm_id *cm_id, + struct iw_cm_event *iw_event) +{ + struct iwcm_work *work; + struct iwcm_id_private *cm_id_priv; + unsigned long flags; + int ret = 0; + + cm_id_priv = container_of(cm_id, struct iwcm_id_private, id); + + spin_lock_irqsave(&cm_id_priv->lock, flags); + work = get_work(cm_id_priv); + if (!work) { + ret = -ENOMEM; + goto out; + } + + INIT_WORK(&work->work, cm_work_handler); + work->cm_id = cm_id_priv; + work->event = *iw_event; + + if ((work->event.event == IW_CM_EVENT_CONNECT_REQUEST || + work->event.event == IW_CM_EVENT_CONNECT_REPLY) && + work->event.private_data_len) { + ret = copy_private_data(&work->event); + if (ret) { + put_work(work); + goto out; + } + } + + atomic_inc(&cm_id_priv->refcount); + if (list_empty(&cm_id_priv->work_list)) { + list_add_tail(&work->list, &cm_id_priv->work_list); + queue_work(iwcm_wq, &work->work); + } else + list_add_tail(&work->list, &cm_id_priv->work_list); +out: + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + return ret; +} + +static int iwcm_init_qp_init_attr(struct iwcm_id_private *cm_id_priv, + struct ib_qp_attr *qp_attr, + int *qp_attr_mask) +{ + unsigned long flags; + int ret; + + spin_lock_irqsave(&cm_id_priv->lock, flags); + switch (cm_id_priv->state) { + case IW_CM_STATE_IDLE: + case IW_CM_STATE_CONN_SENT: + case IW_CM_STATE_CONN_RECV: + case IW_CM_STATE_ESTABLISHED: + *qp_attr_mask = IB_QP_STATE | IB_QP_ACCESS_FLAGS; + qp_attr->qp_access_flags = IB_ACCESS_REMOTE_WRITE| + IB_ACCESS_REMOTE_READ; + ret = 0; + break; + default: + ret = -EINVAL; + break; + } + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + return ret; +} + +static int iwcm_init_qp_rts_attr(struct iwcm_id_private *cm_id_priv, + struct ib_qp_attr *qp_attr, + int *qp_attr_mask) +{ + unsigned long flags; + int ret; + + spin_lock_irqsave(&cm_id_priv->lock, flags); + switch (cm_id_priv->state) { + case IW_CM_STATE_IDLE: + case IW_CM_STATE_CONN_SENT: + case IW_CM_STATE_CONN_RECV: + case IW_CM_STATE_ESTABLISHED: + *qp_attr_mask = 0; + ret = 0; + break; + default: + ret = -EINVAL; + break; + } + spin_unlock_irqrestore(&cm_id_priv->lock, flags); + return ret; +} + +int iw_cm_init_qp_attr(struct iw_cm_id *cm_id, + struct ib_qp_attr *qp_attr, + int *qp_attr_mask) +{ + struct iwcm_id_private *cm_id_priv; + int ret; + + cm_id_priv = container_of(cm_id, struct iwcm_id_private, id); + switch (qp_attr->qp_state) { + case IB_QPS_INIT: + case IB_QPS_RTR: + ret = iwcm_init_qp_init_attr(cm_id_priv, + qp_attr, qp_attr_mask); + break; + case IB_QPS_RTS: + ret = iwcm_init_qp_rts_attr(cm_id_priv, + qp_attr, qp_attr_mask); + break; + default: + ret = -EINVAL; + break; + } + return ret; +} +EXPORT_SYMBOL(iw_cm_init_qp_attr); + +static int __init iw_cm_init(void) +{ + int ret; + + ret = iwpm_init(RDMA_NL_IWCM); + if (ret) + return ret; + + iwcm_wq = alloc_ordered_workqueue("iw_cm_wq", 0); + if (!iwcm_wq) + goto err_alloc; + + iwcm_ctl_table_hdr = register_net_sysctl(&init_net, "net/iw_cm", + iwcm_ctl_table); + if (!iwcm_ctl_table_hdr) { + pr_err("iw_cm: couldn't register sysctl paths\n"); + goto err_sysctl; + } + + rdma_nl_register(RDMA_NL_IWCM, iwcm_nl_cb_table); + return 0; + +err_sysctl: + destroy_workqueue(iwcm_wq); +err_alloc: + iwpm_exit(RDMA_NL_IWCM); + return -ENOMEM; +} + +static void __exit iw_cm_cleanup(void) +{ + rdma_nl_unregister(RDMA_NL_IWCM); + unregister_net_sysctl_table(iwcm_ctl_table_hdr); + destroy_workqueue(iwcm_wq); + iwpm_exit(RDMA_NL_IWCM); +} + +MODULE_ALIAS_RDMA_NETLINK(RDMA_NL_IWCM, 2); + +module_init(iw_cm_init); +module_exit(iw_cm_cleanup); diff --git a/drivers/infiniband/core/iwcm.h b/drivers/infiniband/core/iwcm.h new file mode 100644 index 000000000..82c2cd1b0 --- /dev/null +++ b/drivers/infiniband/core/iwcm.h @@ -0,0 +1,62 @@ +/* + * Copyright (c) 2005 Network Appliance, Inc. All rights reserved. + * Copyright (c) 2005 Open Grid Computing, Inc. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ +#ifndef IWCM_H +#define IWCM_H + +enum iw_cm_state { + IW_CM_STATE_IDLE, /* unbound, inactive */ + IW_CM_STATE_LISTEN, /* listen waiting for connect */ + IW_CM_STATE_CONN_RECV, /* inbound waiting for user accept */ + IW_CM_STATE_CONN_SENT, /* outbound waiting for peer accept */ + IW_CM_STATE_ESTABLISHED, /* established */ + IW_CM_STATE_CLOSING, /* disconnect */ + IW_CM_STATE_DESTROYING /* object being deleted */ +}; + +struct iwcm_id_private { + struct iw_cm_id id; + enum iw_cm_state state; + unsigned long flags; + struct ib_qp *qp; + struct completion destroy_comp; + wait_queue_head_t connect_wait; + struct list_head work_list; + spinlock_t lock; + atomic_t refcount; + struct list_head work_free_list; +}; + +#define IWCM_F_DROP_EVENTS 1 +#define IWCM_F_CONNECT_WAIT 2 + +#endif /* IWCM_H */ diff --git a/drivers/infiniband/core/iwpm_msg.c b/drivers/infiniband/core/iwpm_msg.c new file mode 100644 index 000000000..46686990a --- /dev/null +++ b/drivers/infiniband/core/iwpm_msg.c @@ -0,0 +1,877 @@ +/* + * Copyright (c) 2014 Intel Corporation. All rights reserved. + * Copyright (c) 2014 Chelsio, Inc. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include "iwpm_util.h" + +static const char iwpm_ulib_name[IWPM_ULIBNAME_SIZE] = "iWarpPortMapperUser"; +u16 iwpm_ulib_version = IWPM_UABI_VERSION_MIN; +static int iwpm_user_pid = IWPM_PID_UNDEFINED; +static atomic_t echo_nlmsg_seq; + +/** + * iwpm_valid_pid - Check if the userspace iwarp port mapper pid is valid + * + * Returns true if the pid is greater than zero, otherwise returns false + */ +int iwpm_valid_pid(void) +{ + return iwpm_user_pid > 0; +} + +/** + * iwpm_register_pid - Send a netlink query to userspace + * to get the iwarp port mapper pid + * @pm_msg: Contains driver info to send to the userspace port mapper + * @nl_client: The index of the netlink client + * + * nlmsg attributes: + * [IWPM_NLA_REG_PID_SEQ] + * [IWPM_NLA_REG_IF_NAME] + * [IWPM_NLA_REG_IBDEV_NAME] + * [IWPM_NLA_REG_ULIB_NAME] + */ +int iwpm_register_pid(struct iwpm_dev_data *pm_msg, u8 nl_client) +{ + struct sk_buff *skb = NULL; + struct iwpm_nlmsg_request *nlmsg_request = NULL; + struct nlmsghdr *nlh; + u32 msg_seq; + const char *err_str = ""; + int ret = -EINVAL; + + if (!iwpm_valid_client(nl_client)) { + err_str = "Invalid port mapper client"; + goto pid_query_error; + } + if (iwpm_check_registration(nl_client, IWPM_REG_VALID) || + iwpm_user_pid == IWPM_PID_UNAVAILABLE) + return 0; + skb = iwpm_create_nlmsg(RDMA_NL_IWPM_REG_PID, &nlh, nl_client); + if (!skb) { + err_str = "Unable to create a nlmsg"; + goto pid_query_error; + } + nlh->nlmsg_seq = iwpm_get_nlmsg_seq(); + nlmsg_request = iwpm_get_nlmsg_request(nlh->nlmsg_seq, nl_client, GFP_KERNEL); + if (!nlmsg_request) { + err_str = "Unable to allocate netlink request"; + goto pid_query_error; + } + msg_seq = atomic_read(&echo_nlmsg_seq); + + /* fill in the pid request message */ + err_str = "Unable to put attribute of the nlmsg"; + ret = ibnl_put_attr(skb, nlh, sizeof(u32), &msg_seq, IWPM_NLA_REG_PID_SEQ); + if (ret) + goto pid_query_error; + ret = ibnl_put_attr(skb, nlh, IFNAMSIZ, + pm_msg->if_name, IWPM_NLA_REG_IF_NAME); + if (ret) + goto pid_query_error; + ret = ibnl_put_attr(skb, nlh, IWPM_DEVNAME_SIZE, + pm_msg->dev_name, IWPM_NLA_REG_IBDEV_NAME); + if (ret) + goto pid_query_error; + ret = ibnl_put_attr(skb, nlh, IWPM_ULIBNAME_SIZE, + (char *)iwpm_ulib_name, IWPM_NLA_REG_ULIB_NAME); + if (ret) + goto pid_query_error; + + nlmsg_end(skb, nlh); + + pr_debug("%s: Multicasting a nlmsg (dev = %s ifname = %s iwpm = %s)\n", + __func__, pm_msg->dev_name, pm_msg->if_name, iwpm_ulib_name); + + ret = rdma_nl_multicast(&init_net, skb, RDMA_NL_GROUP_IWPM, GFP_KERNEL); + if (ret) { + skb = NULL; /* skb is freed in the netlink send-op handling */ + iwpm_user_pid = IWPM_PID_UNAVAILABLE; + err_str = "Unable to send a nlmsg"; + goto pid_query_error; + } + nlmsg_request->req_buffer = pm_msg; + ret = iwpm_wait_complete_req(nlmsg_request); + return ret; +pid_query_error: + pr_info("%s: %s (client = %d)\n", __func__, err_str, nl_client); + dev_kfree_skb(skb); + if (nlmsg_request) + iwpm_free_nlmsg_request(&nlmsg_request->kref); + return ret; +} + +/** + * iwpm_add_mapping - Send a netlink add mapping request to + * the userspace port mapper + * @pm_msg: Contains the local ip/tcp address info to send + * @nl_client: The index of the netlink client + * + * nlmsg attributes: + * [IWPM_NLA_MANAGE_MAPPING_SEQ] + * [IWPM_NLA_MANAGE_ADDR] + * [IWPM_NLA_MANAGE_FLAGS] + * + * If the request is successful, the pm_msg stores + * the port mapper response (mapped address info) + */ +int iwpm_add_mapping(struct iwpm_sa_data *pm_msg, u8 nl_client) +{ + struct sk_buff *skb = NULL; + struct iwpm_nlmsg_request *nlmsg_request = NULL; + struct nlmsghdr *nlh; + u32 msg_seq; + const char *err_str = ""; + int ret = -EINVAL; + + if (!iwpm_valid_client(nl_client)) { + err_str = "Invalid port mapper client"; + goto add_mapping_error; + } + if (!iwpm_valid_pid()) + return 0; + if (!iwpm_check_registration(nl_client, IWPM_REG_VALID)) { + err_str = "Unregistered port mapper client"; + goto add_mapping_error; + } + skb = iwpm_create_nlmsg(RDMA_NL_IWPM_ADD_MAPPING, &nlh, nl_client); + if (!skb) { + err_str = "Unable to create a nlmsg"; + goto add_mapping_error; + } + nlh->nlmsg_seq = iwpm_get_nlmsg_seq(); + nlmsg_request = iwpm_get_nlmsg_request(nlh->nlmsg_seq, nl_client, GFP_KERNEL); + if (!nlmsg_request) { + err_str = "Unable to allocate netlink request"; + goto add_mapping_error; + } + msg_seq = atomic_read(&echo_nlmsg_seq); + /* fill in the add mapping message */ + err_str = "Unable to put attribute of the nlmsg"; + ret = ibnl_put_attr(skb, nlh, sizeof(u32), &msg_seq, + IWPM_NLA_MANAGE_MAPPING_SEQ); + if (ret) + goto add_mapping_error; + ret = ibnl_put_attr(skb, nlh, sizeof(struct sockaddr_storage), + &pm_msg->loc_addr, IWPM_NLA_MANAGE_ADDR); + if (ret) + goto add_mapping_error; + + /* If flags are required and we're not V4, then return a quiet error */ + if (pm_msg->flags && iwpm_ulib_version == IWPM_UABI_VERSION_MIN) { + ret = -EINVAL; + goto add_mapping_error_nowarn; + } + if (iwpm_ulib_version > IWPM_UABI_VERSION_MIN) { + ret = ibnl_put_attr(skb, nlh, sizeof(u32), &pm_msg->flags, + IWPM_NLA_MANAGE_FLAGS); + if (ret) + goto add_mapping_error; + } + + nlmsg_end(skb, nlh); + nlmsg_request->req_buffer = pm_msg; + + ret = rdma_nl_unicast_wait(&init_net, skb, iwpm_user_pid); + if (ret) { + skb = NULL; /* skb is freed in the netlink send-op handling */ + iwpm_user_pid = IWPM_PID_UNDEFINED; + err_str = "Unable to send a nlmsg"; + goto add_mapping_error; + } + ret = iwpm_wait_complete_req(nlmsg_request); + return ret; +add_mapping_error: + pr_info("%s: %s (client = %d)\n", __func__, err_str, nl_client); +add_mapping_error_nowarn: + dev_kfree_skb(skb); + if (nlmsg_request) + iwpm_free_nlmsg_request(&nlmsg_request->kref); + return ret; +} + +/** + * iwpm_add_and_query_mapping - Process the port mapper response to + * iwpm_add_and_query_mapping request + * @pm_msg: Contains the local ip/tcp address info to send + * @nl_client: The index of the netlink client + * + * nlmsg attributes: + * [IWPM_NLA_QUERY_MAPPING_SEQ] + * [IWPM_NLA_QUERY_LOCAL_ADDR] + * [IWPM_NLA_QUERY_REMOTE_ADDR] + * [IWPM_NLA_QUERY_FLAGS] + */ +int iwpm_add_and_query_mapping(struct iwpm_sa_data *pm_msg, u8 nl_client) +{ + struct sk_buff *skb = NULL; + struct iwpm_nlmsg_request *nlmsg_request = NULL; + struct nlmsghdr *nlh; + u32 msg_seq; + const char *err_str = ""; + int ret = -EINVAL; + + if (!iwpm_valid_client(nl_client)) { + err_str = "Invalid port mapper client"; + goto query_mapping_error; + } + if (!iwpm_valid_pid()) + return 0; + if (!iwpm_check_registration(nl_client, IWPM_REG_VALID)) { + err_str = "Unregistered port mapper client"; + goto query_mapping_error; + } + ret = -ENOMEM; + skb = iwpm_create_nlmsg(RDMA_NL_IWPM_QUERY_MAPPING, &nlh, nl_client); + if (!skb) { + err_str = "Unable to create a nlmsg"; + goto query_mapping_error; + } + nlh->nlmsg_seq = iwpm_get_nlmsg_seq(); + nlmsg_request = iwpm_get_nlmsg_request(nlh->nlmsg_seq, + nl_client, GFP_KERNEL); + if (!nlmsg_request) { + err_str = "Unable to allocate netlink request"; + goto query_mapping_error; + } + msg_seq = atomic_read(&echo_nlmsg_seq); + + /* fill in the query message */ + err_str = "Unable to put attribute of the nlmsg"; + ret = ibnl_put_attr(skb, nlh, sizeof(u32), &msg_seq, + IWPM_NLA_QUERY_MAPPING_SEQ); + if (ret) + goto query_mapping_error; + ret = ibnl_put_attr(skb, nlh, sizeof(struct sockaddr_storage), + &pm_msg->loc_addr, IWPM_NLA_QUERY_LOCAL_ADDR); + if (ret) + goto query_mapping_error; + ret = ibnl_put_attr(skb, nlh, sizeof(struct sockaddr_storage), + &pm_msg->rem_addr, IWPM_NLA_QUERY_REMOTE_ADDR); + if (ret) + goto query_mapping_error; + + /* If flags are required and we're not V4, then return a quite error */ + if (pm_msg->flags && iwpm_ulib_version == IWPM_UABI_VERSION_MIN) { + ret = -EINVAL; + goto query_mapping_error_nowarn; + } + if (iwpm_ulib_version > IWPM_UABI_VERSION_MIN) { + ret = ibnl_put_attr(skb, nlh, sizeof(u32), &pm_msg->flags, + IWPM_NLA_QUERY_FLAGS); + if (ret) + goto query_mapping_error; + } + + nlmsg_end(skb, nlh); + nlmsg_request->req_buffer = pm_msg; + + ret = rdma_nl_unicast_wait(&init_net, skb, iwpm_user_pid); + if (ret) { + skb = NULL; /* skb is freed in the netlink send-op handling */ + err_str = "Unable to send a nlmsg"; + goto query_mapping_error; + } + ret = iwpm_wait_complete_req(nlmsg_request); + return ret; +query_mapping_error: + pr_info("%s: %s (client = %d)\n", __func__, err_str, nl_client); +query_mapping_error_nowarn: + dev_kfree_skb(skb); + if (nlmsg_request) + iwpm_free_nlmsg_request(&nlmsg_request->kref); + return ret; +} + +/** + * iwpm_remove_mapping - Send a netlink remove mapping request + * to the userspace port mapper + * + * @local_addr: Local ip/tcp address to remove + * @nl_client: The index of the netlink client + * + * nlmsg attributes: + * [IWPM_NLA_MANAGE_MAPPING_SEQ] + * [IWPM_NLA_MANAGE_ADDR] + */ +int iwpm_remove_mapping(struct sockaddr_storage *local_addr, u8 nl_client) +{ + struct sk_buff *skb = NULL; + struct nlmsghdr *nlh; + u32 msg_seq; + const char *err_str = ""; + int ret = -EINVAL; + + if (!iwpm_valid_client(nl_client)) { + err_str = "Invalid port mapper client"; + goto remove_mapping_error; + } + if (!iwpm_valid_pid()) + return 0; + if (iwpm_check_registration(nl_client, IWPM_REG_UNDEF)) { + err_str = "Unregistered port mapper client"; + goto remove_mapping_error; + } + skb = iwpm_create_nlmsg(RDMA_NL_IWPM_REMOVE_MAPPING, &nlh, nl_client); + if (!skb) { + ret = -ENOMEM; + err_str = "Unable to create a nlmsg"; + goto remove_mapping_error; + } + msg_seq = atomic_read(&echo_nlmsg_seq); + nlh->nlmsg_seq = iwpm_get_nlmsg_seq(); + err_str = "Unable to put attribute of the nlmsg"; + ret = ibnl_put_attr(skb, nlh, sizeof(u32), &msg_seq, + IWPM_NLA_MANAGE_MAPPING_SEQ); + if (ret) + goto remove_mapping_error; + ret = ibnl_put_attr(skb, nlh, sizeof(struct sockaddr_storage), + local_addr, IWPM_NLA_MANAGE_ADDR); + if (ret) + goto remove_mapping_error; + + nlmsg_end(skb, nlh); + + ret = rdma_nl_unicast_wait(&init_net, skb, iwpm_user_pid); + if (ret) { + skb = NULL; /* skb is freed in the netlink send-op handling */ + iwpm_user_pid = IWPM_PID_UNDEFINED; + err_str = "Unable to send a nlmsg"; + goto remove_mapping_error; + } + iwpm_print_sockaddr(local_addr, + "remove_mapping: Local sockaddr:"); + return 0; +remove_mapping_error: + pr_info("%s: %s (client = %d)\n", __func__, err_str, nl_client); + if (skb) + dev_kfree_skb_any(skb); + return ret; +} + +/* netlink attribute policy for the received response to register pid request */ +static const struct nla_policy resp_reg_policy[IWPM_NLA_RREG_PID_MAX] = { + [IWPM_NLA_RREG_PID_SEQ] = { .type = NLA_U32 }, + [IWPM_NLA_RREG_IBDEV_NAME] = { .type = NLA_STRING, + .len = IWPM_DEVNAME_SIZE - 1 }, + [IWPM_NLA_RREG_ULIB_NAME] = { .type = NLA_STRING, + .len = IWPM_ULIBNAME_SIZE - 1 }, + [IWPM_NLA_RREG_ULIB_VER] = { .type = NLA_U16 }, + [IWPM_NLA_RREG_PID_ERR] = { .type = NLA_U16 } +}; + +/** + * iwpm_register_pid_cb - Process the port mapper response to + * iwpm_register_pid query + * @skb: + * @cb: Contains the received message (payload and netlink header) + * + * If successful, the function receives the userspace port mapper pid + * which is used in future communication with the port mapper + */ +int iwpm_register_pid_cb(struct sk_buff *skb, struct netlink_callback *cb) +{ + struct iwpm_nlmsg_request *nlmsg_request = NULL; + struct nlattr *nltb[IWPM_NLA_RREG_PID_MAX]; + struct iwpm_dev_data *pm_msg; + char *dev_name, *iwpm_name; + u32 msg_seq; + u8 nl_client; + u16 iwpm_version; + const char *msg_type = "Register Pid response"; + + if (iwpm_parse_nlmsg(cb, IWPM_NLA_RREG_PID_MAX, + resp_reg_policy, nltb, msg_type)) + return -EINVAL; + + msg_seq = nla_get_u32(nltb[IWPM_NLA_RREG_PID_SEQ]); + nlmsg_request = iwpm_find_nlmsg_request(msg_seq); + if (!nlmsg_request) { + pr_info("%s: Could not find a matching request (seq = %u)\n", + __func__, msg_seq); + return -EINVAL; + } + pm_msg = nlmsg_request->req_buffer; + nl_client = nlmsg_request->nl_client; + dev_name = (char *)nla_data(nltb[IWPM_NLA_RREG_IBDEV_NAME]); + iwpm_name = (char *)nla_data(nltb[IWPM_NLA_RREG_ULIB_NAME]); + iwpm_version = nla_get_u16(nltb[IWPM_NLA_RREG_ULIB_VER]); + + /* check device name, ulib name and version */ + if (strcmp(pm_msg->dev_name, dev_name) || + strcmp(iwpm_ulib_name, iwpm_name) || + iwpm_version < IWPM_UABI_VERSION_MIN) { + + pr_info("%s: Incorrect info (dev = %s name = %s version = %d)\n", + __func__, dev_name, iwpm_name, iwpm_version); + nlmsg_request->err_code = IWPM_USER_LIB_INFO_ERR; + goto register_pid_response_exit; + } + iwpm_user_pid = cb->nlh->nlmsg_pid; + iwpm_ulib_version = iwpm_version; + if (iwpm_ulib_version < IWPM_UABI_VERSION) + pr_warn_once("%s: Down level iwpmd/pid %u. Continuing...", + __func__, iwpm_user_pid); + atomic_set(&echo_nlmsg_seq, cb->nlh->nlmsg_seq); + pr_debug("%s: iWarp Port Mapper (pid = %d) is available!\n", + __func__, iwpm_user_pid); + if (iwpm_valid_client(nl_client)) + iwpm_set_registration(nl_client, IWPM_REG_VALID); +register_pid_response_exit: + nlmsg_request->request_done = 1; + /* always for found nlmsg_request */ + kref_put(&nlmsg_request->kref, iwpm_free_nlmsg_request); + barrier(); + up(&nlmsg_request->sem); + return 0; +} + +/* netlink attribute policy for the received response to add mapping request */ +static const struct nla_policy resp_add_policy[IWPM_NLA_RMANAGE_MAPPING_MAX] = { + [IWPM_NLA_RMANAGE_MAPPING_SEQ] = { .type = NLA_U32 }, + [IWPM_NLA_RMANAGE_ADDR] = { + .len = sizeof(struct sockaddr_storage) }, + [IWPM_NLA_RMANAGE_MAPPED_LOC_ADDR] = { + .len = sizeof(struct sockaddr_storage) }, + [IWPM_NLA_RMANAGE_MAPPING_ERR] = { .type = NLA_U16 } +}; + +/** + * iwpm_add_mapping_cb - Process the port mapper response to + * iwpm_add_mapping request + * @skb: + * @cb: Contains the received message (payload and netlink header) + */ +int iwpm_add_mapping_cb(struct sk_buff *skb, struct netlink_callback *cb) +{ + struct iwpm_sa_data *pm_msg; + struct iwpm_nlmsg_request *nlmsg_request = NULL; + struct nlattr *nltb[IWPM_NLA_RMANAGE_MAPPING_MAX]; + struct sockaddr_storage *local_sockaddr; + struct sockaddr_storage *mapped_sockaddr; + const char *msg_type; + u32 msg_seq; + + msg_type = "Add Mapping response"; + if (iwpm_parse_nlmsg(cb, IWPM_NLA_RMANAGE_MAPPING_MAX, + resp_add_policy, nltb, msg_type)) + return -EINVAL; + + atomic_set(&echo_nlmsg_seq, cb->nlh->nlmsg_seq); + + msg_seq = nla_get_u32(nltb[IWPM_NLA_RMANAGE_MAPPING_SEQ]); + nlmsg_request = iwpm_find_nlmsg_request(msg_seq); + if (!nlmsg_request) { + pr_info("%s: Could not find a matching request (seq = %u)\n", + __func__, msg_seq); + return -EINVAL; + } + pm_msg = nlmsg_request->req_buffer; + local_sockaddr = (struct sockaddr_storage *) + nla_data(nltb[IWPM_NLA_RMANAGE_ADDR]); + mapped_sockaddr = (struct sockaddr_storage *) + nla_data(nltb[IWPM_NLA_RMANAGE_MAPPED_LOC_ADDR]); + + if (iwpm_compare_sockaddr(local_sockaddr, &pm_msg->loc_addr)) { + nlmsg_request->err_code = IWPM_USER_LIB_INFO_ERR; + goto add_mapping_response_exit; + } + if (mapped_sockaddr->ss_family != local_sockaddr->ss_family) { + pr_info("%s: Sockaddr family doesn't match the requested one\n", + __func__); + nlmsg_request->err_code = IWPM_USER_LIB_INFO_ERR; + goto add_mapping_response_exit; + } + memcpy(&pm_msg->mapped_loc_addr, mapped_sockaddr, + sizeof(*mapped_sockaddr)); + iwpm_print_sockaddr(&pm_msg->loc_addr, + "add_mapping: Local sockaddr:"); + iwpm_print_sockaddr(&pm_msg->mapped_loc_addr, + "add_mapping: Mapped local sockaddr:"); + +add_mapping_response_exit: + nlmsg_request->request_done = 1; + /* always for found request */ + kref_put(&nlmsg_request->kref, iwpm_free_nlmsg_request); + barrier(); + up(&nlmsg_request->sem); + return 0; +} + +/* netlink attribute policy for the response to add and query mapping request + * and response with remote address info */ +static const struct nla_policy resp_query_policy[IWPM_NLA_RQUERY_MAPPING_MAX] = { + [IWPM_NLA_RQUERY_MAPPING_SEQ] = { .type = NLA_U32 }, + [IWPM_NLA_RQUERY_LOCAL_ADDR] = { + .len = sizeof(struct sockaddr_storage) }, + [IWPM_NLA_RQUERY_REMOTE_ADDR] = { + .len = sizeof(struct sockaddr_storage) }, + [IWPM_NLA_RQUERY_MAPPED_LOC_ADDR] = { + .len = sizeof(struct sockaddr_storage) }, + [IWPM_NLA_RQUERY_MAPPED_REM_ADDR] = { + .len = sizeof(struct sockaddr_storage) }, + [IWPM_NLA_RQUERY_MAPPING_ERR] = { .type = NLA_U16 } +}; + +/** + * iwpm_add_and_query_mapping_cb - Process the port mapper response to + * iwpm_add_and_query_mapping request + * @skb: + * @cb: Contains the received message (payload and netlink header) + */ +int iwpm_add_and_query_mapping_cb(struct sk_buff *skb, + struct netlink_callback *cb) +{ + struct iwpm_sa_data *pm_msg; + struct iwpm_nlmsg_request *nlmsg_request = NULL; + struct nlattr *nltb[IWPM_NLA_RQUERY_MAPPING_MAX]; + struct sockaddr_storage *local_sockaddr, *remote_sockaddr; + struct sockaddr_storage *mapped_loc_sockaddr, *mapped_rem_sockaddr; + const char *msg_type; + u32 msg_seq; + u16 err_code; + + msg_type = "Query Mapping response"; + if (iwpm_parse_nlmsg(cb, IWPM_NLA_RQUERY_MAPPING_MAX, + resp_query_policy, nltb, msg_type)) + return -EINVAL; + atomic_set(&echo_nlmsg_seq, cb->nlh->nlmsg_seq); + + msg_seq = nla_get_u32(nltb[IWPM_NLA_RQUERY_MAPPING_SEQ]); + nlmsg_request = iwpm_find_nlmsg_request(msg_seq); + if (!nlmsg_request) { + pr_info("%s: Could not find a matching request (seq = %u)\n", + __func__, msg_seq); + return -EINVAL; + } + pm_msg = nlmsg_request->req_buffer; + local_sockaddr = (struct sockaddr_storage *) + nla_data(nltb[IWPM_NLA_RQUERY_LOCAL_ADDR]); + remote_sockaddr = (struct sockaddr_storage *) + nla_data(nltb[IWPM_NLA_RQUERY_REMOTE_ADDR]); + mapped_loc_sockaddr = (struct sockaddr_storage *) + nla_data(nltb[IWPM_NLA_RQUERY_MAPPED_LOC_ADDR]); + mapped_rem_sockaddr = (struct sockaddr_storage *) + nla_data(nltb[IWPM_NLA_RQUERY_MAPPED_REM_ADDR]); + + err_code = nla_get_u16(nltb[IWPM_NLA_RQUERY_MAPPING_ERR]); + if (err_code == IWPM_REMOTE_QUERY_REJECT) { + pr_info("%s: Received a Reject (pid = %u, echo seq = %u)\n", + __func__, cb->nlh->nlmsg_pid, msg_seq); + nlmsg_request->err_code = IWPM_REMOTE_QUERY_REJECT; + } + if (iwpm_compare_sockaddr(local_sockaddr, &pm_msg->loc_addr) || + iwpm_compare_sockaddr(remote_sockaddr, &pm_msg->rem_addr)) { + pr_info("%s: Incorrect local sockaddr\n", __func__); + nlmsg_request->err_code = IWPM_USER_LIB_INFO_ERR; + goto query_mapping_response_exit; + } + if (mapped_loc_sockaddr->ss_family != local_sockaddr->ss_family || + mapped_rem_sockaddr->ss_family != remote_sockaddr->ss_family) { + pr_info("%s: Sockaddr family doesn't match the requested one\n", + __func__); + nlmsg_request->err_code = IWPM_USER_LIB_INFO_ERR; + goto query_mapping_response_exit; + } + memcpy(&pm_msg->mapped_loc_addr, mapped_loc_sockaddr, + sizeof(*mapped_loc_sockaddr)); + memcpy(&pm_msg->mapped_rem_addr, mapped_rem_sockaddr, + sizeof(*mapped_rem_sockaddr)); + + iwpm_print_sockaddr(&pm_msg->loc_addr, + "query_mapping: Local sockaddr:"); + iwpm_print_sockaddr(&pm_msg->mapped_loc_addr, + "query_mapping: Mapped local sockaddr:"); + iwpm_print_sockaddr(&pm_msg->rem_addr, + "query_mapping: Remote sockaddr:"); + iwpm_print_sockaddr(&pm_msg->mapped_rem_addr, + "query_mapping: Mapped remote sockaddr:"); +query_mapping_response_exit: + nlmsg_request->request_done = 1; + /* always for found request */ + kref_put(&nlmsg_request->kref, iwpm_free_nlmsg_request); + barrier(); + up(&nlmsg_request->sem); + return 0; +} + +/** + * iwpm_remote_info_cb - Process remote connecting peer address info, which + * the port mapper has received from the connecting peer + * @skb: + * @cb: Contains the received message (payload and netlink header) + * + * Stores the IPv4/IPv6 address info in a hash table + */ +int iwpm_remote_info_cb(struct sk_buff *skb, struct netlink_callback *cb) +{ + struct nlattr *nltb[IWPM_NLA_RQUERY_MAPPING_MAX]; + struct sockaddr_storage *local_sockaddr, *remote_sockaddr; + struct sockaddr_storage *mapped_loc_sockaddr, *mapped_rem_sockaddr; + struct iwpm_remote_info *rem_info; + const char *msg_type; + u8 nl_client; + int ret = -EINVAL; + + msg_type = "Remote Mapping info"; + if (iwpm_parse_nlmsg(cb, IWPM_NLA_RQUERY_MAPPING_MAX, + resp_query_policy, nltb, msg_type)) + return ret; + + nl_client = RDMA_NL_GET_CLIENT(cb->nlh->nlmsg_type); + if (!iwpm_valid_client(nl_client)) { + pr_info("%s: Invalid port mapper client = %d\n", + __func__, nl_client); + return ret; + } + atomic_set(&echo_nlmsg_seq, cb->nlh->nlmsg_seq); + + local_sockaddr = (struct sockaddr_storage *) + nla_data(nltb[IWPM_NLA_RQUERY_LOCAL_ADDR]); + remote_sockaddr = (struct sockaddr_storage *) + nla_data(nltb[IWPM_NLA_RQUERY_REMOTE_ADDR]); + mapped_loc_sockaddr = (struct sockaddr_storage *) + nla_data(nltb[IWPM_NLA_RQUERY_MAPPED_LOC_ADDR]); + mapped_rem_sockaddr = (struct sockaddr_storage *) + nla_data(nltb[IWPM_NLA_RQUERY_MAPPED_REM_ADDR]); + + if (mapped_loc_sockaddr->ss_family != local_sockaddr->ss_family || + mapped_rem_sockaddr->ss_family != remote_sockaddr->ss_family) { + pr_info("%s: Sockaddr family doesn't match the requested one\n", + __func__); + return ret; + } + rem_info = kzalloc(sizeof(struct iwpm_remote_info), GFP_ATOMIC); + if (!rem_info) { + ret = -ENOMEM; + return ret; + } + memcpy(&rem_info->mapped_loc_sockaddr, mapped_loc_sockaddr, + sizeof(struct sockaddr_storage)); + memcpy(&rem_info->remote_sockaddr, remote_sockaddr, + sizeof(struct sockaddr_storage)); + memcpy(&rem_info->mapped_rem_sockaddr, mapped_rem_sockaddr, + sizeof(struct sockaddr_storage)); + rem_info->nl_client = nl_client; + + iwpm_add_remote_info(rem_info); + + iwpm_print_sockaddr(local_sockaddr, + "remote_info: Local sockaddr:"); + iwpm_print_sockaddr(mapped_loc_sockaddr, + "remote_info: Mapped local sockaddr:"); + iwpm_print_sockaddr(remote_sockaddr, + "remote_info: Remote sockaddr:"); + iwpm_print_sockaddr(mapped_rem_sockaddr, + "remote_info: Mapped remote sockaddr:"); + return ret; +} + +/* netlink attribute policy for the received request for mapping info */ +static const struct nla_policy resp_mapinfo_policy[IWPM_NLA_MAPINFO_REQ_MAX] = { + [IWPM_NLA_MAPINFO_ULIB_NAME] = { .type = NLA_STRING, + .len = IWPM_ULIBNAME_SIZE - 1 }, + [IWPM_NLA_MAPINFO_ULIB_VER] = { .type = NLA_U16 } +}; + +/** + * iwpm_mapping_info_cb - Process a notification that the userspace + * port mapper daemon is started + * @skb: + * @cb: Contains the received message (payload and netlink header) + * + * Using the received port mapper pid, send all the local mapping + * info records to the userspace port mapper + */ +int iwpm_mapping_info_cb(struct sk_buff *skb, struct netlink_callback *cb) +{ + struct nlattr *nltb[IWPM_NLA_MAPINFO_REQ_MAX]; + const char *msg_type = "Mapping Info response"; + u8 nl_client; + char *iwpm_name; + u16 iwpm_version; + int ret = -EINVAL; + + if (iwpm_parse_nlmsg(cb, IWPM_NLA_MAPINFO_REQ_MAX, + resp_mapinfo_policy, nltb, msg_type)) { + pr_info("%s: Unable to parse nlmsg\n", __func__); + return ret; + } + iwpm_name = (char *)nla_data(nltb[IWPM_NLA_MAPINFO_ULIB_NAME]); + iwpm_version = nla_get_u16(nltb[IWPM_NLA_MAPINFO_ULIB_VER]); + if (strcmp(iwpm_ulib_name, iwpm_name) || + iwpm_version < IWPM_UABI_VERSION_MIN) { + pr_info("%s: Invalid port mapper name = %s version = %d\n", + __func__, iwpm_name, iwpm_version); + return ret; + } + nl_client = RDMA_NL_GET_CLIENT(cb->nlh->nlmsg_type); + if (!iwpm_valid_client(nl_client)) { + pr_info("%s: Invalid port mapper client = %d\n", + __func__, nl_client); + return ret; + } + iwpm_set_registration(nl_client, IWPM_REG_INCOMPL); + atomic_set(&echo_nlmsg_seq, cb->nlh->nlmsg_seq); + iwpm_user_pid = cb->nlh->nlmsg_pid; + + if (iwpm_ulib_version < IWPM_UABI_VERSION) + pr_warn_once("%s: Down level iwpmd/pid %u. Continuing...", + __func__, iwpm_user_pid); + + if (!iwpm_mapinfo_available()) + return 0; + pr_debug("%s: iWarp Port Mapper (pid = %d) is available!\n", + __func__, iwpm_user_pid); + ret = iwpm_send_mapinfo(nl_client, iwpm_user_pid); + return ret; +} + +/* netlink attribute policy for the received mapping info ack */ +static const struct nla_policy ack_mapinfo_policy[IWPM_NLA_MAPINFO_NUM_MAX] = { + [IWPM_NLA_MAPINFO_SEQ] = { .type = NLA_U32 }, + [IWPM_NLA_MAPINFO_SEND_NUM] = { .type = NLA_U32 }, + [IWPM_NLA_MAPINFO_ACK_NUM] = { .type = NLA_U32 } +}; + +/** + * iwpm_ack_mapping_info_cb - Process the port mapper ack for + * the provided local mapping info records + * @skb: + * @cb: Contains the received message (payload and netlink header) + */ +int iwpm_ack_mapping_info_cb(struct sk_buff *skb, struct netlink_callback *cb) +{ + struct nlattr *nltb[IWPM_NLA_MAPINFO_NUM_MAX]; + u32 mapinfo_send, mapinfo_ack; + const char *msg_type = "Mapping Info Ack"; + + if (iwpm_parse_nlmsg(cb, IWPM_NLA_MAPINFO_NUM_MAX, + ack_mapinfo_policy, nltb, msg_type)) + return -EINVAL; + mapinfo_send = nla_get_u32(nltb[IWPM_NLA_MAPINFO_SEND_NUM]); + mapinfo_ack = nla_get_u32(nltb[IWPM_NLA_MAPINFO_ACK_NUM]); + if (mapinfo_ack != mapinfo_send) + pr_info("%s: Invalid mapinfo number (sent = %u ack-ed = %u)\n", + __func__, mapinfo_send, mapinfo_ack); + atomic_set(&echo_nlmsg_seq, cb->nlh->nlmsg_seq); + return 0; +} + +/* netlink attribute policy for the received port mapper error message */ +static const struct nla_policy map_error_policy[IWPM_NLA_ERR_MAX] = { + [IWPM_NLA_ERR_SEQ] = { .type = NLA_U32 }, + [IWPM_NLA_ERR_CODE] = { .type = NLA_U16 }, +}; + +/** + * iwpm_mapping_error_cb - Process port mapper notification for error + * + * @skb: + * @cb: Contains the received message (payload and netlink header) + */ +int iwpm_mapping_error_cb(struct sk_buff *skb, struct netlink_callback *cb) +{ + struct iwpm_nlmsg_request *nlmsg_request = NULL; + int nl_client = RDMA_NL_GET_CLIENT(cb->nlh->nlmsg_type); + struct nlattr *nltb[IWPM_NLA_ERR_MAX]; + u32 msg_seq; + u16 err_code; + const char *msg_type = "Mapping Error Msg"; + + if (iwpm_parse_nlmsg(cb, IWPM_NLA_ERR_MAX, + map_error_policy, nltb, msg_type)) + return -EINVAL; + + msg_seq = nla_get_u32(nltb[IWPM_NLA_ERR_SEQ]); + err_code = nla_get_u16(nltb[IWPM_NLA_ERR_CODE]); + pr_info("%s: Received msg seq = %u err code = %u client = %d\n", + __func__, msg_seq, err_code, nl_client); + /* look for nlmsg_request */ + nlmsg_request = iwpm_find_nlmsg_request(msg_seq); + if (!nlmsg_request) { + /* not all errors have associated requests */ + pr_debug("Could not find matching req (seq = %u)\n", msg_seq); + return 0; + } + atomic_set(&echo_nlmsg_seq, cb->nlh->nlmsg_seq); + nlmsg_request->err_code = err_code; + nlmsg_request->request_done = 1; + /* always for found request */ + kref_put(&nlmsg_request->kref, iwpm_free_nlmsg_request); + barrier(); + up(&nlmsg_request->sem); + return 0; +} + +/* netlink attribute policy for the received hello request */ +static const struct nla_policy hello_policy[IWPM_NLA_HELLO_MAX] = { + [IWPM_NLA_HELLO_ABI_VERSION] = { .type = NLA_U16 } +}; + +/** + * iwpm_hello_cb - Process a hello message from iwpmd + * + * @skb: + * @cb: Contains the received message (payload and netlink header) + * + * Using the received port mapper pid, send the kernel's abi_version + * after adjusting it to support the iwpmd version. + */ +int iwpm_hello_cb(struct sk_buff *skb, struct netlink_callback *cb) +{ + struct nlattr *nltb[IWPM_NLA_HELLO_MAX]; + const char *msg_type = "Hello request"; + u8 nl_client; + u16 abi_version; + int ret = -EINVAL; + + if (iwpm_parse_nlmsg(cb, IWPM_NLA_HELLO_MAX, hello_policy, nltb, + msg_type)) { + pr_info("%s: Unable to parse nlmsg\n", __func__); + return ret; + } + abi_version = nla_get_u16(nltb[IWPM_NLA_HELLO_ABI_VERSION]); + nl_client = RDMA_NL_GET_CLIENT(cb->nlh->nlmsg_type); + if (!iwpm_valid_client(nl_client)) { + pr_info("%s: Invalid port mapper client = %d\n", + __func__, nl_client); + return ret; + } + iwpm_set_registration(nl_client, IWPM_REG_INCOMPL); + atomic_set(&echo_nlmsg_seq, cb->nlh->nlmsg_seq); + iwpm_ulib_version = min_t(u16, IWPM_UABI_VERSION, abi_version); + pr_debug("Using ABI version %u\n", iwpm_ulib_version); + iwpm_user_pid = cb->nlh->nlmsg_pid; + ret = iwpm_send_hello(nl_client, iwpm_user_pid, iwpm_ulib_version); + return ret; +} diff --git a/drivers/infiniband/core/iwpm_util.c b/drivers/infiniband/core/iwpm_util.c new file mode 100644 index 000000000..13495b43d --- /dev/null +++ b/drivers/infiniband/core/iwpm_util.c @@ -0,0 +1,837 @@ +/* + * Copyright (c) 2014 Chelsio, Inc. All rights reserved. + * Copyright (c) 2014 Intel Corporation. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include "iwpm_util.h" + +#define IWPM_MAPINFO_HASH_SIZE 512 +#define IWPM_MAPINFO_HASH_MASK (IWPM_MAPINFO_HASH_SIZE - 1) +#define IWPM_REMINFO_HASH_SIZE 64 +#define IWPM_REMINFO_HASH_MASK (IWPM_REMINFO_HASH_SIZE - 1) +#define IWPM_MSG_SIZE 512 + +static LIST_HEAD(iwpm_nlmsg_req_list); +static DEFINE_SPINLOCK(iwpm_nlmsg_req_lock); + +static struct hlist_head *iwpm_hash_bucket; +static DEFINE_SPINLOCK(iwpm_mapinfo_lock); + +static struct hlist_head *iwpm_reminfo_bucket; +static DEFINE_SPINLOCK(iwpm_reminfo_lock); + +static DEFINE_MUTEX(iwpm_admin_lock); +static struct iwpm_admin_data iwpm_admin; + +/** + * iwpm_init - Allocate resources for the iwarp port mapper + * @nl_client: The index of the netlink client + * + * Should be called when network interface goes up. + */ +int iwpm_init(u8 nl_client) +{ + int ret = 0; + mutex_lock(&iwpm_admin_lock); + if (atomic_read(&iwpm_admin.refcount) == 0) { + iwpm_hash_bucket = kcalloc(IWPM_MAPINFO_HASH_SIZE, + sizeof(struct hlist_head), + GFP_KERNEL); + if (!iwpm_hash_bucket) { + ret = -ENOMEM; + goto init_exit; + } + iwpm_reminfo_bucket = kcalloc(IWPM_REMINFO_HASH_SIZE, + sizeof(struct hlist_head), + GFP_KERNEL); + if (!iwpm_reminfo_bucket) { + kfree(iwpm_hash_bucket); + ret = -ENOMEM; + goto init_exit; + } + } + atomic_inc(&iwpm_admin.refcount); +init_exit: + mutex_unlock(&iwpm_admin_lock); + if (!ret) { + iwpm_set_valid(nl_client, 1); + iwpm_set_registration(nl_client, IWPM_REG_UNDEF); + pr_debug("%s: Mapinfo and reminfo tables are created\n", + __func__); + } + return ret; +} + +static void free_hash_bucket(void); +static void free_reminfo_bucket(void); + +/** + * iwpm_exit - Deallocate resources for the iwarp port mapper + * @nl_client: The index of the netlink client + * + * Should be called when network interface goes down. + */ +int iwpm_exit(u8 nl_client) +{ + + if (!iwpm_valid_client(nl_client)) + return -EINVAL; + mutex_lock(&iwpm_admin_lock); + if (atomic_read(&iwpm_admin.refcount) == 0) { + mutex_unlock(&iwpm_admin_lock); + pr_err("%s Incorrect usage - negative refcount\n", __func__); + return -EINVAL; + } + if (atomic_dec_and_test(&iwpm_admin.refcount)) { + free_hash_bucket(); + free_reminfo_bucket(); + pr_debug("%s: Resources are destroyed\n", __func__); + } + mutex_unlock(&iwpm_admin_lock); + iwpm_set_valid(nl_client, 0); + iwpm_set_registration(nl_client, IWPM_REG_UNDEF); + return 0; +} + +static struct hlist_head *get_mapinfo_hash_bucket(struct sockaddr_storage *, + struct sockaddr_storage *); + +/** + * iwpm_create_mapinfo - Store local and mapped IPv4/IPv6 address + * info in a hash table + * @local_addr: Local ip/tcp address + * @mapped_addr: Mapped local ip/tcp address + * @nl_client: The index of the netlink client + * @map_flags: IWPM mapping flags + */ +int iwpm_create_mapinfo(struct sockaddr_storage *local_sockaddr, + struct sockaddr_storage *mapped_sockaddr, + u8 nl_client, u32 map_flags) +{ + struct hlist_head *hash_bucket_head = NULL; + struct iwpm_mapping_info *map_info; + unsigned long flags; + int ret = -EINVAL; + + if (!iwpm_valid_client(nl_client)) + return ret; + map_info = kzalloc(sizeof(struct iwpm_mapping_info), GFP_KERNEL); + if (!map_info) + return -ENOMEM; + + memcpy(&map_info->local_sockaddr, local_sockaddr, + sizeof(struct sockaddr_storage)); + memcpy(&map_info->mapped_sockaddr, mapped_sockaddr, + sizeof(struct sockaddr_storage)); + map_info->nl_client = nl_client; + map_info->map_flags = map_flags; + + spin_lock_irqsave(&iwpm_mapinfo_lock, flags); + if (iwpm_hash_bucket) { + hash_bucket_head = get_mapinfo_hash_bucket( + &map_info->local_sockaddr, + &map_info->mapped_sockaddr); + if (hash_bucket_head) { + hlist_add_head(&map_info->hlist_node, hash_bucket_head); + ret = 0; + } + } + spin_unlock_irqrestore(&iwpm_mapinfo_lock, flags); + + if (!hash_bucket_head) + kfree(map_info); + return ret; +} + +/** + * iwpm_remove_mapinfo - Remove local and mapped IPv4/IPv6 address + * info from the hash table + * @local_addr: Local ip/tcp address + * @mapped_local_addr: Mapped local ip/tcp address + * + * Returns err code if mapping info is not found in the hash table, + * otherwise returns 0 + */ +int iwpm_remove_mapinfo(struct sockaddr_storage *local_sockaddr, + struct sockaddr_storage *mapped_local_addr) +{ + struct hlist_node *tmp_hlist_node; + struct hlist_head *hash_bucket_head; + struct iwpm_mapping_info *map_info = NULL; + unsigned long flags; + int ret = -EINVAL; + + spin_lock_irqsave(&iwpm_mapinfo_lock, flags); + if (iwpm_hash_bucket) { + hash_bucket_head = get_mapinfo_hash_bucket( + local_sockaddr, + mapped_local_addr); + if (!hash_bucket_head) + goto remove_mapinfo_exit; + + hlist_for_each_entry_safe(map_info, tmp_hlist_node, + hash_bucket_head, hlist_node) { + + if (!iwpm_compare_sockaddr(&map_info->mapped_sockaddr, + mapped_local_addr)) { + + hlist_del_init(&map_info->hlist_node); + kfree(map_info); + ret = 0; + break; + } + } + } +remove_mapinfo_exit: + spin_unlock_irqrestore(&iwpm_mapinfo_lock, flags); + return ret; +} + +static void free_hash_bucket(void) +{ + struct hlist_node *tmp_hlist_node; + struct iwpm_mapping_info *map_info; + unsigned long flags; + int i; + + /* remove all the mapinfo data from the list */ + spin_lock_irqsave(&iwpm_mapinfo_lock, flags); + for (i = 0; i < IWPM_MAPINFO_HASH_SIZE; i++) { + hlist_for_each_entry_safe(map_info, tmp_hlist_node, + &iwpm_hash_bucket[i], hlist_node) { + + hlist_del_init(&map_info->hlist_node); + kfree(map_info); + } + } + /* free the hash list */ + kfree(iwpm_hash_bucket); + iwpm_hash_bucket = NULL; + spin_unlock_irqrestore(&iwpm_mapinfo_lock, flags); +} + +static void free_reminfo_bucket(void) +{ + struct hlist_node *tmp_hlist_node; + struct iwpm_remote_info *rem_info; + unsigned long flags; + int i; + + /* remove all the remote info from the list */ + spin_lock_irqsave(&iwpm_reminfo_lock, flags); + for (i = 0; i < IWPM_REMINFO_HASH_SIZE; i++) { + hlist_for_each_entry_safe(rem_info, tmp_hlist_node, + &iwpm_reminfo_bucket[i], hlist_node) { + + hlist_del_init(&rem_info->hlist_node); + kfree(rem_info); + } + } + /* free the hash list */ + kfree(iwpm_reminfo_bucket); + iwpm_reminfo_bucket = NULL; + spin_unlock_irqrestore(&iwpm_reminfo_lock, flags); +} + +static struct hlist_head *get_reminfo_hash_bucket(struct sockaddr_storage *, + struct sockaddr_storage *); + +void iwpm_add_remote_info(struct iwpm_remote_info *rem_info) +{ + struct hlist_head *hash_bucket_head; + unsigned long flags; + + spin_lock_irqsave(&iwpm_reminfo_lock, flags); + if (iwpm_reminfo_bucket) { + hash_bucket_head = get_reminfo_hash_bucket( + &rem_info->mapped_loc_sockaddr, + &rem_info->mapped_rem_sockaddr); + if (hash_bucket_head) + hlist_add_head(&rem_info->hlist_node, hash_bucket_head); + } + spin_unlock_irqrestore(&iwpm_reminfo_lock, flags); +} + +/** + * iwpm_get_remote_info - Get the remote connecting peer address info + * + * @mapped_loc_addr: Mapped local address of the listening peer + * @mapped_rem_addr: Mapped remote address of the connecting peer + * @remote_addr: To store the remote address of the connecting peer + * @nl_client: The index of the netlink client + * + * The remote address info is retrieved and provided to the client in + * the remote_addr. After that it is removed from the hash table + */ +int iwpm_get_remote_info(struct sockaddr_storage *mapped_loc_addr, + struct sockaddr_storage *mapped_rem_addr, + struct sockaddr_storage *remote_addr, + u8 nl_client) +{ + struct hlist_node *tmp_hlist_node; + struct hlist_head *hash_bucket_head; + struct iwpm_remote_info *rem_info = NULL; + unsigned long flags; + int ret = -EINVAL; + + if (!iwpm_valid_client(nl_client)) { + pr_info("%s: Invalid client = %d\n", __func__, nl_client); + return ret; + } + spin_lock_irqsave(&iwpm_reminfo_lock, flags); + if (iwpm_reminfo_bucket) { + hash_bucket_head = get_reminfo_hash_bucket( + mapped_loc_addr, + mapped_rem_addr); + if (!hash_bucket_head) + goto get_remote_info_exit; + hlist_for_each_entry_safe(rem_info, tmp_hlist_node, + hash_bucket_head, hlist_node) { + + if (!iwpm_compare_sockaddr(&rem_info->mapped_loc_sockaddr, + mapped_loc_addr) && + !iwpm_compare_sockaddr(&rem_info->mapped_rem_sockaddr, + mapped_rem_addr)) { + + memcpy(remote_addr, &rem_info->remote_sockaddr, + sizeof(struct sockaddr_storage)); + iwpm_print_sockaddr(remote_addr, + "get_remote_info: Remote sockaddr:"); + + hlist_del_init(&rem_info->hlist_node); + kfree(rem_info); + ret = 0; + break; + } + } + } +get_remote_info_exit: + spin_unlock_irqrestore(&iwpm_reminfo_lock, flags); + return ret; +} + +struct iwpm_nlmsg_request *iwpm_get_nlmsg_request(__u32 nlmsg_seq, + u8 nl_client, gfp_t gfp) +{ + struct iwpm_nlmsg_request *nlmsg_request = NULL; + unsigned long flags; + + nlmsg_request = kzalloc(sizeof(struct iwpm_nlmsg_request), gfp); + if (!nlmsg_request) + return NULL; + + spin_lock_irqsave(&iwpm_nlmsg_req_lock, flags); + list_add_tail(&nlmsg_request->inprocess_list, &iwpm_nlmsg_req_list); + spin_unlock_irqrestore(&iwpm_nlmsg_req_lock, flags); + + kref_init(&nlmsg_request->kref); + kref_get(&nlmsg_request->kref); + nlmsg_request->nlmsg_seq = nlmsg_seq; + nlmsg_request->nl_client = nl_client; + nlmsg_request->request_done = 0; + nlmsg_request->err_code = 0; + sema_init(&nlmsg_request->sem, 1); + down(&nlmsg_request->sem); + return nlmsg_request; +} + +void iwpm_free_nlmsg_request(struct kref *kref) +{ + struct iwpm_nlmsg_request *nlmsg_request; + unsigned long flags; + + nlmsg_request = container_of(kref, struct iwpm_nlmsg_request, kref); + + spin_lock_irqsave(&iwpm_nlmsg_req_lock, flags); + list_del_init(&nlmsg_request->inprocess_list); + spin_unlock_irqrestore(&iwpm_nlmsg_req_lock, flags); + + if (!nlmsg_request->request_done) + pr_debug("%s Freeing incomplete nlmsg request (seq = %u).\n", + __func__, nlmsg_request->nlmsg_seq); + kfree(nlmsg_request); +} + +struct iwpm_nlmsg_request *iwpm_find_nlmsg_request(__u32 echo_seq) +{ + struct iwpm_nlmsg_request *nlmsg_request; + struct iwpm_nlmsg_request *found_request = NULL; + unsigned long flags; + + spin_lock_irqsave(&iwpm_nlmsg_req_lock, flags); + list_for_each_entry(nlmsg_request, &iwpm_nlmsg_req_list, + inprocess_list) { + if (nlmsg_request->nlmsg_seq == echo_seq) { + found_request = nlmsg_request; + kref_get(&nlmsg_request->kref); + break; + } + } + spin_unlock_irqrestore(&iwpm_nlmsg_req_lock, flags); + return found_request; +} + +int iwpm_wait_complete_req(struct iwpm_nlmsg_request *nlmsg_request) +{ + int ret; + + ret = down_timeout(&nlmsg_request->sem, IWPM_NL_TIMEOUT); + if (ret) { + ret = -EINVAL; + pr_info("%s: Timeout %d sec for netlink request (seq = %u)\n", + __func__, (IWPM_NL_TIMEOUT/HZ), nlmsg_request->nlmsg_seq); + } else { + ret = nlmsg_request->err_code; + } + kref_put(&nlmsg_request->kref, iwpm_free_nlmsg_request); + return ret; +} + +int iwpm_get_nlmsg_seq(void) +{ + return atomic_inc_return(&iwpm_admin.nlmsg_seq); +} + +int iwpm_valid_client(u8 nl_client) +{ + return iwpm_admin.client_list[nl_client]; +} + +void iwpm_set_valid(u8 nl_client, int valid) +{ + iwpm_admin.client_list[nl_client] = valid; +} + +/* valid client */ +u32 iwpm_get_registration(u8 nl_client) +{ + return iwpm_admin.reg_list[nl_client]; +} + +/* valid client */ +void iwpm_set_registration(u8 nl_client, u32 reg) +{ + iwpm_admin.reg_list[nl_client] = reg; +} + +/* valid client */ +u32 iwpm_check_registration(u8 nl_client, u32 reg) +{ + return (iwpm_get_registration(nl_client) & reg); +} + +int iwpm_compare_sockaddr(struct sockaddr_storage *a_sockaddr, + struct sockaddr_storage *b_sockaddr) +{ + if (a_sockaddr->ss_family != b_sockaddr->ss_family) + return 1; + if (a_sockaddr->ss_family == AF_INET) { + struct sockaddr_in *a4_sockaddr = + (struct sockaddr_in *)a_sockaddr; + struct sockaddr_in *b4_sockaddr = + (struct sockaddr_in *)b_sockaddr; + if (!memcmp(&a4_sockaddr->sin_addr, + &b4_sockaddr->sin_addr, sizeof(struct in_addr)) + && a4_sockaddr->sin_port == b4_sockaddr->sin_port) + return 0; + + } else if (a_sockaddr->ss_family == AF_INET6) { + struct sockaddr_in6 *a6_sockaddr = + (struct sockaddr_in6 *)a_sockaddr; + struct sockaddr_in6 *b6_sockaddr = + (struct sockaddr_in6 *)b_sockaddr; + if (!memcmp(&a6_sockaddr->sin6_addr, + &b6_sockaddr->sin6_addr, sizeof(struct in6_addr)) + && a6_sockaddr->sin6_port == b6_sockaddr->sin6_port) + return 0; + + } else { + pr_err("%s: Invalid sockaddr family\n", __func__); + } + return 1; +} + +struct sk_buff *iwpm_create_nlmsg(u32 nl_op, struct nlmsghdr **nlh, + int nl_client) +{ + struct sk_buff *skb = NULL; + + skb = dev_alloc_skb(IWPM_MSG_SIZE); + if (!skb) + goto create_nlmsg_exit; + + if (!(ibnl_put_msg(skb, nlh, 0, 0, nl_client, nl_op, + NLM_F_REQUEST))) { + pr_warn("%s: Unable to put the nlmsg header\n", __func__); + dev_kfree_skb(skb); + skb = NULL; + } +create_nlmsg_exit: + return skb; +} + +int iwpm_parse_nlmsg(struct netlink_callback *cb, int policy_max, + const struct nla_policy *nlmsg_policy, + struct nlattr *nltb[], const char *msg_type) +{ + int nlh_len = 0; + int ret; + const char *err_str = ""; + + ret = nlmsg_validate_deprecated(cb->nlh, nlh_len, policy_max - 1, + nlmsg_policy, NULL); + if (ret) { + err_str = "Invalid attribute"; + goto parse_nlmsg_error; + } + ret = nlmsg_parse_deprecated(cb->nlh, nlh_len, nltb, policy_max - 1, + nlmsg_policy, NULL); + if (ret) { + err_str = "Unable to parse the nlmsg"; + goto parse_nlmsg_error; + } + ret = iwpm_validate_nlmsg_attr(nltb, policy_max); + if (ret) { + err_str = "Invalid NULL attribute"; + goto parse_nlmsg_error; + } + return 0; +parse_nlmsg_error: + pr_warn("%s: %s (msg type %s ret = %d)\n", + __func__, err_str, msg_type, ret); + return ret; +} + +void iwpm_print_sockaddr(struct sockaddr_storage *sockaddr, char *msg) +{ + struct sockaddr_in6 *sockaddr_v6; + struct sockaddr_in *sockaddr_v4; + + switch (sockaddr->ss_family) { + case AF_INET: + sockaddr_v4 = (struct sockaddr_in *)sockaddr; + pr_debug("%s IPV4 %pI4: %u(0x%04X)\n", + msg, &sockaddr_v4->sin_addr, + ntohs(sockaddr_v4->sin_port), + ntohs(sockaddr_v4->sin_port)); + break; + case AF_INET6: + sockaddr_v6 = (struct sockaddr_in6 *)sockaddr; + pr_debug("%s IPV6 %pI6: %u(0x%04X)\n", + msg, &sockaddr_v6->sin6_addr, + ntohs(sockaddr_v6->sin6_port), + ntohs(sockaddr_v6->sin6_port)); + break; + default: + break; + } +} + +static u32 iwpm_ipv6_jhash(struct sockaddr_in6 *ipv6_sockaddr) +{ + u32 ipv6_hash = jhash(&ipv6_sockaddr->sin6_addr, sizeof(struct in6_addr), 0); + u32 hash = jhash_2words(ipv6_hash, (__force u32) ipv6_sockaddr->sin6_port, 0); + return hash; +} + +static u32 iwpm_ipv4_jhash(struct sockaddr_in *ipv4_sockaddr) +{ + u32 ipv4_hash = jhash(&ipv4_sockaddr->sin_addr, sizeof(struct in_addr), 0); + u32 hash = jhash_2words(ipv4_hash, (__force u32) ipv4_sockaddr->sin_port, 0); + return hash; +} + +static int get_hash_bucket(struct sockaddr_storage *a_sockaddr, + struct sockaddr_storage *b_sockaddr, u32 *hash) +{ + u32 a_hash, b_hash; + + if (a_sockaddr->ss_family == AF_INET) { + a_hash = iwpm_ipv4_jhash((struct sockaddr_in *) a_sockaddr); + b_hash = iwpm_ipv4_jhash((struct sockaddr_in *) b_sockaddr); + + } else if (a_sockaddr->ss_family == AF_INET6) { + a_hash = iwpm_ipv6_jhash((struct sockaddr_in6 *) a_sockaddr); + b_hash = iwpm_ipv6_jhash((struct sockaddr_in6 *) b_sockaddr); + } else { + pr_err("%s: Invalid sockaddr family\n", __func__); + return -EINVAL; + } + + if (a_hash == b_hash) /* if port mapper isn't available */ + *hash = a_hash; + else + *hash = jhash_2words(a_hash, b_hash, 0); + return 0; +} + +static struct hlist_head *get_mapinfo_hash_bucket(struct sockaddr_storage + *local_sockaddr, struct sockaddr_storage + *mapped_sockaddr) +{ + u32 hash; + int ret; + + ret = get_hash_bucket(local_sockaddr, mapped_sockaddr, &hash); + if (ret) + return NULL; + return &iwpm_hash_bucket[hash & IWPM_MAPINFO_HASH_MASK]; +} + +static struct hlist_head *get_reminfo_hash_bucket(struct sockaddr_storage + *mapped_loc_sockaddr, struct sockaddr_storage + *mapped_rem_sockaddr) +{ + u32 hash; + int ret; + + ret = get_hash_bucket(mapped_loc_sockaddr, mapped_rem_sockaddr, &hash); + if (ret) + return NULL; + return &iwpm_reminfo_bucket[hash & IWPM_REMINFO_HASH_MASK]; +} + +static int send_mapinfo_num(u32 mapping_num, u8 nl_client, int iwpm_pid) +{ + struct sk_buff *skb = NULL; + struct nlmsghdr *nlh; + u32 msg_seq; + const char *err_str = ""; + int ret = -EINVAL; + + skb = iwpm_create_nlmsg(RDMA_NL_IWPM_MAPINFO_NUM, &nlh, nl_client); + if (!skb) { + err_str = "Unable to create a nlmsg"; + goto mapinfo_num_error; + } + nlh->nlmsg_seq = iwpm_get_nlmsg_seq(); + msg_seq = 0; + err_str = "Unable to put attribute of mapinfo number nlmsg"; + ret = ibnl_put_attr(skb, nlh, sizeof(u32), &msg_seq, IWPM_NLA_MAPINFO_SEQ); + if (ret) + goto mapinfo_num_error; + ret = ibnl_put_attr(skb, nlh, sizeof(u32), + &mapping_num, IWPM_NLA_MAPINFO_SEND_NUM); + if (ret) + goto mapinfo_num_error; + + nlmsg_end(skb, nlh); + + ret = rdma_nl_unicast(&init_net, skb, iwpm_pid); + if (ret) { + skb = NULL; + err_str = "Unable to send a nlmsg"; + goto mapinfo_num_error; + } + pr_debug("%s: Sent mapping number = %d\n", __func__, mapping_num); + return 0; +mapinfo_num_error: + pr_info("%s: %s\n", __func__, err_str); + dev_kfree_skb(skb); + return ret; +} + +static int send_nlmsg_done(struct sk_buff *skb, u8 nl_client, int iwpm_pid) +{ + struct nlmsghdr *nlh = NULL; + int ret = 0; + + if (!skb) + return ret; + if (!(ibnl_put_msg(skb, &nlh, 0, 0, nl_client, + RDMA_NL_IWPM_MAPINFO, NLM_F_MULTI))) { + pr_warn("%s Unable to put NLMSG_DONE\n", __func__); + dev_kfree_skb(skb); + return -ENOMEM; + } + nlh->nlmsg_type = NLMSG_DONE; + ret = rdma_nl_unicast(&init_net, skb, iwpm_pid); + if (ret) + pr_warn("%s Unable to send a nlmsg\n", __func__); + return ret; +} + +int iwpm_send_mapinfo(u8 nl_client, int iwpm_pid) +{ + struct iwpm_mapping_info *map_info; + struct sk_buff *skb = NULL; + struct nlmsghdr *nlh; + int skb_num = 0, mapping_num = 0; + int i = 0, nlmsg_bytes = 0; + unsigned long flags; + const char *err_str = ""; + int ret; + + skb = dev_alloc_skb(NLMSG_GOODSIZE); + if (!skb) { + ret = -ENOMEM; + err_str = "Unable to allocate skb"; + goto send_mapping_info_exit; + } + skb_num++; + spin_lock_irqsave(&iwpm_mapinfo_lock, flags); + ret = -EINVAL; + for (i = 0; i < IWPM_MAPINFO_HASH_SIZE; i++) { + hlist_for_each_entry(map_info, &iwpm_hash_bucket[i], + hlist_node) { + if (map_info->nl_client != nl_client) + continue; + nlh = NULL; + if (!(ibnl_put_msg(skb, &nlh, 0, 0, nl_client, + RDMA_NL_IWPM_MAPINFO, NLM_F_MULTI))) { + ret = -ENOMEM; + err_str = "Unable to put the nlmsg header"; + goto send_mapping_info_unlock; + } + err_str = "Unable to put attribute of the nlmsg"; + ret = ibnl_put_attr(skb, nlh, + sizeof(struct sockaddr_storage), + &map_info->local_sockaddr, + IWPM_NLA_MAPINFO_LOCAL_ADDR); + if (ret) + goto send_mapping_info_unlock; + + ret = ibnl_put_attr(skb, nlh, + sizeof(struct sockaddr_storage), + &map_info->mapped_sockaddr, + IWPM_NLA_MAPINFO_MAPPED_ADDR); + if (ret) + goto send_mapping_info_unlock; + + if (iwpm_ulib_version > IWPM_UABI_VERSION_MIN) { + ret = ibnl_put_attr(skb, nlh, sizeof(u32), + &map_info->map_flags, + IWPM_NLA_MAPINFO_FLAGS); + if (ret) + goto send_mapping_info_unlock; + } + + nlmsg_end(skb, nlh); + + iwpm_print_sockaddr(&map_info->local_sockaddr, + "send_mapping_info: Local sockaddr:"); + iwpm_print_sockaddr(&map_info->mapped_sockaddr, + "send_mapping_info: Mapped local sockaddr:"); + mapping_num++; + nlmsg_bytes += nlh->nlmsg_len; + + /* check if all mappings can fit in one skb */ + if (NLMSG_GOODSIZE - nlmsg_bytes < nlh->nlmsg_len * 2) { + /* and leave room for NLMSG_DONE */ + nlmsg_bytes = 0; + skb_num++; + spin_unlock_irqrestore(&iwpm_mapinfo_lock, + flags); + /* send the skb */ + ret = send_nlmsg_done(skb, nl_client, iwpm_pid); + skb = NULL; + if (ret) { + err_str = "Unable to send map info"; + goto send_mapping_info_exit; + } + if (skb_num == IWPM_MAPINFO_SKB_COUNT) { + ret = -ENOMEM; + err_str = "Insufficient skbs for map info"; + goto send_mapping_info_exit; + } + skb = dev_alloc_skb(NLMSG_GOODSIZE); + if (!skb) { + ret = -ENOMEM; + err_str = "Unable to allocate skb"; + goto send_mapping_info_exit; + } + spin_lock_irqsave(&iwpm_mapinfo_lock, flags); + } + } + } +send_mapping_info_unlock: + spin_unlock_irqrestore(&iwpm_mapinfo_lock, flags); +send_mapping_info_exit: + if (ret) { + pr_warn("%s: %s (ret = %d)\n", __func__, err_str, ret); + dev_kfree_skb(skb); + return ret; + } + send_nlmsg_done(skb, nl_client, iwpm_pid); + return send_mapinfo_num(mapping_num, nl_client, iwpm_pid); +} + +int iwpm_mapinfo_available(void) +{ + unsigned long flags; + int full_bucket = 0, i = 0; + + spin_lock_irqsave(&iwpm_mapinfo_lock, flags); + if (iwpm_hash_bucket) { + for (i = 0; i < IWPM_MAPINFO_HASH_SIZE; i++) { + if (!hlist_empty(&iwpm_hash_bucket[i])) { + full_bucket = 1; + break; + } + } + } + spin_unlock_irqrestore(&iwpm_mapinfo_lock, flags); + return full_bucket; +} + +int iwpm_send_hello(u8 nl_client, int iwpm_pid, u16 abi_version) +{ + struct sk_buff *skb = NULL; + struct nlmsghdr *nlh; + const char *err_str = ""; + int ret = -EINVAL; + + skb = iwpm_create_nlmsg(RDMA_NL_IWPM_HELLO, &nlh, nl_client); + if (!skb) { + err_str = "Unable to create a nlmsg"; + goto hello_num_error; + } + nlh->nlmsg_seq = iwpm_get_nlmsg_seq(); + err_str = "Unable to put attribute of abi_version into nlmsg"; + ret = ibnl_put_attr(skb, nlh, sizeof(u16), &abi_version, + IWPM_NLA_HELLO_ABI_VERSION); + if (ret) + goto hello_num_error; + nlmsg_end(skb, nlh); + + ret = rdma_nl_unicast(&init_net, skb, iwpm_pid); + if (ret) { + skb = NULL; + err_str = "Unable to send a nlmsg"; + goto hello_num_error; + } + pr_debug("%s: Sent hello abi_version = %u\n", __func__, abi_version); + return 0; +hello_num_error: + pr_info("%s: %s\n", __func__, err_str); + dev_kfree_skb(skb); + return ret; +} diff --git a/drivers/infiniband/core/iwpm_util.h b/drivers/infiniband/core/iwpm_util.h new file mode 100644 index 000000000..1bf87d9fd --- /dev/null +++ b/drivers/infiniband/core/iwpm_util.h @@ -0,0 +1,284 @@ +/* + * Copyright (c) 2014 Intel Corporation. All rights reserved. + * Copyright (c) 2014 Chelsio, Inc. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ +#ifndef _IWPM_UTIL_H +#define _IWPM_UTIL_H + +#include <linux/module.h> +#include <linux/io.h> +#include <linux/in.h> +#include <linux/in6.h> +#include <linux/spinlock.h> +#include <linux/kernel.h> +#include <linux/netdevice.h> +#include <linux/delay.h> +#include <linux/workqueue.h> +#include <linux/mutex.h> +#include <linux/jhash.h> +#include <linux/kref.h> +#include <net/netlink.h> +#include <linux/errno.h> +#include <rdma/iw_portmap.h> +#include <rdma/rdma_netlink.h> + + +#define IWPM_NL_RETRANS 3 +#define IWPM_NL_TIMEOUT (10*HZ) +#define IWPM_MAPINFO_SKB_COUNT 20 + +#define IWPM_PID_UNDEFINED -1 +#define IWPM_PID_UNAVAILABLE -2 + +#define IWPM_REG_UNDEF 0x01 +#define IWPM_REG_VALID 0x02 +#define IWPM_REG_INCOMPL 0x04 + +struct iwpm_nlmsg_request { + struct list_head inprocess_list; + __u32 nlmsg_seq; + void *req_buffer; + u8 nl_client; + u8 request_done; + u16 err_code; + struct semaphore sem; + struct kref kref; +}; + +struct iwpm_mapping_info { + struct hlist_node hlist_node; + struct sockaddr_storage local_sockaddr; + struct sockaddr_storage mapped_sockaddr; + u8 nl_client; + u32 map_flags; +}; + +struct iwpm_remote_info { + struct hlist_node hlist_node; + struct sockaddr_storage remote_sockaddr; + struct sockaddr_storage mapped_loc_sockaddr; + struct sockaddr_storage mapped_rem_sockaddr; + u8 nl_client; +}; + +struct iwpm_admin_data { + atomic_t refcount; + atomic_t nlmsg_seq; + int client_list[RDMA_NL_NUM_CLIENTS]; + u32 reg_list[RDMA_NL_NUM_CLIENTS]; +}; + +/** + * iwpm_get_nlmsg_request - Allocate and initialize netlink message request + * @nlmsg_seq: Sequence number of the netlink message + * @nl_client: The index of the netlink client + * @gfp: Indicates how the memory for the request should be allocated + * + * Returns the newly allocated netlink request object if successful, + * otherwise returns NULL + */ +struct iwpm_nlmsg_request *iwpm_get_nlmsg_request(__u32 nlmsg_seq, + u8 nl_client, gfp_t gfp); + +/** + * iwpm_free_nlmsg_request - Deallocate netlink message request + * @kref: Holds reference of netlink message request + */ +void iwpm_free_nlmsg_request(struct kref *kref); + +/** + * iwpm_find_nlmsg_request - Find netlink message request in the request list + * @echo_seq: Sequence number of the netlink request to find + * + * Returns the found netlink message request, + * if not found, returns NULL + */ +struct iwpm_nlmsg_request *iwpm_find_nlmsg_request(__u32 echo_seq); + +/** + * iwpm_wait_complete_req - Block while servicing the netlink request + * @nlmsg_request: Netlink message request to service + * + * Wakes up, after the request is completed or expired + * Returns 0 if the request is complete without error + */ +int iwpm_wait_complete_req(struct iwpm_nlmsg_request *nlmsg_request); + +/** + * iwpm_get_nlmsg_seq - Get the sequence number for a netlink + * message to send to the port mapper + * + * Returns the sequence number for the netlink message. + */ +int iwpm_get_nlmsg_seq(void); + +/** + * iwpm_add_reminfo - Add remote address info of the connecting peer + * to the remote info hash table + * @reminfo: The remote info to be added + */ +void iwpm_add_remote_info(struct iwpm_remote_info *reminfo); + +/** + * iwpm_valid_client - Check if the port mapper client is valid + * @nl_client: The index of the netlink client + * + * Valid clients need to call iwpm_init() before using + * the port mapper + */ +int iwpm_valid_client(u8 nl_client); + +/** + * iwpm_set_valid - Set the port mapper client to valid or not + * @nl_client: The index of the netlink client + * @valid: 1 if valid or 0 if invalid + */ +void iwpm_set_valid(u8 nl_client, int valid); + +/** + * iwpm_check_registration - Check if the client registration + * matches the given one + * @nl_client: The index of the netlink client + * @reg: The given registration type to compare with + * + * Call iwpm_register_pid() to register a client + * Returns true if the client registration matches reg, + * otherwise returns false + */ +u32 iwpm_check_registration(u8 nl_client, u32 reg); + +/** + * iwpm_set_registration - Set the client registration + * @nl_client: The index of the netlink client + * @reg: Registration type to set + */ +void iwpm_set_registration(u8 nl_client, u32 reg); + +/** + * iwpm_get_registration + * @nl_client: The index of the netlink client + * + * Returns the client registration type + */ +u32 iwpm_get_registration(u8 nl_client); + +/** + * iwpm_send_mapinfo - Send local and mapped IPv4/IPv6 address info of + * a client to the user space port mapper + * @nl_client: The index of the netlink client + * @iwpm_pid: The pid of the user space port mapper + * + * If successful, returns the number of sent mapping info records + */ +int iwpm_send_mapinfo(u8 nl_client, int iwpm_pid); + +/** + * iwpm_mapinfo_available - Check if any mapping info records is available + * in the hash table + * + * Returns 1 if mapping information is available, otherwise returns 0 + */ +int iwpm_mapinfo_available(void); + +/** + * iwpm_compare_sockaddr - Compare two sockaddr storage structs + * @a_sockaddr: first sockaddr to compare + * @b_sockaddr: second sockaddr to compare + * + * Return: 0 if they are holding the same ip/tcp address info, + * otherwise returns 1 + */ +int iwpm_compare_sockaddr(struct sockaddr_storage *a_sockaddr, + struct sockaddr_storage *b_sockaddr); + +/** + * iwpm_validate_nlmsg_attr - Check for NULL netlink attributes + * @nltb: Holds address of each netlink message attributes + * @nla_count: Number of netlink message attributes + * + * Returns error if any of the nla_count attributes is NULL + */ +static inline int iwpm_validate_nlmsg_attr(struct nlattr *nltb[], + int nla_count) +{ + int i; + for (i = 1; i < nla_count; i++) { + if (!nltb[i]) + return -EINVAL; + } + return 0; +} + +/** + * iwpm_create_nlmsg - Allocate skb and form a netlink message + * @nl_op: Netlink message opcode + * @nlh: Holds address of the netlink message header in skb + * @nl_client: The index of the netlink client + * + * Returns the newly allcated skb, or NULL if the tailroom of the skb + * is insufficient to store the message header and payload + */ +struct sk_buff *iwpm_create_nlmsg(u32 nl_op, struct nlmsghdr **nlh, + int nl_client); + +/** + * iwpm_parse_nlmsg - Validate and parse the received netlink message + * @cb: Netlink callback structure + * @policy_max: Maximum attribute type to be expected + * @nlmsg_policy: Validation policy + * @nltb: Array to store policy_max parsed elements + * @msg_type: Type of netlink message + * + * Returns 0 on success or a negative error code + */ +int iwpm_parse_nlmsg(struct netlink_callback *cb, int policy_max, + const struct nla_policy *nlmsg_policy, + struct nlattr *nltb[], const char *msg_type); + +/** + * iwpm_print_sockaddr - Print IPv4/IPv6 address and TCP port + * @sockaddr: Socket address to print + * @msg: Message to print + */ +void iwpm_print_sockaddr(struct sockaddr_storage *sockaddr, char *msg); + +/** + * iwpm_send_hello - Send hello response to iwpmd + * + * @nl_client: The index of the netlink client + * @iwpm_pid: The pid of the user space port mapper + * @abi_version: The kernel's abi_version + * + * Returns 0 on success or a negative error code + */ +int iwpm_send_hello(u8 nl_client, int iwpm_pid, u16 abi_version); +extern u16 iwpm_ulib_version; +#endif diff --git a/drivers/infiniband/core/lag.c b/drivers/infiniband/core/lag.c new file mode 100644 index 000000000..7063e41ea --- /dev/null +++ b/drivers/infiniband/core/lag.c @@ -0,0 +1,138 @@ +// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB +/* + * Copyright (c) 2020 Mellanox Technologies. All rights reserved. + */ + +#include <rdma/ib_verbs.h> +#include <rdma/ib_cache.h> +#include <rdma/lag.h> + +static struct sk_buff *rdma_build_skb(struct ib_device *device, + struct net_device *netdev, + struct rdma_ah_attr *ah_attr, + gfp_t flags) +{ + struct ipv6hdr *ip6h; + struct sk_buff *skb; + struct ethhdr *eth; + struct iphdr *iph; + struct udphdr *uh; + u8 smac[ETH_ALEN]; + bool is_ipv4; + int hdr_len; + + is_ipv4 = ipv6_addr_v4mapped((struct in6_addr *)ah_attr->grh.dgid.raw); + hdr_len = ETH_HLEN + sizeof(struct udphdr) + LL_RESERVED_SPACE(netdev); + hdr_len += is_ipv4 ? sizeof(struct iphdr) : sizeof(struct ipv6hdr); + + skb = alloc_skb(hdr_len, flags); + if (!skb) + return NULL; + + skb->dev = netdev; + skb_reserve(skb, hdr_len); + skb_push(skb, sizeof(struct udphdr)); + skb_reset_transport_header(skb); + uh = udp_hdr(skb); + uh->source = + htons(rdma_flow_label_to_udp_sport(ah_attr->grh.flow_label)); + uh->dest = htons(ROCE_V2_UDP_DPORT); + uh->len = htons(sizeof(struct udphdr)); + + if (is_ipv4) { + skb_push(skb, sizeof(struct iphdr)); + skb_reset_network_header(skb); + iph = ip_hdr(skb); + iph->frag_off = 0; + iph->version = 4; + iph->protocol = IPPROTO_UDP; + iph->ihl = 0x5; + iph->tot_len = htons(sizeof(struct udphdr) + sizeof(struct + iphdr)); + memcpy(&iph->saddr, ah_attr->grh.sgid_attr->gid.raw + 12, + sizeof(struct in_addr)); + memcpy(&iph->daddr, ah_attr->grh.dgid.raw + 12, + sizeof(struct in_addr)); + } else { + skb_push(skb, sizeof(struct ipv6hdr)); + skb_reset_network_header(skb); + ip6h = ipv6_hdr(skb); + ip6h->version = 6; + ip6h->nexthdr = IPPROTO_UDP; + memcpy(&ip6h->flow_lbl, &ah_attr->grh.flow_label, + sizeof(*ip6h->flow_lbl)); + memcpy(&ip6h->saddr, ah_attr->grh.sgid_attr->gid.raw, + sizeof(struct in6_addr)); + memcpy(&ip6h->daddr, ah_attr->grh.dgid.raw, + sizeof(struct in6_addr)); + } + + skb_push(skb, sizeof(struct ethhdr)); + skb_reset_mac_header(skb); + eth = eth_hdr(skb); + skb->protocol = eth->h_proto = htons(is_ipv4 ? ETH_P_IP : ETH_P_IPV6); + rdma_read_gid_l2_fields(ah_attr->grh.sgid_attr, NULL, smac); + memcpy(eth->h_source, smac, ETH_ALEN); + memcpy(eth->h_dest, ah_attr->roce.dmac, ETH_ALEN); + + return skb; +} + +static struct net_device *rdma_get_xmit_slave_udp(struct ib_device *device, + struct net_device *master, + struct rdma_ah_attr *ah_attr, + gfp_t flags) +{ + struct net_device *slave; + struct sk_buff *skb; + + skb = rdma_build_skb(device, master, ah_attr, flags); + if (!skb) + return ERR_PTR(-ENOMEM); + + rcu_read_lock(); + slave = netdev_get_xmit_slave(master, skb, + !!(device->lag_flags & + RDMA_LAG_FLAGS_HASH_ALL_SLAVES)); + if (slave) + dev_hold(slave); + rcu_read_unlock(); + kfree_skb(skb); + return slave; +} + +void rdma_lag_put_ah_roce_slave(struct net_device *xmit_slave) +{ + if (xmit_slave) + dev_put(xmit_slave); +} + +struct net_device *rdma_lag_get_ah_roce_slave(struct ib_device *device, + struct rdma_ah_attr *ah_attr, + gfp_t flags) +{ + struct net_device *slave = NULL; + struct net_device *master; + + if (!(ah_attr->type == RDMA_AH_ATTR_TYPE_ROCE && + ah_attr->grh.sgid_attr->gid_type == IB_GID_TYPE_ROCE_UDP_ENCAP && + ah_attr->grh.flow_label)) + return NULL; + + rcu_read_lock(); + master = rdma_read_gid_attr_ndev_rcu(ah_attr->grh.sgid_attr); + if (IS_ERR(master)) { + rcu_read_unlock(); + return master; + } + dev_hold(master); + rcu_read_unlock(); + + if (!netif_is_bond_master(master)) + goto put; + + slave = rdma_get_xmit_slave_udp(device, master, ah_attr, flags); +put: + dev_put(master); + return slave; +} diff --git a/drivers/infiniband/core/mad.c b/drivers/infiniband/core/mad.c new file mode 100644 index 000000000..9355e521d --- /dev/null +++ b/drivers/infiniband/core/mad.c @@ -0,0 +1,3171 @@ +/* + * Copyright (c) 2004-2007 Voltaire, Inc. All rights reserved. + * Copyright (c) 2005 Intel Corporation. All rights reserved. + * Copyright (c) 2005 Mellanox Technologies Ltd. All rights reserved. + * Copyright (c) 2009 HNR Consulting. All rights reserved. + * Copyright (c) 2014,2018 Intel Corporation. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + */ + +#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt + +#include <linux/dma-mapping.h> +#include <linux/slab.h> +#include <linux/module.h> +#include <linux/security.h> +#include <linux/xarray.h> +#include <rdma/ib_cache.h> + +#include "mad_priv.h" +#include "core_priv.h" +#include "mad_rmpp.h" +#include "smi.h" +#include "opa_smi.h" +#include "agent.h" + +#define CREATE_TRACE_POINTS +#include <trace/events/ib_mad.h> + +#ifdef CONFIG_TRACEPOINTS +static void create_mad_addr_info(struct ib_mad_send_wr_private *mad_send_wr, + struct ib_mad_qp_info *qp_info, + struct trace_event_raw_ib_mad_send_template *entry) +{ + u16 pkey; + struct ib_device *dev = qp_info->port_priv->device; + u8 pnum = qp_info->port_priv->port_num; + struct ib_ud_wr *wr = &mad_send_wr->send_wr; + struct rdma_ah_attr attr = {}; + + rdma_query_ah(wr->ah, &attr); + + /* These are common */ + entry->sl = attr.sl; + ib_query_pkey(dev, pnum, wr->pkey_index, &pkey); + entry->pkey = pkey; + entry->rqpn = wr->remote_qpn; + entry->rqkey = wr->remote_qkey; + entry->dlid = rdma_ah_get_dlid(&attr); +} +#endif + +static int mad_sendq_size = IB_MAD_QP_SEND_SIZE; +static int mad_recvq_size = IB_MAD_QP_RECV_SIZE; + +module_param_named(send_queue_size, mad_sendq_size, int, 0444); +MODULE_PARM_DESC(send_queue_size, "Size of send queue in number of work requests"); +module_param_named(recv_queue_size, mad_recvq_size, int, 0444); +MODULE_PARM_DESC(recv_queue_size, "Size of receive queue in number of work requests"); + +static DEFINE_XARRAY_ALLOC1(ib_mad_clients); +static u32 ib_mad_client_next; +static struct list_head ib_mad_port_list; + +/* Port list lock */ +static DEFINE_SPINLOCK(ib_mad_port_list_lock); + +/* Forward declarations */ +static int method_in_use(struct ib_mad_mgmt_method_table **method, + struct ib_mad_reg_req *mad_reg_req); +static void remove_mad_reg_req(struct ib_mad_agent_private *priv); +static struct ib_mad_agent_private *find_mad_agent( + struct ib_mad_port_private *port_priv, + const struct ib_mad_hdr *mad); +static int ib_mad_post_receive_mads(struct ib_mad_qp_info *qp_info, + struct ib_mad_private *mad); +static void cancel_mads(struct ib_mad_agent_private *mad_agent_priv); +static void timeout_sends(struct work_struct *work); +static void local_completions(struct work_struct *work); +static int add_nonoui_reg_req(struct ib_mad_reg_req *mad_reg_req, + struct ib_mad_agent_private *agent_priv, + u8 mgmt_class); +static int add_oui_reg_req(struct ib_mad_reg_req *mad_reg_req, + struct ib_mad_agent_private *agent_priv); +static bool ib_mad_send_error(struct ib_mad_port_private *port_priv, + struct ib_wc *wc); +static void ib_mad_send_done(struct ib_cq *cq, struct ib_wc *wc); + +/* + * Returns a ib_mad_port_private structure or NULL for a device/port + * Assumes ib_mad_port_list_lock is being held + */ +static inline struct ib_mad_port_private * +__ib_get_mad_port(struct ib_device *device, int port_num) +{ + struct ib_mad_port_private *entry; + + list_for_each_entry(entry, &ib_mad_port_list, port_list) { + if (entry->device == device && entry->port_num == port_num) + return entry; + } + return NULL; +} + +/* + * Wrapper function to return a ib_mad_port_private structure or NULL + * for a device/port + */ +static inline struct ib_mad_port_private * +ib_get_mad_port(struct ib_device *device, int port_num) +{ + struct ib_mad_port_private *entry; + unsigned long flags; + + spin_lock_irqsave(&ib_mad_port_list_lock, flags); + entry = __ib_get_mad_port(device, port_num); + spin_unlock_irqrestore(&ib_mad_port_list_lock, flags); + + return entry; +} + +static inline u8 convert_mgmt_class(u8 mgmt_class) +{ + /* Alias IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE to 0 */ + return mgmt_class == IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE ? + 0 : mgmt_class; +} + +static int get_spl_qp_index(enum ib_qp_type qp_type) +{ + switch (qp_type) + { + case IB_QPT_SMI: + return 0; + case IB_QPT_GSI: + return 1; + default: + return -1; + } +} + +static int vendor_class_index(u8 mgmt_class) +{ + return mgmt_class - IB_MGMT_CLASS_VENDOR_RANGE2_START; +} + +static int is_vendor_class(u8 mgmt_class) +{ + if ((mgmt_class < IB_MGMT_CLASS_VENDOR_RANGE2_START) || + (mgmt_class > IB_MGMT_CLASS_VENDOR_RANGE2_END)) + return 0; + return 1; +} + +static int is_vendor_oui(char *oui) +{ + if (oui[0] || oui[1] || oui[2]) + return 1; + return 0; +} + +static int is_vendor_method_in_use( + struct ib_mad_mgmt_vendor_class *vendor_class, + struct ib_mad_reg_req *mad_reg_req) +{ + struct ib_mad_mgmt_method_table *method; + int i; + + for (i = 0; i < MAX_MGMT_OUI; i++) { + if (!memcmp(vendor_class->oui[i], mad_reg_req->oui, 3)) { + method = vendor_class->method_table[i]; + if (method) { + if (method_in_use(&method, mad_reg_req)) + return 1; + else + break; + } + } + } + return 0; +} + +int ib_response_mad(const struct ib_mad_hdr *hdr) +{ + return ((hdr->method & IB_MGMT_METHOD_RESP) || + (hdr->method == IB_MGMT_METHOD_TRAP_REPRESS) || + ((hdr->mgmt_class == IB_MGMT_CLASS_BM) && + (hdr->attr_mod & IB_BM_ATTR_MOD_RESP))); +} +EXPORT_SYMBOL(ib_response_mad); + +/* + * ib_register_mad_agent - Register to send/receive MADs + * + * Context: Process context. + */ +struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device, + u8 port_num, + enum ib_qp_type qp_type, + struct ib_mad_reg_req *mad_reg_req, + u8 rmpp_version, + ib_mad_send_handler send_handler, + ib_mad_recv_handler recv_handler, + void *context, + u32 registration_flags) +{ + struct ib_mad_port_private *port_priv; + struct ib_mad_agent *ret = ERR_PTR(-EINVAL); + struct ib_mad_agent_private *mad_agent_priv; + struct ib_mad_reg_req *reg_req = NULL; + struct ib_mad_mgmt_class_table *class; + struct ib_mad_mgmt_vendor_class_table *vendor; + struct ib_mad_mgmt_vendor_class *vendor_class; + struct ib_mad_mgmt_method_table *method; + int ret2, qpn; + u8 mgmt_class, vclass; + + if ((qp_type == IB_QPT_SMI && !rdma_cap_ib_smi(device, port_num)) || + (qp_type == IB_QPT_GSI && !rdma_cap_ib_cm(device, port_num))) + return ERR_PTR(-EPROTONOSUPPORT); + + /* Validate parameters */ + qpn = get_spl_qp_index(qp_type); + if (qpn == -1) { + dev_dbg_ratelimited(&device->dev, "%s: invalid QP Type %d\n", + __func__, qp_type); + goto error1; + } + + if (rmpp_version && rmpp_version != IB_MGMT_RMPP_VERSION) { + dev_dbg_ratelimited(&device->dev, + "%s: invalid RMPP Version %u\n", + __func__, rmpp_version); + goto error1; + } + + /* Validate MAD registration request if supplied */ + if (mad_reg_req) { + if (mad_reg_req->mgmt_class_version >= MAX_MGMT_VERSION) { + dev_dbg_ratelimited(&device->dev, + "%s: invalid Class Version %u\n", + __func__, + mad_reg_req->mgmt_class_version); + goto error1; + } + if (!recv_handler) { + dev_dbg_ratelimited(&device->dev, + "%s: no recv_handler\n", __func__); + goto error1; + } + if (mad_reg_req->mgmt_class >= MAX_MGMT_CLASS) { + /* + * IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE is the only + * one in this range currently allowed + */ + if (mad_reg_req->mgmt_class != + IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE) { + dev_dbg_ratelimited(&device->dev, + "%s: Invalid Mgmt Class 0x%x\n", + __func__, mad_reg_req->mgmt_class); + goto error1; + } + } else if (mad_reg_req->mgmt_class == 0) { + /* + * Class 0 is reserved in IBA and is used for + * aliasing of IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE + */ + dev_dbg_ratelimited(&device->dev, + "%s: Invalid Mgmt Class 0\n", + __func__); + goto error1; + } else if (is_vendor_class(mad_reg_req->mgmt_class)) { + /* + * If class is in "new" vendor range, + * ensure supplied OUI is not zero + */ + if (!is_vendor_oui(mad_reg_req->oui)) { + dev_dbg_ratelimited(&device->dev, + "%s: No OUI specified for class 0x%x\n", + __func__, + mad_reg_req->mgmt_class); + goto error1; + } + } + /* Make sure class supplied is consistent with RMPP */ + if (!ib_is_mad_class_rmpp(mad_reg_req->mgmt_class)) { + if (rmpp_version) { + dev_dbg_ratelimited(&device->dev, + "%s: RMPP version for non-RMPP class 0x%x\n", + __func__, mad_reg_req->mgmt_class); + goto error1; + } + } + + /* Make sure class supplied is consistent with QP type */ + if (qp_type == IB_QPT_SMI) { + if ((mad_reg_req->mgmt_class != + IB_MGMT_CLASS_SUBN_LID_ROUTED) && + (mad_reg_req->mgmt_class != + IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE)) { + dev_dbg_ratelimited(&device->dev, + "%s: Invalid SM QP type: class 0x%x\n", + __func__, mad_reg_req->mgmt_class); + goto error1; + } + } else { + if ((mad_reg_req->mgmt_class == + IB_MGMT_CLASS_SUBN_LID_ROUTED) || + (mad_reg_req->mgmt_class == + IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE)) { + dev_dbg_ratelimited(&device->dev, + "%s: Invalid GS QP type: class 0x%x\n", + __func__, mad_reg_req->mgmt_class); + goto error1; + } + } + } else { + /* No registration request supplied */ + if (!send_handler) + goto error1; + if (registration_flags & IB_MAD_USER_RMPP) + goto error1; + } + + /* Validate device and port */ + port_priv = ib_get_mad_port(device, port_num); + if (!port_priv) { + dev_dbg_ratelimited(&device->dev, "%s: Invalid port %d\n", + __func__, port_num); + ret = ERR_PTR(-ENODEV); + goto error1; + } + + /* Verify the QP requested is supported. For example, Ethernet devices + * will not have QP0. + */ + if (!port_priv->qp_info[qpn].qp) { + dev_dbg_ratelimited(&device->dev, "%s: QP %d not supported\n", + __func__, qpn); + ret = ERR_PTR(-EPROTONOSUPPORT); + goto error1; + } + + /* Allocate structures */ + mad_agent_priv = kzalloc(sizeof *mad_agent_priv, GFP_KERNEL); + if (!mad_agent_priv) { + ret = ERR_PTR(-ENOMEM); + goto error1; + } + + if (mad_reg_req) { + reg_req = kmemdup(mad_reg_req, sizeof *reg_req, GFP_KERNEL); + if (!reg_req) { + ret = ERR_PTR(-ENOMEM); + goto error3; + } + } + + /* Now, fill in the various structures */ + mad_agent_priv->qp_info = &port_priv->qp_info[qpn]; + mad_agent_priv->reg_req = reg_req; + mad_agent_priv->agent.rmpp_version = rmpp_version; + mad_agent_priv->agent.device = device; + mad_agent_priv->agent.recv_handler = recv_handler; + mad_agent_priv->agent.send_handler = send_handler; + mad_agent_priv->agent.context = context; + mad_agent_priv->agent.qp = port_priv->qp_info[qpn].qp; + mad_agent_priv->agent.port_num = port_num; + mad_agent_priv->agent.flags = registration_flags; + spin_lock_init(&mad_agent_priv->lock); + INIT_LIST_HEAD(&mad_agent_priv->send_list); + INIT_LIST_HEAD(&mad_agent_priv->wait_list); + INIT_LIST_HEAD(&mad_agent_priv->done_list); + INIT_LIST_HEAD(&mad_agent_priv->rmpp_list); + INIT_DELAYED_WORK(&mad_agent_priv->timed_work, timeout_sends); + INIT_LIST_HEAD(&mad_agent_priv->local_list); + INIT_WORK(&mad_agent_priv->local_work, local_completions); + refcount_set(&mad_agent_priv->refcount, 1); + init_completion(&mad_agent_priv->comp); + + ret2 = ib_mad_agent_security_setup(&mad_agent_priv->agent, qp_type); + if (ret2) { + ret = ERR_PTR(ret2); + goto error4; + } + + /* + * The mlx4 driver uses the top byte to distinguish which virtual + * function generated the MAD, so we must avoid using it. + */ + ret2 = xa_alloc_cyclic(&ib_mad_clients, &mad_agent_priv->agent.hi_tid, + mad_agent_priv, XA_LIMIT(0, (1 << 24) - 1), + &ib_mad_client_next, GFP_KERNEL); + if (ret2 < 0) { + ret = ERR_PTR(ret2); + goto error5; + } + + /* + * Make sure MAD registration (if supplied) + * is non overlapping with any existing ones + */ + spin_lock_irq(&port_priv->reg_lock); + if (mad_reg_req) { + mgmt_class = convert_mgmt_class(mad_reg_req->mgmt_class); + if (!is_vendor_class(mgmt_class)) { + class = port_priv->version[mad_reg_req-> + mgmt_class_version].class; + if (class) { + method = class->method_table[mgmt_class]; + if (method) { + if (method_in_use(&method, + mad_reg_req)) + goto error6; + } + } + ret2 = add_nonoui_reg_req(mad_reg_req, mad_agent_priv, + mgmt_class); + } else { + /* "New" vendor class range */ + vendor = port_priv->version[mad_reg_req-> + mgmt_class_version].vendor; + if (vendor) { + vclass = vendor_class_index(mgmt_class); + vendor_class = vendor->vendor_class[vclass]; + if (vendor_class) { + if (is_vendor_method_in_use( + vendor_class, + mad_reg_req)) + goto error6; + } + } + ret2 = add_oui_reg_req(mad_reg_req, mad_agent_priv); + } + if (ret2) { + ret = ERR_PTR(ret2); + goto error6; + } + } + spin_unlock_irq(&port_priv->reg_lock); + + trace_ib_mad_create_agent(mad_agent_priv); + return &mad_agent_priv->agent; +error6: + spin_unlock_irq(&port_priv->reg_lock); + xa_erase(&ib_mad_clients, mad_agent_priv->agent.hi_tid); +error5: + ib_mad_agent_security_cleanup(&mad_agent_priv->agent); +error4: + kfree(reg_req); +error3: + kfree(mad_agent_priv); +error1: + return ret; +} +EXPORT_SYMBOL(ib_register_mad_agent); + +static inline void deref_mad_agent(struct ib_mad_agent_private *mad_agent_priv) +{ + if (refcount_dec_and_test(&mad_agent_priv->refcount)) + complete(&mad_agent_priv->comp); +} + +static void unregister_mad_agent(struct ib_mad_agent_private *mad_agent_priv) +{ + struct ib_mad_port_private *port_priv; + + /* Note that we could still be handling received MADs */ + trace_ib_mad_unregister_agent(mad_agent_priv); + + /* + * Canceling all sends results in dropping received response + * MADs, preventing us from queuing additional work + */ + cancel_mads(mad_agent_priv); + port_priv = mad_agent_priv->qp_info->port_priv; + cancel_delayed_work(&mad_agent_priv->timed_work); + + spin_lock_irq(&port_priv->reg_lock); + remove_mad_reg_req(mad_agent_priv); + spin_unlock_irq(&port_priv->reg_lock); + xa_erase(&ib_mad_clients, mad_agent_priv->agent.hi_tid); + + flush_workqueue(port_priv->wq); + + deref_mad_agent(mad_agent_priv); + wait_for_completion(&mad_agent_priv->comp); + ib_cancel_rmpp_recvs(mad_agent_priv); + + ib_mad_agent_security_cleanup(&mad_agent_priv->agent); + + kfree(mad_agent_priv->reg_req); + kfree_rcu(mad_agent_priv, rcu); +} + +/* + * ib_unregister_mad_agent - Unregisters a client from using MAD services + * + * Context: Process context. + */ +void ib_unregister_mad_agent(struct ib_mad_agent *mad_agent) +{ + struct ib_mad_agent_private *mad_agent_priv; + + mad_agent_priv = container_of(mad_agent, + struct ib_mad_agent_private, + agent); + unregister_mad_agent(mad_agent_priv); +} +EXPORT_SYMBOL(ib_unregister_mad_agent); + +static void dequeue_mad(struct ib_mad_list_head *mad_list) +{ + struct ib_mad_queue *mad_queue; + unsigned long flags; + + mad_queue = mad_list->mad_queue; + spin_lock_irqsave(&mad_queue->lock, flags); + list_del(&mad_list->list); + mad_queue->count--; + spin_unlock_irqrestore(&mad_queue->lock, flags); +} + +static void build_smp_wc(struct ib_qp *qp, struct ib_cqe *cqe, u16 slid, + u16 pkey_index, u8 port_num, struct ib_wc *wc) +{ + memset(wc, 0, sizeof *wc); + wc->wr_cqe = cqe; + wc->status = IB_WC_SUCCESS; + wc->opcode = IB_WC_RECV; + wc->pkey_index = pkey_index; + wc->byte_len = sizeof(struct ib_mad) + sizeof(struct ib_grh); + wc->src_qp = IB_QP0; + wc->qp = qp; + wc->slid = slid; + wc->sl = 0; + wc->dlid_path_bits = 0; + wc->port_num = port_num; +} + +static size_t mad_priv_size(const struct ib_mad_private *mp) +{ + return sizeof(struct ib_mad_private) + mp->mad_size; +} + +static struct ib_mad_private *alloc_mad_private(size_t mad_size, gfp_t flags) +{ + size_t size = sizeof(struct ib_mad_private) + mad_size; + struct ib_mad_private *ret = kzalloc(size, flags); + + if (ret) + ret->mad_size = mad_size; + + return ret; +} + +static size_t port_mad_size(const struct ib_mad_port_private *port_priv) +{ + return rdma_max_mad_size(port_priv->device, port_priv->port_num); +} + +static size_t mad_priv_dma_size(const struct ib_mad_private *mp) +{ + return sizeof(struct ib_grh) + mp->mad_size; +} + +/* + * Return 0 if SMP is to be sent + * Return 1 if SMP was consumed locally (whether or not solicited) + * Return < 0 if error + */ +static int handle_outgoing_dr_smp(struct ib_mad_agent_private *mad_agent_priv, + struct ib_mad_send_wr_private *mad_send_wr) +{ + int ret = 0; + struct ib_smp *smp = mad_send_wr->send_buf.mad; + struct opa_smp *opa_smp = (struct opa_smp *)smp; + unsigned long flags; + struct ib_mad_local_private *local; + struct ib_mad_private *mad_priv; + struct ib_mad_port_private *port_priv; + struct ib_mad_agent_private *recv_mad_agent = NULL; + struct ib_device *device = mad_agent_priv->agent.device; + u8 port_num; + struct ib_wc mad_wc; + struct ib_ud_wr *send_wr = &mad_send_wr->send_wr; + size_t mad_size = port_mad_size(mad_agent_priv->qp_info->port_priv); + u16 out_mad_pkey_index = 0; + u16 drslid; + bool opa = rdma_cap_opa_mad(mad_agent_priv->qp_info->port_priv->device, + mad_agent_priv->qp_info->port_priv->port_num); + + if (rdma_cap_ib_switch(device) && + smp->mgmt_class == IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE) + port_num = send_wr->port_num; + else + port_num = mad_agent_priv->agent.port_num; + + /* + * Directed route handling starts if the initial LID routed part of + * a request or the ending LID routed part of a response is empty. + * If we are at the start of the LID routed part, don't update the + * hop_ptr or hop_cnt. See section 14.2.2, Vol 1 IB spec. + */ + if (opa && smp->class_version == OPA_SM_CLASS_VERSION) { + u32 opa_drslid; + + trace_ib_mad_handle_out_opa_smi(opa_smp); + + if ((opa_get_smp_direction(opa_smp) + ? opa_smp->route.dr.dr_dlid : opa_smp->route.dr.dr_slid) == + OPA_LID_PERMISSIVE && + opa_smi_handle_dr_smp_send(opa_smp, + rdma_cap_ib_switch(device), + port_num) == IB_SMI_DISCARD) { + ret = -EINVAL; + dev_err(&device->dev, "OPA Invalid directed route\n"); + goto out; + } + opa_drslid = be32_to_cpu(opa_smp->route.dr.dr_slid); + if (opa_drslid != be32_to_cpu(OPA_LID_PERMISSIVE) && + opa_drslid & 0xffff0000) { + ret = -EINVAL; + dev_err(&device->dev, "OPA Invalid dr_slid 0x%x\n", + opa_drslid); + goto out; + } + drslid = (u16)(opa_drslid & 0x0000ffff); + + /* Check to post send on QP or process locally */ + if (opa_smi_check_local_smp(opa_smp, device) == IB_SMI_DISCARD && + opa_smi_check_local_returning_smp(opa_smp, device) == IB_SMI_DISCARD) + goto out; + } else { + trace_ib_mad_handle_out_ib_smi(smp); + + if ((ib_get_smp_direction(smp) ? smp->dr_dlid : smp->dr_slid) == + IB_LID_PERMISSIVE && + smi_handle_dr_smp_send(smp, rdma_cap_ib_switch(device), port_num) == + IB_SMI_DISCARD) { + ret = -EINVAL; + dev_err(&device->dev, "Invalid directed route\n"); + goto out; + } + drslid = be16_to_cpu(smp->dr_slid); + + /* Check to post send on QP or process locally */ + if (smi_check_local_smp(smp, device) == IB_SMI_DISCARD && + smi_check_local_returning_smp(smp, device) == IB_SMI_DISCARD) + goto out; + } + + local = kmalloc(sizeof *local, GFP_ATOMIC); + if (!local) { + ret = -ENOMEM; + goto out; + } + local->mad_priv = NULL; + local->recv_mad_agent = NULL; + mad_priv = alloc_mad_private(mad_size, GFP_ATOMIC); + if (!mad_priv) { + ret = -ENOMEM; + kfree(local); + goto out; + } + + build_smp_wc(mad_agent_priv->agent.qp, + send_wr->wr.wr_cqe, drslid, + send_wr->pkey_index, + send_wr->port_num, &mad_wc); + + if (opa && smp->base_version == OPA_MGMT_BASE_VERSION) { + mad_wc.byte_len = mad_send_wr->send_buf.hdr_len + + mad_send_wr->send_buf.data_len + + sizeof(struct ib_grh); + } + + /* No GRH for DR SMP */ + ret = device->ops.process_mad(device, 0, port_num, &mad_wc, NULL, + (const struct ib_mad *)smp, + (struct ib_mad *)mad_priv->mad, &mad_size, + &out_mad_pkey_index); + switch (ret) + { + case IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_REPLY: + if (ib_response_mad((const struct ib_mad_hdr *)mad_priv->mad) && + mad_agent_priv->agent.recv_handler) { + local->mad_priv = mad_priv; + local->recv_mad_agent = mad_agent_priv; + /* + * Reference MAD agent until receive + * side of local completion handled + */ + refcount_inc(&mad_agent_priv->refcount); + } else + kfree(mad_priv); + break; + case IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_CONSUMED: + kfree(mad_priv); + break; + case IB_MAD_RESULT_SUCCESS: + /* Treat like an incoming receive MAD */ + port_priv = ib_get_mad_port(mad_agent_priv->agent.device, + mad_agent_priv->agent.port_num); + if (port_priv) { + memcpy(mad_priv->mad, smp, mad_priv->mad_size); + recv_mad_agent = find_mad_agent(port_priv, + (const struct ib_mad_hdr *)mad_priv->mad); + } + if (!port_priv || !recv_mad_agent) { + /* + * No receiving agent so drop packet and + * generate send completion. + */ + kfree(mad_priv); + break; + } + local->mad_priv = mad_priv; + local->recv_mad_agent = recv_mad_agent; + break; + default: + kfree(mad_priv); + kfree(local); + ret = -EINVAL; + goto out; + } + + local->mad_send_wr = mad_send_wr; + if (opa) { + local->mad_send_wr->send_wr.pkey_index = out_mad_pkey_index; + local->return_wc_byte_len = mad_size; + } + /* Reference MAD agent until send side of local completion handled */ + refcount_inc(&mad_agent_priv->refcount); + /* Queue local completion to local list */ + spin_lock_irqsave(&mad_agent_priv->lock, flags); + list_add_tail(&local->completion_list, &mad_agent_priv->local_list); + spin_unlock_irqrestore(&mad_agent_priv->lock, flags); + queue_work(mad_agent_priv->qp_info->port_priv->wq, + &mad_agent_priv->local_work); + ret = 1; +out: + return ret; +} + +static int get_pad_size(int hdr_len, int data_len, size_t mad_size) +{ + int seg_size, pad; + + seg_size = mad_size - hdr_len; + if (data_len && seg_size) { + pad = seg_size - data_len % seg_size; + return pad == seg_size ? 0 : pad; + } else + return seg_size; +} + +static void free_send_rmpp_list(struct ib_mad_send_wr_private *mad_send_wr) +{ + struct ib_rmpp_segment *s, *t; + + list_for_each_entry_safe(s, t, &mad_send_wr->rmpp_list, list) { + list_del(&s->list); + kfree(s); + } +} + +static int alloc_send_rmpp_list(struct ib_mad_send_wr_private *send_wr, + size_t mad_size, gfp_t gfp_mask) +{ + struct ib_mad_send_buf *send_buf = &send_wr->send_buf; + struct ib_rmpp_mad *rmpp_mad = send_buf->mad; + struct ib_rmpp_segment *seg = NULL; + int left, seg_size, pad; + + send_buf->seg_size = mad_size - send_buf->hdr_len; + send_buf->seg_rmpp_size = mad_size - IB_MGMT_RMPP_HDR; + seg_size = send_buf->seg_size; + pad = send_wr->pad; + + /* Allocate data segments. */ + for (left = send_buf->data_len + pad; left > 0; left -= seg_size) { + seg = kmalloc(sizeof (*seg) + seg_size, gfp_mask); + if (!seg) { + free_send_rmpp_list(send_wr); + return -ENOMEM; + } + seg->num = ++send_buf->seg_count; + list_add_tail(&seg->list, &send_wr->rmpp_list); + } + + /* Zero any padding */ + if (pad) + memset(seg->data + seg_size - pad, 0, pad); + + rmpp_mad->rmpp_hdr.rmpp_version = send_wr->mad_agent_priv-> + agent.rmpp_version; + rmpp_mad->rmpp_hdr.rmpp_type = IB_MGMT_RMPP_TYPE_DATA; + ib_set_rmpp_flags(&rmpp_mad->rmpp_hdr, IB_MGMT_RMPP_FLAG_ACTIVE); + + send_wr->cur_seg = container_of(send_wr->rmpp_list.next, + struct ib_rmpp_segment, list); + send_wr->last_ack_seg = send_wr->cur_seg; + return 0; +} + +int ib_mad_kernel_rmpp_agent(const struct ib_mad_agent *agent) +{ + return agent->rmpp_version && !(agent->flags & IB_MAD_USER_RMPP); +} +EXPORT_SYMBOL(ib_mad_kernel_rmpp_agent); + +struct ib_mad_send_buf * ib_create_send_mad(struct ib_mad_agent *mad_agent, + u32 remote_qpn, u16 pkey_index, + int rmpp_active, + int hdr_len, int data_len, + gfp_t gfp_mask, + u8 base_version) +{ + struct ib_mad_agent_private *mad_agent_priv; + struct ib_mad_send_wr_private *mad_send_wr; + int pad, message_size, ret, size; + void *buf; + size_t mad_size; + bool opa; + + mad_agent_priv = container_of(mad_agent, struct ib_mad_agent_private, + agent); + + opa = rdma_cap_opa_mad(mad_agent->device, mad_agent->port_num); + + if (opa && base_version == OPA_MGMT_BASE_VERSION) + mad_size = sizeof(struct opa_mad); + else + mad_size = sizeof(struct ib_mad); + + pad = get_pad_size(hdr_len, data_len, mad_size); + message_size = hdr_len + data_len + pad; + + if (ib_mad_kernel_rmpp_agent(mad_agent)) { + if (!rmpp_active && message_size > mad_size) + return ERR_PTR(-EINVAL); + } else + if (rmpp_active || message_size > mad_size) + return ERR_PTR(-EINVAL); + + size = rmpp_active ? hdr_len : mad_size; + buf = kzalloc(sizeof *mad_send_wr + size, gfp_mask); + if (!buf) + return ERR_PTR(-ENOMEM); + + mad_send_wr = buf + size; + INIT_LIST_HEAD(&mad_send_wr->rmpp_list); + mad_send_wr->send_buf.mad = buf; + mad_send_wr->send_buf.hdr_len = hdr_len; + mad_send_wr->send_buf.data_len = data_len; + mad_send_wr->pad = pad; + + mad_send_wr->mad_agent_priv = mad_agent_priv; + mad_send_wr->sg_list[0].length = hdr_len; + mad_send_wr->sg_list[0].lkey = mad_agent->qp->pd->local_dma_lkey; + + /* OPA MADs don't have to be the full 2048 bytes */ + if (opa && base_version == OPA_MGMT_BASE_VERSION && + data_len < mad_size - hdr_len) + mad_send_wr->sg_list[1].length = data_len; + else + mad_send_wr->sg_list[1].length = mad_size - hdr_len; + + mad_send_wr->sg_list[1].lkey = mad_agent->qp->pd->local_dma_lkey; + + mad_send_wr->mad_list.cqe.done = ib_mad_send_done; + + mad_send_wr->send_wr.wr.wr_cqe = &mad_send_wr->mad_list.cqe; + mad_send_wr->send_wr.wr.sg_list = mad_send_wr->sg_list; + mad_send_wr->send_wr.wr.num_sge = 2; + mad_send_wr->send_wr.wr.opcode = IB_WR_SEND; + mad_send_wr->send_wr.wr.send_flags = IB_SEND_SIGNALED; + mad_send_wr->send_wr.remote_qpn = remote_qpn; + mad_send_wr->send_wr.remote_qkey = IB_QP_SET_QKEY; + mad_send_wr->send_wr.pkey_index = pkey_index; + + if (rmpp_active) { + ret = alloc_send_rmpp_list(mad_send_wr, mad_size, gfp_mask); + if (ret) { + kfree(buf); + return ERR_PTR(ret); + } + } + + mad_send_wr->send_buf.mad_agent = mad_agent; + refcount_inc(&mad_agent_priv->refcount); + return &mad_send_wr->send_buf; +} +EXPORT_SYMBOL(ib_create_send_mad); + +int ib_get_mad_data_offset(u8 mgmt_class) +{ + if (mgmt_class == IB_MGMT_CLASS_SUBN_ADM) + return IB_MGMT_SA_HDR; + else if ((mgmt_class == IB_MGMT_CLASS_DEVICE_MGMT) || + (mgmt_class == IB_MGMT_CLASS_DEVICE_ADM) || + (mgmt_class == IB_MGMT_CLASS_BIS)) + return IB_MGMT_DEVICE_HDR; + else if ((mgmt_class >= IB_MGMT_CLASS_VENDOR_RANGE2_START) && + (mgmt_class <= IB_MGMT_CLASS_VENDOR_RANGE2_END)) + return IB_MGMT_VENDOR_HDR; + else + return IB_MGMT_MAD_HDR; +} +EXPORT_SYMBOL(ib_get_mad_data_offset); + +int ib_is_mad_class_rmpp(u8 mgmt_class) +{ + if ((mgmt_class == IB_MGMT_CLASS_SUBN_ADM) || + (mgmt_class == IB_MGMT_CLASS_DEVICE_MGMT) || + (mgmt_class == IB_MGMT_CLASS_DEVICE_ADM) || + (mgmt_class == IB_MGMT_CLASS_BIS) || + ((mgmt_class >= IB_MGMT_CLASS_VENDOR_RANGE2_START) && + (mgmt_class <= IB_MGMT_CLASS_VENDOR_RANGE2_END))) + return 1; + return 0; +} +EXPORT_SYMBOL(ib_is_mad_class_rmpp); + +void *ib_get_rmpp_segment(struct ib_mad_send_buf *send_buf, int seg_num) +{ + struct ib_mad_send_wr_private *mad_send_wr; + struct list_head *list; + + mad_send_wr = container_of(send_buf, struct ib_mad_send_wr_private, + send_buf); + list = &mad_send_wr->cur_seg->list; + + if (mad_send_wr->cur_seg->num < seg_num) { + list_for_each_entry(mad_send_wr->cur_seg, list, list) + if (mad_send_wr->cur_seg->num == seg_num) + break; + } else if (mad_send_wr->cur_seg->num > seg_num) { + list_for_each_entry_reverse(mad_send_wr->cur_seg, list, list) + if (mad_send_wr->cur_seg->num == seg_num) + break; + } + return mad_send_wr->cur_seg->data; +} +EXPORT_SYMBOL(ib_get_rmpp_segment); + +static inline void *ib_get_payload(struct ib_mad_send_wr_private *mad_send_wr) +{ + if (mad_send_wr->send_buf.seg_count) + return ib_get_rmpp_segment(&mad_send_wr->send_buf, + mad_send_wr->seg_num); + else + return mad_send_wr->send_buf.mad + + mad_send_wr->send_buf.hdr_len; +} + +void ib_free_send_mad(struct ib_mad_send_buf *send_buf) +{ + struct ib_mad_agent_private *mad_agent_priv; + struct ib_mad_send_wr_private *mad_send_wr; + + mad_agent_priv = container_of(send_buf->mad_agent, + struct ib_mad_agent_private, agent); + mad_send_wr = container_of(send_buf, struct ib_mad_send_wr_private, + send_buf); + + free_send_rmpp_list(mad_send_wr); + kfree(send_buf->mad); + deref_mad_agent(mad_agent_priv); +} +EXPORT_SYMBOL(ib_free_send_mad); + +int ib_send_mad(struct ib_mad_send_wr_private *mad_send_wr) +{ + struct ib_mad_qp_info *qp_info; + struct list_head *list; + struct ib_mad_agent *mad_agent; + struct ib_sge *sge; + unsigned long flags; + int ret; + + /* Set WR ID to find mad_send_wr upon completion */ + qp_info = mad_send_wr->mad_agent_priv->qp_info; + mad_send_wr->mad_list.mad_queue = &qp_info->send_queue; + mad_send_wr->mad_list.cqe.done = ib_mad_send_done; + mad_send_wr->send_wr.wr.wr_cqe = &mad_send_wr->mad_list.cqe; + + mad_agent = mad_send_wr->send_buf.mad_agent; + sge = mad_send_wr->sg_list; + sge[0].addr = ib_dma_map_single(mad_agent->device, + mad_send_wr->send_buf.mad, + sge[0].length, + DMA_TO_DEVICE); + if (unlikely(ib_dma_mapping_error(mad_agent->device, sge[0].addr))) + return -ENOMEM; + + mad_send_wr->header_mapping = sge[0].addr; + + sge[1].addr = ib_dma_map_single(mad_agent->device, + ib_get_payload(mad_send_wr), + sge[1].length, + DMA_TO_DEVICE); + if (unlikely(ib_dma_mapping_error(mad_agent->device, sge[1].addr))) { + ib_dma_unmap_single(mad_agent->device, + mad_send_wr->header_mapping, + sge[0].length, DMA_TO_DEVICE); + return -ENOMEM; + } + mad_send_wr->payload_mapping = sge[1].addr; + + spin_lock_irqsave(&qp_info->send_queue.lock, flags); + if (qp_info->send_queue.count < qp_info->send_queue.max_active) { + trace_ib_mad_ib_send_mad(mad_send_wr, qp_info); + ret = ib_post_send(mad_agent->qp, &mad_send_wr->send_wr.wr, + NULL); + list = &qp_info->send_queue.list; + } else { + ret = 0; + list = &qp_info->overflow_list; + } + + if (!ret) { + qp_info->send_queue.count++; + list_add_tail(&mad_send_wr->mad_list.list, list); + } + spin_unlock_irqrestore(&qp_info->send_queue.lock, flags); + if (ret) { + ib_dma_unmap_single(mad_agent->device, + mad_send_wr->header_mapping, + sge[0].length, DMA_TO_DEVICE); + ib_dma_unmap_single(mad_agent->device, + mad_send_wr->payload_mapping, + sge[1].length, DMA_TO_DEVICE); + } + return ret; +} + +/* + * ib_post_send_mad - Posts MAD(s) to the send queue of the QP associated + * with the registered client + */ +int ib_post_send_mad(struct ib_mad_send_buf *send_buf, + struct ib_mad_send_buf **bad_send_buf) +{ + struct ib_mad_agent_private *mad_agent_priv; + struct ib_mad_send_buf *next_send_buf; + struct ib_mad_send_wr_private *mad_send_wr; + unsigned long flags; + int ret = -EINVAL; + + /* Walk list of send WRs and post each on send list */ + for (; send_buf; send_buf = next_send_buf) { + mad_send_wr = container_of(send_buf, + struct ib_mad_send_wr_private, + send_buf); + mad_agent_priv = mad_send_wr->mad_agent_priv; + + ret = ib_mad_enforce_security(mad_agent_priv, + mad_send_wr->send_wr.pkey_index); + if (ret) + goto error; + + if (!send_buf->mad_agent->send_handler || + (send_buf->timeout_ms && + !send_buf->mad_agent->recv_handler)) { + ret = -EINVAL; + goto error; + } + + if (!ib_is_mad_class_rmpp(((struct ib_mad_hdr *) send_buf->mad)->mgmt_class)) { + if (mad_agent_priv->agent.rmpp_version) { + ret = -EINVAL; + goto error; + } + } + + /* + * Save pointer to next work request to post in case the + * current one completes, and the user modifies the work + * request associated with the completion + */ + next_send_buf = send_buf->next; + mad_send_wr->send_wr.ah = send_buf->ah; + + if (((struct ib_mad_hdr *) send_buf->mad)->mgmt_class == + IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE) { + ret = handle_outgoing_dr_smp(mad_agent_priv, + mad_send_wr); + if (ret < 0) /* error */ + goto error; + else if (ret == 1) /* locally consumed */ + continue; + } + + mad_send_wr->tid = ((struct ib_mad_hdr *) send_buf->mad)->tid; + /* Timeout will be updated after send completes */ + mad_send_wr->timeout = msecs_to_jiffies(send_buf->timeout_ms); + mad_send_wr->max_retries = send_buf->retries; + mad_send_wr->retries_left = send_buf->retries; + send_buf->retries = 0; + /* Reference for work request to QP + response */ + mad_send_wr->refcount = 1 + (mad_send_wr->timeout > 0); + mad_send_wr->status = IB_WC_SUCCESS; + + /* Reference MAD agent until send completes */ + refcount_inc(&mad_agent_priv->refcount); + spin_lock_irqsave(&mad_agent_priv->lock, flags); + list_add_tail(&mad_send_wr->agent_list, + &mad_agent_priv->send_list); + spin_unlock_irqrestore(&mad_agent_priv->lock, flags); + + if (ib_mad_kernel_rmpp_agent(&mad_agent_priv->agent)) { + ret = ib_send_rmpp_mad(mad_send_wr); + if (ret >= 0 && ret != IB_RMPP_RESULT_CONSUMED) + ret = ib_send_mad(mad_send_wr); + } else + ret = ib_send_mad(mad_send_wr); + if (ret < 0) { + /* Fail send request */ + spin_lock_irqsave(&mad_agent_priv->lock, flags); + list_del(&mad_send_wr->agent_list); + spin_unlock_irqrestore(&mad_agent_priv->lock, flags); + deref_mad_agent(mad_agent_priv); + goto error; + } + } + return 0; +error: + if (bad_send_buf) + *bad_send_buf = send_buf; + return ret; +} +EXPORT_SYMBOL(ib_post_send_mad); + +/* + * ib_free_recv_mad - Returns data buffers used to receive + * a MAD to the access layer + */ +void ib_free_recv_mad(struct ib_mad_recv_wc *mad_recv_wc) +{ + struct ib_mad_recv_buf *mad_recv_buf, *temp_recv_buf; + struct ib_mad_private_header *mad_priv_hdr; + struct ib_mad_private *priv; + struct list_head free_list; + + INIT_LIST_HEAD(&free_list); + list_splice_init(&mad_recv_wc->rmpp_list, &free_list); + + list_for_each_entry_safe(mad_recv_buf, temp_recv_buf, + &free_list, list) { + mad_recv_wc = container_of(mad_recv_buf, struct ib_mad_recv_wc, + recv_buf); + mad_priv_hdr = container_of(mad_recv_wc, + struct ib_mad_private_header, + recv_wc); + priv = container_of(mad_priv_hdr, struct ib_mad_private, + header); + kfree(priv); + } +} +EXPORT_SYMBOL(ib_free_recv_mad); + +static int method_in_use(struct ib_mad_mgmt_method_table **method, + struct ib_mad_reg_req *mad_reg_req) +{ + int i; + + for_each_set_bit(i, mad_reg_req->method_mask, IB_MGMT_MAX_METHODS) { + if ((*method)->agent[i]) { + pr_err("Method %d already in use\n", i); + return -EINVAL; + } + } + return 0; +} + +static int allocate_method_table(struct ib_mad_mgmt_method_table **method) +{ + /* Allocate management method table */ + *method = kzalloc(sizeof **method, GFP_ATOMIC); + return (*method) ? 0 : (-ENOMEM); +} + +/* + * Check to see if there are any methods still in use + */ +static int check_method_table(struct ib_mad_mgmt_method_table *method) +{ + int i; + + for (i = 0; i < IB_MGMT_MAX_METHODS; i++) + if (method->agent[i]) + return 1; + return 0; +} + +/* + * Check to see if there are any method tables for this class still in use + */ +static int check_class_table(struct ib_mad_mgmt_class_table *class) +{ + int i; + + for (i = 0; i < MAX_MGMT_CLASS; i++) + if (class->method_table[i]) + return 1; + return 0; +} + +static int check_vendor_class(struct ib_mad_mgmt_vendor_class *vendor_class) +{ + int i; + + for (i = 0; i < MAX_MGMT_OUI; i++) + if (vendor_class->method_table[i]) + return 1; + return 0; +} + +static int find_vendor_oui(struct ib_mad_mgmt_vendor_class *vendor_class, + const char *oui) +{ + int i; + + for (i = 0; i < MAX_MGMT_OUI; i++) + /* Is there matching OUI for this vendor class ? */ + if (!memcmp(vendor_class->oui[i], oui, 3)) + return i; + + return -1; +} + +static int check_vendor_table(struct ib_mad_mgmt_vendor_class_table *vendor) +{ + int i; + + for (i = 0; i < MAX_MGMT_VENDOR_RANGE2; i++) + if (vendor->vendor_class[i]) + return 1; + + return 0; +} + +static void remove_methods_mad_agent(struct ib_mad_mgmt_method_table *method, + struct ib_mad_agent_private *agent) +{ + int i; + + /* Remove any methods for this mad agent */ + for (i = 0; i < IB_MGMT_MAX_METHODS; i++) { + if (method->agent[i] == agent) { + method->agent[i] = NULL; + } + } +} + +static int add_nonoui_reg_req(struct ib_mad_reg_req *mad_reg_req, + struct ib_mad_agent_private *agent_priv, + u8 mgmt_class) +{ + struct ib_mad_port_private *port_priv; + struct ib_mad_mgmt_class_table **class; + struct ib_mad_mgmt_method_table **method; + int i, ret; + + port_priv = agent_priv->qp_info->port_priv; + class = &port_priv->version[mad_reg_req->mgmt_class_version].class; + if (!*class) { + /* Allocate management class table for "new" class version */ + *class = kzalloc(sizeof **class, GFP_ATOMIC); + if (!*class) { + ret = -ENOMEM; + goto error1; + } + + /* Allocate method table for this management class */ + method = &(*class)->method_table[mgmt_class]; + if ((ret = allocate_method_table(method))) + goto error2; + } else { + method = &(*class)->method_table[mgmt_class]; + if (!*method) { + /* Allocate method table for this management class */ + if ((ret = allocate_method_table(method))) + goto error1; + } + } + + /* Now, make sure methods are not already in use */ + if (method_in_use(method, mad_reg_req)) + goto error3; + + /* Finally, add in methods being registered */ + for_each_set_bit(i, mad_reg_req->method_mask, IB_MGMT_MAX_METHODS) + (*method)->agent[i] = agent_priv; + + return 0; + +error3: + /* Remove any methods for this mad agent */ + remove_methods_mad_agent(*method, agent_priv); + /* Now, check to see if there are any methods in use */ + if (!check_method_table(*method)) { + /* If not, release management method table */ + kfree(*method); + *method = NULL; + } + ret = -EINVAL; + goto error1; +error2: + kfree(*class); + *class = NULL; +error1: + return ret; +} + +static int add_oui_reg_req(struct ib_mad_reg_req *mad_reg_req, + struct ib_mad_agent_private *agent_priv) +{ + struct ib_mad_port_private *port_priv; + struct ib_mad_mgmt_vendor_class_table **vendor_table; + struct ib_mad_mgmt_vendor_class_table *vendor = NULL; + struct ib_mad_mgmt_vendor_class *vendor_class = NULL; + struct ib_mad_mgmt_method_table **method; + int i, ret = -ENOMEM; + u8 vclass; + + /* "New" vendor (with OUI) class */ + vclass = vendor_class_index(mad_reg_req->mgmt_class); + port_priv = agent_priv->qp_info->port_priv; + vendor_table = &port_priv->version[ + mad_reg_req->mgmt_class_version].vendor; + if (!*vendor_table) { + /* Allocate mgmt vendor class table for "new" class version */ + vendor = kzalloc(sizeof *vendor, GFP_ATOMIC); + if (!vendor) + goto error1; + + *vendor_table = vendor; + } + if (!(*vendor_table)->vendor_class[vclass]) { + /* Allocate table for this management vendor class */ + vendor_class = kzalloc(sizeof *vendor_class, GFP_ATOMIC); + if (!vendor_class) + goto error2; + + (*vendor_table)->vendor_class[vclass] = vendor_class; + } + for (i = 0; i < MAX_MGMT_OUI; i++) { + /* Is there matching OUI for this vendor class ? */ + if (!memcmp((*vendor_table)->vendor_class[vclass]->oui[i], + mad_reg_req->oui, 3)) { + method = &(*vendor_table)->vendor_class[ + vclass]->method_table[i]; + if (!*method) + goto error3; + goto check_in_use; + } + } + for (i = 0; i < MAX_MGMT_OUI; i++) { + /* OUI slot available ? */ + if (!is_vendor_oui((*vendor_table)->vendor_class[ + vclass]->oui[i])) { + method = &(*vendor_table)->vendor_class[ + vclass]->method_table[i]; + /* Allocate method table for this OUI */ + if (!*method) { + ret = allocate_method_table(method); + if (ret) + goto error3; + } + memcpy((*vendor_table)->vendor_class[vclass]->oui[i], + mad_reg_req->oui, 3); + goto check_in_use; + } + } + dev_err(&agent_priv->agent.device->dev, "All OUI slots in use\n"); + goto error3; + +check_in_use: + /* Now, make sure methods are not already in use */ + if (method_in_use(method, mad_reg_req)) + goto error4; + + /* Finally, add in methods being registered */ + for_each_set_bit(i, mad_reg_req->method_mask, IB_MGMT_MAX_METHODS) + (*method)->agent[i] = agent_priv; + + return 0; + +error4: + /* Remove any methods for this mad agent */ + remove_methods_mad_agent(*method, agent_priv); + /* Now, check to see if there are any methods in use */ + if (!check_method_table(*method)) { + /* If not, release management method table */ + kfree(*method); + *method = NULL; + } + ret = -EINVAL; +error3: + if (vendor_class) { + (*vendor_table)->vendor_class[vclass] = NULL; + kfree(vendor_class); + } +error2: + if (vendor) { + *vendor_table = NULL; + kfree(vendor); + } +error1: + return ret; +} + +static void remove_mad_reg_req(struct ib_mad_agent_private *agent_priv) +{ + struct ib_mad_port_private *port_priv; + struct ib_mad_mgmt_class_table *class; + struct ib_mad_mgmt_method_table *method; + struct ib_mad_mgmt_vendor_class_table *vendor; + struct ib_mad_mgmt_vendor_class *vendor_class; + int index; + u8 mgmt_class; + + /* + * Was MAD registration request supplied + * with original registration ? + */ + if (!agent_priv->reg_req) { + goto out; + } + + port_priv = agent_priv->qp_info->port_priv; + mgmt_class = convert_mgmt_class(agent_priv->reg_req->mgmt_class); + class = port_priv->version[ + agent_priv->reg_req->mgmt_class_version].class; + if (!class) + goto vendor_check; + + method = class->method_table[mgmt_class]; + if (method) { + /* Remove any methods for this mad agent */ + remove_methods_mad_agent(method, agent_priv); + /* Now, check to see if there are any methods still in use */ + if (!check_method_table(method)) { + /* If not, release management method table */ + kfree(method); + class->method_table[mgmt_class] = NULL; + /* Any management classes left ? */ + if (!check_class_table(class)) { + /* If not, release management class table */ + kfree(class); + port_priv->version[ + agent_priv->reg_req-> + mgmt_class_version].class = NULL; + } + } + } + +vendor_check: + if (!is_vendor_class(mgmt_class)) + goto out; + + /* normalize mgmt_class to vendor range 2 */ + mgmt_class = vendor_class_index(agent_priv->reg_req->mgmt_class); + vendor = port_priv->version[ + agent_priv->reg_req->mgmt_class_version].vendor; + + if (!vendor) + goto out; + + vendor_class = vendor->vendor_class[mgmt_class]; + if (vendor_class) { + index = find_vendor_oui(vendor_class, agent_priv->reg_req->oui); + if (index < 0) + goto out; + method = vendor_class->method_table[index]; + if (method) { + /* Remove any methods for this mad agent */ + remove_methods_mad_agent(method, agent_priv); + /* + * Now, check to see if there are + * any methods still in use + */ + if (!check_method_table(method)) { + /* If not, release management method table */ + kfree(method); + vendor_class->method_table[index] = NULL; + memset(vendor_class->oui[index], 0, 3); + /* Any OUIs left ? */ + if (!check_vendor_class(vendor_class)) { + /* If not, release vendor class table */ + kfree(vendor_class); + vendor->vendor_class[mgmt_class] = NULL; + /* Any other vendor classes left ? */ + if (!check_vendor_table(vendor)) { + kfree(vendor); + port_priv->version[ + agent_priv->reg_req-> + mgmt_class_version]. + vendor = NULL; + } + } + } + } + } + +out: + return; +} + +static struct ib_mad_agent_private * +find_mad_agent(struct ib_mad_port_private *port_priv, + const struct ib_mad_hdr *mad_hdr) +{ + struct ib_mad_agent_private *mad_agent = NULL; + unsigned long flags; + + if (ib_response_mad(mad_hdr)) { + u32 hi_tid; + + /* + * Routing is based on high 32 bits of transaction ID + * of MAD. + */ + hi_tid = be64_to_cpu(mad_hdr->tid) >> 32; + rcu_read_lock(); + mad_agent = xa_load(&ib_mad_clients, hi_tid); + if (mad_agent && !refcount_inc_not_zero(&mad_agent->refcount)) + mad_agent = NULL; + rcu_read_unlock(); + } else { + struct ib_mad_mgmt_class_table *class; + struct ib_mad_mgmt_method_table *method; + struct ib_mad_mgmt_vendor_class_table *vendor; + struct ib_mad_mgmt_vendor_class *vendor_class; + const struct ib_vendor_mad *vendor_mad; + int index; + + spin_lock_irqsave(&port_priv->reg_lock, flags); + /* + * Routing is based on version, class, and method + * For "newer" vendor MADs, also based on OUI + */ + if (mad_hdr->class_version >= MAX_MGMT_VERSION) + goto out; + if (!is_vendor_class(mad_hdr->mgmt_class)) { + class = port_priv->version[ + mad_hdr->class_version].class; + if (!class) + goto out; + if (convert_mgmt_class(mad_hdr->mgmt_class) >= + ARRAY_SIZE(class->method_table)) + goto out; + method = class->method_table[convert_mgmt_class( + mad_hdr->mgmt_class)]; + if (method) + mad_agent = method->agent[mad_hdr->method & + ~IB_MGMT_METHOD_RESP]; + } else { + vendor = port_priv->version[ + mad_hdr->class_version].vendor; + if (!vendor) + goto out; + vendor_class = vendor->vendor_class[vendor_class_index( + mad_hdr->mgmt_class)]; + if (!vendor_class) + goto out; + /* Find matching OUI */ + vendor_mad = (const struct ib_vendor_mad *)mad_hdr; + index = find_vendor_oui(vendor_class, vendor_mad->oui); + if (index == -1) + goto out; + method = vendor_class->method_table[index]; + if (method) { + mad_agent = method->agent[mad_hdr->method & + ~IB_MGMT_METHOD_RESP]; + } + } + if (mad_agent) + refcount_inc(&mad_agent->refcount); +out: + spin_unlock_irqrestore(&port_priv->reg_lock, flags); + } + + if (mad_agent && !mad_agent->agent.recv_handler) { + dev_notice(&port_priv->device->dev, + "No receive handler for client %p on port %d\n", + &mad_agent->agent, port_priv->port_num); + deref_mad_agent(mad_agent); + mad_agent = NULL; + } + + return mad_agent; +} + +static int validate_mad(const struct ib_mad_hdr *mad_hdr, + const struct ib_mad_qp_info *qp_info, + bool opa) +{ + int valid = 0; + u32 qp_num = qp_info->qp->qp_num; + + /* Make sure MAD base version is understood */ + if (mad_hdr->base_version != IB_MGMT_BASE_VERSION && + (!opa || mad_hdr->base_version != OPA_MGMT_BASE_VERSION)) { + pr_err("MAD received with unsupported base version %d %s\n", + mad_hdr->base_version, opa ? "(opa)" : ""); + goto out; + } + + /* Filter SMI packets sent to other than QP0 */ + if ((mad_hdr->mgmt_class == IB_MGMT_CLASS_SUBN_LID_ROUTED) || + (mad_hdr->mgmt_class == IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE)) { + if (qp_num == 0) + valid = 1; + } else { + /* CM attributes other than ClassPortInfo only use Send method */ + if ((mad_hdr->mgmt_class == IB_MGMT_CLASS_CM) && + (mad_hdr->attr_id != IB_MGMT_CLASSPORTINFO_ATTR_ID) && + (mad_hdr->method != IB_MGMT_METHOD_SEND)) + goto out; + /* Filter GSI packets sent to QP0 */ + if (qp_num != 0) + valid = 1; + } + +out: + return valid; +} + +static int is_rmpp_data_mad(const struct ib_mad_agent_private *mad_agent_priv, + const struct ib_mad_hdr *mad_hdr) +{ + struct ib_rmpp_mad *rmpp_mad; + + rmpp_mad = (struct ib_rmpp_mad *)mad_hdr; + return !mad_agent_priv->agent.rmpp_version || + !ib_mad_kernel_rmpp_agent(&mad_agent_priv->agent) || + !(ib_get_rmpp_flags(&rmpp_mad->rmpp_hdr) & + IB_MGMT_RMPP_FLAG_ACTIVE) || + (rmpp_mad->rmpp_hdr.rmpp_type == IB_MGMT_RMPP_TYPE_DATA); +} + +static inline int rcv_has_same_class(const struct ib_mad_send_wr_private *wr, + const struct ib_mad_recv_wc *rwc) +{ + return ((struct ib_mad_hdr *)(wr->send_buf.mad))->mgmt_class == + rwc->recv_buf.mad->mad_hdr.mgmt_class; +} + +static inline int rcv_has_same_gid(const struct ib_mad_agent_private *mad_agent_priv, + const struct ib_mad_send_wr_private *wr, + const struct ib_mad_recv_wc *rwc ) +{ + struct rdma_ah_attr attr; + u8 send_resp, rcv_resp; + union ib_gid sgid; + struct ib_device *device = mad_agent_priv->agent.device; + u8 port_num = mad_agent_priv->agent.port_num; + u8 lmc; + bool has_grh; + + send_resp = ib_response_mad((struct ib_mad_hdr *)wr->send_buf.mad); + rcv_resp = ib_response_mad(&rwc->recv_buf.mad->mad_hdr); + + if (send_resp == rcv_resp) + /* both requests, or both responses. GIDs different */ + return 0; + + if (rdma_query_ah(wr->send_buf.ah, &attr)) + /* Assume not equal, to avoid false positives. */ + return 0; + + has_grh = !!(rdma_ah_get_ah_flags(&attr) & IB_AH_GRH); + if (has_grh != !!(rwc->wc->wc_flags & IB_WC_GRH)) + /* one has GID, other does not. Assume different */ + return 0; + + if (!send_resp && rcv_resp) { + /* is request/response. */ + if (!has_grh) { + if (ib_get_cached_lmc(device, port_num, &lmc)) + return 0; + return (!lmc || !((rdma_ah_get_path_bits(&attr) ^ + rwc->wc->dlid_path_bits) & + ((1 << lmc) - 1))); + } else { + const struct ib_global_route *grh = + rdma_ah_read_grh(&attr); + + if (rdma_query_gid(device, port_num, + grh->sgid_index, &sgid)) + return 0; + return !memcmp(sgid.raw, rwc->recv_buf.grh->dgid.raw, + 16); + } + } + + if (!has_grh) + return rdma_ah_get_dlid(&attr) == rwc->wc->slid; + else + return !memcmp(rdma_ah_read_grh(&attr)->dgid.raw, + rwc->recv_buf.grh->sgid.raw, + 16); +} + +static inline int is_direct(u8 class) +{ + return (class == IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE); +} + +struct ib_mad_send_wr_private* +ib_find_send_mad(const struct ib_mad_agent_private *mad_agent_priv, + const struct ib_mad_recv_wc *wc) +{ + struct ib_mad_send_wr_private *wr; + const struct ib_mad_hdr *mad_hdr; + + mad_hdr = &wc->recv_buf.mad->mad_hdr; + + list_for_each_entry(wr, &mad_agent_priv->wait_list, agent_list) { + if ((wr->tid == mad_hdr->tid) && + rcv_has_same_class(wr, wc) && + /* + * Don't check GID for direct routed MADs. + * These might have permissive LIDs. + */ + (is_direct(mad_hdr->mgmt_class) || + rcv_has_same_gid(mad_agent_priv, wr, wc))) + return (wr->status == IB_WC_SUCCESS) ? wr : NULL; + } + + /* + * It's possible to receive the response before we've + * been notified that the send has completed + */ + list_for_each_entry(wr, &mad_agent_priv->send_list, agent_list) { + if (is_rmpp_data_mad(mad_agent_priv, wr->send_buf.mad) && + wr->tid == mad_hdr->tid && + wr->timeout && + rcv_has_same_class(wr, wc) && + /* + * Don't check GID for direct routed MADs. + * These might have permissive LIDs. + */ + (is_direct(mad_hdr->mgmt_class) || + rcv_has_same_gid(mad_agent_priv, wr, wc))) + /* Verify request has not been canceled */ + return (wr->status == IB_WC_SUCCESS) ? wr : NULL; + } + return NULL; +} + +void ib_mark_mad_done(struct ib_mad_send_wr_private *mad_send_wr) +{ + mad_send_wr->timeout = 0; + if (mad_send_wr->refcount == 1) + list_move_tail(&mad_send_wr->agent_list, + &mad_send_wr->mad_agent_priv->done_list); +} + +static void ib_mad_complete_recv(struct ib_mad_agent_private *mad_agent_priv, + struct ib_mad_recv_wc *mad_recv_wc) +{ + struct ib_mad_send_wr_private *mad_send_wr; + struct ib_mad_send_wc mad_send_wc; + unsigned long flags; + int ret; + + INIT_LIST_HEAD(&mad_recv_wc->rmpp_list); + ret = ib_mad_enforce_security(mad_agent_priv, + mad_recv_wc->wc->pkey_index); + if (ret) { + ib_free_recv_mad(mad_recv_wc); + deref_mad_agent(mad_agent_priv); + return; + } + + list_add(&mad_recv_wc->recv_buf.list, &mad_recv_wc->rmpp_list); + if (ib_mad_kernel_rmpp_agent(&mad_agent_priv->agent)) { + mad_recv_wc = ib_process_rmpp_recv_wc(mad_agent_priv, + mad_recv_wc); + if (!mad_recv_wc) { + deref_mad_agent(mad_agent_priv); + return; + } + } + + /* Complete corresponding request */ + if (ib_response_mad(&mad_recv_wc->recv_buf.mad->mad_hdr)) { + spin_lock_irqsave(&mad_agent_priv->lock, flags); + mad_send_wr = ib_find_send_mad(mad_agent_priv, mad_recv_wc); + if (!mad_send_wr) { + spin_unlock_irqrestore(&mad_agent_priv->lock, flags); + if (!ib_mad_kernel_rmpp_agent(&mad_agent_priv->agent) + && ib_is_mad_class_rmpp(mad_recv_wc->recv_buf.mad->mad_hdr.mgmt_class) + && (ib_get_rmpp_flags(&((struct ib_rmpp_mad *)mad_recv_wc->recv_buf.mad)->rmpp_hdr) + & IB_MGMT_RMPP_FLAG_ACTIVE)) { + /* user rmpp is in effect + * and this is an active RMPP MAD + */ + mad_agent_priv->agent.recv_handler( + &mad_agent_priv->agent, NULL, + mad_recv_wc); + deref_mad_agent(mad_agent_priv); + } else { + /* not user rmpp, revert to normal behavior and + * drop the mad */ + ib_free_recv_mad(mad_recv_wc); + deref_mad_agent(mad_agent_priv); + return; + } + } else { + ib_mark_mad_done(mad_send_wr); + spin_unlock_irqrestore(&mad_agent_priv->lock, flags); + + /* Defined behavior is to complete response before request */ + mad_agent_priv->agent.recv_handler( + &mad_agent_priv->agent, + &mad_send_wr->send_buf, + mad_recv_wc); + deref_mad_agent(mad_agent_priv); + + mad_send_wc.status = IB_WC_SUCCESS; + mad_send_wc.vendor_err = 0; + mad_send_wc.send_buf = &mad_send_wr->send_buf; + ib_mad_complete_send_wr(mad_send_wr, &mad_send_wc); + } + } else { + mad_agent_priv->agent.recv_handler(&mad_agent_priv->agent, NULL, + mad_recv_wc); + deref_mad_agent(mad_agent_priv); + } + + return; +} + +static enum smi_action handle_ib_smi(const struct ib_mad_port_private *port_priv, + const struct ib_mad_qp_info *qp_info, + const struct ib_wc *wc, + int port_num, + struct ib_mad_private *recv, + struct ib_mad_private *response) +{ + enum smi_forward_action retsmi; + struct ib_smp *smp = (struct ib_smp *)recv->mad; + + trace_ib_mad_handle_ib_smi(smp); + + if (smi_handle_dr_smp_recv(smp, + rdma_cap_ib_switch(port_priv->device), + port_num, + port_priv->device->phys_port_cnt) == + IB_SMI_DISCARD) + return IB_SMI_DISCARD; + + retsmi = smi_check_forward_dr_smp(smp); + if (retsmi == IB_SMI_LOCAL) + return IB_SMI_HANDLE; + + if (retsmi == IB_SMI_SEND) { /* don't forward */ + if (smi_handle_dr_smp_send(smp, + rdma_cap_ib_switch(port_priv->device), + port_num) == IB_SMI_DISCARD) + return IB_SMI_DISCARD; + + if (smi_check_local_smp(smp, port_priv->device) == IB_SMI_DISCARD) + return IB_SMI_DISCARD; + } else if (rdma_cap_ib_switch(port_priv->device)) { + /* forward case for switches */ + memcpy(response, recv, mad_priv_size(response)); + response->header.recv_wc.wc = &response->header.wc; + response->header.recv_wc.recv_buf.mad = (struct ib_mad *)response->mad; + response->header.recv_wc.recv_buf.grh = &response->grh; + + agent_send_response((const struct ib_mad_hdr *)response->mad, + &response->grh, wc, + port_priv->device, + smi_get_fwd_port(smp), + qp_info->qp->qp_num, + response->mad_size, + false); + + return IB_SMI_DISCARD; + } + return IB_SMI_HANDLE; +} + +static bool generate_unmatched_resp(const struct ib_mad_private *recv, + struct ib_mad_private *response, + size_t *resp_len, bool opa) +{ + const struct ib_mad_hdr *recv_hdr = (const struct ib_mad_hdr *)recv->mad; + struct ib_mad_hdr *resp_hdr = (struct ib_mad_hdr *)response->mad; + + if (recv_hdr->method == IB_MGMT_METHOD_GET || + recv_hdr->method == IB_MGMT_METHOD_SET) { + memcpy(response, recv, mad_priv_size(response)); + response->header.recv_wc.wc = &response->header.wc; + response->header.recv_wc.recv_buf.mad = (struct ib_mad *)response->mad; + response->header.recv_wc.recv_buf.grh = &response->grh; + resp_hdr->method = IB_MGMT_METHOD_GET_RESP; + resp_hdr->status = cpu_to_be16(IB_MGMT_MAD_STATUS_UNSUPPORTED_METHOD_ATTRIB); + if (recv_hdr->mgmt_class == IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE) + resp_hdr->status |= IB_SMP_DIRECTION; + + if (opa && recv_hdr->base_version == OPA_MGMT_BASE_VERSION) { + if (recv_hdr->mgmt_class == + IB_MGMT_CLASS_SUBN_LID_ROUTED || + recv_hdr->mgmt_class == + IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE) + *resp_len = opa_get_smp_header_size( + (struct opa_smp *)recv->mad); + else + *resp_len = sizeof(struct ib_mad_hdr); + } + + return true; + } else { + return false; + } +} + +static enum smi_action +handle_opa_smi(struct ib_mad_port_private *port_priv, + struct ib_mad_qp_info *qp_info, + struct ib_wc *wc, + int port_num, + struct ib_mad_private *recv, + struct ib_mad_private *response) +{ + enum smi_forward_action retsmi; + struct opa_smp *smp = (struct opa_smp *)recv->mad; + + trace_ib_mad_handle_opa_smi(smp); + + if (opa_smi_handle_dr_smp_recv(smp, + rdma_cap_ib_switch(port_priv->device), + port_num, + port_priv->device->phys_port_cnt) == + IB_SMI_DISCARD) + return IB_SMI_DISCARD; + + retsmi = opa_smi_check_forward_dr_smp(smp); + if (retsmi == IB_SMI_LOCAL) + return IB_SMI_HANDLE; + + if (retsmi == IB_SMI_SEND) { /* don't forward */ + if (opa_smi_handle_dr_smp_send(smp, + rdma_cap_ib_switch(port_priv->device), + port_num) == IB_SMI_DISCARD) + return IB_SMI_DISCARD; + + if (opa_smi_check_local_smp(smp, port_priv->device) == + IB_SMI_DISCARD) + return IB_SMI_DISCARD; + + } else if (rdma_cap_ib_switch(port_priv->device)) { + /* forward case for switches */ + memcpy(response, recv, mad_priv_size(response)); + response->header.recv_wc.wc = &response->header.wc; + response->header.recv_wc.recv_buf.opa_mad = + (struct opa_mad *)response->mad; + response->header.recv_wc.recv_buf.grh = &response->grh; + + agent_send_response((const struct ib_mad_hdr *)response->mad, + &response->grh, wc, + port_priv->device, + opa_smi_get_fwd_port(smp), + qp_info->qp->qp_num, + recv->header.wc.byte_len, + true); + + return IB_SMI_DISCARD; + } + + return IB_SMI_HANDLE; +} + +static enum smi_action +handle_smi(struct ib_mad_port_private *port_priv, + struct ib_mad_qp_info *qp_info, + struct ib_wc *wc, + int port_num, + struct ib_mad_private *recv, + struct ib_mad_private *response, + bool opa) +{ + struct ib_mad_hdr *mad_hdr = (struct ib_mad_hdr *)recv->mad; + + if (opa && mad_hdr->base_version == OPA_MGMT_BASE_VERSION && + mad_hdr->class_version == OPA_SM_CLASS_VERSION) + return handle_opa_smi(port_priv, qp_info, wc, port_num, recv, + response); + + return handle_ib_smi(port_priv, qp_info, wc, port_num, recv, response); +} + +static void ib_mad_recv_done(struct ib_cq *cq, struct ib_wc *wc) +{ + struct ib_mad_port_private *port_priv = cq->cq_context; + struct ib_mad_list_head *mad_list = + container_of(wc->wr_cqe, struct ib_mad_list_head, cqe); + struct ib_mad_qp_info *qp_info; + struct ib_mad_private_header *mad_priv_hdr; + struct ib_mad_private *recv, *response = NULL; + struct ib_mad_agent_private *mad_agent; + int port_num; + int ret = IB_MAD_RESULT_SUCCESS; + size_t mad_size; + u16 resp_mad_pkey_index = 0; + bool opa; + + if (list_empty_careful(&port_priv->port_list)) + return; + + if (wc->status != IB_WC_SUCCESS) { + /* + * Receive errors indicate that the QP has entered the error + * state - error handling/shutdown code will cleanup + */ + return; + } + + qp_info = mad_list->mad_queue->qp_info; + dequeue_mad(mad_list); + + opa = rdma_cap_opa_mad(qp_info->port_priv->device, + qp_info->port_priv->port_num); + + mad_priv_hdr = container_of(mad_list, struct ib_mad_private_header, + mad_list); + recv = container_of(mad_priv_hdr, struct ib_mad_private, header); + ib_dma_unmap_single(port_priv->device, + recv->header.mapping, + mad_priv_dma_size(recv), + DMA_FROM_DEVICE); + + /* Setup MAD receive work completion from "normal" work completion */ + recv->header.wc = *wc; + recv->header.recv_wc.wc = &recv->header.wc; + + if (opa && ((struct ib_mad_hdr *)(recv->mad))->base_version == OPA_MGMT_BASE_VERSION) { + recv->header.recv_wc.mad_len = wc->byte_len - sizeof(struct ib_grh); + recv->header.recv_wc.mad_seg_size = sizeof(struct opa_mad); + } else { + recv->header.recv_wc.mad_len = sizeof(struct ib_mad); + recv->header.recv_wc.mad_seg_size = sizeof(struct ib_mad); + } + + recv->header.recv_wc.recv_buf.mad = (struct ib_mad *)recv->mad; + recv->header.recv_wc.recv_buf.grh = &recv->grh; + + /* Validate MAD */ + if (!validate_mad((const struct ib_mad_hdr *)recv->mad, qp_info, opa)) + goto out; + + trace_ib_mad_recv_done_handler(qp_info, wc, + (struct ib_mad_hdr *)recv->mad); + + mad_size = recv->mad_size; + response = alloc_mad_private(mad_size, GFP_KERNEL); + if (!response) + goto out; + + if (rdma_cap_ib_switch(port_priv->device)) + port_num = wc->port_num; + else + port_num = port_priv->port_num; + + if (((struct ib_mad_hdr *)recv->mad)->mgmt_class == + IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE) { + if (handle_smi(port_priv, qp_info, wc, port_num, recv, + response, opa) + == IB_SMI_DISCARD) + goto out; + } + + /* Give driver "right of first refusal" on incoming MAD */ + if (port_priv->device->ops.process_mad) { + ret = port_priv->device->ops.process_mad( + port_priv->device, 0, port_priv->port_num, wc, + &recv->grh, (const struct ib_mad *)recv->mad, + (struct ib_mad *)response->mad, &mad_size, + &resp_mad_pkey_index); + + if (opa) + wc->pkey_index = resp_mad_pkey_index; + + if (ret & IB_MAD_RESULT_SUCCESS) { + if (ret & IB_MAD_RESULT_CONSUMED) + goto out; + if (ret & IB_MAD_RESULT_REPLY) { + agent_send_response((const struct ib_mad_hdr *)response->mad, + &recv->grh, wc, + port_priv->device, + port_num, + qp_info->qp->qp_num, + mad_size, opa); + goto out; + } + } + } + + mad_agent = find_mad_agent(port_priv, (const struct ib_mad_hdr *)recv->mad); + if (mad_agent) { + trace_ib_mad_recv_done_agent(mad_agent); + ib_mad_complete_recv(mad_agent, &recv->header.recv_wc); + /* + * recv is freed up in error cases in ib_mad_complete_recv + * or via recv_handler in ib_mad_complete_recv() + */ + recv = NULL; + } else if ((ret & IB_MAD_RESULT_SUCCESS) && + generate_unmatched_resp(recv, response, &mad_size, opa)) { + agent_send_response((const struct ib_mad_hdr *)response->mad, &recv->grh, wc, + port_priv->device, port_num, + qp_info->qp->qp_num, mad_size, opa); + } + +out: + /* Post another receive request for this QP */ + if (response) { + ib_mad_post_receive_mads(qp_info, response); + kfree(recv); + } else + ib_mad_post_receive_mads(qp_info, recv); +} + +static void adjust_timeout(struct ib_mad_agent_private *mad_agent_priv) +{ + struct ib_mad_send_wr_private *mad_send_wr; + unsigned long delay; + + if (list_empty(&mad_agent_priv->wait_list)) { + cancel_delayed_work(&mad_agent_priv->timed_work); + } else { + mad_send_wr = list_entry(mad_agent_priv->wait_list.next, + struct ib_mad_send_wr_private, + agent_list); + + if (time_after(mad_agent_priv->timeout, + mad_send_wr->timeout)) { + mad_agent_priv->timeout = mad_send_wr->timeout; + delay = mad_send_wr->timeout - jiffies; + if ((long)delay <= 0) + delay = 1; + mod_delayed_work(mad_agent_priv->qp_info->port_priv->wq, + &mad_agent_priv->timed_work, delay); + } + } +} + +static void wait_for_response(struct ib_mad_send_wr_private *mad_send_wr) +{ + struct ib_mad_agent_private *mad_agent_priv; + struct ib_mad_send_wr_private *temp_mad_send_wr; + struct list_head *list_item; + unsigned long delay; + + mad_agent_priv = mad_send_wr->mad_agent_priv; + list_del(&mad_send_wr->agent_list); + + delay = mad_send_wr->timeout; + mad_send_wr->timeout += jiffies; + + if (delay) { + list_for_each_prev(list_item, &mad_agent_priv->wait_list) { + temp_mad_send_wr = list_entry(list_item, + struct ib_mad_send_wr_private, + agent_list); + if (time_after(mad_send_wr->timeout, + temp_mad_send_wr->timeout)) + break; + } + } + else + list_item = &mad_agent_priv->wait_list; + list_add(&mad_send_wr->agent_list, list_item); + + /* Reschedule a work item if we have a shorter timeout */ + if (mad_agent_priv->wait_list.next == &mad_send_wr->agent_list) + mod_delayed_work(mad_agent_priv->qp_info->port_priv->wq, + &mad_agent_priv->timed_work, delay); +} + +void ib_reset_mad_timeout(struct ib_mad_send_wr_private *mad_send_wr, + unsigned long timeout_ms) +{ + mad_send_wr->timeout = msecs_to_jiffies(timeout_ms); + wait_for_response(mad_send_wr); +} + +/* + * Process a send work completion + */ +void ib_mad_complete_send_wr(struct ib_mad_send_wr_private *mad_send_wr, + struct ib_mad_send_wc *mad_send_wc) +{ + struct ib_mad_agent_private *mad_agent_priv; + unsigned long flags; + int ret; + + mad_agent_priv = mad_send_wr->mad_agent_priv; + spin_lock_irqsave(&mad_agent_priv->lock, flags); + if (ib_mad_kernel_rmpp_agent(&mad_agent_priv->agent)) { + ret = ib_process_rmpp_send_wc(mad_send_wr, mad_send_wc); + if (ret == IB_RMPP_RESULT_CONSUMED) + goto done; + } else + ret = IB_RMPP_RESULT_UNHANDLED; + + if (mad_send_wc->status != IB_WC_SUCCESS && + mad_send_wr->status == IB_WC_SUCCESS) { + mad_send_wr->status = mad_send_wc->status; + mad_send_wr->refcount -= (mad_send_wr->timeout > 0); + } + + if (--mad_send_wr->refcount > 0) { + if (mad_send_wr->refcount == 1 && mad_send_wr->timeout && + mad_send_wr->status == IB_WC_SUCCESS) { + wait_for_response(mad_send_wr); + } + goto done; + } + + /* Remove send from MAD agent and notify client of completion */ + list_del(&mad_send_wr->agent_list); + adjust_timeout(mad_agent_priv); + spin_unlock_irqrestore(&mad_agent_priv->lock, flags); + + if (mad_send_wr->status != IB_WC_SUCCESS ) + mad_send_wc->status = mad_send_wr->status; + if (ret == IB_RMPP_RESULT_INTERNAL) + ib_rmpp_send_handler(mad_send_wc); + else + mad_agent_priv->agent.send_handler(&mad_agent_priv->agent, + mad_send_wc); + + /* Release reference on agent taken when sending */ + deref_mad_agent(mad_agent_priv); + return; +done: + spin_unlock_irqrestore(&mad_agent_priv->lock, flags); +} + +static void ib_mad_send_done(struct ib_cq *cq, struct ib_wc *wc) +{ + struct ib_mad_port_private *port_priv = cq->cq_context; + struct ib_mad_list_head *mad_list = + container_of(wc->wr_cqe, struct ib_mad_list_head, cqe); + struct ib_mad_send_wr_private *mad_send_wr, *queued_send_wr; + struct ib_mad_qp_info *qp_info; + struct ib_mad_queue *send_queue; + struct ib_mad_send_wc mad_send_wc; + unsigned long flags; + int ret; + + if (list_empty_careful(&port_priv->port_list)) + return; + + if (wc->status != IB_WC_SUCCESS) { + if (!ib_mad_send_error(port_priv, wc)) + return; + } + + mad_send_wr = container_of(mad_list, struct ib_mad_send_wr_private, + mad_list); + send_queue = mad_list->mad_queue; + qp_info = send_queue->qp_info; + + trace_ib_mad_send_done_agent(mad_send_wr->mad_agent_priv); + trace_ib_mad_send_done_handler(mad_send_wr, wc); + +retry: + ib_dma_unmap_single(mad_send_wr->send_buf.mad_agent->device, + mad_send_wr->header_mapping, + mad_send_wr->sg_list[0].length, DMA_TO_DEVICE); + ib_dma_unmap_single(mad_send_wr->send_buf.mad_agent->device, + mad_send_wr->payload_mapping, + mad_send_wr->sg_list[1].length, DMA_TO_DEVICE); + queued_send_wr = NULL; + spin_lock_irqsave(&send_queue->lock, flags); + list_del(&mad_list->list); + + /* Move queued send to the send queue */ + if (send_queue->count-- > send_queue->max_active) { + mad_list = container_of(qp_info->overflow_list.next, + struct ib_mad_list_head, list); + queued_send_wr = container_of(mad_list, + struct ib_mad_send_wr_private, + mad_list); + list_move_tail(&mad_list->list, &send_queue->list); + } + spin_unlock_irqrestore(&send_queue->lock, flags); + + mad_send_wc.send_buf = &mad_send_wr->send_buf; + mad_send_wc.status = wc->status; + mad_send_wc.vendor_err = wc->vendor_err; + ib_mad_complete_send_wr(mad_send_wr, &mad_send_wc); + + if (queued_send_wr) { + trace_ib_mad_send_done_resend(queued_send_wr, qp_info); + ret = ib_post_send(qp_info->qp, &queued_send_wr->send_wr.wr, + NULL); + if (ret) { + dev_err(&port_priv->device->dev, + "ib_post_send failed: %d\n", ret); + mad_send_wr = queued_send_wr; + wc->status = IB_WC_LOC_QP_OP_ERR; + goto retry; + } + } +} + +static void mark_sends_for_retry(struct ib_mad_qp_info *qp_info) +{ + struct ib_mad_send_wr_private *mad_send_wr; + struct ib_mad_list_head *mad_list; + unsigned long flags; + + spin_lock_irqsave(&qp_info->send_queue.lock, flags); + list_for_each_entry(mad_list, &qp_info->send_queue.list, list) { + mad_send_wr = container_of(mad_list, + struct ib_mad_send_wr_private, + mad_list); + mad_send_wr->retry = 1; + } + spin_unlock_irqrestore(&qp_info->send_queue.lock, flags); +} + +static bool ib_mad_send_error(struct ib_mad_port_private *port_priv, + struct ib_wc *wc) +{ + struct ib_mad_list_head *mad_list = + container_of(wc->wr_cqe, struct ib_mad_list_head, cqe); + struct ib_mad_qp_info *qp_info = mad_list->mad_queue->qp_info; + struct ib_mad_send_wr_private *mad_send_wr; + int ret; + + /* + * Send errors will transition the QP to SQE - move + * QP to RTS and repost flushed work requests + */ + mad_send_wr = container_of(mad_list, struct ib_mad_send_wr_private, + mad_list); + if (wc->status == IB_WC_WR_FLUSH_ERR) { + if (mad_send_wr->retry) { + /* Repost send */ + mad_send_wr->retry = 0; + trace_ib_mad_error_handler(mad_send_wr, qp_info); + ret = ib_post_send(qp_info->qp, &mad_send_wr->send_wr.wr, + NULL); + if (!ret) + return false; + } + } else { + struct ib_qp_attr *attr; + + /* Transition QP to RTS and fail offending send */ + attr = kmalloc(sizeof *attr, GFP_KERNEL); + if (attr) { + attr->qp_state = IB_QPS_RTS; + attr->cur_qp_state = IB_QPS_SQE; + ret = ib_modify_qp(qp_info->qp, attr, + IB_QP_STATE | IB_QP_CUR_STATE); + kfree(attr); + if (ret) + dev_err(&port_priv->device->dev, + "%s - ib_modify_qp to RTS: %d\n", + __func__, ret); + else + mark_sends_for_retry(qp_info); + } + } + + return true; +} + +static void cancel_mads(struct ib_mad_agent_private *mad_agent_priv) +{ + unsigned long flags; + struct ib_mad_send_wr_private *mad_send_wr, *temp_mad_send_wr; + struct ib_mad_send_wc mad_send_wc; + struct list_head cancel_list; + + INIT_LIST_HEAD(&cancel_list); + + spin_lock_irqsave(&mad_agent_priv->lock, flags); + list_for_each_entry_safe(mad_send_wr, temp_mad_send_wr, + &mad_agent_priv->send_list, agent_list) { + if (mad_send_wr->status == IB_WC_SUCCESS) { + mad_send_wr->status = IB_WC_WR_FLUSH_ERR; + mad_send_wr->refcount -= (mad_send_wr->timeout > 0); + } + } + + /* Empty wait list to prevent receives from finding a request */ + list_splice_init(&mad_agent_priv->wait_list, &cancel_list); + spin_unlock_irqrestore(&mad_agent_priv->lock, flags); + + /* Report all cancelled requests */ + mad_send_wc.status = IB_WC_WR_FLUSH_ERR; + mad_send_wc.vendor_err = 0; + + list_for_each_entry_safe(mad_send_wr, temp_mad_send_wr, + &cancel_list, agent_list) { + mad_send_wc.send_buf = &mad_send_wr->send_buf; + list_del(&mad_send_wr->agent_list); + mad_agent_priv->agent.send_handler(&mad_agent_priv->agent, + &mad_send_wc); + deref_mad_agent(mad_agent_priv); + } +} + +static struct ib_mad_send_wr_private* +find_send_wr(struct ib_mad_agent_private *mad_agent_priv, + struct ib_mad_send_buf *send_buf) +{ + struct ib_mad_send_wr_private *mad_send_wr; + + list_for_each_entry(mad_send_wr, &mad_agent_priv->wait_list, + agent_list) { + if (&mad_send_wr->send_buf == send_buf) + return mad_send_wr; + } + + list_for_each_entry(mad_send_wr, &mad_agent_priv->send_list, + agent_list) { + if (is_rmpp_data_mad(mad_agent_priv, + mad_send_wr->send_buf.mad) && + &mad_send_wr->send_buf == send_buf) + return mad_send_wr; + } + return NULL; +} + +int ib_modify_mad(struct ib_mad_agent *mad_agent, + struct ib_mad_send_buf *send_buf, u32 timeout_ms) +{ + struct ib_mad_agent_private *mad_agent_priv; + struct ib_mad_send_wr_private *mad_send_wr; + unsigned long flags; + int active; + + mad_agent_priv = container_of(mad_agent, struct ib_mad_agent_private, + agent); + spin_lock_irqsave(&mad_agent_priv->lock, flags); + mad_send_wr = find_send_wr(mad_agent_priv, send_buf); + if (!mad_send_wr || mad_send_wr->status != IB_WC_SUCCESS) { + spin_unlock_irqrestore(&mad_agent_priv->lock, flags); + return -EINVAL; + } + + active = (!mad_send_wr->timeout || mad_send_wr->refcount > 1); + if (!timeout_ms) { + mad_send_wr->status = IB_WC_WR_FLUSH_ERR; + mad_send_wr->refcount -= (mad_send_wr->timeout > 0); + } + + mad_send_wr->send_buf.timeout_ms = timeout_ms; + if (active) + mad_send_wr->timeout = msecs_to_jiffies(timeout_ms); + else + ib_reset_mad_timeout(mad_send_wr, timeout_ms); + + spin_unlock_irqrestore(&mad_agent_priv->lock, flags); + return 0; +} +EXPORT_SYMBOL(ib_modify_mad); + +void ib_cancel_mad(struct ib_mad_agent *mad_agent, + struct ib_mad_send_buf *send_buf) +{ + ib_modify_mad(mad_agent, send_buf, 0); +} +EXPORT_SYMBOL(ib_cancel_mad); + +static void local_completions(struct work_struct *work) +{ + struct ib_mad_agent_private *mad_agent_priv; + struct ib_mad_local_private *local; + struct ib_mad_agent_private *recv_mad_agent; + unsigned long flags; + int free_mad; + struct ib_wc wc; + struct ib_mad_send_wc mad_send_wc; + bool opa; + + mad_agent_priv = + container_of(work, struct ib_mad_agent_private, local_work); + + opa = rdma_cap_opa_mad(mad_agent_priv->qp_info->port_priv->device, + mad_agent_priv->qp_info->port_priv->port_num); + + spin_lock_irqsave(&mad_agent_priv->lock, flags); + while (!list_empty(&mad_agent_priv->local_list)) { + local = list_entry(mad_agent_priv->local_list.next, + struct ib_mad_local_private, + completion_list); + list_del(&local->completion_list); + spin_unlock_irqrestore(&mad_agent_priv->lock, flags); + free_mad = 0; + if (local->mad_priv) { + u8 base_version; + recv_mad_agent = local->recv_mad_agent; + if (!recv_mad_agent) { + dev_err(&mad_agent_priv->agent.device->dev, + "No receive MAD agent for local completion\n"); + free_mad = 1; + goto local_send_completion; + } + + /* + * Defined behavior is to complete response + * before request + */ + build_smp_wc(recv_mad_agent->agent.qp, + local->mad_send_wr->send_wr.wr.wr_cqe, + be16_to_cpu(IB_LID_PERMISSIVE), + local->mad_send_wr->send_wr.pkey_index, + recv_mad_agent->agent.port_num, &wc); + + local->mad_priv->header.recv_wc.wc = &wc; + + base_version = ((struct ib_mad_hdr *)(local->mad_priv->mad))->base_version; + if (opa && base_version == OPA_MGMT_BASE_VERSION) { + local->mad_priv->header.recv_wc.mad_len = local->return_wc_byte_len; + local->mad_priv->header.recv_wc.mad_seg_size = sizeof(struct opa_mad); + } else { + local->mad_priv->header.recv_wc.mad_len = sizeof(struct ib_mad); + local->mad_priv->header.recv_wc.mad_seg_size = sizeof(struct ib_mad); + } + + INIT_LIST_HEAD(&local->mad_priv->header.recv_wc.rmpp_list); + list_add(&local->mad_priv->header.recv_wc.recv_buf.list, + &local->mad_priv->header.recv_wc.rmpp_list); + local->mad_priv->header.recv_wc.recv_buf.grh = NULL; + local->mad_priv->header.recv_wc.recv_buf.mad = + (struct ib_mad *)local->mad_priv->mad; + recv_mad_agent->agent.recv_handler( + &recv_mad_agent->agent, + &local->mad_send_wr->send_buf, + &local->mad_priv->header.recv_wc); + spin_lock_irqsave(&recv_mad_agent->lock, flags); + deref_mad_agent(recv_mad_agent); + spin_unlock_irqrestore(&recv_mad_agent->lock, flags); + } + +local_send_completion: + /* Complete send */ + mad_send_wc.status = IB_WC_SUCCESS; + mad_send_wc.vendor_err = 0; + mad_send_wc.send_buf = &local->mad_send_wr->send_buf; + mad_agent_priv->agent.send_handler(&mad_agent_priv->agent, + &mad_send_wc); + + spin_lock_irqsave(&mad_agent_priv->lock, flags); + deref_mad_agent(mad_agent_priv); + if (free_mad) + kfree(local->mad_priv); + kfree(local); + } + spin_unlock_irqrestore(&mad_agent_priv->lock, flags); +} + +static int retry_send(struct ib_mad_send_wr_private *mad_send_wr) +{ + int ret; + + if (!mad_send_wr->retries_left) + return -ETIMEDOUT; + + mad_send_wr->retries_left--; + mad_send_wr->send_buf.retries++; + + mad_send_wr->timeout = msecs_to_jiffies(mad_send_wr->send_buf.timeout_ms); + + if (ib_mad_kernel_rmpp_agent(&mad_send_wr->mad_agent_priv->agent)) { + ret = ib_retry_rmpp(mad_send_wr); + switch (ret) { + case IB_RMPP_RESULT_UNHANDLED: + ret = ib_send_mad(mad_send_wr); + break; + case IB_RMPP_RESULT_CONSUMED: + ret = 0; + break; + default: + ret = -ECOMM; + break; + } + } else + ret = ib_send_mad(mad_send_wr); + + if (!ret) { + mad_send_wr->refcount++; + list_add_tail(&mad_send_wr->agent_list, + &mad_send_wr->mad_agent_priv->send_list); + } + return ret; +} + +static void timeout_sends(struct work_struct *work) +{ + struct ib_mad_agent_private *mad_agent_priv; + struct ib_mad_send_wr_private *mad_send_wr; + struct ib_mad_send_wc mad_send_wc; + unsigned long flags, delay; + + mad_agent_priv = container_of(work, struct ib_mad_agent_private, + timed_work.work); + mad_send_wc.vendor_err = 0; + + spin_lock_irqsave(&mad_agent_priv->lock, flags); + while (!list_empty(&mad_agent_priv->wait_list)) { + mad_send_wr = list_entry(mad_agent_priv->wait_list.next, + struct ib_mad_send_wr_private, + agent_list); + + if (time_after(mad_send_wr->timeout, jiffies)) { + delay = mad_send_wr->timeout - jiffies; + if ((long)delay <= 0) + delay = 1; + queue_delayed_work(mad_agent_priv->qp_info-> + port_priv->wq, + &mad_agent_priv->timed_work, delay); + break; + } + + list_del(&mad_send_wr->agent_list); + if (mad_send_wr->status == IB_WC_SUCCESS && + !retry_send(mad_send_wr)) + continue; + + spin_unlock_irqrestore(&mad_agent_priv->lock, flags); + + if (mad_send_wr->status == IB_WC_SUCCESS) + mad_send_wc.status = IB_WC_RESP_TIMEOUT_ERR; + else + mad_send_wc.status = mad_send_wr->status; + mad_send_wc.send_buf = &mad_send_wr->send_buf; + mad_agent_priv->agent.send_handler(&mad_agent_priv->agent, + &mad_send_wc); + + deref_mad_agent(mad_agent_priv); + spin_lock_irqsave(&mad_agent_priv->lock, flags); + } + spin_unlock_irqrestore(&mad_agent_priv->lock, flags); +} + +/* + * Allocate receive MADs and post receive WRs for them + */ +static int ib_mad_post_receive_mads(struct ib_mad_qp_info *qp_info, + struct ib_mad_private *mad) +{ + unsigned long flags; + int post, ret; + struct ib_mad_private *mad_priv; + struct ib_sge sg_list; + struct ib_recv_wr recv_wr; + struct ib_mad_queue *recv_queue = &qp_info->recv_queue; + + /* Initialize common scatter list fields */ + sg_list.lkey = qp_info->port_priv->pd->local_dma_lkey; + + /* Initialize common receive WR fields */ + recv_wr.next = NULL; + recv_wr.sg_list = &sg_list; + recv_wr.num_sge = 1; + + do { + /* Allocate and map receive buffer */ + if (mad) { + mad_priv = mad; + mad = NULL; + } else { + mad_priv = alloc_mad_private(port_mad_size(qp_info->port_priv), + GFP_ATOMIC); + if (!mad_priv) { + ret = -ENOMEM; + break; + } + } + sg_list.length = mad_priv_dma_size(mad_priv); + sg_list.addr = ib_dma_map_single(qp_info->port_priv->device, + &mad_priv->grh, + mad_priv_dma_size(mad_priv), + DMA_FROM_DEVICE); + if (unlikely(ib_dma_mapping_error(qp_info->port_priv->device, + sg_list.addr))) { + kfree(mad_priv); + ret = -ENOMEM; + break; + } + mad_priv->header.mapping = sg_list.addr; + mad_priv->header.mad_list.mad_queue = recv_queue; + mad_priv->header.mad_list.cqe.done = ib_mad_recv_done; + recv_wr.wr_cqe = &mad_priv->header.mad_list.cqe; + + /* Post receive WR */ + spin_lock_irqsave(&recv_queue->lock, flags); + post = (++recv_queue->count < recv_queue->max_active); + list_add_tail(&mad_priv->header.mad_list.list, &recv_queue->list); + spin_unlock_irqrestore(&recv_queue->lock, flags); + ret = ib_post_recv(qp_info->qp, &recv_wr, NULL); + if (ret) { + spin_lock_irqsave(&recv_queue->lock, flags); + list_del(&mad_priv->header.mad_list.list); + recv_queue->count--; + spin_unlock_irqrestore(&recv_queue->lock, flags); + ib_dma_unmap_single(qp_info->port_priv->device, + mad_priv->header.mapping, + mad_priv_dma_size(mad_priv), + DMA_FROM_DEVICE); + kfree(mad_priv); + dev_err(&qp_info->port_priv->device->dev, + "ib_post_recv failed: %d\n", ret); + break; + } + } while (post); + + return ret; +} + +/* + * Return all the posted receive MADs + */ +static void cleanup_recv_queue(struct ib_mad_qp_info *qp_info) +{ + struct ib_mad_private_header *mad_priv_hdr; + struct ib_mad_private *recv; + struct ib_mad_list_head *mad_list; + + if (!qp_info->qp) + return; + + while (!list_empty(&qp_info->recv_queue.list)) { + + mad_list = list_entry(qp_info->recv_queue.list.next, + struct ib_mad_list_head, list); + mad_priv_hdr = container_of(mad_list, + struct ib_mad_private_header, + mad_list); + recv = container_of(mad_priv_hdr, struct ib_mad_private, + header); + + /* Remove from posted receive MAD list */ + list_del(&mad_list->list); + + ib_dma_unmap_single(qp_info->port_priv->device, + recv->header.mapping, + mad_priv_dma_size(recv), + DMA_FROM_DEVICE); + kfree(recv); + } + + qp_info->recv_queue.count = 0; +} + +/* + * Start the port + */ +static int ib_mad_port_start(struct ib_mad_port_private *port_priv) +{ + int ret, i; + struct ib_qp_attr *attr; + struct ib_qp *qp; + u16 pkey_index; + + attr = kmalloc(sizeof *attr, GFP_KERNEL); + if (!attr) + return -ENOMEM; + + ret = ib_find_pkey(port_priv->device, port_priv->port_num, + IB_DEFAULT_PKEY_FULL, &pkey_index); + if (ret) + pkey_index = 0; + + for (i = 0; i < IB_MAD_QPS_CORE; i++) { + qp = port_priv->qp_info[i].qp; + if (!qp) + continue; + + /* + * PKey index for QP1 is irrelevant but + * one is needed for the Reset to Init transition + */ + attr->qp_state = IB_QPS_INIT; + attr->pkey_index = pkey_index; + attr->qkey = (qp->qp_num == 0) ? 0 : IB_QP1_QKEY; + ret = ib_modify_qp(qp, attr, IB_QP_STATE | + IB_QP_PKEY_INDEX | IB_QP_QKEY); + if (ret) { + dev_err(&port_priv->device->dev, + "Couldn't change QP%d state to INIT: %d\n", + i, ret); + goto out; + } + + attr->qp_state = IB_QPS_RTR; + ret = ib_modify_qp(qp, attr, IB_QP_STATE); + if (ret) { + dev_err(&port_priv->device->dev, + "Couldn't change QP%d state to RTR: %d\n", + i, ret); + goto out; + } + + attr->qp_state = IB_QPS_RTS; + attr->sq_psn = IB_MAD_SEND_Q_PSN; + ret = ib_modify_qp(qp, attr, IB_QP_STATE | IB_QP_SQ_PSN); + if (ret) { + dev_err(&port_priv->device->dev, + "Couldn't change QP%d state to RTS: %d\n", + i, ret); + goto out; + } + } + + ret = ib_req_notify_cq(port_priv->cq, IB_CQ_NEXT_COMP); + if (ret) { + dev_err(&port_priv->device->dev, + "Failed to request completion notification: %d\n", + ret); + goto out; + } + + for (i = 0; i < IB_MAD_QPS_CORE; i++) { + if (!port_priv->qp_info[i].qp) + continue; + + ret = ib_mad_post_receive_mads(&port_priv->qp_info[i], NULL); + if (ret) { + dev_err(&port_priv->device->dev, + "Couldn't post receive WRs\n"); + goto out; + } + } +out: + kfree(attr); + return ret; +} + +static void qp_event_handler(struct ib_event *event, void *qp_context) +{ + struct ib_mad_qp_info *qp_info = qp_context; + + /* It's worse than that! He's dead, Jim! */ + dev_err(&qp_info->port_priv->device->dev, + "Fatal error (%d) on MAD QP (%d)\n", + event->event, qp_info->qp->qp_num); +} + +static void init_mad_queue(struct ib_mad_qp_info *qp_info, + struct ib_mad_queue *mad_queue) +{ + mad_queue->qp_info = qp_info; + mad_queue->count = 0; + spin_lock_init(&mad_queue->lock); + INIT_LIST_HEAD(&mad_queue->list); +} + +static void init_mad_qp(struct ib_mad_port_private *port_priv, + struct ib_mad_qp_info *qp_info) +{ + qp_info->port_priv = port_priv; + init_mad_queue(qp_info, &qp_info->send_queue); + init_mad_queue(qp_info, &qp_info->recv_queue); + INIT_LIST_HEAD(&qp_info->overflow_list); +} + +static int create_mad_qp(struct ib_mad_qp_info *qp_info, + enum ib_qp_type qp_type) +{ + struct ib_qp_init_attr qp_init_attr; + int ret; + + memset(&qp_init_attr, 0, sizeof qp_init_attr); + qp_init_attr.send_cq = qp_info->port_priv->cq; + qp_init_attr.recv_cq = qp_info->port_priv->cq; + qp_init_attr.sq_sig_type = IB_SIGNAL_ALL_WR; + qp_init_attr.cap.max_send_wr = mad_sendq_size; + qp_init_attr.cap.max_recv_wr = mad_recvq_size; + qp_init_attr.cap.max_send_sge = IB_MAD_SEND_REQ_MAX_SG; + qp_init_attr.cap.max_recv_sge = IB_MAD_RECV_REQ_MAX_SG; + qp_init_attr.qp_type = qp_type; + qp_init_attr.port_num = qp_info->port_priv->port_num; + qp_init_attr.qp_context = qp_info; + qp_init_attr.event_handler = qp_event_handler; + qp_info->qp = ib_create_qp(qp_info->port_priv->pd, &qp_init_attr); + if (IS_ERR(qp_info->qp)) { + dev_err(&qp_info->port_priv->device->dev, + "Couldn't create ib_mad QP%d\n", + get_spl_qp_index(qp_type)); + ret = PTR_ERR(qp_info->qp); + goto error; + } + /* Use minimum queue sizes unless the CQ is resized */ + qp_info->send_queue.max_active = mad_sendq_size; + qp_info->recv_queue.max_active = mad_recvq_size; + return 0; + +error: + return ret; +} + +static void destroy_mad_qp(struct ib_mad_qp_info *qp_info) +{ + if (!qp_info->qp) + return; + + ib_destroy_qp(qp_info->qp); +} + +/* + * Open the port + * Create the QP, PD, MR, and CQ if needed + */ +static int ib_mad_port_open(struct ib_device *device, + int port_num) +{ + int ret, cq_size; + struct ib_mad_port_private *port_priv; + unsigned long flags; + char name[sizeof "ib_mad123"]; + int has_smi; + + if (WARN_ON(rdma_max_mad_size(device, port_num) < IB_MGMT_MAD_SIZE)) + return -EFAULT; + + if (WARN_ON(rdma_cap_opa_mad(device, port_num) && + rdma_max_mad_size(device, port_num) < OPA_MGMT_MAD_SIZE)) + return -EFAULT; + + /* Create new device info */ + port_priv = kzalloc(sizeof *port_priv, GFP_KERNEL); + if (!port_priv) + return -ENOMEM; + + port_priv->device = device; + port_priv->port_num = port_num; + spin_lock_init(&port_priv->reg_lock); + init_mad_qp(port_priv, &port_priv->qp_info[0]); + init_mad_qp(port_priv, &port_priv->qp_info[1]); + + cq_size = mad_sendq_size + mad_recvq_size; + has_smi = rdma_cap_ib_smi(device, port_num); + if (has_smi) + cq_size *= 2; + + port_priv->pd = ib_alloc_pd(device, 0); + if (IS_ERR(port_priv->pd)) { + dev_err(&device->dev, "Couldn't create ib_mad PD\n"); + ret = PTR_ERR(port_priv->pd); + goto error3; + } + + port_priv->cq = ib_alloc_cq(port_priv->device, port_priv, cq_size, 0, + IB_POLL_UNBOUND_WORKQUEUE); + if (IS_ERR(port_priv->cq)) { + dev_err(&device->dev, "Couldn't create ib_mad CQ\n"); + ret = PTR_ERR(port_priv->cq); + goto error4; + } + + if (has_smi) { + ret = create_mad_qp(&port_priv->qp_info[0], IB_QPT_SMI); + if (ret) + goto error6; + } + ret = create_mad_qp(&port_priv->qp_info[1], IB_QPT_GSI); + if (ret) + goto error7; + + snprintf(name, sizeof name, "ib_mad%d", port_num); + port_priv->wq = alloc_ordered_workqueue(name, WQ_MEM_RECLAIM); + if (!port_priv->wq) { + ret = -ENOMEM; + goto error8; + } + + spin_lock_irqsave(&ib_mad_port_list_lock, flags); + list_add_tail(&port_priv->port_list, &ib_mad_port_list); + spin_unlock_irqrestore(&ib_mad_port_list_lock, flags); + + ret = ib_mad_port_start(port_priv); + if (ret) { + dev_err(&device->dev, "Couldn't start port\n"); + goto error9; + } + + return 0; + +error9: + spin_lock_irqsave(&ib_mad_port_list_lock, flags); + list_del_init(&port_priv->port_list); + spin_unlock_irqrestore(&ib_mad_port_list_lock, flags); + + destroy_workqueue(port_priv->wq); +error8: + destroy_mad_qp(&port_priv->qp_info[1]); +error7: + destroy_mad_qp(&port_priv->qp_info[0]); +error6: + ib_free_cq(port_priv->cq); + cleanup_recv_queue(&port_priv->qp_info[1]); + cleanup_recv_queue(&port_priv->qp_info[0]); +error4: + ib_dealloc_pd(port_priv->pd); +error3: + kfree(port_priv); + + return ret; +} + +/* + * Close the port + * If there are no classes using the port, free the port + * resources (CQ, MR, PD, QP) and remove the port's info structure + */ +static int ib_mad_port_close(struct ib_device *device, int port_num) +{ + struct ib_mad_port_private *port_priv; + unsigned long flags; + + spin_lock_irqsave(&ib_mad_port_list_lock, flags); + port_priv = __ib_get_mad_port(device, port_num); + if (port_priv == NULL) { + spin_unlock_irqrestore(&ib_mad_port_list_lock, flags); + dev_err(&device->dev, "Port %d not found\n", port_num); + return -ENODEV; + } + list_del_init(&port_priv->port_list); + spin_unlock_irqrestore(&ib_mad_port_list_lock, flags); + + destroy_workqueue(port_priv->wq); + destroy_mad_qp(&port_priv->qp_info[1]); + destroy_mad_qp(&port_priv->qp_info[0]); + ib_free_cq(port_priv->cq); + ib_dealloc_pd(port_priv->pd); + cleanup_recv_queue(&port_priv->qp_info[1]); + cleanup_recv_queue(&port_priv->qp_info[0]); + /* XXX: Handle deallocation of MAD registration tables */ + + kfree(port_priv); + + return 0; +} + +static int ib_mad_init_device(struct ib_device *device) +{ + int start, i; + unsigned int count = 0; + int ret; + + start = rdma_start_port(device); + + for (i = start; i <= rdma_end_port(device); i++) { + if (!rdma_cap_ib_mad(device, i)) + continue; + + ret = ib_mad_port_open(device, i); + if (ret) { + dev_err(&device->dev, "Couldn't open port %d\n", i); + goto error; + } + ret = ib_agent_port_open(device, i); + if (ret) { + dev_err(&device->dev, + "Couldn't open port %d for agents\n", i); + goto error_agent; + } + count++; + } + if (!count) + return -EOPNOTSUPP; + + return 0; + +error_agent: + if (ib_mad_port_close(device, i)) + dev_err(&device->dev, "Couldn't close port %d\n", i); + +error: + while (--i >= start) { + if (!rdma_cap_ib_mad(device, i)) + continue; + + if (ib_agent_port_close(device, i)) + dev_err(&device->dev, + "Couldn't close port %d for agents\n", i); + if (ib_mad_port_close(device, i)) + dev_err(&device->dev, "Couldn't close port %d\n", i); + } + return ret; +} + +static void ib_mad_remove_device(struct ib_device *device, void *client_data) +{ + unsigned int i; + + rdma_for_each_port (device, i) { + if (!rdma_cap_ib_mad(device, i)) + continue; + + if (ib_agent_port_close(device, i)) + dev_err(&device->dev, + "Couldn't close port %d for agents\n", i); + if (ib_mad_port_close(device, i)) + dev_err(&device->dev, "Couldn't close port %d\n", i); + } +} + +static struct ib_client mad_client = { + .name = "mad", + .add = ib_mad_init_device, + .remove = ib_mad_remove_device +}; + +int ib_mad_init(void) +{ + mad_recvq_size = min(mad_recvq_size, IB_MAD_QP_MAX_SIZE); + mad_recvq_size = max(mad_recvq_size, IB_MAD_QP_MIN_SIZE); + + mad_sendq_size = min(mad_sendq_size, IB_MAD_QP_MAX_SIZE); + mad_sendq_size = max(mad_sendq_size, IB_MAD_QP_MIN_SIZE); + + INIT_LIST_HEAD(&ib_mad_port_list); + + if (ib_register_client(&mad_client)) { + pr_err("Couldn't register ib_mad client\n"); + return -EINVAL; + } + + return 0; +} + +void ib_mad_cleanup(void) +{ + ib_unregister_client(&mad_client); +} diff --git a/drivers/infiniband/core/mad_priv.h b/drivers/infiniband/core/mad_priv.h new file mode 100644 index 000000000..4aa16b35d --- /dev/null +++ b/drivers/infiniband/core/mad_priv.h @@ -0,0 +1,226 @@ +/* + * Copyright (c) 2004, 2005, Voltaire, Inc. All rights reserved. + * Copyright (c) 2005 Intel Corporation. All rights reserved. + * Copyright (c) 2005 Sun Microsystems, Inc. All rights reserved. + * Copyright (c) 2009 HNR Consulting. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#ifndef __IB_MAD_PRIV_H__ +#define __IB_MAD_PRIV_H__ + +#include <linux/completion.h> +#include <linux/err.h> +#include <linux/workqueue.h> +#include <rdma/ib_mad.h> +#include <rdma/ib_smi.h> +#include <rdma/opa_smi.h> + +#define IB_MAD_QPS_CORE 2 /* Always QP0 and QP1 as a minimum */ + +/* QP and CQ parameters */ +#define IB_MAD_QP_SEND_SIZE 128 +#define IB_MAD_QP_RECV_SIZE 512 +#define IB_MAD_QP_MIN_SIZE 64 +#define IB_MAD_QP_MAX_SIZE 8192 +#define IB_MAD_SEND_REQ_MAX_SG 2 +#define IB_MAD_RECV_REQ_MAX_SG 1 + +#define IB_MAD_SEND_Q_PSN 0 + +/* Registration table sizes */ +#define MAX_MGMT_CLASS 80 +#define MAX_MGMT_VERSION 0x83 +#define MAX_MGMT_OUI 8 +#define MAX_MGMT_VENDOR_RANGE2 (IB_MGMT_CLASS_VENDOR_RANGE2_END - \ + IB_MGMT_CLASS_VENDOR_RANGE2_START + 1) + +struct ib_mad_list_head { + struct list_head list; + struct ib_cqe cqe; + struct ib_mad_queue *mad_queue; +}; + +struct ib_mad_private_header { + struct ib_mad_list_head mad_list; + struct ib_mad_recv_wc recv_wc; + struct ib_wc wc; + u64 mapping; +} __packed; + +struct ib_mad_private { + struct ib_mad_private_header header; + size_t mad_size; + struct ib_grh grh; + u8 mad[]; +} __packed; + +struct ib_rmpp_segment { + struct list_head list; + u32 num; + u8 data[]; +}; + +struct ib_mad_agent_private { + struct ib_mad_agent agent; + struct ib_mad_reg_req *reg_req; + struct ib_mad_qp_info *qp_info; + + spinlock_t lock; + struct list_head send_list; + struct list_head wait_list; + struct list_head done_list; + struct delayed_work timed_work; + unsigned long timeout; + struct list_head local_list; + struct work_struct local_work; + struct list_head rmpp_list; + + refcount_t refcount; + union { + struct completion comp; + struct rcu_head rcu; + }; +}; + +struct ib_mad_snoop_private { + struct ib_mad_agent agent; + struct ib_mad_qp_info *qp_info; + int snoop_index; + int mad_snoop_flags; + atomic_t refcount; + struct completion comp; +}; + +struct ib_mad_send_wr_private { + struct ib_mad_list_head mad_list; + struct list_head agent_list; + struct ib_mad_agent_private *mad_agent_priv; + struct ib_mad_send_buf send_buf; + u64 header_mapping; + u64 payload_mapping; + struct ib_ud_wr send_wr; + struct ib_sge sg_list[IB_MAD_SEND_REQ_MAX_SG]; + __be64 tid; + unsigned long timeout; + int max_retries; + int retries_left; + int retry; + int refcount; + enum ib_wc_status status; + + /* RMPP control */ + struct list_head rmpp_list; + struct ib_rmpp_segment *last_ack_seg; + struct ib_rmpp_segment *cur_seg; + int last_ack; + int seg_num; + int newwin; + int pad; +}; + +struct ib_mad_local_private { + struct list_head completion_list; + struct ib_mad_private *mad_priv; + struct ib_mad_agent_private *recv_mad_agent; + struct ib_mad_send_wr_private *mad_send_wr; + size_t return_wc_byte_len; +}; + +struct ib_mad_mgmt_method_table { + struct ib_mad_agent_private *agent[IB_MGMT_MAX_METHODS]; +}; + +struct ib_mad_mgmt_class_table { + struct ib_mad_mgmt_method_table *method_table[MAX_MGMT_CLASS]; +}; + +struct ib_mad_mgmt_vendor_class { + u8 oui[MAX_MGMT_OUI][3]; + struct ib_mad_mgmt_method_table *method_table[MAX_MGMT_OUI]; +}; + +struct ib_mad_mgmt_vendor_class_table { + struct ib_mad_mgmt_vendor_class *vendor_class[MAX_MGMT_VENDOR_RANGE2]; +}; + +struct ib_mad_mgmt_version_table { + struct ib_mad_mgmt_class_table *class; + struct ib_mad_mgmt_vendor_class_table *vendor; +}; + +struct ib_mad_queue { + spinlock_t lock; + struct list_head list; + int count; + int max_active; + struct ib_mad_qp_info *qp_info; +}; + +struct ib_mad_qp_info { + struct ib_mad_port_private *port_priv; + struct ib_qp *qp; + struct ib_mad_queue send_queue; + struct ib_mad_queue recv_queue; + struct list_head overflow_list; + spinlock_t snoop_lock; + struct ib_mad_snoop_private **snoop_table; + int snoop_table_size; + atomic_t snoop_count; +}; + +struct ib_mad_port_private { + struct list_head port_list; + struct ib_device *device; + int port_num; + struct ib_cq *cq; + struct ib_pd *pd; + + spinlock_t reg_lock; + struct ib_mad_mgmt_version_table version[MAX_MGMT_VERSION]; + struct workqueue_struct *wq; + struct ib_mad_qp_info qp_info[IB_MAD_QPS_CORE]; +}; + +int ib_send_mad(struct ib_mad_send_wr_private *mad_send_wr); + +struct ib_mad_send_wr_private * +ib_find_send_mad(const struct ib_mad_agent_private *mad_agent_priv, + const struct ib_mad_recv_wc *mad_recv_wc); + +void ib_mad_complete_send_wr(struct ib_mad_send_wr_private *mad_send_wr, + struct ib_mad_send_wc *mad_send_wc); + +void ib_mark_mad_done(struct ib_mad_send_wr_private *mad_send_wr); + +void ib_reset_mad_timeout(struct ib_mad_send_wr_private *mad_send_wr, + unsigned long timeout_ms); + +#endif /* __IB_MAD_PRIV_H__ */ diff --git a/drivers/infiniband/core/mad_rmpp.c b/drivers/infiniband/core/mad_rmpp.c new file mode 100644 index 000000000..e0573e4d0 --- /dev/null +++ b/drivers/infiniband/core/mad_rmpp.c @@ -0,0 +1,960 @@ +/* + * Copyright (c) 2005 Intel Inc. All rights reserved. + * Copyright (c) 2005-2006 Voltaire, Inc. All rights reserved. + * Copyright (c) 2014 Intel Corporation. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include <linux/slab.h> + +#include "mad_priv.h" +#include "mad_rmpp.h" + +enum rmpp_state { + RMPP_STATE_ACTIVE, + RMPP_STATE_TIMEOUT, + RMPP_STATE_COMPLETE +}; + +struct mad_rmpp_recv { + struct ib_mad_agent_private *agent; + struct list_head list; + struct delayed_work timeout_work; + struct delayed_work cleanup_work; + struct completion comp; + enum rmpp_state state; + spinlock_t lock; + refcount_t refcount; + + struct ib_ah *ah; + struct ib_mad_recv_wc *rmpp_wc; + struct ib_mad_recv_buf *cur_seg_buf; + int last_ack; + int seg_num; + int newwin; + int repwin; + + __be64 tid; + u32 src_qp; + u32 slid; + u8 mgmt_class; + u8 class_version; + u8 method; + u8 base_version; +}; + +static inline void deref_rmpp_recv(struct mad_rmpp_recv *rmpp_recv) +{ + if (refcount_dec_and_test(&rmpp_recv->refcount)) + complete(&rmpp_recv->comp); +} + +static void destroy_rmpp_recv(struct mad_rmpp_recv *rmpp_recv) +{ + deref_rmpp_recv(rmpp_recv); + wait_for_completion(&rmpp_recv->comp); + rdma_destroy_ah(rmpp_recv->ah, RDMA_DESTROY_AH_SLEEPABLE); + kfree(rmpp_recv); +} + +void ib_cancel_rmpp_recvs(struct ib_mad_agent_private *agent) +{ + struct mad_rmpp_recv *rmpp_recv, *temp_rmpp_recv; + unsigned long flags; + + spin_lock_irqsave(&agent->lock, flags); + list_for_each_entry(rmpp_recv, &agent->rmpp_list, list) { + cancel_delayed_work(&rmpp_recv->timeout_work); + cancel_delayed_work(&rmpp_recv->cleanup_work); + } + spin_unlock_irqrestore(&agent->lock, flags); + + flush_workqueue(agent->qp_info->port_priv->wq); + + list_for_each_entry_safe(rmpp_recv, temp_rmpp_recv, + &agent->rmpp_list, list) { + list_del(&rmpp_recv->list); + if (rmpp_recv->state != RMPP_STATE_COMPLETE) + ib_free_recv_mad(rmpp_recv->rmpp_wc); + destroy_rmpp_recv(rmpp_recv); + } +} + +static void format_ack(struct ib_mad_send_buf *msg, + struct ib_rmpp_mad *data, + struct mad_rmpp_recv *rmpp_recv) +{ + struct ib_rmpp_mad *ack = msg->mad; + unsigned long flags; + + memcpy(ack, &data->mad_hdr, msg->hdr_len); + + ack->mad_hdr.method ^= IB_MGMT_METHOD_RESP; + ack->rmpp_hdr.rmpp_type = IB_MGMT_RMPP_TYPE_ACK; + ib_set_rmpp_flags(&ack->rmpp_hdr, IB_MGMT_RMPP_FLAG_ACTIVE); + + spin_lock_irqsave(&rmpp_recv->lock, flags); + rmpp_recv->last_ack = rmpp_recv->seg_num; + ack->rmpp_hdr.seg_num = cpu_to_be32(rmpp_recv->seg_num); + ack->rmpp_hdr.paylen_newwin = cpu_to_be32(rmpp_recv->newwin); + spin_unlock_irqrestore(&rmpp_recv->lock, flags); +} + +static void ack_recv(struct mad_rmpp_recv *rmpp_recv, + struct ib_mad_recv_wc *recv_wc) +{ + struct ib_mad_send_buf *msg; + int ret, hdr_len; + + hdr_len = ib_get_mad_data_offset(recv_wc->recv_buf.mad->mad_hdr.mgmt_class); + msg = ib_create_send_mad(&rmpp_recv->agent->agent, recv_wc->wc->src_qp, + recv_wc->wc->pkey_index, 1, hdr_len, + 0, GFP_KERNEL, + IB_MGMT_BASE_VERSION); + if (IS_ERR(msg)) + return; + + format_ack(msg, (struct ib_rmpp_mad *) recv_wc->recv_buf.mad, rmpp_recv); + msg->ah = rmpp_recv->ah; + ret = ib_post_send_mad(msg, NULL); + if (ret) + ib_free_send_mad(msg); +} + +static struct ib_mad_send_buf *alloc_response_msg(struct ib_mad_agent *agent, + struct ib_mad_recv_wc *recv_wc) +{ + struct ib_mad_send_buf *msg; + struct ib_ah *ah; + int hdr_len; + + ah = ib_create_ah_from_wc(agent->qp->pd, recv_wc->wc, + recv_wc->recv_buf.grh, agent->port_num); + if (IS_ERR(ah)) + return (void *) ah; + + hdr_len = ib_get_mad_data_offset(recv_wc->recv_buf.mad->mad_hdr.mgmt_class); + msg = ib_create_send_mad(agent, recv_wc->wc->src_qp, + recv_wc->wc->pkey_index, 1, + hdr_len, 0, GFP_KERNEL, + IB_MGMT_BASE_VERSION); + if (IS_ERR(msg)) + rdma_destroy_ah(ah, RDMA_DESTROY_AH_SLEEPABLE); + else { + msg->ah = ah; + msg->context[0] = ah; + } + + return msg; +} + +static void ack_ds_ack(struct ib_mad_agent_private *agent, + struct ib_mad_recv_wc *recv_wc) +{ + struct ib_mad_send_buf *msg; + struct ib_rmpp_mad *rmpp_mad; + int ret; + + msg = alloc_response_msg(&agent->agent, recv_wc); + if (IS_ERR(msg)) + return; + + rmpp_mad = msg->mad; + memcpy(rmpp_mad, recv_wc->recv_buf.mad, msg->hdr_len); + + rmpp_mad->mad_hdr.method ^= IB_MGMT_METHOD_RESP; + ib_set_rmpp_flags(&rmpp_mad->rmpp_hdr, IB_MGMT_RMPP_FLAG_ACTIVE); + rmpp_mad->rmpp_hdr.seg_num = 0; + rmpp_mad->rmpp_hdr.paylen_newwin = cpu_to_be32(1); + + ret = ib_post_send_mad(msg, NULL); + if (ret) { + rdma_destroy_ah(msg->ah, RDMA_DESTROY_AH_SLEEPABLE); + ib_free_send_mad(msg); + } +} + +void ib_rmpp_send_handler(struct ib_mad_send_wc *mad_send_wc) +{ + if (mad_send_wc->send_buf->context[0] == mad_send_wc->send_buf->ah) + rdma_destroy_ah(mad_send_wc->send_buf->ah, + RDMA_DESTROY_AH_SLEEPABLE); + ib_free_send_mad(mad_send_wc->send_buf); +} + +static void nack_recv(struct ib_mad_agent_private *agent, + struct ib_mad_recv_wc *recv_wc, u8 rmpp_status) +{ + struct ib_mad_send_buf *msg; + struct ib_rmpp_mad *rmpp_mad; + int ret; + + msg = alloc_response_msg(&agent->agent, recv_wc); + if (IS_ERR(msg)) + return; + + rmpp_mad = msg->mad; + memcpy(rmpp_mad, recv_wc->recv_buf.mad, msg->hdr_len); + + rmpp_mad->mad_hdr.method ^= IB_MGMT_METHOD_RESP; + rmpp_mad->rmpp_hdr.rmpp_version = IB_MGMT_RMPP_VERSION; + rmpp_mad->rmpp_hdr.rmpp_type = IB_MGMT_RMPP_TYPE_ABORT; + ib_set_rmpp_flags(&rmpp_mad->rmpp_hdr, IB_MGMT_RMPP_FLAG_ACTIVE); + rmpp_mad->rmpp_hdr.rmpp_status = rmpp_status; + rmpp_mad->rmpp_hdr.seg_num = 0; + rmpp_mad->rmpp_hdr.paylen_newwin = 0; + + ret = ib_post_send_mad(msg, NULL); + if (ret) { + rdma_destroy_ah(msg->ah, RDMA_DESTROY_AH_SLEEPABLE); + ib_free_send_mad(msg); + } +} + +static void recv_timeout_handler(struct work_struct *work) +{ + struct mad_rmpp_recv *rmpp_recv = + container_of(work, struct mad_rmpp_recv, timeout_work.work); + struct ib_mad_recv_wc *rmpp_wc; + unsigned long flags; + + spin_lock_irqsave(&rmpp_recv->agent->lock, flags); + if (rmpp_recv->state != RMPP_STATE_ACTIVE) { + spin_unlock_irqrestore(&rmpp_recv->agent->lock, flags); + return; + } + rmpp_recv->state = RMPP_STATE_TIMEOUT; + list_del(&rmpp_recv->list); + spin_unlock_irqrestore(&rmpp_recv->agent->lock, flags); + + rmpp_wc = rmpp_recv->rmpp_wc; + nack_recv(rmpp_recv->agent, rmpp_wc, IB_MGMT_RMPP_STATUS_T2L); + destroy_rmpp_recv(rmpp_recv); + ib_free_recv_mad(rmpp_wc); +} + +static void recv_cleanup_handler(struct work_struct *work) +{ + struct mad_rmpp_recv *rmpp_recv = + container_of(work, struct mad_rmpp_recv, cleanup_work.work); + unsigned long flags; + + spin_lock_irqsave(&rmpp_recv->agent->lock, flags); + list_del(&rmpp_recv->list); + spin_unlock_irqrestore(&rmpp_recv->agent->lock, flags); + destroy_rmpp_recv(rmpp_recv); +} + +static struct mad_rmpp_recv * +create_rmpp_recv(struct ib_mad_agent_private *agent, + struct ib_mad_recv_wc *mad_recv_wc) +{ + struct mad_rmpp_recv *rmpp_recv; + struct ib_mad_hdr *mad_hdr; + + rmpp_recv = kmalloc(sizeof *rmpp_recv, GFP_KERNEL); + if (!rmpp_recv) + return NULL; + + rmpp_recv->ah = ib_create_ah_from_wc(agent->agent.qp->pd, + mad_recv_wc->wc, + mad_recv_wc->recv_buf.grh, + agent->agent.port_num); + if (IS_ERR(rmpp_recv->ah)) + goto error; + + rmpp_recv->agent = agent; + init_completion(&rmpp_recv->comp); + INIT_DELAYED_WORK(&rmpp_recv->timeout_work, recv_timeout_handler); + INIT_DELAYED_WORK(&rmpp_recv->cleanup_work, recv_cleanup_handler); + spin_lock_init(&rmpp_recv->lock); + rmpp_recv->state = RMPP_STATE_ACTIVE; + refcount_set(&rmpp_recv->refcount, 1); + + rmpp_recv->rmpp_wc = mad_recv_wc; + rmpp_recv->cur_seg_buf = &mad_recv_wc->recv_buf; + rmpp_recv->newwin = 1; + rmpp_recv->seg_num = 1; + rmpp_recv->last_ack = 0; + rmpp_recv->repwin = 1; + + mad_hdr = &mad_recv_wc->recv_buf.mad->mad_hdr; + rmpp_recv->tid = mad_hdr->tid; + rmpp_recv->src_qp = mad_recv_wc->wc->src_qp; + rmpp_recv->slid = mad_recv_wc->wc->slid; + rmpp_recv->mgmt_class = mad_hdr->mgmt_class; + rmpp_recv->class_version = mad_hdr->class_version; + rmpp_recv->method = mad_hdr->method; + rmpp_recv->base_version = mad_hdr->base_version; + return rmpp_recv; + +error: kfree(rmpp_recv); + return NULL; +} + +static struct mad_rmpp_recv * +find_rmpp_recv(struct ib_mad_agent_private *agent, + struct ib_mad_recv_wc *mad_recv_wc) +{ + struct mad_rmpp_recv *rmpp_recv; + struct ib_mad_hdr *mad_hdr = &mad_recv_wc->recv_buf.mad->mad_hdr; + + list_for_each_entry(rmpp_recv, &agent->rmpp_list, list) { + if (rmpp_recv->tid == mad_hdr->tid && + rmpp_recv->src_qp == mad_recv_wc->wc->src_qp && + rmpp_recv->slid == mad_recv_wc->wc->slid && + rmpp_recv->mgmt_class == mad_hdr->mgmt_class && + rmpp_recv->class_version == mad_hdr->class_version && + rmpp_recv->method == mad_hdr->method) + return rmpp_recv; + } + return NULL; +} + +static struct mad_rmpp_recv * +acquire_rmpp_recv(struct ib_mad_agent_private *agent, + struct ib_mad_recv_wc *mad_recv_wc) +{ + struct mad_rmpp_recv *rmpp_recv; + unsigned long flags; + + spin_lock_irqsave(&agent->lock, flags); + rmpp_recv = find_rmpp_recv(agent, mad_recv_wc); + if (rmpp_recv) + refcount_inc(&rmpp_recv->refcount); + spin_unlock_irqrestore(&agent->lock, flags); + return rmpp_recv; +} + +static struct mad_rmpp_recv * +insert_rmpp_recv(struct ib_mad_agent_private *agent, + struct mad_rmpp_recv *rmpp_recv) +{ + struct mad_rmpp_recv *cur_rmpp_recv; + + cur_rmpp_recv = find_rmpp_recv(agent, rmpp_recv->rmpp_wc); + if (!cur_rmpp_recv) + list_add_tail(&rmpp_recv->list, &agent->rmpp_list); + + return cur_rmpp_recv; +} + +static inline int get_last_flag(struct ib_mad_recv_buf *seg) +{ + struct ib_rmpp_mad *rmpp_mad; + + rmpp_mad = (struct ib_rmpp_mad *) seg->mad; + return ib_get_rmpp_flags(&rmpp_mad->rmpp_hdr) & IB_MGMT_RMPP_FLAG_LAST; +} + +static inline int get_seg_num(struct ib_mad_recv_buf *seg) +{ + struct ib_rmpp_mad *rmpp_mad; + + rmpp_mad = (struct ib_rmpp_mad *) seg->mad; + return be32_to_cpu(rmpp_mad->rmpp_hdr.seg_num); +} + +static inline struct ib_mad_recv_buf * get_next_seg(struct list_head *rmpp_list, + struct ib_mad_recv_buf *seg) +{ + if (seg->list.next == rmpp_list) + return NULL; + + return container_of(seg->list.next, struct ib_mad_recv_buf, list); +} + +static inline int window_size(struct ib_mad_agent_private *agent) +{ + return max(agent->qp_info->recv_queue.max_active >> 3, 1); +} + +static struct ib_mad_recv_buf * find_seg_location(struct list_head *rmpp_list, + int seg_num) +{ + struct ib_mad_recv_buf *seg_buf; + int cur_seg_num; + + list_for_each_entry_reverse(seg_buf, rmpp_list, list) { + cur_seg_num = get_seg_num(seg_buf); + if (seg_num > cur_seg_num) + return seg_buf; + if (seg_num == cur_seg_num) + break; + } + return NULL; +} + +static void update_seg_num(struct mad_rmpp_recv *rmpp_recv, + struct ib_mad_recv_buf *new_buf) +{ + struct list_head *rmpp_list = &rmpp_recv->rmpp_wc->rmpp_list; + + while (new_buf && (get_seg_num(new_buf) == rmpp_recv->seg_num + 1)) { + rmpp_recv->cur_seg_buf = new_buf; + rmpp_recv->seg_num++; + new_buf = get_next_seg(rmpp_list, new_buf); + } +} + +static inline int get_mad_len(struct mad_rmpp_recv *rmpp_recv) +{ + struct ib_rmpp_mad *rmpp_mad; + int hdr_size, data_size, pad; + bool opa = rdma_cap_opa_mad(rmpp_recv->agent->qp_info->port_priv->device, + rmpp_recv->agent->qp_info->port_priv->port_num); + + rmpp_mad = (struct ib_rmpp_mad *)rmpp_recv->cur_seg_buf->mad; + + hdr_size = ib_get_mad_data_offset(rmpp_mad->mad_hdr.mgmt_class); + if (opa && rmpp_recv->base_version == OPA_MGMT_BASE_VERSION) { + data_size = sizeof(struct opa_rmpp_mad) - hdr_size; + pad = OPA_MGMT_RMPP_DATA - be32_to_cpu(rmpp_mad->rmpp_hdr.paylen_newwin); + if (pad > OPA_MGMT_RMPP_DATA || pad < 0) + pad = 0; + } else { + data_size = sizeof(struct ib_rmpp_mad) - hdr_size; + pad = IB_MGMT_RMPP_DATA - be32_to_cpu(rmpp_mad->rmpp_hdr.paylen_newwin); + if (pad > IB_MGMT_RMPP_DATA || pad < 0) + pad = 0; + } + + return hdr_size + rmpp_recv->seg_num * data_size - pad; +} + +static struct ib_mad_recv_wc * complete_rmpp(struct mad_rmpp_recv *rmpp_recv) +{ + struct ib_mad_recv_wc *rmpp_wc; + + ack_recv(rmpp_recv, rmpp_recv->rmpp_wc); + if (rmpp_recv->seg_num > 1) + cancel_delayed_work(&rmpp_recv->timeout_work); + + rmpp_wc = rmpp_recv->rmpp_wc; + rmpp_wc->mad_len = get_mad_len(rmpp_recv); + /* 10 seconds until we can find the packet lifetime */ + queue_delayed_work(rmpp_recv->agent->qp_info->port_priv->wq, + &rmpp_recv->cleanup_work, msecs_to_jiffies(10000)); + return rmpp_wc; +} + +static struct ib_mad_recv_wc * +continue_rmpp(struct ib_mad_agent_private *agent, + struct ib_mad_recv_wc *mad_recv_wc) +{ + struct mad_rmpp_recv *rmpp_recv; + struct ib_mad_recv_buf *prev_buf; + struct ib_mad_recv_wc *done_wc; + int seg_num; + unsigned long flags; + + rmpp_recv = acquire_rmpp_recv(agent, mad_recv_wc); + if (!rmpp_recv) + goto drop1; + + seg_num = get_seg_num(&mad_recv_wc->recv_buf); + + spin_lock_irqsave(&rmpp_recv->lock, flags); + if ((rmpp_recv->state == RMPP_STATE_TIMEOUT) || + (seg_num > rmpp_recv->newwin)) + goto drop3; + + if ((seg_num <= rmpp_recv->last_ack) || + (rmpp_recv->state == RMPP_STATE_COMPLETE)) { + spin_unlock_irqrestore(&rmpp_recv->lock, flags); + ack_recv(rmpp_recv, mad_recv_wc); + goto drop2; + } + + prev_buf = find_seg_location(&rmpp_recv->rmpp_wc->rmpp_list, seg_num); + if (!prev_buf) + goto drop3; + + done_wc = NULL; + list_add(&mad_recv_wc->recv_buf.list, &prev_buf->list); + if (rmpp_recv->cur_seg_buf == prev_buf) { + update_seg_num(rmpp_recv, &mad_recv_wc->recv_buf); + if (get_last_flag(rmpp_recv->cur_seg_buf)) { + rmpp_recv->state = RMPP_STATE_COMPLETE; + spin_unlock_irqrestore(&rmpp_recv->lock, flags); + done_wc = complete_rmpp(rmpp_recv); + goto out; + } else if (rmpp_recv->seg_num == rmpp_recv->newwin) { + rmpp_recv->newwin += window_size(agent); + spin_unlock_irqrestore(&rmpp_recv->lock, flags); + ack_recv(rmpp_recv, mad_recv_wc); + goto out; + } + } + spin_unlock_irqrestore(&rmpp_recv->lock, flags); +out: + deref_rmpp_recv(rmpp_recv); + return done_wc; + +drop3: spin_unlock_irqrestore(&rmpp_recv->lock, flags); +drop2: deref_rmpp_recv(rmpp_recv); +drop1: ib_free_recv_mad(mad_recv_wc); + return NULL; +} + +static struct ib_mad_recv_wc * +start_rmpp(struct ib_mad_agent_private *agent, + struct ib_mad_recv_wc *mad_recv_wc) +{ + struct mad_rmpp_recv *rmpp_recv; + unsigned long flags; + + rmpp_recv = create_rmpp_recv(agent, mad_recv_wc); + if (!rmpp_recv) { + ib_free_recv_mad(mad_recv_wc); + return NULL; + } + + spin_lock_irqsave(&agent->lock, flags); + if (insert_rmpp_recv(agent, rmpp_recv)) { + spin_unlock_irqrestore(&agent->lock, flags); + /* duplicate first MAD */ + destroy_rmpp_recv(rmpp_recv); + return continue_rmpp(agent, mad_recv_wc); + } + refcount_inc(&rmpp_recv->refcount); + + if (get_last_flag(&mad_recv_wc->recv_buf)) { + rmpp_recv->state = RMPP_STATE_COMPLETE; + spin_unlock_irqrestore(&agent->lock, flags); + complete_rmpp(rmpp_recv); + } else { + spin_unlock_irqrestore(&agent->lock, flags); + /* 40 seconds until we can find the packet lifetimes */ + queue_delayed_work(agent->qp_info->port_priv->wq, + &rmpp_recv->timeout_work, + msecs_to_jiffies(40000)); + rmpp_recv->newwin += window_size(agent); + ack_recv(rmpp_recv, mad_recv_wc); + mad_recv_wc = NULL; + } + deref_rmpp_recv(rmpp_recv); + return mad_recv_wc; +} + +static int send_next_seg(struct ib_mad_send_wr_private *mad_send_wr) +{ + struct ib_rmpp_mad *rmpp_mad; + int timeout; + u32 paylen = 0; + + rmpp_mad = mad_send_wr->send_buf.mad; + ib_set_rmpp_flags(&rmpp_mad->rmpp_hdr, IB_MGMT_RMPP_FLAG_ACTIVE); + rmpp_mad->rmpp_hdr.seg_num = cpu_to_be32(++mad_send_wr->seg_num); + + if (mad_send_wr->seg_num == 1) { + rmpp_mad->rmpp_hdr.rmpp_rtime_flags |= IB_MGMT_RMPP_FLAG_FIRST; + paylen = (mad_send_wr->send_buf.seg_count * + mad_send_wr->send_buf.seg_rmpp_size) - + mad_send_wr->pad; + } + + if (mad_send_wr->seg_num == mad_send_wr->send_buf.seg_count) { + rmpp_mad->rmpp_hdr.rmpp_rtime_flags |= IB_MGMT_RMPP_FLAG_LAST; + paylen = mad_send_wr->send_buf.seg_rmpp_size - mad_send_wr->pad; + } + rmpp_mad->rmpp_hdr.paylen_newwin = cpu_to_be32(paylen); + + /* 2 seconds for an ACK until we can find the packet lifetime */ + timeout = mad_send_wr->send_buf.timeout_ms; + if (!timeout || timeout > 2000) + mad_send_wr->timeout = msecs_to_jiffies(2000); + + return ib_send_mad(mad_send_wr); +} + +static void abort_send(struct ib_mad_agent_private *agent, + struct ib_mad_recv_wc *mad_recv_wc, u8 rmpp_status) +{ + struct ib_mad_send_wr_private *mad_send_wr; + struct ib_mad_send_wc wc; + unsigned long flags; + + spin_lock_irqsave(&agent->lock, flags); + mad_send_wr = ib_find_send_mad(agent, mad_recv_wc); + if (!mad_send_wr) + goto out; /* Unmatched send */ + + if ((mad_send_wr->last_ack == mad_send_wr->send_buf.seg_count) || + (!mad_send_wr->timeout) || (mad_send_wr->status != IB_WC_SUCCESS)) + goto out; /* Send is already done */ + + ib_mark_mad_done(mad_send_wr); + spin_unlock_irqrestore(&agent->lock, flags); + + wc.status = IB_WC_REM_ABORT_ERR; + wc.vendor_err = rmpp_status; + wc.send_buf = &mad_send_wr->send_buf; + ib_mad_complete_send_wr(mad_send_wr, &wc); + return; +out: + spin_unlock_irqrestore(&agent->lock, flags); +} + +static inline void adjust_last_ack(struct ib_mad_send_wr_private *wr, + int seg_num) +{ + struct list_head *list; + + wr->last_ack = seg_num; + list = &wr->last_ack_seg->list; + list_for_each_entry(wr->last_ack_seg, list, list) + if (wr->last_ack_seg->num == seg_num) + break; +} + +static void process_ds_ack(struct ib_mad_agent_private *agent, + struct ib_mad_recv_wc *mad_recv_wc, int newwin) +{ + struct mad_rmpp_recv *rmpp_recv; + + rmpp_recv = find_rmpp_recv(agent, mad_recv_wc); + if (rmpp_recv && rmpp_recv->state == RMPP_STATE_COMPLETE) + rmpp_recv->repwin = newwin; +} + +static void process_rmpp_ack(struct ib_mad_agent_private *agent, + struct ib_mad_recv_wc *mad_recv_wc) +{ + struct ib_mad_send_wr_private *mad_send_wr; + struct ib_rmpp_mad *rmpp_mad; + unsigned long flags; + int seg_num, newwin, ret; + + rmpp_mad = (struct ib_rmpp_mad *)mad_recv_wc->recv_buf.mad; + if (rmpp_mad->rmpp_hdr.rmpp_status) { + abort_send(agent, mad_recv_wc, IB_MGMT_RMPP_STATUS_BAD_STATUS); + nack_recv(agent, mad_recv_wc, IB_MGMT_RMPP_STATUS_BAD_STATUS); + return; + } + + seg_num = be32_to_cpu(rmpp_mad->rmpp_hdr.seg_num); + newwin = be32_to_cpu(rmpp_mad->rmpp_hdr.paylen_newwin); + if (newwin < seg_num) { + abort_send(agent, mad_recv_wc, IB_MGMT_RMPP_STATUS_W2S); + nack_recv(agent, mad_recv_wc, IB_MGMT_RMPP_STATUS_W2S); + return; + } + + spin_lock_irqsave(&agent->lock, flags); + mad_send_wr = ib_find_send_mad(agent, mad_recv_wc); + if (!mad_send_wr) { + if (!seg_num) + process_ds_ack(agent, mad_recv_wc, newwin); + goto out; /* Unmatched or DS RMPP ACK */ + } + + if ((mad_send_wr->last_ack == mad_send_wr->send_buf.seg_count) && + (mad_send_wr->timeout)) { + spin_unlock_irqrestore(&agent->lock, flags); + ack_ds_ack(agent, mad_recv_wc); + return; /* Repeated ACK for DS RMPP transaction */ + } + + if ((mad_send_wr->last_ack == mad_send_wr->send_buf.seg_count) || + (!mad_send_wr->timeout) || (mad_send_wr->status != IB_WC_SUCCESS)) + goto out; /* Send is already done */ + + if (seg_num > mad_send_wr->send_buf.seg_count || + seg_num > mad_send_wr->newwin) { + spin_unlock_irqrestore(&agent->lock, flags); + abort_send(agent, mad_recv_wc, IB_MGMT_RMPP_STATUS_S2B); + nack_recv(agent, mad_recv_wc, IB_MGMT_RMPP_STATUS_S2B); + return; + } + + if (newwin < mad_send_wr->newwin || seg_num < mad_send_wr->last_ack) + goto out; /* Old ACK */ + + if (seg_num > mad_send_wr->last_ack) { + adjust_last_ack(mad_send_wr, seg_num); + mad_send_wr->retries_left = mad_send_wr->max_retries; + } + mad_send_wr->newwin = newwin; + if (mad_send_wr->last_ack == mad_send_wr->send_buf.seg_count) { + /* If no response is expected, the ACK completes the send */ + if (!mad_send_wr->send_buf.timeout_ms) { + struct ib_mad_send_wc wc; + + ib_mark_mad_done(mad_send_wr); + spin_unlock_irqrestore(&agent->lock, flags); + + wc.status = IB_WC_SUCCESS; + wc.vendor_err = 0; + wc.send_buf = &mad_send_wr->send_buf; + ib_mad_complete_send_wr(mad_send_wr, &wc); + return; + } + if (mad_send_wr->refcount == 1) + ib_reset_mad_timeout(mad_send_wr, + mad_send_wr->send_buf.timeout_ms); + spin_unlock_irqrestore(&agent->lock, flags); + ack_ds_ack(agent, mad_recv_wc); + return; + } else if (mad_send_wr->refcount == 1 && + mad_send_wr->seg_num < mad_send_wr->newwin && + mad_send_wr->seg_num < mad_send_wr->send_buf.seg_count) { + /* Send failure will just result in a timeout/retry */ + ret = send_next_seg(mad_send_wr); + if (ret) + goto out; + + mad_send_wr->refcount++; + list_move_tail(&mad_send_wr->agent_list, + &mad_send_wr->mad_agent_priv->send_list); + } +out: + spin_unlock_irqrestore(&agent->lock, flags); +} + +static struct ib_mad_recv_wc * +process_rmpp_data(struct ib_mad_agent_private *agent, + struct ib_mad_recv_wc *mad_recv_wc) +{ + struct ib_rmpp_hdr *rmpp_hdr; + u8 rmpp_status; + + rmpp_hdr = &((struct ib_rmpp_mad *)mad_recv_wc->recv_buf.mad)->rmpp_hdr; + + if (rmpp_hdr->rmpp_status) { + rmpp_status = IB_MGMT_RMPP_STATUS_BAD_STATUS; + goto bad; + } + + if (rmpp_hdr->seg_num == cpu_to_be32(1)) { + if (!(ib_get_rmpp_flags(rmpp_hdr) & IB_MGMT_RMPP_FLAG_FIRST)) { + rmpp_status = IB_MGMT_RMPP_STATUS_BAD_SEG; + goto bad; + } + return start_rmpp(agent, mad_recv_wc); + } else { + if (ib_get_rmpp_flags(rmpp_hdr) & IB_MGMT_RMPP_FLAG_FIRST) { + rmpp_status = IB_MGMT_RMPP_STATUS_BAD_SEG; + goto bad; + } + return continue_rmpp(agent, mad_recv_wc); + } +bad: + nack_recv(agent, mad_recv_wc, rmpp_status); + ib_free_recv_mad(mad_recv_wc); + return NULL; +} + +static void process_rmpp_stop(struct ib_mad_agent_private *agent, + struct ib_mad_recv_wc *mad_recv_wc) +{ + struct ib_rmpp_mad *rmpp_mad; + + rmpp_mad = (struct ib_rmpp_mad *)mad_recv_wc->recv_buf.mad; + + if (rmpp_mad->rmpp_hdr.rmpp_status != IB_MGMT_RMPP_STATUS_RESX) { + abort_send(agent, mad_recv_wc, IB_MGMT_RMPP_STATUS_BAD_STATUS); + nack_recv(agent, mad_recv_wc, IB_MGMT_RMPP_STATUS_BAD_STATUS); + } else + abort_send(agent, mad_recv_wc, rmpp_mad->rmpp_hdr.rmpp_status); +} + +static void process_rmpp_abort(struct ib_mad_agent_private *agent, + struct ib_mad_recv_wc *mad_recv_wc) +{ + struct ib_rmpp_mad *rmpp_mad; + + rmpp_mad = (struct ib_rmpp_mad *)mad_recv_wc->recv_buf.mad; + + if (rmpp_mad->rmpp_hdr.rmpp_status < IB_MGMT_RMPP_STATUS_ABORT_MIN || + rmpp_mad->rmpp_hdr.rmpp_status > IB_MGMT_RMPP_STATUS_ABORT_MAX) { + abort_send(agent, mad_recv_wc, IB_MGMT_RMPP_STATUS_BAD_STATUS); + nack_recv(agent, mad_recv_wc, IB_MGMT_RMPP_STATUS_BAD_STATUS); + } else + abort_send(agent, mad_recv_wc, rmpp_mad->rmpp_hdr.rmpp_status); +} + +struct ib_mad_recv_wc * +ib_process_rmpp_recv_wc(struct ib_mad_agent_private *agent, + struct ib_mad_recv_wc *mad_recv_wc) +{ + struct ib_rmpp_mad *rmpp_mad; + + rmpp_mad = (struct ib_rmpp_mad *)mad_recv_wc->recv_buf.mad; + if (!(rmpp_mad->rmpp_hdr.rmpp_rtime_flags & IB_MGMT_RMPP_FLAG_ACTIVE)) + return mad_recv_wc; + + if (rmpp_mad->rmpp_hdr.rmpp_version != IB_MGMT_RMPP_VERSION) { + abort_send(agent, mad_recv_wc, IB_MGMT_RMPP_STATUS_UNV); + nack_recv(agent, mad_recv_wc, IB_MGMT_RMPP_STATUS_UNV); + goto out; + } + + switch (rmpp_mad->rmpp_hdr.rmpp_type) { + case IB_MGMT_RMPP_TYPE_DATA: + return process_rmpp_data(agent, mad_recv_wc); + case IB_MGMT_RMPP_TYPE_ACK: + process_rmpp_ack(agent, mad_recv_wc); + break; + case IB_MGMT_RMPP_TYPE_STOP: + process_rmpp_stop(agent, mad_recv_wc); + break; + case IB_MGMT_RMPP_TYPE_ABORT: + process_rmpp_abort(agent, mad_recv_wc); + break; + default: + abort_send(agent, mad_recv_wc, IB_MGMT_RMPP_STATUS_BADT); + nack_recv(agent, mad_recv_wc, IB_MGMT_RMPP_STATUS_BADT); + break; + } +out: + ib_free_recv_mad(mad_recv_wc); + return NULL; +} + +static int init_newwin(struct ib_mad_send_wr_private *mad_send_wr) +{ + struct ib_mad_agent_private *agent = mad_send_wr->mad_agent_priv; + struct ib_mad_hdr *mad_hdr = mad_send_wr->send_buf.mad; + struct mad_rmpp_recv *rmpp_recv; + struct rdma_ah_attr ah_attr; + unsigned long flags; + int newwin = 1; + + if (!(mad_hdr->method & IB_MGMT_METHOD_RESP)) + goto out; + + spin_lock_irqsave(&agent->lock, flags); + list_for_each_entry(rmpp_recv, &agent->rmpp_list, list) { + if (rmpp_recv->tid != mad_hdr->tid || + rmpp_recv->mgmt_class != mad_hdr->mgmt_class || + rmpp_recv->class_version != mad_hdr->class_version || + (rmpp_recv->method & IB_MGMT_METHOD_RESP)) + continue; + + if (rdma_query_ah(mad_send_wr->send_buf.ah, &ah_attr)) + continue; + + if (rmpp_recv->slid == rdma_ah_get_dlid(&ah_attr)) { + newwin = rmpp_recv->repwin; + break; + } + } + spin_unlock_irqrestore(&agent->lock, flags); +out: + return newwin; +} + +int ib_send_rmpp_mad(struct ib_mad_send_wr_private *mad_send_wr) +{ + struct ib_rmpp_mad *rmpp_mad; + int ret; + + rmpp_mad = mad_send_wr->send_buf.mad; + if (!(ib_get_rmpp_flags(&rmpp_mad->rmpp_hdr) & + IB_MGMT_RMPP_FLAG_ACTIVE)) + return IB_RMPP_RESULT_UNHANDLED; + + if (rmpp_mad->rmpp_hdr.rmpp_type != IB_MGMT_RMPP_TYPE_DATA) { + mad_send_wr->seg_num = 1; + return IB_RMPP_RESULT_INTERNAL; + } + + mad_send_wr->newwin = init_newwin(mad_send_wr); + + /* We need to wait for the final ACK even if there isn't a response */ + mad_send_wr->refcount += (mad_send_wr->timeout == 0); + ret = send_next_seg(mad_send_wr); + if (!ret) + return IB_RMPP_RESULT_CONSUMED; + return ret; +} + +int ib_process_rmpp_send_wc(struct ib_mad_send_wr_private *mad_send_wr, + struct ib_mad_send_wc *mad_send_wc) +{ + struct ib_rmpp_mad *rmpp_mad; + int ret; + + rmpp_mad = mad_send_wr->send_buf.mad; + if (!(ib_get_rmpp_flags(&rmpp_mad->rmpp_hdr) & + IB_MGMT_RMPP_FLAG_ACTIVE)) + return IB_RMPP_RESULT_UNHANDLED; /* RMPP not active */ + + if (rmpp_mad->rmpp_hdr.rmpp_type != IB_MGMT_RMPP_TYPE_DATA) + return IB_RMPP_RESULT_INTERNAL; /* ACK, STOP, or ABORT */ + + if (mad_send_wc->status != IB_WC_SUCCESS || + mad_send_wr->status != IB_WC_SUCCESS) + return IB_RMPP_RESULT_PROCESSED; /* Canceled or send error */ + + if (!mad_send_wr->timeout) + return IB_RMPP_RESULT_PROCESSED; /* Response received */ + + if (mad_send_wr->last_ack == mad_send_wr->send_buf.seg_count) { + mad_send_wr->timeout = + msecs_to_jiffies(mad_send_wr->send_buf.timeout_ms); + return IB_RMPP_RESULT_PROCESSED; /* Send done */ + } + + if (mad_send_wr->seg_num == mad_send_wr->newwin || + mad_send_wr->seg_num == mad_send_wr->send_buf.seg_count) + return IB_RMPP_RESULT_PROCESSED; /* Wait for ACK */ + + ret = send_next_seg(mad_send_wr); + if (ret) { + mad_send_wc->status = IB_WC_GENERAL_ERR; + return IB_RMPP_RESULT_PROCESSED; + } + return IB_RMPP_RESULT_CONSUMED; +} + +int ib_retry_rmpp(struct ib_mad_send_wr_private *mad_send_wr) +{ + struct ib_rmpp_mad *rmpp_mad; + int ret; + + rmpp_mad = mad_send_wr->send_buf.mad; + if (!(ib_get_rmpp_flags(&rmpp_mad->rmpp_hdr) & + IB_MGMT_RMPP_FLAG_ACTIVE)) + return IB_RMPP_RESULT_UNHANDLED; /* RMPP not active */ + + if (mad_send_wr->last_ack == mad_send_wr->send_buf.seg_count) + return IB_RMPP_RESULT_PROCESSED; + + mad_send_wr->seg_num = mad_send_wr->last_ack; + mad_send_wr->cur_seg = mad_send_wr->last_ack_seg; + + ret = send_next_seg(mad_send_wr); + if (ret) + return IB_RMPP_RESULT_PROCESSED; + + return IB_RMPP_RESULT_CONSUMED; +} diff --git a/drivers/infiniband/core/mad_rmpp.h b/drivers/infiniband/core/mad_rmpp.h new file mode 100644 index 000000000..3d336bff1 --- /dev/null +++ b/drivers/infiniband/core/mad_rmpp.h @@ -0,0 +1,58 @@ +/* + * Copyright (c) 2005 Intel Inc. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#ifndef __MAD_RMPP_H__ +#define __MAD_RMPP_H__ + +enum { + IB_RMPP_RESULT_PROCESSED, + IB_RMPP_RESULT_CONSUMED, + IB_RMPP_RESULT_INTERNAL, + IB_RMPP_RESULT_UNHANDLED +}; + +int ib_send_rmpp_mad(struct ib_mad_send_wr_private *mad_send_wr); + +struct ib_mad_recv_wc * +ib_process_rmpp_recv_wc(struct ib_mad_agent_private *agent, + struct ib_mad_recv_wc *mad_recv_wc); + +int ib_process_rmpp_send_wc(struct ib_mad_send_wr_private *mad_send_wr, + struct ib_mad_send_wc *mad_send_wc); + +void ib_rmpp_send_handler(struct ib_mad_send_wc *mad_send_wc); + +void ib_cancel_rmpp_recvs(struct ib_mad_agent_private *agent); + +int ib_retry_rmpp(struct ib_mad_send_wr_private *mad_send_wr); + +#endif /* __MAD_RMPP_H__ */ diff --git a/drivers/infiniband/core/mr_pool.c b/drivers/infiniband/core/mr_pool.c new file mode 100644 index 000000000..c0e2df128 --- /dev/null +++ b/drivers/infiniband/core/mr_pool.c @@ -0,0 +1,82 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * Copyright (c) 2016 HGST, a Western Digital Company. + */ +#include <rdma/ib_verbs.h> +#include <rdma/mr_pool.h> + +struct ib_mr *ib_mr_pool_get(struct ib_qp *qp, struct list_head *list) +{ + struct ib_mr *mr; + unsigned long flags; + + spin_lock_irqsave(&qp->mr_lock, flags); + mr = list_first_entry_or_null(list, struct ib_mr, qp_entry); + if (mr) { + list_del(&mr->qp_entry); + qp->mrs_used++; + } + spin_unlock_irqrestore(&qp->mr_lock, flags); + + return mr; +} +EXPORT_SYMBOL(ib_mr_pool_get); + +void ib_mr_pool_put(struct ib_qp *qp, struct list_head *list, struct ib_mr *mr) +{ + unsigned long flags; + + spin_lock_irqsave(&qp->mr_lock, flags); + list_add(&mr->qp_entry, list); + qp->mrs_used--; + spin_unlock_irqrestore(&qp->mr_lock, flags); +} +EXPORT_SYMBOL(ib_mr_pool_put); + +int ib_mr_pool_init(struct ib_qp *qp, struct list_head *list, int nr, + enum ib_mr_type type, u32 max_num_sg, u32 max_num_meta_sg) +{ + struct ib_mr *mr; + unsigned long flags; + int ret, i; + + for (i = 0; i < nr; i++) { + if (type == IB_MR_TYPE_INTEGRITY) + mr = ib_alloc_mr_integrity(qp->pd, max_num_sg, + max_num_meta_sg); + else + mr = ib_alloc_mr(qp->pd, type, max_num_sg); + if (IS_ERR(mr)) { + ret = PTR_ERR(mr); + goto out; + } + + spin_lock_irqsave(&qp->mr_lock, flags); + list_add_tail(&mr->qp_entry, list); + spin_unlock_irqrestore(&qp->mr_lock, flags); + } + + return 0; +out: + ib_mr_pool_destroy(qp, list); + return ret; +} +EXPORT_SYMBOL(ib_mr_pool_init); + +void ib_mr_pool_destroy(struct ib_qp *qp, struct list_head *list) +{ + struct ib_mr *mr; + unsigned long flags; + + spin_lock_irqsave(&qp->mr_lock, flags); + while (!list_empty(list)) { + mr = list_first_entry(list, struct ib_mr, qp_entry); + list_del(&mr->qp_entry); + + spin_unlock_irqrestore(&qp->mr_lock, flags); + ib_dereg_mr(mr); + spin_lock_irqsave(&qp->mr_lock, flags); + } + spin_unlock_irqrestore(&qp->mr_lock, flags); +} +EXPORT_SYMBOL(ib_mr_pool_destroy); diff --git a/drivers/infiniband/core/multicast.c b/drivers/infiniband/core/multicast.c new file mode 100644 index 000000000..740f03ecc --- /dev/null +++ b/drivers/infiniband/core/multicast.c @@ -0,0 +1,905 @@ +/* + * Copyright (c) 2006 Intel Corporation. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include <linux/completion.h> +#include <linux/dma-mapping.h> +#include <linux/err.h> +#include <linux/interrupt.h> +#include <linux/export.h> +#include <linux/slab.h> +#include <linux/bitops.h> +#include <linux/random.h> + +#include <rdma/ib_cache.h> +#include "sa.h" + +static int mcast_add_one(struct ib_device *device); +static void mcast_remove_one(struct ib_device *device, void *client_data); + +static struct ib_client mcast_client = { + .name = "ib_multicast", + .add = mcast_add_one, + .remove = mcast_remove_one +}; + +static struct ib_sa_client sa_client; +static struct workqueue_struct *mcast_wq; +static union ib_gid mgid0; + +struct mcast_device; + +struct mcast_port { + struct mcast_device *dev; + spinlock_t lock; + struct rb_root table; + atomic_t refcount; + struct completion comp; + u8 port_num; +}; + +struct mcast_device { + struct ib_device *device; + struct ib_event_handler event_handler; + int start_port; + int end_port; + struct mcast_port port[]; +}; + +enum mcast_state { + MCAST_JOINING, + MCAST_MEMBER, + MCAST_ERROR, +}; + +enum mcast_group_state { + MCAST_IDLE, + MCAST_BUSY, + MCAST_GROUP_ERROR, + MCAST_PKEY_EVENT +}; + +enum { + MCAST_INVALID_PKEY_INDEX = 0xFFFF +}; + +struct mcast_member; + +struct mcast_group { + struct ib_sa_mcmember_rec rec; + struct rb_node node; + struct mcast_port *port; + spinlock_t lock; + struct work_struct work; + struct list_head pending_list; + struct list_head active_list; + struct mcast_member *last_join; + int members[NUM_JOIN_MEMBERSHIP_TYPES]; + atomic_t refcount; + enum mcast_group_state state; + struct ib_sa_query *query; + u16 pkey_index; + u8 leave_state; + int retries; +}; + +struct mcast_member { + struct ib_sa_multicast multicast; + struct ib_sa_client *client; + struct mcast_group *group; + struct list_head list; + enum mcast_state state; + atomic_t refcount; + struct completion comp; +}; + +static void join_handler(int status, struct ib_sa_mcmember_rec *rec, + void *context); +static void leave_handler(int status, struct ib_sa_mcmember_rec *rec, + void *context); + +static struct mcast_group *mcast_find(struct mcast_port *port, + union ib_gid *mgid) +{ + struct rb_node *node = port->table.rb_node; + struct mcast_group *group; + int ret; + + while (node) { + group = rb_entry(node, struct mcast_group, node); + ret = memcmp(mgid->raw, group->rec.mgid.raw, sizeof *mgid); + if (!ret) + return group; + + if (ret < 0) + node = node->rb_left; + else + node = node->rb_right; + } + return NULL; +} + +static struct mcast_group *mcast_insert(struct mcast_port *port, + struct mcast_group *group, + int allow_duplicates) +{ + struct rb_node **link = &port->table.rb_node; + struct rb_node *parent = NULL; + struct mcast_group *cur_group; + int ret; + + while (*link) { + parent = *link; + cur_group = rb_entry(parent, struct mcast_group, node); + + ret = memcmp(group->rec.mgid.raw, cur_group->rec.mgid.raw, + sizeof group->rec.mgid); + if (ret < 0) + link = &(*link)->rb_left; + else if (ret > 0) + link = &(*link)->rb_right; + else if (allow_duplicates) + link = &(*link)->rb_left; + else + return cur_group; + } + rb_link_node(&group->node, parent, link); + rb_insert_color(&group->node, &port->table); + return NULL; +} + +static void deref_port(struct mcast_port *port) +{ + if (atomic_dec_and_test(&port->refcount)) + complete(&port->comp); +} + +static void release_group(struct mcast_group *group) +{ + struct mcast_port *port = group->port; + unsigned long flags; + + spin_lock_irqsave(&port->lock, flags); + if (atomic_dec_and_test(&group->refcount)) { + rb_erase(&group->node, &port->table); + spin_unlock_irqrestore(&port->lock, flags); + kfree(group); + deref_port(port); + } else + spin_unlock_irqrestore(&port->lock, flags); +} + +static void deref_member(struct mcast_member *member) +{ + if (atomic_dec_and_test(&member->refcount)) + complete(&member->comp); +} + +static void queue_join(struct mcast_member *member) +{ + struct mcast_group *group = member->group; + unsigned long flags; + + spin_lock_irqsave(&group->lock, flags); + list_add_tail(&member->list, &group->pending_list); + if (group->state == MCAST_IDLE) { + group->state = MCAST_BUSY; + atomic_inc(&group->refcount); + queue_work(mcast_wq, &group->work); + } + spin_unlock_irqrestore(&group->lock, flags); +} + +/* + * A multicast group has four types of members: full member, non member, + * sendonly non member and sendonly full member. + * We need to keep track of the number of members of each + * type based on their join state. Adjust the number of members the belong to + * the specified join states. + */ +static void adjust_membership(struct mcast_group *group, u8 join_state, int inc) +{ + int i; + + for (i = 0; i < NUM_JOIN_MEMBERSHIP_TYPES; i++, join_state >>= 1) + if (join_state & 0x1) + group->members[i] += inc; +} + +/* + * If a multicast group has zero members left for a particular join state, but + * the group is still a member with the SA, we need to leave that join state. + * Determine which join states we still belong to, but that do not have any + * active members. + */ +static u8 get_leave_state(struct mcast_group *group) +{ + u8 leave_state = 0; + int i; + + for (i = 0; i < NUM_JOIN_MEMBERSHIP_TYPES; i++) + if (!group->members[i]) + leave_state |= (0x1 << i); + + return leave_state & group->rec.join_state; +} + +static int check_selector(ib_sa_comp_mask comp_mask, + ib_sa_comp_mask selector_mask, + ib_sa_comp_mask value_mask, + u8 selector, u8 src_value, u8 dst_value) +{ + int err; + + if (!(comp_mask & selector_mask) || !(comp_mask & value_mask)) + return 0; + + switch (selector) { + case IB_SA_GT: + err = (src_value <= dst_value); + break; + case IB_SA_LT: + err = (src_value >= dst_value); + break; + case IB_SA_EQ: + err = (src_value != dst_value); + break; + default: + err = 0; + break; + } + + return err; +} + +static int cmp_rec(struct ib_sa_mcmember_rec *src, + struct ib_sa_mcmember_rec *dst, ib_sa_comp_mask comp_mask) +{ + /* MGID must already match */ + + if (comp_mask & IB_SA_MCMEMBER_REC_PORT_GID && + memcmp(&src->port_gid, &dst->port_gid, sizeof src->port_gid)) + return -EINVAL; + if (comp_mask & IB_SA_MCMEMBER_REC_QKEY && src->qkey != dst->qkey) + return -EINVAL; + if (comp_mask & IB_SA_MCMEMBER_REC_MLID && src->mlid != dst->mlid) + return -EINVAL; + if (check_selector(comp_mask, IB_SA_MCMEMBER_REC_MTU_SELECTOR, + IB_SA_MCMEMBER_REC_MTU, dst->mtu_selector, + src->mtu, dst->mtu)) + return -EINVAL; + if (comp_mask & IB_SA_MCMEMBER_REC_TRAFFIC_CLASS && + src->traffic_class != dst->traffic_class) + return -EINVAL; + if (comp_mask & IB_SA_MCMEMBER_REC_PKEY && src->pkey != dst->pkey) + return -EINVAL; + if (check_selector(comp_mask, IB_SA_MCMEMBER_REC_RATE_SELECTOR, + IB_SA_MCMEMBER_REC_RATE, dst->rate_selector, + src->rate, dst->rate)) + return -EINVAL; + if (check_selector(comp_mask, + IB_SA_MCMEMBER_REC_PACKET_LIFE_TIME_SELECTOR, + IB_SA_MCMEMBER_REC_PACKET_LIFE_TIME, + dst->packet_life_time_selector, + src->packet_life_time, dst->packet_life_time)) + return -EINVAL; + if (comp_mask & IB_SA_MCMEMBER_REC_SL && src->sl != dst->sl) + return -EINVAL; + if (comp_mask & IB_SA_MCMEMBER_REC_FLOW_LABEL && + src->flow_label != dst->flow_label) + return -EINVAL; + if (comp_mask & IB_SA_MCMEMBER_REC_HOP_LIMIT && + src->hop_limit != dst->hop_limit) + return -EINVAL; + if (comp_mask & IB_SA_MCMEMBER_REC_SCOPE && src->scope != dst->scope) + return -EINVAL; + + /* join_state checked separately, proxy_join ignored */ + + return 0; +} + +static int send_join(struct mcast_group *group, struct mcast_member *member) +{ + struct mcast_port *port = group->port; + int ret; + + group->last_join = member; + ret = ib_sa_mcmember_rec_query(&sa_client, port->dev->device, + port->port_num, IB_MGMT_METHOD_SET, + &member->multicast.rec, + member->multicast.comp_mask, + 3000, GFP_KERNEL, join_handler, group, + &group->query); + return (ret > 0) ? 0 : ret; +} + +static int send_leave(struct mcast_group *group, u8 leave_state) +{ + struct mcast_port *port = group->port; + struct ib_sa_mcmember_rec rec; + int ret; + + rec = group->rec; + rec.join_state = leave_state; + group->leave_state = leave_state; + + ret = ib_sa_mcmember_rec_query(&sa_client, port->dev->device, + port->port_num, IB_SA_METHOD_DELETE, &rec, + IB_SA_MCMEMBER_REC_MGID | + IB_SA_MCMEMBER_REC_PORT_GID | + IB_SA_MCMEMBER_REC_JOIN_STATE, + 3000, GFP_KERNEL, leave_handler, + group, &group->query); + return (ret > 0) ? 0 : ret; +} + +static void join_group(struct mcast_group *group, struct mcast_member *member, + u8 join_state) +{ + member->state = MCAST_MEMBER; + adjust_membership(group, join_state, 1); + group->rec.join_state |= join_state; + member->multicast.rec = group->rec; + member->multicast.rec.join_state = join_state; + list_move(&member->list, &group->active_list); +} + +static int fail_join(struct mcast_group *group, struct mcast_member *member, + int status) +{ + spin_lock_irq(&group->lock); + list_del_init(&member->list); + spin_unlock_irq(&group->lock); + return member->multicast.callback(status, &member->multicast); +} + +static void process_group_error(struct mcast_group *group) +{ + struct mcast_member *member; + int ret = 0; + u16 pkey_index; + + if (group->state == MCAST_PKEY_EVENT) + ret = ib_find_pkey(group->port->dev->device, + group->port->port_num, + be16_to_cpu(group->rec.pkey), &pkey_index); + + spin_lock_irq(&group->lock); + if (group->state == MCAST_PKEY_EVENT && !ret && + group->pkey_index == pkey_index) + goto out; + + while (!list_empty(&group->active_list)) { + member = list_entry(group->active_list.next, + struct mcast_member, list); + atomic_inc(&member->refcount); + list_del_init(&member->list); + adjust_membership(group, member->multicast.rec.join_state, -1); + member->state = MCAST_ERROR; + spin_unlock_irq(&group->lock); + + ret = member->multicast.callback(-ENETRESET, + &member->multicast); + deref_member(member); + if (ret) + ib_sa_free_multicast(&member->multicast); + spin_lock_irq(&group->lock); + } + + group->rec.join_state = 0; +out: + group->state = MCAST_BUSY; + spin_unlock_irq(&group->lock); +} + +static void mcast_work_handler(struct work_struct *work) +{ + struct mcast_group *group; + struct mcast_member *member; + struct ib_sa_multicast *multicast; + int status, ret; + u8 join_state; + + group = container_of(work, typeof(*group), work); +retest: + spin_lock_irq(&group->lock); + while (!list_empty(&group->pending_list) || + (group->state != MCAST_BUSY)) { + + if (group->state != MCAST_BUSY) { + spin_unlock_irq(&group->lock); + process_group_error(group); + goto retest; + } + + member = list_entry(group->pending_list.next, + struct mcast_member, list); + multicast = &member->multicast; + join_state = multicast->rec.join_state; + atomic_inc(&member->refcount); + + if (join_state == (group->rec.join_state & join_state)) { + status = cmp_rec(&group->rec, &multicast->rec, + multicast->comp_mask); + if (!status) + join_group(group, member, join_state); + else + list_del_init(&member->list); + spin_unlock_irq(&group->lock); + ret = multicast->callback(status, multicast); + } else { + spin_unlock_irq(&group->lock); + status = send_join(group, member); + if (!status) { + deref_member(member); + return; + } + ret = fail_join(group, member, status); + } + + deref_member(member); + if (ret) + ib_sa_free_multicast(&member->multicast); + spin_lock_irq(&group->lock); + } + + join_state = get_leave_state(group); + if (join_state) { + group->rec.join_state &= ~join_state; + spin_unlock_irq(&group->lock); + if (send_leave(group, join_state)) + goto retest; + } else { + group->state = MCAST_IDLE; + spin_unlock_irq(&group->lock); + release_group(group); + } +} + +/* + * Fail a join request if it is still active - at the head of the pending queue. + */ +static void process_join_error(struct mcast_group *group, int status) +{ + struct mcast_member *member; + int ret; + + spin_lock_irq(&group->lock); + member = list_entry(group->pending_list.next, + struct mcast_member, list); + if (group->last_join == member) { + atomic_inc(&member->refcount); + list_del_init(&member->list); + spin_unlock_irq(&group->lock); + ret = member->multicast.callback(status, &member->multicast); + deref_member(member); + if (ret) + ib_sa_free_multicast(&member->multicast); + } else + spin_unlock_irq(&group->lock); +} + +static void join_handler(int status, struct ib_sa_mcmember_rec *rec, + void *context) +{ + struct mcast_group *group = context; + u16 pkey_index = MCAST_INVALID_PKEY_INDEX; + + if (status) + process_join_error(group, status); + else { + int mgids_changed, is_mgid0; + + if (ib_find_pkey(group->port->dev->device, + group->port->port_num, be16_to_cpu(rec->pkey), + &pkey_index)) + pkey_index = MCAST_INVALID_PKEY_INDEX; + + spin_lock_irq(&group->port->lock); + if (group->state == MCAST_BUSY && + group->pkey_index == MCAST_INVALID_PKEY_INDEX) + group->pkey_index = pkey_index; + mgids_changed = memcmp(&rec->mgid, &group->rec.mgid, + sizeof(group->rec.mgid)); + group->rec = *rec; + if (mgids_changed) { + rb_erase(&group->node, &group->port->table); + is_mgid0 = !memcmp(&mgid0, &group->rec.mgid, + sizeof(mgid0)); + mcast_insert(group->port, group, is_mgid0); + } + spin_unlock_irq(&group->port->lock); + } + mcast_work_handler(&group->work); +} + +static void leave_handler(int status, struct ib_sa_mcmember_rec *rec, + void *context) +{ + struct mcast_group *group = context; + + if (status && group->retries > 0 && + !send_leave(group, group->leave_state)) + group->retries--; + else + mcast_work_handler(&group->work); +} + +static struct mcast_group *acquire_group(struct mcast_port *port, + union ib_gid *mgid, gfp_t gfp_mask) +{ + struct mcast_group *group, *cur_group; + unsigned long flags; + int is_mgid0; + + is_mgid0 = !memcmp(&mgid0, mgid, sizeof mgid0); + if (!is_mgid0) { + spin_lock_irqsave(&port->lock, flags); + group = mcast_find(port, mgid); + if (group) + goto found; + spin_unlock_irqrestore(&port->lock, flags); + } + + group = kzalloc(sizeof *group, gfp_mask); + if (!group) + return NULL; + + group->retries = 3; + group->port = port; + group->rec.mgid = *mgid; + group->pkey_index = MCAST_INVALID_PKEY_INDEX; + INIT_LIST_HEAD(&group->pending_list); + INIT_LIST_HEAD(&group->active_list); + INIT_WORK(&group->work, mcast_work_handler); + spin_lock_init(&group->lock); + + spin_lock_irqsave(&port->lock, flags); + cur_group = mcast_insert(port, group, is_mgid0); + if (cur_group) { + kfree(group); + group = cur_group; + } else + atomic_inc(&port->refcount); +found: + atomic_inc(&group->refcount); + spin_unlock_irqrestore(&port->lock, flags); + return group; +} + +/* + * We serialize all join requests to a single group to make our lives much + * easier. Otherwise, two users could try to join the same group + * simultaneously, with different configurations, one could leave while the + * join is in progress, etc., which makes locking around error recovery + * difficult. + */ +struct ib_sa_multicast * +ib_sa_join_multicast(struct ib_sa_client *client, + struct ib_device *device, u8 port_num, + struct ib_sa_mcmember_rec *rec, + ib_sa_comp_mask comp_mask, gfp_t gfp_mask, + int (*callback)(int status, + struct ib_sa_multicast *multicast), + void *context) +{ + struct mcast_device *dev; + struct mcast_member *member; + struct ib_sa_multicast *multicast; + int ret; + + dev = ib_get_client_data(device, &mcast_client); + if (!dev) + return ERR_PTR(-ENODEV); + + member = kmalloc(sizeof *member, gfp_mask); + if (!member) + return ERR_PTR(-ENOMEM); + + ib_sa_client_get(client); + member->client = client; + member->multicast.rec = *rec; + member->multicast.comp_mask = comp_mask; + member->multicast.callback = callback; + member->multicast.context = context; + init_completion(&member->comp); + atomic_set(&member->refcount, 1); + member->state = MCAST_JOINING; + + member->group = acquire_group(&dev->port[port_num - dev->start_port], + &rec->mgid, gfp_mask); + if (!member->group) { + ret = -ENOMEM; + goto err; + } + + /* + * The user will get the multicast structure in their callback. They + * could then free the multicast structure before we can return from + * this routine. So we save the pointer to return before queuing + * any callback. + */ + multicast = &member->multicast; + queue_join(member); + return multicast; + +err: + ib_sa_client_put(client); + kfree(member); + return ERR_PTR(ret); +} +EXPORT_SYMBOL(ib_sa_join_multicast); + +void ib_sa_free_multicast(struct ib_sa_multicast *multicast) +{ + struct mcast_member *member; + struct mcast_group *group; + + member = container_of(multicast, struct mcast_member, multicast); + group = member->group; + + spin_lock_irq(&group->lock); + if (member->state == MCAST_MEMBER) + adjust_membership(group, multicast->rec.join_state, -1); + + list_del_init(&member->list); + + if (group->state == MCAST_IDLE) { + group->state = MCAST_BUSY; + spin_unlock_irq(&group->lock); + /* Continue to hold reference on group until callback */ + queue_work(mcast_wq, &group->work); + } else { + spin_unlock_irq(&group->lock); + release_group(group); + } + + deref_member(member); + wait_for_completion(&member->comp); + ib_sa_client_put(member->client); + kfree(member); +} +EXPORT_SYMBOL(ib_sa_free_multicast); + +int ib_sa_get_mcmember_rec(struct ib_device *device, u8 port_num, + union ib_gid *mgid, struct ib_sa_mcmember_rec *rec) +{ + struct mcast_device *dev; + struct mcast_port *port; + struct mcast_group *group; + unsigned long flags; + int ret = 0; + + dev = ib_get_client_data(device, &mcast_client); + if (!dev) + return -ENODEV; + + port = &dev->port[port_num - dev->start_port]; + spin_lock_irqsave(&port->lock, flags); + group = mcast_find(port, mgid); + if (group) + *rec = group->rec; + else + ret = -EADDRNOTAVAIL; + spin_unlock_irqrestore(&port->lock, flags); + + return ret; +} +EXPORT_SYMBOL(ib_sa_get_mcmember_rec); + +/** + * ib_init_ah_from_mcmember - Initialize AH attribute from multicast + * member record and gid of the device. + * @device: RDMA device + * @port_num: Port of the rdma device to consider + * @ndev: Optional netdevice, applicable only for RoCE + * @gid_type: GID type to consider + * @ah_attr: AH attribute to fillup on successful completion + * + * ib_init_ah_from_mcmember() initializes AH attribute based on multicast + * member record and other device properties. On success the caller is + * responsible to call rdma_destroy_ah_attr on the ah_attr. Returns 0 on + * success or appropriate error code. + * + */ +int ib_init_ah_from_mcmember(struct ib_device *device, u8 port_num, + struct ib_sa_mcmember_rec *rec, + struct net_device *ndev, + enum ib_gid_type gid_type, + struct rdma_ah_attr *ah_attr) +{ + const struct ib_gid_attr *sgid_attr; + + /* GID table is not based on the netdevice for IB link layer, + * so ignore ndev during search. + */ + if (rdma_protocol_ib(device, port_num)) + ndev = NULL; + else if (!rdma_protocol_roce(device, port_num)) + return -EINVAL; + + sgid_attr = rdma_find_gid_by_port(device, &rec->port_gid, + gid_type, port_num, ndev); + if (IS_ERR(sgid_attr)) + return PTR_ERR(sgid_attr); + + memset(ah_attr, 0, sizeof(*ah_attr)); + ah_attr->type = rdma_ah_find_type(device, port_num); + + rdma_ah_set_dlid(ah_attr, be16_to_cpu(rec->mlid)); + rdma_ah_set_sl(ah_attr, rec->sl); + rdma_ah_set_port_num(ah_attr, port_num); + rdma_ah_set_static_rate(ah_attr, rec->rate); + rdma_move_grh_sgid_attr(ah_attr, &rec->mgid, + be32_to_cpu(rec->flow_label), + rec->hop_limit, rec->traffic_class, + sgid_attr); + return 0; +} +EXPORT_SYMBOL(ib_init_ah_from_mcmember); + +static void mcast_groups_event(struct mcast_port *port, + enum mcast_group_state state) +{ + struct mcast_group *group; + struct rb_node *node; + unsigned long flags; + + spin_lock_irqsave(&port->lock, flags); + for (node = rb_first(&port->table); node; node = rb_next(node)) { + group = rb_entry(node, struct mcast_group, node); + spin_lock(&group->lock); + if (group->state == MCAST_IDLE) { + atomic_inc(&group->refcount); + queue_work(mcast_wq, &group->work); + } + if (group->state != MCAST_GROUP_ERROR) + group->state = state; + spin_unlock(&group->lock); + } + spin_unlock_irqrestore(&port->lock, flags); +} + +static void mcast_event_handler(struct ib_event_handler *handler, + struct ib_event *event) +{ + struct mcast_device *dev; + int index; + + dev = container_of(handler, struct mcast_device, event_handler); + if (!rdma_cap_ib_mcast(dev->device, event->element.port_num)) + return; + + index = event->element.port_num - dev->start_port; + + switch (event->event) { + case IB_EVENT_PORT_ERR: + case IB_EVENT_LID_CHANGE: + case IB_EVENT_CLIENT_REREGISTER: + mcast_groups_event(&dev->port[index], MCAST_GROUP_ERROR); + break; + case IB_EVENT_PKEY_CHANGE: + mcast_groups_event(&dev->port[index], MCAST_PKEY_EVENT); + break; + default: + break; + } +} + +static int mcast_add_one(struct ib_device *device) +{ + struct mcast_device *dev; + struct mcast_port *port; + int i; + int count = 0; + + dev = kmalloc(struct_size(dev, port, device->phys_port_cnt), + GFP_KERNEL); + if (!dev) + return -ENOMEM; + + dev->start_port = rdma_start_port(device); + dev->end_port = rdma_end_port(device); + + for (i = 0; i <= dev->end_port - dev->start_port; i++) { + if (!rdma_cap_ib_mcast(device, dev->start_port + i)) + continue; + port = &dev->port[i]; + port->dev = dev; + port->port_num = dev->start_port + i; + spin_lock_init(&port->lock); + port->table = RB_ROOT; + init_completion(&port->comp); + atomic_set(&port->refcount, 1); + ++count; + } + + if (!count) { + kfree(dev); + return -EOPNOTSUPP; + } + + dev->device = device; + ib_set_client_data(device, &mcast_client, dev); + + INIT_IB_EVENT_HANDLER(&dev->event_handler, device, mcast_event_handler); + ib_register_event_handler(&dev->event_handler); + return 0; +} + +static void mcast_remove_one(struct ib_device *device, void *client_data) +{ + struct mcast_device *dev = client_data; + struct mcast_port *port; + int i; + + ib_unregister_event_handler(&dev->event_handler); + flush_workqueue(mcast_wq); + + for (i = 0; i <= dev->end_port - dev->start_port; i++) { + if (rdma_cap_ib_mcast(device, dev->start_port + i)) { + port = &dev->port[i]; + deref_port(port); + wait_for_completion(&port->comp); + } + } + + kfree(dev); +} + +int mcast_init(void) +{ + int ret; + + mcast_wq = alloc_ordered_workqueue("ib_mcast", WQ_MEM_RECLAIM); + if (!mcast_wq) + return -ENOMEM; + + ib_sa_register_client(&sa_client); + + ret = ib_register_client(&mcast_client); + if (ret) + goto err; + return 0; + +err: + ib_sa_unregister_client(&sa_client); + destroy_workqueue(mcast_wq); + return ret; +} + +void mcast_cleanup(void) +{ + ib_unregister_client(&mcast_client); + ib_sa_unregister_client(&sa_client); + destroy_workqueue(mcast_wq); +} diff --git a/drivers/infiniband/core/netlink.c b/drivers/infiniband/core/netlink.c new file mode 100644 index 000000000..8cd31ef25 --- /dev/null +++ b/drivers/infiniband/core/netlink.c @@ -0,0 +1,331 @@ +/* + * Copyright (c) 2017 Mellanox Technologies Inc. All rights reserved. + * Copyright (c) 2010 Voltaire Inc. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#define pr_fmt(fmt) "%s:%s: " fmt, KBUILD_MODNAME, __func__ + +#include <linux/export.h> +#include <net/netlink.h> +#include <net/net_namespace.h> +#include <net/netns/generic.h> +#include <net/sock.h> +#include <rdma/rdma_netlink.h> +#include <linux/module.h> +#include "core_priv.h" + +static struct { + const struct rdma_nl_cbs *cb_table; + /* Synchronizes between ongoing netlink commands and netlink client + * unregistration. + */ + struct rw_semaphore sem; +} rdma_nl_types[RDMA_NL_NUM_CLIENTS]; + +bool rdma_nl_chk_listeners(unsigned int group) +{ + struct rdma_dev_net *rnet = rdma_net_to_dev_net(&init_net); + + return netlink_has_listeners(rnet->nl_sock, group); +} +EXPORT_SYMBOL(rdma_nl_chk_listeners); + +static bool is_nl_msg_valid(unsigned int type, unsigned int op) +{ + static const unsigned int max_num_ops[RDMA_NL_NUM_CLIENTS] = { + [RDMA_NL_IWCM] = RDMA_NL_IWPM_NUM_OPS, + [RDMA_NL_LS] = RDMA_NL_LS_NUM_OPS, + [RDMA_NL_NLDEV] = RDMA_NLDEV_NUM_OPS, + }; + + /* + * This BUILD_BUG_ON is intended to catch addition of new + * RDMA netlink protocol without updating the array above. + */ + BUILD_BUG_ON(RDMA_NL_NUM_CLIENTS != 6); + + if (type >= RDMA_NL_NUM_CLIENTS) + return false; + + return (op < max_num_ops[type]) ? true : false; +} + +static const struct rdma_nl_cbs * +get_cb_table(const struct sk_buff *skb, unsigned int type, unsigned int op) +{ + const struct rdma_nl_cbs *cb_table; + + /* + * Currently only NLDEV client is supporting netlink commands in + * non init_net net namespace. + */ + if (sock_net(skb->sk) != &init_net && type != RDMA_NL_NLDEV) + return NULL; + + cb_table = READ_ONCE(rdma_nl_types[type].cb_table); + if (!cb_table) { + /* + * Didn't get valid reference of the table, attempt module + * load once. + */ + up_read(&rdma_nl_types[type].sem); + + request_module("rdma-netlink-subsys-%d", type); + + down_read(&rdma_nl_types[type].sem); + cb_table = READ_ONCE(rdma_nl_types[type].cb_table); + } + if (!cb_table || (!cb_table[op].dump && !cb_table[op].doit)) + return NULL; + return cb_table; +} + +void rdma_nl_register(unsigned int index, + const struct rdma_nl_cbs cb_table[]) +{ + if (WARN_ON(!is_nl_msg_valid(index, 0)) || + WARN_ON(READ_ONCE(rdma_nl_types[index].cb_table))) + return; + + /* Pairs with the READ_ONCE in is_nl_valid() */ + smp_store_release(&rdma_nl_types[index].cb_table, cb_table); +} +EXPORT_SYMBOL(rdma_nl_register); + +void rdma_nl_unregister(unsigned int index) +{ + down_write(&rdma_nl_types[index].sem); + rdma_nl_types[index].cb_table = NULL; + up_write(&rdma_nl_types[index].sem); +} +EXPORT_SYMBOL(rdma_nl_unregister); + +void *ibnl_put_msg(struct sk_buff *skb, struct nlmsghdr **nlh, int seq, + int len, int client, int op, int flags) +{ + *nlh = nlmsg_put(skb, 0, seq, RDMA_NL_GET_TYPE(client, op), len, flags); + if (!*nlh) + return NULL; + return nlmsg_data(*nlh); +} +EXPORT_SYMBOL(ibnl_put_msg); + +int ibnl_put_attr(struct sk_buff *skb, struct nlmsghdr *nlh, + int len, void *data, int type) +{ + if (nla_put(skb, type, len, data)) { + nlmsg_cancel(skb, nlh); + return -EMSGSIZE; + } + return 0; +} +EXPORT_SYMBOL(ibnl_put_attr); + +static int rdma_nl_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh, + struct netlink_ext_ack *extack) +{ + int type = nlh->nlmsg_type; + unsigned int index = RDMA_NL_GET_CLIENT(type); + unsigned int op = RDMA_NL_GET_OP(type); + const struct rdma_nl_cbs *cb_table; + int err = -EINVAL; + + if (!is_nl_msg_valid(index, op)) + return -EINVAL; + + down_read(&rdma_nl_types[index].sem); + cb_table = get_cb_table(skb, index, op); + if (!cb_table) + goto done; + + if ((cb_table[op].flags & RDMA_NL_ADMIN_PERM) && + !netlink_capable(skb, CAP_NET_ADMIN)) { + err = -EPERM; + goto done; + } + + /* + * LS responses overload the 0x100 (NLM_F_ROOT) flag. Don't + * mistakenly call the .dump() function. + */ + if (index == RDMA_NL_LS) { + if (cb_table[op].doit) + err = cb_table[op].doit(skb, nlh, extack); + goto done; + } + /* FIXME: Convert IWCM to properly handle doit callbacks */ + if ((nlh->nlmsg_flags & NLM_F_DUMP) || index == RDMA_NL_IWCM) { + struct netlink_dump_control c = { + .dump = cb_table[op].dump, + }; + if (c.dump) + err = netlink_dump_start(skb->sk, skb, nlh, &c); + goto done; + } + + if (cb_table[op].doit) + err = cb_table[op].doit(skb, nlh, extack); +done: + up_read(&rdma_nl_types[index].sem); + return err; +} + +/* + * This function is similar to netlink_rcv_skb with one exception: + * It calls to the callback for the netlink messages without NLM_F_REQUEST + * flag. These messages are intended for RDMA_NL_LS consumer, so it is allowed + * for that consumer only. + */ +static int rdma_nl_rcv_skb(struct sk_buff *skb, int (*cb)(struct sk_buff *, + struct nlmsghdr *, + struct netlink_ext_ack *)) +{ + struct netlink_ext_ack extack = {}; + struct nlmsghdr *nlh; + int err; + + while (skb->len >= nlmsg_total_size(0)) { + int msglen; + + nlh = nlmsg_hdr(skb); + err = 0; + + if (nlh->nlmsg_len < NLMSG_HDRLEN || skb->len < nlh->nlmsg_len) + return 0; + + /* + * Generally speaking, the only requests are handled + * by the kernel, but RDMA_NL_LS is different, because it + * runs backward netlink scheme. Kernel initiates messages + * and waits for reply with data to keep pathrecord cache + * in sync. + */ + if (!(nlh->nlmsg_flags & NLM_F_REQUEST) && + (RDMA_NL_GET_CLIENT(nlh->nlmsg_type) != RDMA_NL_LS)) + goto ack; + + /* Skip control messages */ + if (nlh->nlmsg_type < NLMSG_MIN_TYPE) + goto ack; + + err = cb(skb, nlh, &extack); + if (err == -EINTR) + goto skip; + +ack: + if (nlh->nlmsg_flags & NLM_F_ACK || err) + netlink_ack(skb, nlh, err, &extack); + +skip: + msglen = NLMSG_ALIGN(nlh->nlmsg_len); + if (msglen > skb->len) + msglen = skb->len; + skb_pull(skb, msglen); + } + + return 0; +} + +static void rdma_nl_rcv(struct sk_buff *skb) +{ + rdma_nl_rcv_skb(skb, &rdma_nl_rcv_msg); +} + +int rdma_nl_unicast(struct net *net, struct sk_buff *skb, u32 pid) +{ + struct rdma_dev_net *rnet = rdma_net_to_dev_net(net); + int err; + + err = netlink_unicast(rnet->nl_sock, skb, pid, MSG_DONTWAIT); + return (err < 0) ? err : 0; +} +EXPORT_SYMBOL(rdma_nl_unicast); + +int rdma_nl_unicast_wait(struct net *net, struct sk_buff *skb, __u32 pid) +{ + struct rdma_dev_net *rnet = rdma_net_to_dev_net(net); + int err; + + err = netlink_unicast(rnet->nl_sock, skb, pid, 0); + return (err < 0) ? err : 0; +} +EXPORT_SYMBOL(rdma_nl_unicast_wait); + +int rdma_nl_multicast(struct net *net, struct sk_buff *skb, + unsigned int group, gfp_t flags) +{ + struct rdma_dev_net *rnet = rdma_net_to_dev_net(net); + + return nlmsg_multicast(rnet->nl_sock, skb, 0, group, flags); +} +EXPORT_SYMBOL(rdma_nl_multicast); + +void rdma_nl_init(void) +{ + int idx; + + for (idx = 0; idx < RDMA_NL_NUM_CLIENTS; idx++) + init_rwsem(&rdma_nl_types[idx].sem); +} + +void rdma_nl_exit(void) +{ + int idx; + + for (idx = 0; idx < RDMA_NL_NUM_CLIENTS; idx++) + WARN(rdma_nl_types[idx].cb_table, + "Netlink client %d wasn't released prior to unloading %s\n", + idx, KBUILD_MODNAME); +} + +int rdma_nl_net_init(struct rdma_dev_net *rnet) +{ + struct net *net = read_pnet(&rnet->net); + struct netlink_kernel_cfg cfg = { + .input = rdma_nl_rcv, + }; + struct sock *nls; + + nls = netlink_kernel_create(net, NETLINK_RDMA, &cfg); + if (!nls) + return -ENOMEM; + + nls->sk_sndtimeo = 10 * HZ; + rnet->nl_sock = nls; + return 0; +} + +void rdma_nl_net_exit(struct rdma_dev_net *rnet) +{ + netlink_kernel_release(rnet->nl_sock); +} + +MODULE_ALIAS_NET_PF_PROTO(PF_NETLINK, NETLINK_RDMA); diff --git a/drivers/infiniband/core/nldev.c b/drivers/infiniband/core/nldev.c new file mode 100644 index 000000000..f8dfec7ad --- /dev/null +++ b/drivers/infiniband/core/nldev.c @@ -0,0 +1,2192 @@ +/* + * Copyright (c) 2017 Mellanox Technologies. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * 3. Neither the names of the copyright holders nor the names of its + * contributors may be used to endorse or promote products derived from + * this software without specific prior written permission. + * + * Alternatively, this software may be distributed under the terms of the + * GNU General Public License ("GPL") version 2 as published by the Free + * Software Foundation. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" + * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE + * POSSIBILITY OF SUCH DAMAGE. + */ + +#include <linux/module.h> +#include <linux/pid.h> +#include <linux/pid_namespace.h> +#include <linux/mutex.h> +#include <net/netlink.h> +#include <rdma/rdma_cm.h> +#include <rdma/rdma_netlink.h> + +#include "core_priv.h" +#include "cma_priv.h" +#include "restrack.h" +#include "uverbs.h" + +typedef int (*res_fill_func_t)(struct sk_buff*, bool, + struct rdma_restrack_entry*, uint32_t); + +/* + * Sort array elements by the netlink attribute name + */ +static const struct nla_policy nldev_policy[RDMA_NLDEV_ATTR_MAX] = { + [RDMA_NLDEV_ATTR_CHARDEV] = { .type = NLA_U64 }, + [RDMA_NLDEV_ATTR_CHARDEV_ABI] = { .type = NLA_U64 }, + [RDMA_NLDEV_ATTR_CHARDEV_NAME] = { .type = NLA_NUL_STRING, + .len = RDMA_NLDEV_ATTR_EMPTY_STRING }, + [RDMA_NLDEV_ATTR_CHARDEV_TYPE] = { .type = NLA_NUL_STRING, + .len = RDMA_NLDEV_ATTR_CHARDEV_TYPE_SIZE }, + [RDMA_NLDEV_ATTR_DEV_DIM] = { .type = NLA_U8 }, + [RDMA_NLDEV_ATTR_DEV_INDEX] = { .type = NLA_U32 }, + [RDMA_NLDEV_ATTR_DEV_NAME] = { .type = NLA_NUL_STRING, + .len = IB_DEVICE_NAME_MAX }, + [RDMA_NLDEV_ATTR_DEV_NODE_TYPE] = { .type = NLA_U8 }, + [RDMA_NLDEV_ATTR_DEV_PROTOCOL] = { .type = NLA_NUL_STRING, + .len = RDMA_NLDEV_ATTR_EMPTY_STRING }, + [RDMA_NLDEV_ATTR_DRIVER] = { .type = NLA_NESTED }, + [RDMA_NLDEV_ATTR_DRIVER_ENTRY] = { .type = NLA_NESTED }, + [RDMA_NLDEV_ATTR_DRIVER_PRINT_TYPE] = { .type = NLA_U8 }, + [RDMA_NLDEV_ATTR_DRIVER_STRING] = { .type = NLA_NUL_STRING, + .len = RDMA_NLDEV_ATTR_EMPTY_STRING }, + [RDMA_NLDEV_ATTR_DRIVER_S32] = { .type = NLA_S32 }, + [RDMA_NLDEV_ATTR_DRIVER_S64] = { .type = NLA_S64 }, + [RDMA_NLDEV_ATTR_DRIVER_U32] = { .type = NLA_U32 }, + [RDMA_NLDEV_ATTR_DRIVER_U64] = { .type = NLA_U64 }, + [RDMA_NLDEV_ATTR_FW_VERSION] = { .type = NLA_NUL_STRING, + .len = RDMA_NLDEV_ATTR_EMPTY_STRING }, + [RDMA_NLDEV_ATTR_LID] = { .type = NLA_U32 }, + [RDMA_NLDEV_ATTR_LINK_TYPE] = { .type = NLA_NUL_STRING, + .len = IFNAMSIZ }, + [RDMA_NLDEV_ATTR_LMC] = { .type = NLA_U8 }, + [RDMA_NLDEV_ATTR_NDEV_INDEX] = { .type = NLA_U32 }, + [RDMA_NLDEV_ATTR_NDEV_NAME] = { .type = NLA_NUL_STRING, + .len = IFNAMSIZ }, + [RDMA_NLDEV_ATTR_NODE_GUID] = { .type = NLA_U64 }, + [RDMA_NLDEV_ATTR_PORT_INDEX] = { .type = NLA_U32 }, + [RDMA_NLDEV_ATTR_PORT_PHYS_STATE] = { .type = NLA_U8 }, + [RDMA_NLDEV_ATTR_PORT_STATE] = { .type = NLA_U8 }, + [RDMA_NLDEV_ATTR_RES_CM_ID] = { .type = NLA_NESTED }, + [RDMA_NLDEV_ATTR_RES_CM_IDN] = { .type = NLA_U32 }, + [RDMA_NLDEV_ATTR_RES_CM_ID_ENTRY] = { .type = NLA_NESTED }, + [RDMA_NLDEV_ATTR_RES_CQ] = { .type = NLA_NESTED }, + [RDMA_NLDEV_ATTR_RES_CQE] = { .type = NLA_U32 }, + [RDMA_NLDEV_ATTR_RES_CQN] = { .type = NLA_U32 }, + [RDMA_NLDEV_ATTR_RES_CQ_ENTRY] = { .type = NLA_NESTED }, + [RDMA_NLDEV_ATTR_RES_CTXN] = { .type = NLA_U32 }, + [RDMA_NLDEV_ATTR_RES_DST_ADDR] = { + .len = sizeof(struct __kernel_sockaddr_storage) }, + [RDMA_NLDEV_ATTR_RES_IOVA] = { .type = NLA_U64 }, + [RDMA_NLDEV_ATTR_RES_KERN_NAME] = { .type = NLA_NUL_STRING, + .len = RDMA_NLDEV_ATTR_EMPTY_STRING }, + [RDMA_NLDEV_ATTR_RES_LKEY] = { .type = NLA_U32 }, + [RDMA_NLDEV_ATTR_RES_LOCAL_DMA_LKEY] = { .type = NLA_U32 }, + [RDMA_NLDEV_ATTR_RES_LQPN] = { .type = NLA_U32 }, + [RDMA_NLDEV_ATTR_RES_MR] = { .type = NLA_NESTED }, + [RDMA_NLDEV_ATTR_RES_MRLEN] = { .type = NLA_U64 }, + [RDMA_NLDEV_ATTR_RES_MRN] = { .type = NLA_U32 }, + [RDMA_NLDEV_ATTR_RES_MR_ENTRY] = { .type = NLA_NESTED }, + [RDMA_NLDEV_ATTR_RES_PATH_MIG_STATE] = { .type = NLA_U8 }, + [RDMA_NLDEV_ATTR_RES_PD] = { .type = NLA_NESTED }, + [RDMA_NLDEV_ATTR_RES_PDN] = { .type = NLA_U32 }, + [RDMA_NLDEV_ATTR_RES_PD_ENTRY] = { .type = NLA_NESTED }, + [RDMA_NLDEV_ATTR_RES_PID] = { .type = NLA_U32 }, + [RDMA_NLDEV_ATTR_RES_POLL_CTX] = { .type = NLA_U8 }, + [RDMA_NLDEV_ATTR_RES_PS] = { .type = NLA_U32 }, + [RDMA_NLDEV_ATTR_RES_QP] = { .type = NLA_NESTED }, + [RDMA_NLDEV_ATTR_RES_QP_ENTRY] = { .type = NLA_NESTED }, + [RDMA_NLDEV_ATTR_RES_RAW] = { .type = NLA_BINARY }, + [RDMA_NLDEV_ATTR_RES_RKEY] = { .type = NLA_U32 }, + [RDMA_NLDEV_ATTR_RES_RQPN] = { .type = NLA_U32 }, + [RDMA_NLDEV_ATTR_RES_RQ_PSN] = { .type = NLA_U32 }, + [RDMA_NLDEV_ATTR_RES_SQ_PSN] = { .type = NLA_U32 }, + [RDMA_NLDEV_ATTR_RES_SRC_ADDR] = { + .len = sizeof(struct __kernel_sockaddr_storage) }, + [RDMA_NLDEV_ATTR_RES_STATE] = { .type = NLA_U8 }, + [RDMA_NLDEV_ATTR_RES_SUMMARY] = { .type = NLA_NESTED }, + [RDMA_NLDEV_ATTR_RES_SUMMARY_ENTRY] = { .type = NLA_NESTED }, + [RDMA_NLDEV_ATTR_RES_SUMMARY_ENTRY_CURR]= { .type = NLA_U64 }, + [RDMA_NLDEV_ATTR_RES_SUMMARY_ENTRY_NAME]= { .type = NLA_NUL_STRING, + .len = RDMA_NLDEV_ATTR_EMPTY_STRING }, + [RDMA_NLDEV_ATTR_RES_TYPE] = { .type = NLA_U8 }, + [RDMA_NLDEV_ATTR_RES_UNSAFE_GLOBAL_RKEY]= { .type = NLA_U32 }, + [RDMA_NLDEV_ATTR_RES_USECNT] = { .type = NLA_U64 }, + [RDMA_NLDEV_ATTR_SM_LID] = { .type = NLA_U32 }, + [RDMA_NLDEV_ATTR_SUBNET_PREFIX] = { .type = NLA_U64 }, + [RDMA_NLDEV_ATTR_STAT_AUTO_MODE_MASK] = { .type = NLA_U32 }, + [RDMA_NLDEV_ATTR_STAT_MODE] = { .type = NLA_U32 }, + [RDMA_NLDEV_ATTR_STAT_RES] = { .type = NLA_U32 }, + [RDMA_NLDEV_ATTR_STAT_COUNTER] = { .type = NLA_NESTED }, + [RDMA_NLDEV_ATTR_STAT_COUNTER_ENTRY] = { .type = NLA_NESTED }, + [RDMA_NLDEV_ATTR_STAT_COUNTER_ID] = { .type = NLA_U32 }, + [RDMA_NLDEV_ATTR_STAT_HWCOUNTERS] = { .type = NLA_NESTED }, + [RDMA_NLDEV_ATTR_STAT_HWCOUNTER_ENTRY] = { .type = NLA_NESTED }, + [RDMA_NLDEV_ATTR_STAT_HWCOUNTER_ENTRY_NAME] = { .type = NLA_NUL_STRING }, + [RDMA_NLDEV_ATTR_STAT_HWCOUNTER_ENTRY_VALUE] = { .type = NLA_U64 }, + [RDMA_NLDEV_ATTR_SYS_IMAGE_GUID] = { .type = NLA_U64 }, + [RDMA_NLDEV_ATTR_UVERBS_DRIVER_ID] = { .type = NLA_U32 }, + [RDMA_NLDEV_NET_NS_FD] = { .type = NLA_U32 }, + [RDMA_NLDEV_SYS_ATTR_NETNS_MODE] = { .type = NLA_U8 }, +}; + +static int put_driver_name_print_type(struct sk_buff *msg, const char *name, + enum rdma_nldev_print_type print_type) +{ + if (nla_put_string(msg, RDMA_NLDEV_ATTR_DRIVER_STRING, name)) + return -EMSGSIZE; + if (print_type != RDMA_NLDEV_PRINT_TYPE_UNSPEC && + nla_put_u8(msg, RDMA_NLDEV_ATTR_DRIVER_PRINT_TYPE, print_type)) + return -EMSGSIZE; + + return 0; +} + +static int _rdma_nl_put_driver_u32(struct sk_buff *msg, const char *name, + enum rdma_nldev_print_type print_type, + u32 value) +{ + if (put_driver_name_print_type(msg, name, print_type)) + return -EMSGSIZE; + if (nla_put_u32(msg, RDMA_NLDEV_ATTR_DRIVER_U32, value)) + return -EMSGSIZE; + + return 0; +} + +static int _rdma_nl_put_driver_u64(struct sk_buff *msg, const char *name, + enum rdma_nldev_print_type print_type, + u64 value) +{ + if (put_driver_name_print_type(msg, name, print_type)) + return -EMSGSIZE; + if (nla_put_u64_64bit(msg, RDMA_NLDEV_ATTR_DRIVER_U64, value, + RDMA_NLDEV_ATTR_PAD)) + return -EMSGSIZE; + + return 0; +} + +int rdma_nl_put_driver_string(struct sk_buff *msg, const char *name, + const char *str) +{ + if (put_driver_name_print_type(msg, name, + RDMA_NLDEV_PRINT_TYPE_UNSPEC)) + return -EMSGSIZE; + if (nla_put_string(msg, RDMA_NLDEV_ATTR_DRIVER_STRING, str)) + return -EMSGSIZE; + + return 0; +} +EXPORT_SYMBOL(rdma_nl_put_driver_string); + +int rdma_nl_put_driver_u32(struct sk_buff *msg, const char *name, u32 value) +{ + return _rdma_nl_put_driver_u32(msg, name, RDMA_NLDEV_PRINT_TYPE_UNSPEC, + value); +} +EXPORT_SYMBOL(rdma_nl_put_driver_u32); + +int rdma_nl_put_driver_u32_hex(struct sk_buff *msg, const char *name, + u32 value) +{ + return _rdma_nl_put_driver_u32(msg, name, RDMA_NLDEV_PRINT_TYPE_HEX, + value); +} +EXPORT_SYMBOL(rdma_nl_put_driver_u32_hex); + +int rdma_nl_put_driver_u64(struct sk_buff *msg, const char *name, u64 value) +{ + return _rdma_nl_put_driver_u64(msg, name, RDMA_NLDEV_PRINT_TYPE_UNSPEC, + value); +} +EXPORT_SYMBOL(rdma_nl_put_driver_u64); + +int rdma_nl_put_driver_u64_hex(struct sk_buff *msg, const char *name, u64 value) +{ + return _rdma_nl_put_driver_u64(msg, name, RDMA_NLDEV_PRINT_TYPE_HEX, + value); +} +EXPORT_SYMBOL(rdma_nl_put_driver_u64_hex); + +static int fill_nldev_handle(struct sk_buff *msg, struct ib_device *device) +{ + if (nla_put_u32(msg, RDMA_NLDEV_ATTR_DEV_INDEX, device->index)) + return -EMSGSIZE; + if (nla_put_string(msg, RDMA_NLDEV_ATTR_DEV_NAME, + dev_name(&device->dev))) + return -EMSGSIZE; + + return 0; +} + +static int fill_dev_info(struct sk_buff *msg, struct ib_device *device) +{ + char fw[IB_FW_VERSION_NAME_MAX]; + int ret = 0; + u8 port; + + if (fill_nldev_handle(msg, device)) + return -EMSGSIZE; + + if (nla_put_u32(msg, RDMA_NLDEV_ATTR_PORT_INDEX, rdma_end_port(device))) + return -EMSGSIZE; + + BUILD_BUG_ON(sizeof(device->attrs.device_cap_flags) != sizeof(u64)); + if (nla_put_u64_64bit(msg, RDMA_NLDEV_ATTR_CAP_FLAGS, + device->attrs.device_cap_flags, + RDMA_NLDEV_ATTR_PAD)) + return -EMSGSIZE; + + ib_get_device_fw_str(device, fw); + /* Device without FW has strlen(fw) = 0 */ + if (strlen(fw) && nla_put_string(msg, RDMA_NLDEV_ATTR_FW_VERSION, fw)) + return -EMSGSIZE; + + if (nla_put_u64_64bit(msg, RDMA_NLDEV_ATTR_NODE_GUID, + be64_to_cpu(device->node_guid), + RDMA_NLDEV_ATTR_PAD)) + return -EMSGSIZE; + if (nla_put_u64_64bit(msg, RDMA_NLDEV_ATTR_SYS_IMAGE_GUID, + be64_to_cpu(device->attrs.sys_image_guid), + RDMA_NLDEV_ATTR_PAD)) + return -EMSGSIZE; + if (nla_put_u8(msg, RDMA_NLDEV_ATTR_DEV_NODE_TYPE, device->node_type)) + return -EMSGSIZE; + if (nla_put_u8(msg, RDMA_NLDEV_ATTR_DEV_DIM, device->use_cq_dim)) + return -EMSGSIZE; + + /* + * Link type is determined on first port and mlx4 device + * which can potentially have two different link type for the same + * IB device is considered as better to be avoided in the future, + */ + port = rdma_start_port(device); + if (rdma_cap_opa_mad(device, port)) + ret = nla_put_string(msg, RDMA_NLDEV_ATTR_DEV_PROTOCOL, "opa"); + else if (rdma_protocol_ib(device, port)) + ret = nla_put_string(msg, RDMA_NLDEV_ATTR_DEV_PROTOCOL, "ib"); + else if (rdma_protocol_iwarp(device, port)) + ret = nla_put_string(msg, RDMA_NLDEV_ATTR_DEV_PROTOCOL, "iw"); + else if (rdma_protocol_roce(device, port)) + ret = nla_put_string(msg, RDMA_NLDEV_ATTR_DEV_PROTOCOL, "roce"); + else if (rdma_protocol_usnic(device, port)) + ret = nla_put_string(msg, RDMA_NLDEV_ATTR_DEV_PROTOCOL, + "usnic"); + return ret; +} + +static int fill_port_info(struct sk_buff *msg, + struct ib_device *device, u32 port, + const struct net *net) +{ + struct net_device *netdev = NULL; + struct ib_port_attr attr; + int ret; + u64 cap_flags = 0; + + if (fill_nldev_handle(msg, device)) + return -EMSGSIZE; + + if (nla_put_u32(msg, RDMA_NLDEV_ATTR_PORT_INDEX, port)) + return -EMSGSIZE; + + ret = ib_query_port(device, port, &attr); + if (ret) + return ret; + + if (rdma_protocol_ib(device, port)) { + BUILD_BUG_ON((sizeof(attr.port_cap_flags) + + sizeof(attr.port_cap_flags2)) > sizeof(u64)); + cap_flags = attr.port_cap_flags | + ((u64)attr.port_cap_flags2 << 32); + if (nla_put_u64_64bit(msg, RDMA_NLDEV_ATTR_CAP_FLAGS, + cap_flags, RDMA_NLDEV_ATTR_PAD)) + return -EMSGSIZE; + if (nla_put_u64_64bit(msg, RDMA_NLDEV_ATTR_SUBNET_PREFIX, + attr.subnet_prefix, RDMA_NLDEV_ATTR_PAD)) + return -EMSGSIZE; + if (nla_put_u32(msg, RDMA_NLDEV_ATTR_LID, attr.lid)) + return -EMSGSIZE; + if (nla_put_u32(msg, RDMA_NLDEV_ATTR_SM_LID, attr.sm_lid)) + return -EMSGSIZE; + if (nla_put_u8(msg, RDMA_NLDEV_ATTR_LMC, attr.lmc)) + return -EMSGSIZE; + } + if (nla_put_u8(msg, RDMA_NLDEV_ATTR_PORT_STATE, attr.state)) + return -EMSGSIZE; + if (nla_put_u8(msg, RDMA_NLDEV_ATTR_PORT_PHYS_STATE, attr.phys_state)) + return -EMSGSIZE; + + netdev = ib_device_get_netdev(device, port); + if (netdev && net_eq(dev_net(netdev), net)) { + ret = nla_put_u32(msg, + RDMA_NLDEV_ATTR_NDEV_INDEX, netdev->ifindex); + if (ret) + goto out; + ret = nla_put_string(msg, + RDMA_NLDEV_ATTR_NDEV_NAME, netdev->name); + } + +out: + if (netdev) + dev_put(netdev); + return ret; +} + +static int fill_res_info_entry(struct sk_buff *msg, + const char *name, u64 curr) +{ + struct nlattr *entry_attr; + + entry_attr = nla_nest_start_noflag(msg, + RDMA_NLDEV_ATTR_RES_SUMMARY_ENTRY); + if (!entry_attr) + return -EMSGSIZE; + + if (nla_put_string(msg, RDMA_NLDEV_ATTR_RES_SUMMARY_ENTRY_NAME, name)) + goto err; + if (nla_put_u64_64bit(msg, RDMA_NLDEV_ATTR_RES_SUMMARY_ENTRY_CURR, curr, + RDMA_NLDEV_ATTR_PAD)) + goto err; + + nla_nest_end(msg, entry_attr); + return 0; + +err: + nla_nest_cancel(msg, entry_attr); + return -EMSGSIZE; +} + +static int fill_res_info(struct sk_buff *msg, struct ib_device *device) +{ + static const char * const names[RDMA_RESTRACK_MAX] = { + [RDMA_RESTRACK_PD] = "pd", + [RDMA_RESTRACK_CQ] = "cq", + [RDMA_RESTRACK_QP] = "qp", + [RDMA_RESTRACK_CM_ID] = "cm_id", + [RDMA_RESTRACK_MR] = "mr", + [RDMA_RESTRACK_CTX] = "ctx", + }; + + struct nlattr *table_attr; + int ret, i, curr; + + if (fill_nldev_handle(msg, device)) + return -EMSGSIZE; + + table_attr = nla_nest_start_noflag(msg, RDMA_NLDEV_ATTR_RES_SUMMARY); + if (!table_attr) + return -EMSGSIZE; + + for (i = 0; i < RDMA_RESTRACK_MAX; i++) { + if (!names[i]) + continue; + curr = rdma_restrack_count(device, i); + ret = fill_res_info_entry(msg, names[i], curr); + if (ret) + goto err; + } + + nla_nest_end(msg, table_attr); + return 0; + +err: + nla_nest_cancel(msg, table_attr); + return ret; +} + +static int fill_res_name_pid(struct sk_buff *msg, + struct rdma_restrack_entry *res) +{ + int err = 0; + + /* + * For user resources, user is should read /proc/PID/comm to get the + * name of the task file. + */ + if (rdma_is_kernel_res(res)) { + err = nla_put_string(msg, RDMA_NLDEV_ATTR_RES_KERN_NAME, + res->kern_name); + } else { + pid_t pid; + + pid = task_pid_vnr(res->task); + /* + * Task is dead and in zombie state. + * There is no need to print PID anymore. + */ + if (pid) + /* + * This part is racy, task can be killed and PID will + * be zero right here but it is ok, next query won't + * return PID. We don't promise real-time reflection + * of SW objects. + */ + err = nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_PID, pid); + } + + return err ? -EMSGSIZE : 0; +} + +static int fill_res_qp_entry_query(struct sk_buff *msg, + struct rdma_restrack_entry *res, + struct ib_device *dev, + struct ib_qp *qp) +{ + struct ib_qp_init_attr qp_init_attr; + struct ib_qp_attr qp_attr; + int ret; + + ret = ib_query_qp(qp, &qp_attr, 0, &qp_init_attr); + if (ret) + return ret; + + if (qp->qp_type == IB_QPT_RC || qp->qp_type == IB_QPT_UC) { + if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_RQPN, + qp_attr.dest_qp_num)) + goto err; + if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_RQ_PSN, + qp_attr.rq_psn)) + goto err; + } + + if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_SQ_PSN, qp_attr.sq_psn)) + goto err; + + if (qp->qp_type == IB_QPT_RC || qp->qp_type == IB_QPT_UC || + qp->qp_type == IB_QPT_XRC_INI || qp->qp_type == IB_QPT_XRC_TGT) { + if (nla_put_u8(msg, RDMA_NLDEV_ATTR_RES_PATH_MIG_STATE, + qp_attr.path_mig_state)) + goto err; + } + if (nla_put_u8(msg, RDMA_NLDEV_ATTR_RES_TYPE, qp->qp_type)) + goto err; + if (nla_put_u8(msg, RDMA_NLDEV_ATTR_RES_STATE, qp_attr.qp_state)) + goto err; + + if (dev->ops.fill_res_qp_entry) + return dev->ops.fill_res_qp_entry(msg, qp); + return 0; + +err: return -EMSGSIZE; +} + +static int fill_res_qp_entry(struct sk_buff *msg, bool has_cap_net_admin, + struct rdma_restrack_entry *res, uint32_t port) +{ + struct ib_qp *qp = container_of(res, struct ib_qp, res); + struct ib_device *dev = qp->device; + int ret; + + if (port && port != qp->port) + return -EAGAIN; + + /* In create_qp() port is not set yet */ + if (qp->port && nla_put_u32(msg, RDMA_NLDEV_ATTR_PORT_INDEX, qp->port)) + return -EMSGSIZE; + + ret = nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_LQPN, qp->qp_num); + if (ret) + return -EMSGSIZE; + + if (!rdma_is_kernel_res(res) && + nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_PDN, qp->pd->res.id)) + return -EMSGSIZE; + + ret = fill_res_name_pid(msg, res); + if (ret) + return -EMSGSIZE; + + return fill_res_qp_entry_query(msg, res, dev, qp); +} + +static int fill_res_qp_raw_entry(struct sk_buff *msg, bool has_cap_net_admin, + struct rdma_restrack_entry *res, uint32_t port) +{ + struct ib_qp *qp = container_of(res, struct ib_qp, res); + struct ib_device *dev = qp->device; + + if (port && port != qp->port) + return -EAGAIN; + if (!dev->ops.fill_res_qp_entry_raw) + return -EINVAL; + return dev->ops.fill_res_qp_entry_raw(msg, qp); +} + +static int fill_res_cm_id_entry(struct sk_buff *msg, bool has_cap_net_admin, + struct rdma_restrack_entry *res, uint32_t port) +{ + struct rdma_id_private *id_priv = + container_of(res, struct rdma_id_private, res); + struct ib_device *dev = id_priv->id.device; + struct rdma_cm_id *cm_id = &id_priv->id; + + if (port && port != cm_id->port_num) + return -EAGAIN; + + if (cm_id->port_num && + nla_put_u32(msg, RDMA_NLDEV_ATTR_PORT_INDEX, cm_id->port_num)) + goto err; + + if (id_priv->qp_num) { + if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_LQPN, id_priv->qp_num)) + goto err; + if (nla_put_u8(msg, RDMA_NLDEV_ATTR_RES_TYPE, cm_id->qp_type)) + goto err; + } + + if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_PS, cm_id->ps)) + goto err; + + if (nla_put_u8(msg, RDMA_NLDEV_ATTR_RES_STATE, id_priv->state)) + goto err; + + if (cm_id->route.addr.src_addr.ss_family && + nla_put(msg, RDMA_NLDEV_ATTR_RES_SRC_ADDR, + sizeof(cm_id->route.addr.src_addr), + &cm_id->route.addr.src_addr)) + goto err; + if (cm_id->route.addr.dst_addr.ss_family && + nla_put(msg, RDMA_NLDEV_ATTR_RES_DST_ADDR, + sizeof(cm_id->route.addr.dst_addr), + &cm_id->route.addr.dst_addr)) + goto err; + + if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_CM_IDN, res->id)) + goto err; + + if (fill_res_name_pid(msg, res)) + goto err; + + if (dev->ops.fill_res_cm_id_entry) + return dev->ops.fill_res_cm_id_entry(msg, cm_id); + return 0; + +err: return -EMSGSIZE; +} + +static int fill_res_cq_entry(struct sk_buff *msg, bool has_cap_net_admin, + struct rdma_restrack_entry *res, uint32_t port) +{ + struct ib_cq *cq = container_of(res, struct ib_cq, res); + struct ib_device *dev = cq->device; + + if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_CQE, cq->cqe)) + return -EMSGSIZE; + if (nla_put_u64_64bit(msg, RDMA_NLDEV_ATTR_RES_USECNT, + atomic_read(&cq->usecnt), RDMA_NLDEV_ATTR_PAD)) + return -EMSGSIZE; + + /* Poll context is only valid for kernel CQs */ + if (rdma_is_kernel_res(res) && + nla_put_u8(msg, RDMA_NLDEV_ATTR_RES_POLL_CTX, cq->poll_ctx)) + return -EMSGSIZE; + + if (nla_put_u8(msg, RDMA_NLDEV_ATTR_DEV_DIM, (cq->dim != NULL))) + return -EMSGSIZE; + + if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_CQN, res->id)) + return -EMSGSIZE; + if (!rdma_is_kernel_res(res) && + nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_CTXN, + cq->uobject->uevent.uobject.context->res.id)) + return -EMSGSIZE; + + if (fill_res_name_pid(msg, res)) + return -EMSGSIZE; + + return (dev->ops.fill_res_cq_entry) ? + dev->ops.fill_res_cq_entry(msg, cq) : 0; +} + +static int fill_res_cq_raw_entry(struct sk_buff *msg, bool has_cap_net_admin, + struct rdma_restrack_entry *res, uint32_t port) +{ + struct ib_cq *cq = container_of(res, struct ib_cq, res); + struct ib_device *dev = cq->device; + + if (!dev->ops.fill_res_cq_entry_raw) + return -EINVAL; + return dev->ops.fill_res_cq_entry_raw(msg, cq); +} + +static int fill_res_mr_entry(struct sk_buff *msg, bool has_cap_net_admin, + struct rdma_restrack_entry *res, uint32_t port) +{ + struct ib_mr *mr = container_of(res, struct ib_mr, res); + struct ib_device *dev = mr->pd->device; + + if (has_cap_net_admin) { + if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_RKEY, mr->rkey)) + return -EMSGSIZE; + if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_LKEY, mr->lkey)) + return -EMSGSIZE; + } + + if (nla_put_u64_64bit(msg, RDMA_NLDEV_ATTR_RES_MRLEN, mr->length, + RDMA_NLDEV_ATTR_PAD)) + return -EMSGSIZE; + + if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_MRN, res->id)) + return -EMSGSIZE; + + if (!rdma_is_kernel_res(res) && + nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_PDN, mr->pd->res.id)) + return -EMSGSIZE; + + if (fill_res_name_pid(msg, res)) + return -EMSGSIZE; + + return (dev->ops.fill_res_mr_entry) ? + dev->ops.fill_res_mr_entry(msg, mr) : + 0; +} + +static int fill_res_mr_raw_entry(struct sk_buff *msg, bool has_cap_net_admin, + struct rdma_restrack_entry *res, uint32_t port) +{ + struct ib_mr *mr = container_of(res, struct ib_mr, res); + struct ib_device *dev = mr->pd->device; + + if (!dev->ops.fill_res_mr_entry_raw) + return -EINVAL; + return dev->ops.fill_res_mr_entry_raw(msg, mr); +} + +static int fill_res_pd_entry(struct sk_buff *msg, bool has_cap_net_admin, + struct rdma_restrack_entry *res, uint32_t port) +{ + struct ib_pd *pd = container_of(res, struct ib_pd, res); + + if (has_cap_net_admin) { + if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_LOCAL_DMA_LKEY, + pd->local_dma_lkey)) + goto err; + if ((pd->flags & IB_PD_UNSAFE_GLOBAL_RKEY) && + nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_UNSAFE_GLOBAL_RKEY, + pd->unsafe_global_rkey)) + goto err; + } + if (nla_put_u64_64bit(msg, RDMA_NLDEV_ATTR_RES_USECNT, + atomic_read(&pd->usecnt), RDMA_NLDEV_ATTR_PAD)) + goto err; + + if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_PDN, res->id)) + goto err; + + if (!rdma_is_kernel_res(res) && + nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_CTXN, + pd->uobject->context->res.id)) + goto err; + + return fill_res_name_pid(msg, res); + +err: return -EMSGSIZE; +} + +static int fill_stat_counter_mode(struct sk_buff *msg, + struct rdma_counter *counter) +{ + struct rdma_counter_mode *m = &counter->mode; + + if (nla_put_u32(msg, RDMA_NLDEV_ATTR_STAT_MODE, m->mode)) + return -EMSGSIZE; + + if (m->mode == RDMA_COUNTER_MODE_AUTO) { + if ((m->mask & RDMA_COUNTER_MASK_QP_TYPE) && + nla_put_u8(msg, RDMA_NLDEV_ATTR_RES_TYPE, m->param.qp_type)) + return -EMSGSIZE; + + if ((m->mask & RDMA_COUNTER_MASK_PID) && + fill_res_name_pid(msg, &counter->res)) + return -EMSGSIZE; + } + + return 0; +} + +static int fill_stat_counter_qp_entry(struct sk_buff *msg, u32 qpn) +{ + struct nlattr *entry_attr; + + entry_attr = nla_nest_start(msg, RDMA_NLDEV_ATTR_RES_QP_ENTRY); + if (!entry_attr) + return -EMSGSIZE; + + if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_LQPN, qpn)) + goto err; + + nla_nest_end(msg, entry_attr); + return 0; + +err: + nla_nest_cancel(msg, entry_attr); + return -EMSGSIZE; +} + +static int fill_stat_counter_qps(struct sk_buff *msg, + struct rdma_counter *counter) +{ + struct rdma_restrack_entry *res; + struct rdma_restrack_root *rt; + struct nlattr *table_attr; + struct ib_qp *qp = NULL; + unsigned long id = 0; + int ret = 0; + + table_attr = nla_nest_start(msg, RDMA_NLDEV_ATTR_RES_QP); + if (!table_attr) + return -EMSGSIZE; + + rt = &counter->device->res[RDMA_RESTRACK_QP]; + xa_lock(&rt->xa); + xa_for_each(&rt->xa, id, res) { + qp = container_of(res, struct ib_qp, res); + if (!qp->counter || (qp->counter->id != counter->id)) + continue; + + ret = fill_stat_counter_qp_entry(msg, qp->qp_num); + if (ret) + goto err; + } + + xa_unlock(&rt->xa); + nla_nest_end(msg, table_attr); + return 0; + +err: + xa_unlock(&rt->xa); + nla_nest_cancel(msg, table_attr); + return ret; +} + +int rdma_nl_stat_hwcounter_entry(struct sk_buff *msg, const char *name, + u64 value) +{ + struct nlattr *entry_attr; + + entry_attr = nla_nest_start(msg, RDMA_NLDEV_ATTR_STAT_HWCOUNTER_ENTRY); + if (!entry_attr) + return -EMSGSIZE; + + if (nla_put_string(msg, RDMA_NLDEV_ATTR_STAT_HWCOUNTER_ENTRY_NAME, + name)) + goto err; + if (nla_put_u64_64bit(msg, RDMA_NLDEV_ATTR_STAT_HWCOUNTER_ENTRY_VALUE, + value, RDMA_NLDEV_ATTR_PAD)) + goto err; + + nla_nest_end(msg, entry_attr); + return 0; + +err: + nla_nest_cancel(msg, entry_attr); + return -EMSGSIZE; +} +EXPORT_SYMBOL(rdma_nl_stat_hwcounter_entry); + +static int fill_stat_mr_entry(struct sk_buff *msg, bool has_cap_net_admin, + struct rdma_restrack_entry *res, uint32_t port) +{ + struct ib_mr *mr = container_of(res, struct ib_mr, res); + struct ib_device *dev = mr->pd->device; + + if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_MRN, res->id)) + goto err; + + if (dev->ops.fill_stat_mr_entry) + return dev->ops.fill_stat_mr_entry(msg, mr); + return 0; + +err: + return -EMSGSIZE; +} + +static int fill_stat_counter_hwcounters(struct sk_buff *msg, + struct rdma_counter *counter) +{ + struct rdma_hw_stats *st = counter->stats; + struct nlattr *table_attr; + int i; + + table_attr = nla_nest_start(msg, RDMA_NLDEV_ATTR_STAT_HWCOUNTERS); + if (!table_attr) + return -EMSGSIZE; + + for (i = 0; i < st->num_counters; i++) + if (rdma_nl_stat_hwcounter_entry(msg, st->names[i], st->value[i])) + goto err; + + nla_nest_end(msg, table_attr); + return 0; + +err: + nla_nest_cancel(msg, table_attr); + return -EMSGSIZE; +} + +static int fill_res_counter_entry(struct sk_buff *msg, bool has_cap_net_admin, + struct rdma_restrack_entry *res, + uint32_t port) +{ + struct rdma_counter *counter = + container_of(res, struct rdma_counter, res); + + if (port && port != counter->port) + return -EAGAIN; + + /* Dump it even query failed */ + rdma_counter_query_stats(counter); + + if (nla_put_u32(msg, RDMA_NLDEV_ATTR_PORT_INDEX, counter->port) || + nla_put_u32(msg, RDMA_NLDEV_ATTR_STAT_COUNTER_ID, counter->id) || + fill_stat_counter_mode(msg, counter) || + fill_stat_counter_qps(msg, counter) || + fill_stat_counter_hwcounters(msg, counter)) + return -EMSGSIZE; + + return 0; +} + +static int nldev_get_doit(struct sk_buff *skb, struct nlmsghdr *nlh, + struct netlink_ext_ack *extack) +{ + struct nlattr *tb[RDMA_NLDEV_ATTR_MAX]; + struct ib_device *device; + struct sk_buff *msg; + u32 index; + int err; + + err = nlmsg_parse_deprecated(nlh, 0, tb, RDMA_NLDEV_ATTR_MAX - 1, + nldev_policy, extack); + if (err || !tb[RDMA_NLDEV_ATTR_DEV_INDEX]) + return -EINVAL; + + index = nla_get_u32(tb[RDMA_NLDEV_ATTR_DEV_INDEX]); + + device = ib_device_get_by_index(sock_net(skb->sk), index); + if (!device) + return -EINVAL; + + msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); + if (!msg) { + err = -ENOMEM; + goto err; + } + + nlh = nlmsg_put(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, + RDMA_NL_GET_TYPE(RDMA_NL_NLDEV, RDMA_NLDEV_CMD_GET), + 0, 0); + + err = fill_dev_info(msg, device); + if (err) + goto err_free; + + nlmsg_end(msg, nlh); + + ib_device_put(device); + return rdma_nl_unicast(sock_net(skb->sk), msg, NETLINK_CB(skb).portid); + +err_free: + nlmsg_free(msg); +err: + ib_device_put(device); + return err; +} + +static int nldev_set_doit(struct sk_buff *skb, struct nlmsghdr *nlh, + struct netlink_ext_ack *extack) +{ + struct nlattr *tb[RDMA_NLDEV_ATTR_MAX]; + struct ib_device *device; + u32 index; + int err; + + err = nlmsg_parse_deprecated(nlh, 0, tb, RDMA_NLDEV_ATTR_MAX - 1, + nldev_policy, extack); + if (err || !tb[RDMA_NLDEV_ATTR_DEV_INDEX]) + return -EINVAL; + + index = nla_get_u32(tb[RDMA_NLDEV_ATTR_DEV_INDEX]); + device = ib_device_get_by_index(sock_net(skb->sk), index); + if (!device) + return -EINVAL; + + if (tb[RDMA_NLDEV_ATTR_DEV_NAME]) { + char name[IB_DEVICE_NAME_MAX] = {}; + + nla_strlcpy(name, tb[RDMA_NLDEV_ATTR_DEV_NAME], + IB_DEVICE_NAME_MAX); + if (strlen(name) == 0) { + err = -EINVAL; + goto done; + } + err = ib_device_rename(device, name); + goto done; + } + + if (tb[RDMA_NLDEV_NET_NS_FD]) { + u32 ns_fd; + + ns_fd = nla_get_u32(tb[RDMA_NLDEV_NET_NS_FD]); + err = ib_device_set_netns_put(skb, device, ns_fd); + goto put_done; + } + + if (tb[RDMA_NLDEV_ATTR_DEV_DIM]) { + u8 use_dim; + + use_dim = nla_get_u8(tb[RDMA_NLDEV_ATTR_DEV_DIM]); + err = ib_device_set_dim(device, use_dim); + goto done; + } + +done: + ib_device_put(device); +put_done: + return err; +} + +static int _nldev_get_dumpit(struct ib_device *device, + struct sk_buff *skb, + struct netlink_callback *cb, + unsigned int idx) +{ + int start = cb->args[0]; + struct nlmsghdr *nlh; + + if (idx < start) + return 0; + + nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq, + RDMA_NL_GET_TYPE(RDMA_NL_NLDEV, RDMA_NLDEV_CMD_GET), + 0, NLM_F_MULTI); + + if (fill_dev_info(skb, device)) { + nlmsg_cancel(skb, nlh); + goto out; + } + + nlmsg_end(skb, nlh); + + idx++; + +out: cb->args[0] = idx; + return skb->len; +} + +static int nldev_get_dumpit(struct sk_buff *skb, struct netlink_callback *cb) +{ + /* + * There is no need to take lock, because + * we are relying on ib_core's locking. + */ + return ib_enum_all_devs(_nldev_get_dumpit, skb, cb); +} + +static int nldev_port_get_doit(struct sk_buff *skb, struct nlmsghdr *nlh, + struct netlink_ext_ack *extack) +{ + struct nlattr *tb[RDMA_NLDEV_ATTR_MAX]; + struct ib_device *device; + struct sk_buff *msg; + u32 index; + u32 port; + int err; + + err = nlmsg_parse_deprecated(nlh, 0, tb, RDMA_NLDEV_ATTR_MAX - 1, + nldev_policy, extack); + if (err || + !tb[RDMA_NLDEV_ATTR_DEV_INDEX] || + !tb[RDMA_NLDEV_ATTR_PORT_INDEX]) + return -EINVAL; + + index = nla_get_u32(tb[RDMA_NLDEV_ATTR_DEV_INDEX]); + device = ib_device_get_by_index(sock_net(skb->sk), index); + if (!device) + return -EINVAL; + + port = nla_get_u32(tb[RDMA_NLDEV_ATTR_PORT_INDEX]); + if (!rdma_is_port_valid(device, port)) { + err = -EINVAL; + goto err; + } + + msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); + if (!msg) { + err = -ENOMEM; + goto err; + } + + nlh = nlmsg_put(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, + RDMA_NL_GET_TYPE(RDMA_NL_NLDEV, RDMA_NLDEV_CMD_GET), + 0, 0); + + err = fill_port_info(msg, device, port, sock_net(skb->sk)); + if (err) + goto err_free; + + nlmsg_end(msg, nlh); + ib_device_put(device); + + return rdma_nl_unicast(sock_net(skb->sk), msg, NETLINK_CB(skb).portid); + +err_free: + nlmsg_free(msg); +err: + ib_device_put(device); + return err; +} + +static int nldev_port_get_dumpit(struct sk_buff *skb, + struct netlink_callback *cb) +{ + struct nlattr *tb[RDMA_NLDEV_ATTR_MAX]; + struct ib_device *device; + int start = cb->args[0]; + struct nlmsghdr *nlh; + u32 idx = 0; + u32 ifindex; + int err; + unsigned int p; + + err = nlmsg_parse_deprecated(cb->nlh, 0, tb, RDMA_NLDEV_ATTR_MAX - 1, + nldev_policy, NULL); + if (err || !tb[RDMA_NLDEV_ATTR_DEV_INDEX]) + return -EINVAL; + + ifindex = nla_get_u32(tb[RDMA_NLDEV_ATTR_DEV_INDEX]); + device = ib_device_get_by_index(sock_net(skb->sk), ifindex); + if (!device) + return -EINVAL; + + rdma_for_each_port (device, p) { + /* + * The dumpit function returns all information from specific + * index. This specific index is taken from the netlink + * messages request sent by user and it is available + * in cb->args[0]. + * + * Usually, the user doesn't fill this field and it causes + * to return everything. + * + */ + if (idx < start) { + idx++; + continue; + } + + nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, + cb->nlh->nlmsg_seq, + RDMA_NL_GET_TYPE(RDMA_NL_NLDEV, + RDMA_NLDEV_CMD_PORT_GET), + 0, NLM_F_MULTI); + + if (fill_port_info(skb, device, p, sock_net(skb->sk))) { + nlmsg_cancel(skb, nlh); + goto out; + } + idx++; + nlmsg_end(skb, nlh); + } + +out: + ib_device_put(device); + cb->args[0] = idx; + return skb->len; +} + +static int nldev_res_get_doit(struct sk_buff *skb, struct nlmsghdr *nlh, + struct netlink_ext_ack *extack) +{ + struct nlattr *tb[RDMA_NLDEV_ATTR_MAX]; + struct ib_device *device; + struct sk_buff *msg; + u32 index; + int ret; + + ret = nlmsg_parse_deprecated(nlh, 0, tb, RDMA_NLDEV_ATTR_MAX - 1, + nldev_policy, extack); + if (ret || !tb[RDMA_NLDEV_ATTR_DEV_INDEX]) + return -EINVAL; + + index = nla_get_u32(tb[RDMA_NLDEV_ATTR_DEV_INDEX]); + device = ib_device_get_by_index(sock_net(skb->sk), index); + if (!device) + return -EINVAL; + + msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); + if (!msg) { + ret = -ENOMEM; + goto err; + } + + nlh = nlmsg_put(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, + RDMA_NL_GET_TYPE(RDMA_NL_NLDEV, RDMA_NLDEV_CMD_RES_GET), + 0, 0); + + ret = fill_res_info(msg, device); + if (ret) + goto err_free; + + nlmsg_end(msg, nlh); + ib_device_put(device); + return rdma_nl_unicast(sock_net(skb->sk), msg, NETLINK_CB(skb).portid); + +err_free: + nlmsg_free(msg); +err: + ib_device_put(device); + return ret; +} + +static int _nldev_res_get_dumpit(struct ib_device *device, + struct sk_buff *skb, + struct netlink_callback *cb, + unsigned int idx) +{ + int start = cb->args[0]; + struct nlmsghdr *nlh; + + if (idx < start) + return 0; + + nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq, + RDMA_NL_GET_TYPE(RDMA_NL_NLDEV, RDMA_NLDEV_CMD_RES_GET), + 0, NLM_F_MULTI); + + if (fill_res_info(skb, device)) { + nlmsg_cancel(skb, nlh); + goto out; + } + nlmsg_end(skb, nlh); + + idx++; + +out: + cb->args[0] = idx; + return skb->len; +} + +static int nldev_res_get_dumpit(struct sk_buff *skb, + struct netlink_callback *cb) +{ + return ib_enum_all_devs(_nldev_res_get_dumpit, skb, cb); +} + +struct nldev_fill_res_entry { + enum rdma_nldev_attr nldev_attr; + u8 flags; + u32 entry; + u32 id; +}; + +enum nldev_res_flags { + NLDEV_PER_DEV = 1 << 0, +}; + +static const struct nldev_fill_res_entry fill_entries[RDMA_RESTRACK_MAX] = { + [RDMA_RESTRACK_QP] = { + .nldev_attr = RDMA_NLDEV_ATTR_RES_QP, + .entry = RDMA_NLDEV_ATTR_RES_QP_ENTRY, + .id = RDMA_NLDEV_ATTR_RES_LQPN, + }, + [RDMA_RESTRACK_CM_ID] = { + .nldev_attr = RDMA_NLDEV_ATTR_RES_CM_ID, + .entry = RDMA_NLDEV_ATTR_RES_CM_ID_ENTRY, + .id = RDMA_NLDEV_ATTR_RES_CM_IDN, + }, + [RDMA_RESTRACK_CQ] = { + .nldev_attr = RDMA_NLDEV_ATTR_RES_CQ, + .flags = NLDEV_PER_DEV, + .entry = RDMA_NLDEV_ATTR_RES_CQ_ENTRY, + .id = RDMA_NLDEV_ATTR_RES_CQN, + }, + [RDMA_RESTRACK_MR] = { + .nldev_attr = RDMA_NLDEV_ATTR_RES_MR, + .flags = NLDEV_PER_DEV, + .entry = RDMA_NLDEV_ATTR_RES_MR_ENTRY, + .id = RDMA_NLDEV_ATTR_RES_MRN, + }, + [RDMA_RESTRACK_PD] = { + .nldev_attr = RDMA_NLDEV_ATTR_RES_PD, + .flags = NLDEV_PER_DEV, + .entry = RDMA_NLDEV_ATTR_RES_PD_ENTRY, + .id = RDMA_NLDEV_ATTR_RES_PDN, + }, + [RDMA_RESTRACK_COUNTER] = { + .nldev_attr = RDMA_NLDEV_ATTR_STAT_COUNTER, + .entry = RDMA_NLDEV_ATTR_STAT_COUNTER_ENTRY, + .id = RDMA_NLDEV_ATTR_STAT_COUNTER_ID, + }, +}; + +static int res_get_common_doit(struct sk_buff *skb, struct nlmsghdr *nlh, + struct netlink_ext_ack *extack, + enum rdma_restrack_type res_type, + res_fill_func_t fill_func) +{ + const struct nldev_fill_res_entry *fe = &fill_entries[res_type]; + struct nlattr *tb[RDMA_NLDEV_ATTR_MAX]; + struct rdma_restrack_entry *res; + struct ib_device *device; + u32 index, id, port = 0; + bool has_cap_net_admin; + struct sk_buff *msg; + int ret; + + ret = nlmsg_parse_deprecated(nlh, 0, tb, RDMA_NLDEV_ATTR_MAX - 1, + nldev_policy, extack); + if (ret || !tb[RDMA_NLDEV_ATTR_DEV_INDEX] || !fe->id || !tb[fe->id]) + return -EINVAL; + + index = nla_get_u32(tb[RDMA_NLDEV_ATTR_DEV_INDEX]); + device = ib_device_get_by_index(sock_net(skb->sk), index); + if (!device) + return -EINVAL; + + if (tb[RDMA_NLDEV_ATTR_PORT_INDEX]) { + port = nla_get_u32(tb[RDMA_NLDEV_ATTR_PORT_INDEX]); + if (!rdma_is_port_valid(device, port)) { + ret = -EINVAL; + goto err; + } + } + + if ((port && fe->flags & NLDEV_PER_DEV) || + (!port && ~fe->flags & NLDEV_PER_DEV)) { + ret = -EINVAL; + goto err; + } + + id = nla_get_u32(tb[fe->id]); + res = rdma_restrack_get_byid(device, res_type, id); + if (IS_ERR(res)) { + ret = PTR_ERR(res); + goto err; + } + + msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); + if (!msg) { + ret = -ENOMEM; + goto err_get; + } + + nlh = nlmsg_put(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, + RDMA_NL_GET_TYPE(RDMA_NL_NLDEV, + RDMA_NL_GET_OP(nlh->nlmsg_type)), + 0, 0); + + if (fill_nldev_handle(msg, device)) { + ret = -EMSGSIZE; + goto err_free; + } + + has_cap_net_admin = netlink_capable(skb, CAP_NET_ADMIN); + + ret = fill_func(msg, has_cap_net_admin, res, port); + if (ret) + goto err_free; + + rdma_restrack_put(res); + nlmsg_end(msg, nlh); + ib_device_put(device); + return rdma_nl_unicast(sock_net(skb->sk), msg, NETLINK_CB(skb).portid); + +err_free: + nlmsg_free(msg); +err_get: + rdma_restrack_put(res); +err: + ib_device_put(device); + return ret; +} + +static int res_get_common_dumpit(struct sk_buff *skb, + struct netlink_callback *cb, + enum rdma_restrack_type res_type, + res_fill_func_t fill_func) +{ + const struct nldev_fill_res_entry *fe = &fill_entries[res_type]; + struct nlattr *tb[RDMA_NLDEV_ATTR_MAX]; + struct rdma_restrack_entry *res; + struct rdma_restrack_root *rt; + int err, ret = 0, idx = 0; + struct nlattr *table_attr; + struct nlattr *entry_attr; + struct ib_device *device; + int start = cb->args[0]; + bool has_cap_net_admin; + struct nlmsghdr *nlh; + unsigned long id; + u32 index, port = 0; + bool filled = false; + + err = nlmsg_parse_deprecated(cb->nlh, 0, tb, RDMA_NLDEV_ATTR_MAX - 1, + nldev_policy, NULL); + /* + * Right now, we are expecting the device index to get res information, + * but it is possible to extend this code to return all devices in + * one shot by checking the existence of RDMA_NLDEV_ATTR_DEV_INDEX. + * if it doesn't exist, we will iterate over all devices. + * + * But it is not needed for now. + */ + if (err || !tb[RDMA_NLDEV_ATTR_DEV_INDEX]) + return -EINVAL; + + index = nla_get_u32(tb[RDMA_NLDEV_ATTR_DEV_INDEX]); + device = ib_device_get_by_index(sock_net(skb->sk), index); + if (!device) + return -EINVAL; + + /* + * If no PORT_INDEX is supplied, we will return all QPs from that device + */ + if (tb[RDMA_NLDEV_ATTR_PORT_INDEX]) { + port = nla_get_u32(tb[RDMA_NLDEV_ATTR_PORT_INDEX]); + if (!rdma_is_port_valid(device, port)) { + ret = -EINVAL; + goto err_index; + } + } + + nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq, + RDMA_NL_GET_TYPE(RDMA_NL_NLDEV, + RDMA_NL_GET_OP(cb->nlh->nlmsg_type)), + 0, NLM_F_MULTI); + + if (fill_nldev_handle(skb, device)) { + ret = -EMSGSIZE; + goto err; + } + + table_attr = nla_nest_start_noflag(skb, fe->nldev_attr); + if (!table_attr) { + ret = -EMSGSIZE; + goto err; + } + + has_cap_net_admin = netlink_capable(cb->skb, CAP_NET_ADMIN); + + rt = &device->res[res_type]; + xa_lock(&rt->xa); + /* + * FIXME: if the skip ahead is something common this loop should + * use xas_for_each & xas_pause to optimize, we can have a lot of + * objects. + */ + xa_for_each(&rt->xa, id, res) { + if (idx < start || !rdma_restrack_get(res)) + goto next; + + xa_unlock(&rt->xa); + + filled = true; + + entry_attr = nla_nest_start_noflag(skb, fe->entry); + if (!entry_attr) { + ret = -EMSGSIZE; + rdma_restrack_put(res); + goto msg_full; + } + + ret = fill_func(skb, has_cap_net_admin, res, port); + + rdma_restrack_put(res); + + if (ret) { + nla_nest_cancel(skb, entry_attr); + if (ret == -EMSGSIZE) + goto msg_full; + if (ret == -EAGAIN) + goto again; + goto res_err; + } + nla_nest_end(skb, entry_attr); +again: xa_lock(&rt->xa); +next: idx++; + } + xa_unlock(&rt->xa); + +msg_full: + nla_nest_end(skb, table_attr); + nlmsg_end(skb, nlh); + cb->args[0] = idx; + + /* + * No more entries to fill, cancel the message and + * return 0 to mark end of dumpit. + */ + if (!filled) + goto err; + + ib_device_put(device); + return skb->len; + +res_err: + nla_nest_cancel(skb, table_attr); + +err: + nlmsg_cancel(skb, nlh); + +err_index: + ib_device_put(device); + return ret; +} + +#define RES_GET_FUNCS(name, type) \ + static int nldev_res_get_##name##_dumpit(struct sk_buff *skb, \ + struct netlink_callback *cb) \ + { \ + return res_get_common_dumpit(skb, cb, type, \ + fill_res_##name##_entry); \ + } \ + static int nldev_res_get_##name##_doit(struct sk_buff *skb, \ + struct nlmsghdr *nlh, \ + struct netlink_ext_ack *extack) \ + { \ + return res_get_common_doit(skb, nlh, extack, type, \ + fill_res_##name##_entry); \ + } + +RES_GET_FUNCS(qp, RDMA_RESTRACK_QP); +RES_GET_FUNCS(qp_raw, RDMA_RESTRACK_QP); +RES_GET_FUNCS(cm_id, RDMA_RESTRACK_CM_ID); +RES_GET_FUNCS(cq, RDMA_RESTRACK_CQ); +RES_GET_FUNCS(cq_raw, RDMA_RESTRACK_CQ); +RES_GET_FUNCS(pd, RDMA_RESTRACK_PD); +RES_GET_FUNCS(mr, RDMA_RESTRACK_MR); +RES_GET_FUNCS(mr_raw, RDMA_RESTRACK_MR); +RES_GET_FUNCS(counter, RDMA_RESTRACK_COUNTER); + +static LIST_HEAD(link_ops); +static DECLARE_RWSEM(link_ops_rwsem); + +static const struct rdma_link_ops *link_ops_get(const char *type) +{ + const struct rdma_link_ops *ops; + + list_for_each_entry(ops, &link_ops, list) { + if (!strcmp(ops->type, type)) + goto out; + } + ops = NULL; +out: + return ops; +} + +void rdma_link_register(struct rdma_link_ops *ops) +{ + down_write(&link_ops_rwsem); + if (WARN_ON_ONCE(link_ops_get(ops->type))) + goto out; + list_add(&ops->list, &link_ops); +out: + up_write(&link_ops_rwsem); +} +EXPORT_SYMBOL(rdma_link_register); + +void rdma_link_unregister(struct rdma_link_ops *ops) +{ + down_write(&link_ops_rwsem); + list_del(&ops->list); + up_write(&link_ops_rwsem); +} +EXPORT_SYMBOL(rdma_link_unregister); + +static int nldev_newlink(struct sk_buff *skb, struct nlmsghdr *nlh, + struct netlink_ext_ack *extack) +{ + struct nlattr *tb[RDMA_NLDEV_ATTR_MAX]; + char ibdev_name[IB_DEVICE_NAME_MAX]; + const struct rdma_link_ops *ops; + char ndev_name[IFNAMSIZ]; + struct net_device *ndev; + char type[IFNAMSIZ]; + int err; + + err = nlmsg_parse_deprecated(nlh, 0, tb, RDMA_NLDEV_ATTR_MAX - 1, + nldev_policy, extack); + if (err || !tb[RDMA_NLDEV_ATTR_DEV_NAME] || + !tb[RDMA_NLDEV_ATTR_LINK_TYPE] || !tb[RDMA_NLDEV_ATTR_NDEV_NAME]) + return -EINVAL; + + nla_strlcpy(ibdev_name, tb[RDMA_NLDEV_ATTR_DEV_NAME], + sizeof(ibdev_name)); + if (strchr(ibdev_name, '%') || strlen(ibdev_name) == 0) + return -EINVAL; + + nla_strlcpy(type, tb[RDMA_NLDEV_ATTR_LINK_TYPE], sizeof(type)); + nla_strlcpy(ndev_name, tb[RDMA_NLDEV_ATTR_NDEV_NAME], + sizeof(ndev_name)); + + ndev = dev_get_by_name(sock_net(skb->sk), ndev_name); + if (!ndev) + return -ENODEV; + + down_read(&link_ops_rwsem); + ops = link_ops_get(type); +#ifdef CONFIG_MODULES + if (!ops) { + up_read(&link_ops_rwsem); + request_module("rdma-link-%s", type); + down_read(&link_ops_rwsem); + ops = link_ops_get(type); + } +#endif + err = ops ? ops->newlink(ibdev_name, ndev) : -EINVAL; + up_read(&link_ops_rwsem); + dev_put(ndev); + + return err; +} + +static int nldev_dellink(struct sk_buff *skb, struct nlmsghdr *nlh, + struct netlink_ext_ack *extack) +{ + struct nlattr *tb[RDMA_NLDEV_ATTR_MAX]; + struct ib_device *device; + u32 index; + int err; + + err = nlmsg_parse_deprecated(nlh, 0, tb, RDMA_NLDEV_ATTR_MAX - 1, + nldev_policy, extack); + if (err || !tb[RDMA_NLDEV_ATTR_DEV_INDEX]) + return -EINVAL; + + index = nla_get_u32(tb[RDMA_NLDEV_ATTR_DEV_INDEX]); + device = ib_device_get_by_index(sock_net(skb->sk), index); + if (!device) + return -EINVAL; + + if (!(device->attrs.device_cap_flags & IB_DEVICE_ALLOW_USER_UNREG)) { + ib_device_put(device); + return -EINVAL; + } + + ib_unregister_device_and_put(device); + return 0; +} + +static int nldev_get_chardev(struct sk_buff *skb, struct nlmsghdr *nlh, + struct netlink_ext_ack *extack) +{ + struct nlattr *tb[RDMA_NLDEV_ATTR_MAX]; + char client_name[RDMA_NLDEV_ATTR_CHARDEV_TYPE_SIZE]; + struct ib_client_nl_info data = {}; + struct ib_device *ibdev = NULL; + struct sk_buff *msg; + u32 index; + int err; + + err = nlmsg_parse(nlh, 0, tb, RDMA_NLDEV_ATTR_MAX - 1, nldev_policy, + extack); + if (err || !tb[RDMA_NLDEV_ATTR_CHARDEV_TYPE]) + return -EINVAL; + + nla_strlcpy(client_name, tb[RDMA_NLDEV_ATTR_CHARDEV_TYPE], + sizeof(client_name)); + + if (tb[RDMA_NLDEV_ATTR_DEV_INDEX]) { + index = nla_get_u32(tb[RDMA_NLDEV_ATTR_DEV_INDEX]); + ibdev = ib_device_get_by_index(sock_net(skb->sk), index); + if (!ibdev) + return -EINVAL; + + if (tb[RDMA_NLDEV_ATTR_PORT_INDEX]) { + data.port = nla_get_u32(tb[RDMA_NLDEV_ATTR_PORT_INDEX]); + if (!rdma_is_port_valid(ibdev, data.port)) { + err = -EINVAL; + goto out_put; + } + } else { + data.port = -1; + } + } else if (tb[RDMA_NLDEV_ATTR_PORT_INDEX]) { + return -EINVAL; + } + + msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); + if (!msg) { + err = -ENOMEM; + goto out_put; + } + nlh = nlmsg_put(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, + RDMA_NL_GET_TYPE(RDMA_NL_NLDEV, + RDMA_NLDEV_CMD_GET_CHARDEV), + 0, 0); + + data.nl_msg = msg; + err = ib_get_client_nl_info(ibdev, client_name, &data); + if (err) + goto out_nlmsg; + + err = nla_put_u64_64bit(msg, RDMA_NLDEV_ATTR_CHARDEV, + huge_encode_dev(data.cdev->devt), + RDMA_NLDEV_ATTR_PAD); + if (err) + goto out_data; + err = nla_put_u64_64bit(msg, RDMA_NLDEV_ATTR_CHARDEV_ABI, data.abi, + RDMA_NLDEV_ATTR_PAD); + if (err) + goto out_data; + if (nla_put_string(msg, RDMA_NLDEV_ATTR_CHARDEV_NAME, + dev_name(data.cdev))) { + err = -EMSGSIZE; + goto out_data; + } + + nlmsg_end(msg, nlh); + put_device(data.cdev); + if (ibdev) + ib_device_put(ibdev); + return rdma_nl_unicast(sock_net(skb->sk), msg, NETLINK_CB(skb).portid); + +out_data: + put_device(data.cdev); +out_nlmsg: + nlmsg_free(msg); +out_put: + if (ibdev) + ib_device_put(ibdev); + return err; +} + +static int nldev_sys_get_doit(struct sk_buff *skb, struct nlmsghdr *nlh, + struct netlink_ext_ack *extack) +{ + struct nlattr *tb[RDMA_NLDEV_ATTR_MAX]; + struct sk_buff *msg; + int err; + + err = nlmsg_parse(nlh, 0, tb, RDMA_NLDEV_ATTR_MAX - 1, + nldev_policy, extack); + if (err) + return err; + + msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); + if (!msg) + return -ENOMEM; + + nlh = nlmsg_put(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, + RDMA_NL_GET_TYPE(RDMA_NL_NLDEV, + RDMA_NLDEV_CMD_SYS_GET), + 0, 0); + + err = nla_put_u8(msg, RDMA_NLDEV_SYS_ATTR_NETNS_MODE, + (u8)ib_devices_shared_netns); + if (err) { + nlmsg_free(msg); + return err; + } + nlmsg_end(msg, nlh); + return rdma_nl_unicast(sock_net(skb->sk), msg, NETLINK_CB(skb).portid); +} + +static int nldev_set_sys_set_doit(struct sk_buff *skb, struct nlmsghdr *nlh, + struct netlink_ext_ack *extack) +{ + struct nlattr *tb[RDMA_NLDEV_ATTR_MAX]; + u8 enable; + int err; + + err = nlmsg_parse(nlh, 0, tb, RDMA_NLDEV_ATTR_MAX - 1, + nldev_policy, extack); + if (err || !tb[RDMA_NLDEV_SYS_ATTR_NETNS_MODE]) + return -EINVAL; + + enable = nla_get_u8(tb[RDMA_NLDEV_SYS_ATTR_NETNS_MODE]); + /* Only 0 and 1 are supported */ + if (enable > 1) + return -EINVAL; + + err = rdma_compatdev_set(enable); + return err; +} + +static int nldev_stat_set_doit(struct sk_buff *skb, struct nlmsghdr *nlh, + struct netlink_ext_ack *extack) +{ + u32 index, port, mode, mask = 0, qpn, cntn = 0; + struct nlattr *tb[RDMA_NLDEV_ATTR_MAX]; + struct ib_device *device; + struct sk_buff *msg; + int ret; + + ret = nlmsg_parse(nlh, 0, tb, RDMA_NLDEV_ATTR_MAX - 1, + nldev_policy, extack); + /* Currently only counter for QP is supported */ + if (ret || !tb[RDMA_NLDEV_ATTR_STAT_RES] || + !tb[RDMA_NLDEV_ATTR_DEV_INDEX] || + !tb[RDMA_NLDEV_ATTR_PORT_INDEX] || !tb[RDMA_NLDEV_ATTR_STAT_MODE]) + return -EINVAL; + + if (nla_get_u32(tb[RDMA_NLDEV_ATTR_STAT_RES]) != RDMA_NLDEV_ATTR_RES_QP) + return -EINVAL; + + index = nla_get_u32(tb[RDMA_NLDEV_ATTR_DEV_INDEX]); + device = ib_device_get_by_index(sock_net(skb->sk), index); + if (!device) + return -EINVAL; + + port = nla_get_u32(tb[RDMA_NLDEV_ATTR_PORT_INDEX]); + if (!rdma_is_port_valid(device, port)) { + ret = -EINVAL; + goto err; + } + + msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); + if (!msg) { + ret = -ENOMEM; + goto err; + } + nlh = nlmsg_put(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, + RDMA_NL_GET_TYPE(RDMA_NL_NLDEV, + RDMA_NLDEV_CMD_STAT_SET), + 0, 0); + + mode = nla_get_u32(tb[RDMA_NLDEV_ATTR_STAT_MODE]); + if (mode == RDMA_COUNTER_MODE_AUTO) { + if (tb[RDMA_NLDEV_ATTR_STAT_AUTO_MODE_MASK]) + mask = nla_get_u32( + tb[RDMA_NLDEV_ATTR_STAT_AUTO_MODE_MASK]); + + ret = rdma_counter_set_auto_mode(device, port, + mask ? true : false, mask); + if (ret) + goto err_msg; + } else { + if (!tb[RDMA_NLDEV_ATTR_RES_LQPN]) + goto err_msg; + qpn = nla_get_u32(tb[RDMA_NLDEV_ATTR_RES_LQPN]); + if (tb[RDMA_NLDEV_ATTR_STAT_COUNTER_ID]) { + cntn = nla_get_u32(tb[RDMA_NLDEV_ATTR_STAT_COUNTER_ID]); + ret = rdma_counter_bind_qpn(device, port, qpn, cntn); + } else { + ret = rdma_counter_bind_qpn_alloc(device, port, + qpn, &cntn); + } + if (ret) + goto err_msg; + + if (fill_nldev_handle(msg, device) || + nla_put_u32(msg, RDMA_NLDEV_ATTR_PORT_INDEX, port) || + nla_put_u32(msg, RDMA_NLDEV_ATTR_STAT_COUNTER_ID, cntn) || + nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_LQPN, qpn)) { + ret = -EMSGSIZE; + goto err_fill; + } + } + + nlmsg_end(msg, nlh); + ib_device_put(device); + return rdma_nl_unicast(sock_net(skb->sk), msg, NETLINK_CB(skb).portid); + +err_fill: + rdma_counter_unbind_qpn(device, port, qpn, cntn); +err_msg: + nlmsg_free(msg); +err: + ib_device_put(device); + return ret; +} + +static int nldev_stat_del_doit(struct sk_buff *skb, struct nlmsghdr *nlh, + struct netlink_ext_ack *extack) +{ + struct nlattr *tb[RDMA_NLDEV_ATTR_MAX]; + struct ib_device *device; + struct sk_buff *msg; + u32 index, port, qpn, cntn; + int ret; + + ret = nlmsg_parse(nlh, 0, tb, RDMA_NLDEV_ATTR_MAX - 1, + nldev_policy, extack); + if (ret || !tb[RDMA_NLDEV_ATTR_STAT_RES] || + !tb[RDMA_NLDEV_ATTR_DEV_INDEX] || !tb[RDMA_NLDEV_ATTR_PORT_INDEX] || + !tb[RDMA_NLDEV_ATTR_STAT_COUNTER_ID] || + !tb[RDMA_NLDEV_ATTR_RES_LQPN]) + return -EINVAL; + + if (nla_get_u32(tb[RDMA_NLDEV_ATTR_STAT_RES]) != RDMA_NLDEV_ATTR_RES_QP) + return -EINVAL; + + index = nla_get_u32(tb[RDMA_NLDEV_ATTR_DEV_INDEX]); + device = ib_device_get_by_index(sock_net(skb->sk), index); + if (!device) + return -EINVAL; + + port = nla_get_u32(tb[RDMA_NLDEV_ATTR_PORT_INDEX]); + if (!rdma_is_port_valid(device, port)) { + ret = -EINVAL; + goto err; + } + + msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); + if (!msg) { + ret = -ENOMEM; + goto err; + } + nlh = nlmsg_put(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, + RDMA_NL_GET_TYPE(RDMA_NL_NLDEV, + RDMA_NLDEV_CMD_STAT_SET), + 0, 0); + + cntn = nla_get_u32(tb[RDMA_NLDEV_ATTR_STAT_COUNTER_ID]); + qpn = nla_get_u32(tb[RDMA_NLDEV_ATTR_RES_LQPN]); + if (fill_nldev_handle(msg, device) || + nla_put_u32(msg, RDMA_NLDEV_ATTR_PORT_INDEX, port) || + nla_put_u32(msg, RDMA_NLDEV_ATTR_STAT_COUNTER_ID, cntn) || + nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_LQPN, qpn)) { + ret = -EMSGSIZE; + goto err_fill; + } + + ret = rdma_counter_unbind_qpn(device, port, qpn, cntn); + if (ret) + goto err_fill; + + nlmsg_end(msg, nlh); + ib_device_put(device); + return rdma_nl_unicast(sock_net(skb->sk), msg, NETLINK_CB(skb).portid); + +err_fill: + nlmsg_free(msg); +err: + ib_device_put(device); + return ret; +} + +static int stat_get_doit_default_counter(struct sk_buff *skb, + struct nlmsghdr *nlh, + struct netlink_ext_ack *extack, + struct nlattr *tb[]) +{ + struct rdma_hw_stats *stats; + struct nlattr *table_attr; + struct ib_device *device; + int ret, num_cnts, i; + struct sk_buff *msg; + u32 index, port; + u64 v; + + if (!tb[RDMA_NLDEV_ATTR_DEV_INDEX] || !tb[RDMA_NLDEV_ATTR_PORT_INDEX]) + return -EINVAL; + + index = nla_get_u32(tb[RDMA_NLDEV_ATTR_DEV_INDEX]); + device = ib_device_get_by_index(sock_net(skb->sk), index); + if (!device) + return -EINVAL; + + if (!device->ops.alloc_hw_stats || !device->ops.get_hw_stats) { + ret = -EINVAL; + goto err; + } + + port = nla_get_u32(tb[RDMA_NLDEV_ATTR_PORT_INDEX]); + if (!rdma_is_port_valid(device, port)) { + ret = -EINVAL; + goto err; + } + + msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); + if (!msg) { + ret = -ENOMEM; + goto err; + } + + nlh = nlmsg_put(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, + RDMA_NL_GET_TYPE(RDMA_NL_NLDEV, + RDMA_NLDEV_CMD_STAT_GET), + 0, 0); + + if (fill_nldev_handle(msg, device) || + nla_put_u32(msg, RDMA_NLDEV_ATTR_PORT_INDEX, port)) { + ret = -EMSGSIZE; + goto err_msg; + } + + stats = device->port_data ? device->port_data[port].hw_stats : NULL; + if (stats == NULL) { + ret = -EINVAL; + goto err_msg; + } + mutex_lock(&stats->lock); + + num_cnts = device->ops.get_hw_stats(device, stats, port, 0); + if (num_cnts < 0) { + ret = -EINVAL; + goto err_stats; + } + + table_attr = nla_nest_start(msg, RDMA_NLDEV_ATTR_STAT_HWCOUNTERS); + if (!table_attr) { + ret = -EMSGSIZE; + goto err_stats; + } + for (i = 0; i < num_cnts; i++) { + v = stats->value[i] + + rdma_counter_get_hwstat_value(device, port, i); + if (rdma_nl_stat_hwcounter_entry(msg, stats->names[i], v)) { + ret = -EMSGSIZE; + goto err_table; + } + } + nla_nest_end(msg, table_attr); + + mutex_unlock(&stats->lock); + nlmsg_end(msg, nlh); + ib_device_put(device); + return rdma_nl_unicast(sock_net(skb->sk), msg, NETLINK_CB(skb).portid); + +err_table: + nla_nest_cancel(msg, table_attr); +err_stats: + mutex_unlock(&stats->lock); +err_msg: + nlmsg_free(msg); +err: + ib_device_put(device); + return ret; +} + +static int stat_get_doit_qp(struct sk_buff *skb, struct nlmsghdr *nlh, + struct netlink_ext_ack *extack, struct nlattr *tb[]) + +{ + static enum rdma_nl_counter_mode mode; + static enum rdma_nl_counter_mask mask; + struct ib_device *device; + struct sk_buff *msg; + u32 index, port; + int ret; + + if (tb[RDMA_NLDEV_ATTR_STAT_COUNTER_ID]) + return nldev_res_get_counter_doit(skb, nlh, extack); + + if (!tb[RDMA_NLDEV_ATTR_STAT_MODE] || + !tb[RDMA_NLDEV_ATTR_DEV_INDEX] || !tb[RDMA_NLDEV_ATTR_PORT_INDEX]) + return -EINVAL; + + index = nla_get_u32(tb[RDMA_NLDEV_ATTR_DEV_INDEX]); + device = ib_device_get_by_index(sock_net(skb->sk), index); + if (!device) + return -EINVAL; + + port = nla_get_u32(tb[RDMA_NLDEV_ATTR_PORT_INDEX]); + if (!rdma_is_port_valid(device, port)) { + ret = -EINVAL; + goto err; + } + + msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); + if (!msg) { + ret = -ENOMEM; + goto err; + } + + nlh = nlmsg_put(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq, + RDMA_NL_GET_TYPE(RDMA_NL_NLDEV, + RDMA_NLDEV_CMD_STAT_GET), + 0, 0); + + ret = rdma_counter_get_mode(device, port, &mode, &mask); + if (ret) + goto err_msg; + + if (fill_nldev_handle(msg, device) || + nla_put_u32(msg, RDMA_NLDEV_ATTR_PORT_INDEX, port) || + nla_put_u32(msg, RDMA_NLDEV_ATTR_STAT_MODE, mode)) { + ret = -EMSGSIZE; + goto err_msg; + } + + if ((mode == RDMA_COUNTER_MODE_AUTO) && + nla_put_u32(msg, RDMA_NLDEV_ATTR_STAT_AUTO_MODE_MASK, mask)) { + ret = -EMSGSIZE; + goto err_msg; + } + + nlmsg_end(msg, nlh); + ib_device_put(device); + return rdma_nl_unicast(sock_net(skb->sk), msg, NETLINK_CB(skb).portid); + +err_msg: + nlmsg_free(msg); +err: + ib_device_put(device); + return ret; +} + +static int nldev_stat_get_doit(struct sk_buff *skb, struct nlmsghdr *nlh, + struct netlink_ext_ack *extack) +{ + struct nlattr *tb[RDMA_NLDEV_ATTR_MAX]; + int ret; + + ret = nlmsg_parse(nlh, 0, tb, RDMA_NLDEV_ATTR_MAX - 1, + nldev_policy, extack); + if (ret) + return -EINVAL; + + if (!tb[RDMA_NLDEV_ATTR_STAT_RES]) + return stat_get_doit_default_counter(skb, nlh, extack, tb); + + switch (nla_get_u32(tb[RDMA_NLDEV_ATTR_STAT_RES])) { + case RDMA_NLDEV_ATTR_RES_QP: + ret = stat_get_doit_qp(skb, nlh, extack, tb); + break; + case RDMA_NLDEV_ATTR_RES_MR: + ret = res_get_common_doit(skb, nlh, extack, RDMA_RESTRACK_MR, + fill_stat_mr_entry); + break; + default: + ret = -EINVAL; + break; + } + + return ret; +} + +static int nldev_stat_get_dumpit(struct sk_buff *skb, + struct netlink_callback *cb) +{ + struct nlattr *tb[RDMA_NLDEV_ATTR_MAX]; + int ret; + + ret = nlmsg_parse(cb->nlh, 0, tb, RDMA_NLDEV_ATTR_MAX - 1, + nldev_policy, NULL); + if (ret || !tb[RDMA_NLDEV_ATTR_STAT_RES]) + return -EINVAL; + + switch (nla_get_u32(tb[RDMA_NLDEV_ATTR_STAT_RES])) { + case RDMA_NLDEV_ATTR_RES_QP: + ret = nldev_res_get_counter_dumpit(skb, cb); + break; + case RDMA_NLDEV_ATTR_RES_MR: + ret = res_get_common_dumpit(skb, cb, RDMA_RESTRACK_MR, + fill_stat_mr_entry); + break; + default: + ret = -EINVAL; + break; + } + + return ret; +} + +static const struct rdma_nl_cbs nldev_cb_table[RDMA_NLDEV_NUM_OPS] = { + [RDMA_NLDEV_CMD_GET] = { + .doit = nldev_get_doit, + .dump = nldev_get_dumpit, + }, + [RDMA_NLDEV_CMD_GET_CHARDEV] = { + .doit = nldev_get_chardev, + }, + [RDMA_NLDEV_CMD_SET] = { + .doit = nldev_set_doit, + .flags = RDMA_NL_ADMIN_PERM, + }, + [RDMA_NLDEV_CMD_NEWLINK] = { + .doit = nldev_newlink, + .flags = RDMA_NL_ADMIN_PERM, + }, + [RDMA_NLDEV_CMD_DELLINK] = { + .doit = nldev_dellink, + .flags = RDMA_NL_ADMIN_PERM, + }, + [RDMA_NLDEV_CMD_PORT_GET] = { + .doit = nldev_port_get_doit, + .dump = nldev_port_get_dumpit, + }, + [RDMA_NLDEV_CMD_RES_GET] = { + .doit = nldev_res_get_doit, + .dump = nldev_res_get_dumpit, + }, + [RDMA_NLDEV_CMD_RES_QP_GET] = { + .doit = nldev_res_get_qp_doit, + .dump = nldev_res_get_qp_dumpit, + }, + [RDMA_NLDEV_CMD_RES_CM_ID_GET] = { + .doit = nldev_res_get_cm_id_doit, + .dump = nldev_res_get_cm_id_dumpit, + }, + [RDMA_NLDEV_CMD_RES_CQ_GET] = { + .doit = nldev_res_get_cq_doit, + .dump = nldev_res_get_cq_dumpit, + }, + [RDMA_NLDEV_CMD_RES_MR_GET] = { + .doit = nldev_res_get_mr_doit, + .dump = nldev_res_get_mr_dumpit, + }, + [RDMA_NLDEV_CMD_RES_PD_GET] = { + .doit = nldev_res_get_pd_doit, + .dump = nldev_res_get_pd_dumpit, + }, + [RDMA_NLDEV_CMD_SYS_GET] = { + .doit = nldev_sys_get_doit, + }, + [RDMA_NLDEV_CMD_SYS_SET] = { + .doit = nldev_set_sys_set_doit, + .flags = RDMA_NL_ADMIN_PERM, + }, + [RDMA_NLDEV_CMD_STAT_SET] = { + .doit = nldev_stat_set_doit, + .flags = RDMA_NL_ADMIN_PERM, + }, + [RDMA_NLDEV_CMD_STAT_GET] = { + .doit = nldev_stat_get_doit, + .dump = nldev_stat_get_dumpit, + }, + [RDMA_NLDEV_CMD_STAT_DEL] = { + .doit = nldev_stat_del_doit, + .flags = RDMA_NL_ADMIN_PERM, + }, + [RDMA_NLDEV_CMD_RES_QP_GET_RAW] = { + .doit = nldev_res_get_qp_raw_doit, + .dump = nldev_res_get_qp_raw_dumpit, + .flags = RDMA_NL_ADMIN_PERM, + }, + [RDMA_NLDEV_CMD_RES_CQ_GET_RAW] = { + .doit = nldev_res_get_cq_raw_doit, + .dump = nldev_res_get_cq_raw_dumpit, + .flags = RDMA_NL_ADMIN_PERM, + }, + [RDMA_NLDEV_CMD_RES_MR_GET_RAW] = { + .doit = nldev_res_get_mr_raw_doit, + .dump = nldev_res_get_mr_raw_dumpit, + .flags = RDMA_NL_ADMIN_PERM, + }, +}; + +void __init nldev_init(void) +{ + rdma_nl_register(RDMA_NL_NLDEV, nldev_cb_table); +} + +void nldev_exit(void) +{ + rdma_nl_unregister(RDMA_NL_NLDEV); +} + +MODULE_ALIAS_RDMA_NETLINK(RDMA_NL_NLDEV, 5); diff --git a/drivers/infiniband/core/opa_smi.h b/drivers/infiniband/core/opa_smi.h new file mode 100644 index 000000000..af4879bdf --- /dev/null +++ b/drivers/infiniband/core/opa_smi.h @@ -0,0 +1,78 @@ +/* + * Copyright (c) 2014 Intel Corporation. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + */ + +#ifndef __OPA_SMI_H_ +#define __OPA_SMI_H_ + +#include <rdma/ib_smi.h> +#include <rdma/opa_smi.h> + +#include "smi.h" + +enum smi_action opa_smi_handle_dr_smp_recv(struct opa_smp *smp, bool is_switch, + int port_num, int phys_port_cnt); +int opa_smi_get_fwd_port(struct opa_smp *smp); +extern enum smi_forward_action opa_smi_check_forward_dr_smp(struct opa_smp *smp); +extern enum smi_action opa_smi_handle_dr_smp_send(struct opa_smp *smp, + bool is_switch, int port_num); + +/* + * Return IB_SMI_HANDLE if the SMP should be handled by the local SMA/SM + * via process_mad + */ +static inline enum smi_action opa_smi_check_local_smp(struct opa_smp *smp, + struct ib_device *device) +{ + /* C14-9:3 -- We're at the end of the DR segment of path */ + /* C14-9:4 -- Hop Pointer = Hop Count + 1 -> give to SMA/SM */ + return (device->ops.process_mad && + !opa_get_smp_direction(smp) && + (smp->hop_ptr == smp->hop_cnt + 1)) ? + IB_SMI_HANDLE : IB_SMI_DISCARD; +} + +/* + * Return IB_SMI_HANDLE if the SMP should be handled by the local SMA/SM + * via process_mad + */ +static inline enum smi_action opa_smi_check_local_returning_smp(struct opa_smp *smp, + struct ib_device *device) +{ + /* C14-13:3 -- We're at the end of the DR segment of path */ + /* C14-13:4 -- Hop Pointer == 0 -> give to SM */ + return (device->ops.process_mad && + opa_get_smp_direction(smp) && + !smp->hop_ptr) ? IB_SMI_HANDLE : IB_SMI_DISCARD; +} + +#endif /* __OPA_SMI_H_ */ diff --git a/drivers/infiniband/core/packer.c b/drivers/infiniband/core/packer.c new file mode 100644 index 000000000..19b1ee327 --- /dev/null +++ b/drivers/infiniband/core/packer.c @@ -0,0 +1,201 @@ +/* + * Copyright (c) 2004 Topspin Corporation. All rights reserved. + * Copyright (c) 2005 Sun Microsystems, Inc. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include <linux/export.h> +#include <linux/string.h> + +#include <rdma/ib_pack.h> + +static u64 value_read(int offset, int size, void *structure) +{ + switch (size) { + case 1: return *(u8 *) (structure + offset); + case 2: return be16_to_cpup((__be16 *) (structure + offset)); + case 4: return be32_to_cpup((__be32 *) (structure + offset)); + case 8: return be64_to_cpup((__be64 *) (structure + offset)); + default: + pr_warn("Field size %d bits not handled\n", size * 8); + return 0; + } +} + +/** + * ib_pack - Pack a structure into a buffer + * @desc:Array of structure field descriptions + * @desc_len:Number of entries in @desc + * @structure:Structure to pack from + * @buf:Buffer to pack into + * + * ib_pack() packs a list of structure fields into a buffer, + * controlled by the array of fields in @desc. + */ +void ib_pack(const struct ib_field *desc, + int desc_len, + void *structure, + void *buf) +{ + int i; + + for (i = 0; i < desc_len; ++i) { + if (desc[i].size_bits <= 32) { + int shift; + u32 val; + __be32 mask; + __be32 *addr; + + shift = 32 - desc[i].offset_bits - desc[i].size_bits; + if (desc[i].struct_size_bytes) + val = value_read(desc[i].struct_offset_bytes, + desc[i].struct_size_bytes, + structure) << shift; + else + val = 0; + + mask = cpu_to_be32(((1ull << desc[i].size_bits) - 1) << shift); + addr = (__be32 *) buf + desc[i].offset_words; + *addr = (*addr & ~mask) | (cpu_to_be32(val) & mask); + } else if (desc[i].size_bits <= 64) { + int shift; + u64 val; + __be64 mask; + __be64 *addr; + + shift = 64 - desc[i].offset_bits - desc[i].size_bits; + if (desc[i].struct_size_bytes) + val = value_read(desc[i].struct_offset_bytes, + desc[i].struct_size_bytes, + structure) << shift; + else + val = 0; + + mask = cpu_to_be64((~0ull >> (64 - desc[i].size_bits)) << shift); + addr = (__be64 *) ((__be32 *) buf + desc[i].offset_words); + *addr = (*addr & ~mask) | (cpu_to_be64(val) & mask); + } else { + if (desc[i].offset_bits % 8 || + desc[i].size_bits % 8) { + pr_warn("Structure field %s of size %d bits is not byte-aligned\n", + desc[i].field_name, desc[i].size_bits); + } + + if (desc[i].struct_size_bytes) + memcpy(buf + desc[i].offset_words * 4 + + desc[i].offset_bits / 8, + structure + desc[i].struct_offset_bytes, + desc[i].size_bits / 8); + else + memset(buf + desc[i].offset_words * 4 + + desc[i].offset_bits / 8, + 0, + desc[i].size_bits / 8); + } + } +} +EXPORT_SYMBOL(ib_pack); + +static void value_write(int offset, int size, u64 val, void *structure) +{ + switch (size * 8) { + case 8: *( u8 *) (structure + offset) = val; break; + case 16: *(__be16 *) (structure + offset) = cpu_to_be16(val); break; + case 32: *(__be32 *) (structure + offset) = cpu_to_be32(val); break; + case 64: *(__be64 *) (structure + offset) = cpu_to_be64(val); break; + default: + pr_warn("Field size %d bits not handled\n", size * 8); + } +} + +/** + * ib_unpack - Unpack a buffer into a structure + * @desc:Array of structure field descriptions + * @desc_len:Number of entries in @desc + * @buf:Buffer to unpack from + * @structure:Structure to unpack into + * + * ib_pack() unpacks a list of structure fields from a buffer, + * controlled by the array of fields in @desc. + */ +void ib_unpack(const struct ib_field *desc, + int desc_len, + void *buf, + void *structure) +{ + int i; + + for (i = 0; i < desc_len; ++i) { + if (!desc[i].struct_size_bytes) + continue; + + if (desc[i].size_bits <= 32) { + int shift; + u32 val; + u32 mask; + __be32 *addr; + + shift = 32 - desc[i].offset_bits - desc[i].size_bits; + mask = ((1ull << desc[i].size_bits) - 1) << shift; + addr = (__be32 *) buf + desc[i].offset_words; + val = (be32_to_cpup(addr) & mask) >> shift; + value_write(desc[i].struct_offset_bytes, + desc[i].struct_size_bytes, + val, + structure); + } else if (desc[i].size_bits <= 64) { + int shift; + u64 val; + u64 mask; + __be64 *addr; + + shift = 64 - desc[i].offset_bits - desc[i].size_bits; + mask = (~0ull >> (64 - desc[i].size_bits)) << shift; + addr = (__be64 *) buf + desc[i].offset_words; + val = (be64_to_cpup(addr) & mask) >> shift; + value_write(desc[i].struct_offset_bytes, + desc[i].struct_size_bytes, + val, + structure); + } else { + if (desc[i].offset_bits % 8 || + desc[i].size_bits % 8) { + pr_warn("Structure field %s of size %d bits is not byte-aligned\n", + desc[i].field_name, desc[i].size_bits); + } + + memcpy(structure + desc[i].struct_offset_bytes, + buf + desc[i].offset_words * 4 + + desc[i].offset_bits / 8, + desc[i].size_bits / 8); + } + } +} +EXPORT_SYMBOL(ib_unpack); diff --git a/drivers/infiniband/core/rdma_core.c b/drivers/infiniband/core/rdma_core.c new file mode 100644 index 000000000..ffe11b037 --- /dev/null +++ b/drivers/infiniband/core/rdma_core.c @@ -0,0 +1,980 @@ +/* + * Copyright (c) 2016, Mellanox Technologies inc. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include <linux/file.h> +#include <linux/anon_inodes.h> +#include <linux/sched/mm.h> +#include <rdma/ib_verbs.h> +#include <rdma/uverbs_types.h> +#include <linux/rcupdate.h> +#include <rdma/uverbs_ioctl.h> +#include <rdma/rdma_user_ioctl.h> +#include "uverbs.h" +#include "core_priv.h" +#include "rdma_core.h" + +static void uverbs_uobject_free(struct kref *ref) +{ + kfree_rcu(container_of(ref, struct ib_uobject, ref), rcu); +} + +/* + * In order to indicate we no longer needs this uobject, uverbs_uobject_put + * is called. When the reference count is decreased, the uobject is freed. + * For example, this is used when attaching a completion channel to a CQ. + */ +void uverbs_uobject_put(struct ib_uobject *uobject) +{ + kref_put(&uobject->ref, uverbs_uobject_free); +} +EXPORT_SYMBOL(uverbs_uobject_put); + +static int uverbs_try_lock_object(struct ib_uobject *uobj, + enum rdma_lookup_mode mode) +{ + /* + * When a shared access is required, we use a positive counter. Each + * shared access request checks that the value != -1 and increment it. + * Exclusive access is required for operations like write or destroy. + * In exclusive access mode, we check that the counter is zero (nobody + * claimed this object) and we set it to -1. Releasing a shared access + * lock is done simply by decreasing the counter. As for exclusive + * access locks, since only a single one of them is is allowed + * concurrently, setting the counter to zero is enough for releasing + * this lock. + */ + switch (mode) { + case UVERBS_LOOKUP_READ: + return atomic_fetch_add_unless(&uobj->usecnt, 1, -1) == -1 ? + -EBUSY : 0; + case UVERBS_LOOKUP_WRITE: + /* lock is exclusive */ + return atomic_cmpxchg(&uobj->usecnt, 0, -1) == 0 ? 0 : -EBUSY; + case UVERBS_LOOKUP_DESTROY: + return 0; + } + return 0; +} + +static void assert_uverbs_usecnt(struct ib_uobject *uobj, + enum rdma_lookup_mode mode) +{ +#ifdef CONFIG_LOCKDEP + switch (mode) { + case UVERBS_LOOKUP_READ: + WARN_ON(atomic_read(&uobj->usecnt) <= 0); + break; + case UVERBS_LOOKUP_WRITE: + WARN_ON(atomic_read(&uobj->usecnt) != -1); + break; + case UVERBS_LOOKUP_DESTROY: + break; + } +#endif +} + +/* + * This must be called with the hw_destroy_rwsem locked for read or write, + * also the uobject itself must be locked for write. + * + * Upon return the HW object is guaranteed to be destroyed. + * + * For RDMA_REMOVE_ABORT, the hw_destroy_rwsem is not required to be held, + * however the type's allocat_commit function cannot have been called and the + * uobject cannot be on the uobjects_lists + * + * For RDMA_REMOVE_DESTROY the caller shold be holding a kref (eg via + * rdma_lookup_get_uobject) and the object is left in a state where the caller + * needs to call rdma_lookup_put_uobject. + * + * For all other destroy modes this function internally unlocks the uobject + * and consumes the kref on the uobj. + */ +static int uverbs_destroy_uobject(struct ib_uobject *uobj, + enum rdma_remove_reason reason, + struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_file *ufile = attrs->ufile; + unsigned long flags; + int ret; + + lockdep_assert_held(&ufile->hw_destroy_rwsem); + assert_uverbs_usecnt(uobj, UVERBS_LOOKUP_WRITE); + + if (reason == RDMA_REMOVE_ABORT) { + WARN_ON(!list_empty(&uobj->list)); + WARN_ON(!uobj->context); + uobj->uapi_object->type_class->alloc_abort(uobj); + } else if (uobj->object) { + ret = uobj->uapi_object->type_class->destroy_hw(uobj, reason, + attrs); + if (ret) { + if (ib_is_destroy_retryable(ret, reason, uobj)) + return ret; + + /* Nothing to be done, dangle the memory and move on */ + WARN(true, + "ib_uverbs: failed to remove uobject id %d, driver err=%d", + uobj->id, ret); + } + + uobj->object = NULL; + } + + uobj->context = NULL; + + /* + * For DESTROY the usecnt is not changed, the caller is expected to + * manage it via uobj_put_destroy(). Only DESTROY can remove the IDR + * handle. + */ + if (reason != RDMA_REMOVE_DESTROY) + atomic_set(&uobj->usecnt, 0); + else + uobj->uapi_object->type_class->remove_handle(uobj); + + if (!list_empty(&uobj->list)) { + spin_lock_irqsave(&ufile->uobjects_lock, flags); + list_del_init(&uobj->list); + spin_unlock_irqrestore(&ufile->uobjects_lock, flags); + + /* + * Pairs with the get in rdma_alloc_commit_uobject(), could + * destroy uobj. + */ + uverbs_uobject_put(uobj); + } + + /* + * When aborting the stack kref remains owned by the core code, and is + * not transferred into the type. Pairs with the get in alloc_uobj + */ + if (reason == RDMA_REMOVE_ABORT) + uverbs_uobject_put(uobj); + + return 0; +} + +/* + * This calls uverbs_destroy_uobject() using the RDMA_REMOVE_DESTROY + * sequence. It should only be used from command callbacks. On success the + * caller must pair this with uobj_put_destroy(). This + * version requires the caller to have already obtained an + * LOOKUP_DESTROY uobject kref. + */ +int uobj_destroy(struct ib_uobject *uobj, struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_file *ufile = attrs->ufile; + int ret; + + down_read(&ufile->hw_destroy_rwsem); + + /* + * Once the uobject is destroyed by RDMA_REMOVE_DESTROY then it is left + * write locked as the callers put it back with UVERBS_LOOKUP_DESTROY. + * This is because any other concurrent thread can still see the object + * in the xarray due to RCU. Leaving it locked ensures nothing else will + * touch it. + */ + ret = uverbs_try_lock_object(uobj, UVERBS_LOOKUP_WRITE); + if (ret) + goto out_unlock; + + ret = uverbs_destroy_uobject(uobj, RDMA_REMOVE_DESTROY, attrs); + if (ret) { + atomic_set(&uobj->usecnt, 0); + goto out_unlock; + } + +out_unlock: + up_read(&ufile->hw_destroy_rwsem); + return ret; +} + +/* + * uobj_get_destroy destroys the HW object and returns a handle to the uobj + * with a NULL object pointer. The caller must pair this with + * uobj_put_destroy(). + */ +struct ib_uobject *__uobj_get_destroy(const struct uverbs_api_object *obj, + u32 id, struct uverbs_attr_bundle *attrs) +{ + struct ib_uobject *uobj; + int ret; + + uobj = rdma_lookup_get_uobject(obj, attrs->ufile, id, + UVERBS_LOOKUP_DESTROY, attrs); + if (IS_ERR(uobj)) + return uobj; + + ret = uobj_destroy(uobj, attrs); + if (ret) { + rdma_lookup_put_uobject(uobj, UVERBS_LOOKUP_DESTROY); + return ERR_PTR(ret); + } + + return uobj; +} + +/* + * Does both uobj_get_destroy() and uobj_put_destroy(). Returns 0 on success + * (negative errno on failure). For use by callers that do not need the uobj. + */ +int __uobj_perform_destroy(const struct uverbs_api_object *obj, u32 id, + struct uverbs_attr_bundle *attrs) +{ + struct ib_uobject *uobj; + + uobj = __uobj_get_destroy(obj, id, attrs); + if (IS_ERR(uobj)) + return PTR_ERR(uobj); + uobj_put_destroy(uobj); + return 0; +} + +/* alloc_uobj must be undone by uverbs_destroy_uobject() */ +static struct ib_uobject *alloc_uobj(struct uverbs_attr_bundle *attrs, + const struct uverbs_api_object *obj) +{ + struct ib_uverbs_file *ufile = attrs->ufile; + struct ib_uobject *uobj; + + if (!attrs->context) { + struct ib_ucontext *ucontext = + ib_uverbs_get_ucontext_file(ufile); + + if (IS_ERR(ucontext)) + return ERR_CAST(ucontext); + attrs->context = ucontext; + } + + uobj = kzalloc(obj->type_attrs->obj_size, GFP_KERNEL); + if (!uobj) + return ERR_PTR(-ENOMEM); + /* + * user_handle should be filled by the handler, + * The object is added to the list in the commit stage. + */ + uobj->ufile = ufile; + uobj->context = attrs->context; + INIT_LIST_HEAD(&uobj->list); + uobj->uapi_object = obj; + /* + * Allocated objects start out as write locked to deny any other + * syscalls from accessing them until they are committed. See + * rdma_alloc_commit_uobject + */ + atomic_set(&uobj->usecnt, -1); + kref_init(&uobj->ref); + + return uobj; +} + +static int idr_add_uobj(struct ib_uobject *uobj) +{ + /* + * We start with allocating an idr pointing to NULL. This represents an + * object which isn't initialized yet. We'll replace it later on with + * the real object once we commit. + */ + return xa_alloc(&uobj->ufile->idr, &uobj->id, NULL, xa_limit_32b, + GFP_KERNEL); +} + +/* Returns the ib_uobject or an error. The caller should check for IS_ERR. */ +static struct ib_uobject * +lookup_get_idr_uobject(const struct uverbs_api_object *obj, + struct ib_uverbs_file *ufile, s64 id, + enum rdma_lookup_mode mode) +{ + struct ib_uobject *uobj; + + if (id < 0 || id > ULONG_MAX) + return ERR_PTR(-EINVAL); + + rcu_read_lock(); + /* + * The idr_find is guaranteed to return a pointer to something that + * isn't freed yet, or NULL, as the free after idr_remove goes through + * kfree_rcu(). However the object may still have been released and + * kfree() could be called at any time. + */ + uobj = xa_load(&ufile->idr, id); + if (!uobj || !kref_get_unless_zero(&uobj->ref)) + uobj = ERR_PTR(-ENOENT); + rcu_read_unlock(); + return uobj; +} + +static struct ib_uobject * +lookup_get_fd_uobject(const struct uverbs_api_object *obj, + struct ib_uverbs_file *ufile, s64 id, + enum rdma_lookup_mode mode) +{ + const struct uverbs_obj_fd_type *fd_type; + struct file *f; + struct ib_uobject *uobject; + int fdno = id; + + if (fdno != id) + return ERR_PTR(-EINVAL); + + if (mode != UVERBS_LOOKUP_READ) + return ERR_PTR(-EOPNOTSUPP); + + if (!obj->type_attrs) + return ERR_PTR(-EIO); + fd_type = + container_of(obj->type_attrs, struct uverbs_obj_fd_type, type); + + f = fget(fdno); + if (!f) + return ERR_PTR(-EBADF); + + uobject = f->private_data; + /* + * fget(id) ensures we are not currently running + * uverbs_uobject_fd_release(), and the caller is expected to ensure + * that release is never done while a call to lookup is possible. + */ + if (f->f_op != fd_type->fops || uobject->ufile != ufile) { + fput(f); + return ERR_PTR(-EBADF); + } + + uverbs_uobject_get(uobject); + return uobject; +} + +struct ib_uobject *rdma_lookup_get_uobject(const struct uverbs_api_object *obj, + struct ib_uverbs_file *ufile, s64 id, + enum rdma_lookup_mode mode, + struct uverbs_attr_bundle *attrs) +{ + struct ib_uobject *uobj; + int ret; + + if (obj == ERR_PTR(-ENOMSG)) { + /* must be UVERBS_IDR_ANY_OBJECT, see uapi_get_object() */ + uobj = lookup_get_idr_uobject(NULL, ufile, id, mode); + if (IS_ERR(uobj)) + return uobj; + } else { + if (IS_ERR(obj)) + return ERR_PTR(-EINVAL); + + uobj = obj->type_class->lookup_get(obj, ufile, id, mode); + if (IS_ERR(uobj)) + return uobj; + + if (uobj->uapi_object != obj) { + ret = -EINVAL; + goto free; + } + } + + /* + * If we have been disassociated block every command except for + * DESTROY based commands. + */ + if (mode != UVERBS_LOOKUP_DESTROY && + !srcu_dereference(ufile->device->ib_dev, + &ufile->device->disassociate_srcu)) { + ret = -EIO; + goto free; + } + + ret = uverbs_try_lock_object(uobj, mode); + if (ret) + goto free; + if (attrs) + attrs->context = uobj->context; + + return uobj; +free: + uobj->uapi_object->type_class->lookup_put(uobj, mode); + uverbs_uobject_put(uobj); + return ERR_PTR(ret); +} + +static struct ib_uobject * +alloc_begin_idr_uobject(const struct uverbs_api_object *obj, + struct uverbs_attr_bundle *attrs) +{ + int ret; + struct ib_uobject *uobj; + + uobj = alloc_uobj(attrs, obj); + if (IS_ERR(uobj)) + return uobj; + + ret = idr_add_uobj(uobj); + if (ret) + goto uobj_put; + + ret = ib_rdmacg_try_charge(&uobj->cg_obj, uobj->context->device, + RDMACG_RESOURCE_HCA_OBJECT); + if (ret) + goto remove; + + return uobj; + +remove: + xa_erase(&attrs->ufile->idr, uobj->id); +uobj_put: + uverbs_uobject_put(uobj); + return ERR_PTR(ret); +} + +static struct ib_uobject * +alloc_begin_fd_uobject(const struct uverbs_api_object *obj, + struct uverbs_attr_bundle *attrs) +{ + const struct uverbs_obj_fd_type *fd_type; + int new_fd; + struct ib_uobject *uobj, *ret; + struct file *filp; + + uobj = alloc_uobj(attrs, obj); + if (IS_ERR(uobj)) + return uobj; + + fd_type = + container_of(obj->type_attrs, struct uverbs_obj_fd_type, type); + if (WARN_ON(fd_type->fops->release != &uverbs_uobject_fd_release && + fd_type->fops->release != &uverbs_async_event_release)) { + ret = ERR_PTR(-EINVAL); + goto err_fd; + } + + new_fd = get_unused_fd_flags(O_CLOEXEC); + if (new_fd < 0) { + ret = ERR_PTR(new_fd); + goto err_fd; + } + + /* Note that uverbs_uobject_fd_release() is called during abort */ + filp = anon_inode_getfile(fd_type->name, fd_type->fops, NULL, + fd_type->flags); + if (IS_ERR(filp)) { + ret = ERR_CAST(filp); + goto err_getfile; + } + uobj->object = filp; + + uobj->id = new_fd; + return uobj; + +err_getfile: + put_unused_fd(new_fd); +err_fd: + uverbs_uobject_put(uobj); + return ret; +} + +struct ib_uobject *rdma_alloc_begin_uobject(const struct uverbs_api_object *obj, + struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_file *ufile = attrs->ufile; + struct ib_uobject *ret; + + if (IS_ERR(obj)) + return ERR_PTR(-EINVAL); + + /* + * The hw_destroy_rwsem is held across the entire object creation and + * released during rdma_alloc_commit_uobject or + * rdma_alloc_abort_uobject + */ + if (!down_read_trylock(&ufile->hw_destroy_rwsem)) + return ERR_PTR(-EIO); + + ret = obj->type_class->alloc_begin(obj, attrs); + if (IS_ERR(ret)) { + up_read(&ufile->hw_destroy_rwsem); + return ret; + } + return ret; +} + +static void alloc_abort_idr_uobject(struct ib_uobject *uobj) +{ + ib_rdmacg_uncharge(&uobj->cg_obj, uobj->context->device, + RDMACG_RESOURCE_HCA_OBJECT); + + xa_erase(&uobj->ufile->idr, uobj->id); +} + +static int __must_check destroy_hw_idr_uobject(struct ib_uobject *uobj, + enum rdma_remove_reason why, + struct uverbs_attr_bundle *attrs) +{ + const struct uverbs_obj_idr_type *idr_type = + container_of(uobj->uapi_object->type_attrs, + struct uverbs_obj_idr_type, type); + int ret = idr_type->destroy_object(uobj, why, attrs); + + /* + * We can only fail gracefully if the user requested to destroy the + * object or when a retry may be called upon an error. + * In the rest of the cases, just remove whatever you can. + */ + if (ib_is_destroy_retryable(ret, why, uobj)) + return ret; + + if (why == RDMA_REMOVE_ABORT) + return 0; + + ib_rdmacg_uncharge(&uobj->cg_obj, uobj->context->device, + RDMACG_RESOURCE_HCA_OBJECT); + + return 0; +} + +static void remove_handle_idr_uobject(struct ib_uobject *uobj) +{ + xa_erase(&uobj->ufile->idr, uobj->id); + /* Matches the kref in alloc_commit_idr_uobject */ + uverbs_uobject_put(uobj); +} + +static void alloc_abort_fd_uobject(struct ib_uobject *uobj) +{ + struct file *filp = uobj->object; + + fput(filp); + put_unused_fd(uobj->id); +} + +static int __must_check destroy_hw_fd_uobject(struct ib_uobject *uobj, + enum rdma_remove_reason why, + struct uverbs_attr_bundle *attrs) +{ + const struct uverbs_obj_fd_type *fd_type = container_of( + uobj->uapi_object->type_attrs, struct uverbs_obj_fd_type, type); + int ret = fd_type->destroy_object(uobj, why); + + if (ib_is_destroy_retryable(ret, why, uobj)) + return ret; + + return 0; +} + +static void remove_handle_fd_uobject(struct ib_uobject *uobj) +{ +} + +static void alloc_commit_idr_uobject(struct ib_uobject *uobj) +{ + struct ib_uverbs_file *ufile = uobj->ufile; + void *old; + + /* + * We already allocated this IDR with a NULL object, so + * this shouldn't fail. + * + * NOTE: Storing the uobj transfers our kref on uobj to the XArray. + * It will be put by remove_commit_idr_uobject() + */ + old = xa_store(&ufile->idr, uobj->id, uobj, GFP_KERNEL); + WARN_ON(old != NULL); +} + +static void alloc_commit_fd_uobject(struct ib_uobject *uobj) +{ + int fd = uobj->id; + struct file *filp = uobj->object; + + /* Matching put will be done in uverbs_uobject_fd_release() */ + kref_get(&uobj->ufile->ref); + + /* This shouldn't be used anymore. Use the file object instead */ + uobj->id = 0; + + /* + * NOTE: Once we install the file we loose ownership of our kref on + * uobj. It will be put by uverbs_uobject_fd_release() + */ + filp->private_data = uobj; + fd_install(fd, filp); +} + +/* + * In all cases rdma_alloc_commit_uobject() consumes the kref to uobj and the + * caller can no longer assume uobj is valid. If this function fails it + * destroys the uboject, including the attached HW object. + */ +void rdma_alloc_commit_uobject(struct ib_uobject *uobj, + struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_file *ufile = attrs->ufile; + + /* kref is held so long as the uobj is on the uobj list. */ + uverbs_uobject_get(uobj); + spin_lock_irq(&ufile->uobjects_lock); + list_add(&uobj->list, &ufile->uobjects); + spin_unlock_irq(&ufile->uobjects_lock); + + /* matches atomic_set(-1) in alloc_uobj */ + atomic_set(&uobj->usecnt, 0); + + /* alloc_commit consumes the uobj kref */ + uobj->uapi_object->type_class->alloc_commit(uobj); + + /* Matches the down_read in rdma_alloc_begin_uobject */ + up_read(&ufile->hw_destroy_rwsem); +} + +/* + * This consumes the kref for uobj. It is up to the caller to unwind the HW + * object and anything else connected to uobj before calling this. + */ +void rdma_alloc_abort_uobject(struct ib_uobject *uobj, + struct uverbs_attr_bundle *attrs, + bool hw_obj_valid) +{ + struct ib_uverbs_file *ufile = uobj->ufile; + int ret; + + if (hw_obj_valid) { + ret = uobj->uapi_object->type_class->destroy_hw( + uobj, RDMA_REMOVE_ABORT, attrs); + /* + * If the driver couldn't destroy the object then go ahead and + * commit it. Leaking objects that can't be destroyed is only + * done during FD close after the driver has a few more tries to + * destroy it. + */ + if (WARN_ON(ret)) + return rdma_alloc_commit_uobject(uobj, attrs); + } + + uverbs_destroy_uobject(uobj, RDMA_REMOVE_ABORT, attrs); + + /* Matches the down_read in rdma_alloc_begin_uobject */ + up_read(&ufile->hw_destroy_rwsem); +} + +static void lookup_put_idr_uobject(struct ib_uobject *uobj, + enum rdma_lookup_mode mode) +{ +} + +static void lookup_put_fd_uobject(struct ib_uobject *uobj, + enum rdma_lookup_mode mode) +{ + struct file *filp = uobj->object; + + WARN_ON(mode != UVERBS_LOOKUP_READ); + /* + * This indirectly calls uverbs_uobject_fd_release() and free the + * object + */ + fput(filp); +} + +void rdma_lookup_put_uobject(struct ib_uobject *uobj, + enum rdma_lookup_mode mode) +{ + assert_uverbs_usecnt(uobj, mode); + /* + * In order to unlock an object, either decrease its usecnt for + * read access or zero it in case of exclusive access. See + * uverbs_try_lock_object for locking schema information. + */ + switch (mode) { + case UVERBS_LOOKUP_READ: + atomic_dec(&uobj->usecnt); + break; + case UVERBS_LOOKUP_WRITE: + atomic_set(&uobj->usecnt, 0); + break; + case UVERBS_LOOKUP_DESTROY: + break; + } + + uobj->uapi_object->type_class->lookup_put(uobj, mode); + /* Pairs with the kref obtained by type->lookup_get */ + uverbs_uobject_put(uobj); +} + +void setup_ufile_idr_uobject(struct ib_uverbs_file *ufile) +{ + xa_init_flags(&ufile->idr, XA_FLAGS_ALLOC); +} + +void release_ufile_idr_uobject(struct ib_uverbs_file *ufile) +{ + struct ib_uobject *entry; + unsigned long id; + + /* + * At this point uverbs_cleanup_ufile() is guaranteed to have run, and + * there are no HW objects left, however the xarray is still populated + * with anything that has not been cleaned up by userspace. Since the + * kref on ufile is 0, nothing is allowed to call lookup_get. + * + * This is an optimized equivalent to remove_handle_idr_uobject + */ + xa_for_each(&ufile->idr, id, entry) { + WARN_ON(entry->object); + uverbs_uobject_put(entry); + } + + xa_destroy(&ufile->idr); +} + +const struct uverbs_obj_type_class uverbs_idr_class = { + .alloc_begin = alloc_begin_idr_uobject, + .lookup_get = lookup_get_idr_uobject, + .alloc_commit = alloc_commit_idr_uobject, + .alloc_abort = alloc_abort_idr_uobject, + .lookup_put = lookup_put_idr_uobject, + .destroy_hw = destroy_hw_idr_uobject, + .remove_handle = remove_handle_idr_uobject, +}; +EXPORT_SYMBOL(uverbs_idr_class); + +/* + * Users of UVERBS_TYPE_ALLOC_FD should set this function as the struct + * file_operations release method. + */ +int uverbs_uobject_fd_release(struct inode *inode, struct file *filp) +{ + struct ib_uverbs_file *ufile; + struct ib_uobject *uobj; + + /* + * This can only happen if the fput came from alloc_abort_fd_uobject() + */ + if (!filp->private_data) + return 0; + uobj = filp->private_data; + ufile = uobj->ufile; + + if (down_read_trylock(&ufile->hw_destroy_rwsem)) { + struct uverbs_attr_bundle attrs = { + .context = uobj->context, + .ufile = ufile, + }; + + /* + * lookup_get_fd_uobject holds the kref on the struct file any + * time a FD uobj is locked, which prevents this release + * method from being invoked. Meaning we can always get the + * write lock here, or we have a kernel bug. + */ + WARN_ON(uverbs_try_lock_object(uobj, UVERBS_LOOKUP_WRITE)); + uverbs_destroy_uobject(uobj, RDMA_REMOVE_CLOSE, &attrs); + up_read(&ufile->hw_destroy_rwsem); + } + + /* Matches the get in alloc_commit_fd_uobject() */ + kref_put(&ufile->ref, ib_uverbs_release_file); + + /* Pairs with filp->private_data in alloc_begin_fd_uobject */ + uverbs_uobject_put(uobj); + return 0; +} +EXPORT_SYMBOL(uverbs_uobject_fd_release); + +/* + * Drop the ucontext off the ufile and completely disconnect it from the + * ib_device + */ +static void ufile_destroy_ucontext(struct ib_uverbs_file *ufile, + enum rdma_remove_reason reason) +{ + struct ib_ucontext *ucontext = ufile->ucontext; + struct ib_device *ib_dev = ucontext->device; + + /* + * If we are closing the FD then the user mmap VMAs must have + * already been destroyed as they hold on to the filep, otherwise + * they need to be zap'd. + */ + if (reason == RDMA_REMOVE_DRIVER_REMOVE) { + uverbs_user_mmap_disassociate(ufile); + if (ib_dev->ops.disassociate_ucontext) + ib_dev->ops.disassociate_ucontext(ucontext); + } + + ib_rdmacg_uncharge(&ucontext->cg_obj, ib_dev, + RDMACG_RESOURCE_HCA_HANDLE); + + rdma_restrack_del(&ucontext->res); + + ib_dev->ops.dealloc_ucontext(ucontext); + WARN_ON(!xa_empty(&ucontext->mmap_xa)); + kfree(ucontext); + + ufile->ucontext = NULL; +} + +static int __uverbs_cleanup_ufile(struct ib_uverbs_file *ufile, + enum rdma_remove_reason reason) +{ + struct ib_uobject *obj, *next_obj; + int ret = -EINVAL; + struct uverbs_attr_bundle attrs = { .ufile = ufile }; + + /* + * This shouldn't run while executing other commands on this + * context. Thus, the only thing we should take care of is + * releasing a FD while traversing this list. The FD could be + * closed and released from the _release fop of this FD. + * In order to mitigate this, we add a lock. + * We take and release the lock per traversal in order to let + * other threads (which might still use the FDs) chance to run. + */ + list_for_each_entry_safe(obj, next_obj, &ufile->uobjects, list) { + attrs.context = obj->context; + /* + * if we hit this WARN_ON, that means we are + * racing with a lookup_get. + */ + WARN_ON(uverbs_try_lock_object(obj, UVERBS_LOOKUP_WRITE)); + if (!uverbs_destroy_uobject(obj, reason, &attrs)) + ret = 0; + else + atomic_set(&obj->usecnt, 0); + } + return ret; +} + +/* + * Destroy the uncontext and every uobject associated with it. + * + * This is internally locked and can be called in parallel from multiple + * contexts. + */ +void uverbs_destroy_ufile_hw(struct ib_uverbs_file *ufile, + enum rdma_remove_reason reason) +{ + down_write(&ufile->hw_destroy_rwsem); + + /* + * If a ucontext was never created then we can't have any uobjects to + * cleanup, nothing to do. + */ + if (!ufile->ucontext) + goto done; + + ufile->ucontext->cleanup_retryable = true; + while (!list_empty(&ufile->uobjects)) + if (__uverbs_cleanup_ufile(ufile, reason)) { + /* + * No entry was cleaned-up successfully during this + * iteration. It is a driver bug to fail destruction. + */ + WARN_ON(!list_empty(&ufile->uobjects)); + break; + } + + ufile->ucontext->cleanup_retryable = false; + if (!list_empty(&ufile->uobjects)) + __uverbs_cleanup_ufile(ufile, reason); + + ufile_destroy_ucontext(ufile, reason); + +done: + up_write(&ufile->hw_destroy_rwsem); +} + +const struct uverbs_obj_type_class uverbs_fd_class = { + .alloc_begin = alloc_begin_fd_uobject, + .lookup_get = lookup_get_fd_uobject, + .alloc_commit = alloc_commit_fd_uobject, + .alloc_abort = alloc_abort_fd_uobject, + .lookup_put = lookup_put_fd_uobject, + .destroy_hw = destroy_hw_fd_uobject, + .remove_handle = remove_handle_fd_uobject, +}; +EXPORT_SYMBOL(uverbs_fd_class); + +struct ib_uobject * +uverbs_get_uobject_from_file(u16 object_id, enum uverbs_obj_access access, + s64 id, struct uverbs_attr_bundle *attrs) +{ + const struct uverbs_api_object *obj = + uapi_get_object(attrs->ufile->device->uapi, object_id); + + switch (access) { + case UVERBS_ACCESS_READ: + return rdma_lookup_get_uobject(obj, attrs->ufile, id, + UVERBS_LOOKUP_READ, attrs); + case UVERBS_ACCESS_DESTROY: + /* Actual destruction is done inside uverbs_handle_method */ + return rdma_lookup_get_uobject(obj, attrs->ufile, id, + UVERBS_LOOKUP_DESTROY, attrs); + case UVERBS_ACCESS_WRITE: + return rdma_lookup_get_uobject(obj, attrs->ufile, id, + UVERBS_LOOKUP_WRITE, attrs); + case UVERBS_ACCESS_NEW: + return rdma_alloc_begin_uobject(obj, attrs); + default: + WARN_ON(true); + return ERR_PTR(-EOPNOTSUPP); + } +} + +void uverbs_finalize_object(struct ib_uobject *uobj, + enum uverbs_obj_access access, bool hw_obj_valid, + bool commit, struct uverbs_attr_bundle *attrs) +{ + /* + * refcounts should be handled at the object level and not at the + * uobject level. Refcounts of the objects themselves are done in + * handlers. + */ + + switch (access) { + case UVERBS_ACCESS_READ: + rdma_lookup_put_uobject(uobj, UVERBS_LOOKUP_READ); + break; + case UVERBS_ACCESS_WRITE: + rdma_lookup_put_uobject(uobj, UVERBS_LOOKUP_WRITE); + break; + case UVERBS_ACCESS_DESTROY: + if (uobj) + rdma_lookup_put_uobject(uobj, UVERBS_LOOKUP_DESTROY); + break; + case UVERBS_ACCESS_NEW: + if (commit) + rdma_alloc_commit_uobject(uobj, attrs); + else + rdma_alloc_abort_uobject(uobj, attrs, hw_obj_valid); + break; + default: + WARN_ON(true); + } +} diff --git a/drivers/infiniband/core/rdma_core.h b/drivers/infiniband/core/rdma_core.h new file mode 100644 index 000000000..33706dad6 --- /dev/null +++ b/drivers/infiniband/core/rdma_core.h @@ -0,0 +1,191 @@ +/* + * Copyright (c) 2005 Topspin Communications. All rights reserved. + * Copyright (c) 2005, 2006 Cisco Systems. All rights reserved. + * Copyright (c) 2005-2017 Mellanox Technologies. All rights reserved. + * Copyright (c) 2005 Voltaire, Inc. All rights reserved. + * Copyright (c) 2005 PathScale, Inc. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#ifndef RDMA_CORE_H +#define RDMA_CORE_H + +#include <linux/idr.h> +#include <rdma/uverbs_types.h> +#include <rdma/uverbs_ioctl.h> +#include <rdma/ib_verbs.h> +#include <linux/mutex.h> + +struct ib_uverbs_device; + +void uverbs_destroy_ufile_hw(struct ib_uverbs_file *ufile, + enum rdma_remove_reason reason); + +int uobj_destroy(struct ib_uobject *uobj, struct uverbs_attr_bundle *attrs); + +/* + * Get an ib_uobject that corresponds to the given id from ufile, assuming + * the object is from the given type. Lock it to the required access when + * applicable. + * This function could create (access == NEW), destroy (access == DESTROY) + * or unlock (access == READ || access == WRITE) objects if required. + * The action will be finalized only when uverbs_finalize_object or + * uverbs_finalize_objects are called. + */ +struct ib_uobject * +uverbs_get_uobject_from_file(u16 object_id, enum uverbs_obj_access access, + s64 id, struct uverbs_attr_bundle *attrs); + +void uverbs_finalize_object(struct ib_uobject *uobj, + enum uverbs_obj_access access, bool hw_obj_valid, + bool commit, struct uverbs_attr_bundle *attrs); + +int uverbs_output_written(const struct uverbs_attr_bundle *bundle, size_t idx); + +void setup_ufile_idr_uobject(struct ib_uverbs_file *ufile); +void release_ufile_idr_uobject(struct ib_uverbs_file *ufile); + +struct ib_udata *uverbs_get_cleared_udata(struct uverbs_attr_bundle *attrs); + +/* + * This is the runtime description of the uverbs API, used by the syscall + * machinery to validate and dispatch calls. + */ + +/* + * Depending on ID the slot pointer in the radix tree points at one of these + * structs. + */ + +struct uverbs_api_ioctl_method { + int(__rcu *handler)(struct uverbs_attr_bundle *attrs); + DECLARE_BITMAP(attr_mandatory, UVERBS_API_ATTR_BKEY_LEN); + u16 bundle_size; + u8 use_stack:1; + u8 driver_method:1; + u8 disabled:1; + u8 has_udata:1; + u8 key_bitmap_len; + u8 destroy_bkey; +}; + +struct uverbs_api_write_method { + int (*handler)(struct uverbs_attr_bundle *attrs); + u8 disabled:1; + u8 is_ex:1; + u8 has_udata:1; + u8 has_resp:1; + u8 req_size; + u8 resp_size; +}; + +struct uverbs_api_attr { + struct uverbs_attr_spec spec; +}; + +struct uverbs_api { + /* radix tree contains struct uverbs_api_* pointers */ + struct radix_tree_root radix; + enum rdma_driver_id driver_id; + + unsigned int num_write; + unsigned int num_write_ex; + struct uverbs_api_write_method notsupp_method; + const struct uverbs_api_write_method **write_methods; + const struct uverbs_api_write_method **write_ex_methods; +}; + +/* + * Get an uverbs_api_object that corresponds to the given object_id. + * Note: + * -ENOMSG means that any object is allowed to match during lookup. + */ +static inline const struct uverbs_api_object * +uapi_get_object(struct uverbs_api *uapi, u16 object_id) +{ + const struct uverbs_api_object *res; + + if (object_id == UVERBS_IDR_ANY_OBJECT) + return ERR_PTR(-ENOMSG); + + res = radix_tree_lookup(&uapi->radix, uapi_key_obj(object_id)); + if (!res) + return ERR_PTR(-ENOENT); + + return res; +} + +char *uapi_key_format(char *S, unsigned int key); +struct uverbs_api *uverbs_alloc_api(struct ib_device *ibdev); +void uverbs_disassociate_api_pre(struct ib_uverbs_device *uverbs_dev); +void uverbs_disassociate_api(struct uverbs_api *uapi); +void uverbs_destroy_api(struct uverbs_api *uapi); +void uapi_compute_bundle_size(struct uverbs_api_ioctl_method *method_elm, + unsigned int num_attrs); +void uverbs_user_mmap_disassociate(struct ib_uverbs_file *ufile); + +extern const struct uapi_definition uverbs_def_obj_async_fd[]; +extern const struct uapi_definition uverbs_def_obj_counters[]; +extern const struct uapi_definition uverbs_def_obj_cq[]; +extern const struct uapi_definition uverbs_def_obj_device[]; +extern const struct uapi_definition uverbs_def_obj_dm[]; +extern const struct uapi_definition uverbs_def_obj_flow_action[]; +extern const struct uapi_definition uverbs_def_obj_intf[]; +extern const struct uapi_definition uverbs_def_obj_mr[]; +extern const struct uapi_definition uverbs_def_obj_qp[]; +extern const struct uapi_definition uverbs_def_obj_srq[]; +extern const struct uapi_definition uverbs_def_obj_wq[]; +extern const struct uapi_definition uverbs_def_write_intf[]; + +static inline const struct uverbs_api_write_method * +uapi_get_method(const struct uverbs_api *uapi, u32 command) +{ + u32 cmd_idx = command & IB_USER_VERBS_CMD_COMMAND_MASK; + + if (command & ~(u32)(IB_USER_VERBS_CMD_FLAG_EXTENDED | + IB_USER_VERBS_CMD_COMMAND_MASK)) + return ERR_PTR(-EINVAL); + + if (command & IB_USER_VERBS_CMD_FLAG_EXTENDED) { + if (cmd_idx >= uapi->num_write_ex) + return ERR_PTR(-EOPNOTSUPP); + return uapi->write_ex_methods[cmd_idx]; + } + + if (cmd_idx >= uapi->num_write) + return ERR_PTR(-EOPNOTSUPP); + return uapi->write_methods[cmd_idx]; +} + +void uverbs_fill_udata(struct uverbs_attr_bundle *bundle, + struct ib_udata *udata, unsigned int attr_in, + unsigned int attr_out); + +#endif /* RDMA_CORE_H */ diff --git a/drivers/infiniband/core/restrack.c b/drivers/infiniband/core/restrack.c new file mode 100644 index 000000000..bbbbec5b1 --- /dev/null +++ b/drivers/infiniband/core/restrack.c @@ -0,0 +1,337 @@ +// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB +/* + * Copyright (c) 2017-2018 Mellanox Technologies. All rights reserved. + */ + +#include <rdma/rdma_cm.h> +#include <rdma/ib_verbs.h> +#include <rdma/restrack.h> +#include <rdma/rdma_counter.h> +#include <linux/mutex.h> +#include <linux/sched/task.h> +#include <linux/pid_namespace.h> + +#include "cma_priv.h" +#include "restrack.h" + +/** + * rdma_restrack_init() - initialize and allocate resource tracking + * @dev: IB device + * + * Return: 0 on success + */ +int rdma_restrack_init(struct ib_device *dev) +{ + struct rdma_restrack_root *rt; + int i; + + dev->res = kcalloc(RDMA_RESTRACK_MAX, sizeof(*rt), GFP_KERNEL); + if (!dev->res) + return -ENOMEM; + + rt = dev->res; + + for (i = 0; i < RDMA_RESTRACK_MAX; i++) + xa_init_flags(&rt[i].xa, XA_FLAGS_ALLOC); + + return 0; +} + +static const char *type2str(enum rdma_restrack_type type) +{ + static const char * const names[RDMA_RESTRACK_MAX] = { + [RDMA_RESTRACK_PD] = "PD", + [RDMA_RESTRACK_CQ] = "CQ", + [RDMA_RESTRACK_QP] = "QP", + [RDMA_RESTRACK_CM_ID] = "CM_ID", + [RDMA_RESTRACK_MR] = "MR", + [RDMA_RESTRACK_CTX] = "CTX", + [RDMA_RESTRACK_COUNTER] = "COUNTER", + }; + + return names[type]; +}; + +/** + * rdma_restrack_clean() - clean resource tracking + * @dev: IB device + */ +void rdma_restrack_clean(struct ib_device *dev) +{ + struct rdma_restrack_root *rt = dev->res; + struct rdma_restrack_entry *e; + char buf[TASK_COMM_LEN]; + bool found = false; + const char *owner; + int i; + + for (i = 0 ; i < RDMA_RESTRACK_MAX; i++) { + struct xarray *xa = &dev->res[i].xa; + + if (!xa_empty(xa)) { + unsigned long index; + + if (!found) { + pr_err("restrack: %s", CUT_HERE); + dev_err(&dev->dev, "BUG: RESTRACK detected leak of resources\n"); + } + xa_for_each(xa, index, e) { + if (rdma_is_kernel_res(e)) { + owner = e->kern_name; + } else { + /* + * There is no need to call get_task_struct here, + * because we can be here only if there are more + * get_task_struct() call than put_task_struct(). + */ + get_task_comm(buf, e->task); + owner = buf; + } + + pr_err("restrack: %s %s object allocated by %s is not freed\n", + rdma_is_kernel_res(e) ? "Kernel" : + "User", + type2str(e->type), owner); + } + found = true; + } + xa_destroy(xa); + } + if (found) + pr_err("restrack: %s", CUT_HERE); + + kfree(rt); +} + +/** + * rdma_restrack_count() - the current usage of specific object + * @dev: IB device + * @type: actual type of object to operate + */ +int rdma_restrack_count(struct ib_device *dev, enum rdma_restrack_type type) +{ + struct rdma_restrack_root *rt = &dev->res[type]; + struct rdma_restrack_entry *e; + XA_STATE(xas, &rt->xa, 0); + u32 cnt = 0; + + xa_lock(&rt->xa); + xas_for_each(&xas, e, U32_MAX) + cnt++; + xa_unlock(&rt->xa); + return cnt; +} +EXPORT_SYMBOL(rdma_restrack_count); + +static struct ib_device *res_to_dev(struct rdma_restrack_entry *res) +{ + switch (res->type) { + case RDMA_RESTRACK_PD: + return container_of(res, struct ib_pd, res)->device; + case RDMA_RESTRACK_CQ: + return container_of(res, struct ib_cq, res)->device; + case RDMA_RESTRACK_QP: + return container_of(res, struct ib_qp, res)->device; + case RDMA_RESTRACK_CM_ID: + return container_of(res, struct rdma_id_private, + res)->id.device; + case RDMA_RESTRACK_MR: + return container_of(res, struct ib_mr, res)->device; + case RDMA_RESTRACK_CTX: + return container_of(res, struct ib_ucontext, res)->device; + case RDMA_RESTRACK_COUNTER: + return container_of(res, struct rdma_counter, res)->device; + default: + WARN_ONCE(true, "Wrong resource tracking type %u\n", res->type); + return NULL; + } +} + +/** + * rdma_restrack_attach_task() - attach the task onto this resource, + * valid for user space restrack entries. + * @res: resource entry + * @task: the task to attach + */ +static void rdma_restrack_attach_task(struct rdma_restrack_entry *res, + struct task_struct *task) +{ + if (WARN_ON_ONCE(!task)) + return; + + if (res->task) + put_task_struct(res->task); + get_task_struct(task); + res->task = task; + res->user = true; +} + +/** + * rdma_restrack_set_name() - set the task for this resource + * @res: resource entry + * @caller: kernel name, the current task will be used if the caller is NULL. + */ +void rdma_restrack_set_name(struct rdma_restrack_entry *res, const char *caller) +{ + if (caller) { + res->kern_name = caller; + return; + } + + rdma_restrack_attach_task(res, current); +} +EXPORT_SYMBOL(rdma_restrack_set_name); + +/** + * rdma_restrack_parent_name() - set the restrack name properties based + * on parent restrack + * @dst: destination resource entry + * @parent: parent resource entry + */ +void rdma_restrack_parent_name(struct rdma_restrack_entry *dst, + const struct rdma_restrack_entry *parent) +{ + if (rdma_is_kernel_res(parent)) + dst->kern_name = parent->kern_name; + else + rdma_restrack_attach_task(dst, parent->task); +} +EXPORT_SYMBOL(rdma_restrack_parent_name); + +/** + * rdma_restrack_new() - Initializes new restrack entry to allow _put() interface + * to release memory in fully automatic way. + * @res - Entry to initialize + * @type - REstrack type + */ +void rdma_restrack_new(struct rdma_restrack_entry *res, + enum rdma_restrack_type type) +{ + kref_init(&res->kref); + init_completion(&res->comp); + res->type = type; +} +EXPORT_SYMBOL(rdma_restrack_new); + +/** + * rdma_restrack_add() - add object to the reource tracking database + * @res: resource entry + */ +void rdma_restrack_add(struct rdma_restrack_entry *res) +{ + struct ib_device *dev = res_to_dev(res); + struct rdma_restrack_root *rt; + int ret; + + if (!dev) + return; + + rt = &dev->res[res->type]; + + if (res->type == RDMA_RESTRACK_QP) { + /* Special case to ensure that LQPN points to right QP */ + struct ib_qp *qp = container_of(res, struct ib_qp, res); + + ret = xa_insert(&rt->xa, qp->qp_num, res, GFP_KERNEL); + res->id = ret ? 0 : qp->qp_num; + } else if (res->type == RDMA_RESTRACK_COUNTER) { + /* Special case to ensure that cntn points to right counter */ + struct rdma_counter *counter; + + counter = container_of(res, struct rdma_counter, res); + ret = xa_insert(&rt->xa, counter->id, res, GFP_KERNEL); + res->id = ret ? 0 : counter->id; + } else { + ret = xa_alloc_cyclic(&rt->xa, &res->id, res, xa_limit_32b, + &rt->next_id, GFP_KERNEL); + ret = (ret < 0) ? ret : 0; + } + + if (!ret) + res->valid = true; +} +EXPORT_SYMBOL(rdma_restrack_add); + +int __must_check rdma_restrack_get(struct rdma_restrack_entry *res) +{ + return kref_get_unless_zero(&res->kref); +} +EXPORT_SYMBOL(rdma_restrack_get); + +/** + * rdma_restrack_get_byid() - translate from ID to restrack object + * @dev: IB device + * @type: resource track type + * @id: ID to take a look + * + * Return: Pointer to restrack entry or -ENOENT in case of error. + */ +struct rdma_restrack_entry * +rdma_restrack_get_byid(struct ib_device *dev, + enum rdma_restrack_type type, u32 id) +{ + struct rdma_restrack_root *rt = &dev->res[type]; + struct rdma_restrack_entry *res; + + xa_lock(&rt->xa); + res = xa_load(&rt->xa, id); + if (!res || !rdma_restrack_get(res)) + res = ERR_PTR(-ENOENT); + xa_unlock(&rt->xa); + + return res; +} +EXPORT_SYMBOL(rdma_restrack_get_byid); + +static void restrack_release(struct kref *kref) +{ + struct rdma_restrack_entry *res; + + res = container_of(kref, struct rdma_restrack_entry, kref); + if (res->task) { + put_task_struct(res->task); + res->task = NULL; + } + complete(&res->comp); +} + +int rdma_restrack_put(struct rdma_restrack_entry *res) +{ + return kref_put(&res->kref, restrack_release); +} +EXPORT_SYMBOL(rdma_restrack_put); + +/** + * rdma_restrack_del() - delete object from the reource tracking database + * @res: resource entry + */ +void rdma_restrack_del(struct rdma_restrack_entry *res) +{ + struct rdma_restrack_entry *old; + struct rdma_restrack_root *rt; + struct ib_device *dev; + + if (!res->valid) { + if (res->task) { + put_task_struct(res->task); + res->task = NULL; + } + return; + } + + dev = res_to_dev(res); + if (WARN_ON(!dev)) + return; + + rt = &dev->res[res->type]; + + old = xa_erase(&rt->xa, res->id); + if (res->type == RDMA_RESTRACK_MR || res->type == RDMA_RESTRACK_QP) + return; + WARN_ON(old != res); + res->valid = false; + + rdma_restrack_put(res); + wait_for_completion(&res->comp); +} +EXPORT_SYMBOL(rdma_restrack_del); diff --git a/drivers/infiniband/core/restrack.h b/drivers/infiniband/core/restrack.h new file mode 100644 index 000000000..6a04fc41f --- /dev/null +++ b/drivers/infiniband/core/restrack.h @@ -0,0 +1,36 @@ +/* SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB */ +/* + * Copyright (c) 2017-2019 Mellanox Technologies. All rights reserved. + */ + +#ifndef _RDMA_CORE_RESTRACK_H_ +#define _RDMA_CORE_RESTRACK_H_ + +#include <linux/mutex.h> + +/** + * struct rdma_restrack_root - main resource tracking management + * entity, per-device + */ +struct rdma_restrack_root { + /** + * @xa: Array of XArray structure to hold restrack entries. + */ + struct xarray xa; + /** + * @next_id: Next ID to support cyclic allocation + */ + u32 next_id; +}; + +int rdma_restrack_init(struct ib_device *dev); +void rdma_restrack_clean(struct ib_device *dev); +void rdma_restrack_add(struct rdma_restrack_entry *res); +void rdma_restrack_del(struct rdma_restrack_entry *res); +void rdma_restrack_new(struct rdma_restrack_entry *res, + enum rdma_restrack_type type); +void rdma_restrack_set_name(struct rdma_restrack_entry *res, + const char *caller); +void rdma_restrack_parent_name(struct rdma_restrack_entry *dst, + const struct rdma_restrack_entry *parent); +#endif /* _RDMA_CORE_RESTRACK_H_ */ diff --git a/drivers/infiniband/core/roce_gid_mgmt.c b/drivers/infiniband/core/roce_gid_mgmt.c new file mode 100644 index 000000000..6b8364bb0 --- /dev/null +++ b/drivers/infiniband/core/roce_gid_mgmt.c @@ -0,0 +1,928 @@ +/* + * Copyright (c) 2015, Mellanox Technologies inc. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include "core_priv.h" + +#include <linux/in.h> +#include <linux/in6.h> + +/* For in6_dev_get/in6_dev_put */ +#include <net/addrconf.h> +#include <net/bonding.h> + +#include <rdma/ib_cache.h> +#include <rdma/ib_addr.h> + +static struct workqueue_struct *gid_cache_wq; + +enum gid_op_type { + GID_DEL = 0, + GID_ADD +}; + +struct update_gid_event_work { + struct work_struct work; + union ib_gid gid; + struct ib_gid_attr gid_attr; + enum gid_op_type gid_op; +}; + +#define ROCE_NETDEV_CALLBACK_SZ 3 +struct netdev_event_work_cmd { + roce_netdev_callback cb; + roce_netdev_filter filter; + struct net_device *ndev; + struct net_device *filter_ndev; +}; + +struct netdev_event_work { + struct work_struct work; + struct netdev_event_work_cmd cmds[ROCE_NETDEV_CALLBACK_SZ]; +}; + +static const struct { + bool (*is_supported)(const struct ib_device *device, u8 port_num); + enum ib_gid_type gid_type; +} PORT_CAP_TO_GID_TYPE[] = { + {rdma_protocol_roce_eth_encap, IB_GID_TYPE_ROCE}, + {rdma_protocol_roce_udp_encap, IB_GID_TYPE_ROCE_UDP_ENCAP}, +}; + +#define CAP_TO_GID_TABLE_SIZE ARRAY_SIZE(PORT_CAP_TO_GID_TYPE) + +unsigned long roce_gid_type_mask_support(struct ib_device *ib_dev, u8 port) +{ + int i; + unsigned int ret_flags = 0; + + if (!rdma_protocol_roce(ib_dev, port)) + return 1UL << IB_GID_TYPE_IB; + + for (i = 0; i < CAP_TO_GID_TABLE_SIZE; i++) + if (PORT_CAP_TO_GID_TYPE[i].is_supported(ib_dev, port)) + ret_flags |= 1UL << PORT_CAP_TO_GID_TYPE[i].gid_type; + + return ret_flags; +} +EXPORT_SYMBOL(roce_gid_type_mask_support); + +static void update_gid(enum gid_op_type gid_op, struct ib_device *ib_dev, + u8 port, union ib_gid *gid, + struct ib_gid_attr *gid_attr) +{ + int i; + unsigned long gid_type_mask = roce_gid_type_mask_support(ib_dev, port); + + for (i = 0; i < IB_GID_TYPE_SIZE; i++) { + if ((1UL << i) & gid_type_mask) { + gid_attr->gid_type = i; + switch (gid_op) { + case GID_ADD: + ib_cache_gid_add(ib_dev, port, + gid, gid_attr); + break; + case GID_DEL: + ib_cache_gid_del(ib_dev, port, + gid, gid_attr); + break; + } + } + } +} + +enum bonding_slave_state { + BONDING_SLAVE_STATE_ACTIVE = 1UL << 0, + BONDING_SLAVE_STATE_INACTIVE = 1UL << 1, + /* No primary slave or the device isn't a slave in bonding */ + BONDING_SLAVE_STATE_NA = 1UL << 2, +}; + +static enum bonding_slave_state is_eth_active_slave_of_bonding_rcu(struct net_device *dev, + struct net_device *upper) +{ + if (upper && netif_is_bond_master(upper)) { + struct net_device *pdev = + bond_option_active_slave_get_rcu(netdev_priv(upper)); + + if (pdev) + return dev == pdev ? BONDING_SLAVE_STATE_ACTIVE : + BONDING_SLAVE_STATE_INACTIVE; + } + + return BONDING_SLAVE_STATE_NA; +} + +#define REQUIRED_BOND_STATES (BONDING_SLAVE_STATE_ACTIVE | \ + BONDING_SLAVE_STATE_NA) +static bool +is_eth_port_of_netdev_filter(struct ib_device *ib_dev, u8 port, + struct net_device *rdma_ndev, void *cookie) +{ + struct net_device *real_dev; + bool res; + + if (!rdma_ndev) + return false; + + rcu_read_lock(); + real_dev = rdma_vlan_dev_real_dev(cookie); + if (!real_dev) + real_dev = cookie; + + res = ((rdma_is_upper_dev_rcu(rdma_ndev, cookie) && + (is_eth_active_slave_of_bonding_rcu(rdma_ndev, real_dev) & + REQUIRED_BOND_STATES)) || + real_dev == rdma_ndev); + + rcu_read_unlock(); + return res; +} + +static bool +is_eth_port_inactive_slave_filter(struct ib_device *ib_dev, u8 port, + struct net_device *rdma_ndev, void *cookie) +{ + struct net_device *master_dev; + bool res; + + if (!rdma_ndev) + return false; + + rcu_read_lock(); + master_dev = netdev_master_upper_dev_get_rcu(rdma_ndev); + res = is_eth_active_slave_of_bonding_rcu(rdma_ndev, master_dev) == + BONDING_SLAVE_STATE_INACTIVE; + rcu_read_unlock(); + + return res; +} + +/** is_ndev_for_default_gid_filter - Check if a given netdevice + * can be considered for default GIDs or not. + * @ib_dev: IB device to check + * @port: Port to consider for adding default GID + * @rdma_ndev: rdma netdevice pointer + * @cookie_ndev: Netdevice to consider to form a default GID + * + * is_ndev_for_default_gid_filter() returns true if a given netdevice can be + * considered for deriving default RoCE GID, returns false otherwise. + */ +static bool +is_ndev_for_default_gid_filter(struct ib_device *ib_dev, u8 port, + struct net_device *rdma_ndev, void *cookie) +{ + struct net_device *cookie_ndev = cookie; + bool res; + + if (!rdma_ndev) + return false; + + rcu_read_lock(); + + /* + * When rdma netdevice is used in bonding, bonding master netdevice + * should be considered for default GIDs. Therefore, ignore slave rdma + * netdevices when bonding is considered. + * Additionally when event(cookie) netdevice is bond master device, + * make sure that it the upper netdevice of rdma netdevice. + */ + res = ((cookie_ndev == rdma_ndev && !netif_is_bond_slave(rdma_ndev)) || + (netif_is_bond_master(cookie_ndev) && + rdma_is_upper_dev_rcu(rdma_ndev, cookie_ndev))); + + rcu_read_unlock(); + return res; +} + +static bool pass_all_filter(struct ib_device *ib_dev, u8 port, + struct net_device *rdma_ndev, void *cookie) +{ + return true; +} + +static bool upper_device_filter(struct ib_device *ib_dev, u8 port, + struct net_device *rdma_ndev, void *cookie) +{ + bool res; + + if (!rdma_ndev) + return false; + + if (rdma_ndev == cookie) + return true; + + rcu_read_lock(); + res = rdma_is_upper_dev_rcu(rdma_ndev, cookie); + rcu_read_unlock(); + + return res; +} + +/** + * is_upper_ndev_bond_master_filter - Check if a given netdevice + * is bond master device of netdevice of the the RDMA device of port. + * @ib_dev: IB device to check + * @port: Port to consider for adding default GID + * @rdma_ndev: Pointer to rdma netdevice + * @cookie: Netdevice to consider to form a default GID + * + * is_upper_ndev_bond_master_filter() returns true if a cookie_netdev + * is bond master device and rdma_ndev is its lower netdevice. It might + * not have been established as slave device yet. + */ +static bool +is_upper_ndev_bond_master_filter(struct ib_device *ib_dev, u8 port, + struct net_device *rdma_ndev, + void *cookie) +{ + struct net_device *cookie_ndev = cookie; + bool match = false; + + if (!rdma_ndev) + return false; + + rcu_read_lock(); + if (netif_is_bond_master(cookie_ndev) && + rdma_is_upper_dev_rcu(rdma_ndev, cookie_ndev)) + match = true; + rcu_read_unlock(); + return match; +} + +static void update_gid_ip(enum gid_op_type gid_op, + struct ib_device *ib_dev, + u8 port, struct net_device *ndev, + struct sockaddr *addr) +{ + union ib_gid gid; + struct ib_gid_attr gid_attr; + + rdma_ip2gid(addr, &gid); + memset(&gid_attr, 0, sizeof(gid_attr)); + gid_attr.ndev = ndev; + + update_gid(gid_op, ib_dev, port, &gid, &gid_attr); +} + +static void bond_delete_netdev_default_gids(struct ib_device *ib_dev, + u8 port, + struct net_device *rdma_ndev, + struct net_device *event_ndev) +{ + struct net_device *real_dev = rdma_vlan_dev_real_dev(event_ndev); + unsigned long gid_type_mask; + + if (!rdma_ndev) + return; + + if (!real_dev) + real_dev = event_ndev; + + rcu_read_lock(); + + if (((rdma_ndev != event_ndev && + !rdma_is_upper_dev_rcu(rdma_ndev, event_ndev)) || + is_eth_active_slave_of_bonding_rcu(rdma_ndev, real_dev) + == + BONDING_SLAVE_STATE_INACTIVE)) { + rcu_read_unlock(); + return; + } + + rcu_read_unlock(); + + gid_type_mask = roce_gid_type_mask_support(ib_dev, port); + + ib_cache_gid_set_default_gid(ib_dev, port, rdma_ndev, + gid_type_mask, + IB_CACHE_GID_DEFAULT_MODE_DELETE); +} + +static void enum_netdev_ipv4_ips(struct ib_device *ib_dev, + u8 port, struct net_device *ndev) +{ + const struct in_ifaddr *ifa; + struct in_device *in_dev; + struct sin_list { + struct list_head list; + struct sockaddr_in ip; + }; + struct sin_list *sin_iter; + struct sin_list *sin_temp; + + LIST_HEAD(sin_list); + if (ndev->reg_state >= NETREG_UNREGISTERING) + return; + + rcu_read_lock(); + in_dev = __in_dev_get_rcu(ndev); + if (!in_dev) { + rcu_read_unlock(); + return; + } + + in_dev_for_each_ifa_rcu(ifa, in_dev) { + struct sin_list *entry = kzalloc(sizeof(*entry), GFP_ATOMIC); + + if (!entry) + continue; + + entry->ip.sin_family = AF_INET; + entry->ip.sin_addr.s_addr = ifa->ifa_address; + list_add_tail(&entry->list, &sin_list); + } + + rcu_read_unlock(); + + list_for_each_entry_safe(sin_iter, sin_temp, &sin_list, list) { + update_gid_ip(GID_ADD, ib_dev, port, ndev, + (struct sockaddr *)&sin_iter->ip); + list_del(&sin_iter->list); + kfree(sin_iter); + } +} + +static void enum_netdev_ipv6_ips(struct ib_device *ib_dev, + u8 port, struct net_device *ndev) +{ + struct inet6_ifaddr *ifp; + struct inet6_dev *in6_dev; + struct sin6_list { + struct list_head list; + struct sockaddr_in6 sin6; + }; + struct sin6_list *sin6_iter; + struct sin6_list *sin6_temp; + struct ib_gid_attr gid_attr = {.ndev = ndev}; + LIST_HEAD(sin6_list); + + if (ndev->reg_state >= NETREG_UNREGISTERING) + return; + + in6_dev = in6_dev_get(ndev); + if (!in6_dev) + return; + + read_lock_bh(&in6_dev->lock); + list_for_each_entry(ifp, &in6_dev->addr_list, if_list) { + struct sin6_list *entry = kzalloc(sizeof(*entry), GFP_ATOMIC); + + if (!entry) + continue; + + entry->sin6.sin6_family = AF_INET6; + entry->sin6.sin6_addr = ifp->addr; + list_add_tail(&entry->list, &sin6_list); + } + read_unlock_bh(&in6_dev->lock); + + in6_dev_put(in6_dev); + + list_for_each_entry_safe(sin6_iter, sin6_temp, &sin6_list, list) { + union ib_gid gid; + + rdma_ip2gid((struct sockaddr *)&sin6_iter->sin6, &gid); + update_gid(GID_ADD, ib_dev, port, &gid, &gid_attr); + list_del(&sin6_iter->list); + kfree(sin6_iter); + } +} + +static void _add_netdev_ips(struct ib_device *ib_dev, u8 port, + struct net_device *ndev) +{ + enum_netdev_ipv4_ips(ib_dev, port, ndev); + if (IS_ENABLED(CONFIG_IPV6)) + enum_netdev_ipv6_ips(ib_dev, port, ndev); +} + +static void add_netdev_ips(struct ib_device *ib_dev, u8 port, + struct net_device *rdma_ndev, void *cookie) +{ + _add_netdev_ips(ib_dev, port, cookie); +} + +static void del_netdev_ips(struct ib_device *ib_dev, u8 port, + struct net_device *rdma_ndev, void *cookie) +{ + ib_cache_gid_del_all_netdev_gids(ib_dev, port, cookie); +} + +/** + * del_default_gids - Delete default GIDs of the event/cookie netdevice + * @ib_dev: RDMA device pointer + * @port: Port of the RDMA device whose GID table to consider + * @rdma_ndev: Unused rdma netdevice + * @cookie: Pointer to event netdevice + * + * del_default_gids() deletes the default GIDs of the event/cookie netdevice. + */ +static void del_default_gids(struct ib_device *ib_dev, u8 port, + struct net_device *rdma_ndev, void *cookie) +{ + struct net_device *cookie_ndev = cookie; + unsigned long gid_type_mask; + + gid_type_mask = roce_gid_type_mask_support(ib_dev, port); + + ib_cache_gid_set_default_gid(ib_dev, port, cookie_ndev, gid_type_mask, + IB_CACHE_GID_DEFAULT_MODE_DELETE); +} + +static void add_default_gids(struct ib_device *ib_dev, u8 port, + struct net_device *rdma_ndev, void *cookie) +{ + struct net_device *event_ndev = cookie; + unsigned long gid_type_mask; + + gid_type_mask = roce_gid_type_mask_support(ib_dev, port); + ib_cache_gid_set_default_gid(ib_dev, port, event_ndev, gid_type_mask, + IB_CACHE_GID_DEFAULT_MODE_SET); +} + +static void enum_all_gids_of_dev_cb(struct ib_device *ib_dev, + u8 port, + struct net_device *rdma_ndev, + void *cookie) +{ + struct net *net; + struct net_device *ndev; + + /* Lock the rtnl to make sure the netdevs does not move under + * our feet + */ + rtnl_lock(); + down_read(&net_rwsem); + for_each_net(net) + for_each_netdev(net, ndev) { + /* + * Filter and add default GIDs of the primary netdevice + * when not in bonding mode, or add default GIDs + * of bond master device, when in bonding mode. + */ + if (is_ndev_for_default_gid_filter(ib_dev, port, + rdma_ndev, ndev)) + add_default_gids(ib_dev, port, rdma_ndev, ndev); + + if (is_eth_port_of_netdev_filter(ib_dev, port, + rdma_ndev, ndev)) + _add_netdev_ips(ib_dev, port, ndev); + } + up_read(&net_rwsem); + rtnl_unlock(); +} + +/** + * rdma_roce_rescan_device - Rescan all of the network devices in the system + * and add their gids, as needed, to the relevant RoCE devices. + * + * @device: the rdma device + */ +void rdma_roce_rescan_device(struct ib_device *ib_dev) +{ + ib_enum_roce_netdev(ib_dev, pass_all_filter, NULL, + enum_all_gids_of_dev_cb, NULL); +} +EXPORT_SYMBOL(rdma_roce_rescan_device); + +static void callback_for_addr_gid_device_scan(struct ib_device *device, + u8 port, + struct net_device *rdma_ndev, + void *cookie) +{ + struct update_gid_event_work *parsed = cookie; + + return update_gid(parsed->gid_op, device, + port, &parsed->gid, + &parsed->gid_attr); +} + +struct upper_list { + struct list_head list; + struct net_device *upper; +}; + +static int netdev_upper_walk(struct net_device *upper, + struct netdev_nested_priv *priv) +{ + struct upper_list *entry = kmalloc(sizeof(*entry), GFP_ATOMIC); + struct list_head *upper_list = (struct list_head *)priv->data; + + if (!entry) + return 0; + + list_add_tail(&entry->list, upper_list); + dev_hold(upper); + entry->upper = upper; + + return 0; +} + +static void handle_netdev_upper(struct ib_device *ib_dev, u8 port, + void *cookie, + void (*handle_netdev)(struct ib_device *ib_dev, + u8 port, + struct net_device *ndev)) +{ + struct net_device *ndev = cookie; + struct netdev_nested_priv priv; + struct upper_list *upper_iter; + struct upper_list *upper_temp; + LIST_HEAD(upper_list); + + priv.data = &upper_list; + rcu_read_lock(); + netdev_walk_all_upper_dev_rcu(ndev, netdev_upper_walk, &priv); + rcu_read_unlock(); + + handle_netdev(ib_dev, port, ndev); + list_for_each_entry_safe(upper_iter, upper_temp, &upper_list, + list) { + handle_netdev(ib_dev, port, upper_iter->upper); + dev_put(upper_iter->upper); + list_del(&upper_iter->list); + kfree(upper_iter); + } +} + +static void _roce_del_all_netdev_gids(struct ib_device *ib_dev, u8 port, + struct net_device *event_ndev) +{ + ib_cache_gid_del_all_netdev_gids(ib_dev, port, event_ndev); +} + +static void del_netdev_upper_ips(struct ib_device *ib_dev, u8 port, + struct net_device *rdma_ndev, void *cookie) +{ + handle_netdev_upper(ib_dev, port, cookie, _roce_del_all_netdev_gids); +} + +static void add_netdev_upper_ips(struct ib_device *ib_dev, u8 port, + struct net_device *rdma_ndev, void *cookie) +{ + handle_netdev_upper(ib_dev, port, cookie, _add_netdev_ips); +} + +static void del_netdev_default_ips_join(struct ib_device *ib_dev, u8 port, + struct net_device *rdma_ndev, + void *cookie) +{ + struct net_device *master_ndev; + + rcu_read_lock(); + master_ndev = netdev_master_upper_dev_get_rcu(rdma_ndev); + if (master_ndev) + dev_hold(master_ndev); + rcu_read_unlock(); + + if (master_ndev) { + bond_delete_netdev_default_gids(ib_dev, port, rdma_ndev, + master_ndev); + dev_put(master_ndev); + } +} + +/* The following functions operate on all IB devices. netdevice_event and + * addr_event execute ib_enum_all_roce_netdevs through a work. + * ib_enum_all_roce_netdevs iterates through all IB devices. + */ + +static void netdevice_event_work_handler(struct work_struct *_work) +{ + struct netdev_event_work *work = + container_of(_work, struct netdev_event_work, work); + unsigned int i; + + for (i = 0; i < ARRAY_SIZE(work->cmds) && work->cmds[i].cb; i++) { + ib_enum_all_roce_netdevs(work->cmds[i].filter, + work->cmds[i].filter_ndev, + work->cmds[i].cb, + work->cmds[i].ndev); + dev_put(work->cmds[i].ndev); + dev_put(work->cmds[i].filter_ndev); + } + + kfree(work); +} + +static int netdevice_queue_work(struct netdev_event_work_cmd *cmds, + struct net_device *ndev) +{ + unsigned int i; + struct netdev_event_work *ndev_work = + kmalloc(sizeof(*ndev_work), GFP_KERNEL); + + if (!ndev_work) + return NOTIFY_DONE; + + memcpy(ndev_work->cmds, cmds, sizeof(ndev_work->cmds)); + for (i = 0; i < ARRAY_SIZE(ndev_work->cmds) && ndev_work->cmds[i].cb; i++) { + if (!ndev_work->cmds[i].ndev) + ndev_work->cmds[i].ndev = ndev; + if (!ndev_work->cmds[i].filter_ndev) + ndev_work->cmds[i].filter_ndev = ndev; + dev_hold(ndev_work->cmds[i].ndev); + dev_hold(ndev_work->cmds[i].filter_ndev); + } + INIT_WORK(&ndev_work->work, netdevice_event_work_handler); + + queue_work(gid_cache_wq, &ndev_work->work); + + return NOTIFY_DONE; +} + +static const struct netdev_event_work_cmd add_cmd = { + .cb = add_netdev_ips, + .filter = is_eth_port_of_netdev_filter +}; + +static const struct netdev_event_work_cmd add_cmd_upper_ips = { + .cb = add_netdev_upper_ips, + .filter = is_eth_port_of_netdev_filter +}; + +static void +ndev_event_unlink(struct netdev_notifier_changeupper_info *changeupper_info, + struct netdev_event_work_cmd *cmds) +{ + static const struct netdev_event_work_cmd + upper_ips_del_cmd = { + .cb = del_netdev_upper_ips, + .filter = upper_device_filter + }; + + cmds[0] = upper_ips_del_cmd; + cmds[0].ndev = changeupper_info->upper_dev; + cmds[1] = add_cmd; +} + +static const struct netdev_event_work_cmd bonding_default_add_cmd = { + .cb = add_default_gids, + .filter = is_upper_ndev_bond_master_filter +}; + +static void +ndev_event_link(struct net_device *event_ndev, + struct netdev_notifier_changeupper_info *changeupper_info, + struct netdev_event_work_cmd *cmds) +{ + static const struct netdev_event_work_cmd + bonding_default_del_cmd = { + .cb = del_default_gids, + .filter = is_upper_ndev_bond_master_filter + }; + /* + * When a lower netdev is linked to its upper bonding + * netdev, delete lower slave netdev's default GIDs. + */ + cmds[0] = bonding_default_del_cmd; + cmds[0].ndev = event_ndev; + cmds[0].filter_ndev = changeupper_info->upper_dev; + + /* Now add bonding upper device default GIDs */ + cmds[1] = bonding_default_add_cmd; + cmds[1].ndev = changeupper_info->upper_dev; + cmds[1].filter_ndev = changeupper_info->upper_dev; + + /* Now add bonding upper device IP based GIDs */ + cmds[2] = add_cmd_upper_ips; + cmds[2].ndev = changeupper_info->upper_dev; + cmds[2].filter_ndev = changeupper_info->upper_dev; +} + +static void netdevice_event_changeupper(struct net_device *event_ndev, + struct netdev_notifier_changeupper_info *changeupper_info, + struct netdev_event_work_cmd *cmds) +{ + if (changeupper_info->linking) + ndev_event_link(event_ndev, changeupper_info, cmds); + else + ndev_event_unlink(changeupper_info, cmds); +} + +static const struct netdev_event_work_cmd add_default_gid_cmd = { + .cb = add_default_gids, + .filter = is_ndev_for_default_gid_filter, +}; + +static int netdevice_event(struct notifier_block *this, unsigned long event, + void *ptr) +{ + static const struct netdev_event_work_cmd del_cmd = { + .cb = del_netdev_ips, .filter = pass_all_filter}; + static const struct netdev_event_work_cmd + bonding_default_del_cmd_join = { + .cb = del_netdev_default_ips_join, + .filter = is_eth_port_inactive_slave_filter + }; + static const struct netdev_event_work_cmd + netdev_del_cmd = { + .cb = del_netdev_ips, + .filter = is_eth_port_of_netdev_filter + }; + static const struct netdev_event_work_cmd bonding_event_ips_del_cmd = { + .cb = del_netdev_upper_ips, .filter = upper_device_filter}; + struct net_device *ndev = netdev_notifier_info_to_dev(ptr); + struct netdev_event_work_cmd cmds[ROCE_NETDEV_CALLBACK_SZ] = { {NULL} }; + + if (ndev->type != ARPHRD_ETHER) + return NOTIFY_DONE; + + switch (event) { + case NETDEV_REGISTER: + case NETDEV_UP: + cmds[0] = bonding_default_del_cmd_join; + cmds[1] = add_default_gid_cmd; + cmds[2] = add_cmd; + break; + + case NETDEV_UNREGISTER: + if (ndev->reg_state < NETREG_UNREGISTERED) + cmds[0] = del_cmd; + else + return NOTIFY_DONE; + break; + + case NETDEV_CHANGEADDR: + cmds[0] = netdev_del_cmd; + if (ndev->reg_state == NETREG_REGISTERED) { + cmds[1] = add_default_gid_cmd; + cmds[2] = add_cmd; + } + break; + + case NETDEV_CHANGEUPPER: + netdevice_event_changeupper(ndev, + container_of(ptr, struct netdev_notifier_changeupper_info, info), + cmds); + break; + + case NETDEV_BONDING_FAILOVER: + cmds[0] = bonding_event_ips_del_cmd; + /* Add default GIDs of the bond device */ + cmds[1] = bonding_default_add_cmd; + /* Add IP based GIDs of the bond device */ + cmds[2] = add_cmd_upper_ips; + break; + + default: + return NOTIFY_DONE; + } + + return netdevice_queue_work(cmds, ndev); +} + +static void update_gid_event_work_handler(struct work_struct *_work) +{ + struct update_gid_event_work *work = + container_of(_work, struct update_gid_event_work, work); + + ib_enum_all_roce_netdevs(is_eth_port_of_netdev_filter, + work->gid_attr.ndev, + callback_for_addr_gid_device_scan, work); + + dev_put(work->gid_attr.ndev); + kfree(work); +} + +static int addr_event(struct notifier_block *this, unsigned long event, + struct sockaddr *sa, struct net_device *ndev) +{ + struct update_gid_event_work *work; + enum gid_op_type gid_op; + + if (ndev->type != ARPHRD_ETHER) + return NOTIFY_DONE; + + switch (event) { + case NETDEV_UP: + gid_op = GID_ADD; + break; + + case NETDEV_DOWN: + gid_op = GID_DEL; + break; + + default: + return NOTIFY_DONE; + } + + work = kmalloc(sizeof(*work), GFP_ATOMIC); + if (!work) + return NOTIFY_DONE; + + INIT_WORK(&work->work, update_gid_event_work_handler); + + rdma_ip2gid(sa, &work->gid); + work->gid_op = gid_op; + + memset(&work->gid_attr, 0, sizeof(work->gid_attr)); + dev_hold(ndev); + work->gid_attr.ndev = ndev; + + queue_work(gid_cache_wq, &work->work); + + return NOTIFY_DONE; +} + +static int inetaddr_event(struct notifier_block *this, unsigned long event, + void *ptr) +{ + struct sockaddr_in in; + struct net_device *ndev; + struct in_ifaddr *ifa = ptr; + + in.sin_family = AF_INET; + in.sin_addr.s_addr = ifa->ifa_address; + ndev = ifa->ifa_dev->dev; + + return addr_event(this, event, (struct sockaddr *)&in, ndev); +} + +static int inet6addr_event(struct notifier_block *this, unsigned long event, + void *ptr) +{ + struct sockaddr_in6 in6; + struct net_device *ndev; + struct inet6_ifaddr *ifa6 = ptr; + + in6.sin6_family = AF_INET6; + in6.sin6_addr = ifa6->addr; + ndev = ifa6->idev->dev; + + return addr_event(this, event, (struct sockaddr *)&in6, ndev); +} + +static struct notifier_block nb_netdevice = { + .notifier_call = netdevice_event +}; + +static struct notifier_block nb_inetaddr = { + .notifier_call = inetaddr_event +}; + +static struct notifier_block nb_inet6addr = { + .notifier_call = inet6addr_event +}; + +int __init roce_gid_mgmt_init(void) +{ + gid_cache_wq = alloc_ordered_workqueue("gid-cache-wq", 0); + if (!gid_cache_wq) + return -ENOMEM; + + register_inetaddr_notifier(&nb_inetaddr); + if (IS_ENABLED(CONFIG_IPV6)) + register_inet6addr_notifier(&nb_inet6addr); + /* We relay on the netdevice notifier to enumerate all + * existing devices in the system. Register to this notifier + * last to make sure we will not miss any IP add/del + * callbacks. + */ + register_netdevice_notifier(&nb_netdevice); + + return 0; +} + +void __exit roce_gid_mgmt_cleanup(void) +{ + if (IS_ENABLED(CONFIG_IPV6)) + unregister_inet6addr_notifier(&nb_inet6addr); + unregister_inetaddr_notifier(&nb_inetaddr); + unregister_netdevice_notifier(&nb_netdevice); + /* Ensure all gid deletion tasks complete before we go down, + * to avoid any reference to free'd memory. By the time + * ib-core is removed, all physical devices have been removed, + * so no issue with remaining hardware contexts. + */ + destroy_workqueue(gid_cache_wq); +} diff --git a/drivers/infiniband/core/rw.c b/drivers/infiniband/core/rw.c new file mode 100644 index 000000000..a96030b78 --- /dev/null +++ b/drivers/infiniband/core/rw.c @@ -0,0 +1,743 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * Copyright (c) 2016 HGST, a Western Digital Company. + */ +#include <linux/moduleparam.h> +#include <linux/slab.h> +#include <linux/pci-p2pdma.h> +#include <rdma/mr_pool.h> +#include <rdma/rw.h> + +enum { + RDMA_RW_SINGLE_WR, + RDMA_RW_MULTI_WR, + RDMA_RW_MR, + RDMA_RW_SIG_MR, +}; + +static bool rdma_rw_force_mr; +module_param_named(force_mr, rdma_rw_force_mr, bool, 0); +MODULE_PARM_DESC(force_mr, "Force usage of MRs for RDMA READ/WRITE operations"); + +/* + * Report whether memory registration should be used. Memory registration must + * be used for iWarp devices because of iWARP-specific limitations. Memory + * registration is also enabled if registering memory might yield better + * performance than using multiple SGE entries, see rdma_rw_io_needs_mr() + */ +static inline bool rdma_rw_can_use_mr(struct ib_device *dev, u8 port_num) +{ + if (rdma_protocol_iwarp(dev, port_num)) + return true; + if (dev->attrs.max_sgl_rd) + return true; + if (unlikely(rdma_rw_force_mr)) + return true; + return false; +} + +/* + * Check if the device will use memory registration for this RW operation. + * For RDMA READs we must use MRs on iWarp and can optionally use them as an + * optimization otherwise. Additionally we have a debug option to force usage + * of MRs to help testing this code path. + */ +static inline bool rdma_rw_io_needs_mr(struct ib_device *dev, u8 port_num, + enum dma_data_direction dir, int dma_nents) +{ + if (dir == DMA_FROM_DEVICE) { + if (rdma_protocol_iwarp(dev, port_num)) + return true; + if (dev->attrs.max_sgl_rd && dma_nents > dev->attrs.max_sgl_rd) + return true; + } + if (unlikely(rdma_rw_force_mr)) + return true; + return false; +} + +static inline u32 rdma_rw_fr_page_list_len(struct ib_device *dev, + bool pi_support) +{ + u32 max_pages; + + if (pi_support) + max_pages = dev->attrs.max_pi_fast_reg_page_list_len; + else + max_pages = dev->attrs.max_fast_reg_page_list_len; + + /* arbitrary limit to avoid allocating gigantic resources */ + return min_t(u32, max_pages, 256); +} + +static inline int rdma_rw_inv_key(struct rdma_rw_reg_ctx *reg) +{ + int count = 0; + + if (reg->mr->need_inval) { + reg->inv_wr.opcode = IB_WR_LOCAL_INV; + reg->inv_wr.ex.invalidate_rkey = reg->mr->lkey; + reg->inv_wr.next = ®->reg_wr.wr; + count++; + } else { + reg->inv_wr.next = NULL; + } + + return count; +} + +/* Caller must have zero-initialized *reg. */ +static int rdma_rw_init_one_mr(struct ib_qp *qp, u8 port_num, + struct rdma_rw_reg_ctx *reg, struct scatterlist *sg, + u32 sg_cnt, u32 offset) +{ + u32 pages_per_mr = rdma_rw_fr_page_list_len(qp->pd->device, + qp->integrity_en); + u32 nents = min(sg_cnt, pages_per_mr); + int count = 0, ret; + + reg->mr = ib_mr_pool_get(qp, &qp->rdma_mrs); + if (!reg->mr) + return -EAGAIN; + + count += rdma_rw_inv_key(reg); + + ret = ib_map_mr_sg(reg->mr, sg, nents, &offset, PAGE_SIZE); + if (ret < 0 || ret < nents) { + ib_mr_pool_put(qp, &qp->rdma_mrs, reg->mr); + return -EINVAL; + } + + reg->reg_wr.wr.opcode = IB_WR_REG_MR; + reg->reg_wr.mr = reg->mr; + reg->reg_wr.access = IB_ACCESS_LOCAL_WRITE; + if (rdma_protocol_iwarp(qp->device, port_num)) + reg->reg_wr.access |= IB_ACCESS_REMOTE_WRITE; + count++; + + reg->sge.addr = reg->mr->iova; + reg->sge.length = reg->mr->length; + return count; +} + +static int rdma_rw_init_mr_wrs(struct rdma_rw_ctx *ctx, struct ib_qp *qp, + u8 port_num, struct scatterlist *sg, u32 sg_cnt, u32 offset, + u64 remote_addr, u32 rkey, enum dma_data_direction dir) +{ + struct rdma_rw_reg_ctx *prev = NULL; + u32 pages_per_mr = rdma_rw_fr_page_list_len(qp->pd->device, + qp->integrity_en); + int i, j, ret = 0, count = 0; + + ctx->nr_ops = DIV_ROUND_UP(sg_cnt, pages_per_mr); + ctx->reg = kcalloc(ctx->nr_ops, sizeof(*ctx->reg), GFP_KERNEL); + if (!ctx->reg) { + ret = -ENOMEM; + goto out; + } + + for (i = 0; i < ctx->nr_ops; i++) { + struct rdma_rw_reg_ctx *reg = &ctx->reg[i]; + u32 nents = min(sg_cnt, pages_per_mr); + + ret = rdma_rw_init_one_mr(qp, port_num, reg, sg, sg_cnt, + offset); + if (ret < 0) + goto out_free; + count += ret; + + if (prev) { + if (reg->mr->need_inval) + prev->wr.wr.next = ®->inv_wr; + else + prev->wr.wr.next = ®->reg_wr.wr; + } + + reg->reg_wr.wr.next = ®->wr.wr; + + reg->wr.wr.sg_list = ®->sge; + reg->wr.wr.num_sge = 1; + reg->wr.remote_addr = remote_addr; + reg->wr.rkey = rkey; + if (dir == DMA_TO_DEVICE) { + reg->wr.wr.opcode = IB_WR_RDMA_WRITE; + } else if (!rdma_cap_read_inv(qp->device, port_num)) { + reg->wr.wr.opcode = IB_WR_RDMA_READ; + } else { + reg->wr.wr.opcode = IB_WR_RDMA_READ_WITH_INV; + reg->wr.wr.ex.invalidate_rkey = reg->mr->lkey; + } + count++; + + remote_addr += reg->sge.length; + sg_cnt -= nents; + for (j = 0; j < nents; j++) + sg = sg_next(sg); + prev = reg; + offset = 0; + } + + if (prev) + prev->wr.wr.next = NULL; + + ctx->type = RDMA_RW_MR; + return count; + +out_free: + while (--i >= 0) + ib_mr_pool_put(qp, &qp->rdma_mrs, ctx->reg[i].mr); + kfree(ctx->reg); +out: + return ret; +} + +static int rdma_rw_init_map_wrs(struct rdma_rw_ctx *ctx, struct ib_qp *qp, + struct scatterlist *sg, u32 sg_cnt, u32 offset, + u64 remote_addr, u32 rkey, enum dma_data_direction dir) +{ + u32 max_sge = dir == DMA_TO_DEVICE ? qp->max_write_sge : + qp->max_read_sge; + struct ib_sge *sge; + u32 total_len = 0, i, j; + + ctx->nr_ops = DIV_ROUND_UP(sg_cnt, max_sge); + + ctx->map.sges = sge = kcalloc(sg_cnt, sizeof(*sge), GFP_KERNEL); + if (!ctx->map.sges) + goto out; + + ctx->map.wrs = kcalloc(ctx->nr_ops, sizeof(*ctx->map.wrs), GFP_KERNEL); + if (!ctx->map.wrs) + goto out_free_sges; + + for (i = 0; i < ctx->nr_ops; i++) { + struct ib_rdma_wr *rdma_wr = &ctx->map.wrs[i]; + u32 nr_sge = min(sg_cnt, max_sge); + + if (dir == DMA_TO_DEVICE) + rdma_wr->wr.opcode = IB_WR_RDMA_WRITE; + else + rdma_wr->wr.opcode = IB_WR_RDMA_READ; + rdma_wr->remote_addr = remote_addr + total_len; + rdma_wr->rkey = rkey; + rdma_wr->wr.num_sge = nr_sge; + rdma_wr->wr.sg_list = sge; + + for (j = 0; j < nr_sge; j++, sg = sg_next(sg)) { + sge->addr = sg_dma_address(sg) + offset; + sge->length = sg_dma_len(sg) - offset; + sge->lkey = qp->pd->local_dma_lkey; + + total_len += sge->length; + sge++; + sg_cnt--; + offset = 0; + } + + rdma_wr->wr.next = i + 1 < ctx->nr_ops ? + &ctx->map.wrs[i + 1].wr : NULL; + } + + ctx->type = RDMA_RW_MULTI_WR; + return ctx->nr_ops; + +out_free_sges: + kfree(ctx->map.sges); +out: + return -ENOMEM; +} + +static int rdma_rw_init_single_wr(struct rdma_rw_ctx *ctx, struct ib_qp *qp, + struct scatterlist *sg, u32 offset, u64 remote_addr, u32 rkey, + enum dma_data_direction dir) +{ + struct ib_rdma_wr *rdma_wr = &ctx->single.wr; + + ctx->nr_ops = 1; + + ctx->single.sge.lkey = qp->pd->local_dma_lkey; + ctx->single.sge.addr = sg_dma_address(sg) + offset; + ctx->single.sge.length = sg_dma_len(sg) - offset; + + memset(rdma_wr, 0, sizeof(*rdma_wr)); + if (dir == DMA_TO_DEVICE) + rdma_wr->wr.opcode = IB_WR_RDMA_WRITE; + else + rdma_wr->wr.opcode = IB_WR_RDMA_READ; + rdma_wr->wr.sg_list = &ctx->single.sge; + rdma_wr->wr.num_sge = 1; + rdma_wr->remote_addr = remote_addr; + rdma_wr->rkey = rkey; + + ctx->type = RDMA_RW_SINGLE_WR; + return 1; +} + +static void rdma_rw_unmap_sg(struct ib_device *dev, struct scatterlist *sg, + u32 sg_cnt, enum dma_data_direction dir) +{ + if (is_pci_p2pdma_page(sg_page(sg))) + pci_p2pdma_unmap_sg(dev->dma_device, sg, sg_cnt, dir); + else + ib_dma_unmap_sg(dev, sg, sg_cnt, dir); +} + +static int rdma_rw_map_sg(struct ib_device *dev, struct scatterlist *sg, + u32 sg_cnt, enum dma_data_direction dir) +{ + if (is_pci_p2pdma_page(sg_page(sg))) { + if (WARN_ON_ONCE(ib_uses_virt_dma(dev))) + return 0; + return pci_p2pdma_map_sg(dev->dma_device, sg, sg_cnt, dir); + } + return ib_dma_map_sg(dev, sg, sg_cnt, dir); +} + +/** + * rdma_rw_ctx_init - initialize a RDMA READ/WRITE context + * @ctx: context to initialize + * @qp: queue pair to operate on + * @port_num: port num to which the connection is bound + * @sg: scatterlist to READ/WRITE from/to + * @sg_cnt: number of entries in @sg + * @sg_offset: current byte offset into @sg + * @remote_addr:remote address to read/write (relative to @rkey) + * @rkey: remote key to operate on + * @dir: %DMA_TO_DEVICE for RDMA WRITE, %DMA_FROM_DEVICE for RDMA READ + * + * Returns the number of WQEs that will be needed on the workqueue if + * successful, or a negative error code. + */ +int rdma_rw_ctx_init(struct rdma_rw_ctx *ctx, struct ib_qp *qp, u8 port_num, + struct scatterlist *sg, u32 sg_cnt, u32 sg_offset, + u64 remote_addr, u32 rkey, enum dma_data_direction dir) +{ + struct ib_device *dev = qp->pd->device; + int ret; + + ret = rdma_rw_map_sg(dev, sg, sg_cnt, dir); + if (!ret) + return -ENOMEM; + sg_cnt = ret; + + /* + * Skip to the S/G entry that sg_offset falls into: + */ + for (;;) { + u32 len = sg_dma_len(sg); + + if (sg_offset < len) + break; + + sg = sg_next(sg); + sg_offset -= len; + sg_cnt--; + } + + ret = -EIO; + if (WARN_ON_ONCE(sg_cnt == 0)) + goto out_unmap_sg; + + if (rdma_rw_io_needs_mr(qp->device, port_num, dir, sg_cnt)) { + ret = rdma_rw_init_mr_wrs(ctx, qp, port_num, sg, sg_cnt, + sg_offset, remote_addr, rkey, dir); + } else if (sg_cnt > 1) { + ret = rdma_rw_init_map_wrs(ctx, qp, sg, sg_cnt, sg_offset, + remote_addr, rkey, dir); + } else { + ret = rdma_rw_init_single_wr(ctx, qp, sg, sg_offset, + remote_addr, rkey, dir); + } + + if (ret < 0) + goto out_unmap_sg; + return ret; + +out_unmap_sg: + rdma_rw_unmap_sg(dev, sg, sg_cnt, dir); + return ret; +} +EXPORT_SYMBOL(rdma_rw_ctx_init); + +/** + * rdma_rw_ctx_signature_init - initialize a RW context with signature offload + * @ctx: context to initialize + * @qp: queue pair to operate on + * @port_num: port num to which the connection is bound + * @sg: scatterlist to READ/WRITE from/to + * @sg_cnt: number of entries in @sg + * @prot_sg: scatterlist to READ/WRITE protection information from/to + * @prot_sg_cnt: number of entries in @prot_sg + * @sig_attrs: signature offloading algorithms + * @remote_addr:remote address to read/write (relative to @rkey) + * @rkey: remote key to operate on + * @dir: %DMA_TO_DEVICE for RDMA WRITE, %DMA_FROM_DEVICE for RDMA READ + * + * Returns the number of WQEs that will be needed on the workqueue if + * successful, or a negative error code. + */ +int rdma_rw_ctx_signature_init(struct rdma_rw_ctx *ctx, struct ib_qp *qp, + u8 port_num, struct scatterlist *sg, u32 sg_cnt, + struct scatterlist *prot_sg, u32 prot_sg_cnt, + struct ib_sig_attrs *sig_attrs, + u64 remote_addr, u32 rkey, enum dma_data_direction dir) +{ + struct ib_device *dev = qp->pd->device; + u32 pages_per_mr = rdma_rw_fr_page_list_len(qp->pd->device, + qp->integrity_en); + struct ib_rdma_wr *rdma_wr; + int count = 0, ret; + + if (sg_cnt > pages_per_mr || prot_sg_cnt > pages_per_mr) { + pr_err("SG count too large: sg_cnt=%d, prot_sg_cnt=%d, pages_per_mr=%d\n", + sg_cnt, prot_sg_cnt, pages_per_mr); + return -EINVAL; + } + + ret = rdma_rw_map_sg(dev, sg, sg_cnt, dir); + if (!ret) + return -ENOMEM; + sg_cnt = ret; + + if (prot_sg_cnt) { + ret = rdma_rw_map_sg(dev, prot_sg, prot_sg_cnt, dir); + if (!ret) { + ret = -ENOMEM; + goto out_unmap_sg; + } + prot_sg_cnt = ret; + } + + ctx->type = RDMA_RW_SIG_MR; + ctx->nr_ops = 1; + ctx->reg = kcalloc(1, sizeof(*ctx->reg), GFP_KERNEL); + if (!ctx->reg) { + ret = -ENOMEM; + goto out_unmap_prot_sg; + } + + ctx->reg->mr = ib_mr_pool_get(qp, &qp->sig_mrs); + if (!ctx->reg->mr) { + ret = -EAGAIN; + goto out_free_ctx; + } + + count += rdma_rw_inv_key(ctx->reg); + + memcpy(ctx->reg->mr->sig_attrs, sig_attrs, sizeof(struct ib_sig_attrs)); + + ret = ib_map_mr_sg_pi(ctx->reg->mr, sg, sg_cnt, NULL, prot_sg, + prot_sg_cnt, NULL, SZ_4K); + if (unlikely(ret)) { + pr_err("failed to map PI sg (%d)\n", sg_cnt + prot_sg_cnt); + goto out_destroy_sig_mr; + } + + ctx->reg->reg_wr.wr.opcode = IB_WR_REG_MR_INTEGRITY; + ctx->reg->reg_wr.wr.wr_cqe = NULL; + ctx->reg->reg_wr.wr.num_sge = 0; + ctx->reg->reg_wr.wr.send_flags = 0; + ctx->reg->reg_wr.access = IB_ACCESS_LOCAL_WRITE; + if (rdma_protocol_iwarp(qp->device, port_num)) + ctx->reg->reg_wr.access |= IB_ACCESS_REMOTE_WRITE; + ctx->reg->reg_wr.mr = ctx->reg->mr; + ctx->reg->reg_wr.key = ctx->reg->mr->lkey; + count++; + + ctx->reg->sge.addr = ctx->reg->mr->iova; + ctx->reg->sge.length = ctx->reg->mr->length; + if (sig_attrs->wire.sig_type == IB_SIG_TYPE_NONE) + ctx->reg->sge.length -= ctx->reg->mr->sig_attrs->meta_length; + + rdma_wr = &ctx->reg->wr; + rdma_wr->wr.sg_list = &ctx->reg->sge; + rdma_wr->wr.num_sge = 1; + rdma_wr->remote_addr = remote_addr; + rdma_wr->rkey = rkey; + if (dir == DMA_TO_DEVICE) + rdma_wr->wr.opcode = IB_WR_RDMA_WRITE; + else + rdma_wr->wr.opcode = IB_WR_RDMA_READ; + ctx->reg->reg_wr.wr.next = &rdma_wr->wr; + count++; + + return count; + +out_destroy_sig_mr: + ib_mr_pool_put(qp, &qp->sig_mrs, ctx->reg->mr); +out_free_ctx: + kfree(ctx->reg); +out_unmap_prot_sg: + if (prot_sg_cnt) + rdma_rw_unmap_sg(dev, prot_sg, prot_sg_cnt, dir); +out_unmap_sg: + rdma_rw_unmap_sg(dev, sg, sg_cnt, dir); + return ret; +} +EXPORT_SYMBOL(rdma_rw_ctx_signature_init); + +/* + * Now that we are going to post the WRs we can update the lkey and need_inval + * state on the MRs. If we were doing this at init time, we would get double + * or missing invalidations if a context was initialized but not actually + * posted. + */ +static void rdma_rw_update_lkey(struct rdma_rw_reg_ctx *reg, bool need_inval) +{ + reg->mr->need_inval = need_inval; + ib_update_fast_reg_key(reg->mr, ib_inc_rkey(reg->mr->lkey)); + reg->reg_wr.key = reg->mr->lkey; + reg->sge.lkey = reg->mr->lkey; +} + +/** + * rdma_rw_ctx_wrs - return chain of WRs for a RDMA READ or WRITE operation + * @ctx: context to operate on + * @qp: queue pair to operate on + * @port_num: port num to which the connection is bound + * @cqe: completion queue entry for the last WR + * @chain_wr: WR to append to the posted chain + * + * Return the WR chain for the set of RDMA READ/WRITE operations described by + * @ctx, as well as any memory registration operations needed. If @chain_wr + * is non-NULL the WR it points to will be appended to the chain of WRs posted. + * If @chain_wr is not set @cqe must be set so that the caller gets a + * completion notification. + */ +struct ib_send_wr *rdma_rw_ctx_wrs(struct rdma_rw_ctx *ctx, struct ib_qp *qp, + u8 port_num, struct ib_cqe *cqe, struct ib_send_wr *chain_wr) +{ + struct ib_send_wr *first_wr, *last_wr; + int i; + + switch (ctx->type) { + case RDMA_RW_SIG_MR: + case RDMA_RW_MR: + for (i = 0; i < ctx->nr_ops; i++) { + rdma_rw_update_lkey(&ctx->reg[i], + ctx->reg[i].wr.wr.opcode != + IB_WR_RDMA_READ_WITH_INV); + } + + if (ctx->reg[0].inv_wr.next) + first_wr = &ctx->reg[0].inv_wr; + else + first_wr = &ctx->reg[0].reg_wr.wr; + last_wr = &ctx->reg[ctx->nr_ops - 1].wr.wr; + break; + case RDMA_RW_MULTI_WR: + first_wr = &ctx->map.wrs[0].wr; + last_wr = &ctx->map.wrs[ctx->nr_ops - 1].wr; + break; + case RDMA_RW_SINGLE_WR: + first_wr = &ctx->single.wr.wr; + last_wr = &ctx->single.wr.wr; + break; + default: + BUG(); + } + + if (chain_wr) { + last_wr->next = chain_wr; + } else { + last_wr->wr_cqe = cqe; + last_wr->send_flags |= IB_SEND_SIGNALED; + } + + return first_wr; +} +EXPORT_SYMBOL(rdma_rw_ctx_wrs); + +/** + * rdma_rw_ctx_post - post a RDMA READ or RDMA WRITE operation + * @ctx: context to operate on + * @qp: queue pair to operate on + * @port_num: port num to which the connection is bound + * @cqe: completion queue entry for the last WR + * @chain_wr: WR to append to the posted chain + * + * Post the set of RDMA READ/WRITE operations described by @ctx, as well as + * any memory registration operations needed. If @chain_wr is non-NULL the + * WR it points to will be appended to the chain of WRs posted. If @chain_wr + * is not set @cqe must be set so that the caller gets a completion + * notification. + */ +int rdma_rw_ctx_post(struct rdma_rw_ctx *ctx, struct ib_qp *qp, u8 port_num, + struct ib_cqe *cqe, struct ib_send_wr *chain_wr) +{ + struct ib_send_wr *first_wr; + + first_wr = rdma_rw_ctx_wrs(ctx, qp, port_num, cqe, chain_wr); + return ib_post_send(qp, first_wr, NULL); +} +EXPORT_SYMBOL(rdma_rw_ctx_post); + +/** + * rdma_rw_ctx_destroy - release all resources allocated by rdma_rw_ctx_init + * @ctx: context to release + * @qp: queue pair to operate on + * @port_num: port num to which the connection is bound + * @sg: scatterlist that was used for the READ/WRITE + * @sg_cnt: number of entries in @sg + * @dir: %DMA_TO_DEVICE for RDMA WRITE, %DMA_FROM_DEVICE for RDMA READ + */ +void rdma_rw_ctx_destroy(struct rdma_rw_ctx *ctx, struct ib_qp *qp, u8 port_num, + struct scatterlist *sg, u32 sg_cnt, enum dma_data_direction dir) +{ + int i; + + switch (ctx->type) { + case RDMA_RW_MR: + for (i = 0; i < ctx->nr_ops; i++) + ib_mr_pool_put(qp, &qp->rdma_mrs, ctx->reg[i].mr); + kfree(ctx->reg); + break; + case RDMA_RW_MULTI_WR: + kfree(ctx->map.wrs); + kfree(ctx->map.sges); + break; + case RDMA_RW_SINGLE_WR: + break; + default: + BUG(); + break; + } + + rdma_rw_unmap_sg(qp->pd->device, sg, sg_cnt, dir); +} +EXPORT_SYMBOL(rdma_rw_ctx_destroy); + +/** + * rdma_rw_ctx_destroy_signature - release all resources allocated by + * rdma_rw_ctx_signature_init + * @ctx: context to release + * @qp: queue pair to operate on + * @port_num: port num to which the connection is bound + * @sg: scatterlist that was used for the READ/WRITE + * @sg_cnt: number of entries in @sg + * @prot_sg: scatterlist that was used for the READ/WRITE of the PI + * @prot_sg_cnt: number of entries in @prot_sg + * @dir: %DMA_TO_DEVICE for RDMA WRITE, %DMA_FROM_DEVICE for RDMA READ + */ +void rdma_rw_ctx_destroy_signature(struct rdma_rw_ctx *ctx, struct ib_qp *qp, + u8 port_num, struct scatterlist *sg, u32 sg_cnt, + struct scatterlist *prot_sg, u32 prot_sg_cnt, + enum dma_data_direction dir) +{ + if (WARN_ON_ONCE(ctx->type != RDMA_RW_SIG_MR)) + return; + + ib_mr_pool_put(qp, &qp->sig_mrs, ctx->reg->mr); + kfree(ctx->reg); + + if (prot_sg_cnt) + rdma_rw_unmap_sg(qp->pd->device, prot_sg, prot_sg_cnt, dir); + rdma_rw_unmap_sg(qp->pd->device, sg, sg_cnt, dir); +} +EXPORT_SYMBOL(rdma_rw_ctx_destroy_signature); + +/** + * rdma_rw_mr_factor - return number of MRs required for a payload + * @device: device handling the connection + * @port_num: port num to which the connection is bound + * @maxpages: maximum payload pages per rdma_rw_ctx + * + * Returns the number of MRs the device requires to move @maxpayload + * bytes. The returned value is used during transport creation to + * compute max_rdma_ctxts and the size of the transport's Send and + * Send Completion Queues. + */ +unsigned int rdma_rw_mr_factor(struct ib_device *device, u8 port_num, + unsigned int maxpages) +{ + unsigned int mr_pages; + + if (rdma_rw_can_use_mr(device, port_num)) + mr_pages = rdma_rw_fr_page_list_len(device, false); + else + mr_pages = device->attrs.max_sge_rd; + return DIV_ROUND_UP(maxpages, mr_pages); +} +EXPORT_SYMBOL(rdma_rw_mr_factor); + +void rdma_rw_init_qp(struct ib_device *dev, struct ib_qp_init_attr *attr) +{ + u32 factor; + + WARN_ON_ONCE(attr->port_num == 0); + + /* + * Each context needs at least one RDMA READ or WRITE WR. + * + * For some hardware we might need more, eventually we should ask the + * HCA driver for a multiplier here. + */ + factor = 1; + + /* + * If the devices needs MRs to perform RDMA READ or WRITE operations, + * we'll need two additional MRs for the registrations and the + * invalidation. + */ + if (attr->create_flags & IB_QP_CREATE_INTEGRITY_EN || + rdma_rw_can_use_mr(dev, attr->port_num)) + factor += 2; /* inv + reg */ + + attr->cap.max_send_wr += factor * attr->cap.max_rdma_ctxs; + + /* + * But maybe we were just too high in the sky and the device doesn't + * even support all we need, and we'll have to live with what we get.. + */ + attr->cap.max_send_wr = + min_t(u32, attr->cap.max_send_wr, dev->attrs.max_qp_wr); +} + +int rdma_rw_init_mrs(struct ib_qp *qp, struct ib_qp_init_attr *attr) +{ + struct ib_device *dev = qp->pd->device; + u32 nr_mrs = 0, nr_sig_mrs = 0, max_num_sg = 0; + int ret = 0; + + if (attr->create_flags & IB_QP_CREATE_INTEGRITY_EN) { + nr_sig_mrs = attr->cap.max_rdma_ctxs; + nr_mrs = attr->cap.max_rdma_ctxs; + max_num_sg = rdma_rw_fr_page_list_len(dev, true); + } else if (rdma_rw_can_use_mr(dev, attr->port_num)) { + nr_mrs = attr->cap.max_rdma_ctxs; + max_num_sg = rdma_rw_fr_page_list_len(dev, false); + } + + if (nr_mrs) { + ret = ib_mr_pool_init(qp, &qp->rdma_mrs, nr_mrs, + IB_MR_TYPE_MEM_REG, + max_num_sg, 0); + if (ret) { + pr_err("%s: failed to allocated %d MRs\n", + __func__, nr_mrs); + return ret; + } + } + + if (nr_sig_mrs) { + ret = ib_mr_pool_init(qp, &qp->sig_mrs, nr_sig_mrs, + IB_MR_TYPE_INTEGRITY, max_num_sg, max_num_sg); + if (ret) { + pr_err("%s: failed to allocated %d SIG MRs\n", + __func__, nr_sig_mrs); + goto out_free_rdma_mrs; + } + } + + return 0; + +out_free_rdma_mrs: + ib_mr_pool_destroy(qp, &qp->rdma_mrs); + return ret; +} + +void rdma_rw_cleanup_mrs(struct ib_qp *qp) +{ + ib_mr_pool_destroy(qp, &qp->sig_mrs); + ib_mr_pool_destroy(qp, &qp->rdma_mrs); +} diff --git a/drivers/infiniband/core/sa.h b/drivers/infiniband/core/sa.h new file mode 100644 index 000000000..cbaaaa92f --- /dev/null +++ b/drivers/infiniband/core/sa.h @@ -0,0 +1,64 @@ +/* + * Copyright (c) 2004 Topspin Communications. All rights reserved. + * Copyright (c) 2005 Voltaire, Inc. All rights reserved. + * Copyright (c) 2006 Intel Corporation. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#ifndef SA_H +#define SA_H + +#include <rdma/ib_sa.h> + +static inline void ib_sa_client_get(struct ib_sa_client *client) +{ + atomic_inc(&client->users); +} + +static inline void ib_sa_client_put(struct ib_sa_client *client) +{ + if (atomic_dec_and_test(&client->users)) + complete(&client->comp); +} + +int ib_sa_mcmember_rec_query(struct ib_sa_client *client, + struct ib_device *device, u8 port_num, u8 method, + struct ib_sa_mcmember_rec *rec, + ib_sa_comp_mask comp_mask, + unsigned long timeout_ms, gfp_t gfp_mask, + void (*callback)(int status, + struct ib_sa_mcmember_rec *resp, + void *context), + void *context, struct ib_sa_query **sa_query); + +int mcast_init(void); +void mcast_cleanup(void); + +#endif /* SA_H */ diff --git a/drivers/infiniband/core/sa_query.c b/drivers/infiniband/core/sa_query.c new file mode 100644 index 000000000..de88f472e --- /dev/null +++ b/drivers/infiniband/core/sa_query.c @@ -0,0 +1,2467 @@ +/* + * Copyright (c) 2004 Topspin Communications. All rights reserved. + * Copyright (c) 2005 Voltaire, Inc. All rights reserved. + * Copyright (c) 2006 Intel Corporation. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include <linux/module.h> +#include <linux/init.h> +#include <linux/err.h> +#include <linux/random.h> +#include <linux/spinlock.h> +#include <linux/slab.h> +#include <linux/dma-mapping.h> +#include <linux/kref.h> +#include <linux/xarray.h> +#include <linux/workqueue.h> +#include <uapi/linux/if_ether.h> +#include <rdma/ib_pack.h> +#include <rdma/ib_cache.h> +#include <rdma/rdma_netlink.h> +#include <net/netlink.h> +#include <uapi/rdma/ib_user_sa.h> +#include <rdma/ib_marshall.h> +#include <rdma/ib_addr.h> +#include <rdma/opa_addr.h> +#include "sa.h" +#include "core_priv.h" + +#define IB_SA_LOCAL_SVC_TIMEOUT_MIN 100 +#define IB_SA_LOCAL_SVC_TIMEOUT_DEFAULT 2000 +#define IB_SA_LOCAL_SVC_TIMEOUT_MAX 200000 +#define IB_SA_CPI_MAX_RETRY_CNT 3 +#define IB_SA_CPI_RETRY_WAIT 1000 /*msecs */ +static int sa_local_svc_timeout_ms = IB_SA_LOCAL_SVC_TIMEOUT_DEFAULT; + +struct ib_sa_sm_ah { + struct ib_ah *ah; + struct kref ref; + u16 pkey_index; + u8 src_path_mask; +}; + +enum rdma_class_port_info_type { + RDMA_CLASS_PORT_INFO_IB, + RDMA_CLASS_PORT_INFO_OPA +}; + +struct rdma_class_port_info { + enum rdma_class_port_info_type type; + union { + struct ib_class_port_info ib; + struct opa_class_port_info opa; + }; +}; + +struct ib_sa_classport_cache { + bool valid; + int retry_cnt; + struct rdma_class_port_info data; +}; + +struct ib_sa_port { + struct ib_mad_agent *agent; + struct ib_sa_sm_ah *sm_ah; + struct work_struct update_task; + struct ib_sa_classport_cache classport_info; + struct delayed_work ib_cpi_work; + spinlock_t classport_lock; /* protects class port info set */ + spinlock_t ah_lock; + u8 port_num; +}; + +struct ib_sa_device { + int start_port, end_port; + struct ib_event_handler event_handler; + struct ib_sa_port port[]; +}; + +struct ib_sa_query { + void (*callback)(struct ib_sa_query *, int, struct ib_sa_mad *); + void (*release)(struct ib_sa_query *); + struct ib_sa_client *client; + struct ib_sa_port *port; + struct ib_mad_send_buf *mad_buf; + struct ib_sa_sm_ah *sm_ah; + int id; + u32 flags; + struct list_head list; /* Local svc request list */ + u32 seq; /* Local svc request sequence number */ + unsigned long timeout; /* Local svc timeout */ + u8 path_use; /* How will the pathrecord be used */ +}; + +#define IB_SA_ENABLE_LOCAL_SERVICE 0x00000001 +#define IB_SA_CANCEL 0x00000002 +#define IB_SA_QUERY_OPA 0x00000004 + +struct ib_sa_service_query { + void (*callback)(int, struct ib_sa_service_rec *, void *); + void *context; + struct ib_sa_query sa_query; +}; + +struct ib_sa_path_query { + void (*callback)(int, struct sa_path_rec *, void *); + void *context; + struct ib_sa_query sa_query; + struct sa_path_rec *conv_pr; +}; + +struct ib_sa_guidinfo_query { + void (*callback)(int, struct ib_sa_guidinfo_rec *, void *); + void *context; + struct ib_sa_query sa_query; +}; + +struct ib_sa_classport_info_query { + void (*callback)(void *); + void *context; + struct ib_sa_query sa_query; +}; + +struct ib_sa_mcmember_query { + void (*callback)(int, struct ib_sa_mcmember_rec *, void *); + void *context; + struct ib_sa_query sa_query; +}; + +static LIST_HEAD(ib_nl_request_list); +static DEFINE_SPINLOCK(ib_nl_request_lock); +static atomic_t ib_nl_sa_request_seq; +static struct workqueue_struct *ib_nl_wq; +static struct delayed_work ib_nl_timed_work; +static const struct nla_policy ib_nl_policy[LS_NLA_TYPE_MAX] = { + [LS_NLA_TYPE_PATH_RECORD] = {.type = NLA_BINARY, + .len = sizeof(struct ib_path_rec_data)}, + [LS_NLA_TYPE_TIMEOUT] = {.type = NLA_U32}, + [LS_NLA_TYPE_SERVICE_ID] = {.type = NLA_U64}, + [LS_NLA_TYPE_DGID] = {.type = NLA_BINARY, + .len = sizeof(struct rdma_nla_ls_gid)}, + [LS_NLA_TYPE_SGID] = {.type = NLA_BINARY, + .len = sizeof(struct rdma_nla_ls_gid)}, + [LS_NLA_TYPE_TCLASS] = {.type = NLA_U8}, + [LS_NLA_TYPE_PKEY] = {.type = NLA_U16}, + [LS_NLA_TYPE_QOS_CLASS] = {.type = NLA_U16}, +}; + + +static int ib_sa_add_one(struct ib_device *device); +static void ib_sa_remove_one(struct ib_device *device, void *client_data); + +static struct ib_client sa_client = { + .name = "sa", + .add = ib_sa_add_one, + .remove = ib_sa_remove_one +}; + +static DEFINE_XARRAY_FLAGS(queries, XA_FLAGS_ALLOC | XA_FLAGS_LOCK_IRQ); + +static DEFINE_SPINLOCK(tid_lock); +static u32 tid; + +#define PATH_REC_FIELD(field) \ + .struct_offset_bytes = offsetof(struct sa_path_rec, field), \ + .struct_size_bytes = sizeof_field(struct sa_path_rec, field), \ + .field_name = "sa_path_rec:" #field + +static const struct ib_field path_rec_table[] = { + { PATH_REC_FIELD(service_id), + .offset_words = 0, + .offset_bits = 0, + .size_bits = 64 }, + { PATH_REC_FIELD(dgid), + .offset_words = 2, + .offset_bits = 0, + .size_bits = 128 }, + { PATH_REC_FIELD(sgid), + .offset_words = 6, + .offset_bits = 0, + .size_bits = 128 }, + { PATH_REC_FIELD(ib.dlid), + .offset_words = 10, + .offset_bits = 0, + .size_bits = 16 }, + { PATH_REC_FIELD(ib.slid), + .offset_words = 10, + .offset_bits = 16, + .size_bits = 16 }, + { PATH_REC_FIELD(ib.raw_traffic), + .offset_words = 11, + .offset_bits = 0, + .size_bits = 1 }, + { RESERVED, + .offset_words = 11, + .offset_bits = 1, + .size_bits = 3 }, + { PATH_REC_FIELD(flow_label), + .offset_words = 11, + .offset_bits = 4, + .size_bits = 20 }, + { PATH_REC_FIELD(hop_limit), + .offset_words = 11, + .offset_bits = 24, + .size_bits = 8 }, + { PATH_REC_FIELD(traffic_class), + .offset_words = 12, + .offset_bits = 0, + .size_bits = 8 }, + { PATH_REC_FIELD(reversible), + .offset_words = 12, + .offset_bits = 8, + .size_bits = 1 }, + { PATH_REC_FIELD(numb_path), + .offset_words = 12, + .offset_bits = 9, + .size_bits = 7 }, + { PATH_REC_FIELD(pkey), + .offset_words = 12, + .offset_bits = 16, + .size_bits = 16 }, + { PATH_REC_FIELD(qos_class), + .offset_words = 13, + .offset_bits = 0, + .size_bits = 12 }, + { PATH_REC_FIELD(sl), + .offset_words = 13, + .offset_bits = 12, + .size_bits = 4 }, + { PATH_REC_FIELD(mtu_selector), + .offset_words = 13, + .offset_bits = 16, + .size_bits = 2 }, + { PATH_REC_FIELD(mtu), + .offset_words = 13, + .offset_bits = 18, + .size_bits = 6 }, + { PATH_REC_FIELD(rate_selector), + .offset_words = 13, + .offset_bits = 24, + .size_bits = 2 }, + { PATH_REC_FIELD(rate), + .offset_words = 13, + .offset_bits = 26, + .size_bits = 6 }, + { PATH_REC_FIELD(packet_life_time_selector), + .offset_words = 14, + .offset_bits = 0, + .size_bits = 2 }, + { PATH_REC_FIELD(packet_life_time), + .offset_words = 14, + .offset_bits = 2, + .size_bits = 6 }, + { PATH_REC_FIELD(preference), + .offset_words = 14, + .offset_bits = 8, + .size_bits = 8 }, + { RESERVED, + .offset_words = 14, + .offset_bits = 16, + .size_bits = 48 }, +}; + +#define OPA_PATH_REC_FIELD(field) \ + .struct_offset_bytes = \ + offsetof(struct sa_path_rec, field), \ + .struct_size_bytes = \ + sizeof_field(struct sa_path_rec, field), \ + .field_name = "sa_path_rec:" #field + +static const struct ib_field opa_path_rec_table[] = { + { OPA_PATH_REC_FIELD(service_id), + .offset_words = 0, + .offset_bits = 0, + .size_bits = 64 }, + { OPA_PATH_REC_FIELD(dgid), + .offset_words = 2, + .offset_bits = 0, + .size_bits = 128 }, + { OPA_PATH_REC_FIELD(sgid), + .offset_words = 6, + .offset_bits = 0, + .size_bits = 128 }, + { OPA_PATH_REC_FIELD(opa.dlid), + .offset_words = 10, + .offset_bits = 0, + .size_bits = 32 }, + { OPA_PATH_REC_FIELD(opa.slid), + .offset_words = 11, + .offset_bits = 0, + .size_bits = 32 }, + { OPA_PATH_REC_FIELD(opa.raw_traffic), + .offset_words = 12, + .offset_bits = 0, + .size_bits = 1 }, + { RESERVED, + .offset_words = 12, + .offset_bits = 1, + .size_bits = 3 }, + { OPA_PATH_REC_FIELD(flow_label), + .offset_words = 12, + .offset_bits = 4, + .size_bits = 20 }, + { OPA_PATH_REC_FIELD(hop_limit), + .offset_words = 12, + .offset_bits = 24, + .size_bits = 8 }, + { OPA_PATH_REC_FIELD(traffic_class), + .offset_words = 13, + .offset_bits = 0, + .size_bits = 8 }, + { OPA_PATH_REC_FIELD(reversible), + .offset_words = 13, + .offset_bits = 8, + .size_bits = 1 }, + { OPA_PATH_REC_FIELD(numb_path), + .offset_words = 13, + .offset_bits = 9, + .size_bits = 7 }, + { OPA_PATH_REC_FIELD(pkey), + .offset_words = 13, + .offset_bits = 16, + .size_bits = 16 }, + { OPA_PATH_REC_FIELD(opa.l2_8B), + .offset_words = 14, + .offset_bits = 0, + .size_bits = 1 }, + { OPA_PATH_REC_FIELD(opa.l2_10B), + .offset_words = 14, + .offset_bits = 1, + .size_bits = 1 }, + { OPA_PATH_REC_FIELD(opa.l2_9B), + .offset_words = 14, + .offset_bits = 2, + .size_bits = 1 }, + { OPA_PATH_REC_FIELD(opa.l2_16B), + .offset_words = 14, + .offset_bits = 3, + .size_bits = 1 }, + { RESERVED, + .offset_words = 14, + .offset_bits = 4, + .size_bits = 2 }, + { OPA_PATH_REC_FIELD(opa.qos_type), + .offset_words = 14, + .offset_bits = 6, + .size_bits = 2 }, + { OPA_PATH_REC_FIELD(opa.qos_priority), + .offset_words = 14, + .offset_bits = 8, + .size_bits = 8 }, + { RESERVED, + .offset_words = 14, + .offset_bits = 16, + .size_bits = 3 }, + { OPA_PATH_REC_FIELD(sl), + .offset_words = 14, + .offset_bits = 19, + .size_bits = 5 }, + { RESERVED, + .offset_words = 14, + .offset_bits = 24, + .size_bits = 8 }, + { OPA_PATH_REC_FIELD(mtu_selector), + .offset_words = 15, + .offset_bits = 0, + .size_bits = 2 }, + { OPA_PATH_REC_FIELD(mtu), + .offset_words = 15, + .offset_bits = 2, + .size_bits = 6 }, + { OPA_PATH_REC_FIELD(rate_selector), + .offset_words = 15, + .offset_bits = 8, + .size_bits = 2 }, + { OPA_PATH_REC_FIELD(rate), + .offset_words = 15, + .offset_bits = 10, + .size_bits = 6 }, + { OPA_PATH_REC_FIELD(packet_life_time_selector), + .offset_words = 15, + .offset_bits = 16, + .size_bits = 2 }, + { OPA_PATH_REC_FIELD(packet_life_time), + .offset_words = 15, + .offset_bits = 18, + .size_bits = 6 }, + { OPA_PATH_REC_FIELD(preference), + .offset_words = 15, + .offset_bits = 24, + .size_bits = 8 }, +}; + +#define MCMEMBER_REC_FIELD(field) \ + .struct_offset_bytes = offsetof(struct ib_sa_mcmember_rec, field), \ + .struct_size_bytes = sizeof_field(struct ib_sa_mcmember_rec, field), \ + .field_name = "sa_mcmember_rec:" #field + +static const struct ib_field mcmember_rec_table[] = { + { MCMEMBER_REC_FIELD(mgid), + .offset_words = 0, + .offset_bits = 0, + .size_bits = 128 }, + { MCMEMBER_REC_FIELD(port_gid), + .offset_words = 4, + .offset_bits = 0, + .size_bits = 128 }, + { MCMEMBER_REC_FIELD(qkey), + .offset_words = 8, + .offset_bits = 0, + .size_bits = 32 }, + { MCMEMBER_REC_FIELD(mlid), + .offset_words = 9, + .offset_bits = 0, + .size_bits = 16 }, + { MCMEMBER_REC_FIELD(mtu_selector), + .offset_words = 9, + .offset_bits = 16, + .size_bits = 2 }, + { MCMEMBER_REC_FIELD(mtu), + .offset_words = 9, + .offset_bits = 18, + .size_bits = 6 }, + { MCMEMBER_REC_FIELD(traffic_class), + .offset_words = 9, + .offset_bits = 24, + .size_bits = 8 }, + { MCMEMBER_REC_FIELD(pkey), + .offset_words = 10, + .offset_bits = 0, + .size_bits = 16 }, + { MCMEMBER_REC_FIELD(rate_selector), + .offset_words = 10, + .offset_bits = 16, + .size_bits = 2 }, + { MCMEMBER_REC_FIELD(rate), + .offset_words = 10, + .offset_bits = 18, + .size_bits = 6 }, + { MCMEMBER_REC_FIELD(packet_life_time_selector), + .offset_words = 10, + .offset_bits = 24, + .size_bits = 2 }, + { MCMEMBER_REC_FIELD(packet_life_time), + .offset_words = 10, + .offset_bits = 26, + .size_bits = 6 }, + { MCMEMBER_REC_FIELD(sl), + .offset_words = 11, + .offset_bits = 0, + .size_bits = 4 }, + { MCMEMBER_REC_FIELD(flow_label), + .offset_words = 11, + .offset_bits = 4, + .size_bits = 20 }, + { MCMEMBER_REC_FIELD(hop_limit), + .offset_words = 11, + .offset_bits = 24, + .size_bits = 8 }, + { MCMEMBER_REC_FIELD(scope), + .offset_words = 12, + .offset_bits = 0, + .size_bits = 4 }, + { MCMEMBER_REC_FIELD(join_state), + .offset_words = 12, + .offset_bits = 4, + .size_bits = 4 }, + { MCMEMBER_REC_FIELD(proxy_join), + .offset_words = 12, + .offset_bits = 8, + .size_bits = 1 }, + { RESERVED, + .offset_words = 12, + .offset_bits = 9, + .size_bits = 23 }, +}; + +#define SERVICE_REC_FIELD(field) \ + .struct_offset_bytes = offsetof(struct ib_sa_service_rec, field), \ + .struct_size_bytes = sizeof_field(struct ib_sa_service_rec, field), \ + .field_name = "sa_service_rec:" #field + +static const struct ib_field service_rec_table[] = { + { SERVICE_REC_FIELD(id), + .offset_words = 0, + .offset_bits = 0, + .size_bits = 64 }, + { SERVICE_REC_FIELD(gid), + .offset_words = 2, + .offset_bits = 0, + .size_bits = 128 }, + { SERVICE_REC_FIELD(pkey), + .offset_words = 6, + .offset_bits = 0, + .size_bits = 16 }, + { SERVICE_REC_FIELD(lease), + .offset_words = 7, + .offset_bits = 0, + .size_bits = 32 }, + { SERVICE_REC_FIELD(key), + .offset_words = 8, + .offset_bits = 0, + .size_bits = 128 }, + { SERVICE_REC_FIELD(name), + .offset_words = 12, + .offset_bits = 0, + .size_bits = 64*8 }, + { SERVICE_REC_FIELD(data8), + .offset_words = 28, + .offset_bits = 0, + .size_bits = 16*8 }, + { SERVICE_REC_FIELD(data16), + .offset_words = 32, + .offset_bits = 0, + .size_bits = 8*16 }, + { SERVICE_REC_FIELD(data32), + .offset_words = 36, + .offset_bits = 0, + .size_bits = 4*32 }, + { SERVICE_REC_FIELD(data64), + .offset_words = 40, + .offset_bits = 0, + .size_bits = 2*64 }, +}; + +#define CLASSPORTINFO_REC_FIELD(field) \ + .struct_offset_bytes = offsetof(struct ib_class_port_info, field), \ + .struct_size_bytes = sizeof_field(struct ib_class_port_info, field), \ + .field_name = "ib_class_port_info:" #field + +static const struct ib_field ib_classport_info_rec_table[] = { + { CLASSPORTINFO_REC_FIELD(base_version), + .offset_words = 0, + .offset_bits = 0, + .size_bits = 8 }, + { CLASSPORTINFO_REC_FIELD(class_version), + .offset_words = 0, + .offset_bits = 8, + .size_bits = 8 }, + { CLASSPORTINFO_REC_FIELD(capability_mask), + .offset_words = 0, + .offset_bits = 16, + .size_bits = 16 }, + { CLASSPORTINFO_REC_FIELD(cap_mask2_resp_time), + .offset_words = 1, + .offset_bits = 0, + .size_bits = 32 }, + { CLASSPORTINFO_REC_FIELD(redirect_gid), + .offset_words = 2, + .offset_bits = 0, + .size_bits = 128 }, + { CLASSPORTINFO_REC_FIELD(redirect_tcslfl), + .offset_words = 6, + .offset_bits = 0, + .size_bits = 32 }, + { CLASSPORTINFO_REC_FIELD(redirect_lid), + .offset_words = 7, + .offset_bits = 0, + .size_bits = 16 }, + { CLASSPORTINFO_REC_FIELD(redirect_pkey), + .offset_words = 7, + .offset_bits = 16, + .size_bits = 16 }, + + { CLASSPORTINFO_REC_FIELD(redirect_qp), + .offset_words = 8, + .offset_bits = 0, + .size_bits = 32 }, + { CLASSPORTINFO_REC_FIELD(redirect_qkey), + .offset_words = 9, + .offset_bits = 0, + .size_bits = 32 }, + + { CLASSPORTINFO_REC_FIELD(trap_gid), + .offset_words = 10, + .offset_bits = 0, + .size_bits = 128 }, + { CLASSPORTINFO_REC_FIELD(trap_tcslfl), + .offset_words = 14, + .offset_bits = 0, + .size_bits = 32 }, + + { CLASSPORTINFO_REC_FIELD(trap_lid), + .offset_words = 15, + .offset_bits = 0, + .size_bits = 16 }, + { CLASSPORTINFO_REC_FIELD(trap_pkey), + .offset_words = 15, + .offset_bits = 16, + .size_bits = 16 }, + + { CLASSPORTINFO_REC_FIELD(trap_hlqp), + .offset_words = 16, + .offset_bits = 0, + .size_bits = 32 }, + { CLASSPORTINFO_REC_FIELD(trap_qkey), + .offset_words = 17, + .offset_bits = 0, + .size_bits = 32 }, +}; + +#define OPA_CLASSPORTINFO_REC_FIELD(field) \ + .struct_offset_bytes =\ + offsetof(struct opa_class_port_info, field), \ + .struct_size_bytes = \ + sizeof_field(struct opa_class_port_info, field), \ + .field_name = "opa_class_port_info:" #field + +static const struct ib_field opa_classport_info_rec_table[] = { + { OPA_CLASSPORTINFO_REC_FIELD(base_version), + .offset_words = 0, + .offset_bits = 0, + .size_bits = 8 }, + { OPA_CLASSPORTINFO_REC_FIELD(class_version), + .offset_words = 0, + .offset_bits = 8, + .size_bits = 8 }, + { OPA_CLASSPORTINFO_REC_FIELD(cap_mask), + .offset_words = 0, + .offset_bits = 16, + .size_bits = 16 }, + { OPA_CLASSPORTINFO_REC_FIELD(cap_mask2_resp_time), + .offset_words = 1, + .offset_bits = 0, + .size_bits = 32 }, + { OPA_CLASSPORTINFO_REC_FIELD(redirect_gid), + .offset_words = 2, + .offset_bits = 0, + .size_bits = 128 }, + { OPA_CLASSPORTINFO_REC_FIELD(redirect_tc_fl), + .offset_words = 6, + .offset_bits = 0, + .size_bits = 32 }, + { OPA_CLASSPORTINFO_REC_FIELD(redirect_lid), + .offset_words = 7, + .offset_bits = 0, + .size_bits = 32 }, + { OPA_CLASSPORTINFO_REC_FIELD(redirect_sl_qp), + .offset_words = 8, + .offset_bits = 0, + .size_bits = 32 }, + { OPA_CLASSPORTINFO_REC_FIELD(redirect_qkey), + .offset_words = 9, + .offset_bits = 0, + .size_bits = 32 }, + { OPA_CLASSPORTINFO_REC_FIELD(trap_gid), + .offset_words = 10, + .offset_bits = 0, + .size_bits = 128 }, + { OPA_CLASSPORTINFO_REC_FIELD(trap_tc_fl), + .offset_words = 14, + .offset_bits = 0, + .size_bits = 32 }, + { OPA_CLASSPORTINFO_REC_FIELD(trap_lid), + .offset_words = 15, + .offset_bits = 0, + .size_bits = 32 }, + { OPA_CLASSPORTINFO_REC_FIELD(trap_hl_qp), + .offset_words = 16, + .offset_bits = 0, + .size_bits = 32 }, + { OPA_CLASSPORTINFO_REC_FIELD(trap_qkey), + .offset_words = 17, + .offset_bits = 0, + .size_bits = 32 }, + { OPA_CLASSPORTINFO_REC_FIELD(trap_pkey), + .offset_words = 18, + .offset_bits = 0, + .size_bits = 16 }, + { OPA_CLASSPORTINFO_REC_FIELD(redirect_pkey), + .offset_words = 18, + .offset_bits = 16, + .size_bits = 16 }, + { OPA_CLASSPORTINFO_REC_FIELD(trap_sl_rsvd), + .offset_words = 19, + .offset_bits = 0, + .size_bits = 8 }, + { RESERVED, + .offset_words = 19, + .offset_bits = 8, + .size_bits = 24 }, +}; + +#define GUIDINFO_REC_FIELD(field) \ + .struct_offset_bytes = offsetof(struct ib_sa_guidinfo_rec, field), \ + .struct_size_bytes = sizeof_field(struct ib_sa_guidinfo_rec, field), \ + .field_name = "sa_guidinfo_rec:" #field + +static const struct ib_field guidinfo_rec_table[] = { + { GUIDINFO_REC_FIELD(lid), + .offset_words = 0, + .offset_bits = 0, + .size_bits = 16 }, + { GUIDINFO_REC_FIELD(block_num), + .offset_words = 0, + .offset_bits = 16, + .size_bits = 8 }, + { GUIDINFO_REC_FIELD(res1), + .offset_words = 0, + .offset_bits = 24, + .size_bits = 8 }, + { GUIDINFO_REC_FIELD(res2), + .offset_words = 1, + .offset_bits = 0, + .size_bits = 32 }, + { GUIDINFO_REC_FIELD(guid_info_list), + .offset_words = 2, + .offset_bits = 0, + .size_bits = 512 }, +}; + +static inline void ib_sa_disable_local_svc(struct ib_sa_query *query) +{ + query->flags &= ~IB_SA_ENABLE_LOCAL_SERVICE; +} + +static inline int ib_sa_query_cancelled(struct ib_sa_query *query) +{ + return (query->flags & IB_SA_CANCEL); +} + +static void ib_nl_set_path_rec_attrs(struct sk_buff *skb, + struct ib_sa_query *query) +{ + struct sa_path_rec *sa_rec = query->mad_buf->context[1]; + struct ib_sa_mad *mad = query->mad_buf->mad; + ib_sa_comp_mask comp_mask = mad->sa_hdr.comp_mask; + u16 val16; + u64 val64; + struct rdma_ls_resolve_header *header; + + query->mad_buf->context[1] = NULL; + + /* Construct the family header first */ + header = skb_put(skb, NLMSG_ALIGN(sizeof(*header))); + strscpy_pad(header->device_name, + dev_name(&query->port->agent->device->dev), + LS_DEVICE_NAME_MAX); + header->port_num = query->port->port_num; + + if ((comp_mask & IB_SA_PATH_REC_REVERSIBLE) && + sa_rec->reversible != 0) + query->path_use = LS_RESOLVE_PATH_USE_GMP; + else + query->path_use = LS_RESOLVE_PATH_USE_UNIDIRECTIONAL; + header->path_use = query->path_use; + + /* Now build the attributes */ + if (comp_mask & IB_SA_PATH_REC_SERVICE_ID) { + val64 = be64_to_cpu(sa_rec->service_id); + nla_put(skb, RDMA_NLA_F_MANDATORY | LS_NLA_TYPE_SERVICE_ID, + sizeof(val64), &val64); + } + if (comp_mask & IB_SA_PATH_REC_DGID) + nla_put(skb, RDMA_NLA_F_MANDATORY | LS_NLA_TYPE_DGID, + sizeof(sa_rec->dgid), &sa_rec->dgid); + if (comp_mask & IB_SA_PATH_REC_SGID) + nla_put(skb, RDMA_NLA_F_MANDATORY | LS_NLA_TYPE_SGID, + sizeof(sa_rec->sgid), &sa_rec->sgid); + if (comp_mask & IB_SA_PATH_REC_TRAFFIC_CLASS) + nla_put(skb, RDMA_NLA_F_MANDATORY | LS_NLA_TYPE_TCLASS, + sizeof(sa_rec->traffic_class), &sa_rec->traffic_class); + + if (comp_mask & IB_SA_PATH_REC_PKEY) { + val16 = be16_to_cpu(sa_rec->pkey); + nla_put(skb, RDMA_NLA_F_MANDATORY | LS_NLA_TYPE_PKEY, + sizeof(val16), &val16); + } + if (comp_mask & IB_SA_PATH_REC_QOS_CLASS) { + val16 = be16_to_cpu(sa_rec->qos_class); + nla_put(skb, RDMA_NLA_F_MANDATORY | LS_NLA_TYPE_QOS_CLASS, + sizeof(val16), &val16); + } +} + +static int ib_nl_get_path_rec_attrs_len(ib_sa_comp_mask comp_mask) +{ + int len = 0; + + if (comp_mask & IB_SA_PATH_REC_SERVICE_ID) + len += nla_total_size(sizeof(u64)); + if (comp_mask & IB_SA_PATH_REC_DGID) + len += nla_total_size(sizeof(struct rdma_nla_ls_gid)); + if (comp_mask & IB_SA_PATH_REC_SGID) + len += nla_total_size(sizeof(struct rdma_nla_ls_gid)); + if (comp_mask & IB_SA_PATH_REC_TRAFFIC_CLASS) + len += nla_total_size(sizeof(u8)); + if (comp_mask & IB_SA_PATH_REC_PKEY) + len += nla_total_size(sizeof(u16)); + if (comp_mask & IB_SA_PATH_REC_QOS_CLASS) + len += nla_total_size(sizeof(u16)); + + /* + * Make sure that at least some of the required comp_mask bits are + * set. + */ + if (WARN_ON(len == 0)) + return len; + + /* Add the family header */ + len += NLMSG_ALIGN(sizeof(struct rdma_ls_resolve_header)); + + return len; +} + +static int ib_nl_make_request(struct ib_sa_query *query, gfp_t gfp_mask) +{ + struct sk_buff *skb = NULL; + struct nlmsghdr *nlh; + void *data; + struct ib_sa_mad *mad; + int len; + unsigned long flags; + unsigned long delay; + gfp_t gfp_flag; + int ret; + + INIT_LIST_HEAD(&query->list); + query->seq = (u32)atomic_inc_return(&ib_nl_sa_request_seq); + + mad = query->mad_buf->mad; + len = ib_nl_get_path_rec_attrs_len(mad->sa_hdr.comp_mask); + if (len <= 0) + return -EMSGSIZE; + + skb = nlmsg_new(len, gfp_mask); + if (!skb) + return -ENOMEM; + + /* Put nlmsg header only for now */ + data = ibnl_put_msg(skb, &nlh, query->seq, 0, RDMA_NL_LS, + RDMA_NL_LS_OP_RESOLVE, NLM_F_REQUEST); + if (!data) { + nlmsg_free(skb); + return -EMSGSIZE; + } + + /* Add attributes */ + ib_nl_set_path_rec_attrs(skb, query); + + /* Repair the nlmsg header length */ + nlmsg_end(skb, nlh); + + gfp_flag = ((gfp_mask & GFP_ATOMIC) == GFP_ATOMIC) ? GFP_ATOMIC : + GFP_NOWAIT; + + spin_lock_irqsave(&ib_nl_request_lock, flags); + ret = rdma_nl_multicast(&init_net, skb, RDMA_NL_GROUP_LS, gfp_flag); + + if (ret) + goto out; + + /* Put the request on the list.*/ + delay = msecs_to_jiffies(sa_local_svc_timeout_ms); + query->timeout = delay + jiffies; + list_add_tail(&query->list, &ib_nl_request_list); + /* Start the timeout if this is the only request */ + if (ib_nl_request_list.next == &query->list) + queue_delayed_work(ib_nl_wq, &ib_nl_timed_work, delay); + +out: + spin_unlock_irqrestore(&ib_nl_request_lock, flags); + + return ret; +} + +static int ib_nl_cancel_request(struct ib_sa_query *query) +{ + unsigned long flags; + struct ib_sa_query *wait_query; + int found = 0; + + spin_lock_irqsave(&ib_nl_request_lock, flags); + list_for_each_entry(wait_query, &ib_nl_request_list, list) { + /* Let the timeout to take care of the callback */ + if (query == wait_query) { + query->flags |= IB_SA_CANCEL; + query->timeout = jiffies; + list_move(&query->list, &ib_nl_request_list); + found = 1; + mod_delayed_work(ib_nl_wq, &ib_nl_timed_work, 1); + break; + } + } + spin_unlock_irqrestore(&ib_nl_request_lock, flags); + + return found; +} + +static void send_handler(struct ib_mad_agent *agent, + struct ib_mad_send_wc *mad_send_wc); + +static void ib_nl_process_good_resolve_rsp(struct ib_sa_query *query, + const struct nlmsghdr *nlh) +{ + struct ib_mad_send_wc mad_send_wc; + struct ib_sa_mad *mad = NULL; + const struct nlattr *head, *curr; + struct ib_path_rec_data *rec; + int len, rem; + u32 mask = 0; + int status = -EIO; + + if (query->callback) { + head = (const struct nlattr *) nlmsg_data(nlh); + len = nlmsg_len(nlh); + switch (query->path_use) { + case LS_RESOLVE_PATH_USE_UNIDIRECTIONAL: + mask = IB_PATH_PRIMARY | IB_PATH_OUTBOUND; + break; + + case LS_RESOLVE_PATH_USE_ALL: + case LS_RESOLVE_PATH_USE_GMP: + default: + mask = IB_PATH_PRIMARY | IB_PATH_GMP | + IB_PATH_BIDIRECTIONAL; + break; + } + nla_for_each_attr(curr, head, len, rem) { + if (curr->nla_type == LS_NLA_TYPE_PATH_RECORD) { + rec = nla_data(curr); + /* + * Get the first one. In the future, we may + * need to get up to 6 pathrecords. + */ + if ((rec->flags & mask) == mask) { + mad = query->mad_buf->mad; + mad->mad_hdr.method |= + IB_MGMT_METHOD_RESP; + memcpy(mad->data, rec->path_rec, + sizeof(rec->path_rec)); + status = 0; + break; + } + } + } + query->callback(query, status, mad); + } + + mad_send_wc.send_buf = query->mad_buf; + mad_send_wc.status = IB_WC_SUCCESS; + send_handler(query->mad_buf->mad_agent, &mad_send_wc); +} + +static void ib_nl_request_timeout(struct work_struct *work) +{ + unsigned long flags; + struct ib_sa_query *query; + unsigned long delay; + struct ib_mad_send_wc mad_send_wc; + int ret; + + spin_lock_irqsave(&ib_nl_request_lock, flags); + while (!list_empty(&ib_nl_request_list)) { + query = list_entry(ib_nl_request_list.next, + struct ib_sa_query, list); + + if (time_after(query->timeout, jiffies)) { + delay = query->timeout - jiffies; + if ((long)delay <= 0) + delay = 1; + queue_delayed_work(ib_nl_wq, &ib_nl_timed_work, delay); + break; + } + + list_del(&query->list); + ib_sa_disable_local_svc(query); + /* Hold the lock to protect against query cancellation */ + if (ib_sa_query_cancelled(query)) + ret = -1; + else + ret = ib_post_send_mad(query->mad_buf, NULL); + if (ret) { + mad_send_wc.send_buf = query->mad_buf; + mad_send_wc.status = IB_WC_WR_FLUSH_ERR; + spin_unlock_irqrestore(&ib_nl_request_lock, flags); + send_handler(query->port->agent, &mad_send_wc); + spin_lock_irqsave(&ib_nl_request_lock, flags); + } + } + spin_unlock_irqrestore(&ib_nl_request_lock, flags); +} + +int ib_nl_handle_set_timeout(struct sk_buff *skb, + struct nlmsghdr *nlh, + struct netlink_ext_ack *extack) +{ + int timeout, delta, abs_delta; + const struct nlattr *attr; + unsigned long flags; + struct ib_sa_query *query; + long delay = 0; + struct nlattr *tb[LS_NLA_TYPE_MAX]; + int ret; + + if (!(nlh->nlmsg_flags & NLM_F_REQUEST) || + !(NETLINK_CB(skb).sk)) + return -EPERM; + + ret = nla_parse_deprecated(tb, LS_NLA_TYPE_MAX - 1, nlmsg_data(nlh), + nlmsg_len(nlh), ib_nl_policy, NULL); + attr = (const struct nlattr *)tb[LS_NLA_TYPE_TIMEOUT]; + if (ret || !attr) + goto settimeout_out; + + timeout = *(int *) nla_data(attr); + if (timeout < IB_SA_LOCAL_SVC_TIMEOUT_MIN) + timeout = IB_SA_LOCAL_SVC_TIMEOUT_MIN; + if (timeout > IB_SA_LOCAL_SVC_TIMEOUT_MAX) + timeout = IB_SA_LOCAL_SVC_TIMEOUT_MAX; + + delta = timeout - sa_local_svc_timeout_ms; + if (delta < 0) + abs_delta = -delta; + else + abs_delta = delta; + + if (delta != 0) { + spin_lock_irqsave(&ib_nl_request_lock, flags); + sa_local_svc_timeout_ms = timeout; + list_for_each_entry(query, &ib_nl_request_list, list) { + if (delta < 0 && abs_delta > query->timeout) + query->timeout = 0; + else + query->timeout += delta; + + /* Get the new delay from the first entry */ + if (!delay) { + delay = query->timeout - jiffies; + if (delay <= 0) + delay = 1; + } + } + if (delay) + mod_delayed_work(ib_nl_wq, &ib_nl_timed_work, + (unsigned long)delay); + spin_unlock_irqrestore(&ib_nl_request_lock, flags); + } + +settimeout_out: + return 0; +} + +static inline int ib_nl_is_good_resolve_resp(const struct nlmsghdr *nlh) +{ + struct nlattr *tb[LS_NLA_TYPE_MAX]; + int ret; + + if (nlh->nlmsg_flags & RDMA_NL_LS_F_ERR) + return 0; + + ret = nla_parse_deprecated(tb, LS_NLA_TYPE_MAX - 1, nlmsg_data(nlh), + nlmsg_len(nlh), ib_nl_policy, NULL); + if (ret) + return 0; + + return 1; +} + +int ib_nl_handle_resolve_resp(struct sk_buff *skb, + struct nlmsghdr *nlh, + struct netlink_ext_ack *extack) +{ + unsigned long flags; + struct ib_sa_query *query; + struct ib_mad_send_buf *send_buf; + struct ib_mad_send_wc mad_send_wc; + int found = 0; + int ret; + + if ((nlh->nlmsg_flags & NLM_F_REQUEST) || + !(NETLINK_CB(skb).sk)) + return -EPERM; + + spin_lock_irqsave(&ib_nl_request_lock, flags); + list_for_each_entry(query, &ib_nl_request_list, list) { + /* + * If the query is cancelled, let the timeout routine + * take care of it. + */ + if (nlh->nlmsg_seq == query->seq) { + found = !ib_sa_query_cancelled(query); + if (found) + list_del(&query->list); + break; + } + } + + if (!found) { + spin_unlock_irqrestore(&ib_nl_request_lock, flags); + goto resp_out; + } + + send_buf = query->mad_buf; + + if (!ib_nl_is_good_resolve_resp(nlh)) { + /* if the result is a failure, send out the packet via IB */ + ib_sa_disable_local_svc(query); + ret = ib_post_send_mad(query->mad_buf, NULL); + spin_unlock_irqrestore(&ib_nl_request_lock, flags); + if (ret) { + mad_send_wc.send_buf = send_buf; + mad_send_wc.status = IB_WC_GENERAL_ERR; + send_handler(query->port->agent, &mad_send_wc); + } + } else { + spin_unlock_irqrestore(&ib_nl_request_lock, flags); + ib_nl_process_good_resolve_rsp(query, nlh); + } + +resp_out: + return 0; +} + +static void free_sm_ah(struct kref *kref) +{ + struct ib_sa_sm_ah *sm_ah = container_of(kref, struct ib_sa_sm_ah, ref); + + rdma_destroy_ah(sm_ah->ah, 0); + kfree(sm_ah); +} + +void ib_sa_register_client(struct ib_sa_client *client) +{ + atomic_set(&client->users, 1); + init_completion(&client->comp); +} +EXPORT_SYMBOL(ib_sa_register_client); + +void ib_sa_unregister_client(struct ib_sa_client *client) +{ + ib_sa_client_put(client); + wait_for_completion(&client->comp); +} +EXPORT_SYMBOL(ib_sa_unregister_client); + +/** + * ib_sa_cancel_query - try to cancel an SA query + * @id:ID of query to cancel + * @query:query pointer to cancel + * + * Try to cancel an SA query. If the id and query don't match up or + * the query has already completed, nothing is done. Otherwise the + * query is canceled and will complete with a status of -EINTR. + */ +void ib_sa_cancel_query(int id, struct ib_sa_query *query) +{ + unsigned long flags; + struct ib_mad_agent *agent; + struct ib_mad_send_buf *mad_buf; + + xa_lock_irqsave(&queries, flags); + if (xa_load(&queries, id) != query) { + xa_unlock_irqrestore(&queries, flags); + return; + } + agent = query->port->agent; + mad_buf = query->mad_buf; + xa_unlock_irqrestore(&queries, flags); + + /* + * If the query is still on the netlink request list, schedule + * it to be cancelled by the timeout routine. Otherwise, it has been + * sent to the MAD layer and has to be cancelled from there. + */ + if (!ib_nl_cancel_request(query)) + ib_cancel_mad(agent, mad_buf); +} +EXPORT_SYMBOL(ib_sa_cancel_query); + +static u8 get_src_path_mask(struct ib_device *device, u8 port_num) +{ + struct ib_sa_device *sa_dev; + struct ib_sa_port *port; + unsigned long flags; + u8 src_path_mask; + + sa_dev = ib_get_client_data(device, &sa_client); + if (!sa_dev) + return 0x7f; + + port = &sa_dev->port[port_num - sa_dev->start_port]; + spin_lock_irqsave(&port->ah_lock, flags); + src_path_mask = port->sm_ah ? port->sm_ah->src_path_mask : 0x7f; + spin_unlock_irqrestore(&port->ah_lock, flags); + + return src_path_mask; +} + +static int init_ah_attr_grh_fields(struct ib_device *device, u8 port_num, + struct sa_path_rec *rec, + struct rdma_ah_attr *ah_attr, + const struct ib_gid_attr *gid_attr) +{ + enum ib_gid_type type = sa_conv_pathrec_to_gid_type(rec); + + if (!gid_attr) { + gid_attr = rdma_find_gid_by_port(device, &rec->sgid, type, + port_num, NULL); + if (IS_ERR(gid_attr)) + return PTR_ERR(gid_attr); + } else + rdma_hold_gid_attr(gid_attr); + + rdma_move_grh_sgid_attr(ah_attr, &rec->dgid, + be32_to_cpu(rec->flow_label), + rec->hop_limit, rec->traffic_class, + gid_attr); + return 0; +} + +/** + * ib_init_ah_attr_from_path - Initialize address handle attributes based on + * an SA path record. + * @device: Device associated ah attributes initialization. + * @port_num: Port on the specified device. + * @rec: path record entry to use for ah attributes initialization. + * @ah_attr: address handle attributes to initialization from path record. + * @gid_attr: SGID attribute to consider during initialization. + * + * When ib_init_ah_attr_from_path() returns success, + * (a) for IB link layer it optionally contains a reference to SGID attribute + * when GRH is present for IB link layer. + * (b) for RoCE link layer it contains a reference to SGID attribute. + * User must invoke rdma_destroy_ah_attr() to release reference to SGID + * attributes which are initialized using ib_init_ah_attr_from_path(). + */ +int ib_init_ah_attr_from_path(struct ib_device *device, u8 port_num, + struct sa_path_rec *rec, + struct rdma_ah_attr *ah_attr, + const struct ib_gid_attr *gid_attr) +{ + int ret = 0; + + memset(ah_attr, 0, sizeof(*ah_attr)); + ah_attr->type = rdma_ah_find_type(device, port_num); + rdma_ah_set_sl(ah_attr, rec->sl); + rdma_ah_set_port_num(ah_attr, port_num); + rdma_ah_set_static_rate(ah_attr, rec->rate); + + if (sa_path_is_roce(rec)) { + ret = roce_resolve_route_from_path(rec, gid_attr); + if (ret) + return ret; + + memcpy(ah_attr->roce.dmac, sa_path_get_dmac(rec), ETH_ALEN); + } else { + rdma_ah_set_dlid(ah_attr, be32_to_cpu(sa_path_get_dlid(rec))); + if (sa_path_is_opa(rec) && + rdma_ah_get_dlid(ah_attr) == be16_to_cpu(IB_LID_PERMISSIVE)) + rdma_ah_set_make_grd(ah_attr, true); + + rdma_ah_set_path_bits(ah_attr, + be32_to_cpu(sa_path_get_slid(rec)) & + get_src_path_mask(device, port_num)); + } + + if (rec->hop_limit > 0 || sa_path_is_roce(rec)) + ret = init_ah_attr_grh_fields(device, port_num, + rec, ah_attr, gid_attr); + return ret; +} +EXPORT_SYMBOL(ib_init_ah_attr_from_path); + +static int alloc_mad(struct ib_sa_query *query, gfp_t gfp_mask) +{ + struct rdma_ah_attr ah_attr; + unsigned long flags; + + spin_lock_irqsave(&query->port->ah_lock, flags); + if (!query->port->sm_ah) { + spin_unlock_irqrestore(&query->port->ah_lock, flags); + return -EAGAIN; + } + kref_get(&query->port->sm_ah->ref); + query->sm_ah = query->port->sm_ah; + spin_unlock_irqrestore(&query->port->ah_lock, flags); + + /* + * Always check if sm_ah has valid dlid assigned, + * before querying for class port info + */ + if ((rdma_query_ah(query->sm_ah->ah, &ah_attr) < 0) || + !rdma_is_valid_unicast_lid(&ah_attr)) { + kref_put(&query->sm_ah->ref, free_sm_ah); + return -EAGAIN; + } + query->mad_buf = ib_create_send_mad(query->port->agent, 1, + query->sm_ah->pkey_index, + 0, IB_MGMT_SA_HDR, IB_MGMT_SA_DATA, + gfp_mask, + ((query->flags & IB_SA_QUERY_OPA) ? + OPA_MGMT_BASE_VERSION : + IB_MGMT_BASE_VERSION)); + if (IS_ERR(query->mad_buf)) { + kref_put(&query->sm_ah->ref, free_sm_ah); + return -ENOMEM; + } + + query->mad_buf->ah = query->sm_ah->ah; + + return 0; +} + +static void free_mad(struct ib_sa_query *query) +{ + ib_free_send_mad(query->mad_buf); + kref_put(&query->sm_ah->ref, free_sm_ah); +} + +static void init_mad(struct ib_sa_query *query, struct ib_mad_agent *agent) +{ + struct ib_sa_mad *mad = query->mad_buf->mad; + unsigned long flags; + + memset(mad, 0, sizeof *mad); + + if (query->flags & IB_SA_QUERY_OPA) { + mad->mad_hdr.base_version = OPA_MGMT_BASE_VERSION; + mad->mad_hdr.class_version = OPA_SA_CLASS_VERSION; + } else { + mad->mad_hdr.base_version = IB_MGMT_BASE_VERSION; + mad->mad_hdr.class_version = IB_SA_CLASS_VERSION; + } + mad->mad_hdr.mgmt_class = IB_MGMT_CLASS_SUBN_ADM; + spin_lock_irqsave(&tid_lock, flags); + mad->mad_hdr.tid = + cpu_to_be64(((u64) agent->hi_tid) << 32 | tid++); + spin_unlock_irqrestore(&tid_lock, flags); +} + +static int send_mad(struct ib_sa_query *query, unsigned long timeout_ms, + gfp_t gfp_mask) +{ + unsigned long flags; + int ret, id; + + xa_lock_irqsave(&queries, flags); + ret = __xa_alloc(&queries, &id, query, xa_limit_32b, gfp_mask); + xa_unlock_irqrestore(&queries, flags); + if (ret < 0) + return ret; + + query->mad_buf->timeout_ms = timeout_ms; + query->mad_buf->context[0] = query; + query->id = id; + + if ((query->flags & IB_SA_ENABLE_LOCAL_SERVICE) && + (!(query->flags & IB_SA_QUERY_OPA))) { + if (rdma_nl_chk_listeners(RDMA_NL_GROUP_LS)) { + if (!ib_nl_make_request(query, gfp_mask)) + return id; + } + ib_sa_disable_local_svc(query); + } + + ret = ib_post_send_mad(query->mad_buf, NULL); + if (ret) { + xa_lock_irqsave(&queries, flags); + __xa_erase(&queries, id); + xa_unlock_irqrestore(&queries, flags); + } + + /* + * It's not safe to dereference query any more, because the + * send may already have completed and freed the query in + * another context. + */ + return ret ? ret : id; +} + +void ib_sa_unpack_path(void *attribute, struct sa_path_rec *rec) +{ + ib_unpack(path_rec_table, ARRAY_SIZE(path_rec_table), attribute, rec); +} +EXPORT_SYMBOL(ib_sa_unpack_path); + +void ib_sa_pack_path(struct sa_path_rec *rec, void *attribute) +{ + ib_pack(path_rec_table, ARRAY_SIZE(path_rec_table), rec, attribute); +} +EXPORT_SYMBOL(ib_sa_pack_path); + +static bool ib_sa_opa_pathrecord_support(struct ib_sa_client *client, + struct ib_sa_device *sa_dev, + u8 port_num) +{ + struct ib_sa_port *port; + unsigned long flags; + bool ret = false; + + port = &sa_dev->port[port_num - sa_dev->start_port]; + spin_lock_irqsave(&port->classport_lock, flags); + if (!port->classport_info.valid) + goto ret; + + if (port->classport_info.data.type == RDMA_CLASS_PORT_INFO_OPA) + ret = opa_get_cpi_capmask2(&port->classport_info.data.opa) & + OPA_CLASS_PORT_INFO_PR_SUPPORT; +ret: + spin_unlock_irqrestore(&port->classport_lock, flags); + return ret; +} + +enum opa_pr_supported { + PR_NOT_SUPPORTED, + PR_OPA_SUPPORTED, + PR_IB_SUPPORTED +}; + +/** + * Check if current PR query can be an OPA query. + * Retuns PR_NOT_SUPPORTED if a path record query is not + * possible, PR_OPA_SUPPORTED if an OPA path record query + * is possible and PR_IB_SUPPORTED if an IB path record + * query is possible. + */ +static int opa_pr_query_possible(struct ib_sa_client *client, + struct ib_sa_device *sa_dev, + struct ib_device *device, u8 port_num, + struct sa_path_rec *rec) +{ + struct ib_port_attr port_attr; + + if (ib_query_port(device, port_num, &port_attr)) + return PR_NOT_SUPPORTED; + + if (ib_sa_opa_pathrecord_support(client, sa_dev, port_num)) + return PR_OPA_SUPPORTED; + + if (port_attr.lid >= be16_to_cpu(IB_MULTICAST_LID_BASE)) + return PR_NOT_SUPPORTED; + else + return PR_IB_SUPPORTED; +} + +static void ib_sa_path_rec_callback(struct ib_sa_query *sa_query, + int status, + struct ib_sa_mad *mad) +{ + struct ib_sa_path_query *query = + container_of(sa_query, struct ib_sa_path_query, sa_query); + + if (mad) { + struct sa_path_rec rec; + + if (sa_query->flags & IB_SA_QUERY_OPA) { + ib_unpack(opa_path_rec_table, + ARRAY_SIZE(opa_path_rec_table), + mad->data, &rec); + rec.rec_type = SA_PATH_REC_TYPE_OPA; + query->callback(status, &rec, query->context); + } else { + ib_unpack(path_rec_table, + ARRAY_SIZE(path_rec_table), + mad->data, &rec); + rec.rec_type = SA_PATH_REC_TYPE_IB; + sa_path_set_dmac_zero(&rec); + + if (query->conv_pr) { + struct sa_path_rec opa; + + memset(&opa, 0, sizeof(struct sa_path_rec)); + sa_convert_path_ib_to_opa(&opa, &rec); + query->callback(status, &opa, query->context); + } else { + query->callback(status, &rec, query->context); + } + } + } else + query->callback(status, NULL, query->context); +} + +static void ib_sa_path_rec_release(struct ib_sa_query *sa_query) +{ + struct ib_sa_path_query *query = + container_of(sa_query, struct ib_sa_path_query, sa_query); + + kfree(query->conv_pr); + kfree(query); +} + +/** + * ib_sa_path_rec_get - Start a Path get query + * @client:SA client + * @device:device to send query on + * @port_num: port number to send query on + * @rec:Path Record to send in query + * @comp_mask:component mask to send in query + * @timeout_ms:time to wait for response + * @gfp_mask:GFP mask to use for internal allocations + * @callback:function called when query completes, times out or is + * canceled + * @context:opaque user context passed to callback + * @sa_query:query context, used to cancel query + * + * Send a Path Record Get query to the SA to look up a path. The + * callback function will be called when the query completes (or + * fails); status is 0 for a successful response, -EINTR if the query + * is canceled, -ETIMEDOUT is the query timed out, or -EIO if an error + * occurred sending the query. The resp parameter of the callback is + * only valid if status is 0. + * + * If the return value of ib_sa_path_rec_get() is negative, it is an + * error code. Otherwise it is a query ID that can be used to cancel + * the query. + */ +int ib_sa_path_rec_get(struct ib_sa_client *client, + struct ib_device *device, u8 port_num, + struct sa_path_rec *rec, + ib_sa_comp_mask comp_mask, + unsigned long timeout_ms, gfp_t gfp_mask, + void (*callback)(int status, + struct sa_path_rec *resp, + void *context), + void *context, + struct ib_sa_query **sa_query) +{ + struct ib_sa_path_query *query; + struct ib_sa_device *sa_dev = ib_get_client_data(device, &sa_client); + struct ib_sa_port *port; + struct ib_mad_agent *agent; + struct ib_sa_mad *mad; + enum opa_pr_supported status; + int ret; + + if (!sa_dev) + return -ENODEV; + + if ((rec->rec_type != SA_PATH_REC_TYPE_IB) && + (rec->rec_type != SA_PATH_REC_TYPE_OPA)) + return -EINVAL; + + port = &sa_dev->port[port_num - sa_dev->start_port]; + agent = port->agent; + + query = kzalloc(sizeof(*query), gfp_mask); + if (!query) + return -ENOMEM; + + query->sa_query.port = port; + if (rec->rec_type == SA_PATH_REC_TYPE_OPA) { + status = opa_pr_query_possible(client, sa_dev, device, port_num, + rec); + if (status == PR_NOT_SUPPORTED) { + ret = -EINVAL; + goto err1; + } else if (status == PR_OPA_SUPPORTED) { + query->sa_query.flags |= IB_SA_QUERY_OPA; + } else { + query->conv_pr = + kmalloc(sizeof(*query->conv_pr), gfp_mask); + if (!query->conv_pr) { + ret = -ENOMEM; + goto err1; + } + } + } + + ret = alloc_mad(&query->sa_query, gfp_mask); + if (ret) + goto err2; + + ib_sa_client_get(client); + query->sa_query.client = client; + query->callback = callback; + query->context = context; + + mad = query->sa_query.mad_buf->mad; + init_mad(&query->sa_query, agent); + + query->sa_query.callback = callback ? ib_sa_path_rec_callback : NULL; + query->sa_query.release = ib_sa_path_rec_release; + mad->mad_hdr.method = IB_MGMT_METHOD_GET; + mad->mad_hdr.attr_id = cpu_to_be16(IB_SA_ATTR_PATH_REC); + mad->sa_hdr.comp_mask = comp_mask; + + if (query->sa_query.flags & IB_SA_QUERY_OPA) { + ib_pack(opa_path_rec_table, ARRAY_SIZE(opa_path_rec_table), + rec, mad->data); + } else if (query->conv_pr) { + sa_convert_path_opa_to_ib(query->conv_pr, rec); + ib_pack(path_rec_table, ARRAY_SIZE(path_rec_table), + query->conv_pr, mad->data); + } else { + ib_pack(path_rec_table, ARRAY_SIZE(path_rec_table), + rec, mad->data); + } + + *sa_query = &query->sa_query; + + query->sa_query.flags |= IB_SA_ENABLE_LOCAL_SERVICE; + query->sa_query.mad_buf->context[1] = (query->conv_pr) ? + query->conv_pr : rec; + + ret = send_mad(&query->sa_query, timeout_ms, gfp_mask); + if (ret < 0) + goto err3; + + return ret; + +err3: + *sa_query = NULL; + ib_sa_client_put(query->sa_query.client); + free_mad(&query->sa_query); +err2: + kfree(query->conv_pr); +err1: + kfree(query); + return ret; +} +EXPORT_SYMBOL(ib_sa_path_rec_get); + +static void ib_sa_service_rec_callback(struct ib_sa_query *sa_query, + int status, + struct ib_sa_mad *mad) +{ + struct ib_sa_service_query *query = + container_of(sa_query, struct ib_sa_service_query, sa_query); + + if (mad) { + struct ib_sa_service_rec rec; + + ib_unpack(service_rec_table, ARRAY_SIZE(service_rec_table), + mad->data, &rec); + query->callback(status, &rec, query->context); + } else + query->callback(status, NULL, query->context); +} + +static void ib_sa_service_rec_release(struct ib_sa_query *sa_query) +{ + kfree(container_of(sa_query, struct ib_sa_service_query, sa_query)); +} + +/** + * ib_sa_service_rec_query - Start Service Record operation + * @client:SA client + * @device:device to send request on + * @port_num: port number to send request on + * @method:SA method - should be get, set, or delete + * @rec:Service Record to send in request + * @comp_mask:component mask to send in request + * @timeout_ms:time to wait for response + * @gfp_mask:GFP mask to use for internal allocations + * @callback:function called when request completes, times out or is + * canceled + * @context:opaque user context passed to callback + * @sa_query:request context, used to cancel request + * + * Send a Service Record set/get/delete to the SA to register, + * unregister or query a service record. + * The callback function will be called when the request completes (or + * fails); status is 0 for a successful response, -EINTR if the query + * is canceled, -ETIMEDOUT is the query timed out, or -EIO if an error + * occurred sending the query. The resp parameter of the callback is + * only valid if status is 0. + * + * If the return value of ib_sa_service_rec_query() is negative, it is an + * error code. Otherwise it is a request ID that can be used to cancel + * the query. + */ +int ib_sa_service_rec_query(struct ib_sa_client *client, + struct ib_device *device, u8 port_num, u8 method, + struct ib_sa_service_rec *rec, + ib_sa_comp_mask comp_mask, + unsigned long timeout_ms, gfp_t gfp_mask, + void (*callback)(int status, + struct ib_sa_service_rec *resp, + void *context), + void *context, + struct ib_sa_query **sa_query) +{ + struct ib_sa_service_query *query; + struct ib_sa_device *sa_dev = ib_get_client_data(device, &sa_client); + struct ib_sa_port *port; + struct ib_mad_agent *agent; + struct ib_sa_mad *mad; + int ret; + + if (!sa_dev) + return -ENODEV; + + port = &sa_dev->port[port_num - sa_dev->start_port]; + agent = port->agent; + + if (method != IB_MGMT_METHOD_GET && + method != IB_MGMT_METHOD_SET && + method != IB_SA_METHOD_DELETE) + return -EINVAL; + + query = kzalloc(sizeof(*query), gfp_mask); + if (!query) + return -ENOMEM; + + query->sa_query.port = port; + ret = alloc_mad(&query->sa_query, gfp_mask); + if (ret) + goto err1; + + ib_sa_client_get(client); + query->sa_query.client = client; + query->callback = callback; + query->context = context; + + mad = query->sa_query.mad_buf->mad; + init_mad(&query->sa_query, agent); + + query->sa_query.callback = callback ? ib_sa_service_rec_callback : NULL; + query->sa_query.release = ib_sa_service_rec_release; + mad->mad_hdr.method = method; + mad->mad_hdr.attr_id = cpu_to_be16(IB_SA_ATTR_SERVICE_REC); + mad->sa_hdr.comp_mask = comp_mask; + + ib_pack(service_rec_table, ARRAY_SIZE(service_rec_table), + rec, mad->data); + + *sa_query = &query->sa_query; + + ret = send_mad(&query->sa_query, timeout_ms, gfp_mask); + if (ret < 0) + goto err2; + + return ret; + +err2: + *sa_query = NULL; + ib_sa_client_put(query->sa_query.client); + free_mad(&query->sa_query); + +err1: + kfree(query); + return ret; +} +EXPORT_SYMBOL(ib_sa_service_rec_query); + +static void ib_sa_mcmember_rec_callback(struct ib_sa_query *sa_query, + int status, + struct ib_sa_mad *mad) +{ + struct ib_sa_mcmember_query *query = + container_of(sa_query, struct ib_sa_mcmember_query, sa_query); + + if (mad) { + struct ib_sa_mcmember_rec rec; + + ib_unpack(mcmember_rec_table, ARRAY_SIZE(mcmember_rec_table), + mad->data, &rec); + query->callback(status, &rec, query->context); + } else + query->callback(status, NULL, query->context); +} + +static void ib_sa_mcmember_rec_release(struct ib_sa_query *sa_query) +{ + kfree(container_of(sa_query, struct ib_sa_mcmember_query, sa_query)); +} + +int ib_sa_mcmember_rec_query(struct ib_sa_client *client, + struct ib_device *device, u8 port_num, + u8 method, + struct ib_sa_mcmember_rec *rec, + ib_sa_comp_mask comp_mask, + unsigned long timeout_ms, gfp_t gfp_mask, + void (*callback)(int status, + struct ib_sa_mcmember_rec *resp, + void *context), + void *context, + struct ib_sa_query **sa_query) +{ + struct ib_sa_mcmember_query *query; + struct ib_sa_device *sa_dev = ib_get_client_data(device, &sa_client); + struct ib_sa_port *port; + struct ib_mad_agent *agent; + struct ib_sa_mad *mad; + int ret; + + if (!sa_dev) + return -ENODEV; + + port = &sa_dev->port[port_num - sa_dev->start_port]; + agent = port->agent; + + query = kzalloc(sizeof(*query), gfp_mask); + if (!query) + return -ENOMEM; + + query->sa_query.port = port; + ret = alloc_mad(&query->sa_query, gfp_mask); + if (ret) + goto err1; + + ib_sa_client_get(client); + query->sa_query.client = client; + query->callback = callback; + query->context = context; + + mad = query->sa_query.mad_buf->mad; + init_mad(&query->sa_query, agent); + + query->sa_query.callback = callback ? ib_sa_mcmember_rec_callback : NULL; + query->sa_query.release = ib_sa_mcmember_rec_release; + mad->mad_hdr.method = method; + mad->mad_hdr.attr_id = cpu_to_be16(IB_SA_ATTR_MC_MEMBER_REC); + mad->sa_hdr.comp_mask = comp_mask; + + ib_pack(mcmember_rec_table, ARRAY_SIZE(mcmember_rec_table), + rec, mad->data); + + *sa_query = &query->sa_query; + + ret = send_mad(&query->sa_query, timeout_ms, gfp_mask); + if (ret < 0) + goto err2; + + return ret; + +err2: + *sa_query = NULL; + ib_sa_client_put(query->sa_query.client); + free_mad(&query->sa_query); + +err1: + kfree(query); + return ret; +} + +/* Support GuidInfoRecord */ +static void ib_sa_guidinfo_rec_callback(struct ib_sa_query *sa_query, + int status, + struct ib_sa_mad *mad) +{ + struct ib_sa_guidinfo_query *query = + container_of(sa_query, struct ib_sa_guidinfo_query, sa_query); + + if (mad) { + struct ib_sa_guidinfo_rec rec; + + ib_unpack(guidinfo_rec_table, ARRAY_SIZE(guidinfo_rec_table), + mad->data, &rec); + query->callback(status, &rec, query->context); + } else + query->callback(status, NULL, query->context); +} + +static void ib_sa_guidinfo_rec_release(struct ib_sa_query *sa_query) +{ + kfree(container_of(sa_query, struct ib_sa_guidinfo_query, sa_query)); +} + +int ib_sa_guid_info_rec_query(struct ib_sa_client *client, + struct ib_device *device, u8 port_num, + struct ib_sa_guidinfo_rec *rec, + ib_sa_comp_mask comp_mask, u8 method, + unsigned long timeout_ms, gfp_t gfp_mask, + void (*callback)(int status, + struct ib_sa_guidinfo_rec *resp, + void *context), + void *context, + struct ib_sa_query **sa_query) +{ + struct ib_sa_guidinfo_query *query; + struct ib_sa_device *sa_dev = ib_get_client_data(device, &sa_client); + struct ib_sa_port *port; + struct ib_mad_agent *agent; + struct ib_sa_mad *mad; + int ret; + + if (!sa_dev) + return -ENODEV; + + if (method != IB_MGMT_METHOD_GET && + method != IB_MGMT_METHOD_SET && + method != IB_SA_METHOD_DELETE) { + return -EINVAL; + } + + port = &sa_dev->port[port_num - sa_dev->start_port]; + agent = port->agent; + + query = kzalloc(sizeof(*query), gfp_mask); + if (!query) + return -ENOMEM; + + query->sa_query.port = port; + ret = alloc_mad(&query->sa_query, gfp_mask); + if (ret) + goto err1; + + ib_sa_client_get(client); + query->sa_query.client = client; + query->callback = callback; + query->context = context; + + mad = query->sa_query.mad_buf->mad; + init_mad(&query->sa_query, agent); + + query->sa_query.callback = callback ? ib_sa_guidinfo_rec_callback : NULL; + query->sa_query.release = ib_sa_guidinfo_rec_release; + + mad->mad_hdr.method = method; + mad->mad_hdr.attr_id = cpu_to_be16(IB_SA_ATTR_GUID_INFO_REC); + mad->sa_hdr.comp_mask = comp_mask; + + ib_pack(guidinfo_rec_table, ARRAY_SIZE(guidinfo_rec_table), rec, + mad->data); + + *sa_query = &query->sa_query; + + ret = send_mad(&query->sa_query, timeout_ms, gfp_mask); + if (ret < 0) + goto err2; + + return ret; + +err2: + *sa_query = NULL; + ib_sa_client_put(query->sa_query.client); + free_mad(&query->sa_query); + +err1: + kfree(query); + return ret; +} +EXPORT_SYMBOL(ib_sa_guid_info_rec_query); + +bool ib_sa_sendonly_fullmem_support(struct ib_sa_client *client, + struct ib_device *device, + u8 port_num) +{ + struct ib_sa_device *sa_dev = ib_get_client_data(device, &sa_client); + struct ib_sa_port *port; + bool ret = false; + unsigned long flags; + + if (!sa_dev) + return ret; + + port = &sa_dev->port[port_num - sa_dev->start_port]; + + spin_lock_irqsave(&port->classport_lock, flags); + if ((port->classport_info.valid) && + (port->classport_info.data.type == RDMA_CLASS_PORT_INFO_IB)) + ret = ib_get_cpi_capmask2(&port->classport_info.data.ib) + & IB_SA_CAP_MASK2_SENDONLY_FULL_MEM_SUPPORT; + spin_unlock_irqrestore(&port->classport_lock, flags); + return ret; +} +EXPORT_SYMBOL(ib_sa_sendonly_fullmem_support); + +struct ib_classport_info_context { + struct completion done; + struct ib_sa_query *sa_query; +}; + +static void ib_classportinfo_cb(void *context) +{ + struct ib_classport_info_context *cb_ctx = context; + + complete(&cb_ctx->done); +} + +static void ib_sa_classport_info_rec_callback(struct ib_sa_query *sa_query, + int status, + struct ib_sa_mad *mad) +{ + unsigned long flags; + struct ib_sa_classport_info_query *query = + container_of(sa_query, struct ib_sa_classport_info_query, sa_query); + struct ib_sa_classport_cache *info = &sa_query->port->classport_info; + + if (mad) { + if (sa_query->flags & IB_SA_QUERY_OPA) { + struct opa_class_port_info rec; + + ib_unpack(opa_classport_info_rec_table, + ARRAY_SIZE(opa_classport_info_rec_table), + mad->data, &rec); + + spin_lock_irqsave(&sa_query->port->classport_lock, + flags); + if (!status && !info->valid) { + memcpy(&info->data.opa, &rec, + sizeof(info->data.opa)); + + info->valid = true; + info->data.type = RDMA_CLASS_PORT_INFO_OPA; + } + spin_unlock_irqrestore(&sa_query->port->classport_lock, + flags); + + } else { + struct ib_class_port_info rec; + + ib_unpack(ib_classport_info_rec_table, + ARRAY_SIZE(ib_classport_info_rec_table), + mad->data, &rec); + + spin_lock_irqsave(&sa_query->port->classport_lock, + flags); + if (!status && !info->valid) { + memcpy(&info->data.ib, &rec, + sizeof(info->data.ib)); + + info->valid = true; + info->data.type = RDMA_CLASS_PORT_INFO_IB; + } + spin_unlock_irqrestore(&sa_query->port->classport_lock, + flags); + } + } + query->callback(query->context); +} + +static void ib_sa_classport_info_rec_release(struct ib_sa_query *sa_query) +{ + kfree(container_of(sa_query, struct ib_sa_classport_info_query, + sa_query)); +} + +static int ib_sa_classport_info_rec_query(struct ib_sa_port *port, + unsigned long timeout_ms, + void (*callback)(void *context), + void *context, + struct ib_sa_query **sa_query) +{ + struct ib_mad_agent *agent; + struct ib_sa_classport_info_query *query; + struct ib_sa_mad *mad; + gfp_t gfp_mask = GFP_KERNEL; + int ret; + + agent = port->agent; + + query = kzalloc(sizeof(*query), gfp_mask); + if (!query) + return -ENOMEM; + + query->sa_query.port = port; + query->sa_query.flags |= rdma_cap_opa_ah(port->agent->device, + port->port_num) ? + IB_SA_QUERY_OPA : 0; + ret = alloc_mad(&query->sa_query, gfp_mask); + if (ret) + goto err_free; + + query->callback = callback; + query->context = context; + + mad = query->sa_query.mad_buf->mad; + init_mad(&query->sa_query, agent); + + query->sa_query.callback = ib_sa_classport_info_rec_callback; + query->sa_query.release = ib_sa_classport_info_rec_release; + mad->mad_hdr.method = IB_MGMT_METHOD_GET; + mad->mad_hdr.attr_id = cpu_to_be16(IB_SA_ATTR_CLASS_PORTINFO); + mad->sa_hdr.comp_mask = 0; + *sa_query = &query->sa_query; + + ret = send_mad(&query->sa_query, timeout_ms, gfp_mask); + if (ret < 0) + goto err_free_mad; + + return ret; + +err_free_mad: + *sa_query = NULL; + free_mad(&query->sa_query); + +err_free: + kfree(query); + return ret; +} + +static void update_ib_cpi(struct work_struct *work) +{ + struct ib_sa_port *port = + container_of(work, struct ib_sa_port, ib_cpi_work.work); + struct ib_classport_info_context *cb_context; + unsigned long flags; + int ret; + + /* If the classport info is valid, nothing + * to do here. + */ + spin_lock_irqsave(&port->classport_lock, flags); + if (port->classport_info.valid) { + spin_unlock_irqrestore(&port->classport_lock, flags); + return; + } + spin_unlock_irqrestore(&port->classport_lock, flags); + + cb_context = kmalloc(sizeof(*cb_context), GFP_KERNEL); + if (!cb_context) + goto err_nomem; + + init_completion(&cb_context->done); + + ret = ib_sa_classport_info_rec_query(port, 3000, + ib_classportinfo_cb, cb_context, + &cb_context->sa_query); + if (ret < 0) + goto free_cb_err; + wait_for_completion(&cb_context->done); +free_cb_err: + kfree(cb_context); + spin_lock_irqsave(&port->classport_lock, flags); + + /* If the classport info is still not valid, the query should have + * failed for some reason. Retry issuing the query + */ + if (!port->classport_info.valid) { + port->classport_info.retry_cnt++; + if (port->classport_info.retry_cnt <= + IB_SA_CPI_MAX_RETRY_CNT) { + unsigned long delay = + msecs_to_jiffies(IB_SA_CPI_RETRY_WAIT); + + queue_delayed_work(ib_wq, &port->ib_cpi_work, delay); + } + } + spin_unlock_irqrestore(&port->classport_lock, flags); + +err_nomem: + return; +} + +static void send_handler(struct ib_mad_agent *agent, + struct ib_mad_send_wc *mad_send_wc) +{ + struct ib_sa_query *query = mad_send_wc->send_buf->context[0]; + unsigned long flags; + + if (query->callback) + switch (mad_send_wc->status) { + case IB_WC_SUCCESS: + /* No callback -- already got recv */ + break; + case IB_WC_RESP_TIMEOUT_ERR: + query->callback(query, -ETIMEDOUT, NULL); + break; + case IB_WC_WR_FLUSH_ERR: + query->callback(query, -EINTR, NULL); + break; + default: + query->callback(query, -EIO, NULL); + break; + } + + xa_lock_irqsave(&queries, flags); + __xa_erase(&queries, query->id); + xa_unlock_irqrestore(&queries, flags); + + free_mad(query); + if (query->client) + ib_sa_client_put(query->client); + query->release(query); +} + +static void recv_handler(struct ib_mad_agent *mad_agent, + struct ib_mad_send_buf *send_buf, + struct ib_mad_recv_wc *mad_recv_wc) +{ + struct ib_sa_query *query; + + if (!send_buf) + return; + + query = send_buf->context[0]; + if (query->callback) { + if (mad_recv_wc->wc->status == IB_WC_SUCCESS) + query->callback(query, + mad_recv_wc->recv_buf.mad->mad_hdr.status ? + -EINVAL : 0, + (struct ib_sa_mad *) mad_recv_wc->recv_buf.mad); + else + query->callback(query, -EIO, NULL); + } + + ib_free_recv_mad(mad_recv_wc); +} + +static void update_sm_ah(struct work_struct *work) +{ + struct ib_sa_port *port = + container_of(work, struct ib_sa_port, update_task); + struct ib_sa_sm_ah *new_ah; + struct ib_port_attr port_attr; + struct rdma_ah_attr ah_attr; + bool grh_required; + + if (ib_query_port(port->agent->device, port->port_num, &port_attr)) { + pr_warn("Couldn't query port\n"); + return; + } + + new_ah = kmalloc(sizeof(*new_ah), GFP_KERNEL); + if (!new_ah) + return; + + kref_init(&new_ah->ref); + new_ah->src_path_mask = (1 << port_attr.lmc) - 1; + + new_ah->pkey_index = 0; + if (ib_find_pkey(port->agent->device, port->port_num, + IB_DEFAULT_PKEY_FULL, &new_ah->pkey_index)) + pr_err("Couldn't find index for default PKey\n"); + + memset(&ah_attr, 0, sizeof(ah_attr)); + ah_attr.type = rdma_ah_find_type(port->agent->device, + port->port_num); + rdma_ah_set_dlid(&ah_attr, port_attr.sm_lid); + rdma_ah_set_sl(&ah_attr, port_attr.sm_sl); + rdma_ah_set_port_num(&ah_attr, port->port_num); + + grh_required = rdma_is_grh_required(port->agent->device, + port->port_num); + + /* + * The OPA sm_lid of 0xFFFF needs special handling so that it can be + * differentiated from a permissive LID of 0xFFFF. We set the + * grh_required flag here so the SA can program the DGID in the + * address handle appropriately + */ + if (ah_attr.type == RDMA_AH_ATTR_TYPE_OPA && + (grh_required || + port_attr.sm_lid == be16_to_cpu(IB_LID_PERMISSIVE))) + rdma_ah_set_make_grd(&ah_attr, true); + + if (ah_attr.type == RDMA_AH_ATTR_TYPE_IB && grh_required) { + rdma_ah_set_ah_flags(&ah_attr, IB_AH_GRH); + rdma_ah_set_subnet_prefix(&ah_attr, + cpu_to_be64(port_attr.subnet_prefix)); + rdma_ah_set_interface_id(&ah_attr, + cpu_to_be64(IB_SA_WELL_KNOWN_GUID)); + } + + new_ah->ah = rdma_create_ah(port->agent->qp->pd, &ah_attr, + RDMA_CREATE_AH_SLEEPABLE); + if (IS_ERR(new_ah->ah)) { + pr_warn("Couldn't create new SM AH\n"); + kfree(new_ah); + return; + } + + spin_lock_irq(&port->ah_lock); + if (port->sm_ah) + kref_put(&port->sm_ah->ref, free_sm_ah); + port->sm_ah = new_ah; + spin_unlock_irq(&port->ah_lock); +} + +static void ib_sa_event(struct ib_event_handler *handler, + struct ib_event *event) +{ + if (event->event == IB_EVENT_PORT_ERR || + event->event == IB_EVENT_PORT_ACTIVE || + event->event == IB_EVENT_LID_CHANGE || + event->event == IB_EVENT_PKEY_CHANGE || + event->event == IB_EVENT_SM_CHANGE || + event->event == IB_EVENT_CLIENT_REREGISTER) { + unsigned long flags; + struct ib_sa_device *sa_dev = + container_of(handler, typeof(*sa_dev), event_handler); + u8 port_num = event->element.port_num - sa_dev->start_port; + struct ib_sa_port *port = &sa_dev->port[port_num]; + + if (!rdma_cap_ib_sa(handler->device, port->port_num)) + return; + + spin_lock_irqsave(&port->ah_lock, flags); + if (port->sm_ah) + kref_put(&port->sm_ah->ref, free_sm_ah); + port->sm_ah = NULL; + spin_unlock_irqrestore(&port->ah_lock, flags); + + if (event->event == IB_EVENT_SM_CHANGE || + event->event == IB_EVENT_CLIENT_REREGISTER || + event->event == IB_EVENT_LID_CHANGE || + event->event == IB_EVENT_PORT_ACTIVE) { + unsigned long delay = + msecs_to_jiffies(IB_SA_CPI_RETRY_WAIT); + + spin_lock_irqsave(&port->classport_lock, flags); + port->classport_info.valid = false; + port->classport_info.retry_cnt = 0; + spin_unlock_irqrestore(&port->classport_lock, flags); + queue_delayed_work(ib_wq, + &port->ib_cpi_work, delay); + } + queue_work(ib_wq, &sa_dev->port[port_num].update_task); + } +} + +static int ib_sa_add_one(struct ib_device *device) +{ + struct ib_sa_device *sa_dev; + int s, e, i; + int count = 0; + int ret; + + s = rdma_start_port(device); + e = rdma_end_port(device); + + sa_dev = kzalloc(struct_size(sa_dev, port, e - s + 1), GFP_KERNEL); + if (!sa_dev) + return -ENOMEM; + + sa_dev->start_port = s; + sa_dev->end_port = e; + + for (i = 0; i <= e - s; ++i) { + spin_lock_init(&sa_dev->port[i].ah_lock); + if (!rdma_cap_ib_sa(device, i + 1)) + continue; + + sa_dev->port[i].sm_ah = NULL; + sa_dev->port[i].port_num = i + s; + + spin_lock_init(&sa_dev->port[i].classport_lock); + sa_dev->port[i].classport_info.valid = false; + + sa_dev->port[i].agent = + ib_register_mad_agent(device, i + s, IB_QPT_GSI, + NULL, 0, send_handler, + recv_handler, sa_dev, 0); + if (IS_ERR(sa_dev->port[i].agent)) { + ret = PTR_ERR(sa_dev->port[i].agent); + goto err; + } + + INIT_WORK(&sa_dev->port[i].update_task, update_sm_ah); + INIT_DELAYED_WORK(&sa_dev->port[i].ib_cpi_work, + update_ib_cpi); + + count++; + } + + if (!count) { + ret = -EOPNOTSUPP; + goto free; + } + + ib_set_client_data(device, &sa_client, sa_dev); + + /* + * We register our event handler after everything is set up, + * and then update our cached info after the event handler is + * registered to avoid any problems if a port changes state + * during our initialization. + */ + + INIT_IB_EVENT_HANDLER(&sa_dev->event_handler, device, ib_sa_event); + ib_register_event_handler(&sa_dev->event_handler); + + for (i = 0; i <= e - s; ++i) { + if (rdma_cap_ib_sa(device, i + 1)) + update_sm_ah(&sa_dev->port[i].update_task); + } + + return 0; + +err: + while (--i >= 0) { + if (rdma_cap_ib_sa(device, i + 1)) + ib_unregister_mad_agent(sa_dev->port[i].agent); + } +free: + kfree(sa_dev); + return ret; +} + +static void ib_sa_remove_one(struct ib_device *device, void *client_data) +{ + struct ib_sa_device *sa_dev = client_data; + int i; + + ib_unregister_event_handler(&sa_dev->event_handler); + flush_workqueue(ib_wq); + + for (i = 0; i <= sa_dev->end_port - sa_dev->start_port; ++i) { + if (rdma_cap_ib_sa(device, i + 1)) { + cancel_delayed_work_sync(&sa_dev->port[i].ib_cpi_work); + ib_unregister_mad_agent(sa_dev->port[i].agent); + if (sa_dev->port[i].sm_ah) + kref_put(&sa_dev->port[i].sm_ah->ref, free_sm_ah); + } + + } + + kfree(sa_dev); +} + +int ib_sa_init(void) +{ + int ret; + + get_random_bytes(&tid, sizeof tid); + + atomic_set(&ib_nl_sa_request_seq, 0); + + ret = ib_register_client(&sa_client); + if (ret) { + pr_err("Couldn't register ib_sa client\n"); + goto err1; + } + + ret = mcast_init(); + if (ret) { + pr_err("Couldn't initialize multicast handling\n"); + goto err2; + } + + ib_nl_wq = alloc_ordered_workqueue("ib_nl_sa_wq", WQ_MEM_RECLAIM); + if (!ib_nl_wq) { + ret = -ENOMEM; + goto err3; + } + + INIT_DELAYED_WORK(&ib_nl_timed_work, ib_nl_request_timeout); + + return 0; + +err3: + mcast_cleanup(); +err2: + ib_unregister_client(&sa_client); +err1: + return ret; +} + +void ib_sa_cleanup(void) +{ + cancel_delayed_work(&ib_nl_timed_work); + flush_workqueue(ib_nl_wq); + destroy_workqueue(ib_nl_wq); + mcast_cleanup(); + ib_unregister_client(&sa_client); + WARN_ON(!xa_empty(&queries)); +} diff --git a/drivers/infiniband/core/security.c b/drivers/infiniband/core/security.c new file mode 100644 index 000000000..75e7ec017 --- /dev/null +++ b/drivers/infiniband/core/security.c @@ -0,0 +1,753 @@ +/* + * Copyright (c) 2016 Mellanox Technologies Ltd. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include <linux/security.h> +#include <linux/completion.h> +#include <linux/list.h> + +#include <rdma/ib_verbs.h> +#include <rdma/ib_cache.h> +#include "core_priv.h" +#include "mad_priv.h" + +static LIST_HEAD(mad_agent_list); +/* Lock to protect mad_agent_list */ +static DEFINE_SPINLOCK(mad_agent_list_lock); + +static struct pkey_index_qp_list *get_pkey_idx_qp_list(struct ib_port_pkey *pp) +{ + struct pkey_index_qp_list *pkey = NULL; + struct pkey_index_qp_list *tmp_pkey; + struct ib_device *dev = pp->sec->dev; + + spin_lock(&dev->port_data[pp->port_num].pkey_list_lock); + list_for_each_entry (tmp_pkey, &dev->port_data[pp->port_num].pkey_list, + pkey_index_list) { + if (tmp_pkey->pkey_index == pp->pkey_index) { + pkey = tmp_pkey; + break; + } + } + spin_unlock(&dev->port_data[pp->port_num].pkey_list_lock); + return pkey; +} + +static int get_pkey_and_subnet_prefix(struct ib_port_pkey *pp, + u16 *pkey, + u64 *subnet_prefix) +{ + struct ib_device *dev = pp->sec->dev; + int ret; + + ret = ib_get_cached_pkey(dev, pp->port_num, pp->pkey_index, pkey); + if (ret) + return ret; + + ret = ib_get_cached_subnet_prefix(dev, pp->port_num, subnet_prefix); + + return ret; +} + +static int enforce_qp_pkey_security(u16 pkey, + u64 subnet_prefix, + struct ib_qp_security *qp_sec) +{ + struct ib_qp_security *shared_qp_sec; + int ret; + + ret = security_ib_pkey_access(qp_sec->security, subnet_prefix, pkey); + if (ret) + return ret; + + list_for_each_entry(shared_qp_sec, + &qp_sec->shared_qp_list, + shared_qp_list) { + ret = security_ib_pkey_access(shared_qp_sec->security, + subnet_prefix, + pkey); + if (ret) + return ret; + } + return 0; +} + +/* The caller of this function must hold the QP security + * mutex of the QP of the security structure in *pps. + * + * It takes separate ports_pkeys and security structure + * because in some cases the pps will be for a new settings + * or the pps will be for the real QP and security structure + * will be for a shared QP. + */ +static int check_qp_port_pkey_settings(struct ib_ports_pkeys *pps, + struct ib_qp_security *sec) +{ + u64 subnet_prefix; + u16 pkey; + int ret = 0; + + if (!pps) + return 0; + + if (pps->main.state != IB_PORT_PKEY_NOT_VALID) { + ret = get_pkey_and_subnet_prefix(&pps->main, + &pkey, + &subnet_prefix); + if (ret) + return ret; + + ret = enforce_qp_pkey_security(pkey, + subnet_prefix, + sec); + if (ret) + return ret; + } + + if (pps->alt.state != IB_PORT_PKEY_NOT_VALID) { + ret = get_pkey_and_subnet_prefix(&pps->alt, + &pkey, + &subnet_prefix); + if (ret) + return ret; + + ret = enforce_qp_pkey_security(pkey, + subnet_prefix, + sec); + } + + return ret; +} + +/* The caller of this function must hold the QP security + * mutex. + */ +static void qp_to_error(struct ib_qp_security *sec) +{ + struct ib_qp_security *shared_qp_sec; + struct ib_qp_attr attr = { + .qp_state = IB_QPS_ERR + }; + struct ib_event event = { + .event = IB_EVENT_QP_FATAL + }; + + /* If the QP is in the process of being destroyed + * the qp pointer in the security structure is + * undefined. It cannot be modified now. + */ + if (sec->destroying) + return; + + ib_modify_qp(sec->qp, + &attr, + IB_QP_STATE); + + if (sec->qp->event_handler && sec->qp->qp_context) { + event.element.qp = sec->qp; + sec->qp->event_handler(&event, + sec->qp->qp_context); + } + + list_for_each_entry(shared_qp_sec, + &sec->shared_qp_list, + shared_qp_list) { + struct ib_qp *qp = shared_qp_sec->qp; + + if (qp->event_handler && qp->qp_context) { + event.element.qp = qp; + event.device = qp->device; + qp->event_handler(&event, + qp->qp_context); + } + } +} + +static inline void check_pkey_qps(struct pkey_index_qp_list *pkey, + struct ib_device *device, + u8 port_num, + u64 subnet_prefix) +{ + struct ib_port_pkey *pp, *tmp_pp; + bool comp; + LIST_HEAD(to_error_list); + u16 pkey_val; + + if (!ib_get_cached_pkey(device, + port_num, + pkey->pkey_index, + &pkey_val)) { + spin_lock(&pkey->qp_list_lock); + list_for_each_entry(pp, &pkey->qp_list, qp_list) { + if (atomic_read(&pp->sec->error_list_count)) + continue; + + if (enforce_qp_pkey_security(pkey_val, + subnet_prefix, + pp->sec)) { + atomic_inc(&pp->sec->error_list_count); + list_add(&pp->to_error_list, + &to_error_list); + } + } + spin_unlock(&pkey->qp_list_lock); + } + + list_for_each_entry_safe(pp, + tmp_pp, + &to_error_list, + to_error_list) { + mutex_lock(&pp->sec->mutex); + qp_to_error(pp->sec); + list_del(&pp->to_error_list); + atomic_dec(&pp->sec->error_list_count); + comp = pp->sec->destroying; + mutex_unlock(&pp->sec->mutex); + + if (comp) + complete(&pp->sec->error_complete); + } +} + +/* The caller of this function must hold the QP security + * mutex. + */ +static int port_pkey_list_insert(struct ib_port_pkey *pp) +{ + struct pkey_index_qp_list *tmp_pkey; + struct pkey_index_qp_list *pkey; + struct ib_device *dev; + u8 port_num = pp->port_num; + int ret = 0; + + if (pp->state != IB_PORT_PKEY_VALID) + return 0; + + dev = pp->sec->dev; + + pkey = get_pkey_idx_qp_list(pp); + + if (!pkey) { + bool found = false; + + pkey = kzalloc(sizeof(*pkey), GFP_KERNEL); + if (!pkey) + return -ENOMEM; + + spin_lock(&dev->port_data[port_num].pkey_list_lock); + /* Check for the PKey again. A racing process may + * have created it. + */ + list_for_each_entry(tmp_pkey, + &dev->port_data[port_num].pkey_list, + pkey_index_list) { + if (tmp_pkey->pkey_index == pp->pkey_index) { + kfree(pkey); + pkey = tmp_pkey; + found = true; + break; + } + } + + if (!found) { + pkey->pkey_index = pp->pkey_index; + spin_lock_init(&pkey->qp_list_lock); + INIT_LIST_HEAD(&pkey->qp_list); + list_add(&pkey->pkey_index_list, + &dev->port_data[port_num].pkey_list); + } + spin_unlock(&dev->port_data[port_num].pkey_list_lock); + } + + spin_lock(&pkey->qp_list_lock); + list_add(&pp->qp_list, &pkey->qp_list); + spin_unlock(&pkey->qp_list_lock); + + pp->state = IB_PORT_PKEY_LISTED; + + return ret; +} + +/* The caller of this function must hold the QP security + * mutex. + */ +static void port_pkey_list_remove(struct ib_port_pkey *pp) +{ + struct pkey_index_qp_list *pkey; + + if (pp->state != IB_PORT_PKEY_LISTED) + return; + + pkey = get_pkey_idx_qp_list(pp); + + spin_lock(&pkey->qp_list_lock); + list_del(&pp->qp_list); + spin_unlock(&pkey->qp_list_lock); + + /* The setting may still be valid, i.e. after + * a destroy has failed for example. + */ + pp->state = IB_PORT_PKEY_VALID; +} + +static void destroy_qp_security(struct ib_qp_security *sec) +{ + security_ib_free_security(sec->security); + kfree(sec->ports_pkeys); + kfree(sec); +} + +/* The caller of this function must hold the QP security + * mutex. + */ +static struct ib_ports_pkeys *get_new_pps(const struct ib_qp *qp, + const struct ib_qp_attr *qp_attr, + int qp_attr_mask) +{ + struct ib_ports_pkeys *new_pps; + struct ib_ports_pkeys *qp_pps = qp->qp_sec->ports_pkeys; + + new_pps = kzalloc(sizeof(*new_pps), GFP_KERNEL); + if (!new_pps) + return NULL; + + if (qp_attr_mask & IB_QP_PORT) + new_pps->main.port_num = qp_attr->port_num; + else if (qp_pps) + new_pps->main.port_num = qp_pps->main.port_num; + + if (qp_attr_mask & IB_QP_PKEY_INDEX) + new_pps->main.pkey_index = qp_attr->pkey_index; + else if (qp_pps) + new_pps->main.pkey_index = qp_pps->main.pkey_index; + + if (((qp_attr_mask & IB_QP_PKEY_INDEX) && + (qp_attr_mask & IB_QP_PORT)) || + (qp_pps && qp_pps->main.state != IB_PORT_PKEY_NOT_VALID)) + new_pps->main.state = IB_PORT_PKEY_VALID; + + if (qp_attr_mask & IB_QP_ALT_PATH) { + new_pps->alt.port_num = qp_attr->alt_port_num; + new_pps->alt.pkey_index = qp_attr->alt_pkey_index; + new_pps->alt.state = IB_PORT_PKEY_VALID; + } else if (qp_pps) { + new_pps->alt.port_num = qp_pps->alt.port_num; + new_pps->alt.pkey_index = qp_pps->alt.pkey_index; + if (qp_pps->alt.state != IB_PORT_PKEY_NOT_VALID) + new_pps->alt.state = IB_PORT_PKEY_VALID; + } + + new_pps->main.sec = qp->qp_sec; + new_pps->alt.sec = qp->qp_sec; + return new_pps; +} + +int ib_open_shared_qp_security(struct ib_qp *qp, struct ib_device *dev) +{ + struct ib_qp *real_qp = qp->real_qp; + int ret; + + ret = ib_create_qp_security(qp, dev); + + if (ret) + return ret; + + if (!qp->qp_sec) + return 0; + + mutex_lock(&real_qp->qp_sec->mutex); + ret = check_qp_port_pkey_settings(real_qp->qp_sec->ports_pkeys, + qp->qp_sec); + + if (ret) + goto ret; + + if (qp != real_qp) + list_add(&qp->qp_sec->shared_qp_list, + &real_qp->qp_sec->shared_qp_list); +ret: + mutex_unlock(&real_qp->qp_sec->mutex); + if (ret) + destroy_qp_security(qp->qp_sec); + + return ret; +} + +void ib_close_shared_qp_security(struct ib_qp_security *sec) +{ + struct ib_qp *real_qp = sec->qp->real_qp; + + mutex_lock(&real_qp->qp_sec->mutex); + list_del(&sec->shared_qp_list); + mutex_unlock(&real_qp->qp_sec->mutex); + + destroy_qp_security(sec); +} + +int ib_create_qp_security(struct ib_qp *qp, struct ib_device *dev) +{ + unsigned int i; + bool is_ib = false; + int ret; + + rdma_for_each_port (dev, i) { + is_ib = rdma_protocol_ib(dev, i); + if (is_ib) + break; + } + + /* If this isn't an IB device don't create the security context */ + if (!is_ib) + return 0; + + qp->qp_sec = kzalloc(sizeof(*qp->qp_sec), GFP_KERNEL); + if (!qp->qp_sec) + return -ENOMEM; + + qp->qp_sec->qp = qp; + qp->qp_sec->dev = dev; + mutex_init(&qp->qp_sec->mutex); + INIT_LIST_HEAD(&qp->qp_sec->shared_qp_list); + atomic_set(&qp->qp_sec->error_list_count, 0); + init_completion(&qp->qp_sec->error_complete); + ret = security_ib_alloc_security(&qp->qp_sec->security); + if (ret) { + kfree(qp->qp_sec); + qp->qp_sec = NULL; + } + + return ret; +} +EXPORT_SYMBOL(ib_create_qp_security); + +void ib_destroy_qp_security_begin(struct ib_qp_security *sec) +{ + /* Return if not IB */ + if (!sec) + return; + + mutex_lock(&sec->mutex); + + /* Remove the QP from the lists so it won't get added to + * a to_error_list during the destroy process. + */ + if (sec->ports_pkeys) { + port_pkey_list_remove(&sec->ports_pkeys->main); + port_pkey_list_remove(&sec->ports_pkeys->alt); + } + + /* If the QP is already in one or more of those lists + * the destroying flag will ensure the to error flow + * doesn't operate on an undefined QP. + */ + sec->destroying = true; + + /* Record the error list count to know how many completions + * to wait for. + */ + sec->error_comps_pending = atomic_read(&sec->error_list_count); + + mutex_unlock(&sec->mutex); +} + +void ib_destroy_qp_security_abort(struct ib_qp_security *sec) +{ + int ret; + int i; + + /* Return if not IB */ + if (!sec) + return; + + /* If a concurrent cache update is in progress this + * QP security could be marked for an error state + * transition. Wait for this to complete. + */ + for (i = 0; i < sec->error_comps_pending; i++) + wait_for_completion(&sec->error_complete); + + mutex_lock(&sec->mutex); + sec->destroying = false; + + /* Restore the position in the lists and verify + * access is still allowed in case a cache update + * occurred while attempting to destroy. + * + * Because these setting were listed already + * and removed during ib_destroy_qp_security_begin + * we know the pkey_index_qp_list for the PKey + * already exists so port_pkey_list_insert won't fail. + */ + if (sec->ports_pkeys) { + port_pkey_list_insert(&sec->ports_pkeys->main); + port_pkey_list_insert(&sec->ports_pkeys->alt); + } + + ret = check_qp_port_pkey_settings(sec->ports_pkeys, sec); + if (ret) + qp_to_error(sec); + + mutex_unlock(&sec->mutex); +} + +void ib_destroy_qp_security_end(struct ib_qp_security *sec) +{ + int i; + + /* Return if not IB */ + if (!sec) + return; + + /* If a concurrent cache update is occurring we must + * wait until this QP security structure is processed + * in the QP to error flow before destroying it because + * the to_error_list is in use. + */ + for (i = 0; i < sec->error_comps_pending; i++) + wait_for_completion(&sec->error_complete); + + destroy_qp_security(sec); +} + +void ib_security_cache_change(struct ib_device *device, + u8 port_num, + u64 subnet_prefix) +{ + struct pkey_index_qp_list *pkey; + + list_for_each_entry (pkey, &device->port_data[port_num].pkey_list, + pkey_index_list) { + check_pkey_qps(pkey, + device, + port_num, + subnet_prefix); + } +} + +void ib_security_release_port_pkey_list(struct ib_device *device) +{ + struct pkey_index_qp_list *pkey, *tmp_pkey; + unsigned int i; + + rdma_for_each_port (device, i) { + list_for_each_entry_safe(pkey, + tmp_pkey, + &device->port_data[i].pkey_list, + pkey_index_list) { + list_del(&pkey->pkey_index_list); + kfree(pkey); + } + } +} + +int ib_security_modify_qp(struct ib_qp *qp, + struct ib_qp_attr *qp_attr, + int qp_attr_mask, + struct ib_udata *udata) +{ + int ret = 0; + struct ib_ports_pkeys *tmp_pps; + struct ib_ports_pkeys *new_pps = NULL; + struct ib_qp *real_qp = qp->real_qp; + bool special_qp = (real_qp->qp_type == IB_QPT_SMI || + real_qp->qp_type == IB_QPT_GSI || + real_qp->qp_type >= IB_QPT_RESERVED1); + bool pps_change = ((qp_attr_mask & (IB_QP_PKEY_INDEX | IB_QP_PORT)) || + (qp_attr_mask & IB_QP_ALT_PATH)); + + WARN_ONCE((qp_attr_mask & IB_QP_PORT && + rdma_protocol_ib(real_qp->device, qp_attr->port_num) && + !real_qp->qp_sec), + "%s: QP security is not initialized for IB QP: %d\n", + __func__, real_qp->qp_num); + + /* The port/pkey settings are maintained only for the real QP. Open + * handles on the real QP will be in the shared_qp_list. When + * enforcing security on the real QP all the shared QPs will be + * checked as well. + */ + + if (pps_change && !special_qp && real_qp->qp_sec) { + mutex_lock(&real_qp->qp_sec->mutex); + new_pps = get_new_pps(real_qp, + qp_attr, + qp_attr_mask); + if (!new_pps) { + mutex_unlock(&real_qp->qp_sec->mutex); + return -ENOMEM; + } + /* Add this QP to the lists for the new port + * and pkey settings before checking for permission + * in case there is a concurrent cache update + * occurring. Walking the list for a cache change + * doesn't acquire the security mutex unless it's + * sending the QP to error. + */ + ret = port_pkey_list_insert(&new_pps->main); + + if (!ret) + ret = port_pkey_list_insert(&new_pps->alt); + + if (!ret) + ret = check_qp_port_pkey_settings(new_pps, + real_qp->qp_sec); + } + + if (!ret) + ret = real_qp->device->ops.modify_qp(real_qp, + qp_attr, + qp_attr_mask, + udata); + + if (new_pps) { + /* Clean up the lists and free the appropriate + * ports_pkeys structure. + */ + if (ret) { + tmp_pps = new_pps; + } else { + tmp_pps = real_qp->qp_sec->ports_pkeys; + real_qp->qp_sec->ports_pkeys = new_pps; + } + + if (tmp_pps) { + port_pkey_list_remove(&tmp_pps->main); + port_pkey_list_remove(&tmp_pps->alt); + } + kfree(tmp_pps); + mutex_unlock(&real_qp->qp_sec->mutex); + } + return ret; +} + +static int ib_security_pkey_access(struct ib_device *dev, + u8 port_num, + u16 pkey_index, + void *sec) +{ + u64 subnet_prefix; + u16 pkey; + int ret; + + if (!rdma_protocol_ib(dev, port_num)) + return 0; + + ret = ib_get_cached_pkey(dev, port_num, pkey_index, &pkey); + if (ret) + return ret; + + ret = ib_get_cached_subnet_prefix(dev, port_num, &subnet_prefix); + + if (ret) + return ret; + + return security_ib_pkey_access(sec, subnet_prefix, pkey); +} + +void ib_mad_agent_security_change(void) +{ + struct ib_mad_agent *ag; + + spin_lock(&mad_agent_list_lock); + list_for_each_entry(ag, + &mad_agent_list, + mad_agent_sec_list) + WRITE_ONCE(ag->smp_allowed, + !security_ib_endport_manage_subnet(ag->security, + dev_name(&ag->device->dev), ag->port_num)); + spin_unlock(&mad_agent_list_lock); +} + +int ib_mad_agent_security_setup(struct ib_mad_agent *agent, + enum ib_qp_type qp_type) +{ + int ret; + + if (!rdma_protocol_ib(agent->device, agent->port_num)) + return 0; + + INIT_LIST_HEAD(&agent->mad_agent_sec_list); + + ret = security_ib_alloc_security(&agent->security); + if (ret) + return ret; + + if (qp_type != IB_QPT_SMI) + return 0; + + spin_lock(&mad_agent_list_lock); + ret = security_ib_endport_manage_subnet(agent->security, + dev_name(&agent->device->dev), + agent->port_num); + if (ret) + goto free_security; + + WRITE_ONCE(agent->smp_allowed, true); + list_add(&agent->mad_agent_sec_list, &mad_agent_list); + spin_unlock(&mad_agent_list_lock); + return 0; + +free_security: + spin_unlock(&mad_agent_list_lock); + security_ib_free_security(agent->security); + return ret; +} + +void ib_mad_agent_security_cleanup(struct ib_mad_agent *agent) +{ + if (!rdma_protocol_ib(agent->device, agent->port_num)) + return; + + if (agent->qp->qp_type == IB_QPT_SMI) { + spin_lock(&mad_agent_list_lock); + list_del(&agent->mad_agent_sec_list); + spin_unlock(&mad_agent_list_lock); + } + + security_ib_free_security(agent->security); +} + +int ib_mad_enforce_security(struct ib_mad_agent_private *map, u16 pkey_index) +{ + if (!rdma_protocol_ib(map->agent.device, map->agent.port_num)) + return 0; + + if (map->agent.qp->qp_type == IB_QPT_SMI) { + if (!READ_ONCE(map->agent.smp_allowed)) + return -EACCES; + return 0; + } + + return ib_security_pkey_access(map->agent.device, + map->agent.port_num, + pkey_index, + map->agent.security); +} diff --git a/drivers/infiniband/core/smi.c b/drivers/infiniband/core/smi.c new file mode 100644 index 000000000..f19b23817 --- /dev/null +++ b/drivers/infiniband/core/smi.c @@ -0,0 +1,338 @@ +/* + * Copyright (c) 2004, 2005 Mellanox Technologies Ltd. All rights reserved. + * Copyright (c) 2004, 2005 Infinicon Corporation. All rights reserved. + * Copyright (c) 2004, 2005 Intel Corporation. All rights reserved. + * Copyright (c) 2004, 2005 Topspin Corporation. All rights reserved. + * Copyright (c) 2004-2007 Voltaire Corporation. All rights reserved. + * Copyright (c) 2005 Sun Microsystems, Inc. All rights reserved. + * Copyright (c) 2014 Intel Corporation. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + */ + +#include <rdma/ib_smi.h> +#include "smi.h" +#include "opa_smi.h" + +static enum smi_action __smi_handle_dr_smp_send(bool is_switch, int port_num, + u8 *hop_ptr, u8 hop_cnt, + const u8 *initial_path, + const u8 *return_path, + u8 direction, + bool dr_dlid_is_permissive, + bool dr_slid_is_permissive) +{ + /* See section 14.2.2.2, Vol 1 IB spec */ + /* C14-6 -- valid hop_cnt values are from 0 to 63 */ + if (hop_cnt >= IB_SMP_MAX_PATH_HOPS) + return IB_SMI_DISCARD; + + if (!direction) { + /* C14-9:1 */ + if (hop_cnt && *hop_ptr == 0) { + (*hop_ptr)++; + return (initial_path[*hop_ptr] == + port_num ? IB_SMI_HANDLE : IB_SMI_DISCARD); + } + + /* C14-9:2 */ + if (*hop_ptr && *hop_ptr < hop_cnt) { + if (!is_switch) + return IB_SMI_DISCARD; + + /* return_path set when received */ + (*hop_ptr)++; + return (initial_path[*hop_ptr] == + port_num ? IB_SMI_HANDLE : IB_SMI_DISCARD); + } + + /* C14-9:3 -- We're at the end of the DR segment of path */ + if (*hop_ptr == hop_cnt) { + /* return_path set when received */ + (*hop_ptr)++; + return (is_switch || + dr_dlid_is_permissive ? + IB_SMI_HANDLE : IB_SMI_DISCARD); + } + + /* C14-9:4 -- hop_ptr = hop_cnt + 1 -> give to SMA/SM */ + /* C14-9:5 -- Fail unreasonable hop pointer */ + return (*hop_ptr == hop_cnt + 1 ? IB_SMI_HANDLE : IB_SMI_DISCARD); + + } else { + /* C14-13:1 */ + if (hop_cnt && *hop_ptr == hop_cnt + 1) { + (*hop_ptr)--; + return (return_path[*hop_ptr] == + port_num ? IB_SMI_HANDLE : IB_SMI_DISCARD); + } + + /* C14-13:2 */ + if (2 <= *hop_ptr && *hop_ptr <= hop_cnt) { + if (!is_switch) + return IB_SMI_DISCARD; + + (*hop_ptr)--; + return (return_path[*hop_ptr] == + port_num ? IB_SMI_HANDLE : IB_SMI_DISCARD); + } + + /* C14-13:3 -- at the end of the DR segment of path */ + if (*hop_ptr == 1) { + (*hop_ptr)--; + /* C14-13:3 -- SMPs destined for SM shouldn't be here */ + return (is_switch || + dr_slid_is_permissive ? + IB_SMI_HANDLE : IB_SMI_DISCARD); + } + + /* C14-13:4 -- hop_ptr = 0 -> should have gone to SM */ + if (*hop_ptr == 0) + return IB_SMI_HANDLE; + + /* C14-13:5 -- Check for unreasonable hop pointer */ + return IB_SMI_DISCARD; + } +} + +/* + * Fixup a directed route SMP for sending + * Return IB_SMI_DISCARD if the SMP should be discarded + */ +enum smi_action smi_handle_dr_smp_send(struct ib_smp *smp, + bool is_switch, int port_num) +{ + return __smi_handle_dr_smp_send(is_switch, port_num, + &smp->hop_ptr, smp->hop_cnt, + smp->initial_path, + smp->return_path, + ib_get_smp_direction(smp), + smp->dr_dlid == IB_LID_PERMISSIVE, + smp->dr_slid == IB_LID_PERMISSIVE); +} + +enum smi_action opa_smi_handle_dr_smp_send(struct opa_smp *smp, + bool is_switch, int port_num) +{ + return __smi_handle_dr_smp_send(is_switch, port_num, + &smp->hop_ptr, smp->hop_cnt, + smp->route.dr.initial_path, + smp->route.dr.return_path, + opa_get_smp_direction(smp), + smp->route.dr.dr_dlid == + OPA_LID_PERMISSIVE, + smp->route.dr.dr_slid == + OPA_LID_PERMISSIVE); +} + +static enum smi_action __smi_handle_dr_smp_recv(bool is_switch, int port_num, + int phys_port_cnt, + u8 *hop_ptr, u8 hop_cnt, + const u8 *initial_path, + u8 *return_path, + u8 direction, + bool dr_dlid_is_permissive, + bool dr_slid_is_permissive) +{ + /* See section 14.2.2.2, Vol 1 IB spec */ + /* C14-6 -- valid hop_cnt values are from 0 to 63 */ + if (hop_cnt >= IB_SMP_MAX_PATH_HOPS) + return IB_SMI_DISCARD; + + if (!direction) { + /* C14-9:1 -- sender should have incremented hop_ptr */ + if (hop_cnt && *hop_ptr == 0) + return IB_SMI_DISCARD; + + /* C14-9:2 -- intermediate hop */ + if (*hop_ptr && *hop_ptr < hop_cnt) { + if (!is_switch) + return IB_SMI_DISCARD; + + return_path[*hop_ptr] = port_num; + /* hop_ptr updated when sending */ + return (initial_path[*hop_ptr+1] <= phys_port_cnt ? + IB_SMI_HANDLE : IB_SMI_DISCARD); + } + + /* C14-9:3 -- We're at the end of the DR segment of path */ + if (*hop_ptr == hop_cnt) { + if (hop_cnt) + return_path[*hop_ptr] = port_num; + /* hop_ptr updated when sending */ + + return (is_switch || + dr_dlid_is_permissive ? + IB_SMI_HANDLE : IB_SMI_DISCARD); + } + + /* C14-9:4 -- hop_ptr = hop_cnt + 1 -> give to SMA/SM */ + /* C14-9:5 -- fail unreasonable hop pointer */ + return (*hop_ptr == hop_cnt + 1 ? IB_SMI_HANDLE : IB_SMI_DISCARD); + + } else { + + /* C14-13:1 */ + if (hop_cnt && *hop_ptr == hop_cnt + 1) { + (*hop_ptr)--; + return (return_path[*hop_ptr] == + port_num ? IB_SMI_HANDLE : IB_SMI_DISCARD); + } + + /* C14-13:2 */ + if (2 <= *hop_ptr && *hop_ptr <= hop_cnt) { + if (!is_switch) + return IB_SMI_DISCARD; + + /* hop_ptr updated when sending */ + return (return_path[*hop_ptr-1] <= phys_port_cnt ? + IB_SMI_HANDLE : IB_SMI_DISCARD); + } + + /* C14-13:3 -- We're at the end of the DR segment of path */ + if (*hop_ptr == 1) { + if (dr_slid_is_permissive) { + /* giving SMP to SM - update hop_ptr */ + (*hop_ptr)--; + return IB_SMI_HANDLE; + } + /* hop_ptr updated when sending */ + return (is_switch ? IB_SMI_HANDLE : IB_SMI_DISCARD); + } + + /* C14-13:4 -- hop_ptr = 0 -> give to SM */ + /* C14-13:5 -- Check for unreasonable hop pointer */ + return (*hop_ptr == 0 ? IB_SMI_HANDLE : IB_SMI_DISCARD); + } +} + +/* + * Adjust information for a received SMP + * Return IB_SMI_DISCARD if the SMP should be dropped + */ +enum smi_action smi_handle_dr_smp_recv(struct ib_smp *smp, bool is_switch, + int port_num, int phys_port_cnt) +{ + return __smi_handle_dr_smp_recv(is_switch, port_num, phys_port_cnt, + &smp->hop_ptr, smp->hop_cnt, + smp->initial_path, + smp->return_path, + ib_get_smp_direction(smp), + smp->dr_dlid == IB_LID_PERMISSIVE, + smp->dr_slid == IB_LID_PERMISSIVE); +} + +/* + * Adjust information for a received SMP + * Return IB_SMI_DISCARD if the SMP should be dropped + */ +enum smi_action opa_smi_handle_dr_smp_recv(struct opa_smp *smp, bool is_switch, + int port_num, int phys_port_cnt) +{ + return __smi_handle_dr_smp_recv(is_switch, port_num, phys_port_cnt, + &smp->hop_ptr, smp->hop_cnt, + smp->route.dr.initial_path, + smp->route.dr.return_path, + opa_get_smp_direction(smp), + smp->route.dr.dr_dlid == + OPA_LID_PERMISSIVE, + smp->route.dr.dr_slid == + OPA_LID_PERMISSIVE); +} + +static enum smi_forward_action __smi_check_forward_dr_smp(u8 hop_ptr, u8 hop_cnt, + u8 direction, + bool dr_dlid_is_permissive, + bool dr_slid_is_permissive) +{ + if (!direction) { + /* C14-9:2 -- intermediate hop */ + if (hop_ptr && hop_ptr < hop_cnt) + return IB_SMI_FORWARD; + + /* C14-9:3 -- at the end of the DR segment of path */ + if (hop_ptr == hop_cnt) + return (dr_dlid_is_permissive ? + IB_SMI_SEND : IB_SMI_LOCAL); + + /* C14-9:4 -- hop_ptr = hop_cnt + 1 -> give to SMA/SM */ + if (hop_ptr == hop_cnt + 1) + return IB_SMI_SEND; + } else { + /* C14-13:2 -- intermediate hop */ + if (2 <= hop_ptr && hop_ptr <= hop_cnt) + return IB_SMI_FORWARD; + + /* C14-13:3 -- at the end of the DR segment of path */ + if (hop_ptr == 1) + return (!dr_slid_is_permissive ? + IB_SMI_SEND : IB_SMI_LOCAL); + } + return IB_SMI_LOCAL; + +} + +enum smi_forward_action smi_check_forward_dr_smp(struct ib_smp *smp) +{ + return __smi_check_forward_dr_smp(smp->hop_ptr, smp->hop_cnt, + ib_get_smp_direction(smp), + smp->dr_dlid == IB_LID_PERMISSIVE, + smp->dr_slid == IB_LID_PERMISSIVE); +} + +enum smi_forward_action opa_smi_check_forward_dr_smp(struct opa_smp *smp) +{ + return __smi_check_forward_dr_smp(smp->hop_ptr, smp->hop_cnt, + opa_get_smp_direction(smp), + smp->route.dr.dr_dlid == + OPA_LID_PERMISSIVE, + smp->route.dr.dr_slid == + OPA_LID_PERMISSIVE); +} + +/* + * Return the forwarding port number from initial_path for outgoing SMP and + * from return_path for returning SMP + */ +int smi_get_fwd_port(struct ib_smp *smp) +{ + return (!ib_get_smp_direction(smp) ? smp->initial_path[smp->hop_ptr+1] : + smp->return_path[smp->hop_ptr-1]); +} + +/* + * Return the forwarding port number from initial_path for outgoing SMP and + * from return_path for returning SMP + */ +int opa_smi_get_fwd_port(struct opa_smp *smp) +{ + return !opa_get_smp_direction(smp) ? smp->route.dr.initial_path[smp->hop_ptr+1] : + smp->route.dr.return_path[smp->hop_ptr-1]; +} diff --git a/drivers/infiniband/core/smi.h b/drivers/infiniband/core/smi.h new file mode 100644 index 000000000..91d9b353a --- /dev/null +++ b/drivers/infiniband/core/smi.h @@ -0,0 +1,90 @@ +/* + * Copyright (c) 2004 Mellanox Technologies Ltd. All rights reserved. + * Copyright (c) 2004 Infinicon Corporation. All rights reserved. + * Copyright (c) 2004 Intel Corporation. All rights reserved. + * Copyright (c) 2004 Topspin Corporation. All rights reserved. + * Copyright (c) 2004-2007 Voltaire Corporation. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + */ + +#ifndef __SMI_H_ +#define __SMI_H_ + +#include <rdma/ib_smi.h> + +enum smi_action { + IB_SMI_DISCARD, + IB_SMI_HANDLE +}; + +enum smi_forward_action { + IB_SMI_LOCAL, /* SMP should be completed up the stack */ + IB_SMI_SEND, /* received DR SMP should be forwarded to the send queue */ + IB_SMI_FORWARD /* SMP should be forwarded (for switches only) */ +}; + +enum smi_action smi_handle_dr_smp_recv(struct ib_smp *smp, bool is_switch, + int port_num, int phys_port_cnt); +int smi_get_fwd_port(struct ib_smp *smp); +extern enum smi_forward_action smi_check_forward_dr_smp(struct ib_smp *smp); +extern enum smi_action smi_handle_dr_smp_send(struct ib_smp *smp, + bool is_switch, int port_num); + +/* + * Return IB_SMI_HANDLE if the SMP should be handled by the local SMA/SM + * via process_mad + */ +static inline enum smi_action smi_check_local_smp(struct ib_smp *smp, + struct ib_device *device) +{ + /* C14-9:3 -- We're at the end of the DR segment of path */ + /* C14-9:4 -- Hop Pointer = Hop Count + 1 -> give to SMA/SM */ + return ((device->ops.process_mad && + !ib_get_smp_direction(smp) && + (smp->hop_ptr == smp->hop_cnt + 1)) ? + IB_SMI_HANDLE : IB_SMI_DISCARD); +} + +/* + * Return IB_SMI_HANDLE if the SMP should be handled by the local SMA/SM + * via process_mad + */ +static inline enum smi_action smi_check_local_returning_smp(struct ib_smp *smp, + struct ib_device *device) +{ + /* C14-13:3 -- We're at the end of the DR segment of path */ + /* C14-13:4 -- Hop Pointer == 0 -> give to SM */ + return ((device->ops.process_mad && + ib_get_smp_direction(smp) && + !smp->hop_ptr) ? IB_SMI_HANDLE : IB_SMI_DISCARD); +} + +#endif /* __SMI_H_ */ diff --git a/drivers/infiniband/core/sysfs.c b/drivers/infiniband/core/sysfs.c new file mode 100644 index 000000000..914cddea5 --- /dev/null +++ b/drivers/infiniband/core/sysfs.c @@ -0,0 +1,1445 @@ +/* + * Copyright (c) 2004, 2005 Topspin Communications. All rights reserved. + * Copyright (c) 2005 Mellanox Technologies Ltd. All rights reserved. + * Copyright (c) 2005 Sun Microsystems, Inc. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include "core_priv.h" + +#include <linux/slab.h> +#include <linux/stat.h> +#include <linux/string.h> +#include <linux/netdevice.h> +#include <linux/ethtool.h> + +#include <rdma/ib_mad.h> +#include <rdma/ib_pma.h> +#include <rdma/ib_cache.h> +#include <rdma/rdma_counter.h> + +struct ib_port; + +struct gid_attr_group { + struct ib_port *port; + struct kobject kobj; + struct attribute_group ndev; + struct attribute_group type; +}; +struct ib_port { + struct kobject kobj; + struct ib_device *ibdev; + struct gid_attr_group *gid_attr_group; + struct attribute_group gid_group; + struct attribute_group *pkey_group; + const struct attribute_group *pma_table; + struct attribute_group *hw_stats_ag; + struct rdma_hw_stats *hw_stats; + u8 port_num; +}; + +struct port_attribute { + struct attribute attr; + ssize_t (*show)(struct ib_port *, struct port_attribute *, char *buf); + ssize_t (*store)(struct ib_port *, struct port_attribute *, + const char *buf, size_t count); +}; + +#define PORT_ATTR(_name, _mode, _show, _store) \ +struct port_attribute port_attr_##_name = __ATTR(_name, _mode, _show, _store) + +#define PORT_ATTR_RO(_name) \ +struct port_attribute port_attr_##_name = __ATTR_RO(_name) + +struct port_table_attribute { + struct port_attribute attr; + char name[8]; + int index; + __be16 attr_id; +}; + +struct hw_stats_attribute { + struct attribute attr; + ssize_t (*show)(struct kobject *kobj, + struct attribute *attr, char *buf); + ssize_t (*store)(struct kobject *kobj, + struct attribute *attr, + const char *buf, + size_t count); + int index; + u8 port_num; +}; + +static ssize_t port_attr_show(struct kobject *kobj, + struct attribute *attr, char *buf) +{ + struct port_attribute *port_attr = + container_of(attr, struct port_attribute, attr); + struct ib_port *p = container_of(kobj, struct ib_port, kobj); + + if (!port_attr->show) + return -EIO; + + return port_attr->show(p, port_attr, buf); +} + +static ssize_t port_attr_store(struct kobject *kobj, + struct attribute *attr, + const char *buf, size_t count) +{ + struct port_attribute *port_attr = + container_of(attr, struct port_attribute, attr); + struct ib_port *p = container_of(kobj, struct ib_port, kobj); + + if (!port_attr->store) + return -EIO; + return port_attr->store(p, port_attr, buf, count); +} + +static const struct sysfs_ops port_sysfs_ops = { + .show = port_attr_show, + .store = port_attr_store +}; + +static ssize_t gid_attr_show(struct kobject *kobj, + struct attribute *attr, char *buf) +{ + struct port_attribute *port_attr = + container_of(attr, struct port_attribute, attr); + struct ib_port *p = container_of(kobj, struct gid_attr_group, + kobj)->port; + + if (!port_attr->show) + return -EIO; + + return port_attr->show(p, port_attr, buf); +} + +static const struct sysfs_ops gid_attr_sysfs_ops = { + .show = gid_attr_show +}; + +static ssize_t state_show(struct ib_port *p, struct port_attribute *unused, + char *buf) +{ + struct ib_port_attr attr; + ssize_t ret; + + static const char *state_name[] = { + [IB_PORT_NOP] = "NOP", + [IB_PORT_DOWN] = "DOWN", + [IB_PORT_INIT] = "INIT", + [IB_PORT_ARMED] = "ARMED", + [IB_PORT_ACTIVE] = "ACTIVE", + [IB_PORT_ACTIVE_DEFER] = "ACTIVE_DEFER" + }; + + ret = ib_query_port(p->ibdev, p->port_num, &attr); + if (ret) + return ret; + + return sprintf(buf, "%d: %s\n", attr.state, + attr.state >= 0 && attr.state < ARRAY_SIZE(state_name) ? + state_name[attr.state] : "UNKNOWN"); +} + +static ssize_t lid_show(struct ib_port *p, struct port_attribute *unused, + char *buf) +{ + struct ib_port_attr attr; + ssize_t ret; + + ret = ib_query_port(p->ibdev, p->port_num, &attr); + if (ret) + return ret; + + return sprintf(buf, "0x%x\n", attr.lid); +} + +static ssize_t lid_mask_count_show(struct ib_port *p, + struct port_attribute *unused, + char *buf) +{ + struct ib_port_attr attr; + ssize_t ret; + + ret = ib_query_port(p->ibdev, p->port_num, &attr); + if (ret) + return ret; + + return sprintf(buf, "%d\n", attr.lmc); +} + +static ssize_t sm_lid_show(struct ib_port *p, struct port_attribute *unused, + char *buf) +{ + struct ib_port_attr attr; + ssize_t ret; + + ret = ib_query_port(p->ibdev, p->port_num, &attr); + if (ret) + return ret; + + return sprintf(buf, "0x%x\n", attr.sm_lid); +} + +static ssize_t sm_sl_show(struct ib_port *p, struct port_attribute *unused, + char *buf) +{ + struct ib_port_attr attr; + ssize_t ret; + + ret = ib_query_port(p->ibdev, p->port_num, &attr); + if (ret) + return ret; + + return sprintf(buf, "%d\n", attr.sm_sl); +} + +static ssize_t cap_mask_show(struct ib_port *p, struct port_attribute *unused, + char *buf) +{ + struct ib_port_attr attr; + ssize_t ret; + + ret = ib_query_port(p->ibdev, p->port_num, &attr); + if (ret) + return ret; + + return sprintf(buf, "0x%08x\n", attr.port_cap_flags); +} + +static ssize_t rate_show(struct ib_port *p, struct port_attribute *unused, + char *buf) +{ + struct ib_port_attr attr; + char *speed = ""; + int rate; /* in deci-Gb/sec */ + ssize_t ret; + + ret = ib_query_port(p->ibdev, p->port_num, &attr); + if (ret) + return ret; + + switch (attr.active_speed) { + case IB_SPEED_DDR: + speed = " DDR"; + rate = 50; + break; + case IB_SPEED_QDR: + speed = " QDR"; + rate = 100; + break; + case IB_SPEED_FDR10: + speed = " FDR10"; + rate = 100; + break; + case IB_SPEED_FDR: + speed = " FDR"; + rate = 140; + break; + case IB_SPEED_EDR: + speed = " EDR"; + rate = 250; + break; + case IB_SPEED_HDR: + speed = " HDR"; + rate = 500; + break; + case IB_SPEED_SDR: + default: /* default to SDR for invalid rates */ + speed = " SDR"; + rate = 25; + break; + } + + rate *= ib_width_enum_to_int(attr.active_width); + if (rate < 0) + return -EINVAL; + + return sprintf(buf, "%d%s Gb/sec (%dX%s)\n", + rate / 10, rate % 10 ? ".5" : "", + ib_width_enum_to_int(attr.active_width), speed); +} + +static const char *phys_state_to_str(enum ib_port_phys_state phys_state) +{ + static const char * phys_state_str[] = { + "<unknown>", + "Sleep", + "Polling", + "Disabled", + "PortConfigurationTraining", + "LinkUp", + "LinkErrorRecovery", + "Phy Test", + }; + + if (phys_state < ARRAY_SIZE(phys_state_str)) + return phys_state_str[phys_state]; + return "<unknown>"; +} + +static ssize_t phys_state_show(struct ib_port *p, struct port_attribute *unused, + char *buf) +{ + struct ib_port_attr attr; + + ssize_t ret; + + ret = ib_query_port(p->ibdev, p->port_num, &attr); + if (ret) + return ret; + + return sprintf(buf, "%d: %s\n", attr.phys_state, + phys_state_to_str(attr.phys_state)); +} + +static ssize_t link_layer_show(struct ib_port *p, struct port_attribute *unused, + char *buf) +{ + switch (rdma_port_get_link_layer(p->ibdev, p->port_num)) { + case IB_LINK_LAYER_INFINIBAND: + return sprintf(buf, "%s\n", "InfiniBand"); + case IB_LINK_LAYER_ETHERNET: + return sprintf(buf, "%s\n", "Ethernet"); + default: + return sprintf(buf, "%s\n", "Unknown"); + } +} + +static PORT_ATTR_RO(state); +static PORT_ATTR_RO(lid); +static PORT_ATTR_RO(lid_mask_count); +static PORT_ATTR_RO(sm_lid); +static PORT_ATTR_RO(sm_sl); +static PORT_ATTR_RO(cap_mask); +static PORT_ATTR_RO(rate); +static PORT_ATTR_RO(phys_state); +static PORT_ATTR_RO(link_layer); + +static struct attribute *port_default_attrs[] = { + &port_attr_state.attr, + &port_attr_lid.attr, + &port_attr_lid_mask_count.attr, + &port_attr_sm_lid.attr, + &port_attr_sm_sl.attr, + &port_attr_cap_mask.attr, + &port_attr_rate.attr, + &port_attr_phys_state.attr, + &port_attr_link_layer.attr, + NULL +}; + +static size_t print_ndev(const struct ib_gid_attr *gid_attr, char *buf) +{ + struct net_device *ndev; + size_t ret = -EINVAL; + + rcu_read_lock(); + ndev = rcu_dereference(gid_attr->ndev); + if (ndev) + ret = sprintf(buf, "%s\n", ndev->name); + rcu_read_unlock(); + return ret; +} + +static size_t print_gid_type(const struct ib_gid_attr *gid_attr, char *buf) +{ + return sprintf(buf, "%s\n", ib_cache_gid_type_str(gid_attr->gid_type)); +} + +static ssize_t _show_port_gid_attr( + struct ib_port *p, struct port_attribute *attr, char *buf, + size_t (*print)(const struct ib_gid_attr *gid_attr, char *buf)) +{ + struct port_table_attribute *tab_attr = + container_of(attr, struct port_table_attribute, attr); + const struct ib_gid_attr *gid_attr; + ssize_t ret; + + gid_attr = rdma_get_gid_attr(p->ibdev, p->port_num, tab_attr->index); + if (IS_ERR(gid_attr)) + /* -EINVAL is returned for user space compatibility reasons. */ + return -EINVAL; + + ret = print(gid_attr, buf); + rdma_put_gid_attr(gid_attr); + return ret; +} + +static ssize_t show_port_gid(struct ib_port *p, struct port_attribute *attr, + char *buf) +{ + struct port_table_attribute *tab_attr = + container_of(attr, struct port_table_attribute, attr); + const struct ib_gid_attr *gid_attr; + ssize_t ret; + + gid_attr = rdma_get_gid_attr(p->ibdev, p->port_num, tab_attr->index); + if (IS_ERR(gid_attr)) { + const union ib_gid zgid = {}; + + /* If reading GID fails, it is likely due to GID entry being + * empty (invalid) or reserved GID in the table. User space + * expects to read GID table entries as long as it given index + * is within GID table size. Administrative/debugging tool + * fails to query rest of the GID entries if it hits error + * while querying a GID of the given index. To avoid user + * space throwing such error on fail to read gid, return zero + * GID as before. This maintains backward compatibility. + */ + return sprintf(buf, "%pI6\n", zgid.raw); + } + + ret = sprintf(buf, "%pI6\n", gid_attr->gid.raw); + rdma_put_gid_attr(gid_attr); + return ret; +} + +static ssize_t show_port_gid_attr_ndev(struct ib_port *p, + struct port_attribute *attr, char *buf) +{ + return _show_port_gid_attr(p, attr, buf, print_ndev); +} + +static ssize_t show_port_gid_attr_gid_type(struct ib_port *p, + struct port_attribute *attr, + char *buf) +{ + return _show_port_gid_attr(p, attr, buf, print_gid_type); +} + +static ssize_t show_port_pkey(struct ib_port *p, struct port_attribute *attr, + char *buf) +{ + struct port_table_attribute *tab_attr = + container_of(attr, struct port_table_attribute, attr); + u16 pkey; + ssize_t ret; + + ret = ib_query_pkey(p->ibdev, p->port_num, tab_attr->index, &pkey); + if (ret) + return ret; + + return sprintf(buf, "0x%04x\n", pkey); +} + +#define PORT_PMA_ATTR(_name, _counter, _width, _offset) \ +struct port_table_attribute port_pma_attr_##_name = { \ + .attr = __ATTR(_name, S_IRUGO, show_pma_counter, NULL), \ + .index = (_offset) | ((_width) << 16) | ((_counter) << 24), \ + .attr_id = IB_PMA_PORT_COUNTERS , \ +} + +#define PORT_PMA_ATTR_EXT(_name, _width, _offset) \ +struct port_table_attribute port_pma_attr_ext_##_name = { \ + .attr = __ATTR(_name, S_IRUGO, show_pma_counter, NULL), \ + .index = (_offset) | ((_width) << 16), \ + .attr_id = IB_PMA_PORT_COUNTERS_EXT , \ +} + +/* + * Get a Perfmgmt MAD block of data. + * Returns error code or the number of bytes retrieved. + */ +static int get_perf_mad(struct ib_device *dev, int port_num, __be16 attr, + void *data, int offset, size_t size) +{ + struct ib_mad *in_mad; + struct ib_mad *out_mad; + size_t mad_size = sizeof(*out_mad); + u16 out_mad_pkey_index = 0; + ssize_t ret; + + if (!dev->ops.process_mad) + return -ENOSYS; + + in_mad = kzalloc(sizeof(*in_mad), GFP_KERNEL); + out_mad = kzalloc(sizeof(*out_mad), GFP_KERNEL); + if (!in_mad || !out_mad) { + ret = -ENOMEM; + goto out; + } + + in_mad->mad_hdr.base_version = 1; + in_mad->mad_hdr.mgmt_class = IB_MGMT_CLASS_PERF_MGMT; + in_mad->mad_hdr.class_version = 1; + in_mad->mad_hdr.method = IB_MGMT_METHOD_GET; + in_mad->mad_hdr.attr_id = attr; + + if (attr != IB_PMA_CLASS_PORT_INFO) + in_mad->data[41] = port_num; /* PortSelect field */ + + if ((dev->ops.process_mad(dev, IB_MAD_IGNORE_MKEY, port_num, NULL, NULL, + in_mad, out_mad, &mad_size, + &out_mad_pkey_index) & + (IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_REPLY)) != + (IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_REPLY)) { + ret = -EINVAL; + goto out; + } + memcpy(data, out_mad->data + offset, size); + ret = size; +out: + kfree(in_mad); + kfree(out_mad); + return ret; +} + +static ssize_t show_pma_counter(struct ib_port *p, struct port_attribute *attr, + char *buf) +{ + struct port_table_attribute *tab_attr = + container_of(attr, struct port_table_attribute, attr); + int offset = tab_attr->index & 0xffff; + int width = (tab_attr->index >> 16) & 0xff; + ssize_t ret; + u8 data[8]; + + ret = get_perf_mad(p->ibdev, p->port_num, tab_attr->attr_id, &data, + 40 + offset / 8, sizeof(data)); + if (ret < 0) + return ret; + + switch (width) { + case 4: + ret = sprintf(buf, "%u\n", (*data >> + (4 - (offset % 8))) & 0xf); + break; + case 8: + ret = sprintf(buf, "%u\n", *data); + break; + case 16: + ret = sprintf(buf, "%u\n", + be16_to_cpup((__be16 *)data)); + break; + case 32: + ret = sprintf(buf, "%u\n", + be32_to_cpup((__be32 *)data)); + break; + case 64: + ret = sprintf(buf, "%llu\n", + be64_to_cpup((__be64 *)data)); + break; + + default: + ret = 0; + } + + return ret; +} + +static PORT_PMA_ATTR(symbol_error , 0, 16, 32); +static PORT_PMA_ATTR(link_error_recovery , 1, 8, 48); +static PORT_PMA_ATTR(link_downed , 2, 8, 56); +static PORT_PMA_ATTR(port_rcv_errors , 3, 16, 64); +static PORT_PMA_ATTR(port_rcv_remote_physical_errors, 4, 16, 80); +static PORT_PMA_ATTR(port_rcv_switch_relay_errors , 5, 16, 96); +static PORT_PMA_ATTR(port_xmit_discards , 6, 16, 112); +static PORT_PMA_ATTR(port_xmit_constraint_errors , 7, 8, 128); +static PORT_PMA_ATTR(port_rcv_constraint_errors , 8, 8, 136); +static PORT_PMA_ATTR(local_link_integrity_errors , 9, 4, 152); +static PORT_PMA_ATTR(excessive_buffer_overrun_errors, 10, 4, 156); +static PORT_PMA_ATTR(VL15_dropped , 11, 16, 176); +static PORT_PMA_ATTR(port_xmit_data , 12, 32, 192); +static PORT_PMA_ATTR(port_rcv_data , 13, 32, 224); +static PORT_PMA_ATTR(port_xmit_packets , 14, 32, 256); +static PORT_PMA_ATTR(port_rcv_packets , 15, 32, 288); +static PORT_PMA_ATTR(port_xmit_wait , 0, 32, 320); + +/* + * Counters added by extended set + */ +static PORT_PMA_ATTR_EXT(port_xmit_data , 64, 64); +static PORT_PMA_ATTR_EXT(port_rcv_data , 64, 128); +static PORT_PMA_ATTR_EXT(port_xmit_packets , 64, 192); +static PORT_PMA_ATTR_EXT(port_rcv_packets , 64, 256); +static PORT_PMA_ATTR_EXT(unicast_xmit_packets , 64, 320); +static PORT_PMA_ATTR_EXT(unicast_rcv_packets , 64, 384); +static PORT_PMA_ATTR_EXT(multicast_xmit_packets , 64, 448); +static PORT_PMA_ATTR_EXT(multicast_rcv_packets , 64, 512); + +static struct attribute *pma_attrs[] = { + &port_pma_attr_symbol_error.attr.attr, + &port_pma_attr_link_error_recovery.attr.attr, + &port_pma_attr_link_downed.attr.attr, + &port_pma_attr_port_rcv_errors.attr.attr, + &port_pma_attr_port_rcv_remote_physical_errors.attr.attr, + &port_pma_attr_port_rcv_switch_relay_errors.attr.attr, + &port_pma_attr_port_xmit_discards.attr.attr, + &port_pma_attr_port_xmit_constraint_errors.attr.attr, + &port_pma_attr_port_rcv_constraint_errors.attr.attr, + &port_pma_attr_local_link_integrity_errors.attr.attr, + &port_pma_attr_excessive_buffer_overrun_errors.attr.attr, + &port_pma_attr_VL15_dropped.attr.attr, + &port_pma_attr_port_xmit_data.attr.attr, + &port_pma_attr_port_rcv_data.attr.attr, + &port_pma_attr_port_xmit_packets.attr.attr, + &port_pma_attr_port_rcv_packets.attr.attr, + &port_pma_attr_port_xmit_wait.attr.attr, + NULL +}; + +static struct attribute *pma_attrs_ext[] = { + &port_pma_attr_symbol_error.attr.attr, + &port_pma_attr_link_error_recovery.attr.attr, + &port_pma_attr_link_downed.attr.attr, + &port_pma_attr_port_rcv_errors.attr.attr, + &port_pma_attr_port_rcv_remote_physical_errors.attr.attr, + &port_pma_attr_port_rcv_switch_relay_errors.attr.attr, + &port_pma_attr_port_xmit_discards.attr.attr, + &port_pma_attr_port_xmit_constraint_errors.attr.attr, + &port_pma_attr_port_rcv_constraint_errors.attr.attr, + &port_pma_attr_local_link_integrity_errors.attr.attr, + &port_pma_attr_excessive_buffer_overrun_errors.attr.attr, + &port_pma_attr_VL15_dropped.attr.attr, + &port_pma_attr_ext_port_xmit_data.attr.attr, + &port_pma_attr_ext_port_rcv_data.attr.attr, + &port_pma_attr_ext_port_xmit_packets.attr.attr, + &port_pma_attr_port_xmit_wait.attr.attr, + &port_pma_attr_ext_port_rcv_packets.attr.attr, + &port_pma_attr_ext_unicast_rcv_packets.attr.attr, + &port_pma_attr_ext_unicast_xmit_packets.attr.attr, + &port_pma_attr_ext_multicast_rcv_packets.attr.attr, + &port_pma_attr_ext_multicast_xmit_packets.attr.attr, + NULL +}; + +static struct attribute *pma_attrs_noietf[] = { + &port_pma_attr_symbol_error.attr.attr, + &port_pma_attr_link_error_recovery.attr.attr, + &port_pma_attr_link_downed.attr.attr, + &port_pma_attr_port_rcv_errors.attr.attr, + &port_pma_attr_port_rcv_remote_physical_errors.attr.attr, + &port_pma_attr_port_rcv_switch_relay_errors.attr.attr, + &port_pma_attr_port_xmit_discards.attr.attr, + &port_pma_attr_port_xmit_constraint_errors.attr.attr, + &port_pma_attr_port_rcv_constraint_errors.attr.attr, + &port_pma_attr_local_link_integrity_errors.attr.attr, + &port_pma_attr_excessive_buffer_overrun_errors.attr.attr, + &port_pma_attr_VL15_dropped.attr.attr, + &port_pma_attr_ext_port_xmit_data.attr.attr, + &port_pma_attr_ext_port_rcv_data.attr.attr, + &port_pma_attr_ext_port_xmit_packets.attr.attr, + &port_pma_attr_ext_port_rcv_packets.attr.attr, + &port_pma_attr_port_xmit_wait.attr.attr, + NULL +}; + +static const struct attribute_group pma_group = { + .name = "counters", + .attrs = pma_attrs +}; + +static const struct attribute_group pma_group_ext = { + .name = "counters", + .attrs = pma_attrs_ext +}; + +static const struct attribute_group pma_group_noietf = { + .name = "counters", + .attrs = pma_attrs_noietf +}; + +static void ib_port_release(struct kobject *kobj) +{ + struct ib_port *p = container_of(kobj, struct ib_port, kobj); + struct attribute *a; + int i; + + if (p->gid_group.attrs) { + for (i = 0; (a = p->gid_group.attrs[i]); ++i) + kfree(a); + + kfree(p->gid_group.attrs); + } + + if (p->pkey_group) { + if (p->pkey_group->attrs) { + for (i = 0; (a = p->pkey_group->attrs[i]); ++i) + kfree(a); + + kfree(p->pkey_group->attrs); + } + + kfree(p->pkey_group); + p->pkey_group = NULL; + } + + kfree(p); +} + +static void ib_port_gid_attr_release(struct kobject *kobj) +{ + struct gid_attr_group *g = container_of(kobj, struct gid_attr_group, + kobj); + struct attribute *a; + int i; + + if (g->ndev.attrs) { + for (i = 0; (a = g->ndev.attrs[i]); ++i) + kfree(a); + + kfree(g->ndev.attrs); + } + + if (g->type.attrs) { + for (i = 0; (a = g->type.attrs[i]); ++i) + kfree(a); + + kfree(g->type.attrs); + } + + kfree(g); +} + +static struct kobj_type port_type = { + .release = ib_port_release, + .sysfs_ops = &port_sysfs_ops, + .default_attrs = port_default_attrs +}; + +static struct kobj_type gid_attr_type = { + .sysfs_ops = &gid_attr_sysfs_ops, + .release = ib_port_gid_attr_release +}; + +static struct attribute ** +alloc_group_attrs(ssize_t (*show)(struct ib_port *, + struct port_attribute *, char *buf), + int len) +{ + struct attribute **tab_attr; + struct port_table_attribute *element; + int i; + + tab_attr = kcalloc(1 + len, sizeof(struct attribute *), GFP_KERNEL); + if (!tab_attr) + return NULL; + + for (i = 0; i < len; i++) { + element = kzalloc(sizeof(struct port_table_attribute), + GFP_KERNEL); + if (!element) + goto err; + + if (snprintf(element->name, sizeof(element->name), + "%d", i) >= sizeof(element->name)) { + kfree(element); + goto err; + } + + element->attr.attr.name = element->name; + element->attr.attr.mode = S_IRUGO; + element->attr.show = show; + element->index = i; + sysfs_attr_init(&element->attr.attr); + + tab_attr[i] = &element->attr.attr; + } + + return tab_attr; + +err: + while (--i >= 0) + kfree(tab_attr[i]); + kfree(tab_attr); + return NULL; +} + +/* + * Figure out which counter table to use depending on + * the device capabilities. + */ +static const struct attribute_group *get_counter_table(struct ib_device *dev, + int port_num) +{ + struct ib_class_port_info cpi; + + if (get_perf_mad(dev, port_num, IB_PMA_CLASS_PORT_INFO, + &cpi, 40, sizeof(cpi)) >= 0) { + if (cpi.capability_mask & IB_PMA_CLASS_CAP_EXT_WIDTH) + /* We have extended counters */ + return &pma_group_ext; + + if (cpi.capability_mask & IB_PMA_CLASS_CAP_EXT_WIDTH_NOIETF) + /* But not the IETF ones */ + return &pma_group_noietf; + } + + /* Fall back to normal counters */ + return &pma_group; +} + +static int update_hw_stats(struct ib_device *dev, struct rdma_hw_stats *stats, + u8 port_num, int index) +{ + int ret; + + if (time_is_after_eq_jiffies(stats->timestamp + stats->lifespan)) + return 0; + ret = dev->ops.get_hw_stats(dev, stats, port_num, index); + if (ret < 0) + return ret; + if (ret == stats->num_counters) + stats->timestamp = jiffies; + + return 0; +} + +static ssize_t print_hw_stat(struct ib_device *dev, int port_num, + struct rdma_hw_stats *stats, int index, char *buf) +{ + u64 v = rdma_counter_get_hwstat_value(dev, port_num, index); + + return sprintf(buf, "%llu\n", stats->value[index] + v); +} + +static ssize_t show_hw_stats(struct kobject *kobj, struct attribute *attr, + char *buf) +{ + struct ib_device *dev; + struct ib_port *port; + struct hw_stats_attribute *hsa; + struct rdma_hw_stats *stats; + int ret; + + hsa = container_of(attr, struct hw_stats_attribute, attr); + if (!hsa->port_num) { + dev = container_of((struct device *)kobj, + struct ib_device, dev); + stats = dev->hw_stats; + } else { + port = container_of(kobj, struct ib_port, kobj); + dev = port->ibdev; + stats = port->hw_stats; + } + mutex_lock(&stats->lock); + ret = update_hw_stats(dev, stats, hsa->port_num, hsa->index); + if (ret) + goto unlock; + ret = print_hw_stat(dev, hsa->port_num, stats, hsa->index, buf); +unlock: + mutex_unlock(&stats->lock); + + return ret; +} + +static ssize_t show_stats_lifespan(struct kobject *kobj, + struct attribute *attr, + char *buf) +{ + struct hw_stats_attribute *hsa; + struct rdma_hw_stats *stats; + int msecs; + + hsa = container_of(attr, struct hw_stats_attribute, attr); + if (!hsa->port_num) { + struct ib_device *dev = container_of((struct device *)kobj, + struct ib_device, dev); + + stats = dev->hw_stats; + } else { + struct ib_port *p = container_of(kobj, struct ib_port, kobj); + + stats = p->hw_stats; + } + + mutex_lock(&stats->lock); + msecs = jiffies_to_msecs(stats->lifespan); + mutex_unlock(&stats->lock); + + return sprintf(buf, "%d\n", msecs); +} + +static ssize_t set_stats_lifespan(struct kobject *kobj, + struct attribute *attr, + const char *buf, size_t count) +{ + struct hw_stats_attribute *hsa; + struct rdma_hw_stats *stats; + int msecs; + int jiffies; + int ret; + + ret = kstrtoint(buf, 10, &msecs); + if (ret) + return ret; + if (msecs < 0 || msecs > 10000) + return -EINVAL; + jiffies = msecs_to_jiffies(msecs); + hsa = container_of(attr, struct hw_stats_attribute, attr); + if (!hsa->port_num) { + struct ib_device *dev = container_of((struct device *)kobj, + struct ib_device, dev); + + stats = dev->hw_stats; + } else { + struct ib_port *p = container_of(kobj, struct ib_port, kobj); + + stats = p->hw_stats; + } + + mutex_lock(&stats->lock); + stats->lifespan = jiffies; + mutex_unlock(&stats->lock); + + return count; +} + +static void free_hsag(struct kobject *kobj, struct attribute_group *attr_group) +{ + struct attribute **attr; + + sysfs_remove_group(kobj, attr_group); + + for (attr = attr_group->attrs; *attr; attr++) + kfree(*attr); + kfree(attr_group); +} + +static struct attribute *alloc_hsa(int index, u8 port_num, const char *name) +{ + struct hw_stats_attribute *hsa; + + hsa = kmalloc(sizeof(*hsa), GFP_KERNEL); + if (!hsa) + return NULL; + + hsa->attr.name = (char *)name; + hsa->attr.mode = S_IRUGO; + hsa->show = show_hw_stats; + hsa->store = NULL; + hsa->index = index; + hsa->port_num = port_num; + + return &hsa->attr; +} + +static struct attribute *alloc_hsa_lifespan(char *name, u8 port_num) +{ + struct hw_stats_attribute *hsa; + + hsa = kmalloc(sizeof(*hsa), GFP_KERNEL); + if (!hsa) + return NULL; + + hsa->attr.name = name; + hsa->attr.mode = S_IWUSR | S_IRUGO; + hsa->show = show_stats_lifespan; + hsa->store = set_stats_lifespan; + hsa->index = 0; + hsa->port_num = port_num; + + return &hsa->attr; +} + +static void setup_hw_stats(struct ib_device *device, struct ib_port *port, + u8 port_num) +{ + struct attribute_group *hsag; + struct rdma_hw_stats *stats; + int i, ret; + + stats = device->ops.alloc_hw_stats(device, port_num); + + if (!stats) + return; + + if (!stats->names || stats->num_counters <= 0) + goto err_free_stats; + + /* + * Two extra attribue elements here, one for the lifespan entry and + * one to NULL terminate the list for the sysfs core code + */ + hsag = kzalloc(sizeof(*hsag) + + sizeof(void *) * (stats->num_counters + 2), + GFP_KERNEL); + if (!hsag) + goto err_free_stats; + + ret = device->ops.get_hw_stats(device, stats, port_num, + stats->num_counters); + if (ret != stats->num_counters) + goto err_free_hsag; + + stats->timestamp = jiffies; + + hsag->name = "hw_counters"; + hsag->attrs = (void *)hsag + sizeof(*hsag); + + for (i = 0; i < stats->num_counters; i++) { + hsag->attrs[i] = alloc_hsa(i, port_num, stats->names[i]); + if (!hsag->attrs[i]) + goto err; + sysfs_attr_init(hsag->attrs[i]); + } + + mutex_init(&stats->lock); + /* treat an error here as non-fatal */ + hsag->attrs[i] = alloc_hsa_lifespan("lifespan", port_num); + if (hsag->attrs[i]) + sysfs_attr_init(hsag->attrs[i]); + + if (port) { + struct kobject *kobj = &port->kobj; + ret = sysfs_create_group(kobj, hsag); + if (ret) + goto err; + port->hw_stats_ag = hsag; + port->hw_stats = stats; + if (device->port_data) + device->port_data[port_num].hw_stats = stats; + } else { + struct kobject *kobj = &device->dev.kobj; + ret = sysfs_create_group(kobj, hsag); + if (ret) + goto err; + device->hw_stats_ag = hsag; + device->hw_stats = stats; + } + + return; + +err: + for (; i >= 0; i--) + kfree(hsag->attrs[i]); +err_free_hsag: + kfree(hsag); +err_free_stats: + kfree(stats); + return; +} + +static int add_port(struct ib_core_device *coredev, int port_num) +{ + struct ib_device *device = rdma_device_to_ibdev(&coredev->dev); + bool is_full_dev = &device->coredev == coredev; + struct ib_port *p; + struct ib_port_attr attr; + int i; + int ret; + + ret = ib_query_port(device, port_num, &attr); + if (ret) + return ret; + + p = kzalloc(sizeof *p, GFP_KERNEL); + if (!p) + return -ENOMEM; + + p->ibdev = device; + p->port_num = port_num; + + ret = kobject_init_and_add(&p->kobj, &port_type, + coredev->ports_kobj, + "%d", port_num); + if (ret) { + goto err_put; + } + + p->gid_attr_group = kzalloc(sizeof(*p->gid_attr_group), GFP_KERNEL); + if (!p->gid_attr_group) { + ret = -ENOMEM; + goto err_put; + } + + p->gid_attr_group->port = p; + ret = kobject_init_and_add(&p->gid_attr_group->kobj, &gid_attr_type, + &p->kobj, "gid_attrs"); + if (ret) { + goto err_put_gid_attrs; + } + + if (device->ops.process_mad && is_full_dev) { + p->pma_table = get_counter_table(device, port_num); + ret = sysfs_create_group(&p->kobj, p->pma_table); + if (ret) + goto err_put_gid_attrs; + } + + p->gid_group.name = "gids"; + p->gid_group.attrs = alloc_group_attrs(show_port_gid, attr.gid_tbl_len); + if (!p->gid_group.attrs) { + ret = -ENOMEM; + goto err_remove_pma; + } + + ret = sysfs_create_group(&p->kobj, &p->gid_group); + if (ret) + goto err_free_gid; + + p->gid_attr_group->ndev.name = "ndevs"; + p->gid_attr_group->ndev.attrs = alloc_group_attrs(show_port_gid_attr_ndev, + attr.gid_tbl_len); + if (!p->gid_attr_group->ndev.attrs) { + ret = -ENOMEM; + goto err_remove_gid; + } + + ret = sysfs_create_group(&p->gid_attr_group->kobj, + &p->gid_attr_group->ndev); + if (ret) + goto err_free_gid_ndev; + + p->gid_attr_group->type.name = "types"; + p->gid_attr_group->type.attrs = alloc_group_attrs(show_port_gid_attr_gid_type, + attr.gid_tbl_len); + if (!p->gid_attr_group->type.attrs) { + ret = -ENOMEM; + goto err_remove_gid_ndev; + } + + ret = sysfs_create_group(&p->gid_attr_group->kobj, + &p->gid_attr_group->type); + if (ret) + goto err_free_gid_type; + + if (attr.pkey_tbl_len) { + p->pkey_group = kzalloc(sizeof(*p->pkey_group), GFP_KERNEL); + if (!p->pkey_group) { + ret = -ENOMEM; + goto err_remove_gid_type; + } + + p->pkey_group->name = "pkeys"; + p->pkey_group->attrs = alloc_group_attrs(show_port_pkey, + attr.pkey_tbl_len); + if (!p->pkey_group->attrs) { + ret = -ENOMEM; + goto err_free_pkey_group; + } + + ret = sysfs_create_group(&p->kobj, p->pkey_group); + if (ret) + goto err_free_pkey; + } + + + if (device->ops.init_port && is_full_dev) { + ret = device->ops.init_port(device, port_num, &p->kobj); + if (ret) + goto err_remove_pkey; + } + + /* + * If port == 0, it means hw_counters are per device and not per + * port, so holder should be device. Therefore skip per port conunter + * initialization. + */ + if (device->ops.alloc_hw_stats && port_num && is_full_dev) + setup_hw_stats(device, p, port_num); + + list_add_tail(&p->kobj.entry, &coredev->port_list); + + kobject_uevent(&p->kobj, KOBJ_ADD); + return 0; + +err_remove_pkey: + if (p->pkey_group) + sysfs_remove_group(&p->kobj, p->pkey_group); + +err_free_pkey: + if (p->pkey_group) { + for (i = 0; i < attr.pkey_tbl_len; ++i) + kfree(p->pkey_group->attrs[i]); + + kfree(p->pkey_group->attrs); + p->pkey_group->attrs = NULL; + } + +err_free_pkey_group: + kfree(p->pkey_group); + +err_remove_gid_type: + sysfs_remove_group(&p->gid_attr_group->kobj, + &p->gid_attr_group->type); + +err_free_gid_type: + for (i = 0; i < attr.gid_tbl_len; ++i) + kfree(p->gid_attr_group->type.attrs[i]); + + kfree(p->gid_attr_group->type.attrs); + p->gid_attr_group->type.attrs = NULL; + +err_remove_gid_ndev: + sysfs_remove_group(&p->gid_attr_group->kobj, + &p->gid_attr_group->ndev); + +err_free_gid_ndev: + for (i = 0; i < attr.gid_tbl_len; ++i) + kfree(p->gid_attr_group->ndev.attrs[i]); + + kfree(p->gid_attr_group->ndev.attrs); + p->gid_attr_group->ndev.attrs = NULL; + +err_remove_gid: + sysfs_remove_group(&p->kobj, &p->gid_group); + +err_free_gid: + for (i = 0; i < attr.gid_tbl_len; ++i) + kfree(p->gid_group.attrs[i]); + + kfree(p->gid_group.attrs); + p->gid_group.attrs = NULL; + +err_remove_pma: + if (p->pma_table) + sysfs_remove_group(&p->kobj, p->pma_table); + +err_put_gid_attrs: + kobject_put(&p->gid_attr_group->kobj); + +err_put: + kobject_put(&p->kobj); + return ret; +} + +static ssize_t node_type_show(struct device *device, + struct device_attribute *attr, char *buf) +{ + struct ib_device *dev = rdma_device_to_ibdev(device); + + switch (dev->node_type) { + case RDMA_NODE_IB_CA: return sprintf(buf, "%d: CA\n", dev->node_type); + case RDMA_NODE_RNIC: return sprintf(buf, "%d: RNIC\n", dev->node_type); + case RDMA_NODE_USNIC: return sprintf(buf, "%d: usNIC\n", dev->node_type); + case RDMA_NODE_USNIC_UDP: return sprintf(buf, "%d: usNIC UDP\n", dev->node_type); + case RDMA_NODE_UNSPECIFIED: return sprintf(buf, "%d: unspecified\n", dev->node_type); + case RDMA_NODE_IB_SWITCH: return sprintf(buf, "%d: switch\n", dev->node_type); + case RDMA_NODE_IB_ROUTER: return sprintf(buf, "%d: router\n", dev->node_type); + default: return sprintf(buf, "%d: <unknown>\n", dev->node_type); + } +} +static DEVICE_ATTR_RO(node_type); + +static ssize_t sys_image_guid_show(struct device *device, + struct device_attribute *dev_attr, char *buf) +{ + struct ib_device *dev = rdma_device_to_ibdev(device); + + return sprintf(buf, "%04x:%04x:%04x:%04x\n", + be16_to_cpu(((__be16 *) &dev->attrs.sys_image_guid)[0]), + be16_to_cpu(((__be16 *) &dev->attrs.sys_image_guid)[1]), + be16_to_cpu(((__be16 *) &dev->attrs.sys_image_guid)[2]), + be16_to_cpu(((__be16 *) &dev->attrs.sys_image_guid)[3])); +} +static DEVICE_ATTR_RO(sys_image_guid); + +static ssize_t node_guid_show(struct device *device, + struct device_attribute *attr, char *buf) +{ + struct ib_device *dev = rdma_device_to_ibdev(device); + + return sprintf(buf, "%04x:%04x:%04x:%04x\n", + be16_to_cpu(((__be16 *) &dev->node_guid)[0]), + be16_to_cpu(((__be16 *) &dev->node_guid)[1]), + be16_to_cpu(((__be16 *) &dev->node_guid)[2]), + be16_to_cpu(((__be16 *) &dev->node_guid)[3])); +} +static DEVICE_ATTR_RO(node_guid); + +static ssize_t node_desc_show(struct device *device, + struct device_attribute *attr, char *buf) +{ + struct ib_device *dev = rdma_device_to_ibdev(device); + + return sprintf(buf, "%.64s\n", dev->node_desc); +} + +static ssize_t node_desc_store(struct device *device, + struct device_attribute *attr, + const char *buf, size_t count) +{ + struct ib_device *dev = rdma_device_to_ibdev(device); + struct ib_device_modify desc = {}; + int ret; + + if (!dev->ops.modify_device) + return -EOPNOTSUPP; + + memcpy(desc.node_desc, buf, min_t(int, count, IB_DEVICE_NODE_DESC_MAX)); + ret = ib_modify_device(dev, IB_DEVICE_MODIFY_NODE_DESC, &desc); + if (ret) + return ret; + + return count; +} +static DEVICE_ATTR_RW(node_desc); + +static ssize_t fw_ver_show(struct device *device, struct device_attribute *attr, + char *buf) +{ + struct ib_device *dev = rdma_device_to_ibdev(device); + + ib_get_device_fw_str(dev, buf); + strlcat(buf, "\n", IB_FW_VERSION_NAME_MAX); + return strlen(buf); +} +static DEVICE_ATTR_RO(fw_ver); + +static struct attribute *ib_dev_attrs[] = { + &dev_attr_node_type.attr, + &dev_attr_node_guid.attr, + &dev_attr_sys_image_guid.attr, + &dev_attr_fw_ver.attr, + &dev_attr_node_desc.attr, + NULL, +}; + +const struct attribute_group ib_dev_attr_group = { + .attrs = ib_dev_attrs, +}; + +void ib_free_port_attrs(struct ib_core_device *coredev) +{ + struct ib_device *device = rdma_device_to_ibdev(&coredev->dev); + bool is_full_dev = &device->coredev == coredev; + struct kobject *p, *t; + + list_for_each_entry_safe(p, t, &coredev->port_list, entry) { + struct ib_port *port = container_of(p, struct ib_port, kobj); + + list_del(&p->entry); + if (port->hw_stats_ag) + free_hsag(&port->kobj, port->hw_stats_ag); + kfree(port->hw_stats); + if (device->port_data && is_full_dev) + device->port_data[port->port_num].hw_stats = NULL; + + if (port->pma_table) + sysfs_remove_group(p, port->pma_table); + if (port->pkey_group) + sysfs_remove_group(p, port->pkey_group); + sysfs_remove_group(p, &port->gid_group); + sysfs_remove_group(&port->gid_attr_group->kobj, + &port->gid_attr_group->ndev); + sysfs_remove_group(&port->gid_attr_group->kobj, + &port->gid_attr_group->type); + kobject_put(&port->gid_attr_group->kobj); + kobject_put(p); + } + + kobject_put(coredev->ports_kobj); +} + +int ib_setup_port_attrs(struct ib_core_device *coredev) +{ + struct ib_device *device = rdma_device_to_ibdev(&coredev->dev); + unsigned int port; + int ret; + + coredev->ports_kobj = kobject_create_and_add("ports", + &coredev->dev.kobj); + if (!coredev->ports_kobj) + return -ENOMEM; + + rdma_for_each_port (device, port) { + ret = add_port(coredev, port); + if (ret) + goto err_put; + } + + return 0; + +err_put: + ib_free_port_attrs(coredev); + return ret; +} + +int ib_device_register_sysfs(struct ib_device *device) +{ + int ret; + + ret = ib_setup_port_attrs(&device->coredev); + if (ret) + return ret; + + if (device->ops.alloc_hw_stats) + setup_hw_stats(device, NULL, 0); + + return 0; +} + +void ib_device_unregister_sysfs(struct ib_device *device) +{ + if (device->hw_stats_ag) + free_hsag(&device->dev.kobj, device->hw_stats_ag); + kfree(device->hw_stats); + + ib_free_port_attrs(&device->coredev); +} + +/** + * ib_port_register_module_stat - add module counters under relevant port + * of IB device. + * + * @device: IB device to add counters + * @port_num: valid port number + * @kobj: pointer to the kobject to initialize + * @ktype: pointer to the ktype for this kobject. + * @name: the name of the kobject + */ +int ib_port_register_module_stat(struct ib_device *device, u8 port_num, + struct kobject *kobj, struct kobj_type *ktype, + const char *name) +{ + struct kobject *p, *t; + int ret; + + list_for_each_entry_safe(p, t, &device->coredev.port_list, entry) { + struct ib_port *port = container_of(p, struct ib_port, kobj); + + if (port->port_num != port_num) + continue; + + ret = kobject_init_and_add(kobj, ktype, &port->kobj, "%s", + name); + if (ret) { + kobject_put(kobj); + return ret; + } + } + + return 0; +} +EXPORT_SYMBOL(ib_port_register_module_stat); + +/** + * ib_port_unregister_module_stat - release module counters + * @kobj: pointer to the kobject to release + */ +void ib_port_unregister_module_stat(struct kobject *kobj) +{ + kobject_put(kobj); +} +EXPORT_SYMBOL(ib_port_unregister_module_stat); diff --git a/drivers/infiniband/core/trace.c b/drivers/infiniband/core/trace.c new file mode 100644 index 000000000..31e7860d3 --- /dev/null +++ b/drivers/infiniband/core/trace.c @@ -0,0 +1,12 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * Trace points for core RDMA functions. + * + * Author: Chuck Lever <chuck.lever@oracle.com> + * + * Copyright (c) 2019, Oracle and/or its affiliates. All rights reserved. + */ + +#define CREATE_TRACE_POINTS + +#include <trace/events/rdma_core.h> diff --git a/drivers/infiniband/core/ucma.c b/drivers/infiniband/core/ucma.c new file mode 100644 index 000000000..d12018c4c --- /dev/null +++ b/drivers/infiniband/core/ucma.c @@ -0,0 +1,1897 @@ +/* + * Copyright (c) 2005-2006 Intel Corporation. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include <linux/completion.h> +#include <linux/file.h> +#include <linux/mutex.h> +#include <linux/poll.h> +#include <linux/sched.h> +#include <linux/idr.h> +#include <linux/in.h> +#include <linux/in6.h> +#include <linux/miscdevice.h> +#include <linux/slab.h> +#include <linux/sysctl.h> +#include <linux/module.h> +#include <linux/nsproxy.h> + +#include <linux/nospec.h> + +#include <rdma/rdma_user_cm.h> +#include <rdma/ib_marshall.h> +#include <rdma/rdma_cm.h> +#include <rdma/rdma_cm_ib.h> +#include <rdma/ib_addr.h> +#include <rdma/ib.h> +#include <rdma/ib_cm.h> +#include <rdma/rdma_netlink.h> +#include "core_priv.h" + +MODULE_AUTHOR("Sean Hefty"); +MODULE_DESCRIPTION("RDMA Userspace Connection Manager Access"); +MODULE_LICENSE("Dual BSD/GPL"); + +static unsigned int max_backlog = 1024; + +static struct ctl_table_header *ucma_ctl_table_hdr; +static struct ctl_table ucma_ctl_table[] = { + { + .procname = "max_backlog", + .data = &max_backlog, + .maxlen = sizeof max_backlog, + .mode = 0644, + .proc_handler = proc_dointvec, + }, + { } +}; + +struct ucma_file { + struct mutex mut; + struct file *filp; + struct list_head ctx_list; + struct list_head event_list; + wait_queue_head_t poll_wait; +}; + +struct ucma_context { + u32 id; + struct completion comp; + refcount_t ref; + int events_reported; + atomic_t backlog; + + struct ucma_file *file; + struct rdma_cm_id *cm_id; + struct mutex mutex; + u64 uid; + + struct list_head list; + struct list_head mc_list; + struct work_struct close_work; +}; + +struct ucma_multicast { + struct ucma_context *ctx; + u32 id; + int events_reported; + + u64 uid; + u8 join_state; + struct list_head list; + struct sockaddr_storage addr; +}; + +struct ucma_event { + struct ucma_context *ctx; + struct ucma_context *conn_req_ctx; + struct ucma_multicast *mc; + struct list_head list; + struct rdma_ucm_event_resp resp; +}; + +static DEFINE_XARRAY_ALLOC(ctx_table); +static DEFINE_XARRAY_ALLOC(multicast_table); + +static const struct file_operations ucma_fops; +static int ucma_destroy_private_ctx(struct ucma_context *ctx); + +static inline struct ucma_context *_ucma_find_context(int id, + struct ucma_file *file) +{ + struct ucma_context *ctx; + + ctx = xa_load(&ctx_table, id); + if (!ctx) + ctx = ERR_PTR(-ENOENT); + else if (ctx->file != file) + ctx = ERR_PTR(-EINVAL); + return ctx; +} + +static struct ucma_context *ucma_get_ctx(struct ucma_file *file, int id) +{ + struct ucma_context *ctx; + + xa_lock(&ctx_table); + ctx = _ucma_find_context(id, file); + if (!IS_ERR(ctx)) + if (!refcount_inc_not_zero(&ctx->ref)) + ctx = ERR_PTR(-ENXIO); + xa_unlock(&ctx_table); + return ctx; +} + +static void ucma_put_ctx(struct ucma_context *ctx) +{ + if (refcount_dec_and_test(&ctx->ref)) + complete(&ctx->comp); +} + +/* + * Same as ucm_get_ctx but requires that ->cm_id->device is valid, eg that the + * CM_ID is bound. + */ +static struct ucma_context *ucma_get_ctx_dev(struct ucma_file *file, int id) +{ + struct ucma_context *ctx = ucma_get_ctx(file, id); + + if (IS_ERR(ctx)) + return ctx; + if (!ctx->cm_id->device) { + ucma_put_ctx(ctx); + return ERR_PTR(-EINVAL); + } + return ctx; +} + +static void ucma_close_id(struct work_struct *work) +{ + struct ucma_context *ctx = container_of(work, struct ucma_context, close_work); + + /* once all inflight tasks are finished, we close all underlying + * resources. The context is still alive till its explicit destryoing + * by its creator. This puts back the xarray's reference. + */ + ucma_put_ctx(ctx); + wait_for_completion(&ctx->comp); + /* No new events will be generated after destroying the id. */ + rdma_destroy_id(ctx->cm_id); + + /* Reading the cm_id without holding a positive ref is not allowed */ + ctx->cm_id = NULL; +} + +static struct ucma_context *ucma_alloc_ctx(struct ucma_file *file) +{ + struct ucma_context *ctx; + + ctx = kzalloc(sizeof(*ctx), GFP_KERNEL); + if (!ctx) + return NULL; + + INIT_WORK(&ctx->close_work, ucma_close_id); + init_completion(&ctx->comp); + INIT_LIST_HEAD(&ctx->mc_list); + /* So list_del() will work if we don't do ucma_finish_ctx() */ + INIT_LIST_HEAD(&ctx->list); + ctx->file = file; + mutex_init(&ctx->mutex); + + if (xa_alloc(&ctx_table, &ctx->id, NULL, xa_limit_32b, GFP_KERNEL)) { + kfree(ctx); + return NULL; + } + return ctx; +} + +static void ucma_set_ctx_cm_id(struct ucma_context *ctx, + struct rdma_cm_id *cm_id) +{ + refcount_set(&ctx->ref, 1); + ctx->cm_id = cm_id; +} + +static void ucma_finish_ctx(struct ucma_context *ctx) +{ + lockdep_assert_held(&ctx->file->mut); + list_add_tail(&ctx->list, &ctx->file->ctx_list); + xa_store(&ctx_table, ctx->id, ctx, GFP_KERNEL); +} + +static void ucma_copy_conn_event(struct rdma_ucm_conn_param *dst, + struct rdma_conn_param *src) +{ + if (src->private_data_len) + memcpy(dst->private_data, src->private_data, + src->private_data_len); + dst->private_data_len = src->private_data_len; + dst->responder_resources =src->responder_resources; + dst->initiator_depth = src->initiator_depth; + dst->flow_control = src->flow_control; + dst->retry_count = src->retry_count; + dst->rnr_retry_count = src->rnr_retry_count; + dst->srq = src->srq; + dst->qp_num = src->qp_num; +} + +static void ucma_copy_ud_event(struct ib_device *device, + struct rdma_ucm_ud_param *dst, + struct rdma_ud_param *src) +{ + if (src->private_data_len) + memcpy(dst->private_data, src->private_data, + src->private_data_len); + dst->private_data_len = src->private_data_len; + ib_copy_ah_attr_to_user(device, &dst->ah_attr, &src->ah_attr); + dst->qp_num = src->qp_num; + dst->qkey = src->qkey; +} + +static struct ucma_event *ucma_create_uevent(struct ucma_context *ctx, + struct rdma_cm_event *event) +{ + struct ucma_event *uevent; + + uevent = kzalloc(sizeof(*uevent), GFP_KERNEL); + if (!uevent) + return NULL; + + uevent->ctx = ctx; + switch (event->event) { + case RDMA_CM_EVENT_MULTICAST_JOIN: + case RDMA_CM_EVENT_MULTICAST_ERROR: + uevent->mc = (struct ucma_multicast *) + event->param.ud.private_data; + uevent->resp.uid = uevent->mc->uid; + uevent->resp.id = uevent->mc->id; + break; + default: + uevent->resp.uid = ctx->uid; + uevent->resp.id = ctx->id; + break; + } + uevent->resp.event = event->event; + uevent->resp.status = event->status; + if (ctx->cm_id->qp_type == IB_QPT_UD) + ucma_copy_ud_event(ctx->cm_id->device, &uevent->resp.param.ud, + &event->param.ud); + else + ucma_copy_conn_event(&uevent->resp.param.conn, + &event->param.conn); + + uevent->resp.ece.vendor_id = event->ece.vendor_id; + uevent->resp.ece.attr_mod = event->ece.attr_mod; + return uevent; +} + +static int ucma_connect_event_handler(struct rdma_cm_id *cm_id, + struct rdma_cm_event *event) +{ + struct ucma_context *listen_ctx = cm_id->context; + struct ucma_context *ctx; + struct ucma_event *uevent; + + if (!atomic_add_unless(&listen_ctx->backlog, -1, 0)) + return -ENOMEM; + ctx = ucma_alloc_ctx(listen_ctx->file); + if (!ctx) + goto err_backlog; + ucma_set_ctx_cm_id(ctx, cm_id); + + uevent = ucma_create_uevent(listen_ctx, event); + if (!uevent) + goto err_alloc; + uevent->conn_req_ctx = ctx; + uevent->resp.id = ctx->id; + + ctx->cm_id->context = ctx; + + mutex_lock(&ctx->file->mut); + ucma_finish_ctx(ctx); + list_add_tail(&uevent->list, &ctx->file->event_list); + mutex_unlock(&ctx->file->mut); + wake_up_interruptible(&ctx->file->poll_wait); + return 0; + +err_alloc: + ucma_destroy_private_ctx(ctx); +err_backlog: + atomic_inc(&listen_ctx->backlog); + /* Returning error causes the new ID to be destroyed */ + return -ENOMEM; +} + +static int ucma_event_handler(struct rdma_cm_id *cm_id, + struct rdma_cm_event *event) +{ + struct ucma_event *uevent; + struct ucma_context *ctx = cm_id->context; + + if (event->event == RDMA_CM_EVENT_CONNECT_REQUEST) + return ucma_connect_event_handler(cm_id, event); + + /* + * We ignore events for new connections until userspace has set their + * context. This can only happen if an error occurs on a new connection + * before the user accepts it. This is okay, since the accept will just + * fail later. However, we do need to release the underlying HW + * resources in case of a device removal event. + */ + if (ctx->uid) { + uevent = ucma_create_uevent(ctx, event); + if (!uevent) + return 0; + + mutex_lock(&ctx->file->mut); + list_add_tail(&uevent->list, &ctx->file->event_list); + mutex_unlock(&ctx->file->mut); + wake_up_interruptible(&ctx->file->poll_wait); + } + + if (event->event == RDMA_CM_EVENT_DEVICE_REMOVAL) { + xa_lock(&ctx_table); + if (xa_load(&ctx_table, ctx->id) == ctx) + queue_work(system_unbound_wq, &ctx->close_work); + xa_unlock(&ctx_table); + } + return 0; +} + +static ssize_t ucma_get_event(struct ucma_file *file, const char __user *inbuf, + int in_len, int out_len) +{ + struct rdma_ucm_get_event cmd; + struct ucma_event *uevent; + + /* + * Old 32 bit user space does not send the 4 byte padding in the + * reserved field. We don't care, allow it to keep working. + */ + if (out_len < sizeof(uevent->resp) - sizeof(uevent->resp.reserved) - + sizeof(uevent->resp.ece)) + return -ENOSPC; + + if (copy_from_user(&cmd, inbuf, sizeof(cmd))) + return -EFAULT; + + mutex_lock(&file->mut); + while (list_empty(&file->event_list)) { + mutex_unlock(&file->mut); + + if (file->filp->f_flags & O_NONBLOCK) + return -EAGAIN; + + if (wait_event_interruptible(file->poll_wait, + !list_empty(&file->event_list))) + return -ERESTARTSYS; + + mutex_lock(&file->mut); + } + + uevent = list_first_entry(&file->event_list, struct ucma_event, list); + + if (copy_to_user(u64_to_user_ptr(cmd.response), + &uevent->resp, + min_t(size_t, out_len, sizeof(uevent->resp)))) { + mutex_unlock(&file->mut); + return -EFAULT; + } + + list_del(&uevent->list); + uevent->ctx->events_reported++; + if (uevent->mc) + uevent->mc->events_reported++; + if (uevent->resp.event == RDMA_CM_EVENT_CONNECT_REQUEST) + atomic_inc(&uevent->ctx->backlog); + mutex_unlock(&file->mut); + + kfree(uevent); + return 0; +} + +static int ucma_get_qp_type(struct rdma_ucm_create_id *cmd, enum ib_qp_type *qp_type) +{ + switch (cmd->ps) { + case RDMA_PS_TCP: + *qp_type = IB_QPT_RC; + return 0; + case RDMA_PS_UDP: + case RDMA_PS_IPOIB: + *qp_type = IB_QPT_UD; + return 0; + case RDMA_PS_IB: + *qp_type = cmd->qp_type; + return 0; + default: + return -EINVAL; + } +} + +static ssize_t ucma_create_id(struct ucma_file *file, const char __user *inbuf, + int in_len, int out_len) +{ + struct rdma_ucm_create_id cmd; + struct rdma_ucm_create_id_resp resp; + struct ucma_context *ctx; + struct rdma_cm_id *cm_id; + enum ib_qp_type qp_type; + int ret; + + if (out_len < sizeof(resp)) + return -ENOSPC; + + if (copy_from_user(&cmd, inbuf, sizeof(cmd))) + return -EFAULT; + + ret = ucma_get_qp_type(&cmd, &qp_type); + if (ret) + return ret; + + ctx = ucma_alloc_ctx(file); + if (!ctx) + return -ENOMEM; + + ctx->uid = cmd.uid; + cm_id = rdma_create_user_id(ucma_event_handler, ctx, cmd.ps, qp_type); + if (IS_ERR(cm_id)) { + ret = PTR_ERR(cm_id); + goto err1; + } + ucma_set_ctx_cm_id(ctx, cm_id); + + resp.id = ctx->id; + if (copy_to_user(u64_to_user_ptr(cmd.response), + &resp, sizeof(resp))) { + ucma_destroy_private_ctx(ctx); + return -EFAULT; + } + + mutex_lock(&file->mut); + ucma_finish_ctx(ctx); + mutex_unlock(&file->mut); + return 0; + +err1: + ucma_destroy_private_ctx(ctx); + return ret; +} + +static void ucma_cleanup_multicast(struct ucma_context *ctx) +{ + struct ucma_multicast *mc, *tmp; + + xa_lock(&multicast_table); + list_for_each_entry_safe(mc, tmp, &ctx->mc_list, list) { + list_del(&mc->list); + /* + * At this point mc->ctx->ref is 0 so the mc cannot leave the + * lock on the reader and this is enough serialization + */ + __xa_erase(&multicast_table, mc->id); + kfree(mc); + } + xa_unlock(&multicast_table); +} + +static void ucma_cleanup_mc_events(struct ucma_multicast *mc) +{ + struct ucma_event *uevent, *tmp; + + rdma_lock_handler(mc->ctx->cm_id); + mutex_lock(&mc->ctx->file->mut); + list_for_each_entry_safe(uevent, tmp, &mc->ctx->file->event_list, list) { + if (uevent->mc != mc) + continue; + + list_del(&uevent->list); + kfree(uevent); + } + mutex_unlock(&mc->ctx->file->mut); + rdma_unlock_handler(mc->ctx->cm_id); +} + +static int ucma_cleanup_ctx_events(struct ucma_context *ctx) +{ + int events_reported; + struct ucma_event *uevent, *tmp; + LIST_HEAD(list); + + /* Cleanup events not yet reported to the user.*/ + mutex_lock(&ctx->file->mut); + list_for_each_entry_safe(uevent, tmp, &ctx->file->event_list, list) { + if (uevent->ctx != ctx) + continue; + + if (uevent->resp.event == RDMA_CM_EVENT_CONNECT_REQUEST && + xa_cmpxchg(&ctx_table, uevent->conn_req_ctx->id, + uevent->conn_req_ctx, XA_ZERO_ENTRY, + GFP_KERNEL) == uevent->conn_req_ctx) { + list_move_tail(&uevent->list, &list); + continue; + } + list_del(&uevent->list); + kfree(uevent); + } + list_del(&ctx->list); + events_reported = ctx->events_reported; + mutex_unlock(&ctx->file->mut); + + /* + * If this was a listening ID then any connections spawned from it that + * have not been delivered to userspace are cleaned up too. Must be done + * outside any locks. + */ + list_for_each_entry_safe(uevent, tmp, &list, list) { + ucma_destroy_private_ctx(uevent->conn_req_ctx); + kfree(uevent); + } + return events_reported; +} + +/* + * When this is called the xarray must have a XA_ZERO_ENTRY in the ctx->id (ie + * the ctx is not public to the user). This either because: + * - ucma_finish_ctx() hasn't been called + * - xa_cmpxchg() succeed to remove the entry (only one thread can succeed) + */ +static int ucma_destroy_private_ctx(struct ucma_context *ctx) +{ + int events_reported; + + /* + * Destroy the underlying cm_id. New work queuing is prevented now by + * the removal from the xarray. Once the work is cancled ref will either + * be 0 because the work ran to completion and consumed the ref from the + * xarray, or it will be positive because we still have the ref from the + * xarray. This can also be 0 in cases where cm_id was never set + */ + cancel_work_sync(&ctx->close_work); + if (refcount_read(&ctx->ref)) + ucma_close_id(&ctx->close_work); + + events_reported = ucma_cleanup_ctx_events(ctx); + ucma_cleanup_multicast(ctx); + + WARN_ON(xa_cmpxchg(&ctx_table, ctx->id, XA_ZERO_ENTRY, NULL, + GFP_KERNEL) != NULL); + mutex_destroy(&ctx->mutex); + kfree(ctx); + return events_reported; +} + +static ssize_t ucma_destroy_id(struct ucma_file *file, const char __user *inbuf, + int in_len, int out_len) +{ + struct rdma_ucm_destroy_id cmd; + struct rdma_ucm_destroy_id_resp resp; + struct ucma_context *ctx; + int ret = 0; + + if (out_len < sizeof(resp)) + return -ENOSPC; + + if (copy_from_user(&cmd, inbuf, sizeof(cmd))) + return -EFAULT; + + xa_lock(&ctx_table); + ctx = _ucma_find_context(cmd.id, file); + if (!IS_ERR(ctx)) { + if (__xa_cmpxchg(&ctx_table, ctx->id, ctx, XA_ZERO_ENTRY, + GFP_KERNEL) != ctx) + ctx = ERR_PTR(-ENOENT); + } + xa_unlock(&ctx_table); + + if (IS_ERR(ctx)) + return PTR_ERR(ctx); + + resp.events_reported = ucma_destroy_private_ctx(ctx); + if (copy_to_user(u64_to_user_ptr(cmd.response), + &resp, sizeof(resp))) + ret = -EFAULT; + + return ret; +} + +static ssize_t ucma_bind_ip(struct ucma_file *file, const char __user *inbuf, + int in_len, int out_len) +{ + struct rdma_ucm_bind_ip cmd; + struct ucma_context *ctx; + int ret; + + if (copy_from_user(&cmd, inbuf, sizeof(cmd))) + return -EFAULT; + + if (!rdma_addr_size_in6(&cmd.addr)) + return -EINVAL; + + ctx = ucma_get_ctx(file, cmd.id); + if (IS_ERR(ctx)) + return PTR_ERR(ctx); + + mutex_lock(&ctx->mutex); + ret = rdma_bind_addr(ctx->cm_id, (struct sockaddr *) &cmd.addr); + mutex_unlock(&ctx->mutex); + + ucma_put_ctx(ctx); + return ret; +} + +static ssize_t ucma_bind(struct ucma_file *file, const char __user *inbuf, + int in_len, int out_len) +{ + struct rdma_ucm_bind cmd; + struct ucma_context *ctx; + int ret; + + if (copy_from_user(&cmd, inbuf, sizeof(cmd))) + return -EFAULT; + + if (cmd.reserved || !cmd.addr_size || + cmd.addr_size != rdma_addr_size_kss(&cmd.addr)) + return -EINVAL; + + ctx = ucma_get_ctx(file, cmd.id); + if (IS_ERR(ctx)) + return PTR_ERR(ctx); + + mutex_lock(&ctx->mutex); + ret = rdma_bind_addr(ctx->cm_id, (struct sockaddr *) &cmd.addr); + mutex_unlock(&ctx->mutex); + ucma_put_ctx(ctx); + return ret; +} + +static ssize_t ucma_resolve_ip(struct ucma_file *file, + const char __user *inbuf, + int in_len, int out_len) +{ + struct rdma_ucm_resolve_ip cmd; + struct ucma_context *ctx; + int ret; + + if (copy_from_user(&cmd, inbuf, sizeof(cmd))) + return -EFAULT; + + if ((cmd.src_addr.sin6_family && !rdma_addr_size_in6(&cmd.src_addr)) || + !rdma_addr_size_in6(&cmd.dst_addr)) + return -EINVAL; + + ctx = ucma_get_ctx(file, cmd.id); + if (IS_ERR(ctx)) + return PTR_ERR(ctx); + + mutex_lock(&ctx->mutex); + ret = rdma_resolve_addr(ctx->cm_id, (struct sockaddr *) &cmd.src_addr, + (struct sockaddr *) &cmd.dst_addr, cmd.timeout_ms); + mutex_unlock(&ctx->mutex); + ucma_put_ctx(ctx); + return ret; +} + +static ssize_t ucma_resolve_addr(struct ucma_file *file, + const char __user *inbuf, + int in_len, int out_len) +{ + struct rdma_ucm_resolve_addr cmd; + struct ucma_context *ctx; + int ret; + + if (copy_from_user(&cmd, inbuf, sizeof(cmd))) + return -EFAULT; + + if (cmd.reserved || + (cmd.src_size && (cmd.src_size != rdma_addr_size_kss(&cmd.src_addr))) || + !cmd.dst_size || (cmd.dst_size != rdma_addr_size_kss(&cmd.dst_addr))) + return -EINVAL; + + ctx = ucma_get_ctx(file, cmd.id); + if (IS_ERR(ctx)) + return PTR_ERR(ctx); + + mutex_lock(&ctx->mutex); + ret = rdma_resolve_addr(ctx->cm_id, (struct sockaddr *) &cmd.src_addr, + (struct sockaddr *) &cmd.dst_addr, cmd.timeout_ms); + mutex_unlock(&ctx->mutex); + ucma_put_ctx(ctx); + return ret; +} + +static ssize_t ucma_resolve_route(struct ucma_file *file, + const char __user *inbuf, + int in_len, int out_len) +{ + struct rdma_ucm_resolve_route cmd; + struct ucma_context *ctx; + int ret; + + if (copy_from_user(&cmd, inbuf, sizeof(cmd))) + return -EFAULT; + + ctx = ucma_get_ctx_dev(file, cmd.id); + if (IS_ERR(ctx)) + return PTR_ERR(ctx); + + mutex_lock(&ctx->mutex); + ret = rdma_resolve_route(ctx->cm_id, cmd.timeout_ms); + mutex_unlock(&ctx->mutex); + ucma_put_ctx(ctx); + return ret; +} + +static void ucma_copy_ib_route(struct rdma_ucm_query_route_resp *resp, + struct rdma_route *route) +{ + struct rdma_dev_addr *dev_addr; + + resp->num_paths = route->num_paths; + switch (route->num_paths) { + case 0: + dev_addr = &route->addr.dev_addr; + rdma_addr_get_dgid(dev_addr, + (union ib_gid *) &resp->ib_route[0].dgid); + rdma_addr_get_sgid(dev_addr, + (union ib_gid *) &resp->ib_route[0].sgid); + resp->ib_route[0].pkey = cpu_to_be16(ib_addr_get_pkey(dev_addr)); + break; + case 2: + ib_copy_path_rec_to_user(&resp->ib_route[1], + &route->path_rec[1]); + fallthrough; + case 1: + ib_copy_path_rec_to_user(&resp->ib_route[0], + &route->path_rec[0]); + break; + default: + break; + } +} + +static void ucma_copy_iboe_route(struct rdma_ucm_query_route_resp *resp, + struct rdma_route *route) +{ + + resp->num_paths = route->num_paths; + switch (route->num_paths) { + case 0: + rdma_ip2gid((struct sockaddr *)&route->addr.dst_addr, + (union ib_gid *)&resp->ib_route[0].dgid); + rdma_ip2gid((struct sockaddr *)&route->addr.src_addr, + (union ib_gid *)&resp->ib_route[0].sgid); + resp->ib_route[0].pkey = cpu_to_be16(0xffff); + break; + case 2: + ib_copy_path_rec_to_user(&resp->ib_route[1], + &route->path_rec[1]); + fallthrough; + case 1: + ib_copy_path_rec_to_user(&resp->ib_route[0], + &route->path_rec[0]); + break; + default: + break; + } +} + +static void ucma_copy_iw_route(struct rdma_ucm_query_route_resp *resp, + struct rdma_route *route) +{ + struct rdma_dev_addr *dev_addr; + + dev_addr = &route->addr.dev_addr; + rdma_addr_get_dgid(dev_addr, (union ib_gid *) &resp->ib_route[0].dgid); + rdma_addr_get_sgid(dev_addr, (union ib_gid *) &resp->ib_route[0].sgid); +} + +static ssize_t ucma_query_route(struct ucma_file *file, + const char __user *inbuf, + int in_len, int out_len) +{ + struct rdma_ucm_query cmd; + struct rdma_ucm_query_route_resp resp; + struct ucma_context *ctx; + struct sockaddr *addr; + int ret = 0; + + if (out_len < offsetof(struct rdma_ucm_query_route_resp, ibdev_index)) + return -ENOSPC; + + if (copy_from_user(&cmd, inbuf, sizeof(cmd))) + return -EFAULT; + + ctx = ucma_get_ctx(file, cmd.id); + if (IS_ERR(ctx)) + return PTR_ERR(ctx); + + mutex_lock(&ctx->mutex); + memset(&resp, 0, sizeof resp); + addr = (struct sockaddr *) &ctx->cm_id->route.addr.src_addr; + memcpy(&resp.src_addr, addr, addr->sa_family == AF_INET ? + sizeof(struct sockaddr_in) : + sizeof(struct sockaddr_in6)); + addr = (struct sockaddr *) &ctx->cm_id->route.addr.dst_addr; + memcpy(&resp.dst_addr, addr, addr->sa_family == AF_INET ? + sizeof(struct sockaddr_in) : + sizeof(struct sockaddr_in6)); + if (!ctx->cm_id->device) + goto out; + + resp.node_guid = (__force __u64) ctx->cm_id->device->node_guid; + resp.ibdev_index = ctx->cm_id->device->index; + resp.port_num = ctx->cm_id->port_num; + + if (rdma_cap_ib_sa(ctx->cm_id->device, ctx->cm_id->port_num)) + ucma_copy_ib_route(&resp, &ctx->cm_id->route); + else if (rdma_protocol_roce(ctx->cm_id->device, ctx->cm_id->port_num)) + ucma_copy_iboe_route(&resp, &ctx->cm_id->route); + else if (rdma_protocol_iwarp(ctx->cm_id->device, ctx->cm_id->port_num)) + ucma_copy_iw_route(&resp, &ctx->cm_id->route); + +out: + mutex_unlock(&ctx->mutex); + if (copy_to_user(u64_to_user_ptr(cmd.response), &resp, + min_t(size_t, out_len, sizeof(resp)))) + ret = -EFAULT; + + ucma_put_ctx(ctx); + return ret; +} + +static void ucma_query_device_addr(struct rdma_cm_id *cm_id, + struct rdma_ucm_query_addr_resp *resp) +{ + if (!cm_id->device) + return; + + resp->node_guid = (__force __u64) cm_id->device->node_guid; + resp->ibdev_index = cm_id->device->index; + resp->port_num = cm_id->port_num; + resp->pkey = (__force __u16) cpu_to_be16( + ib_addr_get_pkey(&cm_id->route.addr.dev_addr)); +} + +static ssize_t ucma_query_addr(struct ucma_context *ctx, + void __user *response, int out_len) +{ + struct rdma_ucm_query_addr_resp resp; + struct sockaddr *addr; + int ret = 0; + + if (out_len < offsetof(struct rdma_ucm_query_addr_resp, ibdev_index)) + return -ENOSPC; + + memset(&resp, 0, sizeof resp); + + addr = (struct sockaddr *) &ctx->cm_id->route.addr.src_addr; + resp.src_size = rdma_addr_size(addr); + memcpy(&resp.src_addr, addr, resp.src_size); + + addr = (struct sockaddr *) &ctx->cm_id->route.addr.dst_addr; + resp.dst_size = rdma_addr_size(addr); + memcpy(&resp.dst_addr, addr, resp.dst_size); + + ucma_query_device_addr(ctx->cm_id, &resp); + + if (copy_to_user(response, &resp, min_t(size_t, out_len, sizeof(resp)))) + ret = -EFAULT; + + return ret; +} + +static ssize_t ucma_query_path(struct ucma_context *ctx, + void __user *response, int out_len) +{ + struct rdma_ucm_query_path_resp *resp; + int i, ret = 0; + + if (out_len < sizeof(*resp)) + return -ENOSPC; + + resp = kzalloc(out_len, GFP_KERNEL); + if (!resp) + return -ENOMEM; + + resp->num_paths = ctx->cm_id->route.num_paths; + for (i = 0, out_len -= sizeof(*resp); + i < resp->num_paths && out_len > sizeof(struct ib_path_rec_data); + i++, out_len -= sizeof(struct ib_path_rec_data)) { + struct sa_path_rec *rec = &ctx->cm_id->route.path_rec[i]; + + resp->path_data[i].flags = IB_PATH_GMP | IB_PATH_PRIMARY | + IB_PATH_BIDIRECTIONAL; + if (rec->rec_type == SA_PATH_REC_TYPE_OPA) { + struct sa_path_rec ib; + + sa_convert_path_opa_to_ib(&ib, rec); + ib_sa_pack_path(&ib, &resp->path_data[i].path_rec); + + } else { + ib_sa_pack_path(rec, &resp->path_data[i].path_rec); + } + } + + if (copy_to_user(response, resp, struct_size(resp, path_data, i))) + ret = -EFAULT; + + kfree(resp); + return ret; +} + +static ssize_t ucma_query_gid(struct ucma_context *ctx, + void __user *response, int out_len) +{ + struct rdma_ucm_query_addr_resp resp; + struct sockaddr_ib *addr; + int ret = 0; + + if (out_len < offsetof(struct rdma_ucm_query_addr_resp, ibdev_index)) + return -ENOSPC; + + memset(&resp, 0, sizeof resp); + + ucma_query_device_addr(ctx->cm_id, &resp); + + addr = (struct sockaddr_ib *) &resp.src_addr; + resp.src_size = sizeof(*addr); + if (ctx->cm_id->route.addr.src_addr.ss_family == AF_IB) { + memcpy(addr, &ctx->cm_id->route.addr.src_addr, resp.src_size); + } else { + addr->sib_family = AF_IB; + addr->sib_pkey = (__force __be16) resp.pkey; + rdma_read_gids(ctx->cm_id, (union ib_gid *)&addr->sib_addr, + NULL); + addr->sib_sid = rdma_get_service_id(ctx->cm_id, (struct sockaddr *) + &ctx->cm_id->route.addr.src_addr); + } + + addr = (struct sockaddr_ib *) &resp.dst_addr; + resp.dst_size = sizeof(*addr); + if (ctx->cm_id->route.addr.dst_addr.ss_family == AF_IB) { + memcpy(addr, &ctx->cm_id->route.addr.dst_addr, resp.dst_size); + } else { + addr->sib_family = AF_IB; + addr->sib_pkey = (__force __be16) resp.pkey; + rdma_read_gids(ctx->cm_id, NULL, + (union ib_gid *)&addr->sib_addr); + addr->sib_sid = rdma_get_service_id(ctx->cm_id, (struct sockaddr *) + &ctx->cm_id->route.addr.dst_addr); + } + + if (copy_to_user(response, &resp, min_t(size_t, out_len, sizeof(resp)))) + ret = -EFAULT; + + return ret; +} + +static ssize_t ucma_query(struct ucma_file *file, + const char __user *inbuf, + int in_len, int out_len) +{ + struct rdma_ucm_query cmd; + struct ucma_context *ctx; + void __user *response; + int ret; + + if (copy_from_user(&cmd, inbuf, sizeof(cmd))) + return -EFAULT; + + response = u64_to_user_ptr(cmd.response); + ctx = ucma_get_ctx(file, cmd.id); + if (IS_ERR(ctx)) + return PTR_ERR(ctx); + + mutex_lock(&ctx->mutex); + switch (cmd.option) { + case RDMA_USER_CM_QUERY_ADDR: + ret = ucma_query_addr(ctx, response, out_len); + break; + case RDMA_USER_CM_QUERY_PATH: + ret = ucma_query_path(ctx, response, out_len); + break; + case RDMA_USER_CM_QUERY_GID: + ret = ucma_query_gid(ctx, response, out_len); + break; + default: + ret = -ENOSYS; + break; + } + mutex_unlock(&ctx->mutex); + + ucma_put_ctx(ctx); + return ret; +} + +static void ucma_copy_conn_param(struct rdma_cm_id *id, + struct rdma_conn_param *dst, + struct rdma_ucm_conn_param *src) +{ + dst->private_data = src->private_data; + dst->private_data_len = src->private_data_len; + dst->responder_resources =src->responder_resources; + dst->initiator_depth = src->initiator_depth; + dst->flow_control = src->flow_control; + dst->retry_count = src->retry_count; + dst->rnr_retry_count = src->rnr_retry_count; + dst->srq = src->srq; + dst->qp_num = src->qp_num & 0xFFFFFF; + dst->qkey = (id->route.addr.src_addr.ss_family == AF_IB) ? src->qkey : 0; +} + +static ssize_t ucma_connect(struct ucma_file *file, const char __user *inbuf, + int in_len, int out_len) +{ + struct rdma_conn_param conn_param; + struct rdma_ucm_ece ece = {}; + struct rdma_ucm_connect cmd; + struct ucma_context *ctx; + size_t in_size; + int ret; + + if (in_len < offsetofend(typeof(cmd), reserved)) + return -EINVAL; + in_size = min_t(size_t, in_len, sizeof(cmd)); + if (copy_from_user(&cmd, inbuf, in_size)) + return -EFAULT; + + if (!cmd.conn_param.valid) + return -EINVAL; + + ctx = ucma_get_ctx_dev(file, cmd.id); + if (IS_ERR(ctx)) + return PTR_ERR(ctx); + + ucma_copy_conn_param(ctx->cm_id, &conn_param, &cmd.conn_param); + if (offsetofend(typeof(cmd), ece) <= in_size) { + ece.vendor_id = cmd.ece.vendor_id; + ece.attr_mod = cmd.ece.attr_mod; + } + + mutex_lock(&ctx->mutex); + ret = rdma_connect_ece(ctx->cm_id, &conn_param, &ece); + mutex_unlock(&ctx->mutex); + ucma_put_ctx(ctx); + return ret; +} + +static ssize_t ucma_listen(struct ucma_file *file, const char __user *inbuf, + int in_len, int out_len) +{ + struct rdma_ucm_listen cmd; + struct ucma_context *ctx; + int ret; + + if (copy_from_user(&cmd, inbuf, sizeof(cmd))) + return -EFAULT; + + ctx = ucma_get_ctx(file, cmd.id); + if (IS_ERR(ctx)) + return PTR_ERR(ctx); + + if (cmd.backlog <= 0 || cmd.backlog > max_backlog) + cmd.backlog = max_backlog; + atomic_set(&ctx->backlog, cmd.backlog); + + mutex_lock(&ctx->mutex); + ret = rdma_listen(ctx->cm_id, cmd.backlog); + mutex_unlock(&ctx->mutex); + ucma_put_ctx(ctx); + return ret; +} + +static ssize_t ucma_accept(struct ucma_file *file, const char __user *inbuf, + int in_len, int out_len) +{ + struct rdma_ucm_accept cmd; + struct rdma_conn_param conn_param; + struct rdma_ucm_ece ece = {}; + struct ucma_context *ctx; + size_t in_size; + int ret; + + if (in_len < offsetofend(typeof(cmd), reserved)) + return -EINVAL; + in_size = min_t(size_t, in_len, sizeof(cmd)); + if (copy_from_user(&cmd, inbuf, in_size)) + return -EFAULT; + + ctx = ucma_get_ctx_dev(file, cmd.id); + if (IS_ERR(ctx)) + return PTR_ERR(ctx); + + if (offsetofend(typeof(cmd), ece) <= in_size) { + ece.vendor_id = cmd.ece.vendor_id; + ece.attr_mod = cmd.ece.attr_mod; + } + + if (cmd.conn_param.valid) { + ucma_copy_conn_param(ctx->cm_id, &conn_param, &cmd.conn_param); + mutex_lock(&ctx->mutex); + rdma_lock_handler(ctx->cm_id); + ret = rdma_accept_ece(ctx->cm_id, &conn_param, &ece); + if (!ret) { + /* The uid must be set atomically with the handler */ + ctx->uid = cmd.uid; + } + rdma_unlock_handler(ctx->cm_id); + mutex_unlock(&ctx->mutex); + } else { + mutex_lock(&ctx->mutex); + rdma_lock_handler(ctx->cm_id); + ret = rdma_accept_ece(ctx->cm_id, NULL, &ece); + rdma_unlock_handler(ctx->cm_id); + mutex_unlock(&ctx->mutex); + } + ucma_put_ctx(ctx); + return ret; +} + +static ssize_t ucma_reject(struct ucma_file *file, const char __user *inbuf, + int in_len, int out_len) +{ + struct rdma_ucm_reject cmd; + struct ucma_context *ctx; + int ret; + + if (copy_from_user(&cmd, inbuf, sizeof(cmd))) + return -EFAULT; + + if (!cmd.reason) + cmd.reason = IB_CM_REJ_CONSUMER_DEFINED; + + switch (cmd.reason) { + case IB_CM_REJ_CONSUMER_DEFINED: + case IB_CM_REJ_VENDOR_OPTION_NOT_SUPPORTED: + break; + default: + return -EINVAL; + } + + ctx = ucma_get_ctx_dev(file, cmd.id); + if (IS_ERR(ctx)) + return PTR_ERR(ctx); + + mutex_lock(&ctx->mutex); + ret = rdma_reject(ctx->cm_id, cmd.private_data, cmd.private_data_len, + cmd.reason); + mutex_unlock(&ctx->mutex); + ucma_put_ctx(ctx); + return ret; +} + +static ssize_t ucma_disconnect(struct ucma_file *file, const char __user *inbuf, + int in_len, int out_len) +{ + struct rdma_ucm_disconnect cmd; + struct ucma_context *ctx; + int ret; + + if (copy_from_user(&cmd, inbuf, sizeof(cmd))) + return -EFAULT; + + ctx = ucma_get_ctx_dev(file, cmd.id); + if (IS_ERR(ctx)) + return PTR_ERR(ctx); + + mutex_lock(&ctx->mutex); + ret = rdma_disconnect(ctx->cm_id); + mutex_unlock(&ctx->mutex); + ucma_put_ctx(ctx); + return ret; +} + +static ssize_t ucma_init_qp_attr(struct ucma_file *file, + const char __user *inbuf, + int in_len, int out_len) +{ + struct rdma_ucm_init_qp_attr cmd; + struct ib_uverbs_qp_attr resp; + struct ucma_context *ctx; + struct ib_qp_attr qp_attr; + int ret; + + if (out_len < sizeof(resp)) + return -ENOSPC; + + if (copy_from_user(&cmd, inbuf, sizeof(cmd))) + return -EFAULT; + + if (cmd.qp_state > IB_QPS_ERR) + return -EINVAL; + + ctx = ucma_get_ctx_dev(file, cmd.id); + if (IS_ERR(ctx)) + return PTR_ERR(ctx); + + resp.qp_attr_mask = 0; + memset(&qp_attr, 0, sizeof qp_attr); + qp_attr.qp_state = cmd.qp_state; + mutex_lock(&ctx->mutex); + ret = rdma_init_qp_attr(ctx->cm_id, &qp_attr, &resp.qp_attr_mask); + mutex_unlock(&ctx->mutex); + if (ret) + goto out; + + ib_copy_qp_attr_to_user(ctx->cm_id->device, &resp, &qp_attr); + if (copy_to_user(u64_to_user_ptr(cmd.response), + &resp, sizeof(resp))) + ret = -EFAULT; + +out: + ucma_put_ctx(ctx); + return ret; +} + +static int ucma_set_option_id(struct ucma_context *ctx, int optname, + void *optval, size_t optlen) +{ + int ret = 0; + + switch (optname) { + case RDMA_OPTION_ID_TOS: + if (optlen != sizeof(u8)) { + ret = -EINVAL; + break; + } + rdma_set_service_type(ctx->cm_id, *((u8 *) optval)); + break; + case RDMA_OPTION_ID_REUSEADDR: + if (optlen != sizeof(int)) { + ret = -EINVAL; + break; + } + ret = rdma_set_reuseaddr(ctx->cm_id, *((int *) optval) ? 1 : 0); + break; + case RDMA_OPTION_ID_AFONLY: + if (optlen != sizeof(int)) { + ret = -EINVAL; + break; + } + ret = rdma_set_afonly(ctx->cm_id, *((int *) optval) ? 1 : 0); + break; + case RDMA_OPTION_ID_ACK_TIMEOUT: + if (optlen != sizeof(u8)) { + ret = -EINVAL; + break; + } + ret = rdma_set_ack_timeout(ctx->cm_id, *((u8 *)optval)); + break; + default: + ret = -ENOSYS; + } + + return ret; +} + +static int ucma_set_ib_path(struct ucma_context *ctx, + struct ib_path_rec_data *path_data, size_t optlen) +{ + struct sa_path_rec sa_path; + struct rdma_cm_event event; + int ret; + + if (optlen % sizeof(*path_data)) + return -EINVAL; + + for (; optlen; optlen -= sizeof(*path_data), path_data++) { + if (path_data->flags == (IB_PATH_GMP | IB_PATH_PRIMARY | + IB_PATH_BIDIRECTIONAL)) + break; + } + + if (!optlen) + return -EINVAL; + + if (!ctx->cm_id->device) + return -EINVAL; + + memset(&sa_path, 0, sizeof(sa_path)); + + sa_path.rec_type = SA_PATH_REC_TYPE_IB; + ib_sa_unpack_path(path_data->path_rec, &sa_path); + + if (rdma_cap_opa_ah(ctx->cm_id->device, ctx->cm_id->port_num)) { + struct sa_path_rec opa; + + sa_convert_path_ib_to_opa(&opa, &sa_path); + mutex_lock(&ctx->mutex); + ret = rdma_set_ib_path(ctx->cm_id, &opa); + mutex_unlock(&ctx->mutex); + } else { + mutex_lock(&ctx->mutex); + ret = rdma_set_ib_path(ctx->cm_id, &sa_path); + mutex_unlock(&ctx->mutex); + } + if (ret) + return ret; + + memset(&event, 0, sizeof event); + event.event = RDMA_CM_EVENT_ROUTE_RESOLVED; + return ucma_event_handler(ctx->cm_id, &event); +} + +static int ucma_set_option_ib(struct ucma_context *ctx, int optname, + void *optval, size_t optlen) +{ + int ret; + + switch (optname) { + case RDMA_OPTION_IB_PATH: + ret = ucma_set_ib_path(ctx, optval, optlen); + break; + default: + ret = -ENOSYS; + } + + return ret; +} + +static int ucma_set_option_level(struct ucma_context *ctx, int level, + int optname, void *optval, size_t optlen) +{ + int ret; + + switch (level) { + case RDMA_OPTION_ID: + mutex_lock(&ctx->mutex); + ret = ucma_set_option_id(ctx, optname, optval, optlen); + mutex_unlock(&ctx->mutex); + break; + case RDMA_OPTION_IB: + ret = ucma_set_option_ib(ctx, optname, optval, optlen); + break; + default: + ret = -ENOSYS; + } + + return ret; +} + +static ssize_t ucma_set_option(struct ucma_file *file, const char __user *inbuf, + int in_len, int out_len) +{ + struct rdma_ucm_set_option cmd; + struct ucma_context *ctx; + void *optval; + int ret; + + if (copy_from_user(&cmd, inbuf, sizeof(cmd))) + return -EFAULT; + + if (unlikely(cmd.optlen > KMALLOC_MAX_SIZE)) + return -EINVAL; + + ctx = ucma_get_ctx(file, cmd.id); + if (IS_ERR(ctx)) + return PTR_ERR(ctx); + + optval = memdup_user(u64_to_user_ptr(cmd.optval), + cmd.optlen); + if (IS_ERR(optval)) { + ret = PTR_ERR(optval); + goto out; + } + + ret = ucma_set_option_level(ctx, cmd.level, cmd.optname, optval, + cmd.optlen); + kfree(optval); + +out: + ucma_put_ctx(ctx); + return ret; +} + +static ssize_t ucma_notify(struct ucma_file *file, const char __user *inbuf, + int in_len, int out_len) +{ + struct rdma_ucm_notify cmd; + struct ucma_context *ctx; + int ret = -EINVAL; + + if (copy_from_user(&cmd, inbuf, sizeof(cmd))) + return -EFAULT; + + ctx = ucma_get_ctx(file, cmd.id); + if (IS_ERR(ctx)) + return PTR_ERR(ctx); + + mutex_lock(&ctx->mutex); + if (ctx->cm_id->device) + ret = rdma_notify(ctx->cm_id, (enum ib_event_type)cmd.event); + mutex_unlock(&ctx->mutex); + + ucma_put_ctx(ctx); + return ret; +} + +static ssize_t ucma_process_join(struct ucma_file *file, + struct rdma_ucm_join_mcast *cmd, int out_len) +{ + struct rdma_ucm_create_id_resp resp; + struct ucma_context *ctx; + struct ucma_multicast *mc; + struct sockaddr *addr; + int ret; + u8 join_state; + + if (out_len < sizeof(resp)) + return -ENOSPC; + + addr = (struct sockaddr *) &cmd->addr; + if (cmd->addr_size != rdma_addr_size(addr)) + return -EINVAL; + + if (cmd->join_flags == RDMA_MC_JOIN_FLAG_FULLMEMBER) + join_state = BIT(FULLMEMBER_JOIN); + else if (cmd->join_flags == RDMA_MC_JOIN_FLAG_SENDONLY_FULLMEMBER) + join_state = BIT(SENDONLY_FULLMEMBER_JOIN); + else + return -EINVAL; + + ctx = ucma_get_ctx_dev(file, cmd->id); + if (IS_ERR(ctx)) + return PTR_ERR(ctx); + + mc = kzalloc(sizeof(*mc), GFP_KERNEL); + if (!mc) { + ret = -ENOMEM; + goto err_put_ctx; + } + + mc->ctx = ctx; + mc->join_state = join_state; + mc->uid = cmd->uid; + memcpy(&mc->addr, addr, cmd->addr_size); + + xa_lock(&multicast_table); + if (__xa_alloc(&multicast_table, &mc->id, NULL, xa_limit_32b, + GFP_KERNEL)) { + ret = -ENOMEM; + goto err_free_mc; + } + + list_add_tail(&mc->list, &ctx->mc_list); + xa_unlock(&multicast_table); + + mutex_lock(&ctx->mutex); + ret = rdma_join_multicast(ctx->cm_id, (struct sockaddr *)&mc->addr, + join_state, mc); + mutex_unlock(&ctx->mutex); + if (ret) + goto err_xa_erase; + + resp.id = mc->id; + if (copy_to_user(u64_to_user_ptr(cmd->response), + &resp, sizeof(resp))) { + ret = -EFAULT; + goto err_leave_multicast; + } + + xa_store(&multicast_table, mc->id, mc, 0); + + ucma_put_ctx(ctx); + return 0; + +err_leave_multicast: + mutex_lock(&ctx->mutex); + rdma_leave_multicast(ctx->cm_id, (struct sockaddr *) &mc->addr); + mutex_unlock(&ctx->mutex); + ucma_cleanup_mc_events(mc); +err_xa_erase: + xa_lock(&multicast_table); + list_del(&mc->list); + __xa_erase(&multicast_table, mc->id); +err_free_mc: + xa_unlock(&multicast_table); + kfree(mc); +err_put_ctx: + ucma_put_ctx(ctx); + return ret; +} + +static ssize_t ucma_join_ip_multicast(struct ucma_file *file, + const char __user *inbuf, + int in_len, int out_len) +{ + struct rdma_ucm_join_ip_mcast cmd; + struct rdma_ucm_join_mcast join_cmd; + + if (copy_from_user(&cmd, inbuf, sizeof(cmd))) + return -EFAULT; + + join_cmd.response = cmd.response; + join_cmd.uid = cmd.uid; + join_cmd.id = cmd.id; + join_cmd.addr_size = rdma_addr_size_in6(&cmd.addr); + if (!join_cmd.addr_size) + return -EINVAL; + + join_cmd.join_flags = RDMA_MC_JOIN_FLAG_FULLMEMBER; + memcpy(&join_cmd.addr, &cmd.addr, join_cmd.addr_size); + + return ucma_process_join(file, &join_cmd, out_len); +} + +static ssize_t ucma_join_multicast(struct ucma_file *file, + const char __user *inbuf, + int in_len, int out_len) +{ + struct rdma_ucm_join_mcast cmd; + + if (copy_from_user(&cmd, inbuf, sizeof(cmd))) + return -EFAULT; + + if (!rdma_addr_size_kss(&cmd.addr)) + return -EINVAL; + + return ucma_process_join(file, &cmd, out_len); +} + +static ssize_t ucma_leave_multicast(struct ucma_file *file, + const char __user *inbuf, + int in_len, int out_len) +{ + struct rdma_ucm_destroy_id cmd; + struct rdma_ucm_destroy_id_resp resp; + struct ucma_multicast *mc; + int ret = 0; + + if (out_len < sizeof(resp)) + return -ENOSPC; + + if (copy_from_user(&cmd, inbuf, sizeof(cmd))) + return -EFAULT; + + xa_lock(&multicast_table); + mc = xa_load(&multicast_table, cmd.id); + if (!mc) + mc = ERR_PTR(-ENOENT); + else if (READ_ONCE(mc->ctx->file) != file) + mc = ERR_PTR(-EINVAL); + else if (!refcount_inc_not_zero(&mc->ctx->ref)) + mc = ERR_PTR(-ENXIO); + + if (IS_ERR(mc)) { + xa_unlock(&multicast_table); + ret = PTR_ERR(mc); + goto out; + } + + list_del(&mc->list); + __xa_erase(&multicast_table, mc->id); + xa_unlock(&multicast_table); + + mutex_lock(&mc->ctx->mutex); + rdma_leave_multicast(mc->ctx->cm_id, (struct sockaddr *) &mc->addr); + mutex_unlock(&mc->ctx->mutex); + + ucma_cleanup_mc_events(mc); + + ucma_put_ctx(mc->ctx); + resp.events_reported = mc->events_reported; + kfree(mc); + + if (copy_to_user(u64_to_user_ptr(cmd.response), + &resp, sizeof(resp))) + ret = -EFAULT; +out: + return ret; +} + +static ssize_t ucma_migrate_id(struct ucma_file *new_file, + const char __user *inbuf, + int in_len, int out_len) +{ + struct rdma_ucm_migrate_id cmd; + struct rdma_ucm_migrate_resp resp; + struct ucma_event *uevent, *tmp; + struct ucma_context *ctx; + LIST_HEAD(event_list); + struct fd f; + struct ucma_file *cur_file; + int ret = 0; + + if (copy_from_user(&cmd, inbuf, sizeof(cmd))) + return -EFAULT; + + /* Get current fd to protect against it being closed */ + f = fdget(cmd.fd); + if (!f.file) + return -ENOENT; + if (f.file->f_op != &ucma_fops) { + ret = -EINVAL; + goto file_put; + } + cur_file = f.file->private_data; + + /* Validate current fd and prevent destruction of id. */ + ctx = ucma_get_ctx(cur_file, cmd.id); + if (IS_ERR(ctx)) { + ret = PTR_ERR(ctx); + goto file_put; + } + + rdma_lock_handler(ctx->cm_id); + /* + * ctx->file can only be changed under the handler & xa_lock. xa_load() + * must be checked again to ensure the ctx hasn't begun destruction + * since the ucma_get_ctx(). + */ + xa_lock(&ctx_table); + if (_ucma_find_context(cmd.id, cur_file) != ctx) { + xa_unlock(&ctx_table); + ret = -ENOENT; + goto err_unlock; + } + ctx->file = new_file; + xa_unlock(&ctx_table); + + mutex_lock(&cur_file->mut); + list_del(&ctx->list); + /* + * At this point lock_handler() prevents addition of new uevents for + * this ctx. + */ + list_for_each_entry_safe(uevent, tmp, &cur_file->event_list, list) + if (uevent->ctx == ctx) + list_move_tail(&uevent->list, &event_list); + resp.events_reported = ctx->events_reported; + mutex_unlock(&cur_file->mut); + + mutex_lock(&new_file->mut); + list_add_tail(&ctx->list, &new_file->ctx_list); + list_splice_tail(&event_list, &new_file->event_list); + mutex_unlock(&new_file->mut); + + if (copy_to_user(u64_to_user_ptr(cmd.response), + &resp, sizeof(resp))) + ret = -EFAULT; + +err_unlock: + rdma_unlock_handler(ctx->cm_id); + ucma_put_ctx(ctx); +file_put: + fdput(f); + return ret; +} + +static ssize_t (*ucma_cmd_table[])(struct ucma_file *file, + const char __user *inbuf, + int in_len, int out_len) = { + [RDMA_USER_CM_CMD_CREATE_ID] = ucma_create_id, + [RDMA_USER_CM_CMD_DESTROY_ID] = ucma_destroy_id, + [RDMA_USER_CM_CMD_BIND_IP] = ucma_bind_ip, + [RDMA_USER_CM_CMD_RESOLVE_IP] = ucma_resolve_ip, + [RDMA_USER_CM_CMD_RESOLVE_ROUTE] = ucma_resolve_route, + [RDMA_USER_CM_CMD_QUERY_ROUTE] = ucma_query_route, + [RDMA_USER_CM_CMD_CONNECT] = ucma_connect, + [RDMA_USER_CM_CMD_LISTEN] = ucma_listen, + [RDMA_USER_CM_CMD_ACCEPT] = ucma_accept, + [RDMA_USER_CM_CMD_REJECT] = ucma_reject, + [RDMA_USER_CM_CMD_DISCONNECT] = ucma_disconnect, + [RDMA_USER_CM_CMD_INIT_QP_ATTR] = ucma_init_qp_attr, + [RDMA_USER_CM_CMD_GET_EVENT] = ucma_get_event, + [RDMA_USER_CM_CMD_GET_OPTION] = NULL, + [RDMA_USER_CM_CMD_SET_OPTION] = ucma_set_option, + [RDMA_USER_CM_CMD_NOTIFY] = ucma_notify, + [RDMA_USER_CM_CMD_JOIN_IP_MCAST] = ucma_join_ip_multicast, + [RDMA_USER_CM_CMD_LEAVE_MCAST] = ucma_leave_multicast, + [RDMA_USER_CM_CMD_MIGRATE_ID] = ucma_migrate_id, + [RDMA_USER_CM_CMD_QUERY] = ucma_query, + [RDMA_USER_CM_CMD_BIND] = ucma_bind, + [RDMA_USER_CM_CMD_RESOLVE_ADDR] = ucma_resolve_addr, + [RDMA_USER_CM_CMD_JOIN_MCAST] = ucma_join_multicast +}; + +static ssize_t ucma_write(struct file *filp, const char __user *buf, + size_t len, loff_t *pos) +{ + struct ucma_file *file = filp->private_data; + struct rdma_ucm_cmd_hdr hdr; + ssize_t ret; + + if (!ib_safe_file_access(filp)) { + pr_err_once("ucma_write: process %d (%s) changed security contexts after opening file descriptor, this is not allowed.\n", + task_tgid_vnr(current), current->comm); + return -EACCES; + } + + if (len < sizeof(hdr)) + return -EINVAL; + + if (copy_from_user(&hdr, buf, sizeof(hdr))) + return -EFAULT; + + if (hdr.cmd >= ARRAY_SIZE(ucma_cmd_table)) + return -EINVAL; + hdr.cmd = array_index_nospec(hdr.cmd, ARRAY_SIZE(ucma_cmd_table)); + + if (hdr.in + sizeof(hdr) > len) + return -EINVAL; + + if (!ucma_cmd_table[hdr.cmd]) + return -ENOSYS; + + ret = ucma_cmd_table[hdr.cmd](file, buf + sizeof(hdr), hdr.in, hdr.out); + if (!ret) + ret = len; + + return ret; +} + +static __poll_t ucma_poll(struct file *filp, struct poll_table_struct *wait) +{ + struct ucma_file *file = filp->private_data; + __poll_t mask = 0; + + poll_wait(filp, &file->poll_wait, wait); + + if (!list_empty(&file->event_list)) + mask = EPOLLIN | EPOLLRDNORM; + + return mask; +} + +/* + * ucma_open() does not need the BKL: + * + * - no global state is referred to; + * - there is no ioctl method to race against; + * - no further module initialization is required for open to work + * after the device is registered. + */ +static int ucma_open(struct inode *inode, struct file *filp) +{ + struct ucma_file *file; + + file = kmalloc(sizeof *file, GFP_KERNEL); + if (!file) + return -ENOMEM; + + INIT_LIST_HEAD(&file->event_list); + INIT_LIST_HEAD(&file->ctx_list); + init_waitqueue_head(&file->poll_wait); + mutex_init(&file->mut); + + filp->private_data = file; + file->filp = filp; + + return stream_open(inode, filp); +} + +static int ucma_close(struct inode *inode, struct file *filp) +{ + struct ucma_file *file = filp->private_data; + + /* + * All paths that touch ctx_list or ctx_list starting from write() are + * prevented by this being a FD release function. The list_add_tail() in + * ucma_connect_event_handler() can run concurrently, however it only + * adds to the list *after* a listening ID. By only reading the first of + * the list, and relying on ucma_destroy_private_ctx() to block + * ucma_connect_event_handler(), no additional locking is needed. + */ + while (!list_empty(&file->ctx_list)) { + struct ucma_context *ctx = list_first_entry( + &file->ctx_list, struct ucma_context, list); + + WARN_ON(xa_cmpxchg(&ctx_table, ctx->id, ctx, XA_ZERO_ENTRY, + GFP_KERNEL) != ctx); + ucma_destroy_private_ctx(ctx); + } + kfree(file); + return 0; +} + +static const struct file_operations ucma_fops = { + .owner = THIS_MODULE, + .open = ucma_open, + .release = ucma_close, + .write = ucma_write, + .poll = ucma_poll, + .llseek = no_llseek, +}; + +static struct miscdevice ucma_misc = { + .minor = MISC_DYNAMIC_MINOR, + .name = "rdma_cm", + .nodename = "infiniband/rdma_cm", + .mode = 0666, + .fops = &ucma_fops, +}; + +static int ucma_get_global_nl_info(struct ib_client_nl_info *res) +{ + res->abi = RDMA_USER_CM_ABI_VERSION; + res->cdev = ucma_misc.this_device; + return 0; +} + +static struct ib_client rdma_cma_client = { + .name = "rdma_cm", + .get_global_nl_info = ucma_get_global_nl_info, +}; +MODULE_ALIAS_RDMA_CLIENT("rdma_cm"); + +static ssize_t show_abi_version(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + return sprintf(buf, "%d\n", RDMA_USER_CM_ABI_VERSION); +} +static DEVICE_ATTR(abi_version, S_IRUGO, show_abi_version, NULL); + +static int __init ucma_init(void) +{ + int ret; + + ret = misc_register(&ucma_misc); + if (ret) + return ret; + + ret = device_create_file(ucma_misc.this_device, &dev_attr_abi_version); + if (ret) { + pr_err("rdma_ucm: couldn't create abi_version attr\n"); + goto err1; + } + + ucma_ctl_table_hdr = register_net_sysctl(&init_net, "net/rdma_ucm", ucma_ctl_table); + if (!ucma_ctl_table_hdr) { + pr_err("rdma_ucm: couldn't register sysctl paths\n"); + ret = -ENOMEM; + goto err2; + } + + ret = ib_register_client(&rdma_cma_client); + if (ret) + goto err3; + + return 0; +err3: + unregister_net_sysctl_table(ucma_ctl_table_hdr); +err2: + device_remove_file(ucma_misc.this_device, &dev_attr_abi_version); +err1: + misc_deregister(&ucma_misc); + return ret; +} + +static void __exit ucma_cleanup(void) +{ + ib_unregister_client(&rdma_cma_client); + unregister_net_sysctl_table(ucma_ctl_table_hdr); + device_remove_file(ucma_misc.this_device, &dev_attr_abi_version); + misc_deregister(&ucma_misc); +} + +module_init(ucma_init); +module_exit(ucma_cleanup); diff --git a/drivers/infiniband/core/ud_header.c b/drivers/infiniband/core/ud_header.c new file mode 100644 index 000000000..d65d541b9 --- /dev/null +++ b/drivers/infiniband/core/ud_header.c @@ -0,0 +1,547 @@ +/* + * Copyright (c) 2004 Topspin Corporation. All rights reserved. + * Copyright (c) 2005 Sun Microsystems, Inc. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include <linux/errno.h> +#include <linux/string.h> +#include <linux/export.h> +#include <linux/if_ether.h> +#include <linux/ip.h> + +#include <rdma/ib_pack.h> + +#define STRUCT_FIELD(header, field) \ + .struct_offset_bytes = offsetof(struct ib_unpacked_ ## header, field), \ + .struct_size_bytes = sizeof_field(struct ib_unpacked_ ## header, field), \ + .field_name = #header ":" #field + +static const struct ib_field lrh_table[] = { + { STRUCT_FIELD(lrh, virtual_lane), + .offset_words = 0, + .offset_bits = 0, + .size_bits = 4 }, + { STRUCT_FIELD(lrh, link_version), + .offset_words = 0, + .offset_bits = 4, + .size_bits = 4 }, + { STRUCT_FIELD(lrh, service_level), + .offset_words = 0, + .offset_bits = 8, + .size_bits = 4 }, + { RESERVED, + .offset_words = 0, + .offset_bits = 12, + .size_bits = 2 }, + { STRUCT_FIELD(lrh, link_next_header), + .offset_words = 0, + .offset_bits = 14, + .size_bits = 2 }, + { STRUCT_FIELD(lrh, destination_lid), + .offset_words = 0, + .offset_bits = 16, + .size_bits = 16 }, + { RESERVED, + .offset_words = 1, + .offset_bits = 0, + .size_bits = 5 }, + { STRUCT_FIELD(lrh, packet_length), + .offset_words = 1, + .offset_bits = 5, + .size_bits = 11 }, + { STRUCT_FIELD(lrh, source_lid), + .offset_words = 1, + .offset_bits = 16, + .size_bits = 16 } +}; + +static const struct ib_field eth_table[] = { + { STRUCT_FIELD(eth, dmac_h), + .offset_words = 0, + .offset_bits = 0, + .size_bits = 32 }, + { STRUCT_FIELD(eth, dmac_l), + .offset_words = 1, + .offset_bits = 0, + .size_bits = 16 }, + { STRUCT_FIELD(eth, smac_h), + .offset_words = 1, + .offset_bits = 16, + .size_bits = 16 }, + { STRUCT_FIELD(eth, smac_l), + .offset_words = 2, + .offset_bits = 0, + .size_bits = 32 }, + { STRUCT_FIELD(eth, type), + .offset_words = 3, + .offset_bits = 0, + .size_bits = 16 } +}; + +static const struct ib_field vlan_table[] = { + { STRUCT_FIELD(vlan, tag), + .offset_words = 0, + .offset_bits = 0, + .size_bits = 16 }, + { STRUCT_FIELD(vlan, type), + .offset_words = 0, + .offset_bits = 16, + .size_bits = 16 } +}; + +static const struct ib_field ip4_table[] = { + { STRUCT_FIELD(ip4, ver), + .offset_words = 0, + .offset_bits = 0, + .size_bits = 4 }, + { STRUCT_FIELD(ip4, hdr_len), + .offset_words = 0, + .offset_bits = 4, + .size_bits = 4 }, + { STRUCT_FIELD(ip4, tos), + .offset_words = 0, + .offset_bits = 8, + .size_bits = 8 }, + { STRUCT_FIELD(ip4, tot_len), + .offset_words = 0, + .offset_bits = 16, + .size_bits = 16 }, + { STRUCT_FIELD(ip4, id), + .offset_words = 1, + .offset_bits = 0, + .size_bits = 16 }, + { STRUCT_FIELD(ip4, frag_off), + .offset_words = 1, + .offset_bits = 16, + .size_bits = 16 }, + { STRUCT_FIELD(ip4, ttl), + .offset_words = 2, + .offset_bits = 0, + .size_bits = 8 }, + { STRUCT_FIELD(ip4, protocol), + .offset_words = 2, + .offset_bits = 8, + .size_bits = 8 }, + { STRUCT_FIELD(ip4, check), + .offset_words = 2, + .offset_bits = 16, + .size_bits = 16 }, + { STRUCT_FIELD(ip4, saddr), + .offset_words = 3, + .offset_bits = 0, + .size_bits = 32 }, + { STRUCT_FIELD(ip4, daddr), + .offset_words = 4, + .offset_bits = 0, + .size_bits = 32 } +}; + +static const struct ib_field udp_table[] = { + { STRUCT_FIELD(udp, sport), + .offset_words = 0, + .offset_bits = 0, + .size_bits = 16 }, + { STRUCT_FIELD(udp, dport), + .offset_words = 0, + .offset_bits = 16, + .size_bits = 16 }, + { STRUCT_FIELD(udp, length), + .offset_words = 1, + .offset_bits = 0, + .size_bits = 16 }, + { STRUCT_FIELD(udp, csum), + .offset_words = 1, + .offset_bits = 16, + .size_bits = 16 } +}; + +static const struct ib_field grh_table[] = { + { STRUCT_FIELD(grh, ip_version), + .offset_words = 0, + .offset_bits = 0, + .size_bits = 4 }, + { STRUCT_FIELD(grh, traffic_class), + .offset_words = 0, + .offset_bits = 4, + .size_bits = 8 }, + { STRUCT_FIELD(grh, flow_label), + .offset_words = 0, + .offset_bits = 12, + .size_bits = 20 }, + { STRUCT_FIELD(grh, payload_length), + .offset_words = 1, + .offset_bits = 0, + .size_bits = 16 }, + { STRUCT_FIELD(grh, next_header), + .offset_words = 1, + .offset_bits = 16, + .size_bits = 8 }, + { STRUCT_FIELD(grh, hop_limit), + .offset_words = 1, + .offset_bits = 24, + .size_bits = 8 }, + { STRUCT_FIELD(grh, source_gid), + .offset_words = 2, + .offset_bits = 0, + .size_bits = 128 }, + { STRUCT_FIELD(grh, destination_gid), + .offset_words = 6, + .offset_bits = 0, + .size_bits = 128 } +}; + +static const struct ib_field bth_table[] = { + { STRUCT_FIELD(bth, opcode), + .offset_words = 0, + .offset_bits = 0, + .size_bits = 8 }, + { STRUCT_FIELD(bth, solicited_event), + .offset_words = 0, + .offset_bits = 8, + .size_bits = 1 }, + { STRUCT_FIELD(bth, mig_req), + .offset_words = 0, + .offset_bits = 9, + .size_bits = 1 }, + { STRUCT_FIELD(bth, pad_count), + .offset_words = 0, + .offset_bits = 10, + .size_bits = 2 }, + { STRUCT_FIELD(bth, transport_header_version), + .offset_words = 0, + .offset_bits = 12, + .size_bits = 4 }, + { STRUCT_FIELD(bth, pkey), + .offset_words = 0, + .offset_bits = 16, + .size_bits = 16 }, + { RESERVED, + .offset_words = 1, + .offset_bits = 0, + .size_bits = 8 }, + { STRUCT_FIELD(bth, destination_qpn), + .offset_words = 1, + .offset_bits = 8, + .size_bits = 24 }, + { STRUCT_FIELD(bth, ack_req), + .offset_words = 2, + .offset_bits = 0, + .size_bits = 1 }, + { RESERVED, + .offset_words = 2, + .offset_bits = 1, + .size_bits = 7 }, + { STRUCT_FIELD(bth, psn), + .offset_words = 2, + .offset_bits = 8, + .size_bits = 24 } +}; + +static const struct ib_field deth_table[] = { + { STRUCT_FIELD(deth, qkey), + .offset_words = 0, + .offset_bits = 0, + .size_bits = 32 }, + { RESERVED, + .offset_words = 1, + .offset_bits = 0, + .size_bits = 8 }, + { STRUCT_FIELD(deth, source_qpn), + .offset_words = 1, + .offset_bits = 8, + .size_bits = 24 } +}; + +__sum16 ib_ud_ip4_csum(struct ib_ud_header *header) +{ + struct iphdr iph; + + iph.ihl = 5; + iph.version = 4; + iph.tos = header->ip4.tos; + iph.tot_len = header->ip4.tot_len; + iph.id = header->ip4.id; + iph.frag_off = header->ip4.frag_off; + iph.ttl = header->ip4.ttl; + iph.protocol = header->ip4.protocol; + iph.check = 0; + iph.saddr = header->ip4.saddr; + iph.daddr = header->ip4.daddr; + + return ip_fast_csum((u8 *)&iph, iph.ihl); +} +EXPORT_SYMBOL(ib_ud_ip4_csum); + +/** + * ib_ud_header_init - Initialize UD header structure + * @payload_bytes:Length of packet payload + * @lrh_present: specify if LRH is present + * @eth_present: specify if Eth header is present + * @vlan_present: packet is tagged vlan + * @grh_present: GRH flag (if non-zero, GRH will be included) + * @ip_version: if non-zero, IP header, V4 or V6, will be included + * @udp_present :if non-zero, UDP header will be included + * @immediate_present: specify if immediate data is present + * @header:Structure to initialize + */ +int ib_ud_header_init(int payload_bytes, + int lrh_present, + int eth_present, + int vlan_present, + int grh_present, + int ip_version, + int udp_present, + int immediate_present, + struct ib_ud_header *header) +{ + size_t udp_bytes = udp_present ? IB_UDP_BYTES : 0; + + grh_present = grh_present && !ip_version; + memset(header, 0, sizeof *header); + + /* + * UDP header without IP header doesn't make sense + */ + if (udp_present && ip_version != 4 && ip_version != 6) + return -EINVAL; + + if (lrh_present) { + u16 packet_length; + + header->lrh.link_version = 0; + header->lrh.link_next_header = + grh_present ? IB_LNH_IBA_GLOBAL : IB_LNH_IBA_LOCAL; + packet_length = (IB_LRH_BYTES + + IB_BTH_BYTES + + IB_DETH_BYTES + + (grh_present ? IB_GRH_BYTES : 0) + + payload_bytes + + 4 + /* ICRC */ + 3) / 4; /* round up */ + header->lrh.packet_length = cpu_to_be16(packet_length); + } + + if (vlan_present) + header->eth.type = cpu_to_be16(ETH_P_8021Q); + + if (ip_version == 6 || grh_present) { + header->grh.ip_version = 6; + header->grh.payload_length = + cpu_to_be16((udp_bytes + + IB_BTH_BYTES + + IB_DETH_BYTES + + payload_bytes + + 4 + /* ICRC */ + 3) & ~3); /* round up */ + header->grh.next_header = udp_present ? IPPROTO_UDP : 0x1b; + } + + if (ip_version == 4) { + header->ip4.ver = 4; /* version 4 */ + header->ip4.hdr_len = 5; /* 5 words */ + header->ip4.tot_len = + cpu_to_be16(IB_IP4_BYTES + + udp_bytes + + IB_BTH_BYTES + + IB_DETH_BYTES + + payload_bytes + + 4); /* ICRC */ + header->ip4.protocol = IPPROTO_UDP; + } + if (udp_present && ip_version) + header->udp.length = + cpu_to_be16(IB_UDP_BYTES + + IB_BTH_BYTES + + IB_DETH_BYTES + + payload_bytes + + 4); /* ICRC */ + + if (immediate_present) + header->bth.opcode = IB_OPCODE_UD_SEND_ONLY_WITH_IMMEDIATE; + else + header->bth.opcode = IB_OPCODE_UD_SEND_ONLY; + header->bth.pad_count = (4 - payload_bytes) & 3; + header->bth.transport_header_version = 0; + + header->lrh_present = lrh_present; + header->eth_present = eth_present; + header->vlan_present = vlan_present; + header->grh_present = grh_present || (ip_version == 6); + header->ipv4_present = ip_version == 4; + header->udp_present = udp_present; + header->immediate_present = immediate_present; + return 0; +} +EXPORT_SYMBOL(ib_ud_header_init); + +/** + * ib_ud_header_pack - Pack UD header struct into wire format + * @header:UD header struct + * @buf:Buffer to pack into + * + * ib_ud_header_pack() packs the UD header structure @header into wire + * format in the buffer @buf. + */ +int ib_ud_header_pack(struct ib_ud_header *header, + void *buf) +{ + int len = 0; + + if (header->lrh_present) { + ib_pack(lrh_table, ARRAY_SIZE(lrh_table), + &header->lrh, buf + len); + len += IB_LRH_BYTES; + } + if (header->eth_present) { + ib_pack(eth_table, ARRAY_SIZE(eth_table), + &header->eth, buf + len); + len += IB_ETH_BYTES; + } + if (header->vlan_present) { + ib_pack(vlan_table, ARRAY_SIZE(vlan_table), + &header->vlan, buf + len); + len += IB_VLAN_BYTES; + } + if (header->grh_present) { + ib_pack(grh_table, ARRAY_SIZE(grh_table), + &header->grh, buf + len); + len += IB_GRH_BYTES; + } + if (header->ipv4_present) { + ib_pack(ip4_table, ARRAY_SIZE(ip4_table), + &header->ip4, buf + len); + len += IB_IP4_BYTES; + } + if (header->udp_present) { + ib_pack(udp_table, ARRAY_SIZE(udp_table), + &header->udp, buf + len); + len += IB_UDP_BYTES; + } + + ib_pack(bth_table, ARRAY_SIZE(bth_table), + &header->bth, buf + len); + len += IB_BTH_BYTES; + + ib_pack(deth_table, ARRAY_SIZE(deth_table), + &header->deth, buf + len); + len += IB_DETH_BYTES; + + if (header->immediate_present) { + memcpy(buf + len, &header->immediate_data, sizeof header->immediate_data); + len += sizeof header->immediate_data; + } + + return len; +} +EXPORT_SYMBOL(ib_ud_header_pack); + +/** + * ib_ud_header_unpack - Unpack UD header struct from wire format + * @header:UD header struct + * @buf:Buffer to pack into + * + * ib_ud_header_pack() unpacks the UD header structure @header from wire + * format in the buffer @buf. + */ +int ib_ud_header_unpack(void *buf, + struct ib_ud_header *header) +{ + ib_unpack(lrh_table, ARRAY_SIZE(lrh_table), + buf, &header->lrh); + buf += IB_LRH_BYTES; + + if (header->lrh.link_version != 0) { + pr_warn("Invalid LRH.link_version %d\n", + header->lrh.link_version); + return -EINVAL; + } + + switch (header->lrh.link_next_header) { + case IB_LNH_IBA_LOCAL: + header->grh_present = 0; + break; + + case IB_LNH_IBA_GLOBAL: + header->grh_present = 1; + ib_unpack(grh_table, ARRAY_SIZE(grh_table), + buf, &header->grh); + buf += IB_GRH_BYTES; + + if (header->grh.ip_version != 6) { + pr_warn("Invalid GRH.ip_version %d\n", + header->grh.ip_version); + return -EINVAL; + } + if (header->grh.next_header != 0x1b) { + pr_warn("Invalid GRH.next_header 0x%02x\n", + header->grh.next_header); + return -EINVAL; + } + break; + + default: + pr_warn("Invalid LRH.link_next_header %d\n", + header->lrh.link_next_header); + return -EINVAL; + } + + ib_unpack(bth_table, ARRAY_SIZE(bth_table), + buf, &header->bth); + buf += IB_BTH_BYTES; + + switch (header->bth.opcode) { + case IB_OPCODE_UD_SEND_ONLY: + header->immediate_present = 0; + break; + case IB_OPCODE_UD_SEND_ONLY_WITH_IMMEDIATE: + header->immediate_present = 1; + break; + default: + pr_warn("Invalid BTH.opcode 0x%02x\n", header->bth.opcode); + return -EINVAL; + } + + if (header->bth.transport_header_version != 0) { + pr_warn("Invalid BTH.transport_header_version %d\n", + header->bth.transport_header_version); + return -EINVAL; + } + + ib_unpack(deth_table, ARRAY_SIZE(deth_table), + buf, &header->deth); + buf += IB_DETH_BYTES; + + if (header->immediate_present) + memcpy(&header->immediate_data, buf, sizeof header->immediate_data); + + return 0; +} +EXPORT_SYMBOL(ib_ud_header_unpack); diff --git a/drivers/infiniband/core/umem.c b/drivers/infiniband/core/umem.c new file mode 100644 index 000000000..076526710 --- /dev/null +++ b/drivers/infiniband/core/umem.c @@ -0,0 +1,315 @@ +/* + * Copyright (c) 2005 Topspin Communications. All rights reserved. + * Copyright (c) 2005 Cisco Systems. All rights reserved. + * Copyright (c) 2005 Mellanox Technologies. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include <linux/mm.h> +#include <linux/dma-mapping.h> +#include <linux/sched/signal.h> +#include <linux/sched/mm.h> +#include <linux/export.h> +#include <linux/slab.h> +#include <linux/pagemap.h> +#include <linux/count_zeros.h> +#include <rdma/ib_umem_odp.h> + +#include "uverbs.h" + +static void __ib_umem_release(struct ib_device *dev, struct ib_umem *umem, int dirty) +{ + struct sg_page_iter sg_iter; + struct page *page; + + if (umem->nmap > 0) + ib_dma_unmap_sg(dev, umem->sg_head.sgl, umem->sg_nents, + DMA_BIDIRECTIONAL); + + for_each_sg_page(umem->sg_head.sgl, &sg_iter, umem->sg_nents, 0) { + page = sg_page_iter_page(&sg_iter); + unpin_user_pages_dirty_lock(&page, 1, umem->writable && dirty); + } + + sg_free_table(&umem->sg_head); +} + +/** + * ib_umem_find_best_pgsz - Find best HW page size to use for this MR + * + * @umem: umem struct + * @pgsz_bitmap: bitmap of HW supported page sizes + * @virt: IOVA + * + * This helper is intended for HW that support multiple page + * sizes but can do only a single page size in an MR. + * + * Returns 0 if the umem requires page sizes not supported by + * the driver to be mapped. Drivers always supporting PAGE_SIZE + * or smaller will never see a 0 result. + */ +unsigned long ib_umem_find_best_pgsz(struct ib_umem *umem, + unsigned long pgsz_bitmap, + unsigned long virt) +{ + struct scatterlist *sg; + unsigned long va, pgoff; + dma_addr_t mask; + int i; + + /* rdma_for_each_block() has a bug if the page size is smaller than the + * page size used to build the umem. For now prevent smaller page sizes + * from being returned. + */ + pgsz_bitmap &= GENMASK(BITS_PER_LONG - 1, PAGE_SHIFT); + + /* At minimum, drivers must support PAGE_SIZE or smaller */ + if (WARN_ON(!(pgsz_bitmap & GENMASK(PAGE_SHIFT, 0)))) + return 0; + + umem->iova = va = virt; + /* The best result is the smallest page size that results in the minimum + * number of required pages. Compute the largest page size that could + * work based on VA address bits that don't change. + */ + mask = pgsz_bitmap & + GENMASK(BITS_PER_LONG - 1, + bits_per((umem->length - 1 + virt) ^ virt)); + /* offset into first SGL */ + pgoff = umem->address & ~PAGE_MASK; + + for_each_sg(umem->sg_head.sgl, sg, umem->nmap, i) { + /* Walk SGL and reduce max page size if VA/PA bits differ + * for any address. + */ + mask |= (sg_dma_address(sg) + pgoff) ^ va; + va += sg_dma_len(sg) - pgoff; + /* Except for the last entry, the ending iova alignment sets + * the maximum possible page size as the low bits of the iova + * must be zero when starting the next chunk. + */ + if (i != (umem->nmap - 1)) + mask |= va; + pgoff = 0; + } + + /* The mask accumulates 1's in each position where the VA and physical + * address differ, thus the length of trailing 0 is the largest page + * size that can pass the VA through to the physical. + */ + if (mask) + pgsz_bitmap &= GENMASK(count_trailing_zeros(mask), 0); + return pgsz_bitmap ? rounddown_pow_of_two(pgsz_bitmap) : 0; +} +EXPORT_SYMBOL(ib_umem_find_best_pgsz); + +/** + * ib_umem_get - Pin and DMA map userspace memory. + * + * @device: IB device to connect UMEM + * @addr: userspace virtual address to start at + * @size: length of region to pin + * @access: IB_ACCESS_xxx flags for memory being pinned + */ +struct ib_umem *ib_umem_get(struct ib_device *device, unsigned long addr, + size_t size, int access) +{ + struct ib_umem *umem; + struct page **page_list; + unsigned long lock_limit; + unsigned long new_pinned; + unsigned long cur_base; + unsigned long dma_attr = 0; + struct mm_struct *mm; + unsigned long npages; + int ret; + struct scatterlist *sg = NULL; + unsigned int gup_flags = FOLL_WRITE; + + /* + * If the combination of the addr and size requested for this memory + * region causes an integer overflow, return error. + */ + if (((addr + size) < addr) || + PAGE_ALIGN(addr + size) < (addr + size)) + return ERR_PTR(-EINVAL); + + if (!can_do_mlock()) + return ERR_PTR(-EPERM); + + if (access & IB_ACCESS_ON_DEMAND) + return ERR_PTR(-EOPNOTSUPP); + + umem = kzalloc(sizeof(*umem), GFP_KERNEL); + if (!umem) + return ERR_PTR(-ENOMEM); + umem->ibdev = device; + umem->length = size; + umem->address = addr; + /* + * Drivers should call ib_umem_find_best_pgsz() to set the iova + * correctly. + */ + umem->iova = addr; + umem->writable = ib_access_writable(access); + umem->owning_mm = mm = current->mm; + mmgrab(mm); + + page_list = (struct page **) __get_free_page(GFP_KERNEL); + if (!page_list) { + ret = -ENOMEM; + goto umem_kfree; + } + + npages = ib_umem_num_pages(umem); + if (npages == 0 || npages > UINT_MAX) { + ret = -EINVAL; + goto out; + } + + lock_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT; + + new_pinned = atomic64_add_return(npages, &mm->pinned_vm); + if (new_pinned > lock_limit && !capable(CAP_IPC_LOCK)) { + atomic64_sub(npages, &mm->pinned_vm); + ret = -ENOMEM; + goto out; + } + + cur_base = addr & PAGE_MASK; + + if (!umem->writable) + gup_flags |= FOLL_FORCE; + + while (npages) { + cond_resched(); + ret = pin_user_pages_fast(cur_base, + min_t(unsigned long, npages, + PAGE_SIZE / + sizeof(struct page *)), + gup_flags | FOLL_LONGTERM, page_list); + if (ret < 0) + goto umem_release; + + cur_base += ret * PAGE_SIZE; + npages -= ret; + sg = __sg_alloc_table_from_pages(&umem->sg_head, page_list, ret, + 0, ret << PAGE_SHIFT, + ib_dma_max_seg_size(device), sg, npages, + GFP_KERNEL); + umem->sg_nents = umem->sg_head.nents; + if (IS_ERR(sg)) { + unpin_user_pages_dirty_lock(page_list, ret, 0); + ret = PTR_ERR(sg); + goto umem_release; + } + } + + if (access & IB_ACCESS_RELAXED_ORDERING) + dma_attr |= DMA_ATTR_WEAK_ORDERING; + + umem->nmap = + ib_dma_map_sg_attrs(device, umem->sg_head.sgl, umem->sg_nents, + DMA_BIDIRECTIONAL, dma_attr); + + if (!umem->nmap) { + ret = -ENOMEM; + goto umem_release; + } + + ret = 0; + goto out; + +umem_release: + __ib_umem_release(device, umem, 0); + atomic64_sub(ib_umem_num_pages(umem), &mm->pinned_vm); +out: + free_page((unsigned long) page_list); +umem_kfree: + if (ret) { + mmdrop(umem->owning_mm); + kfree(umem); + } + return ret ? ERR_PTR(ret) : umem; +} +EXPORT_SYMBOL(ib_umem_get); + +/** + * ib_umem_release - release memory pinned with ib_umem_get + * @umem: umem struct to release + */ +void ib_umem_release(struct ib_umem *umem) +{ + if (!umem) + return; + if (umem->is_odp) + return ib_umem_odp_release(to_ib_umem_odp(umem)); + + __ib_umem_release(umem->ibdev, umem, 1); + + atomic64_sub(ib_umem_num_pages(umem), &umem->owning_mm->pinned_vm); + mmdrop(umem->owning_mm); + kfree(umem); +} +EXPORT_SYMBOL(ib_umem_release); + +/* + * Copy from the given ib_umem's pages to the given buffer. + * + * umem - the umem to copy from + * offset - offset to start copying from + * dst - destination buffer + * length - buffer length + * + * Returns 0 on success, or an error code. + */ +int ib_umem_copy_from(void *dst, struct ib_umem *umem, size_t offset, + size_t length) +{ + size_t end = offset + length; + int ret; + + if (offset > umem->length || length > umem->length - offset) { + pr_err("ib_umem_copy_from not in range. offset: %zd umem length: %zd end: %zd\n", + offset, umem->length, end); + return -EINVAL; + } + + ret = sg_pcopy_to_buffer(umem->sg_head.sgl, umem->sg_nents, dst, length, + offset + ib_umem_offset(umem)); + + if (ret < 0) + return ret; + else if (ret != length) + return -EINVAL; + else + return 0; +} +EXPORT_SYMBOL(ib_umem_copy_from); diff --git a/drivers/infiniband/core/umem_odp.c b/drivers/infiniband/core/umem_odp.c new file mode 100644 index 000000000..af4af4789 --- /dev/null +++ b/drivers/infiniband/core/umem_odp.c @@ -0,0 +1,521 @@ +/* + * Copyright (c) 2014 Mellanox Technologies. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include <linux/types.h> +#include <linux/sched.h> +#include <linux/sched/mm.h> +#include <linux/sched/task.h> +#include <linux/pid.h> +#include <linux/slab.h> +#include <linux/export.h> +#include <linux/vmalloc.h> +#include <linux/hugetlb.h> +#include <linux/interval_tree.h> +#include <linux/hmm.h> +#include <linux/pagemap.h> + +#include <rdma/ib_verbs.h> +#include <rdma/ib_umem.h> +#include <rdma/ib_umem_odp.h> + +#include "uverbs.h" + +static inline int ib_init_umem_odp(struct ib_umem_odp *umem_odp, + const struct mmu_interval_notifier_ops *ops) +{ + int ret; + + umem_odp->umem.is_odp = 1; + mutex_init(&umem_odp->umem_mutex); + + if (!umem_odp->is_implicit_odp) { + size_t page_size = 1UL << umem_odp->page_shift; + unsigned long start; + unsigned long end; + size_t ndmas, npfns; + + start = ALIGN_DOWN(umem_odp->umem.address, page_size); + if (check_add_overflow(umem_odp->umem.address, + (unsigned long)umem_odp->umem.length, + &end)) + return -EOVERFLOW; + end = ALIGN(end, page_size); + if (unlikely(end < page_size)) + return -EOVERFLOW; + + ndmas = (end - start) >> umem_odp->page_shift; + if (!ndmas) + return -EINVAL; + + npfns = (end - start) >> PAGE_SHIFT; + umem_odp->pfn_list = kvcalloc( + npfns, sizeof(*umem_odp->pfn_list), GFP_KERNEL); + if (!umem_odp->pfn_list) + return -ENOMEM; + + umem_odp->dma_list = kvcalloc( + ndmas, sizeof(*umem_odp->dma_list), GFP_KERNEL); + if (!umem_odp->dma_list) { + ret = -ENOMEM; + goto out_pfn_list; + } + + ret = mmu_interval_notifier_insert(&umem_odp->notifier, + umem_odp->umem.owning_mm, + start, end - start, ops); + if (ret) + goto out_dma_list; + } + + return 0; + +out_dma_list: + kvfree(umem_odp->dma_list); +out_pfn_list: + kvfree(umem_odp->pfn_list); + return ret; +} + +/** + * ib_umem_odp_alloc_implicit - Allocate a parent implicit ODP umem + * + * Implicit ODP umems do not have a VA range and do not have any page lists. + * They exist only to hold the per_mm reference to help the driver create + * children umems. + * + * @device: IB device to create UMEM + * @access: ib_reg_mr access flags + */ +struct ib_umem_odp *ib_umem_odp_alloc_implicit(struct ib_device *device, + int access) +{ + struct ib_umem *umem; + struct ib_umem_odp *umem_odp; + int ret; + + if (access & IB_ACCESS_HUGETLB) + return ERR_PTR(-EINVAL); + + umem_odp = kzalloc(sizeof(*umem_odp), GFP_KERNEL); + if (!umem_odp) + return ERR_PTR(-ENOMEM); + umem = &umem_odp->umem; + umem->ibdev = device; + umem->writable = ib_access_writable(access); + umem->owning_mm = current->mm; + umem_odp->is_implicit_odp = 1; + umem_odp->page_shift = PAGE_SHIFT; + + umem_odp->tgid = get_task_pid(current->group_leader, PIDTYPE_PID); + ret = ib_init_umem_odp(umem_odp, NULL); + if (ret) { + put_pid(umem_odp->tgid); + kfree(umem_odp); + return ERR_PTR(ret); + } + return umem_odp; +} +EXPORT_SYMBOL(ib_umem_odp_alloc_implicit); + +/** + * ib_umem_odp_alloc_child - Allocate a child ODP umem under an implicit + * parent ODP umem + * + * @root: The parent umem enclosing the child. This must be allocated using + * ib_alloc_implicit_odp_umem() + * @addr: The starting userspace VA + * @size: The length of the userspace VA + * @ops: MMU interval ops, currently only @invalidate + */ +struct ib_umem_odp * +ib_umem_odp_alloc_child(struct ib_umem_odp *root, unsigned long addr, + size_t size, + const struct mmu_interval_notifier_ops *ops) +{ + /* + * Caller must ensure that root cannot be freed during the call to + * ib_alloc_odp_umem. + */ + struct ib_umem_odp *odp_data; + struct ib_umem *umem; + int ret; + + if (WARN_ON(!root->is_implicit_odp)) + return ERR_PTR(-EINVAL); + + odp_data = kzalloc(sizeof(*odp_data), GFP_KERNEL); + if (!odp_data) + return ERR_PTR(-ENOMEM); + umem = &odp_data->umem; + umem->ibdev = root->umem.ibdev; + umem->length = size; + umem->address = addr; + umem->writable = root->umem.writable; + umem->owning_mm = root->umem.owning_mm; + odp_data->page_shift = PAGE_SHIFT; + odp_data->notifier.ops = ops; + + /* + * A mmget must be held when registering a notifier, the owming_mm only + * has a mm_grab at this point. + */ + if (!mmget_not_zero(umem->owning_mm)) { + ret = -EFAULT; + goto out_free; + } + + odp_data->tgid = get_pid(root->tgid); + ret = ib_init_umem_odp(odp_data, ops); + if (ret) + goto out_tgid; + mmput(umem->owning_mm); + return odp_data; + +out_tgid: + put_pid(odp_data->tgid); + mmput(umem->owning_mm); +out_free: + kfree(odp_data); + return ERR_PTR(ret); +} +EXPORT_SYMBOL(ib_umem_odp_alloc_child); + +/** + * ib_umem_odp_get - Create a umem_odp for a userspace va + * + * @device: IB device struct to get UMEM + * @addr: userspace virtual address to start at + * @size: length of region to pin + * @access: IB_ACCESS_xxx flags for memory being pinned + * @ops: MMU interval ops, currently only @invalidate + * + * The driver should use when the access flags indicate ODP memory. It avoids + * pinning, instead, stores the mm for future page fault handling in + * conjunction with MMU notifiers. + */ +struct ib_umem_odp *ib_umem_odp_get(struct ib_device *device, + unsigned long addr, size_t size, int access, + const struct mmu_interval_notifier_ops *ops) +{ + struct ib_umem_odp *umem_odp; + struct mm_struct *mm; + int ret; + + if (WARN_ON_ONCE(!(access & IB_ACCESS_ON_DEMAND))) + return ERR_PTR(-EINVAL); + + umem_odp = kzalloc(sizeof(struct ib_umem_odp), GFP_KERNEL); + if (!umem_odp) + return ERR_PTR(-ENOMEM); + + umem_odp->umem.ibdev = device; + umem_odp->umem.length = size; + umem_odp->umem.address = addr; + umem_odp->umem.writable = ib_access_writable(access); + umem_odp->umem.owning_mm = mm = current->mm; + umem_odp->notifier.ops = ops; + + umem_odp->page_shift = PAGE_SHIFT; +#ifdef CONFIG_HUGETLB_PAGE + if (access & IB_ACCESS_HUGETLB) + umem_odp->page_shift = HPAGE_SHIFT; +#endif + + umem_odp->tgid = get_task_pid(current->group_leader, PIDTYPE_PID); + ret = ib_init_umem_odp(umem_odp, ops); + if (ret) + goto err_put_pid; + return umem_odp; + +err_put_pid: + put_pid(umem_odp->tgid); + kfree(umem_odp); + return ERR_PTR(ret); +} +EXPORT_SYMBOL(ib_umem_odp_get); + +void ib_umem_odp_release(struct ib_umem_odp *umem_odp) +{ + /* + * Ensure that no more pages are mapped in the umem. + * + * It is the driver's responsibility to ensure, before calling us, + * that the hardware will not attempt to access the MR any more. + */ + if (!umem_odp->is_implicit_odp) { + mutex_lock(&umem_odp->umem_mutex); + ib_umem_odp_unmap_dma_pages(umem_odp, ib_umem_start(umem_odp), + ib_umem_end(umem_odp)); + mutex_unlock(&umem_odp->umem_mutex); + mmu_interval_notifier_remove(&umem_odp->notifier); + kvfree(umem_odp->dma_list); + kvfree(umem_odp->pfn_list); + } + put_pid(umem_odp->tgid); + kfree(umem_odp); +} +EXPORT_SYMBOL(ib_umem_odp_release); + +/* + * Map for DMA and insert a single page into the on-demand paging page tables. + * + * @umem: the umem to insert the page to. + * @dma_index: index in the umem to add the dma to. + * @page: the page struct to map and add. + * @access_mask: access permissions needed for this page. + * @current_seq: sequence number for synchronization with invalidations. + * the sequence number is taken from + * umem_odp->notifiers_seq. + * + * The function returns -EFAULT if the DMA mapping operation fails. + * + */ +static int ib_umem_odp_map_dma_single_page( + struct ib_umem_odp *umem_odp, + unsigned int dma_index, + struct page *page, + u64 access_mask) +{ + struct ib_device *dev = umem_odp->umem.ibdev; + dma_addr_t *dma_addr = &umem_odp->dma_list[dma_index]; + + if (*dma_addr) { + /* + * If the page is already dma mapped it means it went through + * a non-invalidating trasition, like read-only to writable. + * Resync the flags. + */ + *dma_addr = (*dma_addr & ODP_DMA_ADDR_MASK) | access_mask; + return 0; + } + + *dma_addr = ib_dma_map_page(dev, page, 0, 1 << umem_odp->page_shift, + DMA_BIDIRECTIONAL); + if (ib_dma_mapping_error(dev, *dma_addr)) { + *dma_addr = 0; + return -EFAULT; + } + umem_odp->npages++; + *dma_addr |= access_mask; + return 0; +} + +/** + * ib_umem_odp_map_dma_and_lock - DMA map userspace memory in an ODP MR and lock it. + * + * Maps the range passed in the argument to DMA addresses. + * The DMA addresses of the mapped pages is updated in umem_odp->dma_list. + * Upon success the ODP MR will be locked to let caller complete its device + * page table update. + * + * Returns the number of pages mapped in success, negative error code + * for failure. + * @umem_odp: the umem to map and pin + * @user_virt: the address from which we need to map. + * @bcnt: the minimal number of bytes to pin and map. The mapping might be + * bigger due to alignment, and may also be smaller in case of an error + * pinning or mapping a page. The actual pages mapped is returned in + * the return value. + * @access_mask: bit mask of the requested access permissions for the given + * range. + * @fault: is faulting required for the given range + */ +int ib_umem_odp_map_dma_and_lock(struct ib_umem_odp *umem_odp, u64 user_virt, + u64 bcnt, u64 access_mask, bool fault) + __acquires(&umem_odp->umem_mutex) +{ + struct task_struct *owning_process = NULL; + struct mm_struct *owning_mm = umem_odp->umem.owning_mm; + int pfn_index, dma_index, ret = 0, start_idx; + unsigned int page_shift, hmm_order, pfn_start_idx; + unsigned long num_pfns, current_seq; + struct hmm_range range = {}; + unsigned long timeout; + + if (access_mask == 0) + return -EINVAL; + + if (user_virt < ib_umem_start(umem_odp) || + user_virt + bcnt > ib_umem_end(umem_odp)) + return -EFAULT; + + page_shift = umem_odp->page_shift; + + /* + * owning_process is allowed to be NULL, this means somehow the mm is + * existing beyond the lifetime of the originating process.. Presumably + * mmget_not_zero will fail in this case. + */ + owning_process = get_pid_task(umem_odp->tgid, PIDTYPE_PID); + if (!owning_process || !mmget_not_zero(owning_mm)) { + ret = -EINVAL; + goto out_put_task; + } + + range.notifier = &umem_odp->notifier; + range.start = ALIGN_DOWN(user_virt, 1UL << page_shift); + range.end = ALIGN(user_virt + bcnt, 1UL << page_shift); + pfn_start_idx = (range.start - ib_umem_start(umem_odp)) >> PAGE_SHIFT; + num_pfns = (range.end - range.start) >> PAGE_SHIFT; + if (fault) { + range.default_flags = HMM_PFN_REQ_FAULT; + + if (access_mask & ODP_WRITE_ALLOWED_BIT) + range.default_flags |= HMM_PFN_REQ_WRITE; + } + + range.hmm_pfns = &(umem_odp->pfn_list[pfn_start_idx]); + timeout = jiffies + msecs_to_jiffies(HMM_RANGE_DEFAULT_TIMEOUT); + +retry: + current_seq = range.notifier_seq = + mmu_interval_read_begin(&umem_odp->notifier); + + mmap_read_lock(owning_mm); + ret = hmm_range_fault(&range); + mmap_read_unlock(owning_mm); + if (unlikely(ret)) { + if (ret == -EBUSY && !time_after(jiffies, timeout)) + goto retry; + goto out_put_mm; + } + + start_idx = (range.start - ib_umem_start(umem_odp)) >> page_shift; + dma_index = start_idx; + + mutex_lock(&umem_odp->umem_mutex); + if (mmu_interval_read_retry(&umem_odp->notifier, current_seq)) { + mutex_unlock(&umem_odp->umem_mutex); + goto retry; + } + + for (pfn_index = 0; pfn_index < num_pfns; + pfn_index += 1 << (page_shift - PAGE_SHIFT), dma_index++) { + + if (fault) { + /* + * Since we asked for hmm_range_fault() to populate + * pages it shouldn't return an error entry on success. + */ + WARN_ON(range.hmm_pfns[pfn_index] & HMM_PFN_ERROR); + WARN_ON(!(range.hmm_pfns[pfn_index] & HMM_PFN_VALID)); + } else { + if (!(range.hmm_pfns[pfn_index] & HMM_PFN_VALID)) { + WARN_ON(umem_odp->dma_list[dma_index]); + continue; + } + access_mask = ODP_READ_ALLOWED_BIT; + if (range.hmm_pfns[pfn_index] & HMM_PFN_WRITE) + access_mask |= ODP_WRITE_ALLOWED_BIT; + } + + hmm_order = hmm_pfn_to_map_order(range.hmm_pfns[pfn_index]); + /* If a hugepage was detected and ODP wasn't set for, the umem + * page_shift will be used, the opposite case is an error. + */ + if (hmm_order + PAGE_SHIFT < page_shift) { + ret = -EINVAL; + ibdev_dbg(umem_odp->umem.ibdev, + "%s: un-expected hmm_order %d, page_shift %d\n", + __func__, hmm_order, page_shift); + break; + } + + ret = ib_umem_odp_map_dma_single_page( + umem_odp, dma_index, hmm_pfn_to_page(range.hmm_pfns[pfn_index]), + access_mask); + if (ret < 0) { + ibdev_dbg(umem_odp->umem.ibdev, + "ib_umem_odp_map_dma_single_page failed with error %d\n", ret); + break; + } + } + /* upon sucesss lock should stay on hold for the callee */ + if (!ret) + ret = dma_index - start_idx; + else + mutex_unlock(&umem_odp->umem_mutex); + +out_put_mm: + mmput_async(owning_mm); +out_put_task: + if (owning_process) + put_task_struct(owning_process); + return ret; +} +EXPORT_SYMBOL(ib_umem_odp_map_dma_and_lock); + +void ib_umem_odp_unmap_dma_pages(struct ib_umem_odp *umem_odp, u64 virt, + u64 bound) +{ + dma_addr_t dma_addr; + dma_addr_t dma; + int idx; + u64 addr; + struct ib_device *dev = umem_odp->umem.ibdev; + + lockdep_assert_held(&umem_odp->umem_mutex); + + virt = max_t(u64, virt, ib_umem_start(umem_odp)); + bound = min_t(u64, bound, ib_umem_end(umem_odp)); + for (addr = virt; addr < bound; addr += BIT(umem_odp->page_shift)) { + idx = (addr - ib_umem_start(umem_odp)) >> umem_odp->page_shift; + dma = umem_odp->dma_list[idx]; + + /* The access flags guaranteed a valid DMA address in case was NULL */ + if (dma) { + unsigned long pfn_idx = (addr - ib_umem_start(umem_odp)) >> PAGE_SHIFT; + struct page *page = hmm_pfn_to_page(umem_odp->pfn_list[pfn_idx]); + + dma_addr = dma & ODP_DMA_ADDR_MASK; + ib_dma_unmap_page(dev, dma_addr, + BIT(umem_odp->page_shift), + DMA_BIDIRECTIONAL); + if (dma & ODP_WRITE_ALLOWED_BIT) { + struct page *head_page = compound_head(page); + /* + * set_page_dirty prefers being called with + * the page lock. However, MMU notifiers are + * called sometimes with and sometimes without + * the lock. We rely on the umem_mutex instead + * to prevent other mmu notifiers from + * continuing and allowing the page mapping to + * be removed. + */ + set_page_dirty(head_page); + } + umem_odp->dma_list[idx] = 0; + umem_odp->npages--; + } + } +} +EXPORT_SYMBOL(ib_umem_odp_unmap_dma_pages); diff --git a/drivers/infiniband/core/user_mad.c b/drivers/infiniband/core/user_mad.c new file mode 100644 index 000000000..3bd0dcde8 --- /dev/null +++ b/drivers/infiniband/core/user_mad.c @@ -0,0 +1,1504 @@ +/* + * Copyright (c) 2004 Topspin Communications. All rights reserved. + * Copyright (c) 2005 Voltaire, Inc. All rights reserved. + * Copyright (c) 2005 Sun Microsystems, Inc. All rights reserved. + * Copyright (c) 2008 Cisco. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#define pr_fmt(fmt) "user_mad: " fmt + +#include <linux/module.h> +#include <linux/init.h> +#include <linux/device.h> +#include <linux/err.h> +#include <linux/fs.h> +#include <linux/cdev.h> +#include <linux/dma-mapping.h> +#include <linux/poll.h> +#include <linux/mutex.h> +#include <linux/kref.h> +#include <linux/compat.h> +#include <linux/sched.h> +#include <linux/semaphore.h> +#include <linux/slab.h> +#include <linux/nospec.h> + +#include <linux/uaccess.h> + +#include <rdma/ib_mad.h> +#include <rdma/ib_user_mad.h> +#include <rdma/rdma_netlink.h> + +#include "core_priv.h" + +MODULE_AUTHOR("Roland Dreier"); +MODULE_DESCRIPTION("InfiniBand userspace MAD packet access"); +MODULE_LICENSE("Dual BSD/GPL"); + +enum { + IB_UMAD_MAX_PORTS = RDMA_MAX_PORTS, + IB_UMAD_MAX_AGENTS = 32, + + IB_UMAD_MAJOR = 231, + IB_UMAD_MINOR_BASE = 0, + IB_UMAD_NUM_FIXED_MINOR = 64, + IB_UMAD_NUM_DYNAMIC_MINOR = IB_UMAD_MAX_PORTS - IB_UMAD_NUM_FIXED_MINOR, + IB_ISSM_MINOR_BASE = IB_UMAD_NUM_FIXED_MINOR, +}; + +/* + * Our lifetime rules for these structs are the following: + * device special file is opened, we take a reference on the + * ib_umad_port's struct ib_umad_device. We drop these + * references in the corresponding close(). + * + * In addition to references coming from open character devices, there + * is one more reference to each ib_umad_device representing the + * module's reference taken when allocating the ib_umad_device in + * ib_umad_add_one(). + * + * When destroying an ib_umad_device, we drop the module's reference. + */ + +struct ib_umad_port { + struct cdev cdev; + struct device dev; + struct cdev sm_cdev; + struct device sm_dev; + struct semaphore sm_sem; + + struct mutex file_mutex; + struct list_head file_list; + + struct ib_device *ib_dev; + struct ib_umad_device *umad_dev; + int dev_num; + u8 port_num; +}; + +struct ib_umad_device { + struct kref kref; + struct ib_umad_port ports[]; +}; + +struct ib_umad_file { + struct mutex mutex; + struct ib_umad_port *port; + struct list_head recv_list; + struct list_head send_list; + struct list_head port_list; + spinlock_t send_lock; + wait_queue_head_t recv_wait; + struct ib_mad_agent *agent[IB_UMAD_MAX_AGENTS]; + int agents_dead; + u8 use_pkey_index; + u8 already_used; +}; + +struct ib_umad_packet { + struct ib_mad_send_buf *msg; + struct ib_mad_recv_wc *recv_wc; + struct list_head list; + int length; + struct ib_user_mad mad; +}; + +struct ib_rmpp_mad_hdr { + struct ib_mad_hdr mad_hdr; + struct ib_rmpp_hdr rmpp_hdr; +} __packed; + +#define CREATE_TRACE_POINTS +#include <trace/events/ib_umad.h> + +static const dev_t base_umad_dev = MKDEV(IB_UMAD_MAJOR, IB_UMAD_MINOR_BASE); +static const dev_t base_issm_dev = MKDEV(IB_UMAD_MAJOR, IB_UMAD_MINOR_BASE) + + IB_UMAD_NUM_FIXED_MINOR; +static dev_t dynamic_umad_dev; +static dev_t dynamic_issm_dev; + +static DEFINE_IDA(umad_ida); + +static int ib_umad_add_one(struct ib_device *device); +static void ib_umad_remove_one(struct ib_device *device, void *client_data); + +static void ib_umad_dev_free(struct kref *kref) +{ + struct ib_umad_device *dev = + container_of(kref, struct ib_umad_device, kref); + + kfree(dev); +} + +static void ib_umad_dev_get(struct ib_umad_device *dev) +{ + kref_get(&dev->kref); +} + +static void ib_umad_dev_put(struct ib_umad_device *dev) +{ + kref_put(&dev->kref, ib_umad_dev_free); +} + +static int hdr_size(struct ib_umad_file *file) +{ + return file->use_pkey_index ? sizeof (struct ib_user_mad_hdr) : + sizeof (struct ib_user_mad_hdr_old); +} + +/* caller must hold file->mutex */ +static struct ib_mad_agent *__get_agent(struct ib_umad_file *file, int id) +{ + return file->agents_dead ? NULL : file->agent[id]; +} + +static int queue_packet(struct ib_umad_file *file, + struct ib_mad_agent *agent, + struct ib_umad_packet *packet) +{ + int ret = 1; + + mutex_lock(&file->mutex); + + for (packet->mad.hdr.id = 0; + packet->mad.hdr.id < IB_UMAD_MAX_AGENTS; + packet->mad.hdr.id++) + if (agent == __get_agent(file, packet->mad.hdr.id)) { + list_add_tail(&packet->list, &file->recv_list); + wake_up_interruptible(&file->recv_wait); + ret = 0; + break; + } + + mutex_unlock(&file->mutex); + + return ret; +} + +static void dequeue_send(struct ib_umad_file *file, + struct ib_umad_packet *packet) +{ + spin_lock_irq(&file->send_lock); + list_del(&packet->list); + spin_unlock_irq(&file->send_lock); +} + +static void send_handler(struct ib_mad_agent *agent, + struct ib_mad_send_wc *send_wc) +{ + struct ib_umad_file *file = agent->context; + struct ib_umad_packet *packet = send_wc->send_buf->context[0]; + + dequeue_send(file, packet); + rdma_destroy_ah(packet->msg->ah, RDMA_DESTROY_AH_SLEEPABLE); + ib_free_send_mad(packet->msg); + + if (send_wc->status == IB_WC_RESP_TIMEOUT_ERR) { + packet->length = IB_MGMT_MAD_HDR; + packet->mad.hdr.status = ETIMEDOUT; + if (!queue_packet(file, agent, packet)) + return; + } + kfree(packet); +} + +static void recv_handler(struct ib_mad_agent *agent, + struct ib_mad_send_buf *send_buf, + struct ib_mad_recv_wc *mad_recv_wc) +{ + struct ib_umad_file *file = agent->context; + struct ib_umad_packet *packet; + + if (mad_recv_wc->wc->status != IB_WC_SUCCESS) + goto err1; + + packet = kzalloc(sizeof *packet, GFP_KERNEL); + if (!packet) + goto err1; + + packet->length = mad_recv_wc->mad_len; + packet->recv_wc = mad_recv_wc; + + packet->mad.hdr.status = 0; + packet->mad.hdr.length = hdr_size(file) + mad_recv_wc->mad_len; + packet->mad.hdr.qpn = cpu_to_be32(mad_recv_wc->wc->src_qp); + /* + * On OPA devices it is okay to lose the upper 16 bits of LID as this + * information is obtained elsewhere. Mask off the upper 16 bits. + */ + if (rdma_cap_opa_mad(agent->device, agent->port_num)) + packet->mad.hdr.lid = ib_lid_be16(0xFFFF & + mad_recv_wc->wc->slid); + else + packet->mad.hdr.lid = ib_lid_be16(mad_recv_wc->wc->slid); + packet->mad.hdr.sl = mad_recv_wc->wc->sl; + packet->mad.hdr.path_bits = mad_recv_wc->wc->dlid_path_bits; + packet->mad.hdr.pkey_index = mad_recv_wc->wc->pkey_index; + packet->mad.hdr.grh_present = !!(mad_recv_wc->wc->wc_flags & IB_WC_GRH); + if (packet->mad.hdr.grh_present) { + struct rdma_ah_attr ah_attr; + const struct ib_global_route *grh; + int ret; + + ret = ib_init_ah_attr_from_wc(agent->device, agent->port_num, + mad_recv_wc->wc, + mad_recv_wc->recv_buf.grh, + &ah_attr); + if (ret) + goto err2; + + grh = rdma_ah_read_grh(&ah_attr); + packet->mad.hdr.gid_index = grh->sgid_index; + packet->mad.hdr.hop_limit = grh->hop_limit; + packet->mad.hdr.traffic_class = grh->traffic_class; + memcpy(packet->mad.hdr.gid, &grh->dgid, 16); + packet->mad.hdr.flow_label = cpu_to_be32(grh->flow_label); + rdma_destroy_ah_attr(&ah_attr); + } + + if (queue_packet(file, agent, packet)) + goto err2; + return; + +err2: + kfree(packet); +err1: + ib_free_recv_mad(mad_recv_wc); +} + +static ssize_t copy_recv_mad(struct ib_umad_file *file, char __user *buf, + struct ib_umad_packet *packet, size_t count) +{ + struct ib_mad_recv_buf *recv_buf; + int left, seg_payload, offset, max_seg_payload; + size_t seg_size; + + recv_buf = &packet->recv_wc->recv_buf; + seg_size = packet->recv_wc->mad_seg_size; + + /* We need enough room to copy the first (or only) MAD segment. */ + if ((packet->length <= seg_size && + count < hdr_size(file) + packet->length) || + (packet->length > seg_size && + count < hdr_size(file) + seg_size)) + return -EINVAL; + + if (copy_to_user(buf, &packet->mad, hdr_size(file))) + return -EFAULT; + + buf += hdr_size(file); + seg_payload = min_t(int, packet->length, seg_size); + if (copy_to_user(buf, recv_buf->mad, seg_payload)) + return -EFAULT; + + if (seg_payload < packet->length) { + /* + * Multipacket RMPP MAD message. Copy remainder of message. + * Note that last segment may have a shorter payload. + */ + if (count < hdr_size(file) + packet->length) { + /* + * The buffer is too small, return the first RMPP segment, + * which includes the RMPP message length. + */ + return -ENOSPC; + } + offset = ib_get_mad_data_offset(recv_buf->mad->mad_hdr.mgmt_class); + max_seg_payload = seg_size - offset; + + for (left = packet->length - seg_payload, buf += seg_payload; + left; left -= seg_payload, buf += seg_payload) { + recv_buf = container_of(recv_buf->list.next, + struct ib_mad_recv_buf, list); + seg_payload = min(left, max_seg_payload); + if (copy_to_user(buf, ((void *) recv_buf->mad) + offset, + seg_payload)) + return -EFAULT; + } + } + + trace_ib_umad_read_recv(file, &packet->mad.hdr, &recv_buf->mad->mad_hdr); + + return hdr_size(file) + packet->length; +} + +static ssize_t copy_send_mad(struct ib_umad_file *file, char __user *buf, + struct ib_umad_packet *packet, size_t count) +{ + ssize_t size = hdr_size(file) + packet->length; + + if (count < size) + return -EINVAL; + + if (copy_to_user(buf, &packet->mad, hdr_size(file))) + return -EFAULT; + + buf += hdr_size(file); + + if (copy_to_user(buf, packet->mad.data, packet->length)) + return -EFAULT; + + trace_ib_umad_read_send(file, &packet->mad.hdr, + (struct ib_mad_hdr *)&packet->mad.data); + + return size; +} + +static ssize_t ib_umad_read(struct file *filp, char __user *buf, + size_t count, loff_t *pos) +{ + struct ib_umad_file *file = filp->private_data; + struct ib_umad_packet *packet; + ssize_t ret; + + if (count < hdr_size(file)) + return -EINVAL; + + mutex_lock(&file->mutex); + + if (file->agents_dead) { + mutex_unlock(&file->mutex); + return -EIO; + } + + while (list_empty(&file->recv_list)) { + mutex_unlock(&file->mutex); + + if (filp->f_flags & O_NONBLOCK) + return -EAGAIN; + + if (wait_event_interruptible(file->recv_wait, + !list_empty(&file->recv_list))) + return -ERESTARTSYS; + + mutex_lock(&file->mutex); + } + + if (file->agents_dead) { + mutex_unlock(&file->mutex); + return -EIO; + } + + packet = list_entry(file->recv_list.next, struct ib_umad_packet, list); + list_del(&packet->list); + + mutex_unlock(&file->mutex); + + if (packet->recv_wc) + ret = copy_recv_mad(file, buf, packet, count); + else + ret = copy_send_mad(file, buf, packet, count); + + if (ret < 0) { + /* Requeue packet */ + mutex_lock(&file->mutex); + list_add(&packet->list, &file->recv_list); + mutex_unlock(&file->mutex); + } else { + if (packet->recv_wc) + ib_free_recv_mad(packet->recv_wc); + kfree(packet); + } + return ret; +} + +static int copy_rmpp_mad(struct ib_mad_send_buf *msg, const char __user *buf) +{ + int left, seg; + + /* Copy class specific header */ + if ((msg->hdr_len > IB_MGMT_RMPP_HDR) && + copy_from_user(msg->mad + IB_MGMT_RMPP_HDR, buf + IB_MGMT_RMPP_HDR, + msg->hdr_len - IB_MGMT_RMPP_HDR)) + return -EFAULT; + + /* All headers are in place. Copy data segments. */ + for (seg = 1, left = msg->data_len, buf += msg->hdr_len; left > 0; + seg++, left -= msg->seg_size, buf += msg->seg_size) { + if (copy_from_user(ib_get_rmpp_segment(msg, seg), buf, + min(left, msg->seg_size))) + return -EFAULT; + } + return 0; +} + +static int same_destination(struct ib_user_mad_hdr *hdr1, + struct ib_user_mad_hdr *hdr2) +{ + if (!hdr1->grh_present && !hdr2->grh_present) + return (hdr1->lid == hdr2->lid); + + if (hdr1->grh_present && hdr2->grh_present) + return !memcmp(hdr1->gid, hdr2->gid, 16); + + return 0; +} + +static int is_duplicate(struct ib_umad_file *file, + struct ib_umad_packet *packet) +{ + struct ib_umad_packet *sent_packet; + struct ib_mad_hdr *sent_hdr, *hdr; + + hdr = (struct ib_mad_hdr *) packet->mad.data; + list_for_each_entry(sent_packet, &file->send_list, list) { + sent_hdr = (struct ib_mad_hdr *) sent_packet->mad.data; + + if ((hdr->tid != sent_hdr->tid) || + (hdr->mgmt_class != sent_hdr->mgmt_class)) + continue; + + /* + * No need to be overly clever here. If two new operations have + * the same TID, reject the second as a duplicate. This is more + * restrictive than required by the spec. + */ + if (!ib_response_mad(hdr)) { + if (!ib_response_mad(sent_hdr)) + return 1; + continue; + } else if (!ib_response_mad(sent_hdr)) + continue; + + if (same_destination(&packet->mad.hdr, &sent_packet->mad.hdr)) + return 1; + } + + return 0; +} + +static ssize_t ib_umad_write(struct file *filp, const char __user *buf, + size_t count, loff_t *pos) +{ + struct ib_umad_file *file = filp->private_data; + struct ib_rmpp_mad_hdr *rmpp_mad_hdr; + struct ib_umad_packet *packet; + struct ib_mad_agent *agent; + struct rdma_ah_attr ah_attr; + struct ib_ah *ah; + __be64 *tid; + int ret, data_len, hdr_len, copy_offset, rmpp_active; + u8 base_version; + + if (count < hdr_size(file) + IB_MGMT_RMPP_HDR) + return -EINVAL; + + packet = kzalloc(sizeof(*packet) + IB_MGMT_RMPP_HDR, GFP_KERNEL); + if (!packet) + return -ENOMEM; + + if (copy_from_user(&packet->mad, buf, hdr_size(file))) { + ret = -EFAULT; + goto err; + } + + if (packet->mad.hdr.id >= IB_UMAD_MAX_AGENTS) { + ret = -EINVAL; + goto err; + } + + buf += hdr_size(file); + + if (copy_from_user(packet->mad.data, buf, IB_MGMT_RMPP_HDR)) { + ret = -EFAULT; + goto err; + } + + mutex_lock(&file->mutex); + + trace_ib_umad_write(file, &packet->mad.hdr, + (struct ib_mad_hdr *)&packet->mad.data); + + agent = __get_agent(file, packet->mad.hdr.id); + if (!agent) { + ret = -EIO; + goto err_up; + } + + memset(&ah_attr, 0, sizeof ah_attr); + ah_attr.type = rdma_ah_find_type(agent->device, + file->port->port_num); + rdma_ah_set_dlid(&ah_attr, be16_to_cpu(packet->mad.hdr.lid)); + rdma_ah_set_sl(&ah_attr, packet->mad.hdr.sl); + rdma_ah_set_path_bits(&ah_attr, packet->mad.hdr.path_bits); + rdma_ah_set_port_num(&ah_attr, file->port->port_num); + if (packet->mad.hdr.grh_present) { + rdma_ah_set_grh(&ah_attr, NULL, + be32_to_cpu(packet->mad.hdr.flow_label), + packet->mad.hdr.gid_index, + packet->mad.hdr.hop_limit, + packet->mad.hdr.traffic_class); + rdma_ah_set_dgid_raw(&ah_attr, packet->mad.hdr.gid); + } + + ah = rdma_create_user_ah(agent->qp->pd, &ah_attr, NULL); + if (IS_ERR(ah)) { + ret = PTR_ERR(ah); + goto err_up; + } + + rmpp_mad_hdr = (struct ib_rmpp_mad_hdr *)packet->mad.data; + hdr_len = ib_get_mad_data_offset(rmpp_mad_hdr->mad_hdr.mgmt_class); + + if (ib_is_mad_class_rmpp(rmpp_mad_hdr->mad_hdr.mgmt_class) + && ib_mad_kernel_rmpp_agent(agent)) { + copy_offset = IB_MGMT_RMPP_HDR; + rmpp_active = ib_get_rmpp_flags(&rmpp_mad_hdr->rmpp_hdr) & + IB_MGMT_RMPP_FLAG_ACTIVE; + } else { + copy_offset = IB_MGMT_MAD_HDR; + rmpp_active = 0; + } + + base_version = ((struct ib_mad_hdr *)&packet->mad.data)->base_version; + data_len = count - hdr_size(file) - hdr_len; + packet->msg = ib_create_send_mad(agent, + be32_to_cpu(packet->mad.hdr.qpn), + packet->mad.hdr.pkey_index, rmpp_active, + hdr_len, data_len, GFP_KERNEL, + base_version); + if (IS_ERR(packet->msg)) { + ret = PTR_ERR(packet->msg); + goto err_ah; + } + + packet->msg->ah = ah; + packet->msg->timeout_ms = packet->mad.hdr.timeout_ms; + packet->msg->retries = packet->mad.hdr.retries; + packet->msg->context[0] = packet; + + /* Copy MAD header. Any RMPP header is already in place. */ + memcpy(packet->msg->mad, packet->mad.data, IB_MGMT_MAD_HDR); + + if (!rmpp_active) { + if (copy_from_user(packet->msg->mad + copy_offset, + buf + copy_offset, + hdr_len + data_len - copy_offset)) { + ret = -EFAULT; + goto err_msg; + } + } else { + ret = copy_rmpp_mad(packet->msg, buf); + if (ret) + goto err_msg; + } + + /* + * Set the high-order part of the transaction ID to make MADs from + * different agents unique, and allow routing responses back to the + * original requestor. + */ + if (!ib_response_mad(packet->msg->mad)) { + tid = &((struct ib_mad_hdr *) packet->msg->mad)->tid; + *tid = cpu_to_be64(((u64) agent->hi_tid) << 32 | + (be64_to_cpup(tid) & 0xffffffff)); + rmpp_mad_hdr->mad_hdr.tid = *tid; + } + + if (!ib_mad_kernel_rmpp_agent(agent) + && ib_is_mad_class_rmpp(rmpp_mad_hdr->mad_hdr.mgmt_class) + && (ib_get_rmpp_flags(&rmpp_mad_hdr->rmpp_hdr) & IB_MGMT_RMPP_FLAG_ACTIVE)) { + spin_lock_irq(&file->send_lock); + list_add_tail(&packet->list, &file->send_list); + spin_unlock_irq(&file->send_lock); + } else { + spin_lock_irq(&file->send_lock); + ret = is_duplicate(file, packet); + if (!ret) + list_add_tail(&packet->list, &file->send_list); + spin_unlock_irq(&file->send_lock); + if (ret) { + ret = -EINVAL; + goto err_msg; + } + } + + ret = ib_post_send_mad(packet->msg, NULL); + if (ret) + goto err_send; + + mutex_unlock(&file->mutex); + return count; + +err_send: + dequeue_send(file, packet); +err_msg: + ib_free_send_mad(packet->msg); +err_ah: + rdma_destroy_ah(ah, RDMA_DESTROY_AH_SLEEPABLE); +err_up: + mutex_unlock(&file->mutex); +err: + kfree(packet); + return ret; +} + +static __poll_t ib_umad_poll(struct file *filp, struct poll_table_struct *wait) +{ + struct ib_umad_file *file = filp->private_data; + + /* we will always be able to post a MAD send */ + __poll_t mask = EPOLLOUT | EPOLLWRNORM; + + mutex_lock(&file->mutex); + poll_wait(filp, &file->recv_wait, wait); + + if (!list_empty(&file->recv_list)) + mask |= EPOLLIN | EPOLLRDNORM; + if (file->agents_dead) + mask = EPOLLERR; + mutex_unlock(&file->mutex); + + return mask; +} + +static int ib_umad_reg_agent(struct ib_umad_file *file, void __user *arg, + int compat_method_mask) +{ + struct ib_user_mad_reg_req ureq; + struct ib_mad_reg_req req; + struct ib_mad_agent *agent = NULL; + int agent_id; + int ret; + + mutex_lock(&file->port->file_mutex); + mutex_lock(&file->mutex); + + if (!file->port->ib_dev) { + dev_notice(&file->port->dev, + "ib_umad_reg_agent: invalid device\n"); + ret = -EPIPE; + goto out; + } + + if (copy_from_user(&ureq, arg, sizeof ureq)) { + ret = -EFAULT; + goto out; + } + + if (ureq.qpn != 0 && ureq.qpn != 1) { + dev_notice(&file->port->dev, + "ib_umad_reg_agent: invalid QPN %d specified\n", + ureq.qpn); + ret = -EINVAL; + goto out; + } + + for (agent_id = 0; agent_id < IB_UMAD_MAX_AGENTS; ++agent_id) + if (!__get_agent(file, agent_id)) + goto found; + + dev_notice(&file->port->dev, + "ib_umad_reg_agent: Max Agents (%u) reached\n", + IB_UMAD_MAX_AGENTS); + ret = -ENOMEM; + goto out; + +found: + if (ureq.mgmt_class) { + memset(&req, 0, sizeof(req)); + req.mgmt_class = ureq.mgmt_class; + req.mgmt_class_version = ureq.mgmt_class_version; + memcpy(req.oui, ureq.oui, sizeof req.oui); + + if (compat_method_mask) { + u32 *umm = (u32 *) ureq.method_mask; + int i; + + for (i = 0; i < BITS_TO_LONGS(IB_MGMT_MAX_METHODS); ++i) + req.method_mask[i] = + umm[i * 2] | ((u64) umm[i * 2 + 1] << 32); + } else + memcpy(req.method_mask, ureq.method_mask, + sizeof req.method_mask); + } + + agent = ib_register_mad_agent(file->port->ib_dev, file->port->port_num, + ureq.qpn ? IB_QPT_GSI : IB_QPT_SMI, + ureq.mgmt_class ? &req : NULL, + ureq.rmpp_version, + send_handler, recv_handler, file, 0); + if (IS_ERR(agent)) { + ret = PTR_ERR(agent); + agent = NULL; + goto out; + } + + if (put_user(agent_id, + (u32 __user *) (arg + offsetof(struct ib_user_mad_reg_req, id)))) { + ret = -EFAULT; + goto out; + } + + if (!file->already_used) { + file->already_used = 1; + if (!file->use_pkey_index) { + dev_warn(&file->port->dev, + "process %s did not enable P_Key index support.\n", + current->comm); + dev_warn(&file->port->dev, + " Documentation/infiniband/user_mad.rst has info on the new ABI.\n"); + } + } + + file->agent[agent_id] = agent; + ret = 0; + +out: + mutex_unlock(&file->mutex); + + if (ret && agent) + ib_unregister_mad_agent(agent); + + mutex_unlock(&file->port->file_mutex); + + return ret; +} + +static int ib_umad_reg_agent2(struct ib_umad_file *file, void __user *arg) +{ + struct ib_user_mad_reg_req2 ureq; + struct ib_mad_reg_req req; + struct ib_mad_agent *agent = NULL; + int agent_id; + int ret; + + mutex_lock(&file->port->file_mutex); + mutex_lock(&file->mutex); + + if (!file->port->ib_dev) { + dev_notice(&file->port->dev, + "ib_umad_reg_agent2: invalid device\n"); + ret = -EPIPE; + goto out; + } + + if (copy_from_user(&ureq, arg, sizeof(ureq))) { + ret = -EFAULT; + goto out; + } + + if (ureq.qpn != 0 && ureq.qpn != 1) { + dev_notice(&file->port->dev, + "ib_umad_reg_agent2: invalid QPN %d specified\n", + ureq.qpn); + ret = -EINVAL; + goto out; + } + + if (ureq.flags & ~IB_USER_MAD_REG_FLAGS_CAP) { + dev_notice(&file->port->dev, + "ib_umad_reg_agent2 failed: invalid registration flags specified 0x%x; supported 0x%x\n", + ureq.flags, IB_USER_MAD_REG_FLAGS_CAP); + ret = -EINVAL; + + if (put_user((u32)IB_USER_MAD_REG_FLAGS_CAP, + (u32 __user *) (arg + offsetof(struct + ib_user_mad_reg_req2, flags)))) + ret = -EFAULT; + + goto out; + } + + for (agent_id = 0; agent_id < IB_UMAD_MAX_AGENTS; ++agent_id) + if (!__get_agent(file, agent_id)) + goto found; + + dev_notice(&file->port->dev, + "ib_umad_reg_agent2: Max Agents (%u) reached\n", + IB_UMAD_MAX_AGENTS); + ret = -ENOMEM; + goto out; + +found: + if (ureq.mgmt_class) { + memset(&req, 0, sizeof(req)); + req.mgmt_class = ureq.mgmt_class; + req.mgmt_class_version = ureq.mgmt_class_version; + if (ureq.oui & 0xff000000) { + dev_notice(&file->port->dev, + "ib_umad_reg_agent2 failed: oui invalid 0x%08x\n", + ureq.oui); + ret = -EINVAL; + goto out; + } + req.oui[2] = ureq.oui & 0x0000ff; + req.oui[1] = (ureq.oui & 0x00ff00) >> 8; + req.oui[0] = (ureq.oui & 0xff0000) >> 16; + memcpy(req.method_mask, ureq.method_mask, + sizeof(req.method_mask)); + } + + agent = ib_register_mad_agent(file->port->ib_dev, file->port->port_num, + ureq.qpn ? IB_QPT_GSI : IB_QPT_SMI, + ureq.mgmt_class ? &req : NULL, + ureq.rmpp_version, + send_handler, recv_handler, file, + ureq.flags); + if (IS_ERR(agent)) { + ret = PTR_ERR(agent); + agent = NULL; + goto out; + } + + if (put_user(agent_id, + (u32 __user *)(arg + + offsetof(struct ib_user_mad_reg_req2, id)))) { + ret = -EFAULT; + goto out; + } + + if (!file->already_used) { + file->already_used = 1; + file->use_pkey_index = 1; + } + + file->agent[agent_id] = agent; + ret = 0; + +out: + mutex_unlock(&file->mutex); + + if (ret && agent) + ib_unregister_mad_agent(agent); + + mutex_unlock(&file->port->file_mutex); + + return ret; +} + + +static int ib_umad_unreg_agent(struct ib_umad_file *file, u32 __user *arg) +{ + struct ib_mad_agent *agent = NULL; + u32 id; + int ret = 0; + + if (get_user(id, arg)) + return -EFAULT; + if (id >= IB_UMAD_MAX_AGENTS) + return -EINVAL; + + mutex_lock(&file->port->file_mutex); + mutex_lock(&file->mutex); + + id = array_index_nospec(id, IB_UMAD_MAX_AGENTS); + if (!__get_agent(file, id)) { + ret = -EINVAL; + goto out; + } + + agent = file->agent[id]; + file->agent[id] = NULL; + +out: + mutex_unlock(&file->mutex); + + if (agent) + ib_unregister_mad_agent(agent); + + mutex_unlock(&file->port->file_mutex); + + return ret; +} + +static long ib_umad_enable_pkey(struct ib_umad_file *file) +{ + int ret = 0; + + mutex_lock(&file->mutex); + if (file->already_used) + ret = -EINVAL; + else + file->use_pkey_index = 1; + mutex_unlock(&file->mutex); + + return ret; +} + +static long ib_umad_ioctl(struct file *filp, unsigned int cmd, + unsigned long arg) +{ + switch (cmd) { + case IB_USER_MAD_REGISTER_AGENT: + return ib_umad_reg_agent(filp->private_data, (void __user *) arg, 0); + case IB_USER_MAD_UNREGISTER_AGENT: + return ib_umad_unreg_agent(filp->private_data, (__u32 __user *) arg); + case IB_USER_MAD_ENABLE_PKEY: + return ib_umad_enable_pkey(filp->private_data); + case IB_USER_MAD_REGISTER_AGENT2: + return ib_umad_reg_agent2(filp->private_data, (void __user *) arg); + default: + return -ENOIOCTLCMD; + } +} + +#ifdef CONFIG_COMPAT +static long ib_umad_compat_ioctl(struct file *filp, unsigned int cmd, + unsigned long arg) +{ + switch (cmd) { + case IB_USER_MAD_REGISTER_AGENT: + return ib_umad_reg_agent(filp->private_data, compat_ptr(arg), 1); + case IB_USER_MAD_UNREGISTER_AGENT: + return ib_umad_unreg_agent(filp->private_data, compat_ptr(arg)); + case IB_USER_MAD_ENABLE_PKEY: + return ib_umad_enable_pkey(filp->private_data); + case IB_USER_MAD_REGISTER_AGENT2: + return ib_umad_reg_agent2(filp->private_data, compat_ptr(arg)); + default: + return -ENOIOCTLCMD; + } +} +#endif + +/* + * ib_umad_open() does not need the BKL: + * + * - the ib_umad_port structures are properly reference counted, and + * everything else is purely local to the file being created, so + * races against other open calls are not a problem; + * - the ioctl method does not affect any global state outside of the + * file structure being operated on; + */ +static int ib_umad_open(struct inode *inode, struct file *filp) +{ + struct ib_umad_port *port; + struct ib_umad_file *file; + int ret = 0; + + port = container_of(inode->i_cdev, struct ib_umad_port, cdev); + + mutex_lock(&port->file_mutex); + + if (!port->ib_dev) { + ret = -ENXIO; + goto out; + } + + if (!rdma_dev_access_netns(port->ib_dev, current->nsproxy->net_ns)) { + ret = -EPERM; + goto out; + } + + file = kzalloc(sizeof(*file), GFP_KERNEL); + if (!file) { + ret = -ENOMEM; + goto out; + } + + mutex_init(&file->mutex); + spin_lock_init(&file->send_lock); + INIT_LIST_HEAD(&file->recv_list); + INIT_LIST_HEAD(&file->send_list); + init_waitqueue_head(&file->recv_wait); + + file->port = port; + filp->private_data = file; + + list_add_tail(&file->port_list, &port->file_list); + + stream_open(inode, filp); +out: + mutex_unlock(&port->file_mutex); + return ret; +} + +static int ib_umad_close(struct inode *inode, struct file *filp) +{ + struct ib_umad_file *file = filp->private_data; + struct ib_umad_packet *packet, *tmp; + int already_dead; + int i; + + mutex_lock(&file->port->file_mutex); + mutex_lock(&file->mutex); + + already_dead = file->agents_dead; + file->agents_dead = 1; + + list_for_each_entry_safe(packet, tmp, &file->recv_list, list) { + if (packet->recv_wc) + ib_free_recv_mad(packet->recv_wc); + kfree(packet); + } + + list_del(&file->port_list); + + mutex_unlock(&file->mutex); + + if (!already_dead) + for (i = 0; i < IB_UMAD_MAX_AGENTS; ++i) + if (file->agent[i]) + ib_unregister_mad_agent(file->agent[i]); + + mutex_unlock(&file->port->file_mutex); + mutex_destroy(&file->mutex); + kfree(file); + return 0; +} + +static const struct file_operations umad_fops = { + .owner = THIS_MODULE, + .read = ib_umad_read, + .write = ib_umad_write, + .poll = ib_umad_poll, + .unlocked_ioctl = ib_umad_ioctl, +#ifdef CONFIG_COMPAT + .compat_ioctl = ib_umad_compat_ioctl, +#endif + .open = ib_umad_open, + .release = ib_umad_close, + .llseek = no_llseek, +}; + +static int ib_umad_sm_open(struct inode *inode, struct file *filp) +{ + struct ib_umad_port *port; + struct ib_port_modify props = { + .set_port_cap_mask = IB_PORT_SM + }; + int ret; + + port = container_of(inode->i_cdev, struct ib_umad_port, sm_cdev); + + if (filp->f_flags & O_NONBLOCK) { + if (down_trylock(&port->sm_sem)) { + ret = -EAGAIN; + goto fail; + } + } else { + if (down_interruptible(&port->sm_sem)) { + ret = -ERESTARTSYS; + goto fail; + } + } + + if (!rdma_dev_access_netns(port->ib_dev, current->nsproxy->net_ns)) { + ret = -EPERM; + goto err_up_sem; + } + + ret = ib_modify_port(port->ib_dev, port->port_num, 0, &props); + if (ret) + goto err_up_sem; + + filp->private_data = port; + + nonseekable_open(inode, filp); + return 0; + +err_up_sem: + up(&port->sm_sem); + +fail: + return ret; +} + +static int ib_umad_sm_close(struct inode *inode, struct file *filp) +{ + struct ib_umad_port *port = filp->private_data; + struct ib_port_modify props = { + .clr_port_cap_mask = IB_PORT_SM + }; + int ret = 0; + + mutex_lock(&port->file_mutex); + if (port->ib_dev) + ret = ib_modify_port(port->ib_dev, port->port_num, 0, &props); + mutex_unlock(&port->file_mutex); + + up(&port->sm_sem); + + return ret; +} + +static const struct file_operations umad_sm_fops = { + .owner = THIS_MODULE, + .open = ib_umad_sm_open, + .release = ib_umad_sm_close, + .llseek = no_llseek, +}; + +static struct ib_umad_port *get_port(struct ib_device *ibdev, + struct ib_umad_device *umad_dev, + unsigned int port) +{ + if (!umad_dev) + return ERR_PTR(-EOPNOTSUPP); + if (!rdma_is_port_valid(ibdev, port)) + return ERR_PTR(-EINVAL); + if (!rdma_cap_ib_mad(ibdev, port)) + return ERR_PTR(-EOPNOTSUPP); + + return &umad_dev->ports[port - rdma_start_port(ibdev)]; +} + +static int ib_umad_get_nl_info(struct ib_device *ibdev, void *client_data, + struct ib_client_nl_info *res) +{ + struct ib_umad_port *port = get_port(ibdev, client_data, res->port); + + if (IS_ERR(port)) + return PTR_ERR(port); + + res->abi = IB_USER_MAD_ABI_VERSION; + res->cdev = &port->dev; + return 0; +} + +static struct ib_client umad_client = { + .name = "umad", + .add = ib_umad_add_one, + .remove = ib_umad_remove_one, + .get_nl_info = ib_umad_get_nl_info, +}; +MODULE_ALIAS_RDMA_CLIENT("umad"); + +static int ib_issm_get_nl_info(struct ib_device *ibdev, void *client_data, + struct ib_client_nl_info *res) +{ + struct ib_umad_port *port = get_port(ibdev, client_data, res->port); + + if (IS_ERR(port)) + return PTR_ERR(port); + + res->abi = IB_USER_MAD_ABI_VERSION; + res->cdev = &port->sm_dev; + return 0; +} + +static struct ib_client issm_client = { + .name = "issm", + .get_nl_info = ib_issm_get_nl_info, +}; +MODULE_ALIAS_RDMA_CLIENT("issm"); + +static ssize_t ibdev_show(struct device *dev, struct device_attribute *attr, + char *buf) +{ + struct ib_umad_port *port = dev_get_drvdata(dev); + + if (!port) + return -ENODEV; + + return sprintf(buf, "%s\n", dev_name(&port->ib_dev->dev)); +} +static DEVICE_ATTR_RO(ibdev); + +static ssize_t port_show(struct device *dev, struct device_attribute *attr, + char *buf) +{ + struct ib_umad_port *port = dev_get_drvdata(dev); + + if (!port) + return -ENODEV; + + return sprintf(buf, "%d\n", port->port_num); +} +static DEVICE_ATTR_RO(port); + +static struct attribute *umad_class_dev_attrs[] = { + &dev_attr_ibdev.attr, + &dev_attr_port.attr, + NULL, +}; +ATTRIBUTE_GROUPS(umad_class_dev); + +static char *umad_devnode(struct device *dev, umode_t *mode) +{ + return kasprintf(GFP_KERNEL, "infiniband/%s", dev_name(dev)); +} + +static ssize_t abi_version_show(struct class *class, + struct class_attribute *attr, char *buf) +{ + return sprintf(buf, "%d\n", IB_USER_MAD_ABI_VERSION); +} +static CLASS_ATTR_RO(abi_version); + +static struct attribute *umad_class_attrs[] = { + &class_attr_abi_version.attr, + NULL, +}; +ATTRIBUTE_GROUPS(umad_class); + +static struct class umad_class = { + .name = "infiniband_mad", + .devnode = umad_devnode, + .class_groups = umad_class_groups, + .dev_groups = umad_class_dev_groups, +}; + +static void ib_umad_release_port(struct device *device) +{ + struct ib_umad_port *port = dev_get_drvdata(device); + struct ib_umad_device *umad_dev = port->umad_dev; + + ib_umad_dev_put(umad_dev); +} + +static void ib_umad_init_port_dev(struct device *dev, + struct ib_umad_port *port, + const struct ib_device *device) +{ + device_initialize(dev); + ib_umad_dev_get(port->umad_dev); + dev->class = &umad_class; + dev->parent = device->dev.parent; + dev_set_drvdata(dev, port); + dev->release = ib_umad_release_port; +} + +static int ib_umad_init_port(struct ib_device *device, int port_num, + struct ib_umad_device *umad_dev, + struct ib_umad_port *port) +{ + int devnum; + dev_t base_umad; + dev_t base_issm; + int ret; + + devnum = ida_alloc_max(&umad_ida, IB_UMAD_MAX_PORTS - 1, GFP_KERNEL); + if (devnum < 0) + return -1; + port->dev_num = devnum; + if (devnum >= IB_UMAD_NUM_FIXED_MINOR) { + base_umad = dynamic_umad_dev + devnum - IB_UMAD_NUM_FIXED_MINOR; + base_issm = dynamic_issm_dev + devnum - IB_UMAD_NUM_FIXED_MINOR; + } else { + base_umad = devnum + base_umad_dev; + base_issm = devnum + base_issm_dev; + } + + port->ib_dev = device; + port->umad_dev = umad_dev; + port->port_num = port_num; + sema_init(&port->sm_sem, 1); + mutex_init(&port->file_mutex); + INIT_LIST_HEAD(&port->file_list); + + ib_umad_init_port_dev(&port->dev, port, device); + port->dev.devt = base_umad; + dev_set_name(&port->dev, "umad%d", port->dev_num); + cdev_init(&port->cdev, &umad_fops); + port->cdev.owner = THIS_MODULE; + + ret = cdev_device_add(&port->cdev, &port->dev); + if (ret) + goto err_cdev; + + ib_umad_init_port_dev(&port->sm_dev, port, device); + port->sm_dev.devt = base_issm; + dev_set_name(&port->sm_dev, "issm%d", port->dev_num); + cdev_init(&port->sm_cdev, &umad_sm_fops); + port->sm_cdev.owner = THIS_MODULE; + + ret = cdev_device_add(&port->sm_cdev, &port->sm_dev); + if (ret) + goto err_dev; + + return 0; + +err_dev: + put_device(&port->sm_dev); + cdev_device_del(&port->cdev, &port->dev); +err_cdev: + put_device(&port->dev); + ida_free(&umad_ida, devnum); + return ret; +} + +static void ib_umad_kill_port(struct ib_umad_port *port) +{ + struct ib_umad_file *file; + int id; + + cdev_device_del(&port->sm_cdev, &port->sm_dev); + cdev_device_del(&port->cdev, &port->dev); + + mutex_lock(&port->file_mutex); + + /* Mark ib_dev NULL and block ioctl or other file ops to progress + * further. + */ + port->ib_dev = NULL; + + list_for_each_entry(file, &port->file_list, port_list) { + mutex_lock(&file->mutex); + file->agents_dead = 1; + wake_up_interruptible(&file->recv_wait); + mutex_unlock(&file->mutex); + + for (id = 0; id < IB_UMAD_MAX_AGENTS; ++id) + if (file->agent[id]) + ib_unregister_mad_agent(file->agent[id]); + } + + mutex_unlock(&port->file_mutex); + + ida_free(&umad_ida, port->dev_num); + + /* balances device_initialize() */ + put_device(&port->sm_dev); + put_device(&port->dev); +} + +static int ib_umad_add_one(struct ib_device *device) +{ + struct ib_umad_device *umad_dev; + int s, e, i; + int count = 0; + int ret; + + s = rdma_start_port(device); + e = rdma_end_port(device); + + umad_dev = kzalloc(struct_size(umad_dev, ports, e - s + 1), GFP_KERNEL); + if (!umad_dev) + return -ENOMEM; + + kref_init(&umad_dev->kref); + for (i = s; i <= e; ++i) { + if (!rdma_cap_ib_mad(device, i)) + continue; + + ret = ib_umad_init_port(device, i, umad_dev, + &umad_dev->ports[i - s]); + if (ret) + goto err; + + count++; + } + + if (!count) { + ret = -EOPNOTSUPP; + goto free; + } + + ib_set_client_data(device, &umad_client, umad_dev); + + return 0; + +err: + while (--i >= s) { + if (!rdma_cap_ib_mad(device, i)) + continue; + + ib_umad_kill_port(&umad_dev->ports[i - s]); + } +free: + /* balances kref_init */ + ib_umad_dev_put(umad_dev); + return ret; +} + +static void ib_umad_remove_one(struct ib_device *device, void *client_data) +{ + struct ib_umad_device *umad_dev = client_data; + unsigned int i; + + rdma_for_each_port (device, i) { + if (rdma_cap_ib_mad(device, i)) + ib_umad_kill_port( + &umad_dev->ports[i - rdma_start_port(device)]); + } + /* balances kref_init() */ + ib_umad_dev_put(umad_dev); +} + +static int __init ib_umad_init(void) +{ + int ret; + + ret = register_chrdev_region(base_umad_dev, + IB_UMAD_NUM_FIXED_MINOR * 2, + umad_class.name); + if (ret) { + pr_err("couldn't register device number\n"); + goto out; + } + + ret = alloc_chrdev_region(&dynamic_umad_dev, 0, + IB_UMAD_NUM_DYNAMIC_MINOR * 2, + umad_class.name); + if (ret) { + pr_err("couldn't register dynamic device number\n"); + goto out_alloc; + } + dynamic_issm_dev = dynamic_umad_dev + IB_UMAD_NUM_DYNAMIC_MINOR; + + ret = class_register(&umad_class); + if (ret) { + pr_err("couldn't create class infiniband_mad\n"); + goto out_chrdev; + } + + ret = ib_register_client(&umad_client); + if (ret) + goto out_class; + + ret = ib_register_client(&issm_client); + if (ret) + goto out_client; + + return 0; + +out_client: + ib_unregister_client(&umad_client); +out_class: + class_unregister(&umad_class); + +out_chrdev: + unregister_chrdev_region(dynamic_umad_dev, + IB_UMAD_NUM_DYNAMIC_MINOR * 2); + +out_alloc: + unregister_chrdev_region(base_umad_dev, + IB_UMAD_NUM_FIXED_MINOR * 2); + +out: + return ret; +} + +static void __exit ib_umad_cleanup(void) +{ + ib_unregister_client(&issm_client); + ib_unregister_client(&umad_client); + class_unregister(&umad_class); + unregister_chrdev_region(base_umad_dev, + IB_UMAD_NUM_FIXED_MINOR * 2); + unregister_chrdev_region(dynamic_umad_dev, + IB_UMAD_NUM_DYNAMIC_MINOR * 2); +} + +module_init(ib_umad_init); +module_exit(ib_umad_cleanup); diff --git a/drivers/infiniband/core/uverbs.h b/drivers/infiniband/core/uverbs.h new file mode 100644 index 000000000..53a104799 --- /dev/null +++ b/drivers/infiniband/core/uverbs.h @@ -0,0 +1,322 @@ +/* + * Copyright (c) 2005 Topspin Communications. All rights reserved. + * Copyright (c) 2005, 2006 Cisco Systems. All rights reserved. + * Copyright (c) 2005 Mellanox Technologies. All rights reserved. + * Copyright (c) 2005 Voltaire, Inc. All rights reserved. + * Copyright (c) 2005 PathScale, Inc. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#ifndef UVERBS_H +#define UVERBS_H + +#include <linux/kref.h> +#include <linux/idr.h> +#include <linux/mutex.h> +#include <linux/completion.h> +#include <linux/cdev.h> + +#include <rdma/ib_verbs.h> +#include <rdma/ib_umem.h> +#include <rdma/ib_user_verbs.h> +#include <rdma/uverbs_std_types.h> + +#define UVERBS_MODULE_NAME ib_uverbs +#include <rdma/uverbs_named_ioctl.h> + +static inline void +ib_uverbs_init_udata(struct ib_udata *udata, + const void __user *ibuf, + void __user *obuf, + size_t ilen, size_t olen) +{ + udata->inbuf = ibuf; + udata->outbuf = obuf; + udata->inlen = ilen; + udata->outlen = olen; +} + +static inline void +ib_uverbs_init_udata_buf_or_null(struct ib_udata *udata, + const void __user *ibuf, + void __user *obuf, + size_t ilen, size_t olen) +{ + ib_uverbs_init_udata(udata, + ilen ? ibuf : NULL, olen ? obuf : NULL, + ilen, olen); +} + +/* + * Our lifetime rules for these structs are the following: + * + * struct ib_uverbs_device: One reference is held by the module and + * released in ib_uverbs_remove_one(). Another reference is taken by + * ib_uverbs_open() each time the character special file is opened, + * and released in ib_uverbs_release_file() when the file is released. + * + * struct ib_uverbs_file: One reference is held by the VFS and + * released when the file is closed. Another reference is taken when + * an asynchronous event queue file is created and released when the + * event file is closed. + * + * struct ib_uverbs_event_queue: Base structure for + * struct ib_uverbs_async_event_file and struct ib_uverbs_completion_event_file. + * One reference is held by the VFS and released when the file is closed. + * For asynchronous event files, another reference is held by the corresponding + * main context file and released when that file is closed. For completion + * event files, a reference is taken when a CQ is created that uses the file, + * and released when the CQ is destroyed. + */ + +struct ib_uverbs_device { + atomic_t refcount; + u32 num_comp_vectors; + struct completion comp; + struct device dev; + /* First group for device attributes, NULL terminated array */ + const struct attribute_group *groups[2]; + struct ib_device __rcu *ib_dev; + int devnum; + struct cdev cdev; + struct rb_root xrcd_tree; + struct mutex xrcd_tree_mutex; + struct srcu_struct disassociate_srcu; + struct mutex lists_mutex; /* protect lists */ + struct list_head uverbs_file_list; + struct uverbs_api *uapi; +}; + +struct ib_uverbs_event_queue { + spinlock_t lock; + int is_closed; + wait_queue_head_t poll_wait; + struct fasync_struct *async_queue; + struct list_head event_list; +}; + +struct ib_uverbs_async_event_file { + struct ib_uobject uobj; + struct ib_uverbs_event_queue ev_queue; + struct ib_event_handler event_handler; +}; + +struct ib_uverbs_completion_event_file { + struct ib_uobject uobj; + struct ib_uverbs_event_queue ev_queue; +}; + +struct ib_uverbs_file { + struct kref ref; + struct ib_uverbs_device *device; + struct mutex ucontext_lock; + /* + * ucontext must be accessed via ib_uverbs_get_ucontext() or with + * ucontext_lock held + */ + struct ib_ucontext *ucontext; + struct ib_uverbs_async_event_file *default_async_file; + struct list_head list; + + /* + * To access the uobjects list hw_destroy_rwsem must be held for write + * OR hw_destroy_rwsem held for read AND uobjects_lock held. + * hw_destroy_rwsem should be called across any destruction of the HW + * object of an associated uobject. + */ + struct rw_semaphore hw_destroy_rwsem; + spinlock_t uobjects_lock; + struct list_head uobjects; + + struct mutex umap_lock; + struct list_head umaps; + struct page *disassociate_page; + + struct xarray idr; +}; + +struct ib_uverbs_event { + union { + struct ib_uverbs_async_event_desc async; + struct ib_uverbs_comp_event_desc comp; + } desc; + struct list_head list; + struct list_head obj_list; + u32 *counter; +}; + +struct ib_uverbs_mcast_entry { + struct list_head list; + union ib_gid gid; + u16 lid; +}; + +struct ib_uevent_object { + struct ib_uobject uobject; + struct ib_uverbs_async_event_file *event_file; + /* List member for ib_uverbs_async_event_file list */ + struct list_head event_list; + u32 events_reported; +}; + +struct ib_uxrcd_object { + struct ib_uobject uobject; + atomic_t refcnt; +}; + +struct ib_usrq_object { + struct ib_uevent_object uevent; + struct ib_uxrcd_object *uxrcd; +}; + +struct ib_uqp_object { + struct ib_uevent_object uevent; + /* lock for mcast list */ + struct mutex mcast_lock; + struct list_head mcast_list; + struct ib_uxrcd_object *uxrcd; +}; + +struct ib_uwq_object { + struct ib_uevent_object uevent; +}; + +struct ib_ucq_object { + struct ib_uevent_object uevent; + struct list_head comp_list; + u32 comp_events_reported; +}; + +extern const struct file_operations uverbs_event_fops; +extern const struct file_operations uverbs_async_event_fops; +void ib_uverbs_init_event_queue(struct ib_uverbs_event_queue *ev_queue); +void ib_uverbs_init_async_event_file(struct ib_uverbs_async_event_file *ev_file); +void ib_uverbs_free_event_queue(struct ib_uverbs_event_queue *event_queue); +void ib_uverbs_flow_resources_free(struct ib_uflow_resources *uflow_res); +int uverbs_async_event_release(struct inode *inode, struct file *filp); + +int ib_alloc_ucontext(struct uverbs_attr_bundle *attrs); +int ib_init_ucontext(struct uverbs_attr_bundle *attrs); + +void ib_uverbs_release_ucq(struct ib_uverbs_completion_event_file *ev_file, + struct ib_ucq_object *uobj); +void ib_uverbs_release_uevent(struct ib_uevent_object *uobj); +void ib_uverbs_release_file(struct kref *ref); +void ib_uverbs_async_handler(struct ib_uverbs_async_event_file *async_file, + __u64 element, __u64 event, + struct list_head *obj_list, u32 *counter); + +void ib_uverbs_comp_handler(struct ib_cq *cq, void *cq_context); +void ib_uverbs_cq_event_handler(struct ib_event *event, void *context_ptr); +void ib_uverbs_qp_event_handler(struct ib_event *event, void *context_ptr); +void ib_uverbs_wq_event_handler(struct ib_event *event, void *context_ptr); +void ib_uverbs_srq_event_handler(struct ib_event *event, void *context_ptr); +int ib_uverbs_dealloc_xrcd(struct ib_uobject *uobject, struct ib_xrcd *xrcd, + enum rdma_remove_reason why, + struct uverbs_attr_bundle *attrs); + +int uverbs_dealloc_mw(struct ib_mw *mw); +void ib_uverbs_detach_umcast(struct ib_qp *qp, + struct ib_uqp_object *uobj); + +long ib_uverbs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg); + +struct ib_uverbs_flow_spec { + union { + union { + struct ib_uverbs_flow_spec_hdr hdr; + struct { + __u32 type; + __u16 size; + __u16 reserved; + }; + }; + struct ib_uverbs_flow_spec_eth eth; + struct ib_uverbs_flow_spec_ipv4 ipv4; + struct ib_uverbs_flow_spec_esp esp; + struct ib_uverbs_flow_spec_tcp_udp tcp_udp; + struct ib_uverbs_flow_spec_ipv6 ipv6; + struct ib_uverbs_flow_spec_action_tag flow_tag; + struct ib_uverbs_flow_spec_action_drop drop; + struct ib_uverbs_flow_spec_action_handle action; + struct ib_uverbs_flow_spec_action_count flow_count; + }; +}; + +int ib_uverbs_kern_spec_to_ib_spec_filter(enum ib_flow_spec_type type, + const void *kern_spec_mask, + const void *kern_spec_val, + size_t kern_filter_sz, + union ib_flow_spec *ib_spec); + +/* + * ib_uverbs_query_port_resp.port_cap_flags started out as just a copy of the + * PortInfo CapabilityMask, but was extended with unique bits. + */ +static inline u32 make_port_cap_flags(const struct ib_port_attr *attr) +{ + u32 res; + + /* All IBA CapabilityMask bits are passed through here, except bit 26, + * which is overridden with IP_BASED_GIDS. This is due to a historical + * mistake in the implementation of IP_BASED_GIDS. Otherwise all other + * bits match the IBA definition across all kernel versions. + */ + res = attr->port_cap_flags & ~(u32)IB_UVERBS_PCF_IP_BASED_GIDS; + + if (attr->ip_gids) + res |= IB_UVERBS_PCF_IP_BASED_GIDS; + + return res; +} + +static inline struct ib_uverbs_async_event_file * +ib_uverbs_get_async_event(struct uverbs_attr_bundle *attrs, + u16 id) +{ + struct ib_uobject *async_ev_file_uobj; + struct ib_uverbs_async_event_file *async_ev_file; + + async_ev_file_uobj = uverbs_attr_get_uobject(attrs, id); + if (IS_ERR(async_ev_file_uobj)) + async_ev_file = READ_ONCE(attrs->ufile->default_async_file); + else + async_ev_file = container_of(async_ev_file_uobj, + struct ib_uverbs_async_event_file, + uobj); + if (async_ev_file) + uverbs_uobject_get(&async_ev_file->uobj); + return async_ev_file; +} + +void copy_port_attr_to_resp(struct ib_port_attr *attr, + struct ib_uverbs_query_port_resp *resp, + struct ib_device *ib_dev, u8 port_num); +#endif /* UVERBS_H */ diff --git a/drivers/infiniband/core/uverbs_cmd.c b/drivers/infiniband/core/uverbs_cmd.c new file mode 100644 index 000000000..158f9eadc --- /dev/null +++ b/drivers/infiniband/core/uverbs_cmd.c @@ -0,0 +1,4042 @@ +/* + * Copyright (c) 2005 Topspin Communications. All rights reserved. + * Copyright (c) 2005, 2006, 2007 Cisco Systems. All rights reserved. + * Copyright (c) 2005 PathScale, Inc. All rights reserved. + * Copyright (c) 2006 Mellanox Technologies. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include <linux/file.h> +#include <linux/fs.h> +#include <linux/slab.h> +#include <linux/sched.h> + +#include <linux/uaccess.h> + +#include <rdma/uverbs_types.h> +#include <rdma/uverbs_std_types.h> +#include "rdma_core.h" + +#include "uverbs.h" +#include "core_priv.h" + +/* + * Copy a response to userspace. If the provided 'resp' is larger than the + * user buffer it is silently truncated. If the user provided a larger buffer + * then the trailing portion is zero filled. + * + * These semantics are intended to support future extension of the output + * structures. + */ +static int uverbs_response(struct uverbs_attr_bundle *attrs, const void *resp, + size_t resp_len) +{ + int ret; + + if (uverbs_attr_is_valid(attrs, UVERBS_ATTR_CORE_OUT)) + return uverbs_copy_to_struct_or_zero( + attrs, UVERBS_ATTR_CORE_OUT, resp, resp_len); + + if (copy_to_user(attrs->ucore.outbuf, resp, + min(attrs->ucore.outlen, resp_len))) + return -EFAULT; + + if (resp_len < attrs->ucore.outlen) { + /* + * Zero fill any extra memory that user + * space might have provided. + */ + ret = clear_user(attrs->ucore.outbuf + resp_len, + attrs->ucore.outlen - resp_len); + if (ret) + return -EFAULT; + } + + return 0; +} + +/* + * Copy a request from userspace. If the provided 'req' is larger than the + * user buffer then the user buffer is zero extended into the 'req'. If 'req' + * is smaller than the user buffer then the uncopied bytes in the user buffer + * must be zero. + */ +static int uverbs_request(struct uverbs_attr_bundle *attrs, void *req, + size_t req_len) +{ + if (copy_from_user(req, attrs->ucore.inbuf, + min(attrs->ucore.inlen, req_len))) + return -EFAULT; + + if (attrs->ucore.inlen < req_len) { + memset(req + attrs->ucore.inlen, 0, + req_len - attrs->ucore.inlen); + } else if (attrs->ucore.inlen > req_len) { + if (!ib_is_buffer_cleared(attrs->ucore.inbuf + req_len, + attrs->ucore.inlen - req_len)) + return -EOPNOTSUPP; + } + return 0; +} + +/* + * Generate the value for the 'response_length' protocol used by write_ex. + * This is the number of bytes the kernel actually wrote. Userspace can use + * this to detect what structure members in the response the kernel + * understood. + */ +static u32 uverbs_response_length(struct uverbs_attr_bundle *attrs, + size_t resp_len) +{ + return min_t(size_t, attrs->ucore.outlen, resp_len); +} + +/* + * The iterator version of the request interface is for handlers that need to + * step over a flex array at the end of a command header. + */ +struct uverbs_req_iter { + const void __user *cur; + const void __user *end; +}; + +static int uverbs_request_start(struct uverbs_attr_bundle *attrs, + struct uverbs_req_iter *iter, + void *req, + size_t req_len) +{ + if (attrs->ucore.inlen < req_len) + return -ENOSPC; + + if (copy_from_user(req, attrs->ucore.inbuf, req_len)) + return -EFAULT; + + iter->cur = attrs->ucore.inbuf + req_len; + iter->end = attrs->ucore.inbuf + attrs->ucore.inlen; + return 0; +} + +static int uverbs_request_next(struct uverbs_req_iter *iter, void *val, + size_t len) +{ + if (iter->cur + len > iter->end) + return -ENOSPC; + + if (copy_from_user(val, iter->cur, len)) + return -EFAULT; + + iter->cur += len; + return 0; +} + +static const void __user *uverbs_request_next_ptr(struct uverbs_req_iter *iter, + size_t len) +{ + const void __user *res = iter->cur; + + if (iter->cur + len > iter->end) + return (void __force __user *)ERR_PTR(-ENOSPC); + iter->cur += len; + return res; +} + +static int uverbs_request_finish(struct uverbs_req_iter *iter) +{ + if (!ib_is_buffer_cleared(iter->cur, iter->end - iter->cur)) + return -EOPNOTSUPP; + return 0; +} + +/* + * When calling a destroy function during an error unwind we need to pass in + * the udata that is sanitized of all user arguments. Ie from the driver + * perspective it looks like no udata was passed. + */ +struct ib_udata *uverbs_get_cleared_udata(struct uverbs_attr_bundle *attrs) +{ + attrs->driver_udata = (struct ib_udata){}; + return &attrs->driver_udata; +} + +static struct ib_uverbs_completion_event_file * +_ib_uverbs_lookup_comp_file(s32 fd, struct uverbs_attr_bundle *attrs) +{ + struct ib_uobject *uobj = ufd_get_read(UVERBS_OBJECT_COMP_CHANNEL, + fd, attrs); + + if (IS_ERR(uobj)) + return (void *)uobj; + + uverbs_uobject_get(uobj); + uobj_put_read(uobj); + + return container_of(uobj, struct ib_uverbs_completion_event_file, + uobj); +} +#define ib_uverbs_lookup_comp_file(_fd, _ufile) \ + _ib_uverbs_lookup_comp_file((_fd)*typecheck(s32, _fd), _ufile) + +int ib_alloc_ucontext(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_file *ufile = attrs->ufile; + struct ib_ucontext *ucontext; + struct ib_device *ib_dev; + + ib_dev = srcu_dereference(ufile->device->ib_dev, + &ufile->device->disassociate_srcu); + if (!ib_dev) + return -EIO; + + ucontext = rdma_zalloc_drv_obj(ib_dev, ib_ucontext); + if (!ucontext) + return -ENOMEM; + + ucontext->device = ib_dev; + ucontext->ufile = ufile; + xa_init_flags(&ucontext->mmap_xa, XA_FLAGS_ALLOC); + + rdma_restrack_new(&ucontext->res, RDMA_RESTRACK_CTX); + rdma_restrack_set_name(&ucontext->res, NULL); + attrs->context = ucontext; + return 0; +} + +int ib_init_ucontext(struct uverbs_attr_bundle *attrs) +{ + struct ib_ucontext *ucontext = attrs->context; + struct ib_uverbs_file *file = attrs->ufile; + int ret; + + if (!down_read_trylock(&file->hw_destroy_rwsem)) + return -EIO; + mutex_lock(&file->ucontext_lock); + if (file->ucontext) { + ret = -EINVAL; + goto err; + } + + ret = ib_rdmacg_try_charge(&ucontext->cg_obj, ucontext->device, + RDMACG_RESOURCE_HCA_HANDLE); + if (ret) + goto err; + + ret = ucontext->device->ops.alloc_ucontext(ucontext, + &attrs->driver_udata); + if (ret) + goto err_uncharge; + + rdma_restrack_add(&ucontext->res); + + /* + * Make sure that ib_uverbs_get_ucontext() sees the pointer update + * only after all writes to setup the ucontext have completed + */ + smp_store_release(&file->ucontext, ucontext); + + mutex_unlock(&file->ucontext_lock); + up_read(&file->hw_destroy_rwsem); + return 0; + +err_uncharge: + ib_rdmacg_uncharge(&ucontext->cg_obj, ucontext->device, + RDMACG_RESOURCE_HCA_HANDLE); +err: + mutex_unlock(&file->ucontext_lock); + up_read(&file->hw_destroy_rwsem); + return ret; +} + +static int ib_uverbs_get_context(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_get_context_resp resp; + struct ib_uverbs_get_context cmd; + struct ib_device *ib_dev; + struct ib_uobject *uobj; + int ret; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + ret = ib_alloc_ucontext(attrs); + if (ret) + return ret; + + uobj = uobj_alloc(UVERBS_OBJECT_ASYNC_EVENT, attrs, &ib_dev); + if (IS_ERR(uobj)) { + ret = PTR_ERR(uobj); + goto err_ucontext; + } + + resp = (struct ib_uverbs_get_context_resp){ + .num_comp_vectors = attrs->ufile->device->num_comp_vectors, + .async_fd = uobj->id, + }; + ret = uverbs_response(attrs, &resp, sizeof(resp)); + if (ret) + goto err_uobj; + + ret = ib_init_ucontext(attrs); + if (ret) + goto err_uobj; + + ib_uverbs_init_async_event_file( + container_of(uobj, struct ib_uverbs_async_event_file, uobj)); + rdma_alloc_commit_uobject(uobj, attrs); + return 0; + +err_uobj: + rdma_alloc_abort_uobject(uobj, attrs, false); +err_ucontext: + rdma_restrack_put(&attrs->context->res); + kfree(attrs->context); + attrs->context = NULL; + return ret; +} + +static void copy_query_dev_fields(struct ib_ucontext *ucontext, + struct ib_uverbs_query_device_resp *resp, + struct ib_device_attr *attr) +{ + struct ib_device *ib_dev = ucontext->device; + + resp->fw_ver = attr->fw_ver; + resp->node_guid = ib_dev->node_guid; + resp->sys_image_guid = attr->sys_image_guid; + resp->max_mr_size = attr->max_mr_size; + resp->page_size_cap = attr->page_size_cap; + resp->vendor_id = attr->vendor_id; + resp->vendor_part_id = attr->vendor_part_id; + resp->hw_ver = attr->hw_ver; + resp->max_qp = attr->max_qp; + resp->max_qp_wr = attr->max_qp_wr; + resp->device_cap_flags = lower_32_bits(attr->device_cap_flags); + resp->max_sge = min(attr->max_send_sge, attr->max_recv_sge); + resp->max_sge_rd = attr->max_sge_rd; + resp->max_cq = attr->max_cq; + resp->max_cqe = attr->max_cqe; + resp->max_mr = attr->max_mr; + resp->max_pd = attr->max_pd; + resp->max_qp_rd_atom = attr->max_qp_rd_atom; + resp->max_ee_rd_atom = attr->max_ee_rd_atom; + resp->max_res_rd_atom = attr->max_res_rd_atom; + resp->max_qp_init_rd_atom = attr->max_qp_init_rd_atom; + resp->max_ee_init_rd_atom = attr->max_ee_init_rd_atom; + resp->atomic_cap = attr->atomic_cap; + resp->max_ee = attr->max_ee; + resp->max_rdd = attr->max_rdd; + resp->max_mw = attr->max_mw; + resp->max_raw_ipv6_qp = attr->max_raw_ipv6_qp; + resp->max_raw_ethy_qp = attr->max_raw_ethy_qp; + resp->max_mcast_grp = attr->max_mcast_grp; + resp->max_mcast_qp_attach = attr->max_mcast_qp_attach; + resp->max_total_mcast_qp_attach = attr->max_total_mcast_qp_attach; + resp->max_ah = attr->max_ah; + resp->max_srq = attr->max_srq; + resp->max_srq_wr = attr->max_srq_wr; + resp->max_srq_sge = attr->max_srq_sge; + resp->max_pkeys = attr->max_pkeys; + resp->local_ca_ack_delay = attr->local_ca_ack_delay; + resp->phys_port_cnt = ib_dev->phys_port_cnt; +} + +static int ib_uverbs_query_device(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_query_device cmd; + struct ib_uverbs_query_device_resp resp; + struct ib_ucontext *ucontext; + int ret; + + ucontext = ib_uverbs_get_ucontext(attrs); + if (IS_ERR(ucontext)) + return PTR_ERR(ucontext); + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + memset(&resp, 0, sizeof resp); + copy_query_dev_fields(ucontext, &resp, &ucontext->device->attrs); + + return uverbs_response(attrs, &resp, sizeof(resp)); +} + +static int ib_uverbs_query_port(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_query_port cmd; + struct ib_uverbs_query_port_resp resp; + struct ib_port_attr attr; + int ret; + struct ib_ucontext *ucontext; + struct ib_device *ib_dev; + + ucontext = ib_uverbs_get_ucontext(attrs); + if (IS_ERR(ucontext)) + return PTR_ERR(ucontext); + ib_dev = ucontext->device; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + ret = ib_query_port(ib_dev, cmd.port_num, &attr); + if (ret) + return ret; + + memset(&resp, 0, sizeof resp); + copy_port_attr_to_resp(&attr, &resp, ib_dev, cmd.port_num); + + return uverbs_response(attrs, &resp, sizeof(resp)); +} + +static int ib_uverbs_alloc_pd(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_alloc_pd_resp resp = {}; + struct ib_uverbs_alloc_pd cmd; + struct ib_uobject *uobj; + struct ib_pd *pd; + int ret; + struct ib_device *ib_dev; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + uobj = uobj_alloc(UVERBS_OBJECT_PD, attrs, &ib_dev); + if (IS_ERR(uobj)) + return PTR_ERR(uobj); + + pd = rdma_zalloc_drv_obj(ib_dev, ib_pd); + if (!pd) { + ret = -ENOMEM; + goto err; + } + + pd->device = ib_dev; + pd->uobject = uobj; + atomic_set(&pd->usecnt, 0); + + rdma_restrack_new(&pd->res, RDMA_RESTRACK_PD); + rdma_restrack_set_name(&pd->res, NULL); + + ret = ib_dev->ops.alloc_pd(pd, &attrs->driver_udata); + if (ret) + goto err_alloc; + rdma_restrack_add(&pd->res); + + uobj->object = pd; + uobj_finalize_uobj_create(uobj, attrs); + + resp.pd_handle = uobj->id; + return uverbs_response(attrs, &resp, sizeof(resp)); + +err_alloc: + rdma_restrack_put(&pd->res); + kfree(pd); +err: + uobj_alloc_abort(uobj, attrs); + return ret; +} + +static int ib_uverbs_dealloc_pd(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_dealloc_pd cmd; + int ret; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + return uobj_perform_destroy(UVERBS_OBJECT_PD, cmd.pd_handle, attrs); +} + +struct xrcd_table_entry { + struct rb_node node; + struct ib_xrcd *xrcd; + struct inode *inode; +}; + +static int xrcd_table_insert(struct ib_uverbs_device *dev, + struct inode *inode, + struct ib_xrcd *xrcd) +{ + struct xrcd_table_entry *entry, *scan; + struct rb_node **p = &dev->xrcd_tree.rb_node; + struct rb_node *parent = NULL; + + entry = kmalloc(sizeof *entry, GFP_KERNEL); + if (!entry) + return -ENOMEM; + + entry->xrcd = xrcd; + entry->inode = inode; + + while (*p) { + parent = *p; + scan = rb_entry(parent, struct xrcd_table_entry, node); + + if (inode < scan->inode) { + p = &(*p)->rb_left; + } else if (inode > scan->inode) { + p = &(*p)->rb_right; + } else { + kfree(entry); + return -EEXIST; + } + } + + rb_link_node(&entry->node, parent, p); + rb_insert_color(&entry->node, &dev->xrcd_tree); + igrab(inode); + return 0; +} + +static struct xrcd_table_entry *xrcd_table_search(struct ib_uverbs_device *dev, + struct inode *inode) +{ + struct xrcd_table_entry *entry; + struct rb_node *p = dev->xrcd_tree.rb_node; + + while (p) { + entry = rb_entry(p, struct xrcd_table_entry, node); + + if (inode < entry->inode) + p = p->rb_left; + else if (inode > entry->inode) + p = p->rb_right; + else + return entry; + } + + return NULL; +} + +static struct ib_xrcd *find_xrcd(struct ib_uverbs_device *dev, struct inode *inode) +{ + struct xrcd_table_entry *entry; + + entry = xrcd_table_search(dev, inode); + if (!entry) + return NULL; + + return entry->xrcd; +} + +static void xrcd_table_delete(struct ib_uverbs_device *dev, + struct inode *inode) +{ + struct xrcd_table_entry *entry; + + entry = xrcd_table_search(dev, inode); + if (entry) { + iput(inode); + rb_erase(&entry->node, &dev->xrcd_tree); + kfree(entry); + } +} + +static int ib_uverbs_open_xrcd(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_device *ibudev = attrs->ufile->device; + struct ib_uverbs_open_xrcd_resp resp = {}; + struct ib_uverbs_open_xrcd cmd; + struct ib_uxrcd_object *obj; + struct ib_xrcd *xrcd = NULL; + struct inode *inode = NULL; + int new_xrcd = 0; + struct ib_device *ib_dev; + struct fd f = {}; + int ret; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + mutex_lock(&ibudev->xrcd_tree_mutex); + + if (cmd.fd != -1) { + /* search for file descriptor */ + f = fdget(cmd.fd); + if (!f.file) { + ret = -EBADF; + goto err_tree_mutex_unlock; + } + + inode = file_inode(f.file); + xrcd = find_xrcd(ibudev, inode); + if (!xrcd && !(cmd.oflags & O_CREAT)) { + /* no file descriptor. Need CREATE flag */ + ret = -EAGAIN; + goto err_tree_mutex_unlock; + } + + if (xrcd && cmd.oflags & O_EXCL) { + ret = -EINVAL; + goto err_tree_mutex_unlock; + } + } + + obj = (struct ib_uxrcd_object *)uobj_alloc(UVERBS_OBJECT_XRCD, attrs, + &ib_dev); + if (IS_ERR(obj)) { + ret = PTR_ERR(obj); + goto err_tree_mutex_unlock; + } + + if (!xrcd) { + xrcd = ib_alloc_xrcd_user(ib_dev, inode, &attrs->driver_udata); + if (IS_ERR(xrcd)) { + ret = PTR_ERR(xrcd); + goto err; + } + new_xrcd = 1; + } + + atomic_set(&obj->refcnt, 0); + obj->uobject.object = xrcd; + + if (inode) { + if (new_xrcd) { + /* create new inode/xrcd table entry */ + ret = xrcd_table_insert(ibudev, inode, xrcd); + if (ret) + goto err_dealloc_xrcd; + } + atomic_inc(&xrcd->usecnt); + } + + if (f.file) + fdput(f); + + mutex_unlock(&ibudev->xrcd_tree_mutex); + uobj_finalize_uobj_create(&obj->uobject, attrs); + + resp.xrcd_handle = obj->uobject.id; + return uverbs_response(attrs, &resp, sizeof(resp)); + +err_dealloc_xrcd: + ib_dealloc_xrcd_user(xrcd, uverbs_get_cleared_udata(attrs)); + +err: + uobj_alloc_abort(&obj->uobject, attrs); + +err_tree_mutex_unlock: + if (f.file) + fdput(f); + + mutex_unlock(&ibudev->xrcd_tree_mutex); + + return ret; +} + +static int ib_uverbs_close_xrcd(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_close_xrcd cmd; + int ret; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + return uobj_perform_destroy(UVERBS_OBJECT_XRCD, cmd.xrcd_handle, attrs); +} + +int ib_uverbs_dealloc_xrcd(struct ib_uobject *uobject, struct ib_xrcd *xrcd, + enum rdma_remove_reason why, + struct uverbs_attr_bundle *attrs) +{ + struct inode *inode; + int ret; + struct ib_uverbs_device *dev = attrs->ufile->device; + + inode = xrcd->inode; + if (inode && !atomic_dec_and_test(&xrcd->usecnt)) + return 0; + + ret = ib_dealloc_xrcd_user(xrcd, &attrs->driver_udata); + + if (ib_is_destroy_retryable(ret, why, uobject)) { + atomic_inc(&xrcd->usecnt); + return ret; + } + + if (inode) + xrcd_table_delete(dev, inode); + + return ret; +} + +static int ib_uverbs_reg_mr(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_reg_mr_resp resp = {}; + struct ib_uverbs_reg_mr cmd; + struct ib_uobject *uobj; + struct ib_pd *pd; + struct ib_mr *mr; + int ret; + struct ib_device *ib_dev; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + if ((cmd.start & ~PAGE_MASK) != (cmd.hca_va & ~PAGE_MASK)) + return -EINVAL; + + ret = ib_check_mr_access(cmd.access_flags); + if (ret) + return ret; + + uobj = uobj_alloc(UVERBS_OBJECT_MR, attrs, &ib_dev); + if (IS_ERR(uobj)) + return PTR_ERR(uobj); + + pd = uobj_get_obj_read(pd, UVERBS_OBJECT_PD, cmd.pd_handle, attrs); + if (!pd) { + ret = -EINVAL; + goto err_free; + } + + if (cmd.access_flags & IB_ACCESS_ON_DEMAND) { + if (!(pd->device->attrs.device_cap_flags & + IB_DEVICE_ON_DEMAND_PAGING)) { + pr_debug("ODP support not available\n"); + ret = -EINVAL; + goto err_put; + } + } + + mr = pd->device->ops.reg_user_mr(pd, cmd.start, cmd.length, cmd.hca_va, + cmd.access_flags, + &attrs->driver_udata); + if (IS_ERR(mr)) { + ret = PTR_ERR(mr); + goto err_put; + } + + mr->device = pd->device; + mr->pd = pd; + mr->type = IB_MR_TYPE_USER; + mr->dm = NULL; + mr->sig_attrs = NULL; + mr->uobject = uobj; + atomic_inc(&pd->usecnt); + mr->iova = cmd.hca_va; + mr->length = cmd.length; + + rdma_restrack_new(&mr->res, RDMA_RESTRACK_MR); + rdma_restrack_set_name(&mr->res, NULL); + rdma_restrack_add(&mr->res); + + uobj->object = mr; + uobj_put_obj_read(pd); + uobj_finalize_uobj_create(uobj, attrs); + + resp.lkey = mr->lkey; + resp.rkey = mr->rkey; + resp.mr_handle = uobj->id; + return uverbs_response(attrs, &resp, sizeof(resp)); + +err_put: + uobj_put_obj_read(pd); +err_free: + uobj_alloc_abort(uobj, attrs); + return ret; +} + +static int ib_uverbs_rereg_mr(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_rereg_mr cmd; + struct ib_uverbs_rereg_mr_resp resp; + struct ib_pd *pd = NULL; + struct ib_mr *mr; + struct ib_pd *old_pd; + int ret; + struct ib_uobject *uobj; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + if (cmd.flags & ~IB_MR_REREG_SUPPORTED || !cmd.flags) + return -EINVAL; + + if ((cmd.flags & IB_MR_REREG_TRANS) && + (!cmd.start || !cmd.hca_va || 0 >= cmd.length || + (cmd.start & ~PAGE_MASK) != (cmd.hca_va & ~PAGE_MASK))) + return -EINVAL; + + uobj = uobj_get_write(UVERBS_OBJECT_MR, cmd.mr_handle, attrs); + if (IS_ERR(uobj)) + return PTR_ERR(uobj); + + mr = uobj->object; + + if (mr->dm) { + ret = -EINVAL; + goto put_uobjs; + } + + if (cmd.flags & IB_MR_REREG_ACCESS) { + ret = ib_check_mr_access(cmd.access_flags); + if (ret) + goto put_uobjs; + } + + if (cmd.flags & IB_MR_REREG_PD) { + pd = uobj_get_obj_read(pd, UVERBS_OBJECT_PD, cmd.pd_handle, + attrs); + if (!pd) { + ret = -EINVAL; + goto put_uobjs; + } + } + + old_pd = mr->pd; + ret = mr->device->ops.rereg_user_mr(mr, cmd.flags, cmd.start, + cmd.length, cmd.hca_va, + cmd.access_flags, pd, + &attrs->driver_udata); + if (ret) + goto put_uobj_pd; + + if (cmd.flags & IB_MR_REREG_PD) { + atomic_inc(&pd->usecnt); + mr->pd = pd; + atomic_dec(&old_pd->usecnt); + } + + if (cmd.flags & IB_MR_REREG_TRANS) { + mr->iova = cmd.hca_va; + mr->length = cmd.length; + } + + memset(&resp, 0, sizeof(resp)); + resp.lkey = mr->lkey; + resp.rkey = mr->rkey; + + ret = uverbs_response(attrs, &resp, sizeof(resp)); + +put_uobj_pd: + if (cmd.flags & IB_MR_REREG_PD) + uobj_put_obj_read(pd); + +put_uobjs: + uobj_put_write(uobj); + + return ret; +} + +static int ib_uverbs_dereg_mr(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_dereg_mr cmd; + int ret; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + return uobj_perform_destroy(UVERBS_OBJECT_MR, cmd.mr_handle, attrs); +} + +static int ib_uverbs_alloc_mw(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_alloc_mw cmd; + struct ib_uverbs_alloc_mw_resp resp = {}; + struct ib_uobject *uobj; + struct ib_pd *pd; + struct ib_mw *mw; + int ret; + struct ib_device *ib_dev; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + uobj = uobj_alloc(UVERBS_OBJECT_MW, attrs, &ib_dev); + if (IS_ERR(uobj)) + return PTR_ERR(uobj); + + pd = uobj_get_obj_read(pd, UVERBS_OBJECT_PD, cmd.pd_handle, attrs); + if (!pd) { + ret = -EINVAL; + goto err_free; + } + + if (cmd.mw_type != IB_MW_TYPE_1 && cmd.mw_type != IB_MW_TYPE_2) { + ret = -EINVAL; + goto err_put; + } + + mw = rdma_zalloc_drv_obj(ib_dev, ib_mw); + if (!mw) { + ret = -ENOMEM; + goto err_put; + } + + mw->device = ib_dev; + mw->pd = pd; + mw->uobject = uobj; + mw->type = cmd.mw_type; + + ret = pd->device->ops.alloc_mw(mw, &attrs->driver_udata); + if (ret) + goto err_alloc; + + atomic_inc(&pd->usecnt); + + uobj->object = mw; + uobj_put_obj_read(pd); + uobj_finalize_uobj_create(uobj, attrs); + + resp.rkey = mw->rkey; + resp.mw_handle = uobj->id; + return uverbs_response(attrs, &resp, sizeof(resp)); + +err_alloc: + kfree(mw); +err_put: + uobj_put_obj_read(pd); +err_free: + uobj_alloc_abort(uobj, attrs); + return ret; +} + +static int ib_uverbs_dealloc_mw(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_dealloc_mw cmd; + int ret; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + return uobj_perform_destroy(UVERBS_OBJECT_MW, cmd.mw_handle, attrs); +} + +static int ib_uverbs_create_comp_channel(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_create_comp_channel cmd; + struct ib_uverbs_create_comp_channel_resp resp; + struct ib_uobject *uobj; + struct ib_uverbs_completion_event_file *ev_file; + struct ib_device *ib_dev; + int ret; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + uobj = uobj_alloc(UVERBS_OBJECT_COMP_CHANNEL, attrs, &ib_dev); + if (IS_ERR(uobj)) + return PTR_ERR(uobj); + + ev_file = container_of(uobj, struct ib_uverbs_completion_event_file, + uobj); + ib_uverbs_init_event_queue(&ev_file->ev_queue); + uobj_finalize_uobj_create(uobj, attrs); + + resp.fd = uobj->id; + return uverbs_response(attrs, &resp, sizeof(resp)); +} + +static int create_cq(struct uverbs_attr_bundle *attrs, + struct ib_uverbs_ex_create_cq *cmd) +{ + struct ib_ucq_object *obj; + struct ib_uverbs_completion_event_file *ev_file = NULL; + struct ib_cq *cq; + int ret; + struct ib_uverbs_ex_create_cq_resp resp = {}; + struct ib_cq_init_attr attr = {}; + struct ib_device *ib_dev; + + if (cmd->comp_vector >= attrs->ufile->device->num_comp_vectors) + return -EINVAL; + + obj = (struct ib_ucq_object *)uobj_alloc(UVERBS_OBJECT_CQ, attrs, + &ib_dev); + if (IS_ERR(obj)) + return PTR_ERR(obj); + + if (cmd->comp_channel >= 0) { + ev_file = ib_uverbs_lookup_comp_file(cmd->comp_channel, attrs); + if (IS_ERR(ev_file)) { + ret = PTR_ERR(ev_file); + goto err; + } + } + + obj->uevent.uobject.user_handle = cmd->user_handle; + INIT_LIST_HEAD(&obj->comp_list); + INIT_LIST_HEAD(&obj->uevent.event_list); + + attr.cqe = cmd->cqe; + attr.comp_vector = cmd->comp_vector; + attr.flags = cmd->flags; + + cq = rdma_zalloc_drv_obj(ib_dev, ib_cq); + if (!cq) { + ret = -ENOMEM; + goto err_file; + } + cq->device = ib_dev; + cq->uobject = obj; + cq->comp_handler = ib_uverbs_comp_handler; + cq->event_handler = ib_uverbs_cq_event_handler; + cq->cq_context = ev_file ? &ev_file->ev_queue : NULL; + atomic_set(&cq->usecnt, 0); + + rdma_restrack_new(&cq->res, RDMA_RESTRACK_CQ); + rdma_restrack_set_name(&cq->res, NULL); + + ret = ib_dev->ops.create_cq(cq, &attr, &attrs->driver_udata); + if (ret) + goto err_free; + rdma_restrack_add(&cq->res); + + obj->uevent.uobject.object = cq; + obj->uevent.event_file = READ_ONCE(attrs->ufile->default_async_file); + if (obj->uevent.event_file) + uverbs_uobject_get(&obj->uevent.event_file->uobj); + uobj_finalize_uobj_create(&obj->uevent.uobject, attrs); + + resp.base.cq_handle = obj->uevent.uobject.id; + resp.base.cqe = cq->cqe; + resp.response_length = uverbs_response_length(attrs, sizeof(resp)); + return uverbs_response(attrs, &resp, sizeof(resp)); + +err_free: + rdma_restrack_put(&cq->res); + kfree(cq); +err_file: + if (ev_file) + ib_uverbs_release_ucq(ev_file, obj); +err: + uobj_alloc_abort(&obj->uevent.uobject, attrs); + return ret; +} + +static int ib_uverbs_create_cq(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_create_cq cmd; + struct ib_uverbs_ex_create_cq cmd_ex; + int ret; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + memset(&cmd_ex, 0, sizeof(cmd_ex)); + cmd_ex.user_handle = cmd.user_handle; + cmd_ex.cqe = cmd.cqe; + cmd_ex.comp_vector = cmd.comp_vector; + cmd_ex.comp_channel = cmd.comp_channel; + + return create_cq(attrs, &cmd_ex); +} + +static int ib_uverbs_ex_create_cq(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_ex_create_cq cmd; + int ret; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + if (cmd.comp_mask) + return -EINVAL; + + if (cmd.reserved) + return -EINVAL; + + return create_cq(attrs, &cmd); +} + +static int ib_uverbs_resize_cq(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_resize_cq cmd; + struct ib_uverbs_resize_cq_resp resp = {}; + struct ib_cq *cq; + int ret; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + cq = uobj_get_obj_read(cq, UVERBS_OBJECT_CQ, cmd.cq_handle, attrs); + if (!cq) + return -EINVAL; + + ret = cq->device->ops.resize_cq(cq, cmd.cqe, &attrs->driver_udata); + if (ret) + goto out; + + resp.cqe = cq->cqe; + + ret = uverbs_response(attrs, &resp, sizeof(resp)); +out: + rdma_lookup_put_uobject(&cq->uobject->uevent.uobject, + UVERBS_LOOKUP_READ); + + return ret; +} + +static int copy_wc_to_user(struct ib_device *ib_dev, void __user *dest, + struct ib_wc *wc) +{ + struct ib_uverbs_wc tmp; + + tmp.wr_id = wc->wr_id; + tmp.status = wc->status; + tmp.opcode = wc->opcode; + tmp.vendor_err = wc->vendor_err; + tmp.byte_len = wc->byte_len; + tmp.ex.imm_data = wc->ex.imm_data; + tmp.qp_num = wc->qp->qp_num; + tmp.src_qp = wc->src_qp; + tmp.wc_flags = wc->wc_flags; + tmp.pkey_index = wc->pkey_index; + if (rdma_cap_opa_ah(ib_dev, wc->port_num)) + tmp.slid = OPA_TO_IB_UCAST_LID(wc->slid); + else + tmp.slid = ib_lid_cpu16(wc->slid); + tmp.sl = wc->sl; + tmp.dlid_path_bits = wc->dlid_path_bits; + tmp.port_num = wc->port_num; + tmp.reserved = 0; + + if (copy_to_user(dest, &tmp, sizeof tmp)) + return -EFAULT; + + return 0; +} + +static int ib_uverbs_poll_cq(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_poll_cq cmd; + struct ib_uverbs_poll_cq_resp resp; + u8 __user *header_ptr; + u8 __user *data_ptr; + struct ib_cq *cq; + struct ib_wc wc; + int ret; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + cq = uobj_get_obj_read(cq, UVERBS_OBJECT_CQ, cmd.cq_handle, attrs); + if (!cq) + return -EINVAL; + + /* we copy a struct ib_uverbs_poll_cq_resp to user space */ + header_ptr = attrs->ucore.outbuf; + data_ptr = header_ptr + sizeof resp; + + memset(&resp, 0, sizeof resp); + while (resp.count < cmd.ne) { + ret = ib_poll_cq(cq, 1, &wc); + if (ret < 0) + goto out_put; + if (!ret) + break; + + ret = copy_wc_to_user(cq->device, data_ptr, &wc); + if (ret) + goto out_put; + + data_ptr += sizeof(struct ib_uverbs_wc); + ++resp.count; + } + + if (copy_to_user(header_ptr, &resp, sizeof resp)) { + ret = -EFAULT; + goto out_put; + } + ret = 0; + + if (uverbs_attr_is_valid(attrs, UVERBS_ATTR_CORE_OUT)) + ret = uverbs_output_written(attrs, UVERBS_ATTR_CORE_OUT); + +out_put: + rdma_lookup_put_uobject(&cq->uobject->uevent.uobject, + UVERBS_LOOKUP_READ); + return ret; +} + +static int ib_uverbs_req_notify_cq(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_req_notify_cq cmd; + struct ib_cq *cq; + int ret; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + cq = uobj_get_obj_read(cq, UVERBS_OBJECT_CQ, cmd.cq_handle, attrs); + if (!cq) + return -EINVAL; + + ib_req_notify_cq(cq, cmd.solicited_only ? + IB_CQ_SOLICITED : IB_CQ_NEXT_COMP); + + rdma_lookup_put_uobject(&cq->uobject->uevent.uobject, + UVERBS_LOOKUP_READ); + return 0; +} + +static int ib_uverbs_destroy_cq(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_destroy_cq cmd; + struct ib_uverbs_destroy_cq_resp resp; + struct ib_uobject *uobj; + struct ib_ucq_object *obj; + int ret; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + uobj = uobj_get_destroy(UVERBS_OBJECT_CQ, cmd.cq_handle, attrs); + if (IS_ERR(uobj)) + return PTR_ERR(uobj); + + obj = container_of(uobj, struct ib_ucq_object, uevent.uobject); + memset(&resp, 0, sizeof(resp)); + resp.comp_events_reported = obj->comp_events_reported; + resp.async_events_reported = obj->uevent.events_reported; + + uobj_put_destroy(uobj); + + return uverbs_response(attrs, &resp, sizeof(resp)); +} + +static int create_qp(struct uverbs_attr_bundle *attrs, + struct ib_uverbs_ex_create_qp *cmd) +{ + struct ib_uqp_object *obj; + struct ib_device *device; + struct ib_pd *pd = NULL; + struct ib_xrcd *xrcd = NULL; + struct ib_uobject *xrcd_uobj = ERR_PTR(-ENOENT); + struct ib_cq *scq = NULL, *rcq = NULL; + struct ib_srq *srq = NULL; + struct ib_qp *qp; + struct ib_qp_init_attr attr = {}; + struct ib_uverbs_ex_create_qp_resp resp = {}; + int ret; + struct ib_rwq_ind_table *ind_tbl = NULL; + bool has_sq = true; + struct ib_device *ib_dev; + + switch (cmd->qp_type) { + case IB_QPT_RAW_PACKET: + if (!capable(CAP_NET_RAW)) + return -EPERM; + break; + case IB_QPT_RC: + case IB_QPT_UC: + case IB_QPT_UD: + case IB_QPT_XRC_INI: + case IB_QPT_XRC_TGT: + case IB_QPT_DRIVER: + break; + default: + return -EINVAL; + } + + obj = (struct ib_uqp_object *)uobj_alloc(UVERBS_OBJECT_QP, attrs, + &ib_dev); + if (IS_ERR(obj)) + return PTR_ERR(obj); + obj->uxrcd = NULL; + obj->uevent.uobject.user_handle = cmd->user_handle; + mutex_init(&obj->mcast_lock); + + if (cmd->comp_mask & IB_UVERBS_CREATE_QP_MASK_IND_TABLE) { + ind_tbl = uobj_get_obj_read(rwq_ind_table, + UVERBS_OBJECT_RWQ_IND_TBL, + cmd->rwq_ind_tbl_handle, attrs); + if (!ind_tbl) { + ret = -EINVAL; + goto err_put; + } + + attr.rwq_ind_tbl = ind_tbl; + } + + if (ind_tbl && (cmd->max_recv_wr || cmd->max_recv_sge || cmd->is_srq)) { + ret = -EINVAL; + goto err_put; + } + + if (ind_tbl && !cmd->max_send_wr) + has_sq = false; + + if (cmd->qp_type == IB_QPT_XRC_TGT) { + xrcd_uobj = uobj_get_read(UVERBS_OBJECT_XRCD, cmd->pd_handle, + attrs); + + if (IS_ERR(xrcd_uobj)) { + ret = -EINVAL; + goto err_put; + } + + xrcd = (struct ib_xrcd *)xrcd_uobj->object; + if (!xrcd) { + ret = -EINVAL; + goto err_put; + } + device = xrcd->device; + } else { + if (cmd->qp_type == IB_QPT_XRC_INI) { + cmd->max_recv_wr = 0; + cmd->max_recv_sge = 0; + } else { + if (cmd->is_srq) { + srq = uobj_get_obj_read(srq, UVERBS_OBJECT_SRQ, + cmd->srq_handle, attrs); + if (!srq || srq->srq_type == IB_SRQT_XRC) { + ret = -EINVAL; + goto err_put; + } + } + + if (!ind_tbl) { + if (cmd->recv_cq_handle != cmd->send_cq_handle) { + rcq = uobj_get_obj_read( + cq, UVERBS_OBJECT_CQ, + cmd->recv_cq_handle, attrs); + if (!rcq) { + ret = -EINVAL; + goto err_put; + } + } + } + } + + if (has_sq) + scq = uobj_get_obj_read(cq, UVERBS_OBJECT_CQ, + cmd->send_cq_handle, attrs); + if (!ind_tbl) + rcq = rcq ?: scq; + pd = uobj_get_obj_read(pd, UVERBS_OBJECT_PD, cmd->pd_handle, + attrs); + if (!pd || (!scq && has_sq)) { + ret = -EINVAL; + goto err_put; + } + + device = pd->device; + } + + attr.event_handler = ib_uverbs_qp_event_handler; + attr.send_cq = scq; + attr.recv_cq = rcq; + attr.srq = srq; + attr.xrcd = xrcd; + attr.sq_sig_type = cmd->sq_sig_all ? IB_SIGNAL_ALL_WR : + IB_SIGNAL_REQ_WR; + attr.qp_type = cmd->qp_type; + attr.create_flags = 0; + + attr.cap.max_send_wr = cmd->max_send_wr; + attr.cap.max_recv_wr = cmd->max_recv_wr; + attr.cap.max_send_sge = cmd->max_send_sge; + attr.cap.max_recv_sge = cmd->max_recv_sge; + attr.cap.max_inline_data = cmd->max_inline_data; + + INIT_LIST_HEAD(&obj->uevent.event_list); + INIT_LIST_HEAD(&obj->mcast_list); + + attr.create_flags = cmd->create_flags; + if (attr.create_flags & ~(IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK | + IB_QP_CREATE_CROSS_CHANNEL | + IB_QP_CREATE_MANAGED_SEND | + IB_QP_CREATE_MANAGED_RECV | + IB_QP_CREATE_SCATTER_FCS | + IB_QP_CREATE_CVLAN_STRIPPING | + IB_QP_CREATE_SOURCE_QPN | + IB_QP_CREATE_PCI_WRITE_END_PADDING)) { + ret = -EINVAL; + goto err_put; + } + + if (attr.create_flags & IB_QP_CREATE_SOURCE_QPN) { + if (!capable(CAP_NET_RAW)) { + ret = -EPERM; + goto err_put; + } + + attr.source_qpn = cmd->source_qpn; + } + + if (cmd->qp_type == IB_QPT_XRC_TGT) + qp = ib_create_qp(pd, &attr); + else + qp = _ib_create_qp(device, pd, &attr, &attrs->driver_udata, + obj); + + if (IS_ERR(qp)) { + ret = PTR_ERR(qp); + goto err_put; + } + + if (cmd->qp_type != IB_QPT_XRC_TGT) { + ret = ib_create_qp_security(qp, device); + if (ret) + goto err_cb; + + atomic_inc(&pd->usecnt); + if (attr.send_cq) + atomic_inc(&attr.send_cq->usecnt); + if (attr.recv_cq) + atomic_inc(&attr.recv_cq->usecnt); + if (attr.srq) + atomic_inc(&attr.srq->usecnt); + if (ind_tbl) + atomic_inc(&ind_tbl->usecnt); + } else { + /* It is done in _ib_create_qp for other QP types */ + qp->uobject = obj; + } + + obj->uevent.uobject.object = qp; + obj->uevent.event_file = READ_ONCE(attrs->ufile->default_async_file); + if (obj->uevent.event_file) + uverbs_uobject_get(&obj->uevent.event_file->uobj); + + if (xrcd) { + obj->uxrcd = container_of(xrcd_uobj, struct ib_uxrcd_object, + uobject); + atomic_inc(&obj->uxrcd->refcnt); + uobj_put_read(xrcd_uobj); + } + + if (pd) + uobj_put_obj_read(pd); + if (scq) + rdma_lookup_put_uobject(&scq->uobject->uevent.uobject, + UVERBS_LOOKUP_READ); + if (rcq && rcq != scq) + rdma_lookup_put_uobject(&rcq->uobject->uevent.uobject, + UVERBS_LOOKUP_READ); + if (srq) + rdma_lookup_put_uobject(&srq->uobject->uevent.uobject, + UVERBS_LOOKUP_READ); + if (ind_tbl) + uobj_put_obj_read(ind_tbl); + uobj_finalize_uobj_create(&obj->uevent.uobject, attrs); + + resp.base.qpn = qp->qp_num; + resp.base.qp_handle = obj->uevent.uobject.id; + resp.base.max_recv_sge = attr.cap.max_recv_sge; + resp.base.max_send_sge = attr.cap.max_send_sge; + resp.base.max_recv_wr = attr.cap.max_recv_wr; + resp.base.max_send_wr = attr.cap.max_send_wr; + resp.base.max_inline_data = attr.cap.max_inline_data; + resp.response_length = uverbs_response_length(attrs, sizeof(resp)); + return uverbs_response(attrs, &resp, sizeof(resp)); + +err_cb: + ib_destroy_qp_user(qp, uverbs_get_cleared_udata(attrs)); + +err_put: + if (!IS_ERR(xrcd_uobj)) + uobj_put_read(xrcd_uobj); + if (pd) + uobj_put_obj_read(pd); + if (scq) + rdma_lookup_put_uobject(&scq->uobject->uevent.uobject, + UVERBS_LOOKUP_READ); + if (rcq && rcq != scq) + rdma_lookup_put_uobject(&rcq->uobject->uevent.uobject, + UVERBS_LOOKUP_READ); + if (srq) + rdma_lookup_put_uobject(&srq->uobject->uevent.uobject, + UVERBS_LOOKUP_READ); + if (ind_tbl) + uobj_put_obj_read(ind_tbl); + + uobj_alloc_abort(&obj->uevent.uobject, attrs); + return ret; +} + +static int ib_uverbs_create_qp(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_create_qp cmd; + struct ib_uverbs_ex_create_qp cmd_ex; + int ret; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + memset(&cmd_ex, 0, sizeof(cmd_ex)); + cmd_ex.user_handle = cmd.user_handle; + cmd_ex.pd_handle = cmd.pd_handle; + cmd_ex.send_cq_handle = cmd.send_cq_handle; + cmd_ex.recv_cq_handle = cmd.recv_cq_handle; + cmd_ex.srq_handle = cmd.srq_handle; + cmd_ex.max_send_wr = cmd.max_send_wr; + cmd_ex.max_recv_wr = cmd.max_recv_wr; + cmd_ex.max_send_sge = cmd.max_send_sge; + cmd_ex.max_recv_sge = cmd.max_recv_sge; + cmd_ex.max_inline_data = cmd.max_inline_data; + cmd_ex.sq_sig_all = cmd.sq_sig_all; + cmd_ex.qp_type = cmd.qp_type; + cmd_ex.is_srq = cmd.is_srq; + + return create_qp(attrs, &cmd_ex); +} + +static int ib_uverbs_ex_create_qp(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_ex_create_qp cmd; + int ret; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + if (cmd.comp_mask & ~IB_UVERBS_CREATE_QP_SUP_COMP_MASK) + return -EINVAL; + + if (cmd.reserved) + return -EINVAL; + + return create_qp(attrs, &cmd); +} + +static int ib_uverbs_open_qp(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_create_qp_resp resp = {}; + struct ib_uverbs_open_qp cmd; + struct ib_uqp_object *obj; + struct ib_xrcd *xrcd; + struct ib_qp *qp; + struct ib_qp_open_attr attr = {}; + int ret; + struct ib_uobject *xrcd_uobj; + struct ib_device *ib_dev; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + obj = (struct ib_uqp_object *)uobj_alloc(UVERBS_OBJECT_QP, attrs, + &ib_dev); + if (IS_ERR(obj)) + return PTR_ERR(obj); + + xrcd_uobj = uobj_get_read(UVERBS_OBJECT_XRCD, cmd.pd_handle, attrs); + if (IS_ERR(xrcd_uobj)) { + ret = -EINVAL; + goto err_put; + } + + xrcd = (struct ib_xrcd *)xrcd_uobj->object; + if (!xrcd) { + ret = -EINVAL; + goto err_xrcd; + } + + attr.event_handler = ib_uverbs_qp_event_handler; + attr.qp_num = cmd.qpn; + attr.qp_type = cmd.qp_type; + + INIT_LIST_HEAD(&obj->uevent.event_list); + INIT_LIST_HEAD(&obj->mcast_list); + + qp = ib_open_qp(xrcd, &attr); + if (IS_ERR(qp)) { + ret = PTR_ERR(qp); + goto err_xrcd; + } + + obj->uevent.uobject.object = qp; + obj->uevent.uobject.user_handle = cmd.user_handle; + + obj->uxrcd = container_of(xrcd_uobj, struct ib_uxrcd_object, uobject); + atomic_inc(&obj->uxrcd->refcnt); + qp->uobject = obj; + uobj_put_read(xrcd_uobj); + uobj_finalize_uobj_create(&obj->uevent.uobject, attrs); + + resp.qpn = qp->qp_num; + resp.qp_handle = obj->uevent.uobject.id; + return uverbs_response(attrs, &resp, sizeof(resp)); + +err_xrcd: + uobj_put_read(xrcd_uobj); +err_put: + uobj_alloc_abort(&obj->uevent.uobject, attrs); + return ret; +} + +static void copy_ah_attr_to_uverbs(struct ib_uverbs_qp_dest *uverb_attr, + struct rdma_ah_attr *rdma_attr) +{ + const struct ib_global_route *grh; + + uverb_attr->dlid = rdma_ah_get_dlid(rdma_attr); + uverb_attr->sl = rdma_ah_get_sl(rdma_attr); + uverb_attr->src_path_bits = rdma_ah_get_path_bits(rdma_attr); + uverb_attr->static_rate = rdma_ah_get_static_rate(rdma_attr); + uverb_attr->is_global = !!(rdma_ah_get_ah_flags(rdma_attr) & + IB_AH_GRH); + if (uverb_attr->is_global) { + grh = rdma_ah_read_grh(rdma_attr); + memcpy(uverb_attr->dgid, grh->dgid.raw, 16); + uverb_attr->flow_label = grh->flow_label; + uverb_attr->sgid_index = grh->sgid_index; + uverb_attr->hop_limit = grh->hop_limit; + uverb_attr->traffic_class = grh->traffic_class; + } + uverb_attr->port_num = rdma_ah_get_port_num(rdma_attr); +} + +static int ib_uverbs_query_qp(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_query_qp cmd; + struct ib_uverbs_query_qp_resp resp; + struct ib_qp *qp; + struct ib_qp_attr *attr; + struct ib_qp_init_attr *init_attr; + int ret; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + attr = kmalloc(sizeof *attr, GFP_KERNEL); + init_attr = kmalloc(sizeof *init_attr, GFP_KERNEL); + if (!attr || !init_attr) { + ret = -ENOMEM; + goto out; + } + + qp = uobj_get_obj_read(qp, UVERBS_OBJECT_QP, cmd.qp_handle, attrs); + if (!qp) { + ret = -EINVAL; + goto out; + } + + ret = ib_query_qp(qp, attr, cmd.attr_mask, init_attr); + + rdma_lookup_put_uobject(&qp->uobject->uevent.uobject, + UVERBS_LOOKUP_READ); + + if (ret) + goto out; + + memset(&resp, 0, sizeof resp); + + resp.qp_state = attr->qp_state; + resp.cur_qp_state = attr->cur_qp_state; + resp.path_mtu = attr->path_mtu; + resp.path_mig_state = attr->path_mig_state; + resp.qkey = attr->qkey; + resp.rq_psn = attr->rq_psn; + resp.sq_psn = attr->sq_psn; + resp.dest_qp_num = attr->dest_qp_num; + resp.qp_access_flags = attr->qp_access_flags; + resp.pkey_index = attr->pkey_index; + resp.alt_pkey_index = attr->alt_pkey_index; + resp.sq_draining = attr->sq_draining; + resp.max_rd_atomic = attr->max_rd_atomic; + resp.max_dest_rd_atomic = attr->max_dest_rd_atomic; + resp.min_rnr_timer = attr->min_rnr_timer; + resp.port_num = attr->port_num; + resp.timeout = attr->timeout; + resp.retry_cnt = attr->retry_cnt; + resp.rnr_retry = attr->rnr_retry; + resp.alt_port_num = attr->alt_port_num; + resp.alt_timeout = attr->alt_timeout; + + copy_ah_attr_to_uverbs(&resp.dest, &attr->ah_attr); + copy_ah_attr_to_uverbs(&resp.alt_dest, &attr->alt_ah_attr); + + resp.max_send_wr = init_attr->cap.max_send_wr; + resp.max_recv_wr = init_attr->cap.max_recv_wr; + resp.max_send_sge = init_attr->cap.max_send_sge; + resp.max_recv_sge = init_attr->cap.max_recv_sge; + resp.max_inline_data = init_attr->cap.max_inline_data; + resp.sq_sig_all = init_attr->sq_sig_type == IB_SIGNAL_ALL_WR; + + ret = uverbs_response(attrs, &resp, sizeof(resp)); + +out: + kfree(attr); + kfree(init_attr); + + return ret; +} + +/* Remove ignored fields set in the attribute mask */ +static int modify_qp_mask(enum ib_qp_type qp_type, int mask) +{ + switch (qp_type) { + case IB_QPT_XRC_INI: + return mask & ~(IB_QP_MAX_DEST_RD_ATOMIC | IB_QP_MIN_RNR_TIMER); + case IB_QPT_XRC_TGT: + return mask & ~(IB_QP_MAX_QP_RD_ATOMIC | IB_QP_RETRY_CNT | + IB_QP_RNR_RETRY); + default: + return mask; + } +} + +static void copy_ah_attr_from_uverbs(struct ib_device *dev, + struct rdma_ah_attr *rdma_attr, + struct ib_uverbs_qp_dest *uverb_attr) +{ + rdma_attr->type = rdma_ah_find_type(dev, uverb_attr->port_num); + if (uverb_attr->is_global) { + rdma_ah_set_grh(rdma_attr, NULL, + uverb_attr->flow_label, + uverb_attr->sgid_index, + uverb_attr->hop_limit, + uverb_attr->traffic_class); + rdma_ah_set_dgid_raw(rdma_attr, uverb_attr->dgid); + } else { + rdma_ah_set_ah_flags(rdma_attr, 0); + } + rdma_ah_set_dlid(rdma_attr, uverb_attr->dlid); + rdma_ah_set_sl(rdma_attr, uverb_attr->sl); + rdma_ah_set_path_bits(rdma_attr, uverb_attr->src_path_bits); + rdma_ah_set_static_rate(rdma_attr, uverb_attr->static_rate); + rdma_ah_set_port_num(rdma_attr, uverb_attr->port_num); + rdma_ah_set_make_grd(rdma_attr, false); +} + +static int modify_qp(struct uverbs_attr_bundle *attrs, + struct ib_uverbs_ex_modify_qp *cmd) +{ + struct ib_qp_attr *attr; + struct ib_qp *qp; + int ret; + + attr = kzalloc(sizeof(*attr), GFP_KERNEL); + if (!attr) + return -ENOMEM; + + qp = uobj_get_obj_read(qp, UVERBS_OBJECT_QP, cmd->base.qp_handle, + attrs); + if (!qp) { + ret = -EINVAL; + goto out; + } + + if ((cmd->base.attr_mask & IB_QP_PORT) && + !rdma_is_port_valid(qp->device, cmd->base.port_num)) { + ret = -EINVAL; + goto release_qp; + } + + if ((cmd->base.attr_mask & IB_QP_AV)) { + if (!rdma_is_port_valid(qp->device, cmd->base.dest.port_num)) { + ret = -EINVAL; + goto release_qp; + } + + if (cmd->base.attr_mask & IB_QP_STATE && + cmd->base.qp_state == IB_QPS_RTR) { + /* We are in INIT->RTR TRANSITION (if we are not, + * this transition will be rejected in subsequent checks). + * In the INIT->RTR transition, we cannot have IB_QP_PORT set, + * but the IB_QP_STATE flag is required. + * + * Since kernel 3.14 (commit dbf727de7440), the uverbs driver, + * when IB_QP_AV is set, has required inclusion of a valid + * port number in the primary AV. (AVs are created and handled + * differently for infiniband and ethernet (RoCE) ports). + * + * Check the port number included in the primary AV against + * the port number in the qp struct, which was set (and saved) + * in the RST->INIT transition. + */ + if (cmd->base.dest.port_num != qp->real_qp->port) { + ret = -EINVAL; + goto release_qp; + } + } else { + /* We are in SQD->SQD. (If we are not, this transition will + * be rejected later in the verbs layer checks). + * Check for both IB_QP_PORT and IB_QP_AV, these can be set + * together in the SQD->SQD transition. + * + * If only IP_QP_AV was set, add in IB_QP_PORT as well (the + * verbs layer driver does not track primary port changes + * resulting from path migration. Thus, in SQD, if the primary + * AV is modified, the primary port should also be modified). + * + * Note that in this transition, the IB_QP_STATE flag + * is not allowed. + */ + if (((cmd->base.attr_mask & (IB_QP_AV | IB_QP_PORT)) + == (IB_QP_AV | IB_QP_PORT)) && + cmd->base.port_num != cmd->base.dest.port_num) { + ret = -EINVAL; + goto release_qp; + } + if ((cmd->base.attr_mask & (IB_QP_AV | IB_QP_PORT)) + == IB_QP_AV) { + cmd->base.attr_mask |= IB_QP_PORT; + cmd->base.port_num = cmd->base.dest.port_num; + } + } + } + + if ((cmd->base.attr_mask & IB_QP_ALT_PATH) && + (!rdma_is_port_valid(qp->device, cmd->base.alt_port_num) || + !rdma_is_port_valid(qp->device, cmd->base.alt_dest.port_num) || + cmd->base.alt_port_num != cmd->base.alt_dest.port_num)) { + ret = -EINVAL; + goto release_qp; + } + + if ((cmd->base.attr_mask & IB_QP_CUR_STATE && + cmd->base.cur_qp_state > IB_QPS_ERR) || + (cmd->base.attr_mask & IB_QP_STATE && + cmd->base.qp_state > IB_QPS_ERR)) { + ret = -EINVAL; + goto release_qp; + } + + if (cmd->base.attr_mask & IB_QP_STATE) + attr->qp_state = cmd->base.qp_state; + if (cmd->base.attr_mask & IB_QP_CUR_STATE) + attr->cur_qp_state = cmd->base.cur_qp_state; + if (cmd->base.attr_mask & IB_QP_PATH_MTU) + attr->path_mtu = cmd->base.path_mtu; + if (cmd->base.attr_mask & IB_QP_PATH_MIG_STATE) + attr->path_mig_state = cmd->base.path_mig_state; + if (cmd->base.attr_mask & IB_QP_QKEY) { + if (cmd->base.qkey & IB_QP_SET_QKEY && !capable(CAP_NET_RAW)) { + ret = -EPERM; + goto release_qp; + } + attr->qkey = cmd->base.qkey; + } + if (cmd->base.attr_mask & IB_QP_RQ_PSN) + attr->rq_psn = cmd->base.rq_psn; + if (cmd->base.attr_mask & IB_QP_SQ_PSN) + attr->sq_psn = cmd->base.sq_psn; + if (cmd->base.attr_mask & IB_QP_DEST_QPN) + attr->dest_qp_num = cmd->base.dest_qp_num; + if (cmd->base.attr_mask & IB_QP_ACCESS_FLAGS) + attr->qp_access_flags = cmd->base.qp_access_flags; + if (cmd->base.attr_mask & IB_QP_PKEY_INDEX) + attr->pkey_index = cmd->base.pkey_index; + if (cmd->base.attr_mask & IB_QP_EN_SQD_ASYNC_NOTIFY) + attr->en_sqd_async_notify = cmd->base.en_sqd_async_notify; + if (cmd->base.attr_mask & IB_QP_MAX_QP_RD_ATOMIC) + attr->max_rd_atomic = cmd->base.max_rd_atomic; + if (cmd->base.attr_mask & IB_QP_MAX_DEST_RD_ATOMIC) + attr->max_dest_rd_atomic = cmd->base.max_dest_rd_atomic; + if (cmd->base.attr_mask & IB_QP_MIN_RNR_TIMER) + attr->min_rnr_timer = cmd->base.min_rnr_timer; + if (cmd->base.attr_mask & IB_QP_PORT) + attr->port_num = cmd->base.port_num; + if (cmd->base.attr_mask & IB_QP_TIMEOUT) + attr->timeout = cmd->base.timeout; + if (cmd->base.attr_mask & IB_QP_RETRY_CNT) + attr->retry_cnt = cmd->base.retry_cnt; + if (cmd->base.attr_mask & IB_QP_RNR_RETRY) + attr->rnr_retry = cmd->base.rnr_retry; + if (cmd->base.attr_mask & IB_QP_ALT_PATH) { + attr->alt_port_num = cmd->base.alt_port_num; + attr->alt_timeout = cmd->base.alt_timeout; + attr->alt_pkey_index = cmd->base.alt_pkey_index; + } + if (cmd->base.attr_mask & IB_QP_RATE_LIMIT) + attr->rate_limit = cmd->rate_limit; + + if (cmd->base.attr_mask & IB_QP_AV) + copy_ah_attr_from_uverbs(qp->device, &attr->ah_attr, + &cmd->base.dest); + + if (cmd->base.attr_mask & IB_QP_ALT_PATH) + copy_ah_attr_from_uverbs(qp->device, &attr->alt_ah_attr, + &cmd->base.alt_dest); + + ret = ib_modify_qp_with_udata(qp, attr, + modify_qp_mask(qp->qp_type, + cmd->base.attr_mask), + &attrs->driver_udata); + +release_qp: + rdma_lookup_put_uobject(&qp->uobject->uevent.uobject, + UVERBS_LOOKUP_READ); +out: + kfree(attr); + + return ret; +} + +static int ib_uverbs_modify_qp(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_ex_modify_qp cmd; + int ret; + + ret = uverbs_request(attrs, &cmd.base, sizeof(cmd.base)); + if (ret) + return ret; + + if (cmd.base.attr_mask & + ~((IB_USER_LEGACY_LAST_QP_ATTR_MASK << 1) - 1)) + return -EOPNOTSUPP; + + return modify_qp(attrs, &cmd); +} + +static int ib_uverbs_ex_modify_qp(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_ex_modify_qp cmd; + struct ib_uverbs_ex_modify_qp_resp resp = { + .response_length = uverbs_response_length(attrs, sizeof(resp)) + }; + int ret; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + /* + * Last bit is reserved for extending the attr_mask by + * using another field. + */ + BUILD_BUG_ON(IB_USER_LAST_QP_ATTR_MASK == (1ULL << 31)); + + if (cmd.base.attr_mask & + ~((IB_USER_LAST_QP_ATTR_MASK << 1) - 1)) + return -EOPNOTSUPP; + + ret = modify_qp(attrs, &cmd); + if (ret) + return ret; + + return uverbs_response(attrs, &resp, sizeof(resp)); +} + +static int ib_uverbs_destroy_qp(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_destroy_qp cmd; + struct ib_uverbs_destroy_qp_resp resp; + struct ib_uobject *uobj; + struct ib_uqp_object *obj; + int ret; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + uobj = uobj_get_destroy(UVERBS_OBJECT_QP, cmd.qp_handle, attrs); + if (IS_ERR(uobj)) + return PTR_ERR(uobj); + + obj = container_of(uobj, struct ib_uqp_object, uevent.uobject); + memset(&resp, 0, sizeof(resp)); + resp.events_reported = obj->uevent.events_reported; + + uobj_put_destroy(uobj); + + return uverbs_response(attrs, &resp, sizeof(resp)); +} + +static void *alloc_wr(size_t wr_size, __u32 num_sge) +{ + if (num_sge >= (U32_MAX - ALIGN(wr_size, sizeof (struct ib_sge))) / + sizeof (struct ib_sge)) + return NULL; + + return kmalloc(ALIGN(wr_size, sizeof (struct ib_sge)) + + num_sge * sizeof (struct ib_sge), GFP_KERNEL); +} + +static int ib_uverbs_post_send(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_post_send cmd; + struct ib_uverbs_post_send_resp resp; + struct ib_uverbs_send_wr *user_wr; + struct ib_send_wr *wr = NULL, *last, *next; + const struct ib_send_wr *bad_wr; + struct ib_qp *qp; + int i, sg_ind; + int is_ud; + int ret, ret2; + size_t next_size; + const struct ib_sge __user *sgls; + const void __user *wqes; + struct uverbs_req_iter iter; + + ret = uverbs_request_start(attrs, &iter, &cmd, sizeof(cmd)); + if (ret) + return ret; + wqes = uverbs_request_next_ptr(&iter, cmd.wqe_size * cmd.wr_count); + if (IS_ERR(wqes)) + return PTR_ERR(wqes); + sgls = uverbs_request_next_ptr( + &iter, cmd.sge_count * sizeof(struct ib_uverbs_sge)); + if (IS_ERR(sgls)) + return PTR_ERR(sgls); + ret = uverbs_request_finish(&iter); + if (ret) + return ret; + + user_wr = kmalloc(cmd.wqe_size, GFP_KERNEL); + if (!user_wr) + return -ENOMEM; + + qp = uobj_get_obj_read(qp, UVERBS_OBJECT_QP, cmd.qp_handle, attrs); + if (!qp) { + ret = -EINVAL; + goto out; + } + + is_ud = qp->qp_type == IB_QPT_UD; + sg_ind = 0; + last = NULL; + for (i = 0; i < cmd.wr_count; ++i) { + if (copy_from_user(user_wr, wqes + i * cmd.wqe_size, + cmd.wqe_size)) { + ret = -EFAULT; + goto out_put; + } + + if (user_wr->num_sge + sg_ind > cmd.sge_count) { + ret = -EINVAL; + goto out_put; + } + + if (is_ud) { + struct ib_ud_wr *ud; + + if (user_wr->opcode != IB_WR_SEND && + user_wr->opcode != IB_WR_SEND_WITH_IMM) { + ret = -EINVAL; + goto out_put; + } + + next_size = sizeof(*ud); + ud = alloc_wr(next_size, user_wr->num_sge); + if (!ud) { + ret = -ENOMEM; + goto out_put; + } + + ud->ah = uobj_get_obj_read(ah, UVERBS_OBJECT_AH, + user_wr->wr.ud.ah, attrs); + if (!ud->ah) { + kfree(ud); + ret = -EINVAL; + goto out_put; + } + ud->remote_qpn = user_wr->wr.ud.remote_qpn; + ud->remote_qkey = user_wr->wr.ud.remote_qkey; + + next = &ud->wr; + } else if (user_wr->opcode == IB_WR_RDMA_WRITE_WITH_IMM || + user_wr->opcode == IB_WR_RDMA_WRITE || + user_wr->opcode == IB_WR_RDMA_READ) { + struct ib_rdma_wr *rdma; + + next_size = sizeof(*rdma); + rdma = alloc_wr(next_size, user_wr->num_sge); + if (!rdma) { + ret = -ENOMEM; + goto out_put; + } + + rdma->remote_addr = user_wr->wr.rdma.remote_addr; + rdma->rkey = user_wr->wr.rdma.rkey; + + next = &rdma->wr; + } else if (user_wr->opcode == IB_WR_ATOMIC_CMP_AND_SWP || + user_wr->opcode == IB_WR_ATOMIC_FETCH_AND_ADD) { + struct ib_atomic_wr *atomic; + + next_size = sizeof(*atomic); + atomic = alloc_wr(next_size, user_wr->num_sge); + if (!atomic) { + ret = -ENOMEM; + goto out_put; + } + + atomic->remote_addr = user_wr->wr.atomic.remote_addr; + atomic->compare_add = user_wr->wr.atomic.compare_add; + atomic->swap = user_wr->wr.atomic.swap; + atomic->rkey = user_wr->wr.atomic.rkey; + + next = &atomic->wr; + } else if (user_wr->opcode == IB_WR_SEND || + user_wr->opcode == IB_WR_SEND_WITH_IMM || + user_wr->opcode == IB_WR_SEND_WITH_INV) { + next_size = sizeof(*next); + next = alloc_wr(next_size, user_wr->num_sge); + if (!next) { + ret = -ENOMEM; + goto out_put; + } + } else { + ret = -EINVAL; + goto out_put; + } + + if (user_wr->opcode == IB_WR_SEND_WITH_IMM || + user_wr->opcode == IB_WR_RDMA_WRITE_WITH_IMM) { + next->ex.imm_data = + (__be32 __force) user_wr->ex.imm_data; + } else if (user_wr->opcode == IB_WR_SEND_WITH_INV) { + next->ex.invalidate_rkey = user_wr->ex.invalidate_rkey; + } + + if (!last) + wr = next; + else + last->next = next; + last = next; + + next->next = NULL; + next->wr_id = user_wr->wr_id; + next->num_sge = user_wr->num_sge; + next->opcode = user_wr->opcode; + next->send_flags = user_wr->send_flags; + + if (next->num_sge) { + next->sg_list = (void *) next + + ALIGN(next_size, sizeof(struct ib_sge)); + if (copy_from_user(next->sg_list, sgls + sg_ind, + next->num_sge * + sizeof(struct ib_sge))) { + ret = -EFAULT; + goto out_put; + } + sg_ind += next->num_sge; + } else + next->sg_list = NULL; + } + + resp.bad_wr = 0; + ret = qp->device->ops.post_send(qp->real_qp, wr, &bad_wr); + if (ret) + for (next = wr; next; next = next->next) { + ++resp.bad_wr; + if (next == bad_wr) + break; + } + + ret2 = uverbs_response(attrs, &resp, sizeof(resp)); + if (ret2) + ret = ret2; + +out_put: + rdma_lookup_put_uobject(&qp->uobject->uevent.uobject, + UVERBS_LOOKUP_READ); + + while (wr) { + if (is_ud && ud_wr(wr)->ah) + uobj_put_obj_read(ud_wr(wr)->ah); + next = wr->next; + kfree(wr); + wr = next; + } + +out: + kfree(user_wr); + + return ret; +} + +static struct ib_recv_wr * +ib_uverbs_unmarshall_recv(struct uverbs_req_iter *iter, u32 wr_count, + u32 wqe_size, u32 sge_count) +{ + struct ib_uverbs_recv_wr *user_wr; + struct ib_recv_wr *wr = NULL, *last, *next; + int sg_ind; + int i; + int ret; + const struct ib_sge __user *sgls; + const void __user *wqes; + + if (wqe_size < sizeof (struct ib_uverbs_recv_wr)) + return ERR_PTR(-EINVAL); + + wqes = uverbs_request_next_ptr(iter, wqe_size * wr_count); + if (IS_ERR(wqes)) + return ERR_CAST(wqes); + sgls = uverbs_request_next_ptr( + iter, sge_count * sizeof(struct ib_uverbs_sge)); + if (IS_ERR(sgls)) + return ERR_CAST(sgls); + ret = uverbs_request_finish(iter); + if (ret) + return ERR_PTR(ret); + + user_wr = kmalloc(wqe_size, GFP_KERNEL); + if (!user_wr) + return ERR_PTR(-ENOMEM); + + sg_ind = 0; + last = NULL; + for (i = 0; i < wr_count; ++i) { + if (copy_from_user(user_wr, wqes + i * wqe_size, + wqe_size)) { + ret = -EFAULT; + goto err; + } + + if (user_wr->num_sge + sg_ind > sge_count) { + ret = -EINVAL; + goto err; + } + + if (user_wr->num_sge >= + (U32_MAX - ALIGN(sizeof *next, sizeof (struct ib_sge))) / + sizeof (struct ib_sge)) { + ret = -EINVAL; + goto err; + } + + next = kmalloc(ALIGN(sizeof *next, sizeof (struct ib_sge)) + + user_wr->num_sge * sizeof (struct ib_sge), + GFP_KERNEL); + if (!next) { + ret = -ENOMEM; + goto err; + } + + if (!last) + wr = next; + else + last->next = next; + last = next; + + next->next = NULL; + next->wr_id = user_wr->wr_id; + next->num_sge = user_wr->num_sge; + + if (next->num_sge) { + next->sg_list = (void *) next + + ALIGN(sizeof *next, sizeof (struct ib_sge)); + if (copy_from_user(next->sg_list, sgls + sg_ind, + next->num_sge * + sizeof(struct ib_sge))) { + ret = -EFAULT; + goto err; + } + sg_ind += next->num_sge; + } else + next->sg_list = NULL; + } + + kfree(user_wr); + return wr; + +err: + kfree(user_wr); + + while (wr) { + next = wr->next; + kfree(wr); + wr = next; + } + + return ERR_PTR(ret); +} + +static int ib_uverbs_post_recv(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_post_recv cmd; + struct ib_uverbs_post_recv_resp resp; + struct ib_recv_wr *wr, *next; + const struct ib_recv_wr *bad_wr; + struct ib_qp *qp; + int ret, ret2; + struct uverbs_req_iter iter; + + ret = uverbs_request_start(attrs, &iter, &cmd, sizeof(cmd)); + if (ret) + return ret; + + wr = ib_uverbs_unmarshall_recv(&iter, cmd.wr_count, cmd.wqe_size, + cmd.sge_count); + if (IS_ERR(wr)) + return PTR_ERR(wr); + + qp = uobj_get_obj_read(qp, UVERBS_OBJECT_QP, cmd.qp_handle, attrs); + if (!qp) { + ret = -EINVAL; + goto out; + } + + resp.bad_wr = 0; + ret = qp->device->ops.post_recv(qp->real_qp, wr, &bad_wr); + + rdma_lookup_put_uobject(&qp->uobject->uevent.uobject, + UVERBS_LOOKUP_READ); + if (ret) { + for (next = wr; next; next = next->next) { + ++resp.bad_wr; + if (next == bad_wr) + break; + } + } + + ret2 = uverbs_response(attrs, &resp, sizeof(resp)); + if (ret2) + ret = ret2; +out: + while (wr) { + next = wr->next; + kfree(wr); + wr = next; + } + + return ret; +} + +static int ib_uverbs_post_srq_recv(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_post_srq_recv cmd; + struct ib_uverbs_post_srq_recv_resp resp; + struct ib_recv_wr *wr, *next; + const struct ib_recv_wr *bad_wr; + struct ib_srq *srq; + int ret, ret2; + struct uverbs_req_iter iter; + + ret = uverbs_request_start(attrs, &iter, &cmd, sizeof(cmd)); + if (ret) + return ret; + + wr = ib_uverbs_unmarshall_recv(&iter, cmd.wr_count, cmd.wqe_size, + cmd.sge_count); + if (IS_ERR(wr)) + return PTR_ERR(wr); + + srq = uobj_get_obj_read(srq, UVERBS_OBJECT_SRQ, cmd.srq_handle, attrs); + if (!srq) { + ret = -EINVAL; + goto out; + } + + resp.bad_wr = 0; + ret = srq->device->ops.post_srq_recv(srq, wr, &bad_wr); + + rdma_lookup_put_uobject(&srq->uobject->uevent.uobject, + UVERBS_LOOKUP_READ); + + if (ret) + for (next = wr; next; next = next->next) { + ++resp.bad_wr; + if (next == bad_wr) + break; + } + + ret2 = uverbs_response(attrs, &resp, sizeof(resp)); + if (ret2) + ret = ret2; + +out: + while (wr) { + next = wr->next; + kfree(wr); + wr = next; + } + + return ret; +} + +static int ib_uverbs_create_ah(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_create_ah cmd; + struct ib_uverbs_create_ah_resp resp; + struct ib_uobject *uobj; + struct ib_pd *pd; + struct ib_ah *ah; + struct rdma_ah_attr attr = {}; + int ret; + struct ib_device *ib_dev; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + uobj = uobj_alloc(UVERBS_OBJECT_AH, attrs, &ib_dev); + if (IS_ERR(uobj)) + return PTR_ERR(uobj); + + if (!rdma_is_port_valid(ib_dev, cmd.attr.port_num)) { + ret = -EINVAL; + goto err; + } + + pd = uobj_get_obj_read(pd, UVERBS_OBJECT_PD, cmd.pd_handle, attrs); + if (!pd) { + ret = -EINVAL; + goto err; + } + + attr.type = rdma_ah_find_type(ib_dev, cmd.attr.port_num); + rdma_ah_set_make_grd(&attr, false); + rdma_ah_set_dlid(&attr, cmd.attr.dlid); + rdma_ah_set_sl(&attr, cmd.attr.sl); + rdma_ah_set_path_bits(&attr, cmd.attr.src_path_bits); + rdma_ah_set_static_rate(&attr, cmd.attr.static_rate); + rdma_ah_set_port_num(&attr, cmd.attr.port_num); + + if (cmd.attr.is_global) { + rdma_ah_set_grh(&attr, NULL, cmd.attr.grh.flow_label, + cmd.attr.grh.sgid_index, + cmd.attr.grh.hop_limit, + cmd.attr.grh.traffic_class); + rdma_ah_set_dgid_raw(&attr, cmd.attr.grh.dgid); + } else { + rdma_ah_set_ah_flags(&attr, 0); + } + + ah = rdma_create_user_ah(pd, &attr, &attrs->driver_udata); + if (IS_ERR(ah)) { + ret = PTR_ERR(ah); + goto err_put; + } + + ah->uobject = uobj; + uobj->user_handle = cmd.user_handle; + uobj->object = ah; + uobj_put_obj_read(pd); + uobj_finalize_uobj_create(uobj, attrs); + + resp.ah_handle = uobj->id; + return uverbs_response(attrs, &resp, sizeof(resp)); + +err_put: + uobj_put_obj_read(pd); +err: + uobj_alloc_abort(uobj, attrs); + return ret; +} + +static int ib_uverbs_destroy_ah(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_destroy_ah cmd; + int ret; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + return uobj_perform_destroy(UVERBS_OBJECT_AH, cmd.ah_handle, attrs); +} + +static int ib_uverbs_attach_mcast(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_attach_mcast cmd; + struct ib_qp *qp; + struct ib_uqp_object *obj; + struct ib_uverbs_mcast_entry *mcast; + int ret; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + qp = uobj_get_obj_read(qp, UVERBS_OBJECT_QP, cmd.qp_handle, attrs); + if (!qp) + return -EINVAL; + + obj = qp->uobject; + + mutex_lock(&obj->mcast_lock); + list_for_each_entry(mcast, &obj->mcast_list, list) + if (cmd.mlid == mcast->lid && + !memcmp(cmd.gid, mcast->gid.raw, sizeof mcast->gid.raw)) { + ret = 0; + goto out_put; + } + + mcast = kmalloc(sizeof *mcast, GFP_KERNEL); + if (!mcast) { + ret = -ENOMEM; + goto out_put; + } + + mcast->lid = cmd.mlid; + memcpy(mcast->gid.raw, cmd.gid, sizeof mcast->gid.raw); + + ret = ib_attach_mcast(qp, &mcast->gid, cmd.mlid); + if (!ret) + list_add_tail(&mcast->list, &obj->mcast_list); + else + kfree(mcast); + +out_put: + mutex_unlock(&obj->mcast_lock); + rdma_lookup_put_uobject(&qp->uobject->uevent.uobject, + UVERBS_LOOKUP_READ); + + return ret; +} + +static int ib_uverbs_detach_mcast(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_detach_mcast cmd; + struct ib_uqp_object *obj; + struct ib_qp *qp; + struct ib_uverbs_mcast_entry *mcast; + int ret; + bool found = false; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + qp = uobj_get_obj_read(qp, UVERBS_OBJECT_QP, cmd.qp_handle, attrs); + if (!qp) + return -EINVAL; + + obj = qp->uobject; + mutex_lock(&obj->mcast_lock); + + list_for_each_entry(mcast, &obj->mcast_list, list) + if (cmd.mlid == mcast->lid && + !memcmp(cmd.gid, mcast->gid.raw, sizeof mcast->gid.raw)) { + list_del(&mcast->list); + kfree(mcast); + found = true; + break; + } + + if (!found) { + ret = -EINVAL; + goto out_put; + } + + ret = ib_detach_mcast(qp, (union ib_gid *)cmd.gid, cmd.mlid); + +out_put: + mutex_unlock(&obj->mcast_lock); + rdma_lookup_put_uobject(&qp->uobject->uevent.uobject, + UVERBS_LOOKUP_READ); + return ret; +} + +struct ib_uflow_resources *flow_resources_alloc(size_t num_specs) +{ + struct ib_uflow_resources *resources; + + resources = kzalloc(sizeof(*resources), GFP_KERNEL); + + if (!resources) + return NULL; + + if (!num_specs) + goto out; + + resources->counters = + kcalloc(num_specs, sizeof(*resources->counters), GFP_KERNEL); + resources->collection = + kcalloc(num_specs, sizeof(*resources->collection), GFP_KERNEL); + + if (!resources->counters || !resources->collection) + goto err; + +out: + resources->max = num_specs; + return resources; + +err: + kfree(resources->counters); + kfree(resources); + + return NULL; +} +EXPORT_SYMBOL(flow_resources_alloc); + +void ib_uverbs_flow_resources_free(struct ib_uflow_resources *uflow_res) +{ + unsigned int i; + + if (!uflow_res) + return; + + for (i = 0; i < uflow_res->collection_num; i++) + atomic_dec(&uflow_res->collection[i]->usecnt); + + for (i = 0; i < uflow_res->counters_num; i++) + atomic_dec(&uflow_res->counters[i]->usecnt); + + kfree(uflow_res->collection); + kfree(uflow_res->counters); + kfree(uflow_res); +} +EXPORT_SYMBOL(ib_uverbs_flow_resources_free); + +void flow_resources_add(struct ib_uflow_resources *uflow_res, + enum ib_flow_spec_type type, + void *ibobj) +{ + WARN_ON(uflow_res->num >= uflow_res->max); + + switch (type) { + case IB_FLOW_SPEC_ACTION_HANDLE: + atomic_inc(&((struct ib_flow_action *)ibobj)->usecnt); + uflow_res->collection[uflow_res->collection_num++] = + (struct ib_flow_action *)ibobj; + break; + case IB_FLOW_SPEC_ACTION_COUNT: + atomic_inc(&((struct ib_counters *)ibobj)->usecnt); + uflow_res->counters[uflow_res->counters_num++] = + (struct ib_counters *)ibobj; + break; + default: + WARN_ON(1); + } + + uflow_res->num++; +} +EXPORT_SYMBOL(flow_resources_add); + +static int kern_spec_to_ib_spec_action(struct uverbs_attr_bundle *attrs, + struct ib_uverbs_flow_spec *kern_spec, + union ib_flow_spec *ib_spec, + struct ib_uflow_resources *uflow_res) +{ + ib_spec->type = kern_spec->type; + switch (ib_spec->type) { + case IB_FLOW_SPEC_ACTION_TAG: + if (kern_spec->flow_tag.size != + sizeof(struct ib_uverbs_flow_spec_action_tag)) + return -EINVAL; + + ib_spec->flow_tag.size = sizeof(struct ib_flow_spec_action_tag); + ib_spec->flow_tag.tag_id = kern_spec->flow_tag.tag_id; + break; + case IB_FLOW_SPEC_ACTION_DROP: + if (kern_spec->drop.size != + sizeof(struct ib_uverbs_flow_spec_action_drop)) + return -EINVAL; + + ib_spec->drop.size = sizeof(struct ib_flow_spec_action_drop); + break; + case IB_FLOW_SPEC_ACTION_HANDLE: + if (kern_spec->action.size != + sizeof(struct ib_uverbs_flow_spec_action_handle)) + return -EOPNOTSUPP; + ib_spec->action.act = uobj_get_obj_read(flow_action, + UVERBS_OBJECT_FLOW_ACTION, + kern_spec->action.handle, + attrs); + if (!ib_spec->action.act) + return -EINVAL; + ib_spec->action.size = + sizeof(struct ib_flow_spec_action_handle); + flow_resources_add(uflow_res, + IB_FLOW_SPEC_ACTION_HANDLE, + ib_spec->action.act); + uobj_put_obj_read(ib_spec->action.act); + break; + case IB_FLOW_SPEC_ACTION_COUNT: + if (kern_spec->flow_count.size != + sizeof(struct ib_uverbs_flow_spec_action_count)) + return -EINVAL; + ib_spec->flow_count.counters = + uobj_get_obj_read(counters, + UVERBS_OBJECT_COUNTERS, + kern_spec->flow_count.handle, + attrs); + if (!ib_spec->flow_count.counters) + return -EINVAL; + ib_spec->flow_count.size = + sizeof(struct ib_flow_spec_action_count); + flow_resources_add(uflow_res, + IB_FLOW_SPEC_ACTION_COUNT, + ib_spec->flow_count.counters); + uobj_put_obj_read(ib_spec->flow_count.counters); + break; + default: + return -EINVAL; + } + return 0; +} + +static ssize_t spec_filter_size(const void *kern_spec_filter, u16 kern_filter_size, + u16 ib_real_filter_sz) +{ + /* + * User space filter structures must be 64 bit aligned, otherwise this + * may pass, but we won't handle additional new attributes. + */ + + if (kern_filter_size > ib_real_filter_sz) { + if (memchr_inv(kern_spec_filter + + ib_real_filter_sz, 0, + kern_filter_size - ib_real_filter_sz)) + return -EINVAL; + return ib_real_filter_sz; + } + return kern_filter_size; +} + +int ib_uverbs_kern_spec_to_ib_spec_filter(enum ib_flow_spec_type type, + const void *kern_spec_mask, + const void *kern_spec_val, + size_t kern_filter_sz, + union ib_flow_spec *ib_spec) +{ + ssize_t actual_filter_sz; + ssize_t ib_filter_sz; + + /* User flow spec size must be aligned to 4 bytes */ + if (kern_filter_sz != ALIGN(kern_filter_sz, 4)) + return -EINVAL; + + ib_spec->type = type; + + if (ib_spec->type == (IB_FLOW_SPEC_INNER | IB_FLOW_SPEC_VXLAN_TUNNEL)) + return -EINVAL; + + switch (ib_spec->type & ~IB_FLOW_SPEC_INNER) { + case IB_FLOW_SPEC_ETH: + ib_filter_sz = offsetof(struct ib_flow_eth_filter, real_sz); + actual_filter_sz = spec_filter_size(kern_spec_mask, + kern_filter_sz, + ib_filter_sz); + if (actual_filter_sz <= 0) + return -EINVAL; + ib_spec->size = sizeof(struct ib_flow_spec_eth); + memcpy(&ib_spec->eth.val, kern_spec_val, actual_filter_sz); + memcpy(&ib_spec->eth.mask, kern_spec_mask, actual_filter_sz); + break; + case IB_FLOW_SPEC_IPV4: + ib_filter_sz = offsetof(struct ib_flow_ipv4_filter, real_sz); + actual_filter_sz = spec_filter_size(kern_spec_mask, + kern_filter_sz, + ib_filter_sz); + if (actual_filter_sz <= 0) + return -EINVAL; + ib_spec->size = sizeof(struct ib_flow_spec_ipv4); + memcpy(&ib_spec->ipv4.val, kern_spec_val, actual_filter_sz); + memcpy(&ib_spec->ipv4.mask, kern_spec_mask, actual_filter_sz); + break; + case IB_FLOW_SPEC_IPV6: + ib_filter_sz = offsetof(struct ib_flow_ipv6_filter, real_sz); + actual_filter_sz = spec_filter_size(kern_spec_mask, + kern_filter_sz, + ib_filter_sz); + if (actual_filter_sz <= 0) + return -EINVAL; + ib_spec->size = sizeof(struct ib_flow_spec_ipv6); + memcpy(&ib_spec->ipv6.val, kern_spec_val, actual_filter_sz); + memcpy(&ib_spec->ipv6.mask, kern_spec_mask, actual_filter_sz); + + if ((ntohl(ib_spec->ipv6.mask.flow_label)) >= BIT(20) || + (ntohl(ib_spec->ipv6.val.flow_label)) >= BIT(20)) + return -EINVAL; + break; + case IB_FLOW_SPEC_TCP: + case IB_FLOW_SPEC_UDP: + ib_filter_sz = offsetof(struct ib_flow_tcp_udp_filter, real_sz); + actual_filter_sz = spec_filter_size(kern_spec_mask, + kern_filter_sz, + ib_filter_sz); + if (actual_filter_sz <= 0) + return -EINVAL; + ib_spec->size = sizeof(struct ib_flow_spec_tcp_udp); + memcpy(&ib_spec->tcp_udp.val, kern_spec_val, actual_filter_sz); + memcpy(&ib_spec->tcp_udp.mask, kern_spec_mask, actual_filter_sz); + break; + case IB_FLOW_SPEC_VXLAN_TUNNEL: + ib_filter_sz = offsetof(struct ib_flow_tunnel_filter, real_sz); + actual_filter_sz = spec_filter_size(kern_spec_mask, + kern_filter_sz, + ib_filter_sz); + if (actual_filter_sz <= 0) + return -EINVAL; + ib_spec->tunnel.size = sizeof(struct ib_flow_spec_tunnel); + memcpy(&ib_spec->tunnel.val, kern_spec_val, actual_filter_sz); + memcpy(&ib_spec->tunnel.mask, kern_spec_mask, actual_filter_sz); + + if ((ntohl(ib_spec->tunnel.mask.tunnel_id)) >= BIT(24) || + (ntohl(ib_spec->tunnel.val.tunnel_id)) >= BIT(24)) + return -EINVAL; + break; + case IB_FLOW_SPEC_ESP: + ib_filter_sz = offsetof(struct ib_flow_esp_filter, real_sz); + actual_filter_sz = spec_filter_size(kern_spec_mask, + kern_filter_sz, + ib_filter_sz); + if (actual_filter_sz <= 0) + return -EINVAL; + ib_spec->esp.size = sizeof(struct ib_flow_spec_esp); + memcpy(&ib_spec->esp.val, kern_spec_val, actual_filter_sz); + memcpy(&ib_spec->esp.mask, kern_spec_mask, actual_filter_sz); + break; + case IB_FLOW_SPEC_GRE: + ib_filter_sz = offsetof(struct ib_flow_gre_filter, real_sz); + actual_filter_sz = spec_filter_size(kern_spec_mask, + kern_filter_sz, + ib_filter_sz); + if (actual_filter_sz <= 0) + return -EINVAL; + ib_spec->gre.size = sizeof(struct ib_flow_spec_gre); + memcpy(&ib_spec->gre.val, kern_spec_val, actual_filter_sz); + memcpy(&ib_spec->gre.mask, kern_spec_mask, actual_filter_sz); + break; + case IB_FLOW_SPEC_MPLS: + ib_filter_sz = offsetof(struct ib_flow_mpls_filter, real_sz); + actual_filter_sz = spec_filter_size(kern_spec_mask, + kern_filter_sz, + ib_filter_sz); + if (actual_filter_sz <= 0) + return -EINVAL; + ib_spec->mpls.size = sizeof(struct ib_flow_spec_mpls); + memcpy(&ib_spec->mpls.val, kern_spec_val, actual_filter_sz); + memcpy(&ib_spec->mpls.mask, kern_spec_mask, actual_filter_sz); + break; + default: + return -EINVAL; + } + return 0; +} + +static int kern_spec_to_ib_spec_filter(struct ib_uverbs_flow_spec *kern_spec, + union ib_flow_spec *ib_spec) +{ + size_t kern_filter_sz; + void *kern_spec_mask; + void *kern_spec_val; + + if (check_sub_overflow((size_t)kern_spec->hdr.size, + sizeof(struct ib_uverbs_flow_spec_hdr), + &kern_filter_sz)) + return -EINVAL; + + kern_filter_sz /= 2; + + kern_spec_val = (void *)kern_spec + + sizeof(struct ib_uverbs_flow_spec_hdr); + kern_spec_mask = kern_spec_val + kern_filter_sz; + + return ib_uverbs_kern_spec_to_ib_spec_filter(kern_spec->type, + kern_spec_mask, + kern_spec_val, + kern_filter_sz, ib_spec); +} + +static int kern_spec_to_ib_spec(struct uverbs_attr_bundle *attrs, + struct ib_uverbs_flow_spec *kern_spec, + union ib_flow_spec *ib_spec, + struct ib_uflow_resources *uflow_res) +{ + if (kern_spec->reserved) + return -EINVAL; + + if (kern_spec->type >= IB_FLOW_SPEC_ACTION_TAG) + return kern_spec_to_ib_spec_action(attrs, kern_spec, ib_spec, + uflow_res); + else + return kern_spec_to_ib_spec_filter(kern_spec, ib_spec); +} + +static int ib_uverbs_ex_create_wq(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_ex_create_wq cmd; + struct ib_uverbs_ex_create_wq_resp resp = {}; + struct ib_uwq_object *obj; + int err = 0; + struct ib_cq *cq; + struct ib_pd *pd; + struct ib_wq *wq; + struct ib_wq_init_attr wq_init_attr = {}; + struct ib_device *ib_dev; + + err = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (err) + return err; + + if (cmd.comp_mask) + return -EOPNOTSUPP; + + obj = (struct ib_uwq_object *)uobj_alloc(UVERBS_OBJECT_WQ, attrs, + &ib_dev); + if (IS_ERR(obj)) + return PTR_ERR(obj); + + pd = uobj_get_obj_read(pd, UVERBS_OBJECT_PD, cmd.pd_handle, attrs); + if (!pd) { + err = -EINVAL; + goto err_uobj; + } + + cq = uobj_get_obj_read(cq, UVERBS_OBJECT_CQ, cmd.cq_handle, attrs); + if (!cq) { + err = -EINVAL; + goto err_put_pd; + } + + wq_init_attr.cq = cq; + wq_init_attr.max_sge = cmd.max_sge; + wq_init_attr.max_wr = cmd.max_wr; + wq_init_attr.wq_type = cmd.wq_type; + wq_init_attr.event_handler = ib_uverbs_wq_event_handler; + wq_init_attr.create_flags = cmd.create_flags; + INIT_LIST_HEAD(&obj->uevent.event_list); + obj->uevent.uobject.user_handle = cmd.user_handle; + + wq = pd->device->ops.create_wq(pd, &wq_init_attr, &attrs->driver_udata); + if (IS_ERR(wq)) { + err = PTR_ERR(wq); + goto err_put_cq; + } + + wq->uobject = obj; + obj->uevent.uobject.object = wq; + wq->wq_type = wq_init_attr.wq_type; + wq->cq = cq; + wq->pd = pd; + wq->device = pd->device; + atomic_set(&wq->usecnt, 0); + atomic_inc(&pd->usecnt); + atomic_inc(&cq->usecnt); + obj->uevent.event_file = READ_ONCE(attrs->ufile->default_async_file); + if (obj->uevent.event_file) + uverbs_uobject_get(&obj->uevent.event_file->uobj); + + uobj_put_obj_read(pd); + rdma_lookup_put_uobject(&cq->uobject->uevent.uobject, + UVERBS_LOOKUP_READ); + uobj_finalize_uobj_create(&obj->uevent.uobject, attrs); + + resp.wq_handle = obj->uevent.uobject.id; + resp.max_sge = wq_init_attr.max_sge; + resp.max_wr = wq_init_attr.max_wr; + resp.wqn = wq->wq_num; + resp.response_length = uverbs_response_length(attrs, sizeof(resp)); + return uverbs_response(attrs, &resp, sizeof(resp)); + +err_put_cq: + rdma_lookup_put_uobject(&cq->uobject->uevent.uobject, + UVERBS_LOOKUP_READ); +err_put_pd: + uobj_put_obj_read(pd); +err_uobj: + uobj_alloc_abort(&obj->uevent.uobject, attrs); + + return err; +} + +static int ib_uverbs_ex_destroy_wq(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_ex_destroy_wq cmd; + struct ib_uverbs_ex_destroy_wq_resp resp = {}; + struct ib_uobject *uobj; + struct ib_uwq_object *obj; + int ret; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + if (cmd.comp_mask) + return -EOPNOTSUPP; + + resp.response_length = uverbs_response_length(attrs, sizeof(resp)); + uobj = uobj_get_destroy(UVERBS_OBJECT_WQ, cmd.wq_handle, attrs); + if (IS_ERR(uobj)) + return PTR_ERR(uobj); + + obj = container_of(uobj, struct ib_uwq_object, uevent.uobject); + resp.events_reported = obj->uevent.events_reported; + + uobj_put_destroy(uobj); + + return uverbs_response(attrs, &resp, sizeof(resp)); +} + +static int ib_uverbs_ex_modify_wq(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_ex_modify_wq cmd; + struct ib_wq *wq; + struct ib_wq_attr wq_attr = {}; + int ret; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + if (!cmd.attr_mask) + return -EINVAL; + + if (cmd.attr_mask > (IB_WQ_STATE | IB_WQ_CUR_STATE | IB_WQ_FLAGS)) + return -EINVAL; + + wq = uobj_get_obj_read(wq, UVERBS_OBJECT_WQ, cmd.wq_handle, attrs); + if (!wq) + return -EINVAL; + + if (cmd.attr_mask & IB_WQ_FLAGS) { + wq_attr.flags = cmd.flags; + wq_attr.flags_mask = cmd.flags_mask; + } + + if (cmd.attr_mask & IB_WQ_CUR_STATE) { + if (cmd.curr_wq_state > IB_WQS_ERR) + return -EINVAL; + + wq_attr.curr_wq_state = cmd.curr_wq_state; + } else { + wq_attr.curr_wq_state = wq->state; + } + + if (cmd.attr_mask & IB_WQ_STATE) { + if (cmd.wq_state > IB_WQS_ERR) + return -EINVAL; + + wq_attr.wq_state = cmd.wq_state; + } else { + wq_attr.wq_state = wq_attr.curr_wq_state; + } + + ret = wq->device->ops.modify_wq(wq, &wq_attr, cmd.attr_mask, + &attrs->driver_udata); + rdma_lookup_put_uobject(&wq->uobject->uevent.uobject, + UVERBS_LOOKUP_READ); + return ret; +} + +static int ib_uverbs_ex_create_rwq_ind_table(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_ex_create_rwq_ind_table cmd; + struct ib_uverbs_ex_create_rwq_ind_table_resp resp = {}; + struct ib_uobject *uobj; + int err; + struct ib_rwq_ind_table_init_attr init_attr = {}; + struct ib_rwq_ind_table *rwq_ind_tbl; + struct ib_wq **wqs = NULL; + u32 *wqs_handles = NULL; + struct ib_wq *wq = NULL; + int i, num_read_wqs; + u32 num_wq_handles; + struct uverbs_req_iter iter; + struct ib_device *ib_dev; + + err = uverbs_request_start(attrs, &iter, &cmd, sizeof(cmd)); + if (err) + return err; + + if (cmd.comp_mask) + return -EOPNOTSUPP; + + if (cmd.log_ind_tbl_size > IB_USER_VERBS_MAX_LOG_IND_TBL_SIZE) + return -EINVAL; + + num_wq_handles = 1 << cmd.log_ind_tbl_size; + wqs_handles = kcalloc(num_wq_handles, sizeof(*wqs_handles), + GFP_KERNEL); + if (!wqs_handles) + return -ENOMEM; + + err = uverbs_request_next(&iter, wqs_handles, + num_wq_handles * sizeof(__u32)); + if (err) + goto err_free; + + err = uverbs_request_finish(&iter); + if (err) + goto err_free; + + wqs = kcalloc(num_wq_handles, sizeof(*wqs), GFP_KERNEL); + if (!wqs) { + err = -ENOMEM; + goto err_free; + } + + for (num_read_wqs = 0; num_read_wqs < num_wq_handles; + num_read_wqs++) { + wq = uobj_get_obj_read(wq, UVERBS_OBJECT_WQ, + wqs_handles[num_read_wqs], attrs); + if (!wq) { + err = -EINVAL; + goto put_wqs; + } + + wqs[num_read_wqs] = wq; + atomic_inc(&wqs[num_read_wqs]->usecnt); + } + + uobj = uobj_alloc(UVERBS_OBJECT_RWQ_IND_TBL, attrs, &ib_dev); + if (IS_ERR(uobj)) { + err = PTR_ERR(uobj); + goto put_wqs; + } + + rwq_ind_tbl = rdma_zalloc_drv_obj(ib_dev, ib_rwq_ind_table); + if (!rwq_ind_tbl) { + err = -ENOMEM; + goto err_uobj; + } + + init_attr.log_ind_tbl_size = cmd.log_ind_tbl_size; + init_attr.ind_tbl = wqs; + + rwq_ind_tbl->ind_tbl = wqs; + rwq_ind_tbl->log_ind_tbl_size = init_attr.log_ind_tbl_size; + rwq_ind_tbl->uobject = uobj; + uobj->object = rwq_ind_tbl; + rwq_ind_tbl->device = ib_dev; + atomic_set(&rwq_ind_tbl->usecnt, 0); + + err = ib_dev->ops.create_rwq_ind_table(rwq_ind_tbl, &init_attr, + &attrs->driver_udata); + if (err) + goto err_create; + + for (i = 0; i < num_wq_handles; i++) + rdma_lookup_put_uobject(&wqs[i]->uobject->uevent.uobject, + UVERBS_LOOKUP_READ); + kfree(wqs_handles); + uobj_finalize_uobj_create(uobj, attrs); + + resp.ind_tbl_handle = uobj->id; + resp.ind_tbl_num = rwq_ind_tbl->ind_tbl_num; + resp.response_length = uverbs_response_length(attrs, sizeof(resp)); + return uverbs_response(attrs, &resp, sizeof(resp)); + +err_create: + kfree(rwq_ind_tbl); +err_uobj: + uobj_alloc_abort(uobj, attrs); +put_wqs: + for (i = 0; i < num_read_wqs; i++) { + rdma_lookup_put_uobject(&wqs[i]->uobject->uevent.uobject, + UVERBS_LOOKUP_READ); + atomic_dec(&wqs[i]->usecnt); + } +err_free: + kfree(wqs_handles); + kfree(wqs); + return err; +} + +static int ib_uverbs_ex_destroy_rwq_ind_table(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_ex_destroy_rwq_ind_table cmd; + int ret; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + if (cmd.comp_mask) + return -EOPNOTSUPP; + + return uobj_perform_destroy(UVERBS_OBJECT_RWQ_IND_TBL, + cmd.ind_tbl_handle, attrs); +} + +static int ib_uverbs_ex_create_flow(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_create_flow cmd; + struct ib_uverbs_create_flow_resp resp = {}; + struct ib_uobject *uobj; + struct ib_flow *flow_id; + struct ib_uverbs_flow_attr *kern_flow_attr; + struct ib_flow_attr *flow_attr; + struct ib_qp *qp; + struct ib_uflow_resources *uflow_res; + struct ib_uverbs_flow_spec_hdr *kern_spec; + struct uverbs_req_iter iter; + int err; + void *ib_spec; + int i; + struct ib_device *ib_dev; + + err = uverbs_request_start(attrs, &iter, &cmd, sizeof(cmd)); + if (err) + return err; + + if (cmd.comp_mask) + return -EINVAL; + + if (!capable(CAP_NET_RAW)) + return -EPERM; + + if (cmd.flow_attr.flags >= IB_FLOW_ATTR_FLAGS_RESERVED) + return -EINVAL; + + if ((cmd.flow_attr.flags & IB_FLOW_ATTR_FLAGS_DONT_TRAP) && + ((cmd.flow_attr.type == IB_FLOW_ATTR_ALL_DEFAULT) || + (cmd.flow_attr.type == IB_FLOW_ATTR_MC_DEFAULT))) + return -EINVAL; + + if (cmd.flow_attr.num_of_specs > IB_FLOW_SPEC_SUPPORT_LAYERS) + return -EINVAL; + + if (cmd.flow_attr.size > + (cmd.flow_attr.num_of_specs * sizeof(struct ib_uverbs_flow_spec))) + return -EINVAL; + + if (cmd.flow_attr.reserved[0] || + cmd.flow_attr.reserved[1]) + return -EINVAL; + + if (cmd.flow_attr.num_of_specs) { + kern_flow_attr = kmalloc(sizeof(*kern_flow_attr) + cmd.flow_attr.size, + GFP_KERNEL); + if (!kern_flow_attr) + return -ENOMEM; + + *kern_flow_attr = cmd.flow_attr; + err = uverbs_request_next(&iter, &kern_flow_attr->flow_specs, + cmd.flow_attr.size); + if (err) + goto err_free_attr; + } else { + kern_flow_attr = &cmd.flow_attr; + } + + err = uverbs_request_finish(&iter); + if (err) + goto err_free_attr; + + uobj = uobj_alloc(UVERBS_OBJECT_FLOW, attrs, &ib_dev); + if (IS_ERR(uobj)) { + err = PTR_ERR(uobj); + goto err_free_attr; + } + + qp = uobj_get_obj_read(qp, UVERBS_OBJECT_QP, cmd.qp_handle, attrs); + if (!qp) { + err = -EINVAL; + goto err_uobj; + } + + if (qp->qp_type != IB_QPT_UD && qp->qp_type != IB_QPT_RAW_PACKET) { + err = -EINVAL; + goto err_put; + } + + flow_attr = kzalloc(struct_size(flow_attr, flows, + cmd.flow_attr.num_of_specs), GFP_KERNEL); + if (!flow_attr) { + err = -ENOMEM; + goto err_put; + } + uflow_res = flow_resources_alloc(cmd.flow_attr.num_of_specs); + if (!uflow_res) { + err = -ENOMEM; + goto err_free_flow_attr; + } + + flow_attr->type = kern_flow_attr->type; + flow_attr->priority = kern_flow_attr->priority; + flow_attr->num_of_specs = kern_flow_attr->num_of_specs; + flow_attr->port = kern_flow_attr->port; + flow_attr->flags = kern_flow_attr->flags; + flow_attr->size = sizeof(*flow_attr); + + kern_spec = kern_flow_attr->flow_specs; + ib_spec = flow_attr + 1; + for (i = 0; i < flow_attr->num_of_specs && + cmd.flow_attr.size >= sizeof(*kern_spec) && + cmd.flow_attr.size >= kern_spec->size; + i++) { + err = kern_spec_to_ib_spec( + attrs, (struct ib_uverbs_flow_spec *)kern_spec, + ib_spec, uflow_res); + if (err) + goto err_free; + + flow_attr->size += + ((union ib_flow_spec *) ib_spec)->size; + cmd.flow_attr.size -= kern_spec->size; + kern_spec = ((void *)kern_spec) + kern_spec->size; + ib_spec += ((union ib_flow_spec *) ib_spec)->size; + } + if (cmd.flow_attr.size || (i != flow_attr->num_of_specs)) { + pr_warn("create flow failed, flow %d: %d bytes left from uverb cmd\n", + i, cmd.flow_attr.size); + err = -EINVAL; + goto err_free; + } + + flow_id = qp->device->ops.create_flow(qp, flow_attr, + &attrs->driver_udata); + + if (IS_ERR(flow_id)) { + err = PTR_ERR(flow_id); + goto err_free; + } + + ib_set_flow(uobj, flow_id, qp, qp->device, uflow_res); + + rdma_lookup_put_uobject(&qp->uobject->uevent.uobject, + UVERBS_LOOKUP_READ); + kfree(flow_attr); + + if (cmd.flow_attr.num_of_specs) + kfree(kern_flow_attr); + uobj_finalize_uobj_create(uobj, attrs); + + resp.flow_handle = uobj->id; + return uverbs_response(attrs, &resp, sizeof(resp)); + +err_free: + ib_uverbs_flow_resources_free(uflow_res); +err_free_flow_attr: + kfree(flow_attr); +err_put: + rdma_lookup_put_uobject(&qp->uobject->uevent.uobject, + UVERBS_LOOKUP_READ); +err_uobj: + uobj_alloc_abort(uobj, attrs); +err_free_attr: + if (cmd.flow_attr.num_of_specs) + kfree(kern_flow_attr); + return err; +} + +static int ib_uverbs_ex_destroy_flow(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_destroy_flow cmd; + int ret; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + if (cmd.comp_mask) + return -EINVAL; + + return uobj_perform_destroy(UVERBS_OBJECT_FLOW, cmd.flow_handle, attrs); +} + +static int __uverbs_create_xsrq(struct uverbs_attr_bundle *attrs, + struct ib_uverbs_create_xsrq *cmd, + struct ib_udata *udata) +{ + struct ib_uverbs_create_srq_resp resp = {}; + struct ib_usrq_object *obj; + struct ib_pd *pd; + struct ib_srq *srq; + struct ib_srq_init_attr attr; + int ret; + struct ib_uobject *xrcd_uobj; + struct ib_device *ib_dev; + + obj = (struct ib_usrq_object *)uobj_alloc(UVERBS_OBJECT_SRQ, attrs, + &ib_dev); + if (IS_ERR(obj)) + return PTR_ERR(obj); + + if (cmd->srq_type == IB_SRQT_TM) + attr.ext.tag_matching.max_num_tags = cmd->max_num_tags; + + if (cmd->srq_type == IB_SRQT_XRC) { + xrcd_uobj = uobj_get_read(UVERBS_OBJECT_XRCD, cmd->xrcd_handle, + attrs); + if (IS_ERR(xrcd_uobj)) { + ret = -EINVAL; + goto err; + } + + attr.ext.xrc.xrcd = (struct ib_xrcd *)xrcd_uobj->object; + if (!attr.ext.xrc.xrcd) { + ret = -EINVAL; + goto err_put_xrcd; + } + + obj->uxrcd = container_of(xrcd_uobj, struct ib_uxrcd_object, uobject); + atomic_inc(&obj->uxrcd->refcnt); + } + + if (ib_srq_has_cq(cmd->srq_type)) { + attr.ext.cq = uobj_get_obj_read(cq, UVERBS_OBJECT_CQ, + cmd->cq_handle, attrs); + if (!attr.ext.cq) { + ret = -EINVAL; + goto err_put_xrcd; + } + } + + pd = uobj_get_obj_read(pd, UVERBS_OBJECT_PD, cmd->pd_handle, attrs); + if (!pd) { + ret = -EINVAL; + goto err_put_cq; + } + + attr.event_handler = ib_uverbs_srq_event_handler; + attr.srq_type = cmd->srq_type; + attr.attr.max_wr = cmd->max_wr; + attr.attr.max_sge = cmd->max_sge; + attr.attr.srq_limit = cmd->srq_limit; + + INIT_LIST_HEAD(&obj->uevent.event_list); + obj->uevent.uobject.user_handle = cmd->user_handle; + + srq = ib_create_srq_user(pd, &attr, obj, udata); + if (IS_ERR(srq)) { + ret = PTR_ERR(srq); + goto err_put_pd; + } + + obj->uevent.uobject.object = srq; + obj->uevent.uobject.user_handle = cmd->user_handle; + obj->uevent.event_file = READ_ONCE(attrs->ufile->default_async_file); + if (obj->uevent.event_file) + uverbs_uobject_get(&obj->uevent.event_file->uobj); + + if (cmd->srq_type == IB_SRQT_XRC) + resp.srqn = srq->ext.xrc.srq_num; + + if (cmd->srq_type == IB_SRQT_XRC) + uobj_put_read(xrcd_uobj); + + if (ib_srq_has_cq(cmd->srq_type)) + rdma_lookup_put_uobject(&attr.ext.cq->uobject->uevent.uobject, + UVERBS_LOOKUP_READ); + + uobj_put_obj_read(pd); + uobj_finalize_uobj_create(&obj->uevent.uobject, attrs); + + resp.srq_handle = obj->uevent.uobject.id; + resp.max_wr = attr.attr.max_wr; + resp.max_sge = attr.attr.max_sge; + return uverbs_response(attrs, &resp, sizeof(resp)); + +err_put_pd: + uobj_put_obj_read(pd); +err_put_cq: + if (ib_srq_has_cq(cmd->srq_type)) + rdma_lookup_put_uobject(&attr.ext.cq->uobject->uevent.uobject, + UVERBS_LOOKUP_READ); + +err_put_xrcd: + if (cmd->srq_type == IB_SRQT_XRC) { + atomic_dec(&obj->uxrcd->refcnt); + uobj_put_read(xrcd_uobj); + } + +err: + uobj_alloc_abort(&obj->uevent.uobject, attrs); + return ret; +} + +static int ib_uverbs_create_srq(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_create_srq cmd; + struct ib_uverbs_create_xsrq xcmd; + int ret; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + memset(&xcmd, 0, sizeof(xcmd)); + xcmd.response = cmd.response; + xcmd.user_handle = cmd.user_handle; + xcmd.srq_type = IB_SRQT_BASIC; + xcmd.pd_handle = cmd.pd_handle; + xcmd.max_wr = cmd.max_wr; + xcmd.max_sge = cmd.max_sge; + xcmd.srq_limit = cmd.srq_limit; + + return __uverbs_create_xsrq(attrs, &xcmd, &attrs->driver_udata); +} + +static int ib_uverbs_create_xsrq(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_create_xsrq cmd; + int ret; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + return __uverbs_create_xsrq(attrs, &cmd, &attrs->driver_udata); +} + +static int ib_uverbs_modify_srq(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_modify_srq cmd; + struct ib_srq *srq; + struct ib_srq_attr attr; + int ret; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + srq = uobj_get_obj_read(srq, UVERBS_OBJECT_SRQ, cmd.srq_handle, attrs); + if (!srq) + return -EINVAL; + + attr.max_wr = cmd.max_wr; + attr.srq_limit = cmd.srq_limit; + + ret = srq->device->ops.modify_srq(srq, &attr, cmd.attr_mask, + &attrs->driver_udata); + + rdma_lookup_put_uobject(&srq->uobject->uevent.uobject, + UVERBS_LOOKUP_READ); + + return ret; +} + +static int ib_uverbs_query_srq(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_query_srq cmd; + struct ib_uverbs_query_srq_resp resp; + struct ib_srq_attr attr; + struct ib_srq *srq; + int ret; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + srq = uobj_get_obj_read(srq, UVERBS_OBJECT_SRQ, cmd.srq_handle, attrs); + if (!srq) + return -EINVAL; + + ret = ib_query_srq(srq, &attr); + + rdma_lookup_put_uobject(&srq->uobject->uevent.uobject, + UVERBS_LOOKUP_READ); + + if (ret) + return ret; + + memset(&resp, 0, sizeof resp); + + resp.max_wr = attr.max_wr; + resp.max_sge = attr.max_sge; + resp.srq_limit = attr.srq_limit; + + return uverbs_response(attrs, &resp, sizeof(resp)); +} + +static int ib_uverbs_destroy_srq(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_destroy_srq cmd; + struct ib_uverbs_destroy_srq_resp resp; + struct ib_uobject *uobj; + struct ib_uevent_object *obj; + int ret; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + uobj = uobj_get_destroy(UVERBS_OBJECT_SRQ, cmd.srq_handle, attrs); + if (IS_ERR(uobj)) + return PTR_ERR(uobj); + + obj = container_of(uobj, struct ib_uevent_object, uobject); + memset(&resp, 0, sizeof(resp)); + resp.events_reported = obj->events_reported; + + uobj_put_destroy(uobj); + + return uverbs_response(attrs, &resp, sizeof(resp)); +} + +static int ib_uverbs_ex_query_device(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_ex_query_device_resp resp = {}; + struct ib_uverbs_ex_query_device cmd; + struct ib_device_attr attr = {0}; + struct ib_ucontext *ucontext; + struct ib_device *ib_dev; + int err; + + ucontext = ib_uverbs_get_ucontext(attrs); + if (IS_ERR(ucontext)) + return PTR_ERR(ucontext); + ib_dev = ucontext->device; + + err = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (err) + return err; + + if (cmd.comp_mask) + return -EINVAL; + + if (cmd.reserved) + return -EINVAL; + + err = ib_dev->ops.query_device(ib_dev, &attr, &attrs->driver_udata); + if (err) + return err; + + copy_query_dev_fields(ucontext, &resp.base, &attr); + + resp.odp_caps.general_caps = attr.odp_caps.general_caps; + resp.odp_caps.per_transport_caps.rc_odp_caps = + attr.odp_caps.per_transport_caps.rc_odp_caps; + resp.odp_caps.per_transport_caps.uc_odp_caps = + attr.odp_caps.per_transport_caps.uc_odp_caps; + resp.odp_caps.per_transport_caps.ud_odp_caps = + attr.odp_caps.per_transport_caps.ud_odp_caps; + resp.xrc_odp_caps = attr.odp_caps.per_transport_caps.xrc_odp_caps; + + resp.timestamp_mask = attr.timestamp_mask; + resp.hca_core_clock = attr.hca_core_clock; + resp.device_cap_flags_ex = attr.device_cap_flags; + resp.rss_caps.supported_qpts = attr.rss_caps.supported_qpts; + resp.rss_caps.max_rwq_indirection_tables = + attr.rss_caps.max_rwq_indirection_tables; + resp.rss_caps.max_rwq_indirection_table_size = + attr.rss_caps.max_rwq_indirection_table_size; + resp.max_wq_type_rq = attr.max_wq_type_rq; + resp.raw_packet_caps = attr.raw_packet_caps; + resp.tm_caps.max_rndv_hdr_size = attr.tm_caps.max_rndv_hdr_size; + resp.tm_caps.max_num_tags = attr.tm_caps.max_num_tags; + resp.tm_caps.max_ops = attr.tm_caps.max_ops; + resp.tm_caps.max_sge = attr.tm_caps.max_sge; + resp.tm_caps.flags = attr.tm_caps.flags; + resp.cq_moderation_caps.max_cq_moderation_count = + attr.cq_caps.max_cq_moderation_count; + resp.cq_moderation_caps.max_cq_moderation_period = + attr.cq_caps.max_cq_moderation_period; + resp.max_dm_size = attr.max_dm_size; + resp.response_length = uverbs_response_length(attrs, sizeof(resp)); + + return uverbs_response(attrs, &resp, sizeof(resp)); +} + +static int ib_uverbs_ex_modify_cq(struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_ex_modify_cq cmd; + struct ib_cq *cq; + int ret; + + ret = uverbs_request(attrs, &cmd, sizeof(cmd)); + if (ret) + return ret; + + if (!cmd.attr_mask || cmd.reserved) + return -EINVAL; + + if (cmd.attr_mask > IB_CQ_MODERATE) + return -EOPNOTSUPP; + + cq = uobj_get_obj_read(cq, UVERBS_OBJECT_CQ, cmd.cq_handle, attrs); + if (!cq) + return -EINVAL; + + ret = rdma_set_cq_moderation(cq, cmd.attr.cq_count, cmd.attr.cq_period); + + rdma_lookup_put_uobject(&cq->uobject->uevent.uobject, + UVERBS_LOOKUP_READ); + return ret; +} + +/* + * Describe the input structs for write(). Some write methods have an input + * only struct, most have an input and output. If the struct has an output then + * the 'response' u64 must be the first field in the request structure. + * + * If udata is present then both the request and response structs have a + * trailing driver_data flex array. In this case the size of the base struct + * cannot be changed. + */ +#define UAPI_DEF_WRITE_IO(req, resp) \ + .write.has_resp = 1 + \ + BUILD_BUG_ON_ZERO(offsetof(req, response) != 0) + \ + BUILD_BUG_ON_ZERO(sizeof_field(req, response) != \ + sizeof(u64)), \ + .write.req_size = sizeof(req), .write.resp_size = sizeof(resp) + +#define UAPI_DEF_WRITE_I(req) .write.req_size = sizeof(req) + +#define UAPI_DEF_WRITE_UDATA_IO(req, resp) \ + UAPI_DEF_WRITE_IO(req, resp), \ + .write.has_udata = \ + 1 + \ + BUILD_BUG_ON_ZERO(offsetof(req, driver_data) != \ + sizeof(req)) + \ + BUILD_BUG_ON_ZERO(offsetof(resp, driver_data) != \ + sizeof(resp)) + +#define UAPI_DEF_WRITE_UDATA_I(req) \ + UAPI_DEF_WRITE_I(req), \ + .write.has_udata = \ + 1 + BUILD_BUG_ON_ZERO(offsetof(req, driver_data) != \ + sizeof(req)) + +/* + * The _EX versions are for use with WRITE_EX and allow the last struct member + * to be specified. Buffers that do not include that member will be rejected. + */ +#define UAPI_DEF_WRITE_IO_EX(req, req_last_member, resp, resp_last_member) \ + .write.has_resp = 1, \ + .write.req_size = offsetofend(req, req_last_member), \ + .write.resp_size = offsetofend(resp, resp_last_member) + +#define UAPI_DEF_WRITE_I_EX(req, req_last_member) \ + .write.req_size = offsetofend(req, req_last_member) + +const struct uapi_definition uverbs_def_write_intf[] = { + DECLARE_UVERBS_OBJECT( + UVERBS_OBJECT_AH, + DECLARE_UVERBS_WRITE(IB_USER_VERBS_CMD_CREATE_AH, + ib_uverbs_create_ah, + UAPI_DEF_WRITE_UDATA_IO( + struct ib_uverbs_create_ah, + struct ib_uverbs_create_ah_resp), + UAPI_DEF_METHOD_NEEDS_FN(create_ah)), + DECLARE_UVERBS_WRITE( + IB_USER_VERBS_CMD_DESTROY_AH, + ib_uverbs_destroy_ah, + UAPI_DEF_WRITE_I(struct ib_uverbs_destroy_ah), + UAPI_DEF_METHOD_NEEDS_FN(destroy_ah))), + + DECLARE_UVERBS_OBJECT( + UVERBS_OBJECT_COMP_CHANNEL, + DECLARE_UVERBS_WRITE( + IB_USER_VERBS_CMD_CREATE_COMP_CHANNEL, + ib_uverbs_create_comp_channel, + UAPI_DEF_WRITE_IO( + struct ib_uverbs_create_comp_channel, + struct ib_uverbs_create_comp_channel_resp))), + + DECLARE_UVERBS_OBJECT( + UVERBS_OBJECT_CQ, + DECLARE_UVERBS_WRITE(IB_USER_VERBS_CMD_CREATE_CQ, + ib_uverbs_create_cq, + UAPI_DEF_WRITE_UDATA_IO( + struct ib_uverbs_create_cq, + struct ib_uverbs_create_cq_resp), + UAPI_DEF_METHOD_NEEDS_FN(create_cq)), + DECLARE_UVERBS_WRITE( + IB_USER_VERBS_CMD_DESTROY_CQ, + ib_uverbs_destroy_cq, + UAPI_DEF_WRITE_IO(struct ib_uverbs_destroy_cq, + struct ib_uverbs_destroy_cq_resp), + UAPI_DEF_METHOD_NEEDS_FN(destroy_cq)), + DECLARE_UVERBS_WRITE( + IB_USER_VERBS_CMD_POLL_CQ, + ib_uverbs_poll_cq, + UAPI_DEF_WRITE_IO(struct ib_uverbs_poll_cq, + struct ib_uverbs_poll_cq_resp), + UAPI_DEF_METHOD_NEEDS_FN(poll_cq)), + DECLARE_UVERBS_WRITE( + IB_USER_VERBS_CMD_REQ_NOTIFY_CQ, + ib_uverbs_req_notify_cq, + UAPI_DEF_WRITE_I(struct ib_uverbs_req_notify_cq), + UAPI_DEF_METHOD_NEEDS_FN(req_notify_cq)), + DECLARE_UVERBS_WRITE(IB_USER_VERBS_CMD_RESIZE_CQ, + ib_uverbs_resize_cq, + UAPI_DEF_WRITE_UDATA_IO( + struct ib_uverbs_resize_cq, + struct ib_uverbs_resize_cq_resp), + UAPI_DEF_METHOD_NEEDS_FN(resize_cq)), + DECLARE_UVERBS_WRITE_EX( + IB_USER_VERBS_EX_CMD_CREATE_CQ, + ib_uverbs_ex_create_cq, + UAPI_DEF_WRITE_IO_EX(struct ib_uverbs_ex_create_cq, + reserved, + struct ib_uverbs_ex_create_cq_resp, + response_length), + UAPI_DEF_METHOD_NEEDS_FN(create_cq)), + DECLARE_UVERBS_WRITE_EX( + IB_USER_VERBS_EX_CMD_MODIFY_CQ, + ib_uverbs_ex_modify_cq, + UAPI_DEF_WRITE_I(struct ib_uverbs_ex_modify_cq), + UAPI_DEF_METHOD_NEEDS_FN(modify_cq))), + + DECLARE_UVERBS_OBJECT( + UVERBS_OBJECT_DEVICE, + DECLARE_UVERBS_WRITE(IB_USER_VERBS_CMD_GET_CONTEXT, + ib_uverbs_get_context, + UAPI_DEF_WRITE_UDATA_IO( + struct ib_uverbs_get_context, + struct ib_uverbs_get_context_resp)), + DECLARE_UVERBS_WRITE( + IB_USER_VERBS_CMD_QUERY_DEVICE, + ib_uverbs_query_device, + UAPI_DEF_WRITE_IO(struct ib_uverbs_query_device, + struct ib_uverbs_query_device_resp)), + DECLARE_UVERBS_WRITE( + IB_USER_VERBS_CMD_QUERY_PORT, + ib_uverbs_query_port, + UAPI_DEF_WRITE_IO(struct ib_uverbs_query_port, + struct ib_uverbs_query_port_resp), + UAPI_DEF_METHOD_NEEDS_FN(query_port)), + DECLARE_UVERBS_WRITE_EX( + IB_USER_VERBS_EX_CMD_QUERY_DEVICE, + ib_uverbs_ex_query_device, + UAPI_DEF_WRITE_IO_EX( + struct ib_uverbs_ex_query_device, + reserved, + struct ib_uverbs_ex_query_device_resp, + response_length), + UAPI_DEF_METHOD_NEEDS_FN(query_device)), + UAPI_DEF_OBJ_NEEDS_FN(alloc_ucontext), + UAPI_DEF_OBJ_NEEDS_FN(dealloc_ucontext)), + + DECLARE_UVERBS_OBJECT( + UVERBS_OBJECT_FLOW, + DECLARE_UVERBS_WRITE_EX( + IB_USER_VERBS_EX_CMD_CREATE_FLOW, + ib_uverbs_ex_create_flow, + UAPI_DEF_WRITE_IO_EX(struct ib_uverbs_create_flow, + flow_attr, + struct ib_uverbs_create_flow_resp, + flow_handle), + UAPI_DEF_METHOD_NEEDS_FN(create_flow)), + DECLARE_UVERBS_WRITE_EX( + IB_USER_VERBS_EX_CMD_DESTROY_FLOW, + ib_uverbs_ex_destroy_flow, + UAPI_DEF_WRITE_I(struct ib_uverbs_destroy_flow), + UAPI_DEF_METHOD_NEEDS_FN(destroy_flow))), + + DECLARE_UVERBS_OBJECT( + UVERBS_OBJECT_MR, + DECLARE_UVERBS_WRITE(IB_USER_VERBS_CMD_DEREG_MR, + ib_uverbs_dereg_mr, + UAPI_DEF_WRITE_I(struct ib_uverbs_dereg_mr), + UAPI_DEF_METHOD_NEEDS_FN(dereg_mr)), + DECLARE_UVERBS_WRITE( + IB_USER_VERBS_CMD_REG_MR, + ib_uverbs_reg_mr, + UAPI_DEF_WRITE_UDATA_IO(struct ib_uverbs_reg_mr, + struct ib_uverbs_reg_mr_resp), + UAPI_DEF_METHOD_NEEDS_FN(reg_user_mr)), + DECLARE_UVERBS_WRITE( + IB_USER_VERBS_CMD_REREG_MR, + ib_uverbs_rereg_mr, + UAPI_DEF_WRITE_UDATA_IO(struct ib_uverbs_rereg_mr, + struct ib_uverbs_rereg_mr_resp), + UAPI_DEF_METHOD_NEEDS_FN(rereg_user_mr))), + + DECLARE_UVERBS_OBJECT( + UVERBS_OBJECT_MW, + DECLARE_UVERBS_WRITE( + IB_USER_VERBS_CMD_ALLOC_MW, + ib_uverbs_alloc_mw, + UAPI_DEF_WRITE_UDATA_IO(struct ib_uverbs_alloc_mw, + struct ib_uverbs_alloc_mw_resp), + UAPI_DEF_METHOD_NEEDS_FN(alloc_mw)), + DECLARE_UVERBS_WRITE( + IB_USER_VERBS_CMD_DEALLOC_MW, + ib_uverbs_dealloc_mw, + UAPI_DEF_WRITE_I(struct ib_uverbs_dealloc_mw), + UAPI_DEF_METHOD_NEEDS_FN(dealloc_mw))), + + DECLARE_UVERBS_OBJECT( + UVERBS_OBJECT_PD, + DECLARE_UVERBS_WRITE( + IB_USER_VERBS_CMD_ALLOC_PD, + ib_uverbs_alloc_pd, + UAPI_DEF_WRITE_UDATA_IO(struct ib_uverbs_alloc_pd, + struct ib_uverbs_alloc_pd_resp), + UAPI_DEF_METHOD_NEEDS_FN(alloc_pd)), + DECLARE_UVERBS_WRITE( + IB_USER_VERBS_CMD_DEALLOC_PD, + ib_uverbs_dealloc_pd, + UAPI_DEF_WRITE_I(struct ib_uverbs_dealloc_pd), + UAPI_DEF_METHOD_NEEDS_FN(dealloc_pd))), + + DECLARE_UVERBS_OBJECT( + UVERBS_OBJECT_QP, + DECLARE_UVERBS_WRITE( + IB_USER_VERBS_CMD_ATTACH_MCAST, + ib_uverbs_attach_mcast, + UAPI_DEF_WRITE_I(struct ib_uverbs_attach_mcast), + UAPI_DEF_METHOD_NEEDS_FN(attach_mcast), + UAPI_DEF_METHOD_NEEDS_FN(detach_mcast)), + DECLARE_UVERBS_WRITE(IB_USER_VERBS_CMD_CREATE_QP, + ib_uverbs_create_qp, + UAPI_DEF_WRITE_UDATA_IO( + struct ib_uverbs_create_qp, + struct ib_uverbs_create_qp_resp), + UAPI_DEF_METHOD_NEEDS_FN(create_qp)), + DECLARE_UVERBS_WRITE( + IB_USER_VERBS_CMD_DESTROY_QP, + ib_uverbs_destroy_qp, + UAPI_DEF_WRITE_IO(struct ib_uverbs_destroy_qp, + struct ib_uverbs_destroy_qp_resp), + UAPI_DEF_METHOD_NEEDS_FN(destroy_qp)), + DECLARE_UVERBS_WRITE( + IB_USER_VERBS_CMD_DETACH_MCAST, + ib_uverbs_detach_mcast, + UAPI_DEF_WRITE_I(struct ib_uverbs_detach_mcast), + UAPI_DEF_METHOD_NEEDS_FN(detach_mcast)), + DECLARE_UVERBS_WRITE( + IB_USER_VERBS_CMD_MODIFY_QP, + ib_uverbs_modify_qp, + UAPI_DEF_WRITE_I(struct ib_uverbs_modify_qp), + UAPI_DEF_METHOD_NEEDS_FN(modify_qp)), + DECLARE_UVERBS_WRITE( + IB_USER_VERBS_CMD_POST_RECV, + ib_uverbs_post_recv, + UAPI_DEF_WRITE_IO(struct ib_uverbs_post_recv, + struct ib_uverbs_post_recv_resp), + UAPI_DEF_METHOD_NEEDS_FN(post_recv)), + DECLARE_UVERBS_WRITE( + IB_USER_VERBS_CMD_POST_SEND, + ib_uverbs_post_send, + UAPI_DEF_WRITE_IO(struct ib_uverbs_post_send, + struct ib_uverbs_post_send_resp), + UAPI_DEF_METHOD_NEEDS_FN(post_send)), + DECLARE_UVERBS_WRITE( + IB_USER_VERBS_CMD_QUERY_QP, + ib_uverbs_query_qp, + UAPI_DEF_WRITE_IO(struct ib_uverbs_query_qp, + struct ib_uverbs_query_qp_resp), + UAPI_DEF_METHOD_NEEDS_FN(query_qp)), + DECLARE_UVERBS_WRITE_EX( + IB_USER_VERBS_EX_CMD_CREATE_QP, + ib_uverbs_ex_create_qp, + UAPI_DEF_WRITE_IO_EX(struct ib_uverbs_ex_create_qp, + comp_mask, + struct ib_uverbs_ex_create_qp_resp, + response_length), + UAPI_DEF_METHOD_NEEDS_FN(create_qp)), + DECLARE_UVERBS_WRITE_EX( + IB_USER_VERBS_EX_CMD_MODIFY_QP, + ib_uverbs_ex_modify_qp, + UAPI_DEF_WRITE_IO_EX(struct ib_uverbs_ex_modify_qp, + base, + struct ib_uverbs_ex_modify_qp_resp, + response_length), + UAPI_DEF_METHOD_NEEDS_FN(modify_qp))), + + DECLARE_UVERBS_OBJECT( + UVERBS_OBJECT_RWQ_IND_TBL, + DECLARE_UVERBS_WRITE_EX( + IB_USER_VERBS_EX_CMD_CREATE_RWQ_IND_TBL, + ib_uverbs_ex_create_rwq_ind_table, + UAPI_DEF_WRITE_IO_EX( + struct ib_uverbs_ex_create_rwq_ind_table, + log_ind_tbl_size, + struct ib_uverbs_ex_create_rwq_ind_table_resp, + ind_tbl_num), + UAPI_DEF_METHOD_NEEDS_FN(create_rwq_ind_table)), + DECLARE_UVERBS_WRITE_EX( + IB_USER_VERBS_EX_CMD_DESTROY_RWQ_IND_TBL, + ib_uverbs_ex_destroy_rwq_ind_table, + UAPI_DEF_WRITE_I( + struct ib_uverbs_ex_destroy_rwq_ind_table), + UAPI_DEF_METHOD_NEEDS_FN(destroy_rwq_ind_table))), + + DECLARE_UVERBS_OBJECT( + UVERBS_OBJECT_WQ, + DECLARE_UVERBS_WRITE_EX( + IB_USER_VERBS_EX_CMD_CREATE_WQ, + ib_uverbs_ex_create_wq, + UAPI_DEF_WRITE_IO_EX(struct ib_uverbs_ex_create_wq, + max_sge, + struct ib_uverbs_ex_create_wq_resp, + wqn), + UAPI_DEF_METHOD_NEEDS_FN(create_wq)), + DECLARE_UVERBS_WRITE_EX( + IB_USER_VERBS_EX_CMD_DESTROY_WQ, + ib_uverbs_ex_destroy_wq, + UAPI_DEF_WRITE_IO_EX(struct ib_uverbs_ex_destroy_wq, + wq_handle, + struct ib_uverbs_ex_destroy_wq_resp, + reserved), + UAPI_DEF_METHOD_NEEDS_FN(destroy_wq)), + DECLARE_UVERBS_WRITE_EX( + IB_USER_VERBS_EX_CMD_MODIFY_WQ, + ib_uverbs_ex_modify_wq, + UAPI_DEF_WRITE_I_EX(struct ib_uverbs_ex_modify_wq, + curr_wq_state), + UAPI_DEF_METHOD_NEEDS_FN(modify_wq))), + + DECLARE_UVERBS_OBJECT( + UVERBS_OBJECT_SRQ, + DECLARE_UVERBS_WRITE(IB_USER_VERBS_CMD_CREATE_SRQ, + ib_uverbs_create_srq, + UAPI_DEF_WRITE_UDATA_IO( + struct ib_uverbs_create_srq, + struct ib_uverbs_create_srq_resp), + UAPI_DEF_METHOD_NEEDS_FN(create_srq)), + DECLARE_UVERBS_WRITE(IB_USER_VERBS_CMD_CREATE_XSRQ, + ib_uverbs_create_xsrq, + UAPI_DEF_WRITE_UDATA_IO( + struct ib_uverbs_create_xsrq, + struct ib_uverbs_create_srq_resp), + UAPI_DEF_METHOD_NEEDS_FN(create_srq)), + DECLARE_UVERBS_WRITE( + IB_USER_VERBS_CMD_DESTROY_SRQ, + ib_uverbs_destroy_srq, + UAPI_DEF_WRITE_IO(struct ib_uverbs_destroy_srq, + struct ib_uverbs_destroy_srq_resp), + UAPI_DEF_METHOD_NEEDS_FN(destroy_srq)), + DECLARE_UVERBS_WRITE( + IB_USER_VERBS_CMD_MODIFY_SRQ, + ib_uverbs_modify_srq, + UAPI_DEF_WRITE_UDATA_I(struct ib_uverbs_modify_srq), + UAPI_DEF_METHOD_NEEDS_FN(modify_srq)), + DECLARE_UVERBS_WRITE( + IB_USER_VERBS_CMD_POST_SRQ_RECV, + ib_uverbs_post_srq_recv, + UAPI_DEF_WRITE_IO(struct ib_uverbs_post_srq_recv, + struct ib_uverbs_post_srq_recv_resp), + UAPI_DEF_METHOD_NEEDS_FN(post_srq_recv)), + DECLARE_UVERBS_WRITE( + IB_USER_VERBS_CMD_QUERY_SRQ, + ib_uverbs_query_srq, + UAPI_DEF_WRITE_IO(struct ib_uverbs_query_srq, + struct ib_uverbs_query_srq_resp), + UAPI_DEF_METHOD_NEEDS_FN(query_srq))), + + DECLARE_UVERBS_OBJECT( + UVERBS_OBJECT_XRCD, + DECLARE_UVERBS_WRITE( + IB_USER_VERBS_CMD_CLOSE_XRCD, + ib_uverbs_close_xrcd, + UAPI_DEF_WRITE_I(struct ib_uverbs_close_xrcd), + UAPI_DEF_METHOD_NEEDS_FN(dealloc_xrcd)), + DECLARE_UVERBS_WRITE(IB_USER_VERBS_CMD_OPEN_QP, + ib_uverbs_open_qp, + UAPI_DEF_WRITE_UDATA_IO( + struct ib_uverbs_open_qp, + struct ib_uverbs_create_qp_resp)), + DECLARE_UVERBS_WRITE(IB_USER_VERBS_CMD_OPEN_XRCD, + ib_uverbs_open_xrcd, + UAPI_DEF_WRITE_UDATA_IO( + struct ib_uverbs_open_xrcd, + struct ib_uverbs_open_xrcd_resp), + UAPI_DEF_METHOD_NEEDS_FN(alloc_xrcd))), + + {}, +}; diff --git a/drivers/infiniband/core/uverbs_ioctl.c b/drivers/infiniband/core/uverbs_ioctl.c new file mode 100644 index 000000000..e47c59490 --- /dev/null +++ b/drivers/infiniband/core/uverbs_ioctl.c @@ -0,0 +1,805 @@ +/* + * Copyright (c) 2017, Mellanox Technologies inc. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include <rdma/rdma_user_ioctl.h> +#include <rdma/uverbs_ioctl.h> +#include "rdma_core.h" +#include "uverbs.h" + +struct bundle_alloc_head { + struct bundle_alloc_head *next; + u8 data[]; +}; + +struct bundle_priv { + /* Must be first */ + struct bundle_alloc_head alloc_head; + struct bundle_alloc_head *allocated_mem; + size_t internal_avail; + size_t internal_used; + + struct radix_tree_root *radix; + const struct uverbs_api_ioctl_method *method_elm; + void __rcu **radix_slots; + unsigned long radix_slots_len; + u32 method_key; + + struct ib_uverbs_attr __user *user_attrs; + struct ib_uverbs_attr *uattrs; + + DECLARE_BITMAP(uobj_finalize, UVERBS_API_ATTR_BKEY_LEN); + DECLARE_BITMAP(spec_finalize, UVERBS_API_ATTR_BKEY_LEN); + DECLARE_BITMAP(uobj_hw_obj_valid, UVERBS_API_ATTR_BKEY_LEN); + + /* + * Must be last. bundle ends in a flex array which overlaps + * internal_buffer. + */ + struct uverbs_attr_bundle bundle; + u64 internal_buffer[32]; +}; + +/* + * Each method has an absolute minimum amount of memory it needs to allocate, + * precompute that amount and determine if the onstack memory can be used or + * if allocation is need. + */ +void uapi_compute_bundle_size(struct uverbs_api_ioctl_method *method_elm, + unsigned int num_attrs) +{ + struct bundle_priv *pbundle; + size_t bundle_size = + offsetof(struct bundle_priv, internal_buffer) + + sizeof(*pbundle->bundle.attrs) * method_elm->key_bitmap_len + + sizeof(*pbundle->uattrs) * num_attrs; + + method_elm->use_stack = bundle_size <= sizeof(*pbundle); + method_elm->bundle_size = + ALIGN(bundle_size + 256, sizeof(*pbundle->internal_buffer)); + + /* Do not want order-2 allocations for this. */ + WARN_ON_ONCE(method_elm->bundle_size > PAGE_SIZE); +} + +/** + * uverbs_alloc() - Quickly allocate memory for use with a bundle + * @bundle: The bundle + * @size: Number of bytes to allocate + * @flags: Allocator flags + * + * The bundle allocator is intended for allocations that are connected with + * processing the system call related to the bundle. The allocated memory is + * always freed once the system call completes, and cannot be freed any other + * way. + * + * This tries to use a small pool of pre-allocated memory for performance. + */ +__malloc void *_uverbs_alloc(struct uverbs_attr_bundle *bundle, size_t size, + gfp_t flags) +{ + struct bundle_priv *pbundle = + container_of(bundle, struct bundle_priv, bundle); + size_t new_used; + void *res; + + if (check_add_overflow(size, pbundle->internal_used, &new_used)) + return ERR_PTR(-EOVERFLOW); + + if (new_used > pbundle->internal_avail) { + struct bundle_alloc_head *buf; + + buf = kvmalloc(struct_size(buf, data, size), flags); + if (!buf) + return ERR_PTR(-ENOMEM); + buf->next = pbundle->allocated_mem; + pbundle->allocated_mem = buf; + return buf->data; + } + + res = (void *)pbundle->internal_buffer + pbundle->internal_used; + pbundle->internal_used = + ALIGN(new_used, sizeof(*pbundle->internal_buffer)); + if (want_init_on_alloc(flags)) + memset(res, 0, size); + return res; +} +EXPORT_SYMBOL(_uverbs_alloc); + +static bool uverbs_is_attr_cleared(const struct ib_uverbs_attr *uattr, + u16 len) +{ + if (uattr->len > sizeof_field(struct ib_uverbs_attr, data)) + return ib_is_buffer_cleared(u64_to_user_ptr(uattr->data) + len, + uattr->len - len); + + return !memchr_inv((const void *)&uattr->data + len, + 0, uattr->len - len); +} + +static int uverbs_set_output(const struct uverbs_attr_bundle *bundle, + const struct uverbs_attr *attr) +{ + struct bundle_priv *pbundle = + container_of(bundle, struct bundle_priv, bundle); + u16 flags; + + flags = pbundle->uattrs[attr->ptr_attr.uattr_idx].flags | + UVERBS_ATTR_F_VALID_OUTPUT; + if (put_user(flags, + &pbundle->user_attrs[attr->ptr_attr.uattr_idx].flags)) + return -EFAULT; + return 0; +} + +static int uverbs_process_idrs_array(struct bundle_priv *pbundle, + const struct uverbs_api_attr *attr_uapi, + struct uverbs_objs_arr_attr *attr, + struct ib_uverbs_attr *uattr, + u32 attr_bkey) +{ + const struct uverbs_attr_spec *spec = &attr_uapi->spec; + size_t array_len; + u32 *idr_vals; + int ret = 0; + size_t i; + + if (uattr->attr_data.reserved) + return -EINVAL; + + if (uattr->len % sizeof(u32)) + return -EINVAL; + + array_len = uattr->len / sizeof(u32); + if (array_len < spec->u2.objs_arr.min_len || + array_len > spec->u2.objs_arr.max_len) + return -EINVAL; + + attr->uobjects = + uverbs_alloc(&pbundle->bundle, + array_size(array_len, sizeof(*attr->uobjects))); + if (IS_ERR(attr->uobjects)) + return PTR_ERR(attr->uobjects); + + /* + * Since idr is 4B and *uobjects is >= 4B, we can use attr->uobjects + * to store idrs array and avoid additional memory allocation. The + * idrs array is offset to the end of the uobjects array so we will be + * able to read idr and replace with a pointer. + */ + idr_vals = (u32 *)(attr->uobjects + array_len) - array_len; + + if (uattr->len > sizeof(uattr->data)) { + ret = copy_from_user(idr_vals, u64_to_user_ptr(uattr->data), + uattr->len); + if (ret) + return -EFAULT; + } else { + memcpy(idr_vals, &uattr->data, uattr->len); + } + + for (i = 0; i != array_len; i++) { + attr->uobjects[i] = uverbs_get_uobject_from_file( + spec->u2.objs_arr.obj_type, spec->u2.objs_arr.access, + idr_vals[i], &pbundle->bundle); + if (IS_ERR(attr->uobjects[i])) { + ret = PTR_ERR(attr->uobjects[i]); + break; + } + } + + attr->len = i; + __set_bit(attr_bkey, pbundle->spec_finalize); + return ret; +} + +static void uverbs_free_idrs_array(const struct uverbs_api_attr *attr_uapi, + struct uverbs_objs_arr_attr *attr, + bool commit, + struct uverbs_attr_bundle *attrs) +{ + const struct uverbs_attr_spec *spec = &attr_uapi->spec; + size_t i; + + for (i = 0; i != attr->len; i++) + uverbs_finalize_object(attr->uobjects[i], + spec->u2.objs_arr.access, false, commit, + attrs); +} + +static int uverbs_process_attr(struct bundle_priv *pbundle, + const struct uverbs_api_attr *attr_uapi, + struct ib_uverbs_attr *uattr, u32 attr_bkey) +{ + const struct uverbs_attr_spec *spec = &attr_uapi->spec; + struct uverbs_attr *e = &pbundle->bundle.attrs[attr_bkey]; + const struct uverbs_attr_spec *val_spec = spec; + struct uverbs_obj_attr *o_attr; + + switch (spec->type) { + case UVERBS_ATTR_TYPE_ENUM_IN: + if (uattr->attr_data.enum_data.elem_id >= spec->u.enum_def.num_elems) + return -EOPNOTSUPP; + + if (uattr->attr_data.enum_data.reserved) + return -EINVAL; + + val_spec = &spec->u2.enum_def.ids[uattr->attr_data.enum_data.elem_id]; + + /* Currently we only support PTR_IN based enums */ + if (val_spec->type != UVERBS_ATTR_TYPE_PTR_IN) + return -EOPNOTSUPP; + + e->ptr_attr.enum_id = uattr->attr_data.enum_data.elem_id; + fallthrough; + case UVERBS_ATTR_TYPE_PTR_IN: + /* Ensure that any data provided by userspace beyond the known + * struct is zero. Userspace that knows how to use some future + * longer struct will fail here if used with an old kernel and + * non-zero content, making ABI compat/discovery simpler. + */ + if (uattr->len > val_spec->u.ptr.len && + val_spec->zero_trailing && + !uverbs_is_attr_cleared(uattr, val_spec->u.ptr.len)) + return -EOPNOTSUPP; + + fallthrough; + case UVERBS_ATTR_TYPE_PTR_OUT: + if (uattr->len < val_spec->u.ptr.min_len || + (!val_spec->zero_trailing && + uattr->len > val_spec->u.ptr.len)) + return -EINVAL; + + if (spec->type != UVERBS_ATTR_TYPE_ENUM_IN && + uattr->attr_data.reserved) + return -EINVAL; + + e->ptr_attr.uattr_idx = uattr - pbundle->uattrs; + e->ptr_attr.len = uattr->len; + + if (val_spec->alloc_and_copy && !uverbs_attr_ptr_is_inline(e)) { + void *p; + + p = uverbs_alloc(&pbundle->bundle, uattr->len); + if (IS_ERR(p)) + return PTR_ERR(p); + + e->ptr_attr.ptr = p; + + if (copy_from_user(p, u64_to_user_ptr(uattr->data), + uattr->len)) + return -EFAULT; + } else { + e->ptr_attr.data = uattr->data; + } + break; + + case UVERBS_ATTR_TYPE_IDR: + case UVERBS_ATTR_TYPE_FD: + if (uattr->attr_data.reserved) + return -EINVAL; + + if (uattr->len != 0) + return -EINVAL; + + o_attr = &e->obj_attr; + o_attr->attr_elm = attr_uapi; + + /* + * The type of uattr->data is u64 for UVERBS_ATTR_TYPE_IDR and + * s64 for UVERBS_ATTR_TYPE_FD. We can cast the u64 to s64 + * here without caring about truncation as we know that the + * IDR implementation today rejects negative IDs + */ + o_attr->uobject = uverbs_get_uobject_from_file( + spec->u.obj.obj_type, spec->u.obj.access, + uattr->data_s64, &pbundle->bundle); + if (IS_ERR(o_attr->uobject)) + return PTR_ERR(o_attr->uobject); + __set_bit(attr_bkey, pbundle->uobj_finalize); + + if (spec->u.obj.access == UVERBS_ACCESS_NEW) { + unsigned int uattr_idx = uattr - pbundle->uattrs; + s64 id = o_attr->uobject->id; + + /* Copy the allocated id to the user-space */ + if (put_user(id, &pbundle->user_attrs[uattr_idx].data)) + return -EFAULT; + } + + break; + + case UVERBS_ATTR_TYPE_IDRS_ARRAY: + return uverbs_process_idrs_array(pbundle, attr_uapi, + &e->objs_arr_attr, uattr, + attr_bkey); + default: + return -EOPNOTSUPP; + } + + return 0; +} + +/* + * We search the radix tree with the method prefix and now we want to fast + * search the suffix bits to get a particular attribute pointer. It is not + * totally clear to me if this breaks the radix tree encasulation or not, but + * it uses the iter data to determine if the method iter points at the same + * chunk that will store the attribute, if so it just derefs it directly. By + * construction in most kernel configs the method and attrs will all fit in a + * single radix chunk, so in most cases this will have no search. Other cases + * this falls back to a full search. + */ +static void __rcu **uapi_get_attr_for_method(struct bundle_priv *pbundle, + u32 attr_key) +{ + void __rcu **slot; + + if (likely(attr_key < pbundle->radix_slots_len)) { + void *entry; + + slot = pbundle->radix_slots + attr_key; + entry = rcu_dereference_raw(*slot); + if (likely(!radix_tree_is_internal_node(entry) && entry)) + return slot; + } + + return radix_tree_lookup_slot(pbundle->radix, + pbundle->method_key | attr_key); +} + +static int uverbs_set_attr(struct bundle_priv *pbundle, + struct ib_uverbs_attr *uattr) +{ + u32 attr_key = uapi_key_attr(uattr->attr_id); + u32 attr_bkey = uapi_bkey_attr(attr_key); + const struct uverbs_api_attr *attr; + void __rcu **slot; + int ret; + + slot = uapi_get_attr_for_method(pbundle, attr_key); + if (!slot) { + /* + * Kernel does not support the attribute but user-space says it + * is mandatory + */ + if (uattr->flags & UVERBS_ATTR_F_MANDATORY) + return -EPROTONOSUPPORT; + return 0; + } + attr = rcu_dereference_protected(*slot, true); + + /* Reject duplicate attributes from user-space */ + if (test_bit(attr_bkey, pbundle->bundle.attr_present)) + return -EINVAL; + + ret = uverbs_process_attr(pbundle, attr, uattr, attr_bkey); + if (ret) + return ret; + + __set_bit(attr_bkey, pbundle->bundle.attr_present); + + return 0; +} + +static int ib_uverbs_run_method(struct bundle_priv *pbundle, + unsigned int num_attrs) +{ + int (*handler)(struct uverbs_attr_bundle *attrs); + size_t uattrs_size = array_size(sizeof(*pbundle->uattrs), num_attrs); + unsigned int destroy_bkey = pbundle->method_elm->destroy_bkey; + unsigned int i; + int ret; + + /* See uverbs_disassociate_api() */ + handler = srcu_dereference( + pbundle->method_elm->handler, + &pbundle->bundle.ufile->device->disassociate_srcu); + if (!handler) + return -EIO; + + pbundle->uattrs = uverbs_alloc(&pbundle->bundle, uattrs_size); + if (IS_ERR(pbundle->uattrs)) + return PTR_ERR(pbundle->uattrs); + if (copy_from_user(pbundle->uattrs, pbundle->user_attrs, uattrs_size)) + return -EFAULT; + + for (i = 0; i != num_attrs; i++) { + ret = uverbs_set_attr(pbundle, &pbundle->uattrs[i]); + if (unlikely(ret)) + return ret; + } + + /* User space did not provide all the mandatory attributes */ + if (unlikely(!bitmap_subset(pbundle->method_elm->attr_mandatory, + pbundle->bundle.attr_present, + pbundle->method_elm->key_bitmap_len))) + return -EINVAL; + + if (pbundle->method_elm->has_udata) + uverbs_fill_udata(&pbundle->bundle, + &pbundle->bundle.driver_udata, + UVERBS_ATTR_UHW_IN, UVERBS_ATTR_UHW_OUT); + else + pbundle->bundle.driver_udata = (struct ib_udata){}; + + if (destroy_bkey != UVERBS_API_ATTR_BKEY_LEN) { + struct uverbs_obj_attr *destroy_attr = + &pbundle->bundle.attrs[destroy_bkey].obj_attr; + + ret = uobj_destroy(destroy_attr->uobject, &pbundle->bundle); + if (ret) + return ret; + __clear_bit(destroy_bkey, pbundle->uobj_finalize); + + ret = handler(&pbundle->bundle); + uobj_put_destroy(destroy_attr->uobject); + } else { + ret = handler(&pbundle->bundle); + } + + /* + * Until the drivers are revised to use the bundle directly we have to + * assume that the driver wrote to its UHW_OUT and flag userspace + * appropriately. + */ + if (!ret && pbundle->method_elm->has_udata) { + const struct uverbs_attr *attr = + uverbs_attr_get(&pbundle->bundle, UVERBS_ATTR_UHW_OUT); + + if (!IS_ERR(attr)) + ret = uverbs_set_output(&pbundle->bundle, attr); + } + + /* + * EPROTONOSUPPORT is ONLY to be returned if the ioctl framework can + * not invoke the method because the request is not supported. No + * other cases should return this code. + */ + if (WARN_ON_ONCE(ret == -EPROTONOSUPPORT)) + return -EINVAL; + + return ret; +} + +static void bundle_destroy(struct bundle_priv *pbundle, bool commit) +{ + unsigned int key_bitmap_len = pbundle->method_elm->key_bitmap_len; + struct bundle_alloc_head *memblock; + unsigned int i; + + /* fast path for simple uobjects */ + i = -1; + while ((i = find_next_bit(pbundle->uobj_finalize, key_bitmap_len, + i + 1)) < key_bitmap_len) { + struct uverbs_attr *attr = &pbundle->bundle.attrs[i]; + + uverbs_finalize_object( + attr->obj_attr.uobject, + attr->obj_attr.attr_elm->spec.u.obj.access, + test_bit(i, pbundle->uobj_hw_obj_valid), + commit, + &pbundle->bundle); + } + + i = -1; + while ((i = find_next_bit(pbundle->spec_finalize, key_bitmap_len, + i + 1)) < key_bitmap_len) { + struct uverbs_attr *attr = &pbundle->bundle.attrs[i]; + const struct uverbs_api_attr *attr_uapi; + void __rcu **slot; + + slot = uapi_get_attr_for_method( + pbundle, + pbundle->method_key | uapi_bkey_to_key_attr(i)); + if (WARN_ON(!slot)) + continue; + + attr_uapi = rcu_dereference_protected(*slot, true); + + if (attr_uapi->spec.type == UVERBS_ATTR_TYPE_IDRS_ARRAY) { + uverbs_free_idrs_array(attr_uapi, &attr->objs_arr_attr, + commit, &pbundle->bundle); + } + } + + for (memblock = pbundle->allocated_mem; memblock;) { + struct bundle_alloc_head *tmp = memblock; + + memblock = memblock->next; + kvfree(tmp); + } +} + +static int ib_uverbs_cmd_verbs(struct ib_uverbs_file *ufile, + struct ib_uverbs_ioctl_hdr *hdr, + struct ib_uverbs_attr __user *user_attrs) +{ + const struct uverbs_api_ioctl_method *method_elm; + struct uverbs_api *uapi = ufile->device->uapi; + struct radix_tree_iter attrs_iter; + struct bundle_priv *pbundle; + struct bundle_priv onstack; + void __rcu **slot; + int ret; + + if (unlikely(hdr->driver_id != uapi->driver_id)) + return -EINVAL; + + slot = radix_tree_iter_lookup( + &uapi->radix, &attrs_iter, + uapi_key_obj(hdr->object_id) | + uapi_key_ioctl_method(hdr->method_id)); + if (unlikely(!slot)) + return -EPROTONOSUPPORT; + method_elm = rcu_dereference_protected(*slot, true); + + if (!method_elm->use_stack) { + pbundle = kmalloc(method_elm->bundle_size, GFP_KERNEL); + if (!pbundle) + return -ENOMEM; + pbundle->internal_avail = + method_elm->bundle_size - + offsetof(struct bundle_priv, internal_buffer); + pbundle->alloc_head.next = NULL; + pbundle->allocated_mem = &pbundle->alloc_head; + } else { + pbundle = &onstack; + pbundle->internal_avail = sizeof(pbundle->internal_buffer); + pbundle->allocated_mem = NULL; + } + + /* Space for the pbundle->bundle.attrs flex array */ + pbundle->method_elm = method_elm; + pbundle->method_key = attrs_iter.index; + pbundle->bundle.ufile = ufile; + pbundle->bundle.context = NULL; /* only valid if bundle has uobject */ + pbundle->radix = &uapi->radix; + pbundle->radix_slots = slot; + pbundle->radix_slots_len = radix_tree_chunk_size(&attrs_iter); + pbundle->user_attrs = user_attrs; + + pbundle->internal_used = ALIGN(pbundle->method_elm->key_bitmap_len * + sizeof(*pbundle->bundle.attrs), + sizeof(*pbundle->internal_buffer)); + memset(pbundle->bundle.attr_present, 0, + sizeof(pbundle->bundle.attr_present)); + memset(pbundle->uobj_finalize, 0, sizeof(pbundle->uobj_finalize)); + memset(pbundle->spec_finalize, 0, sizeof(pbundle->spec_finalize)); + memset(pbundle->uobj_hw_obj_valid, 0, + sizeof(pbundle->uobj_hw_obj_valid)); + + ret = ib_uverbs_run_method(pbundle, hdr->num_attrs); + bundle_destroy(pbundle, ret == 0); + return ret; +} + +long ib_uverbs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) +{ + struct ib_uverbs_file *file = filp->private_data; + struct ib_uverbs_ioctl_hdr __user *user_hdr = + (struct ib_uverbs_ioctl_hdr __user *)arg; + struct ib_uverbs_ioctl_hdr hdr; + int srcu_key; + int err; + + if (unlikely(cmd != RDMA_VERBS_IOCTL)) + return -ENOIOCTLCMD; + + err = copy_from_user(&hdr, user_hdr, sizeof(hdr)); + if (err) + return -EFAULT; + + if (hdr.length > PAGE_SIZE || + hdr.length != struct_size(&hdr, attrs, hdr.num_attrs)) + return -EINVAL; + + if (hdr.reserved1 || hdr.reserved2) + return -EPROTONOSUPPORT; + + srcu_key = srcu_read_lock(&file->device->disassociate_srcu); + err = ib_uverbs_cmd_verbs(file, &hdr, user_hdr->attrs); + srcu_read_unlock(&file->device->disassociate_srcu, srcu_key); + return err; +} + +int uverbs_get_flags64(u64 *to, const struct uverbs_attr_bundle *attrs_bundle, + size_t idx, u64 allowed_bits) +{ + const struct uverbs_attr *attr; + u64 flags; + + attr = uverbs_attr_get(attrs_bundle, idx); + /* Missing attribute means 0 flags */ + if (IS_ERR(attr)) { + *to = 0; + return 0; + } + + /* + * New userspace code should use 8 bytes to pass flags, but we + * transparently support old userspaces that were using 4 bytes as + * well. + */ + if (attr->ptr_attr.len == 8) + flags = attr->ptr_attr.data; + else if (attr->ptr_attr.len == 4) + flags = *(u32 *)&attr->ptr_attr.data; + else + return -EINVAL; + + if (flags & ~allowed_bits) + return -EINVAL; + + *to = flags; + return 0; +} +EXPORT_SYMBOL(uverbs_get_flags64); + +int uverbs_get_flags32(u32 *to, const struct uverbs_attr_bundle *attrs_bundle, + size_t idx, u64 allowed_bits) +{ + u64 flags; + int ret; + + ret = uverbs_get_flags64(&flags, attrs_bundle, idx, allowed_bits); + if (ret) + return ret; + + if (flags > U32_MAX) + return -EINVAL; + *to = flags; + + return 0; +} +EXPORT_SYMBOL(uverbs_get_flags32); + +/* + * Fill a ib_udata struct (core or uhw) using the given attribute IDs. + * This is primarily used to convert the UVERBS_ATTR_UHW() into the + * ib_udata format used by the drivers. + */ +void uverbs_fill_udata(struct uverbs_attr_bundle *bundle, + struct ib_udata *udata, unsigned int attr_in, + unsigned int attr_out) +{ + struct bundle_priv *pbundle = + container_of(bundle, struct bundle_priv, bundle); + const struct uverbs_attr *in = + uverbs_attr_get(&pbundle->bundle, attr_in); + const struct uverbs_attr *out = + uverbs_attr_get(&pbundle->bundle, attr_out); + + if (!IS_ERR(in)) { + udata->inlen = in->ptr_attr.len; + if (uverbs_attr_ptr_is_inline(in)) + udata->inbuf = + &pbundle->user_attrs[in->ptr_attr.uattr_idx] + .data; + else + udata->inbuf = u64_to_user_ptr(in->ptr_attr.data); + } else { + udata->inbuf = NULL; + udata->inlen = 0; + } + + if (!IS_ERR(out)) { + udata->outbuf = u64_to_user_ptr(out->ptr_attr.data); + udata->outlen = out->ptr_attr.len; + } else { + udata->outbuf = NULL; + udata->outlen = 0; + } +} + +int uverbs_copy_to(const struct uverbs_attr_bundle *bundle, size_t idx, + const void *from, size_t size) +{ + const struct uverbs_attr *attr = uverbs_attr_get(bundle, idx); + size_t min_size; + + if (IS_ERR(attr)) + return PTR_ERR(attr); + + min_size = min_t(size_t, attr->ptr_attr.len, size); + if (copy_to_user(u64_to_user_ptr(attr->ptr_attr.data), from, min_size)) + return -EFAULT; + + return uverbs_set_output(bundle, attr); +} +EXPORT_SYMBOL(uverbs_copy_to); + + +/* + * This is only used if the caller has directly used copy_to_use to write the + * data. It signals to user space that the buffer is filled in. + */ +int uverbs_output_written(const struct uverbs_attr_bundle *bundle, size_t idx) +{ + const struct uverbs_attr *attr = uverbs_attr_get(bundle, idx); + + if (IS_ERR(attr)) + return PTR_ERR(attr); + + return uverbs_set_output(bundle, attr); +} + +int _uverbs_get_const(s64 *to, const struct uverbs_attr_bundle *attrs_bundle, + size_t idx, s64 lower_bound, u64 upper_bound, + s64 *def_val) +{ + const struct uverbs_attr *attr; + + attr = uverbs_attr_get(attrs_bundle, idx); + if (IS_ERR(attr)) { + if ((PTR_ERR(attr) != -ENOENT) || !def_val) + return PTR_ERR(attr); + + *to = *def_val; + } else { + *to = attr->ptr_attr.data; + } + + if (*to < lower_bound || (*to > 0 && (u64)*to > upper_bound)) + return -EINVAL; + + return 0; +} +EXPORT_SYMBOL(_uverbs_get_const); + +int uverbs_copy_to_struct_or_zero(const struct uverbs_attr_bundle *bundle, + size_t idx, const void *from, size_t size) +{ + const struct uverbs_attr *attr = uverbs_attr_get(bundle, idx); + + if (IS_ERR(attr)) + return PTR_ERR(attr); + + if (size < attr->ptr_attr.len) { + if (clear_user(u64_to_user_ptr(attr->ptr_attr.data) + size, + attr->ptr_attr.len - size)) + return -EFAULT; + } + return uverbs_copy_to(bundle, idx, from, size); +} +EXPORT_SYMBOL(uverbs_copy_to_struct_or_zero); + +/* Once called an abort will call through to the type's destroy_hw() */ +void uverbs_finalize_uobj_create(const struct uverbs_attr_bundle *bundle, + u16 idx) +{ + struct bundle_priv *pbundle = + container_of(bundle, struct bundle_priv, bundle); + + __set_bit(uapi_bkey_attr(uapi_key_attr(idx)), + pbundle->uobj_hw_obj_valid); +} +EXPORT_SYMBOL(uverbs_finalize_uobj_create); diff --git a/drivers/infiniband/core/uverbs_main.c b/drivers/infiniband/core/uverbs_main.c new file mode 100644 index 000000000..f2fcdb70d --- /dev/null +++ b/drivers/infiniband/core/uverbs_main.c @@ -0,0 +1,1315 @@ +/* + * Copyright (c) 2005 Topspin Communications. All rights reserved. + * Copyright (c) 2005, 2006 Cisco Systems. All rights reserved. + * Copyright (c) 2005 Mellanox Technologies. All rights reserved. + * Copyright (c) 2005 Voltaire, Inc. All rights reserved. + * Copyright (c) 2005 PathScale, Inc. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include <linux/module.h> +#include <linux/init.h> +#include <linux/device.h> +#include <linux/err.h> +#include <linux/fs.h> +#include <linux/poll.h> +#include <linux/sched.h> +#include <linux/file.h> +#include <linux/cdev.h> +#include <linux/anon_inodes.h> +#include <linux/slab.h> +#include <linux/sched/mm.h> + +#include <linux/uaccess.h> + +#include <rdma/ib.h> +#include <rdma/uverbs_std_types.h> +#include <rdma/rdma_netlink.h> + +#include "uverbs.h" +#include "core_priv.h" +#include "rdma_core.h" + +MODULE_AUTHOR("Roland Dreier"); +MODULE_DESCRIPTION("InfiniBand userspace verbs access"); +MODULE_LICENSE("Dual BSD/GPL"); + +enum { + IB_UVERBS_MAJOR = 231, + IB_UVERBS_BASE_MINOR = 192, + IB_UVERBS_MAX_DEVICES = RDMA_MAX_PORTS, + IB_UVERBS_NUM_FIXED_MINOR = 32, + IB_UVERBS_NUM_DYNAMIC_MINOR = IB_UVERBS_MAX_DEVICES - IB_UVERBS_NUM_FIXED_MINOR, +}; + +#define IB_UVERBS_BASE_DEV MKDEV(IB_UVERBS_MAJOR, IB_UVERBS_BASE_MINOR) + +static dev_t dynamic_uverbs_dev; +static struct class *uverbs_class; + +static DEFINE_IDA(uverbs_ida); +static int ib_uverbs_add_one(struct ib_device *device); +static void ib_uverbs_remove_one(struct ib_device *device, void *client_data); + +/* + * Must be called with the ufile->device->disassociate_srcu held, and the lock + * must be held until use of the ucontext is finished. + */ +struct ib_ucontext *ib_uverbs_get_ucontext_file(struct ib_uverbs_file *ufile) +{ + /* + * We do not hold the hw_destroy_rwsem lock for this flow, instead + * srcu is used. It does not matter if someone races this with + * get_context, we get NULL or valid ucontext. + */ + struct ib_ucontext *ucontext = smp_load_acquire(&ufile->ucontext); + + if (!srcu_dereference(ufile->device->ib_dev, + &ufile->device->disassociate_srcu)) + return ERR_PTR(-EIO); + + if (!ucontext) + return ERR_PTR(-EINVAL); + + return ucontext; +} +EXPORT_SYMBOL(ib_uverbs_get_ucontext_file); + +int uverbs_dealloc_mw(struct ib_mw *mw) +{ + struct ib_pd *pd = mw->pd; + int ret; + + ret = mw->device->ops.dealloc_mw(mw); + if (ret) + return ret; + + atomic_dec(&pd->usecnt); + kfree(mw); + return ret; +} + +static void ib_uverbs_release_dev(struct device *device) +{ + struct ib_uverbs_device *dev = + container_of(device, struct ib_uverbs_device, dev); + + uverbs_destroy_api(dev->uapi); + cleanup_srcu_struct(&dev->disassociate_srcu); + mutex_destroy(&dev->lists_mutex); + mutex_destroy(&dev->xrcd_tree_mutex); + kfree(dev); +} + +void ib_uverbs_release_ucq(struct ib_uverbs_completion_event_file *ev_file, + struct ib_ucq_object *uobj) +{ + struct ib_uverbs_event *evt, *tmp; + + if (ev_file) { + spin_lock_irq(&ev_file->ev_queue.lock); + list_for_each_entry_safe(evt, tmp, &uobj->comp_list, obj_list) { + list_del(&evt->list); + kfree(evt); + } + spin_unlock_irq(&ev_file->ev_queue.lock); + + uverbs_uobject_put(&ev_file->uobj); + } + + ib_uverbs_release_uevent(&uobj->uevent); +} + +void ib_uverbs_release_uevent(struct ib_uevent_object *uobj) +{ + struct ib_uverbs_async_event_file *async_file = uobj->event_file; + struct ib_uverbs_event *evt, *tmp; + + if (!async_file) + return; + + spin_lock_irq(&async_file->ev_queue.lock); + list_for_each_entry_safe(evt, tmp, &uobj->event_list, obj_list) { + list_del(&evt->list); + kfree(evt); + } + spin_unlock_irq(&async_file->ev_queue.lock); + uverbs_uobject_put(&async_file->uobj); +} + +void ib_uverbs_detach_umcast(struct ib_qp *qp, + struct ib_uqp_object *uobj) +{ + struct ib_uverbs_mcast_entry *mcast, *tmp; + + list_for_each_entry_safe(mcast, tmp, &uobj->mcast_list, list) { + ib_detach_mcast(qp, &mcast->gid, mcast->lid); + list_del(&mcast->list); + kfree(mcast); + } +} + +static void ib_uverbs_comp_dev(struct ib_uverbs_device *dev) +{ + complete(&dev->comp); +} + +void ib_uverbs_release_file(struct kref *ref) +{ + struct ib_uverbs_file *file = + container_of(ref, struct ib_uverbs_file, ref); + struct ib_device *ib_dev; + int srcu_key; + + release_ufile_idr_uobject(file); + + srcu_key = srcu_read_lock(&file->device->disassociate_srcu); + ib_dev = srcu_dereference(file->device->ib_dev, + &file->device->disassociate_srcu); + if (ib_dev && !ib_dev->ops.disassociate_ucontext) + module_put(ib_dev->ops.owner); + srcu_read_unlock(&file->device->disassociate_srcu, srcu_key); + + if (atomic_dec_and_test(&file->device->refcount)) + ib_uverbs_comp_dev(file->device); + + if (file->default_async_file) + uverbs_uobject_put(&file->default_async_file->uobj); + put_device(&file->device->dev); + + if (file->disassociate_page) + __free_pages(file->disassociate_page, 0); + mutex_destroy(&file->umap_lock); + mutex_destroy(&file->ucontext_lock); + kfree(file); +} + +static ssize_t ib_uverbs_event_read(struct ib_uverbs_event_queue *ev_queue, + struct file *filp, char __user *buf, + size_t count, loff_t *pos, + size_t eventsz) +{ + struct ib_uverbs_event *event; + int ret = 0; + + spin_lock_irq(&ev_queue->lock); + + while (list_empty(&ev_queue->event_list)) { + if (ev_queue->is_closed) { + spin_unlock_irq(&ev_queue->lock); + return -EIO; + } + + spin_unlock_irq(&ev_queue->lock); + if (filp->f_flags & O_NONBLOCK) + return -EAGAIN; + + if (wait_event_interruptible(ev_queue->poll_wait, + (!list_empty(&ev_queue->event_list) || + ev_queue->is_closed))) + return -ERESTARTSYS; + + spin_lock_irq(&ev_queue->lock); + } + + event = list_entry(ev_queue->event_list.next, struct ib_uverbs_event, list); + + if (eventsz > count) { + ret = -EINVAL; + event = NULL; + } else { + list_del(ev_queue->event_list.next); + if (event->counter) { + ++(*event->counter); + list_del(&event->obj_list); + } + } + + spin_unlock_irq(&ev_queue->lock); + + if (event) { + if (copy_to_user(buf, event, eventsz)) + ret = -EFAULT; + else + ret = eventsz; + } + + kfree(event); + + return ret; +} + +static ssize_t ib_uverbs_async_event_read(struct file *filp, char __user *buf, + size_t count, loff_t *pos) +{ + struct ib_uverbs_async_event_file *file = filp->private_data; + + return ib_uverbs_event_read(&file->ev_queue, filp, buf, count, pos, + sizeof(struct ib_uverbs_async_event_desc)); +} + +static ssize_t ib_uverbs_comp_event_read(struct file *filp, char __user *buf, + size_t count, loff_t *pos) +{ + struct ib_uverbs_completion_event_file *comp_ev_file = + filp->private_data; + + return ib_uverbs_event_read(&comp_ev_file->ev_queue, filp, buf, count, + pos, + sizeof(struct ib_uverbs_comp_event_desc)); +} + +static __poll_t ib_uverbs_event_poll(struct ib_uverbs_event_queue *ev_queue, + struct file *filp, + struct poll_table_struct *wait) +{ + __poll_t pollflags = 0; + + poll_wait(filp, &ev_queue->poll_wait, wait); + + spin_lock_irq(&ev_queue->lock); + if (!list_empty(&ev_queue->event_list)) + pollflags = EPOLLIN | EPOLLRDNORM; + else if (ev_queue->is_closed) + pollflags = EPOLLERR; + spin_unlock_irq(&ev_queue->lock); + + return pollflags; +} + +static __poll_t ib_uverbs_async_event_poll(struct file *filp, + struct poll_table_struct *wait) +{ + struct ib_uverbs_async_event_file *file = filp->private_data; + + return ib_uverbs_event_poll(&file->ev_queue, filp, wait); +} + +static __poll_t ib_uverbs_comp_event_poll(struct file *filp, + struct poll_table_struct *wait) +{ + struct ib_uverbs_completion_event_file *comp_ev_file = + filp->private_data; + + return ib_uverbs_event_poll(&comp_ev_file->ev_queue, filp, wait); +} + +static int ib_uverbs_async_event_fasync(int fd, struct file *filp, int on) +{ + struct ib_uverbs_async_event_file *file = filp->private_data; + + return fasync_helper(fd, filp, on, &file->ev_queue.async_queue); +} + +static int ib_uverbs_comp_event_fasync(int fd, struct file *filp, int on) +{ + struct ib_uverbs_completion_event_file *comp_ev_file = + filp->private_data; + + return fasync_helper(fd, filp, on, &comp_ev_file->ev_queue.async_queue); +} + +const struct file_operations uverbs_event_fops = { + .owner = THIS_MODULE, + .read = ib_uverbs_comp_event_read, + .poll = ib_uverbs_comp_event_poll, + .release = uverbs_uobject_fd_release, + .fasync = ib_uverbs_comp_event_fasync, + .llseek = no_llseek, +}; + +const struct file_operations uverbs_async_event_fops = { + .owner = THIS_MODULE, + .read = ib_uverbs_async_event_read, + .poll = ib_uverbs_async_event_poll, + .release = uverbs_async_event_release, + .fasync = ib_uverbs_async_event_fasync, + .llseek = no_llseek, +}; + +void ib_uverbs_comp_handler(struct ib_cq *cq, void *cq_context) +{ + struct ib_uverbs_event_queue *ev_queue = cq_context; + struct ib_ucq_object *uobj; + struct ib_uverbs_event *entry; + unsigned long flags; + + if (!ev_queue) + return; + + spin_lock_irqsave(&ev_queue->lock, flags); + if (ev_queue->is_closed) { + spin_unlock_irqrestore(&ev_queue->lock, flags); + return; + } + + entry = kmalloc(sizeof(*entry), GFP_ATOMIC); + if (!entry) { + spin_unlock_irqrestore(&ev_queue->lock, flags); + return; + } + + uobj = cq->uobject; + + entry->desc.comp.cq_handle = cq->uobject->uevent.uobject.user_handle; + entry->counter = &uobj->comp_events_reported; + + list_add_tail(&entry->list, &ev_queue->event_list); + list_add_tail(&entry->obj_list, &uobj->comp_list); + spin_unlock_irqrestore(&ev_queue->lock, flags); + + wake_up_interruptible(&ev_queue->poll_wait); + kill_fasync(&ev_queue->async_queue, SIGIO, POLL_IN); +} + +void ib_uverbs_async_handler(struct ib_uverbs_async_event_file *async_file, + __u64 element, __u64 event, + struct list_head *obj_list, u32 *counter) +{ + struct ib_uverbs_event *entry; + unsigned long flags; + + if (!async_file) + return; + + spin_lock_irqsave(&async_file->ev_queue.lock, flags); + if (async_file->ev_queue.is_closed) { + spin_unlock_irqrestore(&async_file->ev_queue.lock, flags); + return; + } + + entry = kmalloc(sizeof(*entry), GFP_ATOMIC); + if (!entry) { + spin_unlock_irqrestore(&async_file->ev_queue.lock, flags); + return; + } + + entry->desc.async.element = element; + entry->desc.async.event_type = event; + entry->desc.async.reserved = 0; + entry->counter = counter; + + list_add_tail(&entry->list, &async_file->ev_queue.event_list); + if (obj_list) + list_add_tail(&entry->obj_list, obj_list); + spin_unlock_irqrestore(&async_file->ev_queue.lock, flags); + + wake_up_interruptible(&async_file->ev_queue.poll_wait); + kill_fasync(&async_file->ev_queue.async_queue, SIGIO, POLL_IN); +} + +static void uverbs_uobj_event(struct ib_uevent_object *eobj, + struct ib_event *event) +{ + ib_uverbs_async_handler(eobj->event_file, + eobj->uobject.user_handle, event->event, + &eobj->event_list, &eobj->events_reported); +} + +void ib_uverbs_cq_event_handler(struct ib_event *event, void *context_ptr) +{ + uverbs_uobj_event(&event->element.cq->uobject->uevent, event); +} + +void ib_uverbs_qp_event_handler(struct ib_event *event, void *context_ptr) +{ + /* for XRC target qp's, check that qp is live */ + if (!event->element.qp->uobject) + return; + + uverbs_uobj_event(&event->element.qp->uobject->uevent, event); +} + +void ib_uverbs_wq_event_handler(struct ib_event *event, void *context_ptr) +{ + uverbs_uobj_event(&event->element.wq->uobject->uevent, event); +} + +void ib_uverbs_srq_event_handler(struct ib_event *event, void *context_ptr) +{ + uverbs_uobj_event(&event->element.srq->uobject->uevent, event); +} + +static void ib_uverbs_event_handler(struct ib_event_handler *handler, + struct ib_event *event) +{ + ib_uverbs_async_handler( + container_of(handler, struct ib_uverbs_async_event_file, + event_handler), + event->element.port_num, event->event, NULL, NULL); +} + +void ib_uverbs_init_event_queue(struct ib_uverbs_event_queue *ev_queue) +{ + spin_lock_init(&ev_queue->lock); + INIT_LIST_HEAD(&ev_queue->event_list); + init_waitqueue_head(&ev_queue->poll_wait); + ev_queue->is_closed = 0; + ev_queue->async_queue = NULL; +} + +void ib_uverbs_init_async_event_file( + struct ib_uverbs_async_event_file *async_file) +{ + struct ib_uverbs_file *uverbs_file = async_file->uobj.ufile; + struct ib_device *ib_dev = async_file->uobj.context->device; + + ib_uverbs_init_event_queue(&async_file->ev_queue); + + /* The first async_event_file becomes the default one for the file. */ + mutex_lock(&uverbs_file->ucontext_lock); + if (!uverbs_file->default_async_file) { + /* Pairs with the put in ib_uverbs_release_file */ + uverbs_uobject_get(&async_file->uobj); + smp_store_release(&uverbs_file->default_async_file, async_file); + } + mutex_unlock(&uverbs_file->ucontext_lock); + + INIT_IB_EVENT_HANDLER(&async_file->event_handler, ib_dev, + ib_uverbs_event_handler); + ib_register_event_handler(&async_file->event_handler); +} + +static ssize_t verify_hdr(struct ib_uverbs_cmd_hdr *hdr, + struct ib_uverbs_ex_cmd_hdr *ex_hdr, size_t count, + const struct uverbs_api_write_method *method_elm) +{ + if (method_elm->is_ex) { + count -= sizeof(*hdr) + sizeof(*ex_hdr); + + if ((hdr->in_words + ex_hdr->provider_in_words) * 8 != count) + return -EINVAL; + + if (hdr->in_words * 8 < method_elm->req_size) + return -ENOSPC; + + if (ex_hdr->cmd_hdr_reserved) + return -EINVAL; + + if (ex_hdr->response) { + if (!hdr->out_words && !ex_hdr->provider_out_words) + return -EINVAL; + + if (hdr->out_words * 8 < method_elm->resp_size) + return -ENOSPC; + + if (!access_ok(u64_to_user_ptr(ex_hdr->response), + (hdr->out_words + ex_hdr->provider_out_words) * 8)) + return -EFAULT; + } else { + if (hdr->out_words || ex_hdr->provider_out_words) + return -EINVAL; + } + + return 0; + } + + /* not extended command */ + if (hdr->in_words * 4 != count) + return -EINVAL; + + if (count < method_elm->req_size + sizeof(*hdr)) { + /* + * rdma-core v18 and v19 have a bug where they send DESTROY_CQ + * with a 16 byte write instead of 24. Old kernels didn't + * check the size so they allowed this. Now that the size is + * checked provide a compatibility work around to not break + * those userspaces. + */ + if (hdr->command == IB_USER_VERBS_CMD_DESTROY_CQ && + count == 16) { + hdr->in_words = 6; + return 0; + } + return -ENOSPC; + } + if (hdr->out_words * 4 < method_elm->resp_size) + return -ENOSPC; + + return 0; +} + +static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf, + size_t count, loff_t *pos) +{ + struct ib_uverbs_file *file = filp->private_data; + const struct uverbs_api_write_method *method_elm; + struct uverbs_api *uapi = file->device->uapi; + struct ib_uverbs_ex_cmd_hdr ex_hdr; + struct ib_uverbs_cmd_hdr hdr; + struct uverbs_attr_bundle bundle; + int srcu_key; + ssize_t ret; + + if (!ib_safe_file_access(filp)) { + pr_err_once("uverbs_write: process %d (%s) changed security contexts after opening file descriptor, this is not allowed.\n", + task_tgid_vnr(current), current->comm); + return -EACCES; + } + + if (count < sizeof(hdr)) + return -EINVAL; + + if (copy_from_user(&hdr, buf, sizeof(hdr))) + return -EFAULT; + + method_elm = uapi_get_method(uapi, hdr.command); + if (IS_ERR(method_elm)) + return PTR_ERR(method_elm); + + if (method_elm->is_ex) { + if (count < (sizeof(hdr) + sizeof(ex_hdr))) + return -EINVAL; + if (copy_from_user(&ex_hdr, buf + sizeof(hdr), sizeof(ex_hdr))) + return -EFAULT; + } + + ret = verify_hdr(&hdr, &ex_hdr, count, method_elm); + if (ret) + return ret; + + srcu_key = srcu_read_lock(&file->device->disassociate_srcu); + + buf += sizeof(hdr); + + memset(bundle.attr_present, 0, sizeof(bundle.attr_present)); + bundle.ufile = file; + bundle.context = NULL; /* only valid if bundle has uobject */ + bundle.uobject = NULL; + if (!method_elm->is_ex) { + size_t in_len = hdr.in_words * 4 - sizeof(hdr); + size_t out_len = hdr.out_words * 4; + u64 response = 0; + + if (method_elm->has_udata) { + bundle.driver_udata.inlen = + in_len - method_elm->req_size; + in_len = method_elm->req_size; + if (bundle.driver_udata.inlen) + bundle.driver_udata.inbuf = buf + in_len; + else + bundle.driver_udata.inbuf = NULL; + } else { + memset(&bundle.driver_udata, 0, + sizeof(bundle.driver_udata)); + } + + if (method_elm->has_resp) { + /* + * The macros check that if has_resp is set + * then the command request structure starts + * with a '__aligned u64 response' member. + */ + ret = get_user(response, (const u64 __user *)buf); + if (ret) + goto out_unlock; + + if (method_elm->has_udata) { + bundle.driver_udata.outlen = + out_len - method_elm->resp_size; + out_len = method_elm->resp_size; + if (bundle.driver_udata.outlen) + bundle.driver_udata.outbuf = + u64_to_user_ptr(response + + out_len); + else + bundle.driver_udata.outbuf = NULL; + } + } else { + bundle.driver_udata.outlen = 0; + bundle.driver_udata.outbuf = NULL; + } + + ib_uverbs_init_udata_buf_or_null( + &bundle.ucore, buf, u64_to_user_ptr(response), + in_len, out_len); + } else { + buf += sizeof(ex_hdr); + + ib_uverbs_init_udata_buf_or_null(&bundle.ucore, buf, + u64_to_user_ptr(ex_hdr.response), + hdr.in_words * 8, hdr.out_words * 8); + + ib_uverbs_init_udata_buf_or_null( + &bundle.driver_udata, buf + bundle.ucore.inlen, + u64_to_user_ptr(ex_hdr.response) + bundle.ucore.outlen, + ex_hdr.provider_in_words * 8, + ex_hdr.provider_out_words * 8); + + } + + ret = method_elm->handler(&bundle); + if (bundle.uobject) + uverbs_finalize_object(bundle.uobject, UVERBS_ACCESS_NEW, true, + !ret, &bundle); +out_unlock: + srcu_read_unlock(&file->device->disassociate_srcu, srcu_key); + return (ret) ? : count; +} + +static const struct vm_operations_struct rdma_umap_ops; + +static int ib_uverbs_mmap(struct file *filp, struct vm_area_struct *vma) +{ + struct ib_uverbs_file *file = filp->private_data; + struct ib_ucontext *ucontext; + int ret = 0; + int srcu_key; + + srcu_key = srcu_read_lock(&file->device->disassociate_srcu); + ucontext = ib_uverbs_get_ucontext_file(file); + if (IS_ERR(ucontext)) { + ret = PTR_ERR(ucontext); + goto out; + } + vma->vm_ops = &rdma_umap_ops; + ret = ucontext->device->ops.mmap(ucontext, vma); +out: + srcu_read_unlock(&file->device->disassociate_srcu, srcu_key); + return ret; +} + +/* + * The VMA has been dup'd, initialize the vm_private_data with a new tracking + * struct + */ +static void rdma_umap_open(struct vm_area_struct *vma) +{ + struct ib_uverbs_file *ufile = vma->vm_file->private_data; + struct rdma_umap_priv *opriv = vma->vm_private_data; + struct rdma_umap_priv *priv; + + if (!opriv) + return; + + /* We are racing with disassociation */ + if (!down_read_trylock(&ufile->hw_destroy_rwsem)) + goto out_zap; + /* + * Disassociation already completed, the VMA should already be zapped. + */ + if (!ufile->ucontext) + goto out_unlock; + + priv = kzalloc(sizeof(*priv), GFP_KERNEL); + if (!priv) + goto out_unlock; + rdma_umap_priv_init(priv, vma, opriv->entry); + + up_read(&ufile->hw_destroy_rwsem); + return; + +out_unlock: + up_read(&ufile->hw_destroy_rwsem); +out_zap: + /* + * We can't allow the VMA to be created with the actual IO pages, that + * would break our API contract, and it can't be stopped at this + * point, so zap it. + */ + vma->vm_private_data = NULL; + zap_vma_ptes(vma, vma->vm_start, vma->vm_end - vma->vm_start); +} + +static void rdma_umap_close(struct vm_area_struct *vma) +{ + struct ib_uverbs_file *ufile = vma->vm_file->private_data; + struct rdma_umap_priv *priv = vma->vm_private_data; + + if (!priv) + return; + + /* + * The vma holds a reference on the struct file that created it, which + * in turn means that the ib_uverbs_file is guaranteed to exist at + * this point. + */ + mutex_lock(&ufile->umap_lock); + if (priv->entry) + rdma_user_mmap_entry_put(priv->entry); + + list_del(&priv->list); + mutex_unlock(&ufile->umap_lock); + kfree(priv); +} + +/* + * Once the zap_vma_ptes has been called touches to the VMA will come here and + * we return a dummy writable zero page for all the pfns. + */ +static vm_fault_t rdma_umap_fault(struct vm_fault *vmf) +{ + struct ib_uverbs_file *ufile = vmf->vma->vm_file->private_data; + struct rdma_umap_priv *priv = vmf->vma->vm_private_data; + vm_fault_t ret = 0; + + if (!priv) + return VM_FAULT_SIGBUS; + + /* Read only pages can just use the system zero page. */ + if (!(vmf->vma->vm_flags & (VM_WRITE | VM_MAYWRITE))) { + vmf->page = ZERO_PAGE(vmf->address); + get_page(vmf->page); + return 0; + } + + mutex_lock(&ufile->umap_lock); + if (!ufile->disassociate_page) + ufile->disassociate_page = + alloc_pages(vmf->gfp_mask | __GFP_ZERO, 0); + + if (ufile->disassociate_page) { + /* + * This VMA is forced to always be shared so this doesn't have + * to worry about COW. + */ + vmf->page = ufile->disassociate_page; + get_page(vmf->page); + } else { + ret = VM_FAULT_SIGBUS; + } + mutex_unlock(&ufile->umap_lock); + + return ret; +} + +static const struct vm_operations_struct rdma_umap_ops = { + .open = rdma_umap_open, + .close = rdma_umap_close, + .fault = rdma_umap_fault, +}; + +void uverbs_user_mmap_disassociate(struct ib_uverbs_file *ufile) +{ + struct rdma_umap_priv *priv, *next_priv; + + lockdep_assert_held(&ufile->hw_destroy_rwsem); + + while (1) { + struct mm_struct *mm = NULL; + + /* Get an arbitrary mm pointer that hasn't been cleaned yet */ + mutex_lock(&ufile->umap_lock); + while (!list_empty(&ufile->umaps)) { + int ret; + + priv = list_first_entry(&ufile->umaps, + struct rdma_umap_priv, list); + mm = priv->vma->vm_mm; + ret = mmget_not_zero(mm); + if (!ret) { + list_del_init(&priv->list); + if (priv->entry) { + rdma_user_mmap_entry_put(priv->entry); + priv->entry = NULL; + } + mm = NULL; + continue; + } + break; + } + mutex_unlock(&ufile->umap_lock); + if (!mm) + return; + + /* + * The umap_lock is nested under mmap_lock since it used within + * the vma_ops callbacks, so we have to clean the list one mm + * at a time to get the lock ordering right. Typically there + * will only be one mm, so no big deal. + */ + mmap_read_lock(mm); + mutex_lock(&ufile->umap_lock); + list_for_each_entry_safe (priv, next_priv, &ufile->umaps, + list) { + struct vm_area_struct *vma = priv->vma; + + if (vma->vm_mm != mm) + continue; + list_del_init(&priv->list); + + zap_vma_ptes(vma, vma->vm_start, + vma->vm_end - vma->vm_start); + + if (priv->entry) { + rdma_user_mmap_entry_put(priv->entry); + priv->entry = NULL; + } + } + mutex_unlock(&ufile->umap_lock); + mmap_read_unlock(mm); + mmput(mm); + } +} + +/* + * ib_uverbs_open() does not need the BKL: + * + * - the ib_uverbs_device structures are properly reference counted and + * everything else is purely local to the file being created, so + * races against other open calls are not a problem; + * - there is no ioctl method to race against; + * - the open method will either immediately run -ENXIO, or all + * required initialization will be done. + */ +static int ib_uverbs_open(struct inode *inode, struct file *filp) +{ + struct ib_uverbs_device *dev; + struct ib_uverbs_file *file; + struct ib_device *ib_dev; + int ret; + int module_dependent; + int srcu_key; + + dev = container_of(inode->i_cdev, struct ib_uverbs_device, cdev); + if (!atomic_inc_not_zero(&dev->refcount)) + return -ENXIO; + + get_device(&dev->dev); + srcu_key = srcu_read_lock(&dev->disassociate_srcu); + mutex_lock(&dev->lists_mutex); + ib_dev = srcu_dereference(dev->ib_dev, + &dev->disassociate_srcu); + if (!ib_dev) { + ret = -EIO; + goto err; + } + + if (!rdma_dev_access_netns(ib_dev, current->nsproxy->net_ns)) { + ret = -EPERM; + goto err; + } + + /* In case IB device supports disassociate ucontext, there is no hard + * dependency between uverbs device and its low level device. + */ + module_dependent = !(ib_dev->ops.disassociate_ucontext); + + if (module_dependent) { + if (!try_module_get(ib_dev->ops.owner)) { + ret = -ENODEV; + goto err; + } + } + + file = kzalloc(sizeof(*file), GFP_KERNEL); + if (!file) { + ret = -ENOMEM; + if (module_dependent) + goto err_module; + + goto err; + } + + file->device = dev; + kref_init(&file->ref); + mutex_init(&file->ucontext_lock); + + spin_lock_init(&file->uobjects_lock); + INIT_LIST_HEAD(&file->uobjects); + init_rwsem(&file->hw_destroy_rwsem); + mutex_init(&file->umap_lock); + INIT_LIST_HEAD(&file->umaps); + + filp->private_data = file; + list_add_tail(&file->list, &dev->uverbs_file_list); + mutex_unlock(&dev->lists_mutex); + srcu_read_unlock(&dev->disassociate_srcu, srcu_key); + + setup_ufile_idr_uobject(file); + + return stream_open(inode, filp); + +err_module: + module_put(ib_dev->ops.owner); + +err: + mutex_unlock(&dev->lists_mutex); + srcu_read_unlock(&dev->disassociate_srcu, srcu_key); + if (atomic_dec_and_test(&dev->refcount)) + ib_uverbs_comp_dev(dev); + + put_device(&dev->dev); + return ret; +} + +static int ib_uverbs_close(struct inode *inode, struct file *filp) +{ + struct ib_uverbs_file *file = filp->private_data; + + uverbs_destroy_ufile_hw(file, RDMA_REMOVE_CLOSE); + + mutex_lock(&file->device->lists_mutex); + list_del_init(&file->list); + mutex_unlock(&file->device->lists_mutex); + + kref_put(&file->ref, ib_uverbs_release_file); + + return 0; +} + +static const struct file_operations uverbs_fops = { + .owner = THIS_MODULE, + .write = ib_uverbs_write, + .open = ib_uverbs_open, + .release = ib_uverbs_close, + .llseek = no_llseek, + .unlocked_ioctl = ib_uverbs_ioctl, + .compat_ioctl = compat_ptr_ioctl, +}; + +static const struct file_operations uverbs_mmap_fops = { + .owner = THIS_MODULE, + .write = ib_uverbs_write, + .mmap = ib_uverbs_mmap, + .open = ib_uverbs_open, + .release = ib_uverbs_close, + .llseek = no_llseek, + .unlocked_ioctl = ib_uverbs_ioctl, + .compat_ioctl = compat_ptr_ioctl, +}; + +static int ib_uverbs_get_nl_info(struct ib_device *ibdev, void *client_data, + struct ib_client_nl_info *res) +{ + struct ib_uverbs_device *uverbs_dev = client_data; + int ret; + + if (res->port != -1) + return -EINVAL; + + res->abi = ibdev->ops.uverbs_abi_ver; + res->cdev = &uverbs_dev->dev; + + /* + * To support DRIVER_ID binding in userspace some of the driver need + * upgrading to expose their PCI dependent revision information + * through get_context instead of relying on modalias matching. When + * the drivers are fixed they can drop this flag. + */ + if (!ibdev->ops.uverbs_no_driver_id_binding) { + ret = nla_put_u32(res->nl_msg, RDMA_NLDEV_ATTR_UVERBS_DRIVER_ID, + ibdev->ops.driver_id); + if (ret) + return ret; + } + return 0; +} + +static struct ib_client uverbs_client = { + .name = "uverbs", + .no_kverbs_req = true, + .add = ib_uverbs_add_one, + .remove = ib_uverbs_remove_one, + .get_nl_info = ib_uverbs_get_nl_info, +}; +MODULE_ALIAS_RDMA_CLIENT("uverbs"); + +static ssize_t ibdev_show(struct device *device, struct device_attribute *attr, + char *buf) +{ + struct ib_uverbs_device *dev = + container_of(device, struct ib_uverbs_device, dev); + int ret = -ENODEV; + int srcu_key; + struct ib_device *ib_dev; + + srcu_key = srcu_read_lock(&dev->disassociate_srcu); + ib_dev = srcu_dereference(dev->ib_dev, &dev->disassociate_srcu); + if (ib_dev) + ret = sprintf(buf, "%s\n", dev_name(&ib_dev->dev)); + srcu_read_unlock(&dev->disassociate_srcu, srcu_key); + + return ret; +} +static DEVICE_ATTR_RO(ibdev); + +static ssize_t abi_version_show(struct device *device, + struct device_attribute *attr, char *buf) +{ + struct ib_uverbs_device *dev = + container_of(device, struct ib_uverbs_device, dev); + int ret = -ENODEV; + int srcu_key; + struct ib_device *ib_dev; + + srcu_key = srcu_read_lock(&dev->disassociate_srcu); + ib_dev = srcu_dereference(dev->ib_dev, &dev->disassociate_srcu); + if (ib_dev) + ret = sprintf(buf, "%u\n", ib_dev->ops.uverbs_abi_ver); + srcu_read_unlock(&dev->disassociate_srcu, srcu_key); + + return ret; +} +static DEVICE_ATTR_RO(abi_version); + +static struct attribute *ib_dev_attrs[] = { + &dev_attr_abi_version.attr, + &dev_attr_ibdev.attr, + NULL, +}; + +static const struct attribute_group dev_attr_group = { + .attrs = ib_dev_attrs, +}; + +static CLASS_ATTR_STRING(abi_version, S_IRUGO, + __stringify(IB_USER_VERBS_ABI_VERSION)); + +static int ib_uverbs_create_uapi(struct ib_device *device, + struct ib_uverbs_device *uverbs_dev) +{ + struct uverbs_api *uapi; + + uapi = uverbs_alloc_api(device); + if (IS_ERR(uapi)) + return PTR_ERR(uapi); + + uverbs_dev->uapi = uapi; + return 0; +} + +static int ib_uverbs_add_one(struct ib_device *device) +{ + int devnum; + dev_t base; + struct ib_uverbs_device *uverbs_dev; + int ret; + + if (!device->ops.alloc_ucontext) + return -EOPNOTSUPP; + + uverbs_dev = kzalloc(sizeof(*uverbs_dev), GFP_KERNEL); + if (!uverbs_dev) + return -ENOMEM; + + ret = init_srcu_struct(&uverbs_dev->disassociate_srcu); + if (ret) { + kfree(uverbs_dev); + return -ENOMEM; + } + + device_initialize(&uverbs_dev->dev); + uverbs_dev->dev.class = uverbs_class; + uverbs_dev->dev.parent = device->dev.parent; + uverbs_dev->dev.release = ib_uverbs_release_dev; + uverbs_dev->groups[0] = &dev_attr_group; + uverbs_dev->dev.groups = uverbs_dev->groups; + atomic_set(&uverbs_dev->refcount, 1); + init_completion(&uverbs_dev->comp); + uverbs_dev->xrcd_tree = RB_ROOT; + mutex_init(&uverbs_dev->xrcd_tree_mutex); + mutex_init(&uverbs_dev->lists_mutex); + INIT_LIST_HEAD(&uverbs_dev->uverbs_file_list); + rcu_assign_pointer(uverbs_dev->ib_dev, device); + uverbs_dev->num_comp_vectors = device->num_comp_vectors; + + devnum = ida_alloc_max(&uverbs_ida, IB_UVERBS_MAX_DEVICES - 1, + GFP_KERNEL); + if (devnum < 0) { + ret = -ENOMEM; + goto err; + } + uverbs_dev->devnum = devnum; + if (devnum >= IB_UVERBS_NUM_FIXED_MINOR) + base = dynamic_uverbs_dev + devnum - IB_UVERBS_NUM_FIXED_MINOR; + else + base = IB_UVERBS_BASE_DEV + devnum; + + ret = ib_uverbs_create_uapi(device, uverbs_dev); + if (ret) + goto err_uapi; + + uverbs_dev->dev.devt = base; + dev_set_name(&uverbs_dev->dev, "uverbs%d", uverbs_dev->devnum); + + cdev_init(&uverbs_dev->cdev, + device->ops.mmap ? &uverbs_mmap_fops : &uverbs_fops); + uverbs_dev->cdev.owner = THIS_MODULE; + + ret = cdev_device_add(&uverbs_dev->cdev, &uverbs_dev->dev); + if (ret) + goto err_uapi; + + ib_set_client_data(device, &uverbs_client, uverbs_dev); + return 0; + +err_uapi: + ida_free(&uverbs_ida, devnum); +err: + if (atomic_dec_and_test(&uverbs_dev->refcount)) + ib_uverbs_comp_dev(uverbs_dev); + wait_for_completion(&uverbs_dev->comp); + put_device(&uverbs_dev->dev); + return ret; +} + +static void ib_uverbs_free_hw_resources(struct ib_uverbs_device *uverbs_dev, + struct ib_device *ib_dev) +{ + struct ib_uverbs_file *file; + + /* Pending running commands to terminate */ + uverbs_disassociate_api_pre(uverbs_dev); + + mutex_lock(&uverbs_dev->lists_mutex); + while (!list_empty(&uverbs_dev->uverbs_file_list)) { + file = list_first_entry(&uverbs_dev->uverbs_file_list, + struct ib_uverbs_file, list); + list_del_init(&file->list); + kref_get(&file->ref); + + /* We must release the mutex before going ahead and calling + * uverbs_cleanup_ufile, as it might end up indirectly calling + * uverbs_close, for example due to freeing the resources (e.g + * mmput). + */ + mutex_unlock(&uverbs_dev->lists_mutex); + + uverbs_destroy_ufile_hw(file, RDMA_REMOVE_DRIVER_REMOVE); + kref_put(&file->ref, ib_uverbs_release_file); + + mutex_lock(&uverbs_dev->lists_mutex); + } + mutex_unlock(&uverbs_dev->lists_mutex); + + uverbs_disassociate_api(uverbs_dev->uapi); +} + +static void ib_uverbs_remove_one(struct ib_device *device, void *client_data) +{ + struct ib_uverbs_device *uverbs_dev = client_data; + int wait_clients = 1; + + cdev_device_del(&uverbs_dev->cdev, &uverbs_dev->dev); + ida_free(&uverbs_ida, uverbs_dev->devnum); + + if (device->ops.disassociate_ucontext) { + /* We disassociate HW resources and immediately return. + * Userspace will see a EIO errno for all future access. + * Upon returning, ib_device may be freed internally and is not + * valid any more. + * uverbs_device is still available until all clients close + * their files, then the uverbs device ref count will be zero + * and its resources will be freed. + * Note: At this point no more files can be opened since the + * cdev was deleted, however active clients can still issue + * commands and close their open files. + */ + ib_uverbs_free_hw_resources(uverbs_dev, device); + wait_clients = 0; + } + + if (atomic_dec_and_test(&uverbs_dev->refcount)) + ib_uverbs_comp_dev(uverbs_dev); + if (wait_clients) + wait_for_completion(&uverbs_dev->comp); + + put_device(&uverbs_dev->dev); +} + +static char *uverbs_devnode(struct device *dev, umode_t *mode) +{ + if (mode) + *mode = 0666; + return kasprintf(GFP_KERNEL, "infiniband/%s", dev_name(dev)); +} + +static int __init ib_uverbs_init(void) +{ + int ret; + + ret = register_chrdev_region(IB_UVERBS_BASE_DEV, + IB_UVERBS_NUM_FIXED_MINOR, + "infiniband_verbs"); + if (ret) { + pr_err("user_verbs: couldn't register device number\n"); + goto out; + } + + ret = alloc_chrdev_region(&dynamic_uverbs_dev, 0, + IB_UVERBS_NUM_DYNAMIC_MINOR, + "infiniband_verbs"); + if (ret) { + pr_err("couldn't register dynamic device number\n"); + goto out_alloc; + } + + uverbs_class = class_create(THIS_MODULE, "infiniband_verbs"); + if (IS_ERR(uverbs_class)) { + ret = PTR_ERR(uverbs_class); + pr_err("user_verbs: couldn't create class infiniband_verbs\n"); + goto out_chrdev; + } + + uverbs_class->devnode = uverbs_devnode; + + ret = class_create_file(uverbs_class, &class_attr_abi_version.attr); + if (ret) { + pr_err("user_verbs: couldn't create abi_version attribute\n"); + goto out_class; + } + + ret = ib_register_client(&uverbs_client); + if (ret) { + pr_err("user_verbs: couldn't register client\n"); + goto out_class; + } + + return 0; + +out_class: + class_destroy(uverbs_class); + +out_chrdev: + unregister_chrdev_region(dynamic_uverbs_dev, + IB_UVERBS_NUM_DYNAMIC_MINOR); + +out_alloc: + unregister_chrdev_region(IB_UVERBS_BASE_DEV, + IB_UVERBS_NUM_FIXED_MINOR); + +out: + return ret; +} + +static void __exit ib_uverbs_cleanup(void) +{ + ib_unregister_client(&uverbs_client); + class_destroy(uverbs_class); + unregister_chrdev_region(IB_UVERBS_BASE_DEV, + IB_UVERBS_NUM_FIXED_MINOR); + unregister_chrdev_region(dynamic_uverbs_dev, + IB_UVERBS_NUM_DYNAMIC_MINOR); + mmu_notifier_synchronize(); +} + +module_init(ib_uverbs_init); +module_exit(ib_uverbs_cleanup); diff --git a/drivers/infiniband/core/uverbs_marshall.c b/drivers/infiniband/core/uverbs_marshall.c new file mode 100644 index 000000000..11a080646 --- /dev/null +++ b/drivers/infiniband/core/uverbs_marshall.c @@ -0,0 +1,215 @@ +/* + * Copyright (c) 2005 Intel Corporation. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include <linux/export.h> +#include <rdma/ib_marshall.h> + +#define OPA_DEFAULT_GID_PREFIX cpu_to_be64(0xfe80000000000000ULL) +static int rdma_ah_conv_opa_to_ib(struct ib_device *dev, + struct rdma_ah_attr *ib, + struct rdma_ah_attr *opa) +{ + struct ib_port_attr port_attr; + int ret = 0; + + /* Do structure copy and the over-write fields */ + *ib = *opa; + + ib->type = RDMA_AH_ATTR_TYPE_IB; + rdma_ah_set_grh(ib, NULL, 0, 0, 1, 0); + + if (ib_query_port(dev, opa->port_num, &port_attr)) { + /* Set to default subnet to indicate error */ + rdma_ah_set_subnet_prefix(ib, OPA_DEFAULT_GID_PREFIX); + ret = -EINVAL; + } else { + rdma_ah_set_subnet_prefix(ib, + cpu_to_be64(port_attr.subnet_prefix)); + } + rdma_ah_set_interface_id(ib, OPA_MAKE_ID(rdma_ah_get_dlid(opa))); + return ret; +} + +void ib_copy_ah_attr_to_user(struct ib_device *device, + struct ib_uverbs_ah_attr *dst, + struct rdma_ah_attr *ah_attr) +{ + struct rdma_ah_attr *src = ah_attr; + struct rdma_ah_attr conv_ah; + + memset(&dst->grh, 0, sizeof(dst->grh)); + + if ((ah_attr->type == RDMA_AH_ATTR_TYPE_OPA) && + (rdma_ah_get_dlid(ah_attr) > be16_to_cpu(IB_LID_PERMISSIVE)) && + (!rdma_ah_conv_opa_to_ib(device, &conv_ah, ah_attr))) + src = &conv_ah; + + dst->dlid = rdma_ah_get_dlid(src); + dst->sl = rdma_ah_get_sl(src); + dst->src_path_bits = rdma_ah_get_path_bits(src); + dst->static_rate = rdma_ah_get_static_rate(src); + dst->is_global = rdma_ah_get_ah_flags(src) & + IB_AH_GRH ? 1 : 0; + if (dst->is_global) { + const struct ib_global_route *grh = rdma_ah_read_grh(src); + + memcpy(dst->grh.dgid, grh->dgid.raw, sizeof(grh->dgid)); + dst->grh.flow_label = grh->flow_label; + dst->grh.sgid_index = grh->sgid_index; + dst->grh.hop_limit = grh->hop_limit; + dst->grh.traffic_class = grh->traffic_class; + } + dst->port_num = rdma_ah_get_port_num(src); + dst->reserved = 0; +} +EXPORT_SYMBOL(ib_copy_ah_attr_to_user); + +void ib_copy_qp_attr_to_user(struct ib_device *device, + struct ib_uverbs_qp_attr *dst, + struct ib_qp_attr *src) +{ + dst->qp_state = src->qp_state; + dst->cur_qp_state = src->cur_qp_state; + dst->path_mtu = src->path_mtu; + dst->path_mig_state = src->path_mig_state; + dst->qkey = src->qkey; + dst->rq_psn = src->rq_psn; + dst->sq_psn = src->sq_psn; + dst->dest_qp_num = src->dest_qp_num; + dst->qp_access_flags = src->qp_access_flags; + + dst->max_send_wr = src->cap.max_send_wr; + dst->max_recv_wr = src->cap.max_recv_wr; + dst->max_send_sge = src->cap.max_send_sge; + dst->max_recv_sge = src->cap.max_recv_sge; + dst->max_inline_data = src->cap.max_inline_data; + + ib_copy_ah_attr_to_user(device, &dst->ah_attr, &src->ah_attr); + ib_copy_ah_attr_to_user(device, &dst->alt_ah_attr, &src->alt_ah_attr); + + dst->pkey_index = src->pkey_index; + dst->alt_pkey_index = src->alt_pkey_index; + dst->en_sqd_async_notify = src->en_sqd_async_notify; + dst->sq_draining = src->sq_draining; + dst->max_rd_atomic = src->max_rd_atomic; + dst->max_dest_rd_atomic = src->max_dest_rd_atomic; + dst->min_rnr_timer = src->min_rnr_timer; + dst->port_num = src->port_num; + dst->timeout = src->timeout; + dst->retry_cnt = src->retry_cnt; + dst->rnr_retry = src->rnr_retry; + dst->alt_port_num = src->alt_port_num; + dst->alt_timeout = src->alt_timeout; + memset(dst->reserved, 0, sizeof(dst->reserved)); +} +EXPORT_SYMBOL(ib_copy_qp_attr_to_user); + +static void __ib_copy_path_rec_to_user(struct ib_user_path_rec *dst, + struct sa_path_rec *src) +{ + memcpy(dst->dgid, src->dgid.raw, sizeof(src->dgid)); + memcpy(dst->sgid, src->sgid.raw, sizeof(src->sgid)); + + dst->dlid = htons(ntohl(sa_path_get_dlid(src))); + dst->slid = htons(ntohl(sa_path_get_slid(src))); + dst->raw_traffic = sa_path_get_raw_traffic(src); + dst->flow_label = src->flow_label; + dst->hop_limit = src->hop_limit; + dst->traffic_class = src->traffic_class; + dst->reversible = src->reversible; + dst->numb_path = src->numb_path; + dst->pkey = src->pkey; + dst->sl = src->sl; + dst->mtu_selector = src->mtu_selector; + dst->mtu = src->mtu; + dst->rate_selector = src->rate_selector; + dst->rate = src->rate; + dst->packet_life_time = src->packet_life_time; + dst->preference = src->preference; + dst->packet_life_time_selector = src->packet_life_time_selector; +} + +void ib_copy_path_rec_to_user(struct ib_user_path_rec *dst, + struct sa_path_rec *src) +{ + struct sa_path_rec rec; + + if (src->rec_type == SA_PATH_REC_TYPE_OPA) { + sa_convert_path_opa_to_ib(&rec, src); + __ib_copy_path_rec_to_user(dst, &rec); + return; + } + __ib_copy_path_rec_to_user(dst, src); +} +EXPORT_SYMBOL(ib_copy_path_rec_to_user); + +void ib_copy_path_rec_from_user(struct sa_path_rec *dst, + struct ib_user_path_rec *src) +{ + u32 slid, dlid; + + memset(dst, 0, sizeof(*dst)); + if ((ib_is_opa_gid((union ib_gid *)src->sgid)) || + (ib_is_opa_gid((union ib_gid *)src->dgid))) { + dst->rec_type = SA_PATH_REC_TYPE_OPA; + slid = opa_get_lid_from_gid((union ib_gid *)src->sgid); + dlid = opa_get_lid_from_gid((union ib_gid *)src->dgid); + } else { + dst->rec_type = SA_PATH_REC_TYPE_IB; + slid = ntohs(src->slid); + dlid = ntohs(src->dlid); + } + memcpy(dst->dgid.raw, src->dgid, sizeof dst->dgid); + memcpy(dst->sgid.raw, src->sgid, sizeof dst->sgid); + + sa_path_set_dlid(dst, dlid); + sa_path_set_slid(dst, slid); + sa_path_set_raw_traffic(dst, src->raw_traffic); + dst->flow_label = src->flow_label; + dst->hop_limit = src->hop_limit; + dst->traffic_class = src->traffic_class; + dst->reversible = src->reversible; + dst->numb_path = src->numb_path; + dst->pkey = src->pkey; + dst->sl = src->sl; + dst->mtu_selector = src->mtu_selector; + dst->mtu = src->mtu; + dst->rate_selector = src->rate_selector; + dst->rate = src->rate; + dst->packet_life_time = src->packet_life_time; + dst->preference = src->preference; + dst->packet_life_time_selector = src->packet_life_time_selector; + + /* TODO: No need to set this */ + sa_path_set_dmac_zero(dst); +} +EXPORT_SYMBOL(ib_copy_path_rec_from_user); diff --git a/drivers/infiniband/core/uverbs_std_types.c b/drivers/infiniband/core/uverbs_std_types.c new file mode 100644 index 000000000..0658101fc --- /dev/null +++ b/drivers/infiniband/core/uverbs_std_types.c @@ -0,0 +1,273 @@ +/* + * Copyright (c) 2017, Mellanox Technologies inc. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include <rdma/uverbs_std_types.h> +#include <rdma/ib_user_verbs.h> +#include <rdma/ib_verbs.h> +#include <linux/bug.h> +#include <linux/file.h> +#include <rdma/restrack.h> +#include "rdma_core.h" +#include "uverbs.h" + +static int uverbs_free_ah(struct ib_uobject *uobject, + enum rdma_remove_reason why, + struct uverbs_attr_bundle *attrs) +{ + return rdma_destroy_ah_user((struct ib_ah *)uobject->object, + RDMA_DESTROY_AH_SLEEPABLE, + &attrs->driver_udata); +} + +static int uverbs_free_flow(struct ib_uobject *uobject, + enum rdma_remove_reason why, + struct uverbs_attr_bundle *attrs) +{ + struct ib_flow *flow = (struct ib_flow *)uobject->object; + struct ib_uflow_object *uflow = + container_of(uobject, struct ib_uflow_object, uobject); + struct ib_qp *qp = flow->qp; + int ret; + + ret = flow->device->ops.destroy_flow(flow); + if (!ret) { + if (qp) + atomic_dec(&qp->usecnt); + ib_uverbs_flow_resources_free(uflow->resources); + } + + return ret; +} + +static int uverbs_free_mw(struct ib_uobject *uobject, + enum rdma_remove_reason why, + struct uverbs_attr_bundle *attrs) +{ + return uverbs_dealloc_mw((struct ib_mw *)uobject->object); +} + +static int uverbs_free_rwq_ind_tbl(struct ib_uobject *uobject, + enum rdma_remove_reason why, + struct uverbs_attr_bundle *attrs) +{ + struct ib_rwq_ind_table *rwq_ind_tbl = uobject->object; + struct ib_wq **ind_tbl = rwq_ind_tbl->ind_tbl; + u32 table_size = (1 << rwq_ind_tbl->log_ind_tbl_size); + int ret, i; + + if (atomic_read(&rwq_ind_tbl->usecnt)) + return -EBUSY; + + ret = rwq_ind_tbl->device->ops.destroy_rwq_ind_table(rwq_ind_tbl); + if (ib_is_destroy_retryable(ret, why, uobject)) + return ret; + + for (i = 0; i < table_size; i++) + atomic_dec(&ind_tbl[i]->usecnt); + + kfree(rwq_ind_tbl); + kfree(ind_tbl); + return ret; +} + +static int uverbs_free_xrcd(struct ib_uobject *uobject, + enum rdma_remove_reason why, + struct uverbs_attr_bundle *attrs) +{ + struct ib_xrcd *xrcd = uobject->object; + struct ib_uxrcd_object *uxrcd = + container_of(uobject, struct ib_uxrcd_object, uobject); + int ret; + + ret = ib_destroy_usecnt(&uxrcd->refcnt, why, uobject); + if (ret) + return ret; + + mutex_lock(&attrs->ufile->device->xrcd_tree_mutex); + ret = ib_uverbs_dealloc_xrcd(uobject, xrcd, why, attrs); + mutex_unlock(&attrs->ufile->device->xrcd_tree_mutex); + + return ret; +} + +static int uverbs_free_pd(struct ib_uobject *uobject, + enum rdma_remove_reason why, + struct uverbs_attr_bundle *attrs) +{ + struct ib_pd *pd = uobject->object; + int ret; + + ret = ib_destroy_usecnt(&pd->usecnt, why, uobject); + if (ret) + return ret; + + return ib_dealloc_pd_user(pd, &attrs->driver_udata); +} + +void ib_uverbs_free_event_queue(struct ib_uverbs_event_queue *event_queue) +{ + struct ib_uverbs_event *entry, *tmp; + + spin_lock_irq(&event_queue->lock); + /* + * The user must ensure that no new items are added to the event_list + * once is_closed is set. + */ + event_queue->is_closed = 1; + spin_unlock_irq(&event_queue->lock); + wake_up_interruptible(&event_queue->poll_wait); + kill_fasync(&event_queue->async_queue, SIGIO, POLL_IN); + + spin_lock_irq(&event_queue->lock); + list_for_each_entry_safe(entry, tmp, &event_queue->event_list, list) { + if (entry->counter) + list_del(&entry->obj_list); + list_del(&entry->list); + kfree(entry); + } + spin_unlock_irq(&event_queue->lock); +} + +static int +uverbs_completion_event_file_destroy_uobj(struct ib_uobject *uobj, + enum rdma_remove_reason why) +{ + struct ib_uverbs_completion_event_file *file = + container_of(uobj, struct ib_uverbs_completion_event_file, + uobj); + + ib_uverbs_free_event_queue(&file->ev_queue); + return 0; +} + +int uverbs_destroy_def_handler(struct uverbs_attr_bundle *attrs) +{ + return 0; +} +EXPORT_SYMBOL(uverbs_destroy_def_handler); + +DECLARE_UVERBS_NAMED_OBJECT( + UVERBS_OBJECT_COMP_CHANNEL, + UVERBS_TYPE_ALLOC_FD(sizeof(struct ib_uverbs_completion_event_file), + uverbs_completion_event_file_destroy_uobj, + &uverbs_event_fops, + "[infinibandevent]", + O_RDONLY)); + +DECLARE_UVERBS_NAMED_METHOD_DESTROY( + UVERBS_METHOD_MW_DESTROY, + UVERBS_ATTR_IDR(UVERBS_ATTR_DESTROY_MW_HANDLE, + UVERBS_OBJECT_MW, + UVERBS_ACCESS_DESTROY, + UA_MANDATORY)); + +DECLARE_UVERBS_NAMED_OBJECT(UVERBS_OBJECT_MW, + UVERBS_TYPE_ALLOC_IDR(uverbs_free_mw), + &UVERBS_METHOD(UVERBS_METHOD_MW_DESTROY)); + +DECLARE_UVERBS_NAMED_METHOD_DESTROY( + UVERBS_METHOD_AH_DESTROY, + UVERBS_ATTR_IDR(UVERBS_ATTR_DESTROY_AH_HANDLE, + UVERBS_OBJECT_AH, + UVERBS_ACCESS_DESTROY, + UA_MANDATORY)); + +DECLARE_UVERBS_NAMED_OBJECT(UVERBS_OBJECT_AH, + UVERBS_TYPE_ALLOC_IDR(uverbs_free_ah), + &UVERBS_METHOD(UVERBS_METHOD_AH_DESTROY)); + +DECLARE_UVERBS_NAMED_METHOD_DESTROY( + UVERBS_METHOD_FLOW_DESTROY, + UVERBS_ATTR_IDR(UVERBS_ATTR_DESTROY_FLOW_HANDLE, + UVERBS_OBJECT_FLOW, + UVERBS_ACCESS_DESTROY, + UA_MANDATORY)); + +DECLARE_UVERBS_NAMED_OBJECT( + UVERBS_OBJECT_FLOW, + UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_uflow_object), + uverbs_free_flow), + &UVERBS_METHOD(UVERBS_METHOD_FLOW_DESTROY)); + +DECLARE_UVERBS_NAMED_METHOD_DESTROY( + UVERBS_METHOD_RWQ_IND_TBL_DESTROY, + UVERBS_ATTR_IDR(UVERBS_ATTR_DESTROY_RWQ_IND_TBL_HANDLE, + UVERBS_OBJECT_RWQ_IND_TBL, + UVERBS_ACCESS_DESTROY, + UA_MANDATORY)); + +DECLARE_UVERBS_NAMED_OBJECT(UVERBS_OBJECT_RWQ_IND_TBL, + UVERBS_TYPE_ALLOC_IDR(uverbs_free_rwq_ind_tbl), + &UVERBS_METHOD(UVERBS_METHOD_RWQ_IND_TBL_DESTROY)); + +DECLARE_UVERBS_NAMED_METHOD_DESTROY( + UVERBS_METHOD_XRCD_DESTROY, + UVERBS_ATTR_IDR(UVERBS_ATTR_DESTROY_XRCD_HANDLE, + UVERBS_OBJECT_XRCD, + UVERBS_ACCESS_DESTROY, + UA_MANDATORY)); + +DECLARE_UVERBS_NAMED_OBJECT( + UVERBS_OBJECT_XRCD, + UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_uxrcd_object), + uverbs_free_xrcd), + &UVERBS_METHOD(UVERBS_METHOD_XRCD_DESTROY)); + +DECLARE_UVERBS_NAMED_METHOD_DESTROY( + UVERBS_METHOD_PD_DESTROY, + UVERBS_ATTR_IDR(UVERBS_ATTR_DESTROY_PD_HANDLE, + UVERBS_OBJECT_PD, + UVERBS_ACCESS_DESTROY, + UA_MANDATORY)); + +DECLARE_UVERBS_NAMED_OBJECT(UVERBS_OBJECT_PD, + UVERBS_TYPE_ALLOC_IDR(uverbs_free_pd), + &UVERBS_METHOD(UVERBS_METHOD_PD_DESTROY)); + +const struct uapi_definition uverbs_def_obj_intf[] = { + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_PD, + UAPI_DEF_OBJ_NEEDS_FN(dealloc_pd)), + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_COMP_CHANNEL, + UAPI_DEF_OBJ_NEEDS_FN(dealloc_pd)), + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_AH, + UAPI_DEF_OBJ_NEEDS_FN(destroy_ah)), + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_MW, + UAPI_DEF_OBJ_NEEDS_FN(dealloc_mw)), + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_FLOW, + UAPI_DEF_OBJ_NEEDS_FN(destroy_flow)), + UAPI_DEF_CHAIN_OBJ_TREE_NAMED( + UVERBS_OBJECT_RWQ_IND_TBL, + UAPI_DEF_OBJ_NEEDS_FN(destroy_rwq_ind_table)), + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_XRCD, + UAPI_DEF_OBJ_NEEDS_FN(dealloc_xrcd)), + {} +}; diff --git a/drivers/infiniband/core/uverbs_std_types_async_fd.c b/drivers/infiniband/core/uverbs_std_types_async_fd.c new file mode 100644 index 000000000..61899eaf1 --- /dev/null +++ b/drivers/infiniband/core/uverbs_std_types_async_fd.c @@ -0,0 +1,80 @@ +// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB +/* + * Copyright (c) 2019, Mellanox Technologies inc. All rights reserved. + */ + +#include <rdma/uverbs_std_types.h> +#include <rdma/uverbs_ioctl.h> +#include "rdma_core.h" +#include "uverbs.h" + +static int UVERBS_HANDLER(UVERBS_METHOD_ASYNC_EVENT_ALLOC)( + struct uverbs_attr_bundle *attrs) +{ + struct ib_uobject *uobj = + uverbs_attr_get_uobject(attrs, UVERBS_METHOD_ASYNC_EVENT_ALLOC); + + ib_uverbs_init_async_event_file( + container_of(uobj, struct ib_uverbs_async_event_file, uobj)); + return 0; +} + +static int uverbs_async_event_destroy_uobj(struct ib_uobject *uobj, + enum rdma_remove_reason why) +{ + struct ib_uverbs_async_event_file *event_file = + container_of(uobj, struct ib_uverbs_async_event_file, uobj); + + ib_unregister_event_handler(&event_file->event_handler); + + if (why == RDMA_REMOVE_DRIVER_REMOVE) + ib_uverbs_async_handler(event_file, 0, IB_EVENT_DEVICE_FATAL, + NULL, NULL); + return 0; +} + +int uverbs_async_event_release(struct inode *inode, struct file *filp) +{ + struct ib_uverbs_async_event_file *event_file; + struct ib_uobject *uobj = filp->private_data; + int ret; + + if (!uobj) + return uverbs_uobject_fd_release(inode, filp); + + event_file = + container_of(uobj, struct ib_uverbs_async_event_file, uobj); + + /* + * The async event FD has to deliver IB_EVENT_DEVICE_FATAL even after + * disassociation, so cleaning the event list must only happen after + * release. The user knows it has reached the end of the event stream + * when it sees IB_EVENT_DEVICE_FATAL. + */ + uverbs_uobject_get(uobj); + ret = uverbs_uobject_fd_release(inode, filp); + ib_uverbs_free_event_queue(&event_file->ev_queue); + uverbs_uobject_put(uobj); + return ret; +} + +DECLARE_UVERBS_NAMED_METHOD( + UVERBS_METHOD_ASYNC_EVENT_ALLOC, + UVERBS_ATTR_FD(UVERBS_ATTR_ASYNC_EVENT_ALLOC_FD_HANDLE, + UVERBS_OBJECT_ASYNC_EVENT, + UVERBS_ACCESS_NEW, + UA_MANDATORY)); + +DECLARE_UVERBS_NAMED_OBJECT( + UVERBS_OBJECT_ASYNC_EVENT, + UVERBS_TYPE_ALLOC_FD(sizeof(struct ib_uverbs_async_event_file), + uverbs_async_event_destroy_uobj, + &uverbs_async_event_fops, + "[infinibandevent]", + O_RDONLY), + &UVERBS_METHOD(UVERBS_METHOD_ASYNC_EVENT_ALLOC)); + +const struct uapi_definition uverbs_def_obj_async_fd[] = { + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_ASYNC_EVENT), + {} +}; diff --git a/drivers/infiniband/core/uverbs_std_types_counters.c b/drivers/infiniband/core/uverbs_std_types_counters.c new file mode 100644 index 000000000..c61b10fbf --- /dev/null +++ b/drivers/infiniband/core/uverbs_std_types_counters.c @@ -0,0 +1,164 @@ +// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB +/* + * Copyright (c) 2018, Mellanox Technologies inc. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include "rdma_core.h" +#include "uverbs.h" +#include <rdma/uverbs_std_types.h> + +static int uverbs_free_counters(struct ib_uobject *uobject, + enum rdma_remove_reason why, + struct uverbs_attr_bundle *attrs) +{ + struct ib_counters *counters = uobject->object; + int ret; + + ret = ib_destroy_usecnt(&counters->usecnt, why, uobject); + if (ret) + return ret; + + ret = counters->device->ops.destroy_counters(counters); + if (ret) + return ret; + kfree(counters); + return 0; +} + +static int UVERBS_HANDLER(UVERBS_METHOD_COUNTERS_CREATE)( + struct uverbs_attr_bundle *attrs) +{ + struct ib_uobject *uobj = uverbs_attr_get_uobject( + attrs, UVERBS_ATTR_CREATE_COUNTERS_HANDLE); + struct ib_device *ib_dev = attrs->context->device; + struct ib_counters *counters; + int ret; + + /* + * This check should be removed once the infrastructure + * have the ability to remove methods from parse tree once + * such condition is met. + */ + if (!ib_dev->ops.create_counters) + return -EOPNOTSUPP; + + counters = rdma_zalloc_drv_obj(ib_dev, ib_counters); + if (!counters) + return -ENOMEM; + + counters->device = ib_dev; + counters->uobject = uobj; + uobj->object = counters; + atomic_set(&counters->usecnt, 0); + + ret = ib_dev->ops.create_counters(counters, attrs); + if (ret) + kfree(counters); + + return ret; +} + +static int UVERBS_HANDLER(UVERBS_METHOD_COUNTERS_READ)( + struct uverbs_attr_bundle *attrs) +{ + struct ib_counters_read_attr read_attr = {}; + const struct uverbs_attr *uattr; + struct ib_counters *counters = + uverbs_attr_get_obj(attrs, UVERBS_ATTR_READ_COUNTERS_HANDLE); + int ret; + + if (!counters->device->ops.read_counters) + return -EOPNOTSUPP; + + if (!atomic_read(&counters->usecnt)) + return -EINVAL; + + ret = uverbs_get_flags32(&read_attr.flags, attrs, + UVERBS_ATTR_READ_COUNTERS_FLAGS, + IB_UVERBS_READ_COUNTERS_PREFER_CACHED); + if (ret) + return ret; + + uattr = uverbs_attr_get(attrs, UVERBS_ATTR_READ_COUNTERS_BUFF); + if (IS_ERR(uattr)) + return PTR_ERR(uattr); + read_attr.ncounters = uattr->ptr_attr.len / sizeof(u64); + read_attr.counters_buff = uverbs_zalloc( + attrs, array_size(read_attr.ncounters, sizeof(u64))); + if (IS_ERR(read_attr.counters_buff)) + return PTR_ERR(read_attr.counters_buff); + + ret = counters->device->ops.read_counters(counters, &read_attr, attrs); + if (ret) + return ret; + + return uverbs_copy_to(attrs, UVERBS_ATTR_READ_COUNTERS_BUFF, + read_attr.counters_buff, + read_attr.ncounters * sizeof(u64)); +} + +DECLARE_UVERBS_NAMED_METHOD( + UVERBS_METHOD_COUNTERS_CREATE, + UVERBS_ATTR_IDR(UVERBS_ATTR_CREATE_COUNTERS_HANDLE, + UVERBS_OBJECT_COUNTERS, + UVERBS_ACCESS_NEW, + UA_MANDATORY)); + +DECLARE_UVERBS_NAMED_METHOD_DESTROY( + UVERBS_METHOD_COUNTERS_DESTROY, + UVERBS_ATTR_IDR(UVERBS_ATTR_DESTROY_COUNTERS_HANDLE, + UVERBS_OBJECT_COUNTERS, + UVERBS_ACCESS_DESTROY, + UA_MANDATORY)); + +DECLARE_UVERBS_NAMED_METHOD( + UVERBS_METHOD_COUNTERS_READ, + UVERBS_ATTR_IDR(UVERBS_ATTR_READ_COUNTERS_HANDLE, + UVERBS_OBJECT_COUNTERS, + UVERBS_ACCESS_READ, + UA_MANDATORY), + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_READ_COUNTERS_BUFF, + UVERBS_ATTR_MIN_SIZE(0), + UA_MANDATORY), + UVERBS_ATTR_FLAGS_IN(UVERBS_ATTR_READ_COUNTERS_FLAGS, + enum ib_uverbs_read_counters_flags)); + +DECLARE_UVERBS_NAMED_OBJECT(UVERBS_OBJECT_COUNTERS, + UVERBS_TYPE_ALLOC_IDR(uverbs_free_counters), + &UVERBS_METHOD(UVERBS_METHOD_COUNTERS_CREATE), + &UVERBS_METHOD(UVERBS_METHOD_COUNTERS_DESTROY), + &UVERBS_METHOD(UVERBS_METHOD_COUNTERS_READ)); + +const struct uapi_definition uverbs_def_obj_counters[] = { + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_COUNTERS, + UAPI_DEF_OBJ_NEEDS_FN(destroy_counters)), + {} +}; diff --git a/drivers/infiniband/core/uverbs_std_types_cq.c b/drivers/infiniband/core/uverbs_std_types_cq.c new file mode 100644 index 000000000..8dabd0598 --- /dev/null +++ b/drivers/infiniband/core/uverbs_std_types_cq.c @@ -0,0 +1,222 @@ +/* + * Copyright (c) 2017, Mellanox Technologies inc. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include <rdma/uverbs_std_types.h> +#include "rdma_core.h" +#include "uverbs.h" +#include "restrack.h" + +static int uverbs_free_cq(struct ib_uobject *uobject, + enum rdma_remove_reason why, + struct uverbs_attr_bundle *attrs) +{ + struct ib_cq *cq = uobject->object; + struct ib_uverbs_event_queue *ev_queue = cq->cq_context; + struct ib_ucq_object *ucq = + container_of(uobject, struct ib_ucq_object, uevent.uobject); + int ret; + + ret = ib_destroy_cq_user(cq, &attrs->driver_udata); + if (ib_is_destroy_retryable(ret, why, uobject)) + return ret; + + ib_uverbs_release_ucq( + ev_queue ? container_of(ev_queue, + struct ib_uverbs_completion_event_file, + ev_queue) : + NULL, + ucq); + return ret; +} + +static int UVERBS_HANDLER(UVERBS_METHOD_CQ_CREATE)( + struct uverbs_attr_bundle *attrs) +{ + struct ib_ucq_object *obj = container_of( + uverbs_attr_get_uobject(attrs, UVERBS_ATTR_CREATE_CQ_HANDLE), + typeof(*obj), uevent.uobject); + struct ib_device *ib_dev = attrs->context->device; + int ret; + u64 user_handle; + struct ib_cq_init_attr attr = {}; + struct ib_cq *cq; + struct ib_uverbs_completion_event_file *ev_file = NULL; + struct ib_uobject *ev_file_uobj; + + if (!ib_dev->ops.create_cq || !ib_dev->ops.destroy_cq) + return -EOPNOTSUPP; + + ret = uverbs_copy_from(&attr.comp_vector, attrs, + UVERBS_ATTR_CREATE_CQ_COMP_VECTOR); + if (!ret) + ret = uverbs_copy_from(&attr.cqe, attrs, + UVERBS_ATTR_CREATE_CQ_CQE); + if (!ret) + ret = uverbs_copy_from(&user_handle, attrs, + UVERBS_ATTR_CREATE_CQ_USER_HANDLE); + if (ret) + return ret; + + ret = uverbs_get_flags32(&attr.flags, attrs, + UVERBS_ATTR_CREATE_CQ_FLAGS, + IB_UVERBS_CQ_FLAGS_TIMESTAMP_COMPLETION | + IB_UVERBS_CQ_FLAGS_IGNORE_OVERRUN); + if (ret) + return ret; + + ev_file_uobj = uverbs_attr_get_uobject(attrs, UVERBS_ATTR_CREATE_CQ_COMP_CHANNEL); + if (!IS_ERR(ev_file_uobj)) { + ev_file = container_of(ev_file_uobj, + struct ib_uverbs_completion_event_file, + uobj); + uverbs_uobject_get(ev_file_uobj); + } + + obj->uevent.event_file = ib_uverbs_get_async_event( + attrs, UVERBS_ATTR_CREATE_CQ_EVENT_FD); + + if (attr.comp_vector >= attrs->ufile->device->num_comp_vectors) { + ret = -EINVAL; + goto err_event_file; + } + + INIT_LIST_HEAD(&obj->comp_list); + INIT_LIST_HEAD(&obj->uevent.event_list); + + cq = rdma_zalloc_drv_obj(ib_dev, ib_cq); + if (!cq) { + ret = -ENOMEM; + goto err_event_file; + } + + cq->device = ib_dev; + cq->uobject = obj; + cq->comp_handler = ib_uverbs_comp_handler; + cq->event_handler = ib_uverbs_cq_event_handler; + cq->cq_context = ev_file ? &ev_file->ev_queue : NULL; + atomic_set(&cq->usecnt, 0); + + rdma_restrack_new(&cq->res, RDMA_RESTRACK_CQ); + rdma_restrack_set_name(&cq->res, NULL); + + ret = ib_dev->ops.create_cq(cq, &attr, &attrs->driver_udata); + if (ret) + goto err_free; + + obj->uevent.uobject.object = cq; + obj->uevent.uobject.user_handle = user_handle; + rdma_restrack_add(&cq->res); + uverbs_finalize_uobj_create(attrs, UVERBS_ATTR_CREATE_CQ_HANDLE); + + ret = uverbs_copy_to(attrs, UVERBS_ATTR_CREATE_CQ_RESP_CQE, &cq->cqe, + sizeof(cq->cqe)); + return ret; + +err_free: + rdma_restrack_put(&cq->res); + kfree(cq); +err_event_file: + if (obj->uevent.event_file) + uverbs_uobject_put(&obj->uevent.event_file->uobj); + if (ev_file) + uverbs_uobject_put(ev_file_uobj); + return ret; +}; + +DECLARE_UVERBS_NAMED_METHOD( + UVERBS_METHOD_CQ_CREATE, + UVERBS_ATTR_IDR(UVERBS_ATTR_CREATE_CQ_HANDLE, + UVERBS_OBJECT_CQ, + UVERBS_ACCESS_NEW, + UA_MANDATORY), + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_CREATE_CQ_CQE, + UVERBS_ATTR_TYPE(u32), + UA_MANDATORY), + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_CREATE_CQ_USER_HANDLE, + UVERBS_ATTR_TYPE(u64), + UA_MANDATORY), + UVERBS_ATTR_FD(UVERBS_ATTR_CREATE_CQ_COMP_CHANNEL, + UVERBS_OBJECT_COMP_CHANNEL, + UVERBS_ACCESS_READ, + UA_OPTIONAL), + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_CREATE_CQ_COMP_VECTOR, + UVERBS_ATTR_TYPE(u32), + UA_MANDATORY), + UVERBS_ATTR_FLAGS_IN(UVERBS_ATTR_CREATE_CQ_FLAGS, + enum ib_uverbs_ex_create_cq_flags), + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_CREATE_CQ_RESP_CQE, + UVERBS_ATTR_TYPE(u32), + UA_MANDATORY), + UVERBS_ATTR_FD(UVERBS_ATTR_CREATE_CQ_EVENT_FD, + UVERBS_OBJECT_ASYNC_EVENT, + UVERBS_ACCESS_READ, + UA_OPTIONAL), + UVERBS_ATTR_UHW()); + +static int UVERBS_HANDLER(UVERBS_METHOD_CQ_DESTROY)( + struct uverbs_attr_bundle *attrs) +{ + struct ib_uobject *uobj = + uverbs_attr_get_uobject(attrs, UVERBS_ATTR_DESTROY_CQ_HANDLE); + struct ib_ucq_object *obj = + container_of(uobj, struct ib_ucq_object, uevent.uobject); + struct ib_uverbs_destroy_cq_resp resp = { + .comp_events_reported = obj->comp_events_reported, + .async_events_reported = obj->uevent.events_reported + }; + + return uverbs_copy_to(attrs, UVERBS_ATTR_DESTROY_CQ_RESP, &resp, + sizeof(resp)); +} + +DECLARE_UVERBS_NAMED_METHOD( + UVERBS_METHOD_CQ_DESTROY, + UVERBS_ATTR_IDR(UVERBS_ATTR_DESTROY_CQ_HANDLE, + UVERBS_OBJECT_CQ, + UVERBS_ACCESS_DESTROY, + UA_MANDATORY), + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_DESTROY_CQ_RESP, + UVERBS_ATTR_TYPE(struct ib_uverbs_destroy_cq_resp), + UA_MANDATORY)); + +DECLARE_UVERBS_NAMED_OBJECT( + UVERBS_OBJECT_CQ, + UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_ucq_object), uverbs_free_cq), + &UVERBS_METHOD(UVERBS_METHOD_CQ_CREATE), + &UVERBS_METHOD(UVERBS_METHOD_CQ_DESTROY) +); + +const struct uapi_definition uverbs_def_obj_cq[] = { + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_CQ, + UAPI_DEF_OBJ_NEEDS_FN(destroy_cq)), + {} +}; diff --git a/drivers/infiniband/core/uverbs_std_types_device.c b/drivers/infiniband/core/uverbs_std_types_device.c new file mode 100644 index 000000000..049684880 --- /dev/null +++ b/drivers/infiniband/core/uverbs_std_types_device.c @@ -0,0 +1,503 @@ +// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB +/* + * Copyright (c) 2018, Mellanox Technologies inc. All rights reserved. + */ + +#include <linux/overflow.h> +#include <rdma/uverbs_std_types.h> +#include "rdma_core.h" +#include "uverbs.h" +#include <rdma/uverbs_ioctl.h> +#include <rdma/opa_addr.h> +#include <rdma/ib_cache.h> + +/* + * This ioctl method allows calling any defined write or write_ex + * handler. This essentially replaces the hdr/ex_hdr system with the ioctl + * marshalling, and brings the non-ex path into the same marshalling as the ex + * path. + */ +static int UVERBS_HANDLER(UVERBS_METHOD_INVOKE_WRITE)( + struct uverbs_attr_bundle *attrs) +{ + struct uverbs_api *uapi = attrs->ufile->device->uapi; + const struct uverbs_api_write_method *method_elm; + u32 cmd; + int rc; + + rc = uverbs_get_const(&cmd, attrs, UVERBS_ATTR_WRITE_CMD); + if (rc) + return rc; + + method_elm = uapi_get_method(uapi, cmd); + if (IS_ERR(method_elm)) + return PTR_ERR(method_elm); + + uverbs_fill_udata(attrs, &attrs->ucore, UVERBS_ATTR_CORE_IN, + UVERBS_ATTR_CORE_OUT); + + if (attrs->ucore.inlen < method_elm->req_size || + attrs->ucore.outlen < method_elm->resp_size) + return -ENOSPC; + + attrs->uobject = NULL; + rc = method_elm->handler(attrs); + if (attrs->uobject) + uverbs_finalize_object(attrs->uobject, UVERBS_ACCESS_NEW, true, + !rc, attrs); + return rc; +} + +DECLARE_UVERBS_NAMED_METHOD(UVERBS_METHOD_INVOKE_WRITE, + UVERBS_ATTR_CONST_IN(UVERBS_ATTR_WRITE_CMD, + enum ib_uverbs_write_cmds, + UA_MANDATORY), + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_CORE_IN, + UVERBS_ATTR_MIN_SIZE(sizeof(u32)), + UA_OPTIONAL), + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_CORE_OUT, + UVERBS_ATTR_MIN_SIZE(0), + UA_OPTIONAL), + UVERBS_ATTR_UHW()); + +static uint32_t * +gather_objects_handle(struct ib_uverbs_file *ufile, + const struct uverbs_api_object *uapi_object, + struct uverbs_attr_bundle *attrs, + ssize_t out_len, + u64 *total) +{ + u64 max_count = out_len / sizeof(u32); + struct ib_uobject *obj; + u64 count = 0; + u32 *handles; + + /* Allocated memory that cannot page out where we gather + * all object ids under a spin_lock. + */ + handles = uverbs_zalloc(attrs, out_len); + if (IS_ERR(handles)) + return handles; + + spin_lock_irq(&ufile->uobjects_lock); + list_for_each_entry(obj, &ufile->uobjects, list) { + u32 obj_id = obj->id; + + if (obj->uapi_object != uapi_object) + continue; + + if (count >= max_count) + break; + + handles[count] = obj_id; + count++; + } + spin_unlock_irq(&ufile->uobjects_lock); + + *total = count; + return handles; +} + +static int UVERBS_HANDLER(UVERBS_METHOD_INFO_HANDLES)( + struct uverbs_attr_bundle *attrs) +{ + const struct uverbs_api_object *uapi_object; + ssize_t out_len; + u64 total = 0; + u16 object_id; + u32 *handles; + int ret; + + out_len = uverbs_attr_get_len(attrs, UVERBS_ATTR_INFO_HANDLES_LIST); + if (out_len <= 0 || (out_len % sizeof(u32) != 0)) + return -EINVAL; + + ret = uverbs_get_const(&object_id, attrs, UVERBS_ATTR_INFO_OBJECT_ID); + if (ret) + return ret; + + uapi_object = uapi_get_object(attrs->ufile->device->uapi, object_id); + if (IS_ERR(uapi_object)) + return PTR_ERR(uapi_object); + + handles = gather_objects_handle(attrs->ufile, uapi_object, attrs, + out_len, &total); + if (IS_ERR(handles)) + return PTR_ERR(handles); + + ret = uverbs_copy_to(attrs, UVERBS_ATTR_INFO_HANDLES_LIST, handles, + sizeof(u32) * total); + if (ret) + goto err; + + ret = uverbs_copy_to(attrs, UVERBS_ATTR_INFO_TOTAL_HANDLES, &total, + sizeof(total)); +err: + return ret; +} + +void copy_port_attr_to_resp(struct ib_port_attr *attr, + struct ib_uverbs_query_port_resp *resp, + struct ib_device *ib_dev, u8 port_num) +{ + resp->state = attr->state; + resp->max_mtu = attr->max_mtu; + resp->active_mtu = attr->active_mtu; + resp->gid_tbl_len = attr->gid_tbl_len; + resp->port_cap_flags = make_port_cap_flags(attr); + resp->max_msg_sz = attr->max_msg_sz; + resp->bad_pkey_cntr = attr->bad_pkey_cntr; + resp->qkey_viol_cntr = attr->qkey_viol_cntr; + resp->pkey_tbl_len = attr->pkey_tbl_len; + + if (rdma_is_grh_required(ib_dev, port_num)) + resp->flags |= IB_UVERBS_QPF_GRH_REQUIRED; + + if (rdma_cap_opa_ah(ib_dev, port_num)) { + resp->lid = OPA_TO_IB_UCAST_LID(attr->lid); + resp->sm_lid = OPA_TO_IB_UCAST_LID(attr->sm_lid); + } else { + resp->lid = ib_lid_cpu16(attr->lid); + resp->sm_lid = ib_lid_cpu16(attr->sm_lid); + } + + resp->lmc = attr->lmc; + resp->max_vl_num = attr->max_vl_num; + resp->sm_sl = attr->sm_sl; + resp->subnet_timeout = attr->subnet_timeout; + resp->init_type_reply = attr->init_type_reply; + resp->active_width = attr->active_width; + /* This ABI needs to be extended to provide any speed more than IB_SPEED_NDR */ + resp->active_speed = min_t(u16, attr->active_speed, IB_SPEED_NDR); + resp->phys_state = attr->phys_state; + resp->link_layer = rdma_port_get_link_layer(ib_dev, port_num); +} + +static int UVERBS_HANDLER(UVERBS_METHOD_QUERY_PORT)( + struct uverbs_attr_bundle *attrs) +{ + struct ib_device *ib_dev; + struct ib_port_attr attr = {}; + struct ib_uverbs_query_port_resp_ex resp = {}; + struct ib_ucontext *ucontext; + int ret; + u8 port_num; + + ucontext = ib_uverbs_get_ucontext(attrs); + if (IS_ERR(ucontext)) + return PTR_ERR(ucontext); + ib_dev = ucontext->device; + + /* FIXME: Extend the UAPI_DEF_OBJ_NEEDS_FN stuff.. */ + if (!ib_dev->ops.query_port) + return -EOPNOTSUPP; + + ret = uverbs_get_const(&port_num, attrs, + UVERBS_ATTR_QUERY_PORT_PORT_NUM); + if (ret) + return ret; + + ret = ib_query_port(ib_dev, port_num, &attr); + if (ret) + return ret; + + copy_port_attr_to_resp(&attr, &resp.legacy_resp, ib_dev, port_num); + resp.port_cap_flags2 = attr.port_cap_flags2; + + return uverbs_copy_to_struct_or_zero(attrs, UVERBS_ATTR_QUERY_PORT_RESP, + &resp, sizeof(resp)); +} + +static int UVERBS_HANDLER(UVERBS_METHOD_GET_CONTEXT)( + struct uverbs_attr_bundle *attrs) +{ + u32 num_comp = attrs->ufile->device->num_comp_vectors; + u64 core_support = IB_UVERBS_CORE_SUPPORT_OPTIONAL_MR_ACCESS; + int ret; + + ret = uverbs_copy_to(attrs, UVERBS_ATTR_GET_CONTEXT_NUM_COMP_VECTORS, + &num_comp, sizeof(num_comp)); + if (IS_UVERBS_COPY_ERR(ret)) + return ret; + + ret = uverbs_copy_to(attrs, UVERBS_ATTR_GET_CONTEXT_CORE_SUPPORT, + &core_support, sizeof(core_support)); + if (IS_UVERBS_COPY_ERR(ret)) + return ret; + + ret = ib_alloc_ucontext(attrs); + if (ret) + return ret; + ret = ib_init_ucontext(attrs); + if (ret) { + kfree(attrs->context); + attrs->context = NULL; + return ret; + } + return 0; +} + +static int UVERBS_HANDLER(UVERBS_METHOD_QUERY_CONTEXT)( + struct uverbs_attr_bundle *attrs) +{ + u64 core_support = IB_UVERBS_CORE_SUPPORT_OPTIONAL_MR_ACCESS; + struct ib_ucontext *ucontext; + struct ib_device *ib_dev; + u32 num_comp; + int ret; + + ucontext = ib_uverbs_get_ucontext(attrs); + if (IS_ERR(ucontext)) + return PTR_ERR(ucontext); + ib_dev = ucontext->device; + + if (!ib_dev->ops.query_ucontext) + return -EOPNOTSUPP; + + num_comp = attrs->ufile->device->num_comp_vectors; + ret = uverbs_copy_to(attrs, UVERBS_ATTR_QUERY_CONTEXT_NUM_COMP_VECTORS, + &num_comp, sizeof(num_comp)); + if (IS_UVERBS_COPY_ERR(ret)) + return ret; + + ret = uverbs_copy_to(attrs, UVERBS_ATTR_QUERY_CONTEXT_CORE_SUPPORT, + &core_support, sizeof(core_support)); + if (IS_UVERBS_COPY_ERR(ret)) + return ret; + + return ucontext->device->ops.query_ucontext(ucontext, attrs); +} + +static int copy_gid_entries_to_user(struct uverbs_attr_bundle *attrs, + struct ib_uverbs_gid_entry *entries, + size_t num_entries, size_t user_entry_size) +{ + const struct uverbs_attr *attr; + void __user *user_entries; + size_t copy_len; + int ret; + int i; + + if (user_entry_size == sizeof(*entries)) { + ret = uverbs_copy_to(attrs, + UVERBS_ATTR_QUERY_GID_TABLE_RESP_ENTRIES, + entries, sizeof(*entries) * num_entries); + return ret; + } + + copy_len = min_t(size_t, user_entry_size, sizeof(*entries)); + attr = uverbs_attr_get(attrs, UVERBS_ATTR_QUERY_GID_TABLE_RESP_ENTRIES); + if (IS_ERR(attr)) + return PTR_ERR(attr); + + user_entries = u64_to_user_ptr(attr->ptr_attr.data); + for (i = 0; i < num_entries; i++) { + if (copy_to_user(user_entries, entries, copy_len)) + return -EFAULT; + + if (user_entry_size > sizeof(*entries)) { + if (clear_user(user_entries + sizeof(*entries), + user_entry_size - sizeof(*entries))) + return -EFAULT; + } + + entries++; + user_entries += user_entry_size; + } + + return uverbs_output_written(attrs, + UVERBS_ATTR_QUERY_GID_TABLE_RESP_ENTRIES); +} + +static int UVERBS_HANDLER(UVERBS_METHOD_QUERY_GID_TABLE)( + struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_gid_entry *entries; + struct ib_ucontext *ucontext; + struct ib_device *ib_dev; + size_t user_entry_size; + ssize_t num_entries; + int max_entries; + u32 flags; + int ret; + + ret = uverbs_get_flags32(&flags, attrs, + UVERBS_ATTR_QUERY_GID_TABLE_FLAGS, 0); + if (ret) + return ret; + + ret = uverbs_get_const(&user_entry_size, attrs, + UVERBS_ATTR_QUERY_GID_TABLE_ENTRY_SIZE); + if (ret) + return ret; + + if (!user_entry_size) + return -EINVAL; + + max_entries = uverbs_attr_ptr_get_array_size( + attrs, UVERBS_ATTR_QUERY_GID_TABLE_RESP_ENTRIES, + user_entry_size); + if (max_entries <= 0) + return max_entries ?: -EINVAL; + + ucontext = ib_uverbs_get_ucontext(attrs); + if (IS_ERR(ucontext)) + return PTR_ERR(ucontext); + ib_dev = ucontext->device; + + entries = uverbs_kcalloc(attrs, max_entries, sizeof(*entries)); + if (IS_ERR(entries)) + return PTR_ERR(entries); + + num_entries = rdma_query_gid_table(ib_dev, entries, max_entries); + if (num_entries < 0) + return -EINVAL; + + ret = copy_gid_entries_to_user(attrs, entries, num_entries, + user_entry_size); + if (ret) + return ret; + + ret = uverbs_copy_to(attrs, + UVERBS_ATTR_QUERY_GID_TABLE_RESP_NUM_ENTRIES, + &num_entries, sizeof(num_entries)); + return ret; +} + +static int UVERBS_HANDLER(UVERBS_METHOD_QUERY_GID_ENTRY)( + struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_gid_entry entry = {}; + const struct ib_gid_attr *gid_attr; + struct ib_ucontext *ucontext; + struct ib_device *ib_dev; + struct net_device *ndev; + u32 gid_index; + u32 port_num; + u32 flags; + int ret; + + ret = uverbs_get_flags32(&flags, attrs, + UVERBS_ATTR_QUERY_GID_ENTRY_FLAGS, 0); + if (ret) + return ret; + + ret = uverbs_get_const(&port_num, attrs, + UVERBS_ATTR_QUERY_GID_ENTRY_PORT); + if (ret) + return ret; + + ret = uverbs_get_const(&gid_index, attrs, + UVERBS_ATTR_QUERY_GID_ENTRY_GID_INDEX); + if (ret) + return ret; + + ucontext = ib_uverbs_get_ucontext(attrs); + if (IS_ERR(ucontext)) + return PTR_ERR(ucontext); + ib_dev = ucontext->device; + + if (!rdma_is_port_valid(ib_dev, port_num)) + return -EINVAL; + + gid_attr = rdma_get_gid_attr(ib_dev, port_num, gid_index); + if (IS_ERR(gid_attr)) + return PTR_ERR(gid_attr); + + memcpy(&entry.gid, &gid_attr->gid, sizeof(gid_attr->gid)); + entry.gid_index = gid_attr->index; + entry.port_num = gid_attr->port_num; + entry.gid_type = gid_attr->gid_type; + + rcu_read_lock(); + ndev = rdma_read_gid_attr_ndev_rcu(gid_attr); + if (IS_ERR(ndev)) { + if (PTR_ERR(ndev) != -ENODEV) { + ret = PTR_ERR(ndev); + rcu_read_unlock(); + goto out; + } + } else { + entry.netdev_ifindex = ndev->ifindex; + } + rcu_read_unlock(); + + ret = uverbs_copy_to_struct_or_zero( + attrs, UVERBS_ATTR_QUERY_GID_ENTRY_RESP_ENTRY, &entry, + sizeof(entry)); +out: + rdma_put_gid_attr(gid_attr); + return ret; +} + +DECLARE_UVERBS_NAMED_METHOD( + UVERBS_METHOD_GET_CONTEXT, + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_GET_CONTEXT_NUM_COMP_VECTORS, + UVERBS_ATTR_TYPE(u32), UA_OPTIONAL), + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_GET_CONTEXT_CORE_SUPPORT, + UVERBS_ATTR_TYPE(u64), UA_OPTIONAL), + UVERBS_ATTR_UHW()); + +DECLARE_UVERBS_NAMED_METHOD( + UVERBS_METHOD_QUERY_CONTEXT, + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_QUERY_CONTEXT_NUM_COMP_VECTORS, + UVERBS_ATTR_TYPE(u32), UA_OPTIONAL), + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_QUERY_CONTEXT_CORE_SUPPORT, + UVERBS_ATTR_TYPE(u64), UA_OPTIONAL)); + +DECLARE_UVERBS_NAMED_METHOD( + UVERBS_METHOD_INFO_HANDLES, + /* Also includes any device specific object ids */ + UVERBS_ATTR_CONST_IN(UVERBS_ATTR_INFO_OBJECT_ID, + enum uverbs_default_objects, UA_MANDATORY), + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_INFO_TOTAL_HANDLES, + UVERBS_ATTR_TYPE(u32), UA_OPTIONAL), + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_INFO_HANDLES_LIST, + UVERBS_ATTR_MIN_SIZE(sizeof(u32)), UA_OPTIONAL)); + +DECLARE_UVERBS_NAMED_METHOD( + UVERBS_METHOD_QUERY_PORT, + UVERBS_ATTR_CONST_IN(UVERBS_ATTR_QUERY_PORT_PORT_NUM, u8, UA_MANDATORY), + UVERBS_ATTR_PTR_OUT( + UVERBS_ATTR_QUERY_PORT_RESP, + UVERBS_ATTR_STRUCT(struct ib_uverbs_query_port_resp_ex, + reserved), + UA_MANDATORY)); + +DECLARE_UVERBS_NAMED_METHOD( + UVERBS_METHOD_QUERY_GID_TABLE, + UVERBS_ATTR_CONST_IN(UVERBS_ATTR_QUERY_GID_TABLE_ENTRY_SIZE, u64, + UA_MANDATORY), + UVERBS_ATTR_FLAGS_IN(UVERBS_ATTR_QUERY_GID_TABLE_FLAGS, u32, + UA_OPTIONAL), + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_QUERY_GID_TABLE_RESP_ENTRIES, + UVERBS_ATTR_MIN_SIZE(0), UA_MANDATORY), + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_QUERY_GID_TABLE_RESP_NUM_ENTRIES, + UVERBS_ATTR_TYPE(u64), UA_MANDATORY)); + +DECLARE_UVERBS_NAMED_METHOD( + UVERBS_METHOD_QUERY_GID_ENTRY, + UVERBS_ATTR_CONST_IN(UVERBS_ATTR_QUERY_GID_ENTRY_PORT, u32, + UA_MANDATORY), + UVERBS_ATTR_CONST_IN(UVERBS_ATTR_QUERY_GID_ENTRY_GID_INDEX, u32, + UA_MANDATORY), + UVERBS_ATTR_FLAGS_IN(UVERBS_ATTR_QUERY_GID_ENTRY_FLAGS, u32, + UA_MANDATORY), + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_QUERY_GID_ENTRY_RESP_ENTRY, + UVERBS_ATTR_STRUCT(struct ib_uverbs_gid_entry, + netdev_ifindex), + UA_MANDATORY)); + +DECLARE_UVERBS_GLOBAL_METHODS(UVERBS_OBJECT_DEVICE, + &UVERBS_METHOD(UVERBS_METHOD_GET_CONTEXT), + &UVERBS_METHOD(UVERBS_METHOD_INVOKE_WRITE), + &UVERBS_METHOD(UVERBS_METHOD_INFO_HANDLES), + &UVERBS_METHOD(UVERBS_METHOD_QUERY_PORT), + &UVERBS_METHOD(UVERBS_METHOD_QUERY_CONTEXT), + &UVERBS_METHOD(UVERBS_METHOD_QUERY_GID_TABLE), + &UVERBS_METHOD(UVERBS_METHOD_QUERY_GID_ENTRY)); + +const struct uapi_definition uverbs_def_obj_device[] = { + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_DEVICE), + {}, +}; diff --git a/drivers/infiniband/core/uverbs_std_types_dm.c b/drivers/infiniband/core/uverbs_std_types_dm.c new file mode 100644 index 000000000..d5a1de33c --- /dev/null +++ b/drivers/infiniband/core/uverbs_std_types_dm.c @@ -0,0 +1,118 @@ +/* + * Copyright (c) 2018, Mellanox Technologies inc. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include "rdma_core.h" +#include "uverbs.h" +#include <rdma/uverbs_std_types.h> + +static int uverbs_free_dm(struct ib_uobject *uobject, + enum rdma_remove_reason why, + struct uverbs_attr_bundle *attrs) +{ + struct ib_dm *dm = uobject->object; + int ret; + + ret = ib_destroy_usecnt(&dm->usecnt, why, uobject); + if (ret) + return ret; + + return dm->device->ops.dealloc_dm(dm, attrs); +} + +static int UVERBS_HANDLER(UVERBS_METHOD_DM_ALLOC)( + struct uverbs_attr_bundle *attrs) +{ + struct ib_dm_alloc_attr attr = {}; + struct ib_uobject *uobj = + uverbs_attr_get(attrs, UVERBS_ATTR_ALLOC_DM_HANDLE) + ->obj_attr.uobject; + struct ib_device *ib_dev = attrs->context->device; + struct ib_dm *dm; + int ret; + + if (!ib_dev->ops.alloc_dm) + return -EOPNOTSUPP; + + ret = uverbs_copy_from(&attr.length, attrs, + UVERBS_ATTR_ALLOC_DM_LENGTH); + if (ret) + return ret; + + ret = uverbs_copy_from(&attr.alignment, attrs, + UVERBS_ATTR_ALLOC_DM_ALIGNMENT); + if (ret) + return ret; + + dm = ib_dev->ops.alloc_dm(ib_dev, attrs->context, &attr, attrs); + if (IS_ERR(dm)) + return PTR_ERR(dm); + + dm->device = ib_dev; + dm->length = attr.length; + dm->uobject = uobj; + atomic_set(&dm->usecnt, 0); + + uobj->object = dm; + + return 0; +} + +DECLARE_UVERBS_NAMED_METHOD( + UVERBS_METHOD_DM_ALLOC, + UVERBS_ATTR_IDR(UVERBS_ATTR_ALLOC_DM_HANDLE, + UVERBS_OBJECT_DM, + UVERBS_ACCESS_NEW, + UA_MANDATORY), + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_ALLOC_DM_LENGTH, + UVERBS_ATTR_TYPE(u64), + UA_MANDATORY), + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_ALLOC_DM_ALIGNMENT, + UVERBS_ATTR_TYPE(u32), + UA_MANDATORY)); + +DECLARE_UVERBS_NAMED_METHOD_DESTROY( + UVERBS_METHOD_DM_FREE, + UVERBS_ATTR_IDR(UVERBS_ATTR_FREE_DM_HANDLE, + UVERBS_OBJECT_DM, + UVERBS_ACCESS_DESTROY, + UA_MANDATORY)); + +DECLARE_UVERBS_NAMED_OBJECT(UVERBS_OBJECT_DM, + UVERBS_TYPE_ALLOC_IDR(uverbs_free_dm), + &UVERBS_METHOD(UVERBS_METHOD_DM_ALLOC), + &UVERBS_METHOD(UVERBS_METHOD_DM_FREE)); + +const struct uapi_definition uverbs_def_obj_dm[] = { + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_DM, + UAPI_DEF_OBJ_NEEDS_FN(dealloc_dm)), + {} +}; diff --git a/drivers/infiniband/core/uverbs_std_types_flow_action.c b/drivers/infiniband/core/uverbs_std_types_flow_action.c new file mode 100644 index 000000000..459cf165b --- /dev/null +++ b/drivers/infiniband/core/uverbs_std_types_flow_action.c @@ -0,0 +1,449 @@ +/* + * Copyright (c) 2018, Mellanox Technologies inc. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include "rdma_core.h" +#include "uverbs.h" +#include <rdma/uverbs_std_types.h> + +static int uverbs_free_flow_action(struct ib_uobject *uobject, + enum rdma_remove_reason why, + struct uverbs_attr_bundle *attrs) +{ + struct ib_flow_action *action = uobject->object; + int ret; + + ret = ib_destroy_usecnt(&action->usecnt, why, uobject); + if (ret) + return ret; + + return action->device->ops.destroy_flow_action(action); +} + +static u64 esp_flags_uverbs_to_verbs(struct uverbs_attr_bundle *attrs, + u32 flags, bool is_modify) +{ + u64 verbs_flags = flags; + + if (uverbs_attr_is_valid(attrs, UVERBS_ATTR_FLOW_ACTION_ESP_ESN)) + verbs_flags |= IB_FLOW_ACTION_ESP_FLAGS_ESN_TRIGGERED; + + if (is_modify && uverbs_attr_is_valid(attrs, + UVERBS_ATTR_FLOW_ACTION_ESP_ATTRS)) + verbs_flags |= IB_FLOW_ACTION_ESP_FLAGS_MOD_ESP_ATTRS; + + return verbs_flags; +}; + +static int validate_flow_action_esp_keymat_aes_gcm(struct ib_flow_action_attrs_esp_keymats *keymat) +{ + struct ib_uverbs_flow_action_esp_keymat_aes_gcm *aes_gcm = + &keymat->keymat.aes_gcm; + + if (aes_gcm->iv_algo > IB_UVERBS_FLOW_ACTION_IV_ALGO_SEQ) + return -EOPNOTSUPP; + + if (aes_gcm->key_len != 32 && + aes_gcm->key_len != 24 && + aes_gcm->key_len != 16) + return -EINVAL; + + if (aes_gcm->icv_len != 16 && + aes_gcm->icv_len != 8 && + aes_gcm->icv_len != 12) + return -EINVAL; + + return 0; +} + +static int (* const flow_action_esp_keymat_validate[])(struct ib_flow_action_attrs_esp_keymats *keymat) = { + [IB_UVERBS_FLOW_ACTION_ESP_KEYMAT_AES_GCM] = validate_flow_action_esp_keymat_aes_gcm, +}; + +static int flow_action_esp_replay_none(struct ib_flow_action_attrs_esp_replays *replay, + bool is_modify) +{ + /* This is used in order to modify an esp flow action with an enabled + * replay protection to a disabled one. This is only supported via + * modify, as in create verb we can simply drop the REPLAY attribute and + * achieve the same thing. + */ + return is_modify ? 0 : -EINVAL; +} + +static int flow_action_esp_replay_def_ok(struct ib_flow_action_attrs_esp_replays *replay, + bool is_modify) +{ + /* Some replay protections could always be enabled without validating + * anything. + */ + return 0; +} + +static int (* const flow_action_esp_replay_validate[])(struct ib_flow_action_attrs_esp_replays *replay, + bool is_modify) = { + [IB_UVERBS_FLOW_ACTION_ESP_REPLAY_NONE] = flow_action_esp_replay_none, + [IB_UVERBS_FLOW_ACTION_ESP_REPLAY_BMP] = flow_action_esp_replay_def_ok, +}; + +static int parse_esp_ip(enum ib_flow_spec_type proto, + const void __user *val_ptr, + size_t len, union ib_flow_spec *out) +{ + int ret; + const struct ib_uverbs_flow_ipv4_filter ipv4 = { + .src_ip = cpu_to_be32(0xffffffffUL), + .dst_ip = cpu_to_be32(0xffffffffUL), + .proto = 0xff, + .tos = 0xff, + .ttl = 0xff, + .flags = 0xff, + }; + const struct ib_uverbs_flow_ipv6_filter ipv6 = { + .src_ip = {0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff}, + .dst_ip = {0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff}, + .flow_label = cpu_to_be32(0xffffffffUL), + .next_hdr = 0xff, + .traffic_class = 0xff, + .hop_limit = 0xff, + }; + union { + struct ib_uverbs_flow_ipv4_filter ipv4; + struct ib_uverbs_flow_ipv6_filter ipv6; + } user_val = {}; + const void *user_pmask; + size_t val_len; + + /* If the flow IPv4/IPv6 flow specifications are extended, the mask + * should be changed as well. + */ + BUILD_BUG_ON(offsetof(struct ib_uverbs_flow_ipv4_filter, flags) + + sizeof(ipv4.flags) != sizeof(ipv4)); + BUILD_BUG_ON(offsetof(struct ib_uverbs_flow_ipv6_filter, reserved) + + sizeof(ipv6.reserved) != sizeof(ipv6)); + + switch (proto) { + case IB_FLOW_SPEC_IPV4: + if (len > sizeof(user_val.ipv4) && + !ib_is_buffer_cleared(val_ptr + sizeof(user_val.ipv4), + len - sizeof(user_val.ipv4))) + return -EOPNOTSUPP; + + val_len = min_t(size_t, len, sizeof(user_val.ipv4)); + ret = copy_from_user(&user_val.ipv4, val_ptr, + val_len); + if (ret) + return -EFAULT; + + user_pmask = &ipv4; + break; + case IB_FLOW_SPEC_IPV6: + if (len > sizeof(user_val.ipv6) && + !ib_is_buffer_cleared(val_ptr + sizeof(user_val.ipv6), + len - sizeof(user_val.ipv6))) + return -EOPNOTSUPP; + + val_len = min_t(size_t, len, sizeof(user_val.ipv6)); + ret = copy_from_user(&user_val.ipv6, val_ptr, + val_len); + if (ret) + return -EFAULT; + + user_pmask = &ipv6; + break; + default: + return -EOPNOTSUPP; + } + + return ib_uverbs_kern_spec_to_ib_spec_filter(proto, user_pmask, + &user_val, + val_len, out); +} + +static int flow_action_esp_get_encap(struct ib_flow_spec_list *out, + struct uverbs_attr_bundle *attrs) +{ + struct ib_uverbs_flow_action_esp_encap uverbs_encap; + int ret; + + ret = uverbs_copy_from(&uverbs_encap, attrs, + UVERBS_ATTR_FLOW_ACTION_ESP_ENCAP); + if (ret) + return ret; + + /* We currently support only one encap */ + if (uverbs_encap.next_ptr) + return -EOPNOTSUPP; + + if (uverbs_encap.type != IB_FLOW_SPEC_IPV4 && + uverbs_encap.type != IB_FLOW_SPEC_IPV6) + return -EOPNOTSUPP; + + return parse_esp_ip(uverbs_encap.type, + u64_to_user_ptr(uverbs_encap.val_ptr), + uverbs_encap.len, + &out->spec); +} + +struct ib_flow_action_esp_attr { + struct ib_flow_action_attrs_esp hdr; + struct ib_flow_action_attrs_esp_keymats keymat; + struct ib_flow_action_attrs_esp_replays replay; + /* We currently support only one spec */ + struct ib_flow_spec_list encap; +}; + +#define ESP_LAST_SUPPORTED_FLAG IB_UVERBS_FLOW_ACTION_ESP_FLAGS_ESN_NEW_WINDOW +static int parse_flow_action_esp(struct ib_device *ib_dev, + struct uverbs_attr_bundle *attrs, + struct ib_flow_action_esp_attr *esp_attr, + bool is_modify) +{ + struct ib_uverbs_flow_action_esp uverbs_esp = {}; + int ret; + + /* Optional param, if it doesn't exist, we get -ENOENT and skip it */ + ret = uverbs_copy_from(&esp_attr->hdr.esn, attrs, + UVERBS_ATTR_FLOW_ACTION_ESP_ESN); + if (IS_UVERBS_COPY_ERR(ret)) + return ret; + + /* This can be called from FLOW_ACTION_ESP_MODIFY where + * UVERBS_ATTR_FLOW_ACTION_ESP_ATTRS is optional + */ + if (uverbs_attr_is_valid(attrs, UVERBS_ATTR_FLOW_ACTION_ESP_ATTRS)) { + ret = uverbs_copy_from_or_zero(&uverbs_esp, attrs, + UVERBS_ATTR_FLOW_ACTION_ESP_ATTRS); + if (ret) + return ret; + + if (uverbs_esp.flags & ~((ESP_LAST_SUPPORTED_FLAG << 1) - 1)) + return -EOPNOTSUPP; + + esp_attr->hdr.spi = uverbs_esp.spi; + esp_attr->hdr.seq = uverbs_esp.seq; + esp_attr->hdr.tfc_pad = uverbs_esp.tfc_pad; + esp_attr->hdr.hard_limit_pkts = uverbs_esp.hard_limit_pkts; + } + esp_attr->hdr.flags = esp_flags_uverbs_to_verbs(attrs, uverbs_esp.flags, + is_modify); + + if (uverbs_attr_is_valid(attrs, UVERBS_ATTR_FLOW_ACTION_ESP_KEYMAT)) { + esp_attr->keymat.protocol = + uverbs_attr_get_enum_id(attrs, + UVERBS_ATTR_FLOW_ACTION_ESP_KEYMAT); + ret = uverbs_copy_from_or_zero(&esp_attr->keymat.keymat, + attrs, + UVERBS_ATTR_FLOW_ACTION_ESP_KEYMAT); + if (ret) + return ret; + + ret = flow_action_esp_keymat_validate[esp_attr->keymat.protocol](&esp_attr->keymat); + if (ret) + return ret; + + esp_attr->hdr.keymat = &esp_attr->keymat; + } + + if (uverbs_attr_is_valid(attrs, UVERBS_ATTR_FLOW_ACTION_ESP_REPLAY)) { + esp_attr->replay.protocol = + uverbs_attr_get_enum_id(attrs, + UVERBS_ATTR_FLOW_ACTION_ESP_REPLAY); + + ret = uverbs_copy_from_or_zero(&esp_attr->replay.replay, + attrs, + UVERBS_ATTR_FLOW_ACTION_ESP_REPLAY); + if (ret) + return ret; + + ret = flow_action_esp_replay_validate[esp_attr->replay.protocol](&esp_attr->replay, + is_modify); + if (ret) + return ret; + + esp_attr->hdr.replay = &esp_attr->replay; + } + + if (uverbs_attr_is_valid(attrs, UVERBS_ATTR_FLOW_ACTION_ESP_ENCAP)) { + ret = flow_action_esp_get_encap(&esp_attr->encap, attrs); + if (ret) + return ret; + + esp_attr->hdr.encap = &esp_attr->encap; + } + + return 0; +} + +static int UVERBS_HANDLER(UVERBS_METHOD_FLOW_ACTION_ESP_CREATE)( + struct uverbs_attr_bundle *attrs) +{ + struct ib_uobject *uobj = uverbs_attr_get_uobject( + attrs, UVERBS_ATTR_CREATE_FLOW_ACTION_ESP_HANDLE); + struct ib_device *ib_dev = attrs->context->device; + int ret; + struct ib_flow_action *action; + struct ib_flow_action_esp_attr esp_attr = {}; + + if (!ib_dev->ops.create_flow_action_esp) + return -EOPNOTSUPP; + + ret = parse_flow_action_esp(ib_dev, attrs, &esp_attr, false); + if (ret) + return ret; + + /* No need to check as this attribute is marked as MANDATORY */ + action = ib_dev->ops.create_flow_action_esp(ib_dev, &esp_attr.hdr, + attrs); + if (IS_ERR(action)) + return PTR_ERR(action); + + uverbs_flow_action_fill_action(action, uobj, ib_dev, + IB_FLOW_ACTION_ESP); + + return 0; +} + +static int UVERBS_HANDLER(UVERBS_METHOD_FLOW_ACTION_ESP_MODIFY)( + struct uverbs_attr_bundle *attrs) +{ + struct ib_uobject *uobj = uverbs_attr_get_uobject( + attrs, UVERBS_ATTR_MODIFY_FLOW_ACTION_ESP_HANDLE); + struct ib_flow_action *action = uobj->object; + int ret; + struct ib_flow_action_esp_attr esp_attr = {}; + + if (!action->device->ops.modify_flow_action_esp) + return -EOPNOTSUPP; + + ret = parse_flow_action_esp(action->device, attrs, &esp_attr, true); + if (ret) + return ret; + + if (action->type != IB_FLOW_ACTION_ESP) + return -EINVAL; + + return action->device->ops.modify_flow_action_esp(action, + &esp_attr.hdr, + attrs); +} + +static const struct uverbs_attr_spec uverbs_flow_action_esp_keymat[] = { + [IB_UVERBS_FLOW_ACTION_ESP_KEYMAT_AES_GCM] = { + .type = UVERBS_ATTR_TYPE_PTR_IN, + UVERBS_ATTR_STRUCT( + struct ib_uverbs_flow_action_esp_keymat_aes_gcm, + aes_key), + }, +}; + +static const struct uverbs_attr_spec uverbs_flow_action_esp_replay[] = { + [IB_UVERBS_FLOW_ACTION_ESP_REPLAY_NONE] = { + .type = UVERBS_ATTR_TYPE_PTR_IN, + UVERBS_ATTR_NO_DATA(), + }, + [IB_UVERBS_FLOW_ACTION_ESP_REPLAY_BMP] = { + .type = UVERBS_ATTR_TYPE_PTR_IN, + UVERBS_ATTR_STRUCT(struct ib_uverbs_flow_action_esp_replay_bmp, + size), + }, +}; + +DECLARE_UVERBS_NAMED_METHOD( + UVERBS_METHOD_FLOW_ACTION_ESP_CREATE, + UVERBS_ATTR_IDR(UVERBS_ATTR_CREATE_FLOW_ACTION_ESP_HANDLE, + UVERBS_OBJECT_FLOW_ACTION, + UVERBS_ACCESS_NEW, + UA_MANDATORY), + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_FLOW_ACTION_ESP_ATTRS, + UVERBS_ATTR_STRUCT(struct ib_uverbs_flow_action_esp, + hard_limit_pkts), + UA_MANDATORY), + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_FLOW_ACTION_ESP_ESN, + UVERBS_ATTR_TYPE(__u32), + UA_OPTIONAL), + UVERBS_ATTR_ENUM_IN(UVERBS_ATTR_FLOW_ACTION_ESP_KEYMAT, + uverbs_flow_action_esp_keymat, + UA_MANDATORY), + UVERBS_ATTR_ENUM_IN(UVERBS_ATTR_FLOW_ACTION_ESP_REPLAY, + uverbs_flow_action_esp_replay, + UA_OPTIONAL), + UVERBS_ATTR_PTR_IN( + UVERBS_ATTR_FLOW_ACTION_ESP_ENCAP, + UVERBS_ATTR_TYPE(struct ib_uverbs_flow_action_esp_encap), + UA_OPTIONAL)); + +DECLARE_UVERBS_NAMED_METHOD( + UVERBS_METHOD_FLOW_ACTION_ESP_MODIFY, + UVERBS_ATTR_IDR(UVERBS_ATTR_MODIFY_FLOW_ACTION_ESP_HANDLE, + UVERBS_OBJECT_FLOW_ACTION, + UVERBS_ACCESS_WRITE, + UA_MANDATORY), + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_FLOW_ACTION_ESP_ATTRS, + UVERBS_ATTR_STRUCT(struct ib_uverbs_flow_action_esp, + hard_limit_pkts), + UA_OPTIONAL), + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_FLOW_ACTION_ESP_ESN, + UVERBS_ATTR_TYPE(__u32), + UA_OPTIONAL), + UVERBS_ATTR_ENUM_IN(UVERBS_ATTR_FLOW_ACTION_ESP_KEYMAT, + uverbs_flow_action_esp_keymat, + UA_OPTIONAL), + UVERBS_ATTR_ENUM_IN(UVERBS_ATTR_FLOW_ACTION_ESP_REPLAY, + uverbs_flow_action_esp_replay, + UA_OPTIONAL), + UVERBS_ATTR_PTR_IN( + UVERBS_ATTR_FLOW_ACTION_ESP_ENCAP, + UVERBS_ATTR_TYPE(struct ib_uverbs_flow_action_esp_encap), + UA_OPTIONAL)); + +DECLARE_UVERBS_NAMED_METHOD_DESTROY( + UVERBS_METHOD_FLOW_ACTION_DESTROY, + UVERBS_ATTR_IDR(UVERBS_ATTR_DESTROY_FLOW_ACTION_HANDLE, + UVERBS_OBJECT_FLOW_ACTION, + UVERBS_ACCESS_DESTROY, + UA_MANDATORY)); + +DECLARE_UVERBS_NAMED_OBJECT( + UVERBS_OBJECT_FLOW_ACTION, + UVERBS_TYPE_ALLOC_IDR(uverbs_free_flow_action), + &UVERBS_METHOD(UVERBS_METHOD_FLOW_ACTION_ESP_CREATE), + &UVERBS_METHOD(UVERBS_METHOD_FLOW_ACTION_DESTROY), + &UVERBS_METHOD(UVERBS_METHOD_FLOW_ACTION_ESP_MODIFY)); + +const struct uapi_definition uverbs_def_obj_flow_action[] = { + UAPI_DEF_CHAIN_OBJ_TREE_NAMED( + UVERBS_OBJECT_FLOW_ACTION, + UAPI_DEF_OBJ_NEEDS_FN(destroy_flow_action)), + {} +}; diff --git a/drivers/infiniband/core/uverbs_std_types_mr.c b/drivers/infiniband/core/uverbs_std_types_mr.c new file mode 100644 index 000000000..dc5856441 --- /dev/null +++ b/drivers/infiniband/core/uverbs_std_types_mr.c @@ -0,0 +1,269 @@ +/* + * Copyright (c) 2018, Mellanox Technologies inc. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include "rdma_core.h" +#include "uverbs.h" +#include <rdma/uverbs_std_types.h> +#include "restrack.h" + +static int uverbs_free_mr(struct ib_uobject *uobject, + enum rdma_remove_reason why, + struct uverbs_attr_bundle *attrs) +{ + return ib_dereg_mr_user((struct ib_mr *)uobject->object, + &attrs->driver_udata); +} + +static int UVERBS_HANDLER(UVERBS_METHOD_ADVISE_MR)( + struct uverbs_attr_bundle *attrs) +{ + struct ib_pd *pd = + uverbs_attr_get_obj(attrs, UVERBS_ATTR_ADVISE_MR_PD_HANDLE); + enum ib_uverbs_advise_mr_advice advice; + struct ib_device *ib_dev = pd->device; + struct ib_sge *sg_list; + int num_sge; + u32 flags; + int ret; + + /* FIXME: Extend the UAPI_DEF_OBJ_NEEDS_FN stuff.. */ + if (!ib_dev->ops.advise_mr) + return -EOPNOTSUPP; + + ret = uverbs_get_const(&advice, attrs, UVERBS_ATTR_ADVISE_MR_ADVICE); + if (ret) + return ret; + + ret = uverbs_get_flags32(&flags, attrs, UVERBS_ATTR_ADVISE_MR_FLAGS, + IB_UVERBS_ADVISE_MR_FLAG_FLUSH); + if (ret) + return ret; + + num_sge = uverbs_attr_ptr_get_array_size( + attrs, UVERBS_ATTR_ADVISE_MR_SGE_LIST, sizeof(struct ib_sge)); + if (num_sge <= 0) + return num_sge; + + sg_list = uverbs_attr_get_alloced_ptr(attrs, + UVERBS_ATTR_ADVISE_MR_SGE_LIST); + return ib_dev->ops.advise_mr(pd, advice, flags, sg_list, num_sge, + attrs); +} + +static int UVERBS_HANDLER(UVERBS_METHOD_DM_MR_REG)( + struct uverbs_attr_bundle *attrs) +{ + struct ib_dm_mr_attr attr = {}; + struct ib_uobject *uobj = + uverbs_attr_get_uobject(attrs, UVERBS_ATTR_REG_DM_MR_HANDLE); + struct ib_dm *dm = + uverbs_attr_get_obj(attrs, UVERBS_ATTR_REG_DM_MR_DM_HANDLE); + struct ib_pd *pd = + uverbs_attr_get_obj(attrs, UVERBS_ATTR_REG_DM_MR_PD_HANDLE); + struct ib_device *ib_dev = pd->device; + + struct ib_mr *mr; + int ret; + + if (!ib_dev->ops.reg_dm_mr) + return -EOPNOTSUPP; + + ret = uverbs_copy_from(&attr.offset, attrs, UVERBS_ATTR_REG_DM_MR_OFFSET); + if (ret) + return ret; + + ret = uverbs_copy_from(&attr.length, attrs, + UVERBS_ATTR_REG_DM_MR_LENGTH); + if (ret) + return ret; + + ret = uverbs_get_flags32(&attr.access_flags, attrs, + UVERBS_ATTR_REG_DM_MR_ACCESS_FLAGS, + IB_ACCESS_SUPPORTED); + if (ret) + return ret; + + if (!(attr.access_flags & IB_ZERO_BASED)) + return -EINVAL; + + ret = ib_check_mr_access(attr.access_flags); + if (ret) + return ret; + + if (attr.offset > dm->length || attr.length > dm->length || + attr.length > dm->length - attr.offset) + return -EINVAL; + + mr = pd->device->ops.reg_dm_mr(pd, dm, &attr, attrs); + if (IS_ERR(mr)) + return PTR_ERR(mr); + + mr->device = pd->device; + mr->pd = pd; + mr->type = IB_MR_TYPE_DM; + mr->dm = dm; + mr->uobject = uobj; + atomic_inc(&pd->usecnt); + atomic_inc(&dm->usecnt); + + rdma_restrack_new(&mr->res, RDMA_RESTRACK_MR); + rdma_restrack_set_name(&mr->res, NULL); + rdma_restrack_add(&mr->res); + uobj->object = mr; + + uverbs_finalize_uobj_create(attrs, UVERBS_ATTR_REG_DM_MR_HANDLE); + + ret = uverbs_copy_to(attrs, UVERBS_ATTR_REG_DM_MR_RESP_LKEY, &mr->lkey, + sizeof(mr->lkey)); + if (ret) + return ret; + + ret = uverbs_copy_to(attrs, UVERBS_ATTR_REG_DM_MR_RESP_RKEY, + &mr->rkey, sizeof(mr->rkey)); + return ret; +} + +static int UVERBS_HANDLER(UVERBS_METHOD_QUERY_MR)( + struct uverbs_attr_bundle *attrs) +{ + struct ib_mr *mr = + uverbs_attr_get_obj(attrs, UVERBS_ATTR_QUERY_MR_HANDLE); + int ret; + + ret = uverbs_copy_to(attrs, UVERBS_ATTR_QUERY_MR_RESP_LKEY, &mr->lkey, + sizeof(mr->lkey)); + if (ret) + return ret; + + ret = uverbs_copy_to(attrs, UVERBS_ATTR_QUERY_MR_RESP_RKEY, + &mr->rkey, sizeof(mr->rkey)); + + if (ret) + return ret; + + ret = uverbs_copy_to(attrs, UVERBS_ATTR_QUERY_MR_RESP_LENGTH, + &mr->length, sizeof(mr->length)); + + if (ret) + return ret; + + ret = uverbs_copy_to(attrs, UVERBS_ATTR_QUERY_MR_RESP_IOVA, + &mr->iova, sizeof(mr->iova)); + + return IS_UVERBS_COPY_ERR(ret) ? ret : 0; +} + +DECLARE_UVERBS_NAMED_METHOD( + UVERBS_METHOD_ADVISE_MR, + UVERBS_ATTR_IDR(UVERBS_ATTR_ADVISE_MR_PD_HANDLE, + UVERBS_OBJECT_PD, + UVERBS_ACCESS_READ, + UA_MANDATORY), + UVERBS_ATTR_CONST_IN(UVERBS_ATTR_ADVISE_MR_ADVICE, + enum ib_uverbs_advise_mr_advice, + UA_MANDATORY), + UVERBS_ATTR_FLAGS_IN(UVERBS_ATTR_ADVISE_MR_FLAGS, + enum ib_uverbs_advise_mr_flag, + UA_MANDATORY), + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_ADVISE_MR_SGE_LIST, + UVERBS_ATTR_MIN_SIZE(sizeof(struct ib_uverbs_sge)), + UA_MANDATORY, + UA_ALLOC_AND_COPY)); + +DECLARE_UVERBS_NAMED_METHOD( + UVERBS_METHOD_QUERY_MR, + UVERBS_ATTR_IDR(UVERBS_ATTR_QUERY_MR_HANDLE, + UVERBS_OBJECT_MR, + UVERBS_ACCESS_READ, + UA_MANDATORY), + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_QUERY_MR_RESP_RKEY, + UVERBS_ATTR_TYPE(u32), + UA_MANDATORY), + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_QUERY_MR_RESP_LKEY, + UVERBS_ATTR_TYPE(u32), + UA_MANDATORY), + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_QUERY_MR_RESP_LENGTH, + UVERBS_ATTR_TYPE(u64), + UA_MANDATORY), + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_QUERY_MR_RESP_IOVA, + UVERBS_ATTR_TYPE(u64), + UA_OPTIONAL)); + +DECLARE_UVERBS_NAMED_METHOD( + UVERBS_METHOD_DM_MR_REG, + UVERBS_ATTR_IDR(UVERBS_ATTR_REG_DM_MR_HANDLE, + UVERBS_OBJECT_MR, + UVERBS_ACCESS_NEW, + UA_MANDATORY), + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_REG_DM_MR_OFFSET, + UVERBS_ATTR_TYPE(u64), + UA_MANDATORY), + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_REG_DM_MR_LENGTH, + UVERBS_ATTR_TYPE(u64), + UA_MANDATORY), + UVERBS_ATTR_IDR(UVERBS_ATTR_REG_DM_MR_PD_HANDLE, + UVERBS_OBJECT_PD, + UVERBS_ACCESS_READ, + UA_MANDATORY), + UVERBS_ATTR_FLAGS_IN(UVERBS_ATTR_REG_DM_MR_ACCESS_FLAGS, + enum ib_access_flags), + UVERBS_ATTR_IDR(UVERBS_ATTR_REG_DM_MR_DM_HANDLE, + UVERBS_OBJECT_DM, + UVERBS_ACCESS_READ, + UA_MANDATORY), + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_REG_DM_MR_RESP_LKEY, + UVERBS_ATTR_TYPE(u32), + UA_MANDATORY), + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_REG_DM_MR_RESP_RKEY, + UVERBS_ATTR_TYPE(u32), + UA_MANDATORY)); + +DECLARE_UVERBS_NAMED_METHOD_DESTROY( + UVERBS_METHOD_MR_DESTROY, + UVERBS_ATTR_IDR(UVERBS_ATTR_DESTROY_MR_HANDLE, + UVERBS_OBJECT_MR, + UVERBS_ACCESS_DESTROY, + UA_MANDATORY)); + +DECLARE_UVERBS_NAMED_OBJECT( + UVERBS_OBJECT_MR, + UVERBS_TYPE_ALLOC_IDR(uverbs_free_mr), + &UVERBS_METHOD(UVERBS_METHOD_DM_MR_REG), + &UVERBS_METHOD(UVERBS_METHOD_MR_DESTROY), + &UVERBS_METHOD(UVERBS_METHOD_ADVISE_MR), + &UVERBS_METHOD(UVERBS_METHOD_QUERY_MR)); + +const struct uapi_definition uverbs_def_obj_mr[] = { + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_MR, + UAPI_DEF_OBJ_NEEDS_FN(dereg_mr)), + {} +}; diff --git a/drivers/infiniband/core/uverbs_std_types_qp.c b/drivers/infiniband/core/uverbs_std_types_qp.c new file mode 100644 index 000000000..3bf8dcdfe --- /dev/null +++ b/drivers/infiniband/core/uverbs_std_types_qp.c @@ -0,0 +1,401 @@ +// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB +/* + * Copyright (c) 2020, Mellanox Technologies inc. All rights reserved. + */ + +#include <rdma/uverbs_std_types.h> +#include "rdma_core.h" +#include "uverbs.h" +#include "core_priv.h" + +static int uverbs_free_qp(struct ib_uobject *uobject, + enum rdma_remove_reason why, + struct uverbs_attr_bundle *attrs) +{ + struct ib_qp *qp = uobject->object; + struct ib_uqp_object *uqp = + container_of(uobject, struct ib_uqp_object, uevent.uobject); + int ret; + + /* + * If this is a user triggered destroy then do not allow destruction + * until the user cleans up all the mcast bindings. Unlike in other + * places we forcibly clean up the mcast attachments for !DESTROY + * because the mcast attaches are not ubojects and will not be + * destroyed by anything else during cleanup processing. + */ + if (why == RDMA_REMOVE_DESTROY) { + if (!list_empty(&uqp->mcast_list)) + return -EBUSY; + } else if (qp == qp->real_qp) { + ib_uverbs_detach_umcast(qp, uqp); + } + + ret = ib_destroy_qp_user(qp, &attrs->driver_udata); + if (ib_is_destroy_retryable(ret, why, uobject)) + return ret; + + if (uqp->uxrcd) + atomic_dec(&uqp->uxrcd->refcnt); + + ib_uverbs_release_uevent(&uqp->uevent); + return ret; +} + +static int check_creation_flags(enum ib_qp_type qp_type, + u32 create_flags) +{ + create_flags &= ~IB_UVERBS_QP_CREATE_SQ_SIG_ALL; + + if (!create_flags || qp_type == IB_QPT_DRIVER) + return 0; + + if (qp_type != IB_QPT_RAW_PACKET && qp_type != IB_QPT_UD) + return -EINVAL; + + if ((create_flags & IB_UVERBS_QP_CREATE_SCATTER_FCS || + create_flags & IB_UVERBS_QP_CREATE_CVLAN_STRIPPING) && + qp_type != IB_QPT_RAW_PACKET) + return -EINVAL; + + return 0; +} + +static void set_caps(struct ib_qp_init_attr *attr, + struct ib_uverbs_qp_cap *cap, bool req) +{ + if (req) { + attr->cap.max_send_wr = cap->max_send_wr; + attr->cap.max_recv_wr = cap->max_recv_wr; + attr->cap.max_send_sge = cap->max_send_sge; + attr->cap.max_recv_sge = cap->max_recv_sge; + attr->cap.max_inline_data = cap->max_inline_data; + } else { + cap->max_send_wr = attr->cap.max_send_wr; + cap->max_recv_wr = attr->cap.max_recv_wr; + cap->max_send_sge = attr->cap.max_send_sge; + cap->max_recv_sge = attr->cap.max_recv_sge; + cap->max_inline_data = attr->cap.max_inline_data; + } +} + +static int UVERBS_HANDLER(UVERBS_METHOD_QP_CREATE)( + struct uverbs_attr_bundle *attrs) +{ + struct ib_uqp_object *obj = container_of( + uverbs_attr_get_uobject(attrs, UVERBS_ATTR_CREATE_QP_HANDLE), + typeof(*obj), uevent.uobject); + struct ib_qp_init_attr attr = {}; + struct ib_uverbs_qp_cap cap = {}; + struct ib_rwq_ind_table *rwq_ind_tbl = NULL; + struct ib_qp *qp; + struct ib_pd *pd = NULL; + struct ib_srq *srq = NULL; + struct ib_cq *recv_cq = NULL; + struct ib_cq *send_cq = NULL; + struct ib_xrcd *xrcd = NULL; + struct ib_uobject *xrcd_uobj = NULL; + struct ib_device *device; + u64 user_handle; + int ret; + + ret = uverbs_copy_from_or_zero(&cap, attrs, + UVERBS_ATTR_CREATE_QP_CAP); + if (!ret) + ret = uverbs_copy_from(&user_handle, attrs, + UVERBS_ATTR_CREATE_QP_USER_HANDLE); + if (!ret) + ret = uverbs_get_const(&attr.qp_type, attrs, + UVERBS_ATTR_CREATE_QP_TYPE); + if (ret) + return ret; + + switch (attr.qp_type) { + case IB_QPT_XRC_TGT: + if (uverbs_attr_is_valid(attrs, + UVERBS_ATTR_CREATE_QP_RECV_CQ_HANDLE) || + uverbs_attr_is_valid(attrs, + UVERBS_ATTR_CREATE_QP_SEND_CQ_HANDLE) || + uverbs_attr_is_valid(attrs, + UVERBS_ATTR_CREATE_QP_PD_HANDLE) || + uverbs_attr_is_valid(attrs, + UVERBS_ATTR_CREATE_QP_IND_TABLE_HANDLE)) + return -EINVAL; + + xrcd_uobj = uverbs_attr_get_uobject(attrs, + UVERBS_ATTR_CREATE_QP_XRCD_HANDLE); + if (IS_ERR(xrcd_uobj)) + return PTR_ERR(xrcd_uobj); + + xrcd = (struct ib_xrcd *)xrcd_uobj->object; + if (!xrcd) + return -EINVAL; + device = xrcd->device; + break; + case IB_UVERBS_QPT_RAW_PACKET: + if (!capable(CAP_NET_RAW)) + return -EPERM; + fallthrough; + case IB_UVERBS_QPT_RC: + case IB_UVERBS_QPT_UC: + case IB_UVERBS_QPT_UD: + case IB_UVERBS_QPT_XRC_INI: + case IB_UVERBS_QPT_DRIVER: + if (uverbs_attr_is_valid(attrs, + UVERBS_ATTR_CREATE_QP_XRCD_HANDLE) || + (uverbs_attr_is_valid(attrs, + UVERBS_ATTR_CREATE_QP_SRQ_HANDLE) && + attr.qp_type == IB_QPT_XRC_INI)) + return -EINVAL; + + pd = uverbs_attr_get_obj(attrs, + UVERBS_ATTR_CREATE_QP_PD_HANDLE); + if (IS_ERR(pd)) + return PTR_ERR(pd); + + rwq_ind_tbl = uverbs_attr_get_obj(attrs, + UVERBS_ATTR_CREATE_QP_IND_TABLE_HANDLE); + if (!IS_ERR(rwq_ind_tbl)) { + if (cap.max_recv_wr || cap.max_recv_sge || + uverbs_attr_is_valid(attrs, + UVERBS_ATTR_CREATE_QP_RECV_CQ_HANDLE) || + uverbs_attr_is_valid(attrs, + UVERBS_ATTR_CREATE_QP_SRQ_HANDLE)) + return -EINVAL; + + /* send_cq is optinal */ + if (cap.max_send_wr) { + send_cq = uverbs_attr_get_obj(attrs, + UVERBS_ATTR_CREATE_QP_SEND_CQ_HANDLE); + if (IS_ERR(send_cq)) + return PTR_ERR(send_cq); + } + attr.rwq_ind_tbl = rwq_ind_tbl; + } else { + send_cq = uverbs_attr_get_obj(attrs, + UVERBS_ATTR_CREATE_QP_SEND_CQ_HANDLE); + if (IS_ERR(send_cq)) + return PTR_ERR(send_cq); + + if (attr.qp_type != IB_QPT_XRC_INI) { + recv_cq = uverbs_attr_get_obj(attrs, + UVERBS_ATTR_CREATE_QP_RECV_CQ_HANDLE); + if (IS_ERR(recv_cq)) + return PTR_ERR(recv_cq); + } + } + + device = pd->device; + break; + default: + return -EINVAL; + } + + ret = uverbs_get_flags32(&attr.create_flags, attrs, + UVERBS_ATTR_CREATE_QP_FLAGS, + IB_UVERBS_QP_CREATE_BLOCK_MULTICAST_LOOPBACK | + IB_UVERBS_QP_CREATE_SCATTER_FCS | + IB_UVERBS_QP_CREATE_CVLAN_STRIPPING | + IB_UVERBS_QP_CREATE_PCI_WRITE_END_PADDING | + IB_UVERBS_QP_CREATE_SQ_SIG_ALL); + if (ret) + return ret; + + ret = check_creation_flags(attr.qp_type, attr.create_flags); + if (ret) + return ret; + + if (uverbs_attr_is_valid(attrs, + UVERBS_ATTR_CREATE_QP_SOURCE_QPN)) { + ret = uverbs_copy_from(&attr.source_qpn, attrs, + UVERBS_ATTR_CREATE_QP_SOURCE_QPN); + if (ret) + return ret; + attr.create_flags |= IB_QP_CREATE_SOURCE_QPN; + } + + srq = uverbs_attr_get_obj(attrs, + UVERBS_ATTR_CREATE_QP_SRQ_HANDLE); + if (!IS_ERR(srq)) { + if ((srq->srq_type == IB_SRQT_XRC && + attr.qp_type != IB_QPT_XRC_TGT) || + (srq->srq_type != IB_SRQT_XRC && + attr.qp_type == IB_QPT_XRC_TGT)) + return -EINVAL; + attr.srq = srq; + } + + obj->uevent.event_file = ib_uverbs_get_async_event(attrs, + UVERBS_ATTR_CREATE_QP_EVENT_FD); + INIT_LIST_HEAD(&obj->uevent.event_list); + INIT_LIST_HEAD(&obj->mcast_list); + obj->uevent.uobject.user_handle = user_handle; + attr.event_handler = ib_uverbs_qp_event_handler; + attr.send_cq = send_cq; + attr.recv_cq = recv_cq; + attr.xrcd = xrcd; + if (attr.create_flags & IB_UVERBS_QP_CREATE_SQ_SIG_ALL) { + /* This creation bit is uverbs one, need to mask before + * calling drivers. It was added to prevent an extra user attr + * only for that when using ioctl. + */ + attr.create_flags &= ~IB_UVERBS_QP_CREATE_SQ_SIG_ALL; + attr.sq_sig_type = IB_SIGNAL_ALL_WR; + } else { + attr.sq_sig_type = IB_SIGNAL_REQ_WR; + } + + set_caps(&attr, &cap, true); + mutex_init(&obj->mcast_lock); + + if (attr.qp_type == IB_QPT_XRC_TGT) + qp = ib_create_qp(pd, &attr); + else + qp = _ib_create_qp(device, pd, &attr, &attrs->driver_udata, + obj); + + if (IS_ERR(qp)) { + ret = PTR_ERR(qp); + goto err_put; + } + + if (attr.qp_type != IB_QPT_XRC_TGT) { + atomic_inc(&pd->usecnt); + if (attr.send_cq) + atomic_inc(&attr.send_cq->usecnt); + if (attr.recv_cq) + atomic_inc(&attr.recv_cq->usecnt); + if (attr.srq) + atomic_inc(&attr.srq->usecnt); + if (attr.rwq_ind_tbl) + atomic_inc(&attr.rwq_ind_tbl->usecnt); + } else { + obj->uxrcd = container_of(xrcd_uobj, struct ib_uxrcd_object, + uobject); + atomic_inc(&obj->uxrcd->refcnt); + /* It is done in _ib_create_qp for other QP types */ + qp->uobject = obj; + } + + obj->uevent.uobject.object = qp; + uverbs_finalize_uobj_create(attrs, UVERBS_ATTR_CREATE_QP_HANDLE); + + if (attr.qp_type != IB_QPT_XRC_TGT) { + ret = ib_create_qp_security(qp, device); + if (ret) + return ret; + } + + set_caps(&attr, &cap, false); + ret = uverbs_copy_to_struct_or_zero(attrs, + UVERBS_ATTR_CREATE_QP_RESP_CAP, &cap, + sizeof(cap)); + if (ret) + return ret; + + ret = uverbs_copy_to(attrs, UVERBS_ATTR_CREATE_QP_RESP_QP_NUM, + &qp->qp_num, + sizeof(qp->qp_num)); + + return ret; +err_put: + if (obj->uevent.event_file) + uverbs_uobject_put(&obj->uevent.event_file->uobj); + return ret; +}; + +DECLARE_UVERBS_NAMED_METHOD( + UVERBS_METHOD_QP_CREATE, + UVERBS_ATTR_IDR(UVERBS_ATTR_CREATE_QP_HANDLE, + UVERBS_OBJECT_QP, + UVERBS_ACCESS_NEW, + UA_MANDATORY), + UVERBS_ATTR_IDR(UVERBS_ATTR_CREATE_QP_XRCD_HANDLE, + UVERBS_OBJECT_XRCD, + UVERBS_ACCESS_READ, + UA_OPTIONAL), + UVERBS_ATTR_IDR(UVERBS_ATTR_CREATE_QP_PD_HANDLE, + UVERBS_OBJECT_PD, + UVERBS_ACCESS_READ, + UA_OPTIONAL), + UVERBS_ATTR_IDR(UVERBS_ATTR_CREATE_QP_SRQ_HANDLE, + UVERBS_OBJECT_SRQ, + UVERBS_ACCESS_READ, + UA_OPTIONAL), + UVERBS_ATTR_IDR(UVERBS_ATTR_CREATE_QP_SEND_CQ_HANDLE, + UVERBS_OBJECT_CQ, + UVERBS_ACCESS_READ, + UA_OPTIONAL), + UVERBS_ATTR_IDR(UVERBS_ATTR_CREATE_QP_RECV_CQ_HANDLE, + UVERBS_OBJECT_CQ, + UVERBS_ACCESS_READ, + UA_OPTIONAL), + UVERBS_ATTR_IDR(UVERBS_ATTR_CREATE_QP_IND_TABLE_HANDLE, + UVERBS_OBJECT_RWQ_IND_TBL, + UVERBS_ACCESS_READ, + UA_OPTIONAL), + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_CREATE_QP_USER_HANDLE, + UVERBS_ATTR_TYPE(u64), + UA_MANDATORY), + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_CREATE_QP_CAP, + UVERBS_ATTR_STRUCT(struct ib_uverbs_qp_cap, + max_inline_data), + UA_MANDATORY), + UVERBS_ATTR_CONST_IN(UVERBS_ATTR_CREATE_QP_TYPE, + enum ib_uverbs_qp_type, + UA_MANDATORY), + UVERBS_ATTR_FLAGS_IN(UVERBS_ATTR_CREATE_QP_FLAGS, + enum ib_uverbs_qp_create_flags, + UA_OPTIONAL), + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_CREATE_QP_SOURCE_QPN, + UVERBS_ATTR_TYPE(u32), + UA_OPTIONAL), + UVERBS_ATTR_FD(UVERBS_ATTR_CREATE_QP_EVENT_FD, + UVERBS_OBJECT_ASYNC_EVENT, + UVERBS_ACCESS_READ, + UA_OPTIONAL), + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_CREATE_QP_RESP_CAP, + UVERBS_ATTR_STRUCT(struct ib_uverbs_qp_cap, + max_inline_data), + UA_MANDATORY), + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_CREATE_QP_RESP_QP_NUM, + UVERBS_ATTR_TYPE(u32), + UA_MANDATORY), + UVERBS_ATTR_UHW()); + +static int UVERBS_HANDLER(UVERBS_METHOD_QP_DESTROY)( + struct uverbs_attr_bundle *attrs) +{ + struct ib_uobject *uobj = + uverbs_attr_get_uobject(attrs, UVERBS_ATTR_DESTROY_QP_HANDLE); + struct ib_uqp_object *obj = + container_of(uobj, struct ib_uqp_object, uevent.uobject); + struct ib_uverbs_destroy_qp_resp resp = { + .events_reported = obj->uevent.events_reported + }; + + return uverbs_copy_to(attrs, UVERBS_ATTR_DESTROY_QP_RESP, &resp, + sizeof(resp)); +} + +DECLARE_UVERBS_NAMED_METHOD( + UVERBS_METHOD_QP_DESTROY, + UVERBS_ATTR_IDR(UVERBS_ATTR_DESTROY_QP_HANDLE, + UVERBS_OBJECT_QP, + UVERBS_ACCESS_DESTROY, + UA_MANDATORY), + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_DESTROY_QP_RESP, + UVERBS_ATTR_TYPE(struct ib_uverbs_destroy_qp_resp), + UA_MANDATORY)); + +DECLARE_UVERBS_NAMED_OBJECT( + UVERBS_OBJECT_QP, + UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_uqp_object), uverbs_free_qp), + &UVERBS_METHOD(UVERBS_METHOD_QP_CREATE), + &UVERBS_METHOD(UVERBS_METHOD_QP_DESTROY)); + +const struct uapi_definition uverbs_def_obj_qp[] = { + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_QP, + UAPI_DEF_OBJ_NEEDS_FN(destroy_qp)), + {} +}; diff --git a/drivers/infiniband/core/uverbs_std_types_srq.c b/drivers/infiniband/core/uverbs_std_types_srq.c new file mode 100644 index 000000000..c0ecbba26 --- /dev/null +++ b/drivers/infiniband/core/uverbs_std_types_srq.c @@ -0,0 +1,234 @@ +// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB +/* + * Copyright (c) 2020, Mellanox Technologies inc. All rights reserved. + */ + +#include <rdma/uverbs_std_types.h> +#include "rdma_core.h" +#include "uverbs.h" + +static int uverbs_free_srq(struct ib_uobject *uobject, + enum rdma_remove_reason why, + struct uverbs_attr_bundle *attrs) +{ + struct ib_srq *srq = uobject->object; + struct ib_uevent_object *uevent = + container_of(uobject, struct ib_uevent_object, uobject); + enum ib_srq_type srq_type = srq->srq_type; + int ret; + + ret = ib_destroy_srq_user(srq, &attrs->driver_udata); + if (ib_is_destroy_retryable(ret, why, uobject)) + return ret; + + if (srq_type == IB_SRQT_XRC) { + struct ib_usrq_object *us = + container_of(uobject, struct ib_usrq_object, + uevent.uobject); + + atomic_dec(&us->uxrcd->refcnt); + } + + ib_uverbs_release_uevent(uevent); + return ret; +} + +static int UVERBS_HANDLER(UVERBS_METHOD_SRQ_CREATE)( + struct uverbs_attr_bundle *attrs) +{ + struct ib_usrq_object *obj = container_of( + uverbs_attr_get_uobject(attrs, UVERBS_ATTR_CREATE_SRQ_HANDLE), + typeof(*obj), uevent.uobject); + struct ib_pd *pd = + uverbs_attr_get_obj(attrs, UVERBS_ATTR_CREATE_SRQ_PD_HANDLE); + struct ib_srq_init_attr attr = {}; + struct ib_uobject *xrcd_uobj; + struct ib_srq *srq; + u64 user_handle; + int ret; + + ret = uverbs_copy_from(&attr.attr.max_sge, attrs, + UVERBS_ATTR_CREATE_SRQ_MAX_SGE); + if (!ret) + ret = uverbs_copy_from(&attr.attr.max_wr, attrs, + UVERBS_ATTR_CREATE_SRQ_MAX_WR); + if (!ret) + ret = uverbs_copy_from(&attr.attr.srq_limit, attrs, + UVERBS_ATTR_CREATE_SRQ_LIMIT); + if (!ret) + ret = uverbs_copy_from(&user_handle, attrs, + UVERBS_ATTR_CREATE_SRQ_USER_HANDLE); + if (!ret) + ret = uverbs_get_const(&attr.srq_type, attrs, + UVERBS_ATTR_CREATE_SRQ_TYPE); + if (ret) + return ret; + + if (ib_srq_has_cq(attr.srq_type)) { + attr.ext.cq = uverbs_attr_get_obj(attrs, + UVERBS_ATTR_CREATE_SRQ_CQ_HANDLE); + if (IS_ERR(attr.ext.cq)) + return PTR_ERR(attr.ext.cq); + } + + switch (attr.srq_type) { + case IB_UVERBS_SRQT_XRC: + xrcd_uobj = uverbs_attr_get_uobject(attrs, + UVERBS_ATTR_CREATE_SRQ_XRCD_HANDLE); + if (IS_ERR(xrcd_uobj)) + return PTR_ERR(xrcd_uobj); + + attr.ext.xrc.xrcd = (struct ib_xrcd *)xrcd_uobj->object; + if (!attr.ext.xrc.xrcd) + return -EINVAL; + obj->uxrcd = container_of(xrcd_uobj, struct ib_uxrcd_object, + uobject); + atomic_inc(&obj->uxrcd->refcnt); + break; + case IB_UVERBS_SRQT_TM: + ret = uverbs_copy_from(&attr.ext.tag_matching.max_num_tags, + attrs, + UVERBS_ATTR_CREATE_SRQ_MAX_NUM_TAGS); + if (ret) + return ret; + break; + case IB_UVERBS_SRQT_BASIC: + break; + default: + return -EINVAL; + } + + obj->uevent.event_file = ib_uverbs_get_async_event(attrs, + UVERBS_ATTR_CREATE_SRQ_EVENT_FD); + INIT_LIST_HEAD(&obj->uevent.event_list); + attr.event_handler = ib_uverbs_srq_event_handler; + obj->uevent.uobject.user_handle = user_handle; + + srq = ib_create_srq_user(pd, &attr, obj, &attrs->driver_udata); + if (IS_ERR(srq)) { + ret = PTR_ERR(srq); + goto err; + } + + obj->uevent.uobject.object = srq; + uverbs_finalize_uobj_create(attrs, UVERBS_ATTR_CREATE_SRQ_HANDLE); + + ret = uverbs_copy_to(attrs, UVERBS_ATTR_CREATE_SRQ_RESP_MAX_WR, + &attr.attr.max_wr, + sizeof(attr.attr.max_wr)); + if (ret) + return ret; + + ret = uverbs_copy_to(attrs, UVERBS_ATTR_CREATE_SRQ_RESP_MAX_SGE, + &attr.attr.max_sge, + sizeof(attr.attr.max_sge)); + if (ret) + return ret; + + if (attr.srq_type == IB_SRQT_XRC) { + ret = uverbs_copy_to(attrs, + UVERBS_ATTR_CREATE_SRQ_RESP_SRQ_NUM, + &srq->ext.xrc.srq_num, + sizeof(srq->ext.xrc.srq_num)); + if (ret) + return ret; + } + + return 0; +err: + if (obj->uevent.event_file) + uverbs_uobject_put(&obj->uevent.event_file->uobj); + if (attr.srq_type == IB_SRQT_XRC) + atomic_dec(&obj->uxrcd->refcnt); + return ret; +}; + +DECLARE_UVERBS_NAMED_METHOD( + UVERBS_METHOD_SRQ_CREATE, + UVERBS_ATTR_IDR(UVERBS_ATTR_CREATE_SRQ_HANDLE, + UVERBS_OBJECT_SRQ, + UVERBS_ACCESS_NEW, + UA_MANDATORY), + UVERBS_ATTR_IDR(UVERBS_ATTR_CREATE_SRQ_PD_HANDLE, + UVERBS_OBJECT_PD, + UVERBS_ACCESS_READ, + UA_MANDATORY), + UVERBS_ATTR_CONST_IN(UVERBS_ATTR_CREATE_SRQ_TYPE, + enum ib_uverbs_srq_type, + UA_MANDATORY), + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_CREATE_SRQ_USER_HANDLE, + UVERBS_ATTR_TYPE(u64), + UA_MANDATORY), + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_CREATE_SRQ_MAX_WR, + UVERBS_ATTR_TYPE(u32), + UA_MANDATORY), + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_CREATE_SRQ_MAX_SGE, + UVERBS_ATTR_TYPE(u32), + UA_MANDATORY), + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_CREATE_SRQ_LIMIT, + UVERBS_ATTR_TYPE(u32), + UA_MANDATORY), + UVERBS_ATTR_IDR(UVERBS_ATTR_CREATE_SRQ_XRCD_HANDLE, + UVERBS_OBJECT_XRCD, + UVERBS_ACCESS_READ, + UA_OPTIONAL), + UVERBS_ATTR_IDR(UVERBS_ATTR_CREATE_SRQ_CQ_HANDLE, + UVERBS_OBJECT_CQ, + UVERBS_ACCESS_READ, + UA_OPTIONAL), + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_CREATE_SRQ_MAX_NUM_TAGS, + UVERBS_ATTR_TYPE(u32), + UA_OPTIONAL), + UVERBS_ATTR_FD(UVERBS_ATTR_CREATE_SRQ_EVENT_FD, + UVERBS_OBJECT_ASYNC_EVENT, + UVERBS_ACCESS_READ, + UA_OPTIONAL), + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_CREATE_SRQ_RESP_MAX_WR, + UVERBS_ATTR_TYPE(u32), + UA_MANDATORY), + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_CREATE_SRQ_RESP_MAX_SGE, + UVERBS_ATTR_TYPE(u32), + UA_MANDATORY), + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_CREATE_SRQ_RESP_SRQ_NUM, + UVERBS_ATTR_TYPE(u32), + UA_OPTIONAL), + UVERBS_ATTR_UHW()); + +static int UVERBS_HANDLER(UVERBS_METHOD_SRQ_DESTROY)( + struct uverbs_attr_bundle *attrs) +{ + struct ib_uobject *uobj = + uverbs_attr_get_uobject(attrs, UVERBS_ATTR_DESTROY_SRQ_HANDLE); + struct ib_usrq_object *obj = + container_of(uobj, struct ib_usrq_object, uevent.uobject); + struct ib_uverbs_destroy_srq_resp resp = { + .events_reported = obj->uevent.events_reported + }; + + return uverbs_copy_to(attrs, UVERBS_ATTR_DESTROY_SRQ_RESP, &resp, + sizeof(resp)); +} + +DECLARE_UVERBS_NAMED_METHOD( + UVERBS_METHOD_SRQ_DESTROY, + UVERBS_ATTR_IDR(UVERBS_ATTR_DESTROY_SRQ_HANDLE, + UVERBS_OBJECT_SRQ, + UVERBS_ACCESS_DESTROY, + UA_MANDATORY), + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_DESTROY_SRQ_RESP, + UVERBS_ATTR_TYPE(struct ib_uverbs_destroy_srq_resp), + UA_MANDATORY)); + +DECLARE_UVERBS_NAMED_OBJECT( + UVERBS_OBJECT_SRQ, + UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_usrq_object), + uverbs_free_srq), + &UVERBS_METHOD(UVERBS_METHOD_SRQ_CREATE), + &UVERBS_METHOD(UVERBS_METHOD_SRQ_DESTROY) +); + +const struct uapi_definition uverbs_def_obj_srq[] = { + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_SRQ, + UAPI_DEF_OBJ_NEEDS_FN(destroy_srq)), + {} +}; diff --git a/drivers/infiniband/core/uverbs_std_types_wq.c b/drivers/infiniband/core/uverbs_std_types_wq.c new file mode 100644 index 000000000..f2e6a6257 --- /dev/null +++ b/drivers/infiniband/core/uverbs_std_types_wq.c @@ -0,0 +1,194 @@ +// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB +/* + * Copyright (c) 2020, Mellanox Technologies inc. All rights reserved. + */ + +#include <rdma/uverbs_std_types.h> +#include "rdma_core.h" +#include "uverbs.h" + +static int uverbs_free_wq(struct ib_uobject *uobject, + enum rdma_remove_reason why, + struct uverbs_attr_bundle *attrs) +{ + struct ib_wq *wq = uobject->object; + struct ib_uwq_object *uwq = + container_of(uobject, struct ib_uwq_object, uevent.uobject); + int ret; + + ret = ib_destroy_wq_user(wq, &attrs->driver_udata); + if (ib_is_destroy_retryable(ret, why, uobject)) + return ret; + + ib_uverbs_release_uevent(&uwq->uevent); + return ret; +} + +static int UVERBS_HANDLER(UVERBS_METHOD_WQ_CREATE)( + struct uverbs_attr_bundle *attrs) +{ + struct ib_uwq_object *obj = container_of( + uverbs_attr_get_uobject(attrs, UVERBS_ATTR_CREATE_WQ_HANDLE), + typeof(*obj), uevent.uobject); + struct ib_pd *pd = + uverbs_attr_get_obj(attrs, UVERBS_ATTR_CREATE_WQ_PD_HANDLE); + struct ib_cq *cq = + uverbs_attr_get_obj(attrs, UVERBS_ATTR_CREATE_WQ_CQ_HANDLE); + struct ib_wq_init_attr wq_init_attr = {}; + struct ib_wq *wq; + u64 user_handle; + int ret; + + ret = uverbs_get_flags32(&wq_init_attr.create_flags, attrs, + UVERBS_ATTR_CREATE_WQ_FLAGS, + IB_UVERBS_WQ_FLAGS_CVLAN_STRIPPING | + IB_UVERBS_WQ_FLAGS_SCATTER_FCS | + IB_UVERBS_WQ_FLAGS_DELAY_DROP | + IB_UVERBS_WQ_FLAGS_PCI_WRITE_END_PADDING); + if (!ret) + ret = uverbs_copy_from(&wq_init_attr.max_sge, attrs, + UVERBS_ATTR_CREATE_WQ_MAX_SGE); + if (!ret) + ret = uverbs_copy_from(&wq_init_attr.max_wr, attrs, + UVERBS_ATTR_CREATE_WQ_MAX_WR); + if (!ret) + ret = uverbs_copy_from(&user_handle, attrs, + UVERBS_ATTR_CREATE_WQ_USER_HANDLE); + if (!ret) + ret = uverbs_get_const(&wq_init_attr.wq_type, attrs, + UVERBS_ATTR_CREATE_WQ_TYPE); + if (ret) + return ret; + + if (wq_init_attr.wq_type != IB_WQT_RQ) + return -EINVAL; + + obj->uevent.event_file = ib_uverbs_get_async_event(attrs, + UVERBS_ATTR_CREATE_WQ_EVENT_FD); + obj->uevent.uobject.user_handle = user_handle; + INIT_LIST_HEAD(&obj->uevent.event_list); + wq_init_attr.event_handler = ib_uverbs_wq_event_handler; + wq_init_attr.wq_context = attrs->ufile; + wq_init_attr.cq = cq; + + wq = pd->device->ops.create_wq(pd, &wq_init_attr, &attrs->driver_udata); + if (IS_ERR(wq)) { + ret = PTR_ERR(wq); + goto err; + } + + obj->uevent.uobject.object = wq; + wq->wq_type = wq_init_attr.wq_type; + wq->cq = cq; + wq->pd = pd; + wq->device = pd->device; + wq->wq_context = wq_init_attr.wq_context; + atomic_set(&wq->usecnt, 0); + atomic_inc(&pd->usecnt); + atomic_inc(&cq->usecnt); + wq->uobject = obj; + uverbs_finalize_uobj_create(attrs, UVERBS_ATTR_CREATE_WQ_HANDLE); + + ret = uverbs_copy_to(attrs, UVERBS_ATTR_CREATE_WQ_RESP_MAX_WR, + &wq_init_attr.max_wr, + sizeof(wq_init_attr.max_wr)); + if (ret) + return ret; + + ret = uverbs_copy_to(attrs, UVERBS_ATTR_CREATE_WQ_RESP_MAX_SGE, + &wq_init_attr.max_sge, + sizeof(wq_init_attr.max_sge)); + if (ret) + return ret; + + ret = uverbs_copy_to(attrs, UVERBS_ATTR_CREATE_WQ_RESP_WQ_NUM, + &wq->wq_num, + sizeof(wq->wq_num)); + return ret; + +err: + if (obj->uevent.event_file) + uverbs_uobject_put(&obj->uevent.event_file->uobj); + return ret; +}; + +DECLARE_UVERBS_NAMED_METHOD( + UVERBS_METHOD_WQ_CREATE, + UVERBS_ATTR_IDR(UVERBS_ATTR_CREATE_WQ_HANDLE, + UVERBS_OBJECT_WQ, + UVERBS_ACCESS_NEW, + UA_MANDATORY), + UVERBS_ATTR_IDR(UVERBS_ATTR_CREATE_WQ_PD_HANDLE, + UVERBS_OBJECT_PD, + UVERBS_ACCESS_READ, + UA_MANDATORY), + UVERBS_ATTR_CONST_IN(UVERBS_ATTR_CREATE_WQ_TYPE, + enum ib_wq_type, + UA_MANDATORY), + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_CREATE_WQ_USER_HANDLE, + UVERBS_ATTR_TYPE(u64), + UA_MANDATORY), + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_CREATE_WQ_MAX_WR, + UVERBS_ATTR_TYPE(u32), + UA_MANDATORY), + UVERBS_ATTR_PTR_IN(UVERBS_ATTR_CREATE_WQ_MAX_SGE, + UVERBS_ATTR_TYPE(u32), + UA_MANDATORY), + UVERBS_ATTR_FLAGS_IN(UVERBS_ATTR_CREATE_WQ_FLAGS, + enum ib_uverbs_wq_flags, + UA_MANDATORY), + UVERBS_ATTR_IDR(UVERBS_ATTR_CREATE_WQ_CQ_HANDLE, + UVERBS_OBJECT_CQ, + UVERBS_ACCESS_READ, + UA_OPTIONAL), + UVERBS_ATTR_FD(UVERBS_ATTR_CREATE_WQ_EVENT_FD, + UVERBS_OBJECT_ASYNC_EVENT, + UVERBS_ACCESS_READ, + UA_OPTIONAL), + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_CREATE_WQ_RESP_MAX_WR, + UVERBS_ATTR_TYPE(u32), + UA_MANDATORY), + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_CREATE_WQ_RESP_MAX_SGE, + UVERBS_ATTR_TYPE(u32), + UA_MANDATORY), + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_CREATE_WQ_RESP_WQ_NUM, + UVERBS_ATTR_TYPE(u32), + UA_OPTIONAL), + UVERBS_ATTR_UHW()); + +static int UVERBS_HANDLER(UVERBS_METHOD_WQ_DESTROY)( + struct uverbs_attr_bundle *attrs) +{ + struct ib_uobject *uobj = + uverbs_attr_get_uobject(attrs, UVERBS_ATTR_DESTROY_WQ_HANDLE); + struct ib_uwq_object *obj = + container_of(uobj, struct ib_uwq_object, uevent.uobject); + + return uverbs_copy_to(attrs, UVERBS_ATTR_DESTROY_WQ_RESP, + &obj->uevent.events_reported, + sizeof(obj->uevent.events_reported)); +} + +DECLARE_UVERBS_NAMED_METHOD( + UVERBS_METHOD_WQ_DESTROY, + UVERBS_ATTR_IDR(UVERBS_ATTR_DESTROY_WQ_HANDLE, + UVERBS_OBJECT_WQ, + UVERBS_ACCESS_DESTROY, + UA_MANDATORY), + UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_DESTROY_WQ_RESP, + UVERBS_ATTR_TYPE(u32), + UA_MANDATORY)); + + +DECLARE_UVERBS_NAMED_OBJECT( + UVERBS_OBJECT_WQ, + UVERBS_TYPE_ALLOC_IDR_SZ(sizeof(struct ib_uwq_object), uverbs_free_wq), + &UVERBS_METHOD(UVERBS_METHOD_WQ_CREATE), + &UVERBS_METHOD(UVERBS_METHOD_WQ_DESTROY) +); + +const struct uapi_definition uverbs_def_obj_wq[] = { + UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_WQ, + UAPI_DEF_OBJ_NEEDS_FN(destroy_wq)), + {} +}; diff --git a/drivers/infiniband/core/uverbs_uapi.c b/drivers/infiniband/core/uverbs_uapi.c new file mode 100644 index 000000000..91dbcb3c2 --- /dev/null +++ b/drivers/infiniband/core/uverbs_uapi.c @@ -0,0 +1,737 @@ +// SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB +/* + * Copyright (c) 2017, Mellanox Technologies inc. All rights reserved. + */ +#include <rdma/uverbs_ioctl.h> +#include <rdma/rdma_user_ioctl.h> +#include <linux/bitops.h> +#include "rdma_core.h" +#include "uverbs.h" + +static int ib_uverbs_notsupp(struct uverbs_attr_bundle *attrs) +{ + return -EOPNOTSUPP; +} + +static void *uapi_add_elm(struct uverbs_api *uapi, u32 key, size_t alloc_size) +{ + void *elm; + int rc; + + if (key == UVERBS_API_KEY_ERR) + return ERR_PTR(-EOVERFLOW); + + elm = kzalloc(alloc_size, GFP_KERNEL); + if (!elm) + return ERR_PTR(-ENOMEM); + rc = radix_tree_insert(&uapi->radix, key, elm); + if (rc) { + kfree(elm); + return ERR_PTR(rc); + } + + return elm; +} + +static void *uapi_add_get_elm(struct uverbs_api *uapi, u32 key, + size_t alloc_size, bool *exists) +{ + void *elm; + + elm = uapi_add_elm(uapi, key, alloc_size); + if (!IS_ERR(elm)) { + *exists = false; + return elm; + } + + if (elm != ERR_PTR(-EEXIST)) + return elm; + + elm = radix_tree_lookup(&uapi->radix, key); + if (WARN_ON(!elm)) + return ERR_PTR(-EINVAL); + *exists = true; + return elm; +} + +static int uapi_create_write(struct uverbs_api *uapi, + struct ib_device *ibdev, + const struct uapi_definition *def, + u32 obj_key, + u32 *cur_method_key) +{ + struct uverbs_api_write_method *method_elm; + u32 method_key = obj_key; + bool exists; + + if (def->write.is_ex) + method_key |= uapi_key_write_ex_method(def->write.command_num); + else + method_key |= uapi_key_write_method(def->write.command_num); + + method_elm = uapi_add_get_elm(uapi, method_key, sizeof(*method_elm), + &exists); + if (IS_ERR(method_elm)) + return PTR_ERR(method_elm); + + if (WARN_ON(exists && (def->write.is_ex != method_elm->is_ex))) + return -EINVAL; + + method_elm->is_ex = def->write.is_ex; + method_elm->handler = def->func_write; + if (def->write.is_ex) + method_elm->disabled = !(ibdev->uverbs_ex_cmd_mask & + BIT_ULL(def->write.command_num)); + else + method_elm->disabled = !(ibdev->uverbs_cmd_mask & + BIT_ULL(def->write.command_num)); + + if (!def->write.is_ex && def->func_write) { + method_elm->has_udata = def->write.has_udata; + method_elm->has_resp = def->write.has_resp; + method_elm->req_size = def->write.req_size; + method_elm->resp_size = def->write.resp_size; + } + + *cur_method_key = method_key; + return 0; +} + +static int uapi_merge_method(struct uverbs_api *uapi, + struct uverbs_api_object *obj_elm, u32 obj_key, + const struct uverbs_method_def *method, + bool is_driver) +{ + u32 method_key = obj_key | uapi_key_ioctl_method(method->id); + struct uverbs_api_ioctl_method *method_elm; + unsigned int i; + bool exists; + + if (!method->attrs) + return 0; + + method_elm = uapi_add_get_elm(uapi, method_key, sizeof(*method_elm), + &exists); + if (IS_ERR(method_elm)) + return PTR_ERR(method_elm); + if (exists) { + /* + * This occurs when a driver uses ADD_UVERBS_ATTRIBUTES_SIMPLE + */ + if (WARN_ON(method->handler)) + return -EINVAL; + } else { + WARN_ON(!method->handler); + rcu_assign_pointer(method_elm->handler, method->handler); + if (method->handler != uverbs_destroy_def_handler) + method_elm->driver_method = is_driver; + } + + for (i = 0; i != method->num_attrs; i++) { + const struct uverbs_attr_def *attr = (*method->attrs)[i]; + struct uverbs_api_attr *attr_slot; + + if (!attr) + continue; + + /* + * ENUM_IN contains the 'ids' pointer to the driver's .rodata, + * so if it is specified by a driver then it always makes this + * into a driver method. + */ + if (attr->attr.type == UVERBS_ATTR_TYPE_ENUM_IN) + method_elm->driver_method |= is_driver; + + /* + * Like other uobject based things we only support a single + * uobject being NEW'd or DESTROY'd + */ + if (attr->attr.type == UVERBS_ATTR_TYPE_IDRS_ARRAY) { + u8 access = attr->attr.u2.objs_arr.access; + + if (WARN_ON(access == UVERBS_ACCESS_NEW || + access == UVERBS_ACCESS_DESTROY)) + return -EINVAL; + } + + attr_slot = + uapi_add_elm(uapi, method_key | uapi_key_attr(attr->id), + sizeof(*attr_slot)); + /* Attributes are not allowed to be modified by drivers */ + if (IS_ERR(attr_slot)) + return PTR_ERR(attr_slot); + + attr_slot->spec = attr->attr; + } + + return 0; +} + +static int uapi_merge_obj_tree(struct uverbs_api *uapi, + const struct uverbs_object_def *obj, + bool is_driver) +{ + struct uverbs_api_object *obj_elm; + unsigned int i; + u32 obj_key; + bool exists; + int rc; + + obj_key = uapi_key_obj(obj->id); + obj_elm = uapi_add_get_elm(uapi, obj_key, sizeof(*obj_elm), &exists); + if (IS_ERR(obj_elm)) + return PTR_ERR(obj_elm); + + if (obj->type_attrs) { + if (WARN_ON(obj_elm->type_attrs)) + return -EINVAL; + + obj_elm->id = obj->id; + obj_elm->type_attrs = obj->type_attrs; + obj_elm->type_class = obj->type_attrs->type_class; + /* + * Today drivers are only permitted to use idr_class and + * fd_class types. We can revoke the IDR types during + * disassociation, and the FD types require the driver to use + * struct file_operations.owner to prevent the driver module + * code from unloading while the file is open. This provides + * enough safety that uverbs_uobject_fd_release() will + * continue to work. Drivers using FD are responsible to + * handle disassociation of the device on their own. + */ + if (WARN_ON(is_driver && + obj->type_attrs->type_class != &uverbs_idr_class && + obj->type_attrs->type_class != &uverbs_fd_class)) + return -EINVAL; + } + + if (!obj->methods) + return 0; + + for (i = 0; i != obj->num_methods; i++) { + const struct uverbs_method_def *method = (*obj->methods)[i]; + + if (!method) + continue; + + rc = uapi_merge_method(uapi, obj_elm, obj_key, method, + is_driver); + if (rc) + return rc; + } + + return 0; +} + +static int uapi_disable_elm(struct uverbs_api *uapi, + const struct uapi_definition *def, + u32 obj_key, + u32 method_key) +{ + bool exists; + + if (def->scope == UAPI_SCOPE_OBJECT) { + struct uverbs_api_object *obj_elm; + + obj_elm = uapi_add_get_elm( + uapi, obj_key, sizeof(*obj_elm), &exists); + if (IS_ERR(obj_elm)) + return PTR_ERR(obj_elm); + obj_elm->disabled = 1; + return 0; + } + + if (def->scope == UAPI_SCOPE_METHOD && + uapi_key_is_ioctl_method(method_key)) { + struct uverbs_api_ioctl_method *method_elm; + + method_elm = uapi_add_get_elm(uapi, method_key, + sizeof(*method_elm), &exists); + if (IS_ERR(method_elm)) + return PTR_ERR(method_elm); + method_elm->disabled = 1; + return 0; + } + + if (def->scope == UAPI_SCOPE_METHOD && + (uapi_key_is_write_method(method_key) || + uapi_key_is_write_ex_method(method_key))) { + struct uverbs_api_write_method *write_elm; + + write_elm = uapi_add_get_elm(uapi, method_key, + sizeof(*write_elm), &exists); + if (IS_ERR(write_elm)) + return PTR_ERR(write_elm); + write_elm->disabled = 1; + return 0; + } + + WARN_ON(true); + return -EINVAL; +} + +static int uapi_merge_def(struct uverbs_api *uapi, struct ib_device *ibdev, + const struct uapi_definition *def_list, + bool is_driver) +{ + const struct uapi_definition *def = def_list; + u32 cur_obj_key = UVERBS_API_KEY_ERR; + u32 cur_method_key = UVERBS_API_KEY_ERR; + bool exists; + int rc; + + if (!def_list) + return 0; + + for (;; def++) { + switch ((enum uapi_definition_kind)def->kind) { + case UAPI_DEF_CHAIN: + rc = uapi_merge_def(uapi, ibdev, def->chain, is_driver); + if (rc) + return rc; + continue; + + case UAPI_DEF_CHAIN_OBJ_TREE: + if (WARN_ON(def->object_start.object_id != + def->chain_obj_tree->id)) + return -EINVAL; + + cur_obj_key = uapi_key_obj(def->object_start.object_id); + rc = uapi_merge_obj_tree(uapi, def->chain_obj_tree, + is_driver); + if (rc) + return rc; + continue; + + case UAPI_DEF_END: + return 0; + + case UAPI_DEF_IS_SUPPORTED_DEV_FN: { + void **ibdev_fn = + (void *)(&ibdev->ops) + def->needs_fn_offset; + + if (*ibdev_fn) + continue; + rc = uapi_disable_elm( + uapi, def, cur_obj_key, cur_method_key); + if (rc) + return rc; + continue; + } + + case UAPI_DEF_IS_SUPPORTED_FUNC: + if (def->func_is_supported(ibdev)) + continue; + rc = uapi_disable_elm( + uapi, def, cur_obj_key, cur_method_key); + if (rc) + return rc; + continue; + + case UAPI_DEF_OBJECT_START: { + struct uverbs_api_object *obj_elm; + + cur_obj_key = uapi_key_obj(def->object_start.object_id); + obj_elm = uapi_add_get_elm(uapi, cur_obj_key, + sizeof(*obj_elm), &exists); + if (IS_ERR(obj_elm)) + return PTR_ERR(obj_elm); + continue; + } + + case UAPI_DEF_WRITE: + rc = uapi_create_write( + uapi, ibdev, def, cur_obj_key, &cur_method_key); + if (rc) + return rc; + continue; + } + WARN_ON(true); + return -EINVAL; + } +} + +static int +uapi_finalize_ioctl_method(struct uverbs_api *uapi, + struct uverbs_api_ioctl_method *method_elm, + u32 method_key) +{ + struct radix_tree_iter iter; + unsigned int num_attrs = 0; + unsigned int max_bkey = 0; + bool single_uobj = false; + void __rcu **slot; + + method_elm->destroy_bkey = UVERBS_API_ATTR_BKEY_LEN; + radix_tree_for_each_slot (slot, &uapi->radix, &iter, + uapi_key_attrs_start(method_key)) { + struct uverbs_api_attr *elm = + rcu_dereference_protected(*slot, true); + u32 attr_key = iter.index & UVERBS_API_ATTR_KEY_MASK; + u32 attr_bkey = uapi_bkey_attr(attr_key); + u8 type = elm->spec.type; + + if (uapi_key_attr_to_ioctl_method(iter.index) != + uapi_key_attr_to_ioctl_method(method_key)) + break; + + if (elm->spec.mandatory) + __set_bit(attr_bkey, method_elm->attr_mandatory); + + if (elm->spec.is_udata) + method_elm->has_udata = true; + + if (type == UVERBS_ATTR_TYPE_IDR || + type == UVERBS_ATTR_TYPE_FD) { + u8 access = elm->spec.u.obj.access; + + /* + * Verbs specs may only have one NEW/DESTROY, we don't + * have the infrastructure to abort multiple NEW's or + * cope with multiple DESTROY failure. + */ + if (access == UVERBS_ACCESS_NEW || + access == UVERBS_ACCESS_DESTROY) { + if (WARN_ON(single_uobj)) + return -EINVAL; + + single_uobj = true; + if (WARN_ON(!elm->spec.mandatory)) + return -EINVAL; + } + + if (access == UVERBS_ACCESS_DESTROY) + method_elm->destroy_bkey = attr_bkey; + } + + max_bkey = max(max_bkey, attr_bkey); + num_attrs++; + } + + method_elm->key_bitmap_len = max_bkey + 1; + WARN_ON(method_elm->key_bitmap_len > UVERBS_API_ATTR_BKEY_LEN); + + uapi_compute_bundle_size(method_elm, num_attrs); + return 0; +} + +static int uapi_finalize(struct uverbs_api *uapi) +{ + const struct uverbs_api_write_method **data; + unsigned long max_write_ex = 0; + unsigned long max_write = 0; + struct radix_tree_iter iter; + void __rcu **slot; + int rc; + int i; + + radix_tree_for_each_slot (slot, &uapi->radix, &iter, 0) { + struct uverbs_api_ioctl_method *method_elm = + rcu_dereference_protected(*slot, true); + + if (uapi_key_is_ioctl_method(iter.index)) { + rc = uapi_finalize_ioctl_method(uapi, method_elm, + iter.index); + if (rc) + return rc; + } + + if (uapi_key_is_write_method(iter.index)) + max_write = max(max_write, + iter.index & UVERBS_API_ATTR_KEY_MASK); + if (uapi_key_is_write_ex_method(iter.index)) + max_write_ex = + max(max_write_ex, + iter.index & UVERBS_API_ATTR_KEY_MASK); + } + + uapi->notsupp_method.handler = ib_uverbs_notsupp; + uapi->num_write = max_write + 1; + uapi->num_write_ex = max_write_ex + 1; + data = kmalloc_array(uapi->num_write + uapi->num_write_ex, + sizeof(*uapi->write_methods), GFP_KERNEL); + if (!data) + return -ENOMEM; + + for (i = 0; i != uapi->num_write + uapi->num_write_ex; i++) + data[i] = &uapi->notsupp_method; + uapi->write_methods = data; + uapi->write_ex_methods = data + uapi->num_write; + + radix_tree_for_each_slot (slot, &uapi->radix, &iter, 0) { + if (uapi_key_is_write_method(iter.index)) + uapi->write_methods[iter.index & + UVERBS_API_ATTR_KEY_MASK] = + rcu_dereference_protected(*slot, true); + if (uapi_key_is_write_ex_method(iter.index)) + uapi->write_ex_methods[iter.index & + UVERBS_API_ATTR_KEY_MASK] = + rcu_dereference_protected(*slot, true); + } + + return 0; +} + +static void uapi_remove_range(struct uverbs_api *uapi, u32 start, u32 last) +{ + struct radix_tree_iter iter; + void __rcu **slot; + + radix_tree_for_each_slot (slot, &uapi->radix, &iter, start) { + if (iter.index > last) + return; + kfree(rcu_dereference_protected(*slot, true)); + radix_tree_iter_delete(&uapi->radix, &iter, slot); + } +} + +static void uapi_remove_object(struct uverbs_api *uapi, u32 obj_key) +{ + uapi_remove_range(uapi, obj_key, + obj_key | UVERBS_API_METHOD_KEY_MASK | + UVERBS_API_ATTR_KEY_MASK); +} + +static void uapi_remove_method(struct uverbs_api *uapi, u32 method_key) +{ + uapi_remove_range(uapi, method_key, + method_key | UVERBS_API_ATTR_KEY_MASK); +} + + +static u32 uapi_get_obj_id(struct uverbs_attr_spec *spec) +{ + if (spec->type == UVERBS_ATTR_TYPE_IDR || + spec->type == UVERBS_ATTR_TYPE_FD) + return spec->u.obj.obj_type; + if (spec->type == UVERBS_ATTR_TYPE_IDRS_ARRAY) + return spec->u2.objs_arr.obj_type; + return UVERBS_API_KEY_ERR; +} + +static void uapi_key_okay(u32 key) +{ + unsigned int count = 0; + + if (uapi_key_is_object(key)) + count++; + if (uapi_key_is_ioctl_method(key)) + count++; + if (uapi_key_is_write_method(key)) + count++; + if (uapi_key_is_write_ex_method(key)) + count++; + if (uapi_key_is_attr(key)) + count++; + WARN(count != 1, "Bad count %d key=%x", count, key); +} + +static void uapi_finalize_disable(struct uverbs_api *uapi) +{ + struct radix_tree_iter iter; + u32 starting_key = 0; + bool scan_again = false; + void __rcu **slot; + +again: + radix_tree_for_each_slot (slot, &uapi->radix, &iter, starting_key) { + uapi_key_okay(iter.index); + + if (uapi_key_is_object(iter.index)) { + struct uverbs_api_object *obj_elm = + rcu_dereference_protected(*slot, true); + + if (obj_elm->disabled) { + /* Have to check all the attrs again */ + scan_again = true; + starting_key = iter.index; + uapi_remove_object(uapi, iter.index); + goto again; + } + continue; + } + + if (uapi_key_is_ioctl_method(iter.index)) { + struct uverbs_api_ioctl_method *method_elm = + rcu_dereference_protected(*slot, true); + + if (method_elm->disabled) { + starting_key = iter.index; + uapi_remove_method(uapi, iter.index); + goto again; + } + continue; + } + + if (uapi_key_is_write_method(iter.index) || + uapi_key_is_write_ex_method(iter.index)) { + struct uverbs_api_write_method *method_elm = + rcu_dereference_protected(*slot, true); + + if (method_elm->disabled) { + kfree(method_elm); + radix_tree_iter_delete(&uapi->radix, &iter, slot); + } + continue; + } + + if (uapi_key_is_attr(iter.index)) { + struct uverbs_api_attr *attr_elm = + rcu_dereference_protected(*slot, true); + const struct uverbs_api_object *tmp_obj; + u32 obj_key; + + /* + * If the method has a mandatory object handle + * attribute which relies on an object which is not + * present then the entire method is uncallable. + */ + if (!attr_elm->spec.mandatory) + continue; + obj_key = uapi_get_obj_id(&attr_elm->spec); + if (obj_key == UVERBS_API_KEY_ERR) + continue; + tmp_obj = uapi_get_object(uapi, obj_key); + if (IS_ERR(tmp_obj)) { + if (PTR_ERR(tmp_obj) == -ENOMSG) + continue; + } else { + if (!tmp_obj->disabled) + continue; + } + + starting_key = iter.index; + uapi_remove_method( + uapi, + iter.index & (UVERBS_API_OBJ_KEY_MASK | + UVERBS_API_METHOD_KEY_MASK)); + goto again; + } + + WARN_ON(false); + } + + if (!scan_again) + return; + scan_again = false; + starting_key = 0; + goto again; +} + +void uverbs_destroy_api(struct uverbs_api *uapi) +{ + if (!uapi) + return; + + uapi_remove_range(uapi, 0, U32_MAX); + kfree(uapi->write_methods); + kfree(uapi); +} + +static const struct uapi_definition uverbs_core_api[] = { + UAPI_DEF_CHAIN(uverbs_def_obj_async_fd), + UAPI_DEF_CHAIN(uverbs_def_obj_counters), + UAPI_DEF_CHAIN(uverbs_def_obj_cq), + UAPI_DEF_CHAIN(uverbs_def_obj_device), + UAPI_DEF_CHAIN(uverbs_def_obj_dm), + UAPI_DEF_CHAIN(uverbs_def_obj_flow_action), + UAPI_DEF_CHAIN(uverbs_def_obj_intf), + UAPI_DEF_CHAIN(uverbs_def_obj_mr), + UAPI_DEF_CHAIN(uverbs_def_obj_qp), + UAPI_DEF_CHAIN(uverbs_def_obj_srq), + UAPI_DEF_CHAIN(uverbs_def_obj_wq), + UAPI_DEF_CHAIN(uverbs_def_write_intf), + {}, +}; + +struct uverbs_api *uverbs_alloc_api(struct ib_device *ibdev) +{ + struct uverbs_api *uapi; + int rc; + + uapi = kzalloc(sizeof(*uapi), GFP_KERNEL); + if (!uapi) + return ERR_PTR(-ENOMEM); + + INIT_RADIX_TREE(&uapi->radix, GFP_KERNEL); + uapi->driver_id = ibdev->ops.driver_id; + + rc = uapi_merge_def(uapi, ibdev, uverbs_core_api, false); + if (rc) + goto err; + rc = uapi_merge_def(uapi, ibdev, ibdev->driver_def, true); + if (rc) + goto err; + + uapi_finalize_disable(uapi); + rc = uapi_finalize(uapi); + if (rc) + goto err; + + return uapi; +err: + if (rc != -ENOMEM) + dev_err(&ibdev->dev, + "Setup of uverbs_api failed, kernel parsing tree description is not valid (%d)??\n", + rc); + + uverbs_destroy_api(uapi); + return ERR_PTR(rc); +} + +/* + * The pre version is done before destroying the HW objects, it only blocks + * off method access. All methods that require the ib_dev or the module data + * must test one of these assignments prior to continuing. + */ +void uverbs_disassociate_api_pre(struct ib_uverbs_device *uverbs_dev) +{ + struct uverbs_api *uapi = uverbs_dev->uapi; + struct radix_tree_iter iter; + void __rcu **slot; + + rcu_assign_pointer(uverbs_dev->ib_dev, NULL); + + radix_tree_for_each_slot (slot, &uapi->radix, &iter, 0) { + if (uapi_key_is_ioctl_method(iter.index)) { + struct uverbs_api_ioctl_method *method_elm = + rcu_dereference_protected(*slot, true); + + if (method_elm->driver_method) + rcu_assign_pointer(method_elm->handler, NULL); + } + } + + synchronize_srcu(&uverbs_dev->disassociate_srcu); +} + +/* + * Called when a driver disassociates from the ib_uverbs_device. The + * assumption is that the driver module will unload after. Replace everything + * related to the driver with NULL as a safety measure. + */ +void uverbs_disassociate_api(struct uverbs_api *uapi) +{ + struct radix_tree_iter iter; + void __rcu **slot; + + radix_tree_for_each_slot (slot, &uapi->radix, &iter, 0) { + if (uapi_key_is_object(iter.index)) { + struct uverbs_api_object *object_elm = + rcu_dereference_protected(*slot, true); + + /* + * Some type_attrs are in the driver module. We don't + * bother to keep track of which since there should be + * no use of this after disassociate. + */ + object_elm->type_attrs = NULL; + } else if (uapi_key_is_attr(iter.index)) { + struct uverbs_api_attr *elm = + rcu_dereference_protected(*slot, true); + + if (elm->spec.type == UVERBS_ATTR_TYPE_ENUM_IN) + elm->spec.u2.enum_def.ids = NULL; + } + } +} diff --git a/drivers/infiniband/core/verbs.c b/drivers/infiniband/core/verbs.c new file mode 100644 index 000000000..4fcabe5a8 --- /dev/null +++ b/drivers/infiniband/core/verbs.c @@ -0,0 +1,2935 @@ +/* + * Copyright (c) 2004 Mellanox Technologies Ltd. All rights reserved. + * Copyright (c) 2004 Infinicon Corporation. All rights reserved. + * Copyright (c) 2004 Intel Corporation. All rights reserved. + * Copyright (c) 2004 Topspin Corporation. All rights reserved. + * Copyright (c) 2004 Voltaire Corporation. All rights reserved. + * Copyright (c) 2005 Sun Microsystems, Inc. All rights reserved. + * Copyright (c) 2005, 2006 Cisco Systems. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#include <linux/errno.h> +#include <linux/err.h> +#include <linux/export.h> +#include <linux/string.h> +#include <linux/slab.h> +#include <linux/in.h> +#include <linux/in6.h> +#include <net/addrconf.h> +#include <linux/security.h> + +#include <rdma/ib_verbs.h> +#include <rdma/ib_cache.h> +#include <rdma/ib_addr.h> +#include <rdma/rw.h> +#include <rdma/lag.h> + +#include "core_priv.h" +#include <trace/events/rdma_core.h> + +static int ib_resolve_eth_dmac(struct ib_device *device, + struct rdma_ah_attr *ah_attr); + +static const char * const ib_events[] = { + [IB_EVENT_CQ_ERR] = "CQ error", + [IB_EVENT_QP_FATAL] = "QP fatal error", + [IB_EVENT_QP_REQ_ERR] = "QP request error", + [IB_EVENT_QP_ACCESS_ERR] = "QP access error", + [IB_EVENT_COMM_EST] = "communication established", + [IB_EVENT_SQ_DRAINED] = "send queue drained", + [IB_EVENT_PATH_MIG] = "path migration successful", + [IB_EVENT_PATH_MIG_ERR] = "path migration error", + [IB_EVENT_DEVICE_FATAL] = "device fatal error", + [IB_EVENT_PORT_ACTIVE] = "port active", + [IB_EVENT_PORT_ERR] = "port error", + [IB_EVENT_LID_CHANGE] = "LID change", + [IB_EVENT_PKEY_CHANGE] = "P_key change", + [IB_EVENT_SM_CHANGE] = "SM change", + [IB_EVENT_SRQ_ERR] = "SRQ error", + [IB_EVENT_SRQ_LIMIT_REACHED] = "SRQ limit reached", + [IB_EVENT_QP_LAST_WQE_REACHED] = "last WQE reached", + [IB_EVENT_CLIENT_REREGISTER] = "client reregister", + [IB_EVENT_GID_CHANGE] = "GID changed", +}; + +const char *__attribute_const__ ib_event_msg(enum ib_event_type event) +{ + size_t index = event; + + return (index < ARRAY_SIZE(ib_events) && ib_events[index]) ? + ib_events[index] : "unrecognized event"; +} +EXPORT_SYMBOL(ib_event_msg); + +static const char * const wc_statuses[] = { + [IB_WC_SUCCESS] = "success", + [IB_WC_LOC_LEN_ERR] = "local length error", + [IB_WC_LOC_QP_OP_ERR] = "local QP operation error", + [IB_WC_LOC_EEC_OP_ERR] = "local EE context operation error", + [IB_WC_LOC_PROT_ERR] = "local protection error", + [IB_WC_WR_FLUSH_ERR] = "WR flushed", + [IB_WC_MW_BIND_ERR] = "memory management operation error", + [IB_WC_BAD_RESP_ERR] = "bad response error", + [IB_WC_LOC_ACCESS_ERR] = "local access error", + [IB_WC_REM_INV_REQ_ERR] = "invalid request error", + [IB_WC_REM_ACCESS_ERR] = "remote access error", + [IB_WC_REM_OP_ERR] = "remote operation error", + [IB_WC_RETRY_EXC_ERR] = "transport retry counter exceeded", + [IB_WC_RNR_RETRY_EXC_ERR] = "RNR retry counter exceeded", + [IB_WC_LOC_RDD_VIOL_ERR] = "local RDD violation error", + [IB_WC_REM_INV_RD_REQ_ERR] = "remote invalid RD request", + [IB_WC_REM_ABORT_ERR] = "operation aborted", + [IB_WC_INV_EECN_ERR] = "invalid EE context number", + [IB_WC_INV_EEC_STATE_ERR] = "invalid EE context state", + [IB_WC_FATAL_ERR] = "fatal error", + [IB_WC_RESP_TIMEOUT_ERR] = "response timeout error", + [IB_WC_GENERAL_ERR] = "general error", +}; + +const char *__attribute_const__ ib_wc_status_msg(enum ib_wc_status status) +{ + size_t index = status; + + return (index < ARRAY_SIZE(wc_statuses) && wc_statuses[index]) ? + wc_statuses[index] : "unrecognized status"; +} +EXPORT_SYMBOL(ib_wc_status_msg); + +__attribute_const__ int ib_rate_to_mult(enum ib_rate rate) +{ + switch (rate) { + case IB_RATE_2_5_GBPS: return 1; + case IB_RATE_5_GBPS: return 2; + case IB_RATE_10_GBPS: return 4; + case IB_RATE_20_GBPS: return 8; + case IB_RATE_30_GBPS: return 12; + case IB_RATE_40_GBPS: return 16; + case IB_RATE_60_GBPS: return 24; + case IB_RATE_80_GBPS: return 32; + case IB_RATE_120_GBPS: return 48; + case IB_RATE_14_GBPS: return 6; + case IB_RATE_56_GBPS: return 22; + case IB_RATE_112_GBPS: return 45; + case IB_RATE_168_GBPS: return 67; + case IB_RATE_25_GBPS: return 10; + case IB_RATE_100_GBPS: return 40; + case IB_RATE_200_GBPS: return 80; + case IB_RATE_300_GBPS: return 120; + case IB_RATE_28_GBPS: return 11; + case IB_RATE_50_GBPS: return 20; + case IB_RATE_400_GBPS: return 160; + case IB_RATE_600_GBPS: return 240; + default: return -1; + } +} +EXPORT_SYMBOL(ib_rate_to_mult); + +__attribute_const__ enum ib_rate mult_to_ib_rate(int mult) +{ + switch (mult) { + case 1: return IB_RATE_2_5_GBPS; + case 2: return IB_RATE_5_GBPS; + case 4: return IB_RATE_10_GBPS; + case 8: return IB_RATE_20_GBPS; + case 12: return IB_RATE_30_GBPS; + case 16: return IB_RATE_40_GBPS; + case 24: return IB_RATE_60_GBPS; + case 32: return IB_RATE_80_GBPS; + case 48: return IB_RATE_120_GBPS; + case 6: return IB_RATE_14_GBPS; + case 22: return IB_RATE_56_GBPS; + case 45: return IB_RATE_112_GBPS; + case 67: return IB_RATE_168_GBPS; + case 10: return IB_RATE_25_GBPS; + case 40: return IB_RATE_100_GBPS; + case 80: return IB_RATE_200_GBPS; + case 120: return IB_RATE_300_GBPS; + case 11: return IB_RATE_28_GBPS; + case 20: return IB_RATE_50_GBPS; + case 160: return IB_RATE_400_GBPS; + case 240: return IB_RATE_600_GBPS; + default: return IB_RATE_PORT_CURRENT; + } +} +EXPORT_SYMBOL(mult_to_ib_rate); + +__attribute_const__ int ib_rate_to_mbps(enum ib_rate rate) +{ + switch (rate) { + case IB_RATE_2_5_GBPS: return 2500; + case IB_RATE_5_GBPS: return 5000; + case IB_RATE_10_GBPS: return 10000; + case IB_RATE_20_GBPS: return 20000; + case IB_RATE_30_GBPS: return 30000; + case IB_RATE_40_GBPS: return 40000; + case IB_RATE_60_GBPS: return 60000; + case IB_RATE_80_GBPS: return 80000; + case IB_RATE_120_GBPS: return 120000; + case IB_RATE_14_GBPS: return 14062; + case IB_RATE_56_GBPS: return 56250; + case IB_RATE_112_GBPS: return 112500; + case IB_RATE_168_GBPS: return 168750; + case IB_RATE_25_GBPS: return 25781; + case IB_RATE_100_GBPS: return 103125; + case IB_RATE_200_GBPS: return 206250; + case IB_RATE_300_GBPS: return 309375; + case IB_RATE_28_GBPS: return 28125; + case IB_RATE_50_GBPS: return 53125; + case IB_RATE_400_GBPS: return 425000; + case IB_RATE_600_GBPS: return 637500; + default: return -1; + } +} +EXPORT_SYMBOL(ib_rate_to_mbps); + +__attribute_const__ enum rdma_transport_type +rdma_node_get_transport(unsigned int node_type) +{ + + if (node_type == RDMA_NODE_USNIC) + return RDMA_TRANSPORT_USNIC; + if (node_type == RDMA_NODE_USNIC_UDP) + return RDMA_TRANSPORT_USNIC_UDP; + if (node_type == RDMA_NODE_RNIC) + return RDMA_TRANSPORT_IWARP; + if (node_type == RDMA_NODE_UNSPECIFIED) + return RDMA_TRANSPORT_UNSPECIFIED; + + return RDMA_TRANSPORT_IB; +} +EXPORT_SYMBOL(rdma_node_get_transport); + +enum rdma_link_layer rdma_port_get_link_layer(struct ib_device *device, u8 port_num) +{ + enum rdma_transport_type lt; + if (device->ops.get_link_layer) + return device->ops.get_link_layer(device, port_num); + + lt = rdma_node_get_transport(device->node_type); + if (lt == RDMA_TRANSPORT_IB) + return IB_LINK_LAYER_INFINIBAND; + + return IB_LINK_LAYER_ETHERNET; +} +EXPORT_SYMBOL(rdma_port_get_link_layer); + +/* Protection domains */ + +/** + * ib_alloc_pd - Allocates an unused protection domain. + * @device: The device on which to allocate the protection domain. + * @flags: protection domain flags + * @caller: caller's build-time module name + * + * A protection domain object provides an association between QPs, shared + * receive queues, address handles, memory regions, and memory windows. + * + * Every PD has a local_dma_lkey which can be used as the lkey value for local + * memory operations. + */ +struct ib_pd *__ib_alloc_pd(struct ib_device *device, unsigned int flags, + const char *caller) +{ + struct ib_pd *pd; + int mr_access_flags = 0; + int ret; + + pd = rdma_zalloc_drv_obj(device, ib_pd); + if (!pd) + return ERR_PTR(-ENOMEM); + + pd->device = device; + pd->uobject = NULL; + pd->__internal_mr = NULL; + atomic_set(&pd->usecnt, 0); + pd->flags = flags; + + rdma_restrack_new(&pd->res, RDMA_RESTRACK_PD); + rdma_restrack_set_name(&pd->res, caller); + + ret = device->ops.alloc_pd(pd, NULL); + if (ret) { + rdma_restrack_put(&pd->res); + kfree(pd); + return ERR_PTR(ret); + } + rdma_restrack_add(&pd->res); + + if (device->attrs.device_cap_flags & IB_DEVICE_LOCAL_DMA_LKEY) + pd->local_dma_lkey = device->local_dma_lkey; + else + mr_access_flags |= IB_ACCESS_LOCAL_WRITE; + + if (flags & IB_PD_UNSAFE_GLOBAL_RKEY) { + pr_warn("%s: enabling unsafe global rkey\n", caller); + mr_access_flags |= IB_ACCESS_REMOTE_READ | IB_ACCESS_REMOTE_WRITE; + } + + if (mr_access_flags) { + struct ib_mr *mr; + + mr = pd->device->ops.get_dma_mr(pd, mr_access_flags); + if (IS_ERR(mr)) { + ib_dealloc_pd(pd); + return ERR_CAST(mr); + } + + mr->device = pd->device; + mr->pd = pd; + mr->type = IB_MR_TYPE_DMA; + mr->uobject = NULL; + mr->need_inval = false; + + pd->__internal_mr = mr; + + if (!(device->attrs.device_cap_flags & IB_DEVICE_LOCAL_DMA_LKEY)) + pd->local_dma_lkey = pd->__internal_mr->lkey; + + if (flags & IB_PD_UNSAFE_GLOBAL_RKEY) + pd->unsafe_global_rkey = pd->__internal_mr->rkey; + } + + return pd; +} +EXPORT_SYMBOL(__ib_alloc_pd); + +/** + * ib_dealloc_pd_user - Deallocates a protection domain. + * @pd: The protection domain to deallocate. + * @udata: Valid user data or NULL for kernel object + * + * It is an error to call this function while any resources in the pd still + * exist. The caller is responsible to synchronously destroy them and + * guarantee no new allocations will happen. + */ +int ib_dealloc_pd_user(struct ib_pd *pd, struct ib_udata *udata) +{ + int ret; + + if (pd->__internal_mr) { + ret = pd->device->ops.dereg_mr(pd->__internal_mr, NULL); + WARN_ON(ret); + pd->__internal_mr = NULL; + } + + /* uverbs manipulates usecnt with proper locking, while the kabi + requires the caller to guarantee we can't race here. */ + WARN_ON(atomic_read(&pd->usecnt)); + + ret = pd->device->ops.dealloc_pd(pd, udata); + if (ret) + return ret; + + rdma_restrack_del(&pd->res); + kfree(pd); + return ret; +} +EXPORT_SYMBOL(ib_dealloc_pd_user); + +/* Address handles */ + +/** + * rdma_copy_ah_attr - Copy rdma ah attribute from source to destination. + * @dest: Pointer to destination ah_attr. Contents of the destination + * pointer is assumed to be invalid and attribute are overwritten. + * @src: Pointer to source ah_attr. + */ +void rdma_copy_ah_attr(struct rdma_ah_attr *dest, + const struct rdma_ah_attr *src) +{ + *dest = *src; + if (dest->grh.sgid_attr) + rdma_hold_gid_attr(dest->grh.sgid_attr); +} +EXPORT_SYMBOL(rdma_copy_ah_attr); + +/** + * rdma_replace_ah_attr - Replace valid ah_attr with new new one. + * @old: Pointer to existing ah_attr which needs to be replaced. + * old is assumed to be valid or zero'd + * @new: Pointer to the new ah_attr. + * + * rdma_replace_ah_attr() first releases any reference in the old ah_attr if + * old the ah_attr is valid; after that it copies the new attribute and holds + * the reference to the replaced ah_attr. + */ +void rdma_replace_ah_attr(struct rdma_ah_attr *old, + const struct rdma_ah_attr *new) +{ + rdma_destroy_ah_attr(old); + *old = *new; + if (old->grh.sgid_attr) + rdma_hold_gid_attr(old->grh.sgid_attr); +} +EXPORT_SYMBOL(rdma_replace_ah_attr); + +/** + * rdma_move_ah_attr - Move ah_attr pointed by source to destination. + * @dest: Pointer to destination ah_attr to copy to. + * dest is assumed to be valid or zero'd + * @src: Pointer to the new ah_attr. + * + * rdma_move_ah_attr() first releases any reference in the destination ah_attr + * if it is valid. This also transfers ownership of internal references from + * src to dest, making src invalid in the process. No new reference of the src + * ah_attr is taken. + */ +void rdma_move_ah_attr(struct rdma_ah_attr *dest, struct rdma_ah_attr *src) +{ + rdma_destroy_ah_attr(dest); + *dest = *src; + src->grh.sgid_attr = NULL; +} +EXPORT_SYMBOL(rdma_move_ah_attr); + +/* + * Validate that the rdma_ah_attr is valid for the device before passing it + * off to the driver. + */ +static int rdma_check_ah_attr(struct ib_device *device, + struct rdma_ah_attr *ah_attr) +{ + if (!rdma_is_port_valid(device, ah_attr->port_num)) + return -EINVAL; + + if ((rdma_is_grh_required(device, ah_attr->port_num) || + ah_attr->type == RDMA_AH_ATTR_TYPE_ROCE) && + !(ah_attr->ah_flags & IB_AH_GRH)) + return -EINVAL; + + if (ah_attr->grh.sgid_attr) { + /* + * Make sure the passed sgid_attr is consistent with the + * parameters + */ + if (ah_attr->grh.sgid_attr->index != ah_attr->grh.sgid_index || + ah_attr->grh.sgid_attr->port_num != ah_attr->port_num) + return -EINVAL; + } + return 0; +} + +/* + * If the ah requires a GRH then ensure that sgid_attr pointer is filled in. + * On success the caller is responsible to call rdma_unfill_sgid_attr(). + */ +static int rdma_fill_sgid_attr(struct ib_device *device, + struct rdma_ah_attr *ah_attr, + const struct ib_gid_attr **old_sgid_attr) +{ + const struct ib_gid_attr *sgid_attr; + struct ib_global_route *grh; + int ret; + + *old_sgid_attr = ah_attr->grh.sgid_attr; + + ret = rdma_check_ah_attr(device, ah_attr); + if (ret) + return ret; + + if (!(ah_attr->ah_flags & IB_AH_GRH)) + return 0; + + grh = rdma_ah_retrieve_grh(ah_attr); + if (grh->sgid_attr) + return 0; + + sgid_attr = + rdma_get_gid_attr(device, ah_attr->port_num, grh->sgid_index); + if (IS_ERR(sgid_attr)) + return PTR_ERR(sgid_attr); + + /* Move ownerhip of the kref into the ah_attr */ + grh->sgid_attr = sgid_attr; + return 0; +} + +static void rdma_unfill_sgid_attr(struct rdma_ah_attr *ah_attr, + const struct ib_gid_attr *old_sgid_attr) +{ + /* + * Fill didn't change anything, the caller retains ownership of + * whatever it passed + */ + if (ah_attr->grh.sgid_attr == old_sgid_attr) + return; + + /* + * Otherwise, we need to undo what rdma_fill_sgid_attr so the caller + * doesn't see any change in the rdma_ah_attr. If we get here + * old_sgid_attr is NULL. + */ + rdma_destroy_ah_attr(ah_attr); +} + +static const struct ib_gid_attr * +rdma_update_sgid_attr(struct rdma_ah_attr *ah_attr, + const struct ib_gid_attr *old_attr) +{ + if (old_attr) + rdma_put_gid_attr(old_attr); + if (ah_attr->ah_flags & IB_AH_GRH) { + rdma_hold_gid_attr(ah_attr->grh.sgid_attr); + return ah_attr->grh.sgid_attr; + } + return NULL; +} + +static struct ib_ah *_rdma_create_ah(struct ib_pd *pd, + struct rdma_ah_attr *ah_attr, + u32 flags, + struct ib_udata *udata, + struct net_device *xmit_slave) +{ + struct rdma_ah_init_attr init_attr = {}; + struct ib_device *device = pd->device; + struct ib_ah *ah; + int ret; + + might_sleep_if(flags & RDMA_CREATE_AH_SLEEPABLE); + + if (!device->ops.create_ah) + return ERR_PTR(-EOPNOTSUPP); + + ah = rdma_zalloc_drv_obj_gfp( + device, ib_ah, + (flags & RDMA_CREATE_AH_SLEEPABLE) ? GFP_KERNEL : GFP_ATOMIC); + if (!ah) + return ERR_PTR(-ENOMEM); + + ah->device = device; + ah->pd = pd; + ah->type = ah_attr->type; + ah->sgid_attr = rdma_update_sgid_attr(ah_attr, NULL); + init_attr.ah_attr = ah_attr; + init_attr.flags = flags; + init_attr.xmit_slave = xmit_slave; + + ret = device->ops.create_ah(ah, &init_attr, udata); + if (ret) { + if (ah->sgid_attr) + rdma_put_gid_attr(ah->sgid_attr); + kfree(ah); + return ERR_PTR(ret); + } + + atomic_inc(&pd->usecnt); + return ah; +} + +/** + * rdma_create_ah - Creates an address handle for the + * given address vector. + * @pd: The protection domain associated with the address handle. + * @ah_attr: The attributes of the address vector. + * @flags: Create address handle flags (see enum rdma_create_ah_flags). + * + * It returns 0 on success and returns appropriate error code on error. + * The address handle is used to reference a local or global destination + * in all UD QP post sends. + */ +struct ib_ah *rdma_create_ah(struct ib_pd *pd, struct rdma_ah_attr *ah_attr, + u32 flags) +{ + const struct ib_gid_attr *old_sgid_attr; + struct net_device *slave; + struct ib_ah *ah; + int ret; + + ret = rdma_fill_sgid_attr(pd->device, ah_attr, &old_sgid_attr); + if (ret) + return ERR_PTR(ret); + slave = rdma_lag_get_ah_roce_slave(pd->device, ah_attr, + (flags & RDMA_CREATE_AH_SLEEPABLE) ? + GFP_KERNEL : GFP_ATOMIC); + if (IS_ERR(slave)) { + rdma_unfill_sgid_attr(ah_attr, old_sgid_attr); + return (void *)slave; + } + ah = _rdma_create_ah(pd, ah_attr, flags, NULL, slave); + rdma_lag_put_ah_roce_slave(slave); + rdma_unfill_sgid_attr(ah_attr, old_sgid_attr); + return ah; +} +EXPORT_SYMBOL(rdma_create_ah); + +/** + * rdma_create_user_ah - Creates an address handle for the + * given address vector. + * It resolves destination mac address for ah attribute of RoCE type. + * @pd: The protection domain associated with the address handle. + * @ah_attr: The attributes of the address vector. + * @udata: pointer to user's input output buffer information need by + * provider driver. + * + * It returns 0 on success and returns appropriate error code on error. + * The address handle is used to reference a local or global destination + * in all UD QP post sends. + */ +struct ib_ah *rdma_create_user_ah(struct ib_pd *pd, + struct rdma_ah_attr *ah_attr, + struct ib_udata *udata) +{ + const struct ib_gid_attr *old_sgid_attr; + struct ib_ah *ah; + int err; + + err = rdma_fill_sgid_attr(pd->device, ah_attr, &old_sgid_attr); + if (err) + return ERR_PTR(err); + + if (ah_attr->type == RDMA_AH_ATTR_TYPE_ROCE) { + err = ib_resolve_eth_dmac(pd->device, ah_attr); + if (err) { + ah = ERR_PTR(err); + goto out; + } + } + + ah = _rdma_create_ah(pd, ah_attr, RDMA_CREATE_AH_SLEEPABLE, + udata, NULL); + +out: + rdma_unfill_sgid_attr(ah_attr, old_sgid_attr); + return ah; +} +EXPORT_SYMBOL(rdma_create_user_ah); + +int ib_get_rdma_header_version(const union rdma_network_hdr *hdr) +{ + const struct iphdr *ip4h = (struct iphdr *)&hdr->roce4grh; + struct iphdr ip4h_checked; + const struct ipv6hdr *ip6h = (struct ipv6hdr *)&hdr->ibgrh; + + /* If it's IPv6, the version must be 6, otherwise, the first + * 20 bytes (before the IPv4 header) are garbled. + */ + if (ip6h->version != 6) + return (ip4h->version == 4) ? 4 : 0; + /* version may be 6 or 4 because the first 20 bytes could be garbled */ + + /* RoCE v2 requires no options, thus header length + * must be 5 words + */ + if (ip4h->ihl != 5) + return 6; + + /* Verify checksum. + * We can't write on scattered buffers so we need to copy to + * temp buffer. + */ + memcpy(&ip4h_checked, ip4h, sizeof(ip4h_checked)); + ip4h_checked.check = 0; + ip4h_checked.check = ip_fast_csum((u8 *)&ip4h_checked, 5); + /* if IPv4 header checksum is OK, believe it */ + if (ip4h->check == ip4h_checked.check) + return 4; + return 6; +} +EXPORT_SYMBOL(ib_get_rdma_header_version); + +static enum rdma_network_type ib_get_net_type_by_grh(struct ib_device *device, + u8 port_num, + const struct ib_grh *grh) +{ + int grh_version; + + if (rdma_protocol_ib(device, port_num)) + return RDMA_NETWORK_IB; + + grh_version = ib_get_rdma_header_version((union rdma_network_hdr *)grh); + + if (grh_version == 4) + return RDMA_NETWORK_IPV4; + + if (grh->next_hdr == IPPROTO_UDP) + return RDMA_NETWORK_IPV6; + + return RDMA_NETWORK_ROCE_V1; +} + +struct find_gid_index_context { + u16 vlan_id; + enum ib_gid_type gid_type; +}; + +static bool find_gid_index(const union ib_gid *gid, + const struct ib_gid_attr *gid_attr, + void *context) +{ + struct find_gid_index_context *ctx = context; + u16 vlan_id = 0xffff; + int ret; + + if (ctx->gid_type != gid_attr->gid_type) + return false; + + ret = rdma_read_gid_l2_fields(gid_attr, &vlan_id, NULL); + if (ret) + return false; + + return ctx->vlan_id == vlan_id; +} + +static const struct ib_gid_attr * +get_sgid_attr_from_eth(struct ib_device *device, u8 port_num, + u16 vlan_id, const union ib_gid *sgid, + enum ib_gid_type gid_type) +{ + struct find_gid_index_context context = {.vlan_id = vlan_id, + .gid_type = gid_type}; + + return rdma_find_gid_by_filter(device, sgid, port_num, find_gid_index, + &context); +} + +int ib_get_gids_from_rdma_hdr(const union rdma_network_hdr *hdr, + enum rdma_network_type net_type, + union ib_gid *sgid, union ib_gid *dgid) +{ + struct sockaddr_in src_in; + struct sockaddr_in dst_in; + __be32 src_saddr, dst_saddr; + + if (!sgid || !dgid) + return -EINVAL; + + if (net_type == RDMA_NETWORK_IPV4) { + memcpy(&src_in.sin_addr.s_addr, + &hdr->roce4grh.saddr, 4); + memcpy(&dst_in.sin_addr.s_addr, + &hdr->roce4grh.daddr, 4); + src_saddr = src_in.sin_addr.s_addr; + dst_saddr = dst_in.sin_addr.s_addr; + ipv6_addr_set_v4mapped(src_saddr, + (struct in6_addr *)sgid); + ipv6_addr_set_v4mapped(dst_saddr, + (struct in6_addr *)dgid); + return 0; + } else if (net_type == RDMA_NETWORK_IPV6 || + net_type == RDMA_NETWORK_IB || RDMA_NETWORK_ROCE_V1) { + *dgid = hdr->ibgrh.dgid; + *sgid = hdr->ibgrh.sgid; + return 0; + } else { + return -EINVAL; + } +} +EXPORT_SYMBOL(ib_get_gids_from_rdma_hdr); + +/* Resolve destination mac address and hop limit for unicast destination + * GID entry, considering the source GID entry as well. + * ah_attribute must have have valid port_num, sgid_index. + */ +static int ib_resolve_unicast_gid_dmac(struct ib_device *device, + struct rdma_ah_attr *ah_attr) +{ + struct ib_global_route *grh = rdma_ah_retrieve_grh(ah_attr); + const struct ib_gid_attr *sgid_attr = grh->sgid_attr; + int hop_limit = 0xff; + int ret = 0; + + /* If destination is link local and source GID is RoCEv1, + * IP stack is not used. + */ + if (rdma_link_local_addr((struct in6_addr *)grh->dgid.raw) && + sgid_attr->gid_type == IB_GID_TYPE_ROCE) { + rdma_get_ll_mac((struct in6_addr *)grh->dgid.raw, + ah_attr->roce.dmac); + return ret; + } + + ret = rdma_addr_find_l2_eth_by_grh(&sgid_attr->gid, &grh->dgid, + ah_attr->roce.dmac, + sgid_attr, &hop_limit); + + grh->hop_limit = hop_limit; + return ret; +} + +/* + * This function initializes address handle attributes from the incoming packet. + * Incoming packet has dgid of the receiver node on which this code is + * getting executed and, sgid contains the GID of the sender. + * + * When resolving mac address of destination, the arrived dgid is used + * as sgid and, sgid is used as dgid because sgid contains destinations + * GID whom to respond to. + * + * On success the caller is responsible to call rdma_destroy_ah_attr on the + * attr. + */ +int ib_init_ah_attr_from_wc(struct ib_device *device, u8 port_num, + const struct ib_wc *wc, const struct ib_grh *grh, + struct rdma_ah_attr *ah_attr) +{ + u32 flow_class; + int ret; + enum rdma_network_type net_type = RDMA_NETWORK_IB; + enum ib_gid_type gid_type = IB_GID_TYPE_IB; + const struct ib_gid_attr *sgid_attr; + int hoplimit = 0xff; + union ib_gid dgid; + union ib_gid sgid; + + might_sleep(); + + memset(ah_attr, 0, sizeof *ah_attr); + ah_attr->type = rdma_ah_find_type(device, port_num); + if (rdma_cap_eth_ah(device, port_num)) { + if (wc->wc_flags & IB_WC_WITH_NETWORK_HDR_TYPE) + net_type = wc->network_hdr_type; + else + net_type = ib_get_net_type_by_grh(device, port_num, grh); + gid_type = ib_network_to_gid_type(net_type); + } + ret = ib_get_gids_from_rdma_hdr((union rdma_network_hdr *)grh, net_type, + &sgid, &dgid); + if (ret) + return ret; + + rdma_ah_set_sl(ah_attr, wc->sl); + rdma_ah_set_port_num(ah_attr, port_num); + + if (rdma_protocol_roce(device, port_num)) { + u16 vlan_id = wc->wc_flags & IB_WC_WITH_VLAN ? + wc->vlan_id : 0xffff; + + if (!(wc->wc_flags & IB_WC_GRH)) + return -EPROTOTYPE; + + sgid_attr = get_sgid_attr_from_eth(device, port_num, + vlan_id, &dgid, + gid_type); + if (IS_ERR(sgid_attr)) + return PTR_ERR(sgid_attr); + + flow_class = be32_to_cpu(grh->version_tclass_flow); + rdma_move_grh_sgid_attr(ah_attr, + &sgid, + flow_class & 0xFFFFF, + hoplimit, + (flow_class >> 20) & 0xFF, + sgid_attr); + + ret = ib_resolve_unicast_gid_dmac(device, ah_attr); + if (ret) + rdma_destroy_ah_attr(ah_attr); + + return ret; + } else { + rdma_ah_set_dlid(ah_attr, wc->slid); + rdma_ah_set_path_bits(ah_attr, wc->dlid_path_bits); + + if ((wc->wc_flags & IB_WC_GRH) == 0) + return 0; + + if (dgid.global.interface_id != + cpu_to_be64(IB_SA_WELL_KNOWN_GUID)) { + sgid_attr = rdma_find_gid_by_port( + device, &dgid, IB_GID_TYPE_IB, port_num, NULL); + } else + sgid_attr = rdma_get_gid_attr(device, port_num, 0); + + if (IS_ERR(sgid_attr)) + return PTR_ERR(sgid_attr); + flow_class = be32_to_cpu(grh->version_tclass_flow); + rdma_move_grh_sgid_attr(ah_attr, + &sgid, + flow_class & 0xFFFFF, + hoplimit, + (flow_class >> 20) & 0xFF, + sgid_attr); + + return 0; + } +} +EXPORT_SYMBOL(ib_init_ah_attr_from_wc); + +/** + * rdma_move_grh_sgid_attr - Sets the sgid attribute of GRH, taking ownership + * of the reference + * + * @attr: Pointer to AH attribute structure + * @dgid: Destination GID + * @flow_label: Flow label + * @hop_limit: Hop limit + * @traffic_class: traffic class + * @sgid_attr: Pointer to SGID attribute + * + * This takes ownership of the sgid_attr reference. The caller must ensure + * rdma_destroy_ah_attr() is called before destroying the rdma_ah_attr after + * calling this function. + */ +void rdma_move_grh_sgid_attr(struct rdma_ah_attr *attr, union ib_gid *dgid, + u32 flow_label, u8 hop_limit, u8 traffic_class, + const struct ib_gid_attr *sgid_attr) +{ + rdma_ah_set_grh(attr, dgid, flow_label, sgid_attr->index, hop_limit, + traffic_class); + attr->grh.sgid_attr = sgid_attr; +} +EXPORT_SYMBOL(rdma_move_grh_sgid_attr); + +/** + * rdma_destroy_ah_attr - Release reference to SGID attribute of + * ah attribute. + * @ah_attr: Pointer to ah attribute + * + * Release reference to the SGID attribute of the ah attribute if it is + * non NULL. It is safe to call this multiple times, and safe to call it on + * a zero initialized ah_attr. + */ +void rdma_destroy_ah_attr(struct rdma_ah_attr *ah_attr) +{ + if (ah_attr->grh.sgid_attr) { + rdma_put_gid_attr(ah_attr->grh.sgid_attr); + ah_attr->grh.sgid_attr = NULL; + } +} +EXPORT_SYMBOL(rdma_destroy_ah_attr); + +struct ib_ah *ib_create_ah_from_wc(struct ib_pd *pd, const struct ib_wc *wc, + const struct ib_grh *grh, u8 port_num) +{ + struct rdma_ah_attr ah_attr; + struct ib_ah *ah; + int ret; + + ret = ib_init_ah_attr_from_wc(pd->device, port_num, wc, grh, &ah_attr); + if (ret) + return ERR_PTR(ret); + + ah = rdma_create_ah(pd, &ah_attr, RDMA_CREATE_AH_SLEEPABLE); + + rdma_destroy_ah_attr(&ah_attr); + return ah; +} +EXPORT_SYMBOL(ib_create_ah_from_wc); + +int rdma_modify_ah(struct ib_ah *ah, struct rdma_ah_attr *ah_attr) +{ + const struct ib_gid_attr *old_sgid_attr; + int ret; + + if (ah->type != ah_attr->type) + return -EINVAL; + + ret = rdma_fill_sgid_attr(ah->device, ah_attr, &old_sgid_attr); + if (ret) + return ret; + + ret = ah->device->ops.modify_ah ? + ah->device->ops.modify_ah(ah, ah_attr) : + -EOPNOTSUPP; + + ah->sgid_attr = rdma_update_sgid_attr(ah_attr, ah->sgid_attr); + rdma_unfill_sgid_attr(ah_attr, old_sgid_attr); + return ret; +} +EXPORT_SYMBOL(rdma_modify_ah); + +int rdma_query_ah(struct ib_ah *ah, struct rdma_ah_attr *ah_attr) +{ + ah_attr->grh.sgid_attr = NULL; + + return ah->device->ops.query_ah ? + ah->device->ops.query_ah(ah, ah_attr) : + -EOPNOTSUPP; +} +EXPORT_SYMBOL(rdma_query_ah); + +int rdma_destroy_ah_user(struct ib_ah *ah, u32 flags, struct ib_udata *udata) +{ + const struct ib_gid_attr *sgid_attr = ah->sgid_attr; + struct ib_pd *pd; + int ret; + + might_sleep_if(flags & RDMA_DESTROY_AH_SLEEPABLE); + + pd = ah->pd; + + ret = ah->device->ops.destroy_ah(ah, flags); + if (ret) + return ret; + + atomic_dec(&pd->usecnt); + if (sgid_attr) + rdma_put_gid_attr(sgid_attr); + + kfree(ah); + return ret; +} +EXPORT_SYMBOL(rdma_destroy_ah_user); + +/* Shared receive queues */ + +/** + * ib_create_srq_user - Creates a SRQ associated with the specified protection + * domain. + * @pd: The protection domain associated with the SRQ. + * @srq_init_attr: A list of initial attributes required to create the + * SRQ. If SRQ creation succeeds, then the attributes are updated to + * the actual capabilities of the created SRQ. + * @uobject: uobject pointer if this is not a kernel SRQ + * @udata: udata pointer if this is not a kernel SRQ + * + * srq_attr->max_wr and srq_attr->max_sge are read the determine the + * requested size of the SRQ, and set to the actual values allocated + * on return. If ib_create_srq() succeeds, then max_wr and max_sge + * will always be at least as large as the requested values. + */ +struct ib_srq *ib_create_srq_user(struct ib_pd *pd, + struct ib_srq_init_attr *srq_init_attr, + struct ib_usrq_object *uobject, + struct ib_udata *udata) +{ + struct ib_srq *srq; + int ret; + + srq = rdma_zalloc_drv_obj(pd->device, ib_srq); + if (!srq) + return ERR_PTR(-ENOMEM); + + srq->device = pd->device; + srq->pd = pd; + srq->event_handler = srq_init_attr->event_handler; + srq->srq_context = srq_init_attr->srq_context; + srq->srq_type = srq_init_attr->srq_type; + srq->uobject = uobject; + + if (ib_srq_has_cq(srq->srq_type)) { + srq->ext.cq = srq_init_attr->ext.cq; + atomic_inc(&srq->ext.cq->usecnt); + } + if (srq->srq_type == IB_SRQT_XRC) { + srq->ext.xrc.xrcd = srq_init_attr->ext.xrc.xrcd; + atomic_inc(&srq->ext.xrc.xrcd->usecnt); + } + atomic_inc(&pd->usecnt); + + ret = pd->device->ops.create_srq(srq, srq_init_attr, udata); + if (ret) { + atomic_dec(&srq->pd->usecnt); + if (srq->srq_type == IB_SRQT_XRC) + atomic_dec(&srq->ext.xrc.xrcd->usecnt); + if (ib_srq_has_cq(srq->srq_type)) + atomic_dec(&srq->ext.cq->usecnt); + kfree(srq); + return ERR_PTR(ret); + } + + return srq; +} +EXPORT_SYMBOL(ib_create_srq_user); + +int ib_modify_srq(struct ib_srq *srq, + struct ib_srq_attr *srq_attr, + enum ib_srq_attr_mask srq_attr_mask) +{ + return srq->device->ops.modify_srq ? + srq->device->ops.modify_srq(srq, srq_attr, srq_attr_mask, + NULL) : -EOPNOTSUPP; +} +EXPORT_SYMBOL(ib_modify_srq); + +int ib_query_srq(struct ib_srq *srq, + struct ib_srq_attr *srq_attr) +{ + return srq->device->ops.query_srq ? + srq->device->ops.query_srq(srq, srq_attr) : -EOPNOTSUPP; +} +EXPORT_SYMBOL(ib_query_srq); + +int ib_destroy_srq_user(struct ib_srq *srq, struct ib_udata *udata) +{ + int ret; + + if (atomic_read(&srq->usecnt)) + return -EBUSY; + + ret = srq->device->ops.destroy_srq(srq, udata); + if (ret) + return ret; + + atomic_dec(&srq->pd->usecnt); + if (srq->srq_type == IB_SRQT_XRC) + atomic_dec(&srq->ext.xrc.xrcd->usecnt); + if (ib_srq_has_cq(srq->srq_type)) + atomic_dec(&srq->ext.cq->usecnt); + kfree(srq); + + return ret; +} +EXPORT_SYMBOL(ib_destroy_srq_user); + +/* Queue pairs */ + +static void __ib_shared_qp_event_handler(struct ib_event *event, void *context) +{ + struct ib_qp *qp = context; + unsigned long flags; + + spin_lock_irqsave(&qp->device->qp_open_list_lock, flags); + list_for_each_entry(event->element.qp, &qp->open_list, open_list) + if (event->element.qp->event_handler) + event->element.qp->event_handler(event, event->element.qp->qp_context); + spin_unlock_irqrestore(&qp->device->qp_open_list_lock, flags); +} + +static struct ib_qp *__ib_open_qp(struct ib_qp *real_qp, + void (*event_handler)(struct ib_event *, void *), + void *qp_context) +{ + struct ib_qp *qp; + unsigned long flags; + int err; + + qp = kzalloc(sizeof *qp, GFP_KERNEL); + if (!qp) + return ERR_PTR(-ENOMEM); + + qp->real_qp = real_qp; + err = ib_open_shared_qp_security(qp, real_qp->device); + if (err) { + kfree(qp); + return ERR_PTR(err); + } + + qp->real_qp = real_qp; + atomic_inc(&real_qp->usecnt); + qp->device = real_qp->device; + qp->event_handler = event_handler; + qp->qp_context = qp_context; + qp->qp_num = real_qp->qp_num; + qp->qp_type = real_qp->qp_type; + + spin_lock_irqsave(&real_qp->device->qp_open_list_lock, flags); + list_add(&qp->open_list, &real_qp->open_list); + spin_unlock_irqrestore(&real_qp->device->qp_open_list_lock, flags); + + return qp; +} + +struct ib_qp *ib_open_qp(struct ib_xrcd *xrcd, + struct ib_qp_open_attr *qp_open_attr) +{ + struct ib_qp *qp, *real_qp; + + if (qp_open_attr->qp_type != IB_QPT_XRC_TGT) + return ERR_PTR(-EINVAL); + + down_read(&xrcd->tgt_qps_rwsem); + real_qp = xa_load(&xrcd->tgt_qps, qp_open_attr->qp_num); + if (!real_qp) { + up_read(&xrcd->tgt_qps_rwsem); + return ERR_PTR(-EINVAL); + } + qp = __ib_open_qp(real_qp, qp_open_attr->event_handler, + qp_open_attr->qp_context); + up_read(&xrcd->tgt_qps_rwsem); + return qp; +} +EXPORT_SYMBOL(ib_open_qp); + +static struct ib_qp *create_xrc_qp_user(struct ib_qp *qp, + struct ib_qp_init_attr *qp_init_attr) +{ + struct ib_qp *real_qp = qp; + int err; + + qp->event_handler = __ib_shared_qp_event_handler; + qp->qp_context = qp; + qp->pd = NULL; + qp->send_cq = qp->recv_cq = NULL; + qp->srq = NULL; + qp->xrcd = qp_init_attr->xrcd; + atomic_inc(&qp_init_attr->xrcd->usecnt); + INIT_LIST_HEAD(&qp->open_list); + + qp = __ib_open_qp(real_qp, qp_init_attr->event_handler, + qp_init_attr->qp_context); + if (IS_ERR(qp)) + return qp; + + err = xa_err(xa_store(&qp_init_attr->xrcd->tgt_qps, real_qp->qp_num, + real_qp, GFP_KERNEL)); + if (err) { + ib_close_qp(qp); + return ERR_PTR(err); + } + return qp; +} + +/** + * ib_create_qp - Creates a kernel QP associated with the specified protection + * domain. + * @pd: The protection domain associated with the QP. + * @qp_init_attr: A list of initial attributes required to create the + * QP. If QP creation succeeds, then the attributes are updated to + * the actual capabilities of the created QP. + * + * NOTE: for user qp use ib_create_qp_user with valid udata! + */ +struct ib_qp *ib_create_qp(struct ib_pd *pd, + struct ib_qp_init_attr *qp_init_attr) +{ + struct ib_device *device = pd ? pd->device : qp_init_attr->xrcd->device; + struct ib_qp *qp; + int ret; + + if (qp_init_attr->rwq_ind_tbl && + (qp_init_attr->recv_cq || + qp_init_attr->srq || qp_init_attr->cap.max_recv_wr || + qp_init_attr->cap.max_recv_sge)) + return ERR_PTR(-EINVAL); + + if ((qp_init_attr->create_flags & IB_QP_CREATE_INTEGRITY_EN) && + !(device->attrs.device_cap_flags & IB_DEVICE_INTEGRITY_HANDOVER)) + return ERR_PTR(-EINVAL); + + /* + * If the callers is using the RDMA API calculate the resources + * needed for the RDMA READ/WRITE operations. + * + * Note that these callers need to pass in a port number. + */ + if (qp_init_attr->cap.max_rdma_ctxs) + rdma_rw_init_qp(device, qp_init_attr); + + qp = _ib_create_qp(device, pd, qp_init_attr, NULL, NULL); + if (IS_ERR(qp)) + return qp; + + ret = ib_create_qp_security(qp, device); + if (ret) + goto err; + + if (qp_init_attr->qp_type == IB_QPT_XRC_TGT) { + struct ib_qp *xrc_qp = + create_xrc_qp_user(qp, qp_init_attr); + + if (IS_ERR(xrc_qp)) { + ret = PTR_ERR(xrc_qp); + goto err; + } + return xrc_qp; + } + + qp->event_handler = qp_init_attr->event_handler; + qp->qp_context = qp_init_attr->qp_context; + if (qp_init_attr->qp_type == IB_QPT_XRC_INI) { + qp->recv_cq = NULL; + qp->srq = NULL; + } else { + qp->recv_cq = qp_init_attr->recv_cq; + if (qp_init_attr->recv_cq) + atomic_inc(&qp_init_attr->recv_cq->usecnt); + qp->srq = qp_init_attr->srq; + if (qp->srq) + atomic_inc(&qp_init_attr->srq->usecnt); + } + + qp->send_cq = qp_init_attr->send_cq; + qp->xrcd = NULL; + + atomic_inc(&pd->usecnt); + if (qp_init_attr->send_cq) + atomic_inc(&qp_init_attr->send_cq->usecnt); + if (qp_init_attr->rwq_ind_tbl) + atomic_inc(&qp->rwq_ind_tbl->usecnt); + + if (qp_init_attr->cap.max_rdma_ctxs) { + ret = rdma_rw_init_mrs(qp, qp_init_attr); + if (ret) + goto err; + } + + /* + * Note: all hw drivers guarantee that max_send_sge is lower than + * the device RDMA WRITE SGE limit but not all hw drivers ensure that + * max_send_sge <= max_sge_rd. + */ + qp->max_write_sge = qp_init_attr->cap.max_send_sge; + qp->max_read_sge = min_t(u32, qp_init_attr->cap.max_send_sge, + device->attrs.max_sge_rd); + if (qp_init_attr->create_flags & IB_QP_CREATE_INTEGRITY_EN) + qp->integrity_en = true; + + return qp; + +err: + ib_destroy_qp(qp); + return ERR_PTR(ret); + +} +EXPORT_SYMBOL(ib_create_qp); + +static const struct { + int valid; + enum ib_qp_attr_mask req_param[IB_QPT_MAX]; + enum ib_qp_attr_mask opt_param[IB_QPT_MAX]; +} qp_state_table[IB_QPS_ERR + 1][IB_QPS_ERR + 1] = { + [IB_QPS_RESET] = { + [IB_QPS_RESET] = { .valid = 1 }, + [IB_QPS_INIT] = { + .valid = 1, + .req_param = { + [IB_QPT_UD] = (IB_QP_PKEY_INDEX | + IB_QP_PORT | + IB_QP_QKEY), + [IB_QPT_RAW_PACKET] = IB_QP_PORT, + [IB_QPT_UC] = (IB_QP_PKEY_INDEX | + IB_QP_PORT | + IB_QP_ACCESS_FLAGS), + [IB_QPT_RC] = (IB_QP_PKEY_INDEX | + IB_QP_PORT | + IB_QP_ACCESS_FLAGS), + [IB_QPT_XRC_INI] = (IB_QP_PKEY_INDEX | + IB_QP_PORT | + IB_QP_ACCESS_FLAGS), + [IB_QPT_XRC_TGT] = (IB_QP_PKEY_INDEX | + IB_QP_PORT | + IB_QP_ACCESS_FLAGS), + [IB_QPT_SMI] = (IB_QP_PKEY_INDEX | + IB_QP_QKEY), + [IB_QPT_GSI] = (IB_QP_PKEY_INDEX | + IB_QP_QKEY), + } + }, + }, + [IB_QPS_INIT] = { + [IB_QPS_RESET] = { .valid = 1 }, + [IB_QPS_ERR] = { .valid = 1 }, + [IB_QPS_INIT] = { + .valid = 1, + .opt_param = { + [IB_QPT_UD] = (IB_QP_PKEY_INDEX | + IB_QP_PORT | + IB_QP_QKEY), + [IB_QPT_UC] = (IB_QP_PKEY_INDEX | + IB_QP_PORT | + IB_QP_ACCESS_FLAGS), + [IB_QPT_RC] = (IB_QP_PKEY_INDEX | + IB_QP_PORT | + IB_QP_ACCESS_FLAGS), + [IB_QPT_XRC_INI] = (IB_QP_PKEY_INDEX | + IB_QP_PORT | + IB_QP_ACCESS_FLAGS), + [IB_QPT_XRC_TGT] = (IB_QP_PKEY_INDEX | + IB_QP_PORT | + IB_QP_ACCESS_FLAGS), + [IB_QPT_SMI] = (IB_QP_PKEY_INDEX | + IB_QP_QKEY), + [IB_QPT_GSI] = (IB_QP_PKEY_INDEX | + IB_QP_QKEY), + } + }, + [IB_QPS_RTR] = { + .valid = 1, + .req_param = { + [IB_QPT_UC] = (IB_QP_AV | + IB_QP_PATH_MTU | + IB_QP_DEST_QPN | + IB_QP_RQ_PSN), + [IB_QPT_RC] = (IB_QP_AV | + IB_QP_PATH_MTU | + IB_QP_DEST_QPN | + IB_QP_RQ_PSN | + IB_QP_MAX_DEST_RD_ATOMIC | + IB_QP_MIN_RNR_TIMER), + [IB_QPT_XRC_INI] = (IB_QP_AV | + IB_QP_PATH_MTU | + IB_QP_DEST_QPN | + IB_QP_RQ_PSN), + [IB_QPT_XRC_TGT] = (IB_QP_AV | + IB_QP_PATH_MTU | + IB_QP_DEST_QPN | + IB_QP_RQ_PSN | + IB_QP_MAX_DEST_RD_ATOMIC | + IB_QP_MIN_RNR_TIMER), + }, + .opt_param = { + [IB_QPT_UD] = (IB_QP_PKEY_INDEX | + IB_QP_QKEY), + [IB_QPT_UC] = (IB_QP_ALT_PATH | + IB_QP_ACCESS_FLAGS | + IB_QP_PKEY_INDEX), + [IB_QPT_RC] = (IB_QP_ALT_PATH | + IB_QP_ACCESS_FLAGS | + IB_QP_PKEY_INDEX), + [IB_QPT_XRC_INI] = (IB_QP_ALT_PATH | + IB_QP_ACCESS_FLAGS | + IB_QP_PKEY_INDEX), + [IB_QPT_XRC_TGT] = (IB_QP_ALT_PATH | + IB_QP_ACCESS_FLAGS | + IB_QP_PKEY_INDEX), + [IB_QPT_SMI] = (IB_QP_PKEY_INDEX | + IB_QP_QKEY), + [IB_QPT_GSI] = (IB_QP_PKEY_INDEX | + IB_QP_QKEY), + }, + }, + }, + [IB_QPS_RTR] = { + [IB_QPS_RESET] = { .valid = 1 }, + [IB_QPS_ERR] = { .valid = 1 }, + [IB_QPS_RTS] = { + .valid = 1, + .req_param = { + [IB_QPT_UD] = IB_QP_SQ_PSN, + [IB_QPT_UC] = IB_QP_SQ_PSN, + [IB_QPT_RC] = (IB_QP_TIMEOUT | + IB_QP_RETRY_CNT | + IB_QP_RNR_RETRY | + IB_QP_SQ_PSN | + IB_QP_MAX_QP_RD_ATOMIC), + [IB_QPT_XRC_INI] = (IB_QP_TIMEOUT | + IB_QP_RETRY_CNT | + IB_QP_RNR_RETRY | + IB_QP_SQ_PSN | + IB_QP_MAX_QP_RD_ATOMIC), + [IB_QPT_XRC_TGT] = (IB_QP_TIMEOUT | + IB_QP_SQ_PSN), + [IB_QPT_SMI] = IB_QP_SQ_PSN, + [IB_QPT_GSI] = IB_QP_SQ_PSN, + }, + .opt_param = { + [IB_QPT_UD] = (IB_QP_CUR_STATE | + IB_QP_QKEY), + [IB_QPT_UC] = (IB_QP_CUR_STATE | + IB_QP_ALT_PATH | + IB_QP_ACCESS_FLAGS | + IB_QP_PATH_MIG_STATE), + [IB_QPT_RC] = (IB_QP_CUR_STATE | + IB_QP_ALT_PATH | + IB_QP_ACCESS_FLAGS | + IB_QP_MIN_RNR_TIMER | + IB_QP_PATH_MIG_STATE), + [IB_QPT_XRC_INI] = (IB_QP_CUR_STATE | + IB_QP_ALT_PATH | + IB_QP_ACCESS_FLAGS | + IB_QP_PATH_MIG_STATE), + [IB_QPT_XRC_TGT] = (IB_QP_CUR_STATE | + IB_QP_ALT_PATH | + IB_QP_ACCESS_FLAGS | + IB_QP_MIN_RNR_TIMER | + IB_QP_PATH_MIG_STATE), + [IB_QPT_SMI] = (IB_QP_CUR_STATE | + IB_QP_QKEY), + [IB_QPT_GSI] = (IB_QP_CUR_STATE | + IB_QP_QKEY), + [IB_QPT_RAW_PACKET] = IB_QP_RATE_LIMIT, + } + } + }, + [IB_QPS_RTS] = { + [IB_QPS_RESET] = { .valid = 1 }, + [IB_QPS_ERR] = { .valid = 1 }, + [IB_QPS_RTS] = { + .valid = 1, + .opt_param = { + [IB_QPT_UD] = (IB_QP_CUR_STATE | + IB_QP_QKEY), + [IB_QPT_UC] = (IB_QP_CUR_STATE | + IB_QP_ACCESS_FLAGS | + IB_QP_ALT_PATH | + IB_QP_PATH_MIG_STATE), + [IB_QPT_RC] = (IB_QP_CUR_STATE | + IB_QP_ACCESS_FLAGS | + IB_QP_ALT_PATH | + IB_QP_PATH_MIG_STATE | + IB_QP_MIN_RNR_TIMER), + [IB_QPT_XRC_INI] = (IB_QP_CUR_STATE | + IB_QP_ACCESS_FLAGS | + IB_QP_ALT_PATH | + IB_QP_PATH_MIG_STATE), + [IB_QPT_XRC_TGT] = (IB_QP_CUR_STATE | + IB_QP_ACCESS_FLAGS | + IB_QP_ALT_PATH | + IB_QP_PATH_MIG_STATE | + IB_QP_MIN_RNR_TIMER), + [IB_QPT_SMI] = (IB_QP_CUR_STATE | + IB_QP_QKEY), + [IB_QPT_GSI] = (IB_QP_CUR_STATE | + IB_QP_QKEY), + [IB_QPT_RAW_PACKET] = IB_QP_RATE_LIMIT, + } + }, + [IB_QPS_SQD] = { + .valid = 1, + .opt_param = { + [IB_QPT_UD] = IB_QP_EN_SQD_ASYNC_NOTIFY, + [IB_QPT_UC] = IB_QP_EN_SQD_ASYNC_NOTIFY, + [IB_QPT_RC] = IB_QP_EN_SQD_ASYNC_NOTIFY, + [IB_QPT_XRC_INI] = IB_QP_EN_SQD_ASYNC_NOTIFY, + [IB_QPT_XRC_TGT] = IB_QP_EN_SQD_ASYNC_NOTIFY, /* ??? */ + [IB_QPT_SMI] = IB_QP_EN_SQD_ASYNC_NOTIFY, + [IB_QPT_GSI] = IB_QP_EN_SQD_ASYNC_NOTIFY + } + }, + }, + [IB_QPS_SQD] = { + [IB_QPS_RESET] = { .valid = 1 }, + [IB_QPS_ERR] = { .valid = 1 }, + [IB_QPS_RTS] = { + .valid = 1, + .opt_param = { + [IB_QPT_UD] = (IB_QP_CUR_STATE | + IB_QP_QKEY), + [IB_QPT_UC] = (IB_QP_CUR_STATE | + IB_QP_ALT_PATH | + IB_QP_ACCESS_FLAGS | + IB_QP_PATH_MIG_STATE), + [IB_QPT_RC] = (IB_QP_CUR_STATE | + IB_QP_ALT_PATH | + IB_QP_ACCESS_FLAGS | + IB_QP_MIN_RNR_TIMER | + IB_QP_PATH_MIG_STATE), + [IB_QPT_XRC_INI] = (IB_QP_CUR_STATE | + IB_QP_ALT_PATH | + IB_QP_ACCESS_FLAGS | + IB_QP_PATH_MIG_STATE), + [IB_QPT_XRC_TGT] = (IB_QP_CUR_STATE | + IB_QP_ALT_PATH | + IB_QP_ACCESS_FLAGS | + IB_QP_MIN_RNR_TIMER | + IB_QP_PATH_MIG_STATE), + [IB_QPT_SMI] = (IB_QP_CUR_STATE | + IB_QP_QKEY), + [IB_QPT_GSI] = (IB_QP_CUR_STATE | + IB_QP_QKEY), + } + }, + [IB_QPS_SQD] = { + .valid = 1, + .opt_param = { + [IB_QPT_UD] = (IB_QP_PKEY_INDEX | + IB_QP_QKEY), + [IB_QPT_UC] = (IB_QP_AV | + IB_QP_ALT_PATH | + IB_QP_ACCESS_FLAGS | + IB_QP_PKEY_INDEX | + IB_QP_PATH_MIG_STATE), + [IB_QPT_RC] = (IB_QP_PORT | + IB_QP_AV | + IB_QP_TIMEOUT | + IB_QP_RETRY_CNT | + IB_QP_RNR_RETRY | + IB_QP_MAX_QP_RD_ATOMIC | + IB_QP_MAX_DEST_RD_ATOMIC | + IB_QP_ALT_PATH | + IB_QP_ACCESS_FLAGS | + IB_QP_PKEY_INDEX | + IB_QP_MIN_RNR_TIMER | + IB_QP_PATH_MIG_STATE), + [IB_QPT_XRC_INI] = (IB_QP_PORT | + IB_QP_AV | + IB_QP_TIMEOUT | + IB_QP_RETRY_CNT | + IB_QP_RNR_RETRY | + IB_QP_MAX_QP_RD_ATOMIC | + IB_QP_ALT_PATH | + IB_QP_ACCESS_FLAGS | + IB_QP_PKEY_INDEX | + IB_QP_PATH_MIG_STATE), + [IB_QPT_XRC_TGT] = (IB_QP_PORT | + IB_QP_AV | + IB_QP_TIMEOUT | + IB_QP_MAX_DEST_RD_ATOMIC | + IB_QP_ALT_PATH | + IB_QP_ACCESS_FLAGS | + IB_QP_PKEY_INDEX | + IB_QP_MIN_RNR_TIMER | + IB_QP_PATH_MIG_STATE), + [IB_QPT_SMI] = (IB_QP_PKEY_INDEX | + IB_QP_QKEY), + [IB_QPT_GSI] = (IB_QP_PKEY_INDEX | + IB_QP_QKEY), + } + } + }, + [IB_QPS_SQE] = { + [IB_QPS_RESET] = { .valid = 1 }, + [IB_QPS_ERR] = { .valid = 1 }, + [IB_QPS_RTS] = { + .valid = 1, + .opt_param = { + [IB_QPT_UD] = (IB_QP_CUR_STATE | + IB_QP_QKEY), + [IB_QPT_UC] = (IB_QP_CUR_STATE | + IB_QP_ACCESS_FLAGS), + [IB_QPT_SMI] = (IB_QP_CUR_STATE | + IB_QP_QKEY), + [IB_QPT_GSI] = (IB_QP_CUR_STATE | + IB_QP_QKEY), + } + } + }, + [IB_QPS_ERR] = { + [IB_QPS_RESET] = { .valid = 1 }, + [IB_QPS_ERR] = { .valid = 1 } + } +}; + +bool ib_modify_qp_is_ok(enum ib_qp_state cur_state, enum ib_qp_state next_state, + enum ib_qp_type type, enum ib_qp_attr_mask mask) +{ + enum ib_qp_attr_mask req_param, opt_param; + + if (mask & IB_QP_CUR_STATE && + cur_state != IB_QPS_RTR && cur_state != IB_QPS_RTS && + cur_state != IB_QPS_SQD && cur_state != IB_QPS_SQE) + return false; + + if (!qp_state_table[cur_state][next_state].valid) + return false; + + req_param = qp_state_table[cur_state][next_state].req_param[type]; + opt_param = qp_state_table[cur_state][next_state].opt_param[type]; + + if ((mask & req_param) != req_param) + return false; + + if (mask & ~(req_param | opt_param | IB_QP_STATE)) + return false; + + return true; +} +EXPORT_SYMBOL(ib_modify_qp_is_ok); + +/** + * ib_resolve_eth_dmac - Resolve destination mac address + * @device: Device to consider + * @ah_attr: address handle attribute which describes the + * source and destination parameters + * ib_resolve_eth_dmac() resolves destination mac address and L3 hop limit It + * returns 0 on success or appropriate error code. It initializes the + * necessary ah_attr fields when call is successful. + */ +static int ib_resolve_eth_dmac(struct ib_device *device, + struct rdma_ah_attr *ah_attr) +{ + int ret = 0; + + if (rdma_is_multicast_addr((struct in6_addr *)ah_attr->grh.dgid.raw)) { + if (ipv6_addr_v4mapped((struct in6_addr *)ah_attr->grh.dgid.raw)) { + __be32 addr = 0; + + memcpy(&addr, ah_attr->grh.dgid.raw + 12, 4); + ip_eth_mc_map(addr, (char *)ah_attr->roce.dmac); + } else { + ipv6_eth_mc_map((struct in6_addr *)ah_attr->grh.dgid.raw, + (char *)ah_attr->roce.dmac); + } + } else { + ret = ib_resolve_unicast_gid_dmac(device, ah_attr); + } + return ret; +} + +static bool is_qp_type_connected(const struct ib_qp *qp) +{ + return (qp->qp_type == IB_QPT_UC || + qp->qp_type == IB_QPT_RC || + qp->qp_type == IB_QPT_XRC_INI || + qp->qp_type == IB_QPT_XRC_TGT); +} + +/** + * IB core internal function to perform QP attributes modification. + */ +static int _ib_modify_qp(struct ib_qp *qp, struct ib_qp_attr *attr, + int attr_mask, struct ib_udata *udata) +{ + u8 port = attr_mask & IB_QP_PORT ? attr->port_num : qp->port; + const struct ib_gid_attr *old_sgid_attr_av; + const struct ib_gid_attr *old_sgid_attr_alt_av; + int ret; + + attr->xmit_slave = NULL; + if (attr_mask & IB_QP_AV) { + ret = rdma_fill_sgid_attr(qp->device, &attr->ah_attr, + &old_sgid_attr_av); + if (ret) + return ret; + + if (attr->ah_attr.type == RDMA_AH_ATTR_TYPE_ROCE && + is_qp_type_connected(qp)) { + struct net_device *slave; + + /* + * If the user provided the qp_attr then we have to + * resolve it. Kerne users have to provide already + * resolved rdma_ah_attr's. + */ + if (udata) { + ret = ib_resolve_eth_dmac(qp->device, + &attr->ah_attr); + if (ret) + goto out_av; + } + slave = rdma_lag_get_ah_roce_slave(qp->device, + &attr->ah_attr, + GFP_KERNEL); + if (IS_ERR(slave)) { + ret = PTR_ERR(slave); + goto out_av; + } + attr->xmit_slave = slave; + } + } + if (attr_mask & IB_QP_ALT_PATH) { + /* + * FIXME: This does not track the migration state, so if the + * user loads a new alternate path after the HW has migrated + * from primary->alternate we will keep the wrong + * references. This is OK for IB because the reference + * counting does not serve any functional purpose. + */ + ret = rdma_fill_sgid_attr(qp->device, &attr->alt_ah_attr, + &old_sgid_attr_alt_av); + if (ret) + goto out_av; + + /* + * Today the core code can only handle alternate paths and APM + * for IB. Ban them in roce mode. + */ + if (!(rdma_protocol_ib(qp->device, + attr->alt_ah_attr.port_num) && + rdma_protocol_ib(qp->device, port))) { + ret = -EINVAL; + goto out; + } + } + + if (rdma_ib_or_roce(qp->device, port)) { + if (attr_mask & IB_QP_RQ_PSN && attr->rq_psn & ~0xffffff) { + dev_warn(&qp->device->dev, + "%s rq_psn overflow, masking to 24 bits\n", + __func__); + attr->rq_psn &= 0xffffff; + } + + if (attr_mask & IB_QP_SQ_PSN && attr->sq_psn & ~0xffffff) { + dev_warn(&qp->device->dev, + " %s sq_psn overflow, masking to 24 bits\n", + __func__); + attr->sq_psn &= 0xffffff; + } + } + + /* + * Bind this qp to a counter automatically based on the rdma counter + * rules. This only set in RST2INIT with port specified + */ + if (!qp->counter && (attr_mask & IB_QP_PORT) && + ((attr_mask & IB_QP_STATE) && attr->qp_state == IB_QPS_INIT)) + rdma_counter_bind_qp_auto(qp, attr->port_num); + + ret = ib_security_modify_qp(qp, attr, attr_mask, udata); + if (ret) + goto out; + + if (attr_mask & IB_QP_PORT) + qp->port = attr->port_num; + if (attr_mask & IB_QP_AV) + qp->av_sgid_attr = + rdma_update_sgid_attr(&attr->ah_attr, qp->av_sgid_attr); + if (attr_mask & IB_QP_ALT_PATH) + qp->alt_path_sgid_attr = rdma_update_sgid_attr( + &attr->alt_ah_attr, qp->alt_path_sgid_attr); + +out: + if (attr_mask & IB_QP_ALT_PATH) + rdma_unfill_sgid_attr(&attr->alt_ah_attr, old_sgid_attr_alt_av); +out_av: + if (attr_mask & IB_QP_AV) { + rdma_lag_put_ah_roce_slave(attr->xmit_slave); + rdma_unfill_sgid_attr(&attr->ah_attr, old_sgid_attr_av); + } + return ret; +} + +/** + * ib_modify_qp_with_udata - Modifies the attributes for the specified QP. + * @ib_qp: The QP to modify. + * @attr: On input, specifies the QP attributes to modify. On output, + * the current values of selected QP attributes are returned. + * @attr_mask: A bit-mask used to specify which attributes of the QP + * are being modified. + * @udata: pointer to user's input output buffer information + * are being modified. + * It returns 0 on success and returns appropriate error code on error. + */ +int ib_modify_qp_with_udata(struct ib_qp *ib_qp, struct ib_qp_attr *attr, + int attr_mask, struct ib_udata *udata) +{ + return _ib_modify_qp(ib_qp->real_qp, attr, attr_mask, udata); +} +EXPORT_SYMBOL(ib_modify_qp_with_udata); + +int ib_get_eth_speed(struct ib_device *dev, u8 port_num, u16 *speed, u8 *width) +{ + int rc; + u32 netdev_speed; + struct net_device *netdev; + struct ethtool_link_ksettings lksettings; + + if (rdma_port_get_link_layer(dev, port_num) != IB_LINK_LAYER_ETHERNET) + return -EINVAL; + + netdev = ib_device_get_netdev(dev, port_num); + if (!netdev) + return -ENODEV; + + rtnl_lock(); + rc = __ethtool_get_link_ksettings(netdev, &lksettings); + rtnl_unlock(); + + dev_put(netdev); + + if (!rc && lksettings.base.speed != (u32)SPEED_UNKNOWN) { + netdev_speed = lksettings.base.speed; + } else { + netdev_speed = SPEED_1000; + pr_warn("%s speed is unknown, defaulting to %d\n", netdev->name, + netdev_speed); + } + + if (netdev_speed <= SPEED_1000) { + *width = IB_WIDTH_1X; + *speed = IB_SPEED_SDR; + } else if (netdev_speed <= SPEED_10000) { + *width = IB_WIDTH_1X; + *speed = IB_SPEED_FDR10; + } else if (netdev_speed <= SPEED_20000) { + *width = IB_WIDTH_4X; + *speed = IB_SPEED_DDR; + } else if (netdev_speed <= SPEED_25000) { + *width = IB_WIDTH_1X; + *speed = IB_SPEED_EDR; + } else if (netdev_speed <= SPEED_40000) { + *width = IB_WIDTH_4X; + *speed = IB_SPEED_FDR10; + } else { + *width = IB_WIDTH_4X; + *speed = IB_SPEED_EDR; + } + + return 0; +} +EXPORT_SYMBOL(ib_get_eth_speed); + +int ib_modify_qp(struct ib_qp *qp, + struct ib_qp_attr *qp_attr, + int qp_attr_mask) +{ + return _ib_modify_qp(qp->real_qp, qp_attr, qp_attr_mask, NULL); +} +EXPORT_SYMBOL(ib_modify_qp); + +int ib_query_qp(struct ib_qp *qp, + struct ib_qp_attr *qp_attr, + int qp_attr_mask, + struct ib_qp_init_attr *qp_init_attr) +{ + qp_attr->ah_attr.grh.sgid_attr = NULL; + qp_attr->alt_ah_attr.grh.sgid_attr = NULL; + + return qp->device->ops.query_qp ? + qp->device->ops.query_qp(qp->real_qp, qp_attr, qp_attr_mask, + qp_init_attr) : -EOPNOTSUPP; +} +EXPORT_SYMBOL(ib_query_qp); + +int ib_close_qp(struct ib_qp *qp) +{ + struct ib_qp *real_qp; + unsigned long flags; + + real_qp = qp->real_qp; + if (real_qp == qp) + return -EINVAL; + + spin_lock_irqsave(&real_qp->device->qp_open_list_lock, flags); + list_del(&qp->open_list); + spin_unlock_irqrestore(&real_qp->device->qp_open_list_lock, flags); + + atomic_dec(&real_qp->usecnt); + if (qp->qp_sec) + ib_close_shared_qp_security(qp->qp_sec); + kfree(qp); + + return 0; +} +EXPORT_SYMBOL(ib_close_qp); + +static int __ib_destroy_shared_qp(struct ib_qp *qp) +{ + struct ib_xrcd *xrcd; + struct ib_qp *real_qp; + int ret; + + real_qp = qp->real_qp; + xrcd = real_qp->xrcd; + down_write(&xrcd->tgt_qps_rwsem); + ib_close_qp(qp); + if (atomic_read(&real_qp->usecnt) == 0) + xa_erase(&xrcd->tgt_qps, real_qp->qp_num); + else + real_qp = NULL; + up_write(&xrcd->tgt_qps_rwsem); + + if (real_qp) { + ret = ib_destroy_qp(real_qp); + if (!ret) + atomic_dec(&xrcd->usecnt); + } + + return 0; +} + +int ib_destroy_qp_user(struct ib_qp *qp, struct ib_udata *udata) +{ + const struct ib_gid_attr *alt_path_sgid_attr = qp->alt_path_sgid_attr; + const struct ib_gid_attr *av_sgid_attr = qp->av_sgid_attr; + struct ib_pd *pd; + struct ib_cq *scq, *rcq; + struct ib_srq *srq; + struct ib_rwq_ind_table *ind_tbl; + struct ib_qp_security *sec; + int ret; + + WARN_ON_ONCE(qp->mrs_used > 0); + + if (atomic_read(&qp->usecnt)) + return -EBUSY; + + if (qp->real_qp != qp) + return __ib_destroy_shared_qp(qp); + + pd = qp->pd; + scq = qp->send_cq; + rcq = qp->recv_cq; + srq = qp->srq; + ind_tbl = qp->rwq_ind_tbl; + sec = qp->qp_sec; + if (sec) + ib_destroy_qp_security_begin(sec); + + if (!qp->uobject) + rdma_rw_cleanup_mrs(qp); + + rdma_counter_unbind_qp(qp, true); + rdma_restrack_del(&qp->res); + ret = qp->device->ops.destroy_qp(qp, udata); + if (!ret) { + if (alt_path_sgid_attr) + rdma_put_gid_attr(alt_path_sgid_attr); + if (av_sgid_attr) + rdma_put_gid_attr(av_sgid_attr); + if (pd) + atomic_dec(&pd->usecnt); + if (scq) + atomic_dec(&scq->usecnt); + if (rcq) + atomic_dec(&rcq->usecnt); + if (srq) + atomic_dec(&srq->usecnt); + if (ind_tbl) + atomic_dec(&ind_tbl->usecnt); + if (sec) + ib_destroy_qp_security_end(sec); + } else { + if (sec) + ib_destroy_qp_security_abort(sec); + } + + return ret; +} +EXPORT_SYMBOL(ib_destroy_qp_user); + +/* Completion queues */ + +struct ib_cq *__ib_create_cq(struct ib_device *device, + ib_comp_handler comp_handler, + void (*event_handler)(struct ib_event *, void *), + void *cq_context, + const struct ib_cq_init_attr *cq_attr, + const char *caller) +{ + struct ib_cq *cq; + int ret; + + cq = rdma_zalloc_drv_obj(device, ib_cq); + if (!cq) + return ERR_PTR(-ENOMEM); + + cq->device = device; + cq->uobject = NULL; + cq->comp_handler = comp_handler; + cq->event_handler = event_handler; + cq->cq_context = cq_context; + atomic_set(&cq->usecnt, 0); + + rdma_restrack_new(&cq->res, RDMA_RESTRACK_CQ); + rdma_restrack_set_name(&cq->res, caller); + + ret = device->ops.create_cq(cq, cq_attr, NULL); + if (ret) { + rdma_restrack_put(&cq->res); + kfree(cq); + return ERR_PTR(ret); + } + + rdma_restrack_add(&cq->res); + return cq; +} +EXPORT_SYMBOL(__ib_create_cq); + +int rdma_set_cq_moderation(struct ib_cq *cq, u16 cq_count, u16 cq_period) +{ + if (cq->shared) + return -EOPNOTSUPP; + + return cq->device->ops.modify_cq ? + cq->device->ops.modify_cq(cq, cq_count, + cq_period) : -EOPNOTSUPP; +} +EXPORT_SYMBOL(rdma_set_cq_moderation); + +int ib_destroy_cq_user(struct ib_cq *cq, struct ib_udata *udata) +{ + int ret; + + if (WARN_ON_ONCE(cq->shared)) + return -EOPNOTSUPP; + + if (atomic_read(&cq->usecnt)) + return -EBUSY; + + ret = cq->device->ops.destroy_cq(cq, udata); + if (ret) + return ret; + + rdma_restrack_del(&cq->res); + kfree(cq); + return ret; +} +EXPORT_SYMBOL(ib_destroy_cq_user); + +int ib_resize_cq(struct ib_cq *cq, int cqe) +{ + if (cq->shared) + return -EOPNOTSUPP; + + return cq->device->ops.resize_cq ? + cq->device->ops.resize_cq(cq, cqe, NULL) : -EOPNOTSUPP; +} +EXPORT_SYMBOL(ib_resize_cq); + +/* Memory regions */ + +struct ib_mr *ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length, + u64 virt_addr, int access_flags) +{ + struct ib_mr *mr; + + if (access_flags & IB_ACCESS_ON_DEMAND) { + if (!(pd->device->attrs.device_cap_flags & + IB_DEVICE_ON_DEMAND_PAGING)) { + pr_debug("ODP support not available\n"); + return ERR_PTR(-EINVAL); + } + } + + mr = pd->device->ops.reg_user_mr(pd, start, length, virt_addr, + access_flags, NULL); + + if (IS_ERR(mr)) + return mr; + + mr->device = pd->device; + mr->type = IB_MR_TYPE_USER; + mr->pd = pd; + mr->dm = NULL; + atomic_inc(&pd->usecnt); + mr->iova = virt_addr; + mr->length = length; + + rdma_restrack_new(&mr->res, RDMA_RESTRACK_MR); + rdma_restrack_parent_name(&mr->res, &pd->res); + rdma_restrack_add(&mr->res); + + return mr; +} +EXPORT_SYMBOL(ib_reg_user_mr); + +int ib_advise_mr(struct ib_pd *pd, enum ib_uverbs_advise_mr_advice advice, + u32 flags, struct ib_sge *sg_list, u32 num_sge) +{ + if (!pd->device->ops.advise_mr) + return -EOPNOTSUPP; + + if (!num_sge) + return 0; + + return pd->device->ops.advise_mr(pd, advice, flags, sg_list, num_sge, + NULL); +} +EXPORT_SYMBOL(ib_advise_mr); + +int ib_dereg_mr_user(struct ib_mr *mr, struct ib_udata *udata) +{ + struct ib_pd *pd = mr->pd; + struct ib_dm *dm = mr->dm; + struct ib_sig_attrs *sig_attrs = mr->sig_attrs; + int ret; + + trace_mr_dereg(mr); + rdma_restrack_del(&mr->res); + ret = mr->device->ops.dereg_mr(mr, udata); + if (!ret) { + atomic_dec(&pd->usecnt); + if (dm) + atomic_dec(&dm->usecnt); + kfree(sig_attrs); + } + + return ret; +} +EXPORT_SYMBOL(ib_dereg_mr_user); + +/** + * ib_alloc_mr() - Allocates a memory region + * @pd: protection domain associated with the region + * @mr_type: memory region type + * @max_num_sg: maximum sg entries available for registration. + * + * Notes: + * Memory registeration page/sg lists must not exceed max_num_sg. + * For mr_type IB_MR_TYPE_MEM_REG, the total length cannot exceed + * max_num_sg * used_page_size. + * + */ +struct ib_mr *ib_alloc_mr(struct ib_pd *pd, enum ib_mr_type mr_type, + u32 max_num_sg) +{ + struct ib_mr *mr; + + if (!pd->device->ops.alloc_mr) { + mr = ERR_PTR(-EOPNOTSUPP); + goto out; + } + + if (mr_type == IB_MR_TYPE_INTEGRITY) { + WARN_ON_ONCE(1); + mr = ERR_PTR(-EINVAL); + goto out; + } + + mr = pd->device->ops.alloc_mr(pd, mr_type, max_num_sg); + if (IS_ERR(mr)) + goto out; + + mr->device = pd->device; + mr->pd = pd; + mr->dm = NULL; + mr->uobject = NULL; + atomic_inc(&pd->usecnt); + mr->need_inval = false; + mr->type = mr_type; + mr->sig_attrs = NULL; + + rdma_restrack_new(&mr->res, RDMA_RESTRACK_MR); + rdma_restrack_parent_name(&mr->res, &pd->res); + rdma_restrack_add(&mr->res); +out: + trace_mr_alloc(pd, mr_type, max_num_sg, mr); + return mr; +} +EXPORT_SYMBOL(ib_alloc_mr); + +/** + * ib_alloc_mr_integrity() - Allocates an integrity memory region + * @pd: protection domain associated with the region + * @max_num_data_sg: maximum data sg entries available for registration + * @max_num_meta_sg: maximum metadata sg entries available for + * registration + * + * Notes: + * Memory registration page/sg lists must not exceed max_num_sg, + * also the integrity page/sg lists must not exceed max_num_meta_sg. + * + */ +struct ib_mr *ib_alloc_mr_integrity(struct ib_pd *pd, + u32 max_num_data_sg, + u32 max_num_meta_sg) +{ + struct ib_mr *mr; + struct ib_sig_attrs *sig_attrs; + + if (!pd->device->ops.alloc_mr_integrity || + !pd->device->ops.map_mr_sg_pi) { + mr = ERR_PTR(-EOPNOTSUPP); + goto out; + } + + if (!max_num_meta_sg) { + mr = ERR_PTR(-EINVAL); + goto out; + } + + sig_attrs = kzalloc(sizeof(struct ib_sig_attrs), GFP_KERNEL); + if (!sig_attrs) { + mr = ERR_PTR(-ENOMEM); + goto out; + } + + mr = pd->device->ops.alloc_mr_integrity(pd, max_num_data_sg, + max_num_meta_sg); + if (IS_ERR(mr)) { + kfree(sig_attrs); + goto out; + } + + mr->device = pd->device; + mr->pd = pd; + mr->dm = NULL; + mr->uobject = NULL; + atomic_inc(&pd->usecnt); + mr->need_inval = false; + mr->type = IB_MR_TYPE_INTEGRITY; + mr->sig_attrs = sig_attrs; + + rdma_restrack_new(&mr->res, RDMA_RESTRACK_MR); + rdma_restrack_parent_name(&mr->res, &pd->res); + rdma_restrack_add(&mr->res); +out: + trace_mr_integ_alloc(pd, max_num_data_sg, max_num_meta_sg, mr); + return mr; +} +EXPORT_SYMBOL(ib_alloc_mr_integrity); + +/* Multicast groups */ + +static bool is_valid_mcast_lid(struct ib_qp *qp, u16 lid) +{ + struct ib_qp_init_attr init_attr = {}; + struct ib_qp_attr attr = {}; + int num_eth_ports = 0; + int port; + + /* If QP state >= init, it is assigned to a port and we can check this + * port only. + */ + if (!ib_query_qp(qp, &attr, IB_QP_STATE | IB_QP_PORT, &init_attr)) { + if (attr.qp_state >= IB_QPS_INIT) { + if (rdma_port_get_link_layer(qp->device, attr.port_num) != + IB_LINK_LAYER_INFINIBAND) + return true; + goto lid_check; + } + } + + /* Can't get a quick answer, iterate over all ports */ + for (port = 0; port < qp->device->phys_port_cnt; port++) + if (rdma_port_get_link_layer(qp->device, port) != + IB_LINK_LAYER_INFINIBAND) + num_eth_ports++; + + /* If we have at lease one Ethernet port, RoCE annex declares that + * multicast LID should be ignored. We can't tell at this step if the + * QP belongs to an IB or Ethernet port. + */ + if (num_eth_ports) + return true; + + /* If all the ports are IB, we can check according to IB spec. */ +lid_check: + return !(lid < be16_to_cpu(IB_MULTICAST_LID_BASE) || + lid == be16_to_cpu(IB_LID_PERMISSIVE)); +} + +int ib_attach_mcast(struct ib_qp *qp, union ib_gid *gid, u16 lid) +{ + int ret; + + if (!qp->device->ops.attach_mcast) + return -EOPNOTSUPP; + + if (!rdma_is_multicast_addr((struct in6_addr *)gid->raw) || + qp->qp_type != IB_QPT_UD || !is_valid_mcast_lid(qp, lid)) + return -EINVAL; + + ret = qp->device->ops.attach_mcast(qp, gid, lid); + if (!ret) + atomic_inc(&qp->usecnt); + return ret; +} +EXPORT_SYMBOL(ib_attach_mcast); + +int ib_detach_mcast(struct ib_qp *qp, union ib_gid *gid, u16 lid) +{ + int ret; + + if (!qp->device->ops.detach_mcast) + return -EOPNOTSUPP; + + if (!rdma_is_multicast_addr((struct in6_addr *)gid->raw) || + qp->qp_type != IB_QPT_UD || !is_valid_mcast_lid(qp, lid)) + return -EINVAL; + + ret = qp->device->ops.detach_mcast(qp, gid, lid); + if (!ret) + atomic_dec(&qp->usecnt); + return ret; +} +EXPORT_SYMBOL(ib_detach_mcast); + +/** + * ib_alloc_xrcd_user - Allocates an XRC domain. + * @device: The device on which to allocate the XRC domain. + * @inode: inode to connect XRCD + * @udata: Valid user data or NULL for kernel object + */ +struct ib_xrcd *ib_alloc_xrcd_user(struct ib_device *device, + struct inode *inode, struct ib_udata *udata) +{ + struct ib_xrcd *xrcd; + int ret; + + if (!device->ops.alloc_xrcd) + return ERR_PTR(-EOPNOTSUPP); + + xrcd = rdma_zalloc_drv_obj(device, ib_xrcd); + if (!xrcd) + return ERR_PTR(-ENOMEM); + + xrcd->device = device; + xrcd->inode = inode; + atomic_set(&xrcd->usecnt, 0); + init_rwsem(&xrcd->tgt_qps_rwsem); + xa_init(&xrcd->tgt_qps); + + ret = device->ops.alloc_xrcd(xrcd, udata); + if (ret) + goto err; + return xrcd; +err: + kfree(xrcd); + return ERR_PTR(ret); +} +EXPORT_SYMBOL(ib_alloc_xrcd_user); + +/** + * ib_dealloc_xrcd_user - Deallocates an XRC domain. + * @xrcd: The XRC domain to deallocate. + * @udata: Valid user data or NULL for kernel object + */ +int ib_dealloc_xrcd_user(struct ib_xrcd *xrcd, struct ib_udata *udata) +{ + int ret; + + if (atomic_read(&xrcd->usecnt)) + return -EBUSY; + + WARN_ON(!xa_empty(&xrcd->tgt_qps)); + ret = xrcd->device->ops.dealloc_xrcd(xrcd, udata); + if (ret) + return ret; + kfree(xrcd); + return ret; +} +EXPORT_SYMBOL(ib_dealloc_xrcd_user); + +/** + * ib_create_wq - Creates a WQ associated with the specified protection + * domain. + * @pd: The protection domain associated with the WQ. + * @wq_attr: A list of initial attributes required to create the + * WQ. If WQ creation succeeds, then the attributes are updated to + * the actual capabilities of the created WQ. + * + * wq_attr->max_wr and wq_attr->max_sge determine + * the requested size of the WQ, and set to the actual values allocated + * on return. + * If ib_create_wq() succeeds, then max_wr and max_sge will always be + * at least as large as the requested values. + */ +struct ib_wq *ib_create_wq(struct ib_pd *pd, + struct ib_wq_init_attr *wq_attr) +{ + struct ib_wq *wq; + + if (!pd->device->ops.create_wq) + return ERR_PTR(-EOPNOTSUPP); + + wq = pd->device->ops.create_wq(pd, wq_attr, NULL); + if (!IS_ERR(wq)) { + wq->event_handler = wq_attr->event_handler; + wq->wq_context = wq_attr->wq_context; + wq->wq_type = wq_attr->wq_type; + wq->cq = wq_attr->cq; + wq->device = pd->device; + wq->pd = pd; + wq->uobject = NULL; + atomic_inc(&pd->usecnt); + atomic_inc(&wq_attr->cq->usecnt); + atomic_set(&wq->usecnt, 0); + } + return wq; +} +EXPORT_SYMBOL(ib_create_wq); + +/** + * ib_destroy_wq_user - Destroys the specified user WQ. + * @wq: The WQ to destroy. + * @udata: Valid user data + */ +int ib_destroy_wq_user(struct ib_wq *wq, struct ib_udata *udata) +{ + struct ib_cq *cq = wq->cq; + struct ib_pd *pd = wq->pd; + int ret; + + if (atomic_read(&wq->usecnt)) + return -EBUSY; + + ret = wq->device->ops.destroy_wq(wq, udata); + if (ret) + return ret; + + atomic_dec(&pd->usecnt); + atomic_dec(&cq->usecnt); + return ret; +} +EXPORT_SYMBOL(ib_destroy_wq_user); + +/** + * ib_modify_wq - Modifies the specified WQ. + * @wq: The WQ to modify. + * @wq_attr: On input, specifies the WQ attributes to modify. + * @wq_attr_mask: A bit-mask used to specify which attributes of the WQ + * are being modified. + * On output, the current values of selected WQ attributes are returned. + */ +int ib_modify_wq(struct ib_wq *wq, struct ib_wq_attr *wq_attr, + u32 wq_attr_mask) +{ + int err; + + if (!wq->device->ops.modify_wq) + return -EOPNOTSUPP; + + err = wq->device->ops.modify_wq(wq, wq_attr, wq_attr_mask, NULL); + return err; +} +EXPORT_SYMBOL(ib_modify_wq); + +int ib_check_mr_status(struct ib_mr *mr, u32 check_mask, + struct ib_mr_status *mr_status) +{ + if (!mr->device->ops.check_mr_status) + return -EOPNOTSUPP; + + return mr->device->ops.check_mr_status(mr, check_mask, mr_status); +} +EXPORT_SYMBOL(ib_check_mr_status); + +int ib_set_vf_link_state(struct ib_device *device, int vf, u8 port, + int state) +{ + if (!device->ops.set_vf_link_state) + return -EOPNOTSUPP; + + return device->ops.set_vf_link_state(device, vf, port, state); +} +EXPORT_SYMBOL(ib_set_vf_link_state); + +int ib_get_vf_config(struct ib_device *device, int vf, u8 port, + struct ifla_vf_info *info) +{ + if (!device->ops.get_vf_config) + return -EOPNOTSUPP; + + return device->ops.get_vf_config(device, vf, port, info); +} +EXPORT_SYMBOL(ib_get_vf_config); + +int ib_get_vf_stats(struct ib_device *device, int vf, u8 port, + struct ifla_vf_stats *stats) +{ + if (!device->ops.get_vf_stats) + return -EOPNOTSUPP; + + return device->ops.get_vf_stats(device, vf, port, stats); +} +EXPORT_SYMBOL(ib_get_vf_stats); + +int ib_set_vf_guid(struct ib_device *device, int vf, u8 port, u64 guid, + int type) +{ + if (!device->ops.set_vf_guid) + return -EOPNOTSUPP; + + return device->ops.set_vf_guid(device, vf, port, guid, type); +} +EXPORT_SYMBOL(ib_set_vf_guid); + +int ib_get_vf_guid(struct ib_device *device, int vf, u8 port, + struct ifla_vf_guid *node_guid, + struct ifla_vf_guid *port_guid) +{ + if (!device->ops.get_vf_guid) + return -EOPNOTSUPP; + + return device->ops.get_vf_guid(device, vf, port, node_guid, port_guid); +} +EXPORT_SYMBOL(ib_get_vf_guid); +/** + * ib_map_mr_sg_pi() - Map the dma mapped SG lists for PI (protection + * information) and set an appropriate memory region for registration. + * @mr: memory region + * @data_sg: dma mapped scatterlist for data + * @data_sg_nents: number of entries in data_sg + * @data_sg_offset: offset in bytes into data_sg + * @meta_sg: dma mapped scatterlist for metadata + * @meta_sg_nents: number of entries in meta_sg + * @meta_sg_offset: offset in bytes into meta_sg + * @page_size: page vector desired page size + * + * Constraints: + * - The MR must be allocated with type IB_MR_TYPE_INTEGRITY. + * + * Return: 0 on success. + * + * After this completes successfully, the memory region + * is ready for registration. + */ +int ib_map_mr_sg_pi(struct ib_mr *mr, struct scatterlist *data_sg, + int data_sg_nents, unsigned int *data_sg_offset, + struct scatterlist *meta_sg, int meta_sg_nents, + unsigned int *meta_sg_offset, unsigned int page_size) +{ + if (unlikely(!mr->device->ops.map_mr_sg_pi || + WARN_ON_ONCE(mr->type != IB_MR_TYPE_INTEGRITY))) + return -EOPNOTSUPP; + + mr->page_size = page_size; + + return mr->device->ops.map_mr_sg_pi(mr, data_sg, data_sg_nents, + data_sg_offset, meta_sg, + meta_sg_nents, meta_sg_offset); +} +EXPORT_SYMBOL(ib_map_mr_sg_pi); + +/** + * ib_map_mr_sg() - Map the largest prefix of a dma mapped SG list + * and set it the memory region. + * @mr: memory region + * @sg: dma mapped scatterlist + * @sg_nents: number of entries in sg + * @sg_offset: offset in bytes into sg + * @page_size: page vector desired page size + * + * Constraints: + * + * - The first sg element is allowed to have an offset. + * - Each sg element must either be aligned to page_size or virtually + * contiguous to the previous element. In case an sg element has a + * non-contiguous offset, the mapping prefix will not include it. + * - The last sg element is allowed to have length less than page_size. + * - If sg_nents total byte length exceeds the mr max_num_sge * page_size + * then only max_num_sg entries will be mapped. + * - If the MR was allocated with type IB_MR_TYPE_SG_GAPS, none of these + * constraints holds and the page_size argument is ignored. + * + * Returns the number of sg elements that were mapped to the memory region. + * + * After this completes successfully, the memory region + * is ready for registration. + */ +int ib_map_mr_sg(struct ib_mr *mr, struct scatterlist *sg, int sg_nents, + unsigned int *sg_offset, unsigned int page_size) +{ + if (unlikely(!mr->device->ops.map_mr_sg)) + return -EOPNOTSUPP; + + mr->page_size = page_size; + + return mr->device->ops.map_mr_sg(mr, sg, sg_nents, sg_offset); +} +EXPORT_SYMBOL(ib_map_mr_sg); + +/** + * ib_sg_to_pages() - Convert the largest prefix of a sg list + * to a page vector + * @mr: memory region + * @sgl: dma mapped scatterlist + * @sg_nents: number of entries in sg + * @sg_offset_p: ==== ======================================================= + * IN start offset in bytes into sg + * OUT offset in bytes for element n of the sg of the first + * byte that has not been processed where n is the return + * value of this function. + * ==== ======================================================= + * @set_page: driver page assignment function pointer + * + * Core service helper for drivers to convert the largest + * prefix of given sg list to a page vector. The sg list + * prefix converted is the prefix that meet the requirements + * of ib_map_mr_sg. + * + * Returns the number of sg elements that were assigned to + * a page vector. + */ +int ib_sg_to_pages(struct ib_mr *mr, struct scatterlist *sgl, int sg_nents, + unsigned int *sg_offset_p, int (*set_page)(struct ib_mr *, u64)) +{ + struct scatterlist *sg; + u64 last_end_dma_addr = 0; + unsigned int sg_offset = sg_offset_p ? *sg_offset_p : 0; + unsigned int last_page_off = 0; + u64 page_mask = ~((u64)mr->page_size - 1); + int i, ret; + + if (unlikely(sg_nents <= 0 || sg_offset > sg_dma_len(&sgl[0]))) + return -EINVAL; + + mr->iova = sg_dma_address(&sgl[0]) + sg_offset; + mr->length = 0; + + for_each_sg(sgl, sg, sg_nents, i) { + u64 dma_addr = sg_dma_address(sg) + sg_offset; + u64 prev_addr = dma_addr; + unsigned int dma_len = sg_dma_len(sg) - sg_offset; + u64 end_dma_addr = dma_addr + dma_len; + u64 page_addr = dma_addr & page_mask; + + /* + * For the second and later elements, check whether either the + * end of element i-1 or the start of element i is not aligned + * on a page boundary. + */ + if (i && (last_page_off != 0 || page_addr != dma_addr)) { + /* Stop mapping if there is a gap. */ + if (last_end_dma_addr != dma_addr) + break; + + /* + * Coalesce this element with the last. If it is small + * enough just update mr->length. Otherwise start + * mapping from the next page. + */ + goto next_page; + } + + do { + ret = set_page(mr, page_addr); + if (unlikely(ret < 0)) { + sg_offset = prev_addr - sg_dma_address(sg); + mr->length += prev_addr - dma_addr; + if (sg_offset_p) + *sg_offset_p = sg_offset; + return i || sg_offset ? i : ret; + } + prev_addr = page_addr; +next_page: + page_addr += mr->page_size; + } while (page_addr < end_dma_addr); + + mr->length += dma_len; + last_end_dma_addr = end_dma_addr; + last_page_off = end_dma_addr & ~page_mask; + + sg_offset = 0; + } + + if (sg_offset_p) + *sg_offset_p = 0; + return i; +} +EXPORT_SYMBOL(ib_sg_to_pages); + +struct ib_drain_cqe { + struct ib_cqe cqe; + struct completion done; +}; + +static void ib_drain_qp_done(struct ib_cq *cq, struct ib_wc *wc) +{ + struct ib_drain_cqe *cqe = container_of(wc->wr_cqe, struct ib_drain_cqe, + cqe); + + complete(&cqe->done); +} + +/* + * Post a WR and block until its completion is reaped for the SQ. + */ +static void __ib_drain_sq(struct ib_qp *qp) +{ + struct ib_cq *cq = qp->send_cq; + struct ib_qp_attr attr = { .qp_state = IB_QPS_ERR }; + struct ib_drain_cqe sdrain; + struct ib_rdma_wr swr = { + .wr = { + .next = NULL, + { .wr_cqe = &sdrain.cqe, }, + .opcode = IB_WR_RDMA_WRITE, + }, + }; + int ret; + + ret = ib_modify_qp(qp, &attr, IB_QP_STATE); + if (ret) { + WARN_ONCE(ret, "failed to drain send queue: %d\n", ret); + return; + } + + sdrain.cqe.done = ib_drain_qp_done; + init_completion(&sdrain.done); + + ret = ib_post_send(qp, &swr.wr, NULL); + if (ret) { + WARN_ONCE(ret, "failed to drain send queue: %d\n", ret); + return; + } + + if (cq->poll_ctx == IB_POLL_DIRECT) + while (wait_for_completion_timeout(&sdrain.done, HZ / 10) <= 0) + ib_process_cq_direct(cq, -1); + else + wait_for_completion(&sdrain.done); +} + +/* + * Post a WR and block until its completion is reaped for the RQ. + */ +static void __ib_drain_rq(struct ib_qp *qp) +{ + struct ib_cq *cq = qp->recv_cq; + struct ib_qp_attr attr = { .qp_state = IB_QPS_ERR }; + struct ib_drain_cqe rdrain; + struct ib_recv_wr rwr = {}; + int ret; + + ret = ib_modify_qp(qp, &attr, IB_QP_STATE); + if (ret) { + WARN_ONCE(ret, "failed to drain recv queue: %d\n", ret); + return; + } + + rwr.wr_cqe = &rdrain.cqe; + rdrain.cqe.done = ib_drain_qp_done; + init_completion(&rdrain.done); + + ret = ib_post_recv(qp, &rwr, NULL); + if (ret) { + WARN_ONCE(ret, "failed to drain recv queue: %d\n", ret); + return; + } + + if (cq->poll_ctx == IB_POLL_DIRECT) + while (wait_for_completion_timeout(&rdrain.done, HZ / 10) <= 0) + ib_process_cq_direct(cq, -1); + else + wait_for_completion(&rdrain.done); +} + +/** + * ib_drain_sq() - Block until all SQ CQEs have been consumed by the + * application. + * @qp: queue pair to drain + * + * If the device has a provider-specific drain function, then + * call that. Otherwise call the generic drain function + * __ib_drain_sq(). + * + * The caller must: + * + * ensure there is room in the CQ and SQ for the drain work request and + * completion. + * + * allocate the CQ using ib_alloc_cq(). + * + * ensure that there are no other contexts that are posting WRs concurrently. + * Otherwise the drain is not guaranteed. + */ +void ib_drain_sq(struct ib_qp *qp) +{ + if (qp->device->ops.drain_sq) + qp->device->ops.drain_sq(qp); + else + __ib_drain_sq(qp); + trace_cq_drain_complete(qp->send_cq); +} +EXPORT_SYMBOL(ib_drain_sq); + +/** + * ib_drain_rq() - Block until all RQ CQEs have been consumed by the + * application. + * @qp: queue pair to drain + * + * If the device has a provider-specific drain function, then + * call that. Otherwise call the generic drain function + * __ib_drain_rq(). + * + * The caller must: + * + * ensure there is room in the CQ and RQ for the drain work request and + * completion. + * + * allocate the CQ using ib_alloc_cq(). + * + * ensure that there are no other contexts that are posting WRs concurrently. + * Otherwise the drain is not guaranteed. + */ +void ib_drain_rq(struct ib_qp *qp) +{ + if (qp->device->ops.drain_rq) + qp->device->ops.drain_rq(qp); + else + __ib_drain_rq(qp); + trace_cq_drain_complete(qp->recv_cq); +} +EXPORT_SYMBOL(ib_drain_rq); + +/** + * ib_drain_qp() - Block until all CQEs have been consumed by the + * application on both the RQ and SQ. + * @qp: queue pair to drain + * + * The caller must: + * + * ensure there is room in the CQ(s), SQ, and RQ for drain work requests + * and completions. + * + * allocate the CQs using ib_alloc_cq(). + * + * ensure that there are no other contexts that are posting WRs concurrently. + * Otherwise the drain is not guaranteed. + */ +void ib_drain_qp(struct ib_qp *qp) +{ + ib_drain_sq(qp); + if (!qp->srq) + ib_drain_rq(qp); +} +EXPORT_SYMBOL(ib_drain_qp); + +struct net_device *rdma_alloc_netdev(struct ib_device *device, u8 port_num, + enum rdma_netdev_t type, const char *name, + unsigned char name_assign_type, + void (*setup)(struct net_device *)) +{ + struct rdma_netdev_alloc_params params; + struct net_device *netdev; + int rc; + + if (!device->ops.rdma_netdev_get_params) + return ERR_PTR(-EOPNOTSUPP); + + rc = device->ops.rdma_netdev_get_params(device, port_num, type, + ¶ms); + if (rc) + return ERR_PTR(rc); + + netdev = alloc_netdev_mqs(params.sizeof_priv, name, name_assign_type, + setup, params.txqs, params.rxqs); + if (!netdev) + return ERR_PTR(-ENOMEM); + + return netdev; +} +EXPORT_SYMBOL(rdma_alloc_netdev); + +int rdma_init_netdev(struct ib_device *device, u8 port_num, + enum rdma_netdev_t type, const char *name, + unsigned char name_assign_type, + void (*setup)(struct net_device *), + struct net_device *netdev) +{ + struct rdma_netdev_alloc_params params; + int rc; + + if (!device->ops.rdma_netdev_get_params) + return -EOPNOTSUPP; + + rc = device->ops.rdma_netdev_get_params(device, port_num, type, + ¶ms); + if (rc) + return rc; + + return params.initialize_rdma_netdev(device, port_num, + netdev, params.param); +} +EXPORT_SYMBOL(rdma_init_netdev); + +void __rdma_block_iter_start(struct ib_block_iter *biter, + struct scatterlist *sglist, unsigned int nents, + unsigned long pgsz) +{ + memset(biter, 0, sizeof(struct ib_block_iter)); + biter->__sg = sglist; + biter->__sg_nents = nents; + + /* Driver provides best block size to use */ + biter->__pg_bit = __fls(pgsz); +} +EXPORT_SYMBOL(__rdma_block_iter_start); + +bool __rdma_block_iter_next(struct ib_block_iter *biter) +{ + unsigned int block_offset; + unsigned int sg_delta; + + if (!biter->__sg_nents || !biter->__sg) + return false; + + biter->__dma_addr = sg_dma_address(biter->__sg) + biter->__sg_advance; + block_offset = biter->__dma_addr & (BIT_ULL(biter->__pg_bit) - 1); + sg_delta = BIT_ULL(biter->__pg_bit) - block_offset; + + if (sg_dma_len(biter->__sg) - biter->__sg_advance > sg_delta) { + biter->__sg_advance += sg_delta; + } else { + biter->__sg_advance = 0; + biter->__sg = sg_next(biter->__sg); + biter->__sg_nents--; + } + + return true; +} +EXPORT_SYMBOL(__rdma_block_iter_next); |