diff options
Diffstat (limited to 'drivers/infiniband/ulp/rtrs')
| -rw-r--r-- | drivers/infiniband/ulp/rtrs/Kconfig | 27 | ||||
| -rw-r--r-- | drivers/infiniband/ulp/rtrs/Makefile | 15 | ||||
| -rw-r--r-- | drivers/infiniband/ulp/rtrs/README | 213 | ||||
| -rw-r--r-- | drivers/infiniband/ulp/rtrs/rtrs-clt-stats.c | 200 | ||||
| -rw-r--r-- | drivers/infiniband/ulp/rtrs/rtrs-clt-sysfs.c | 485 | ||||
| -rw-r--r-- | drivers/infiniband/ulp/rtrs/rtrs-clt.c | 3021 | ||||
| -rw-r--r-- | drivers/infiniband/ulp/rtrs/rtrs-clt.h | 253 | ||||
| -rw-r--r-- | drivers/infiniband/ulp/rtrs/rtrs-log.h | 28 | ||||
| -rw-r--r-- | drivers/infiniband/ulp/rtrs/rtrs-pri.h | 405 | ||||
| -rw-r--r-- | drivers/infiniband/ulp/rtrs/rtrs-srv-stats.c | 38 | ||||
| -rw-r--r-- | drivers/infiniband/ulp/rtrs/rtrs-srv-sysfs.c | 319 | ||||
| -rw-r--r-- | drivers/infiniband/ulp/rtrs/rtrs-srv.c | 2267 | ||||
| -rw-r--r-- | drivers/infiniband/ulp/rtrs/rtrs-srv.h | 155 | ||||
| -rw-r--r-- | drivers/infiniband/ulp/rtrs/rtrs.c | 600 | ||||
| -rw-r--r-- | drivers/infiniband/ulp/rtrs/rtrs.h | 196 |
15 files changed, 8222 insertions, 0 deletions
diff --git a/drivers/infiniband/ulp/rtrs/Kconfig b/drivers/infiniband/ulp/rtrs/Kconfig new file mode 100644 index 000000000..9092b62e6 --- /dev/null +++ b/drivers/infiniband/ulp/rtrs/Kconfig @@ -0,0 +1,27 @@ +# SPDX-License-Identifier: GPL-2.0-or-later + +config INFINIBAND_RTRS + tristate + depends on INFINIBAND_ADDR_TRANS + +config INFINIBAND_RTRS_CLIENT + tristate "RTRS client module" + depends on INFINIBAND_ADDR_TRANS + select INFINIBAND_RTRS + help + RDMA transport client module. + + RDMA Transport (RTRS) client implements a reliable transport layer + and also multipathing functionality and that it is intended to be + the base layer for a block storage initiator over RDMA. + +config INFINIBAND_RTRS_SERVER + tristate "RTRS server module" + depends on INFINIBAND_ADDR_TRANS + select INFINIBAND_RTRS + help + RDMA transport server module. + + RDMA Transport (RTRS) server module processing connection and IO + requests received from the RTRS client module, it will pass the + IO requests to its user eg. RNBD_server. diff --git a/drivers/infiniband/ulp/rtrs/Makefile b/drivers/infiniband/ulp/rtrs/Makefile new file mode 100644 index 000000000..3898509be --- /dev/null +++ b/drivers/infiniband/ulp/rtrs/Makefile @@ -0,0 +1,15 @@ +# SPDX-License-Identifier: GPL-2.0-or-later + +rtrs-client-y := rtrs-clt.o \ + rtrs-clt-stats.o \ + rtrs-clt-sysfs.o + +rtrs-server-y := rtrs-srv.o \ + rtrs-srv-stats.o \ + rtrs-srv-sysfs.o + +rtrs-core-y := rtrs.o + +obj-$(CONFIG_INFINIBAND_RTRS) += rtrs-core.o +obj-$(CONFIG_INFINIBAND_RTRS_CLIENT) += rtrs-client.o +obj-$(CONFIG_INFINIBAND_RTRS_SERVER) += rtrs-server.o diff --git a/drivers/infiniband/ulp/rtrs/README b/drivers/infiniband/ulp/rtrs/README new file mode 100644 index 000000000..5d9ea142e --- /dev/null +++ b/drivers/infiniband/ulp/rtrs/README @@ -0,0 +1,213 @@ +**************************** +RDMA Transport (RTRS) +**************************** + +RTRS (RDMA Transport) is a reliable high speed transport library +which provides support to establish optimal number of connections +between client and server machines using RDMA (InfiniBand, RoCE, iWarp) +transport. It is optimized to transfer (read/write) IO blocks. + +In its core interface it follows the BIO semantics of providing the +possibility to either write data from an sg list to the remote side +or to request ("read") data transfer from the remote side into a given +sg list. + +RTRS provides I/O fail-over and load-balancing capabilities by using +multipath I/O (see "add_path" and "mp_policy" configuration entries in +Documentation/ABI/testing/sysfs-class-rtrs-client). + +RTRS is used by the RNBD (RDMA Network Block Device) modules. + +================== +Transport protocol +================== + +Overview +-------- +An established connection between a client and a server is called rtrs +session. A session is associated with a set of memory chunks reserved on the +server side for a given client for rdma transfer. A session +consists of multiple paths, each representing a separate physical link +between client and server. Those are used for load balancing and failover. +Each path consists of as many connections (QPs) as there are cpus on +the client. + +When processing an incoming write or read request, rtrs client uses memory +chunks reserved for him on the server side. Their number, size and addresses +need to be exchanged between client and server during the connection +establishment phase. Apart from the memory related information client needs to +inform the server about the session name and identify each path and connection +individually. + +On an established session client sends to server write or read messages. +Server uses immediate field to tell the client which request is being +acknowledged and for errno. Client uses immediate field to tell the server +which of the memory chunks has been accessed and at which offset the message +can be found. + +Module parameter always_invalidate is introduced for the security problem +discussed in LPC RDMA MC 2019. When always_invalidate=Y, on the server side we +invalidate each rdma buffer before we hand it over to RNBD server and +then pass it to the block layer. A new rkey is generated and registered for the +buffer after it returns back from the block layer and RNBD server. +The new rkey is sent back to the client along with the IO result. +The procedure is the default behaviour of the driver. This invalidation and +registration on each IO causes performance drop of up to 20%. A user of the +driver may choose to load the modules with this mechanism switched off +(always_invalidate=N), if he understands and can take the risk of a malicious +client being able to corrupt memory of a server it is connected to. This might +be a reasonable option in a scenario where all the clients and all the servers +are located within a secure datacenter. + + +Connection establishment +------------------------ + +1. Client starts establishing connections belonging to a path of a session one +by one via attaching RTRS_MSG_CON_REQ messages to the rdma_connect requests. +Those include uuid of the session and uuid of the path to be +established. They are used by the server to find a persisting session/path or +to create a new one when necessary. The message also contains the protocol +version and magic for compatibility, total number of connections per session +(as many as cpus on the client), the id of the current connection and +the reconnect counter, which is used to resolve the situations where +client is trying to reconnect a path, while server is still destroying the old +one. + +2. Server accepts the connection requests one by one and attaches +RTRS_MSG_CONN_RSP messages to the rdma_accept. Apart from magic and +protocol version, the messages include error code, queue depth supported by +the server (number of memory chunks which are going to be allocated for that +session) and the maximum size of one io, RTRS_MSG_NEW_RKEY_F flags is set +when always_invalidate=Y. + +3. After all connections of a path are established client sends to server the +RTRS_MSG_INFO_REQ message, containing the name of the session. This message +requests the address information from the server. + +4. Server replies to the session info request message with RTRS_MSG_INFO_RSP, +which contains the addresses and keys of the RDMA buffers allocated for that +session. + +5. Session becomes connected after all paths to be established are connected +(i.e. steps 1-4 finished for all paths requested for a session) + +6. Server and client exchange periodically heartbeat messages (empty rdma +messages with an immediate field) which are used to detect a crash on remote +side or network outage in an absence of IO. + +7. On any RDMA related error or in the case of a heartbeat timeout, the +corresponding path is disconnected, all the inflight IO are failed over to a +healthy path, if any, and the reconnect mechanism is triggered. + +CLT SRV +*for each connection belonging to a path and for each path: +RTRS_MSG_CON_REQ -------------------> + <------------------- RTRS_MSG_CON_RSP +... +*after all connections are established: +RTRS_MSG_INFO_REQ -------------------> + <------------------- RTRS_MSG_INFO_RSP +*heartbeat is started from both sides: + -------------------> [RTRS_HB_MSG_IMM] +[RTRS_HB_MSG_ACK] <------------------- +[RTRS_HB_MSG_IMM] <------------------- + -------------------> [RTRS_HB_MSG_ACK] + +IO path +------- + +* Write (always_invalidate=N) * + +1. When processing a write request client selects one of the memory chunks +on the server side and rdma writes there the user data, user header and the +RTRS_MSG_RDMA_WRITE message. Apart from the type (write), the message only +contains size of the user header. The client tells the server which chunk has +been accessed and at what offset the RTRS_MSG_RDMA_WRITE can be found by +using the IMM field. + +2. When confirming a write request server sends an "empty" rdma message with +an immediate field. The 32 bit field is used to specify the outstanding +inflight IO and for the error code. + +CLT SRV +usr_data + usr_hdr + rtrs_msg_rdma_write -----------------> [RTRS_IO_REQ_IMM] +[RTRS_IO_RSP_IMM] <----------------- (id + errno) + +* Write (always_invalidate=Y) * + +1. When processing a write request client selects one of the memory chunks +on the server side and rdma writes there the user data, user header and the +RTRS_MSG_RDMA_WRITE message. Apart from the type (write), the message only +contains size of the user header. The client tells the server which chunk has +been accessed and at what offset the RTRS_MSG_RDMA_WRITE can be found by +using the IMM field, Server invalidate rkey associated to the memory chunks +first, when it finishes, pass the IO to RNBD server module. + +2. When confirming a write request server sends an "empty" rdma message with +an immediate field. The 32 bit field is used to specify the outstanding +inflight IO and for the error code. The new rkey is sent back using +SEND_WITH_IMM WR, client When it recived new rkey message, it validates +the message and finished IO after update rkey for the rbuffer, then post +back the recv buffer for later use. + +CLT SRV +usr_data + usr_hdr + rtrs_msg_rdma_write -----------------> [RTRS_IO_REQ_IMM] +[RTRS_MSG_RKEY_RSP] <----------------- (RTRS_MSG_RKEY_RSP) +[RTRS_IO_RSP_IMM] <----------------- (id + errno) + + +* Read (always_invalidate=N)* + +1. When processing a read request client selects one of the memory chunks +on the server side and rdma writes there the user header and the +RTRS_MSG_RDMA_READ message. This message contains the type (read), size of +the user header, flags (specifying if memory invalidation is necessary) and the +list of addresses along with keys for the data to be read into. + +2. When confirming a read request server transfers the requested data first, +attaches an invalidation message if requested and finally an "empty" rdma +message with an immediate field. The 32 bit field is used to specify the +outstanding inflight IO and the error code. + +CLT SRV +usr_hdr + rtrs_msg_rdma_read --------------> [RTRS_IO_REQ_IMM] +[RTRS_IO_RSP_IMM] <-------------- usr_data + (id + errno) +or in case client requested invalidation: +[RTRS_IO_RSP_IMM_W_INV] <-------------- usr_data + (INV) + (id + errno) + +* Read (always_invalidate=Y)* + +1. When processing a read request client selects one of the memory chunks +on the server side and rdma writes there the user header and the +RTRS_MSG_RDMA_READ message. This message contains the type (read), size of +the user header, flags (specifying if memory invalidation is necessary) and the +list of addresses along with keys for the data to be read into. +Server invalidate rkey associated to the memory chunks first, when it finishes, +passes the IO to RNBD server module. + +2. When confirming a read request server transfers the requested data first, +attaches an invalidation message if requested and finally an "empty" rdma +message with an immediate field. The 32 bit field is used to specify the +outstanding inflight IO and the error code. The new rkey is sent back using +SEND_WITH_IMM WR, client When it recived new rkey message, it validates +the message and finished IO after update rkey for the rbuffer, then post +back the recv buffer for later use. + +CLT SRV +usr_hdr + rtrs_msg_rdma_read --------------> [RTRS_IO_REQ_IMM] +[RTRS_IO_RSP_IMM] <-------------- usr_data + (id + errno) +[RTRS_MSG_RKEY_RSP] <----------------- (RTRS_MSG_RKEY_RSP) +or in case client requested invalidation: +[RTRS_IO_RSP_IMM_W_INV] <-------------- usr_data + (INV) + (id + errno) +========================================= +Contributors List(in alphabetical order) +========================================= +Danil Kipnis <danil.kipnis@profitbricks.com> +Fabian Holler <mail@fholler.de> +Guoqing Jiang <guoqing.jiang@cloud.ionos.com> +Jack Wang <jinpu.wang@profitbricks.com> +Kleber Souza <kleber.souza@profitbricks.com> +Lutz Pogrell <lutz.pogrell@cloud.ionos.com> +Milind Dumbare <Milind.dumbare@gmail.com> +Roman Penyaev <roman.penyaev@profitbricks.com> diff --git a/drivers/infiniband/ulp/rtrs/rtrs-clt-stats.c b/drivers/infiniband/ulp/rtrs/rtrs-clt-stats.c new file mode 100644 index 000000000..26bbe5d6d --- /dev/null +++ b/drivers/infiniband/ulp/rtrs/rtrs-clt-stats.c @@ -0,0 +1,200 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * RDMA Transport Layer + * + * Copyright (c) 2014 - 2018 ProfitBricks GmbH. All rights reserved. + * Copyright (c) 2018 - 2019 1&1 IONOS Cloud GmbH. All rights reserved. + * Copyright (c) 2019 - 2020 1&1 IONOS SE. All rights reserved. + */ +#undef pr_fmt +#define pr_fmt(fmt) KBUILD_MODNAME " L" __stringify(__LINE__) ": " fmt + +#include "rtrs-clt.h" + +void rtrs_clt_update_wc_stats(struct rtrs_clt_con *con) +{ + struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess); + struct rtrs_clt_stats *stats = sess->stats; + struct rtrs_clt_stats_pcpu *s; + int cpu; + + cpu = raw_smp_processor_id(); + s = this_cpu_ptr(stats->pcpu_stats); + if (unlikely(con->cpu != cpu)) { + s->cpu_migr.to++; + + /* Careful here, override s pointer */ + s = per_cpu_ptr(stats->pcpu_stats, con->cpu); + atomic_inc(&s->cpu_migr.from); + } +} + +void rtrs_clt_inc_failover_cnt(struct rtrs_clt_stats *stats) +{ + struct rtrs_clt_stats_pcpu *s; + + s = this_cpu_ptr(stats->pcpu_stats); + s->rdma.failover_cnt++; +} + +int rtrs_clt_stats_migration_cnt_to_str(struct rtrs_clt_stats *stats, + char *buf, size_t len) +{ + struct rtrs_clt_stats_pcpu *s; + + size_t used; + int cpu; + + used = scnprintf(buf, len, " "); + for_each_possible_cpu(cpu) + used += scnprintf(buf + used, len - used, " CPU%u", cpu); + + used += scnprintf(buf + used, len - used, "\nfrom:"); + for_each_possible_cpu(cpu) { + s = per_cpu_ptr(stats->pcpu_stats, cpu); + used += scnprintf(buf + used, len - used, " %d", + atomic_read(&s->cpu_migr.from)); + } + + used += scnprintf(buf + used, len - used, "\nto :"); + for_each_possible_cpu(cpu) { + s = per_cpu_ptr(stats->pcpu_stats, cpu); + used += scnprintf(buf + used, len - used, " %d", + s->cpu_migr.to); + } + used += scnprintf(buf + used, len - used, "\n"); + + return used; +} + +int rtrs_clt_stats_reconnects_to_str(struct rtrs_clt_stats *stats, char *buf, + size_t len) +{ + return scnprintf(buf, len, "%d %d\n", + stats->reconnects.successful_cnt, + stats->reconnects.fail_cnt); +} + +ssize_t rtrs_clt_stats_rdma_to_str(struct rtrs_clt_stats *stats, + char *page, size_t len) +{ + struct rtrs_clt_stats_rdma sum; + struct rtrs_clt_stats_rdma *r; + int cpu; + + memset(&sum, 0, sizeof(sum)); + + for_each_possible_cpu(cpu) { + r = &per_cpu_ptr(stats->pcpu_stats, cpu)->rdma; + + sum.dir[READ].cnt += r->dir[READ].cnt; + sum.dir[READ].size_total += r->dir[READ].size_total; + sum.dir[WRITE].cnt += r->dir[WRITE].cnt; + sum.dir[WRITE].size_total += r->dir[WRITE].size_total; + sum.failover_cnt += r->failover_cnt; + } + + return scnprintf(page, len, "%llu %llu %llu %llu %u %llu\n", + sum.dir[READ].cnt, sum.dir[READ].size_total, + sum.dir[WRITE].cnt, sum.dir[WRITE].size_total, + atomic_read(&stats->inflight), sum.failover_cnt); +} + +ssize_t rtrs_clt_reset_all_help(struct rtrs_clt_stats *s, + char *page, size_t len) +{ + return scnprintf(page, len, "echo 1 to reset all statistics\n"); +} + +int rtrs_clt_reset_rdma_stats(struct rtrs_clt_stats *stats, bool enable) +{ + struct rtrs_clt_stats_pcpu *s; + int cpu; + + if (!enable) + return -EINVAL; + + for_each_possible_cpu(cpu) { + s = per_cpu_ptr(stats->pcpu_stats, cpu); + memset(&s->rdma, 0, sizeof(s->rdma)); + } + + return 0; +} + +int rtrs_clt_reset_cpu_migr_stats(struct rtrs_clt_stats *stats, bool enable) +{ + struct rtrs_clt_stats_pcpu *s; + int cpu; + + if (!enable) + return -EINVAL; + + for_each_possible_cpu(cpu) { + s = per_cpu_ptr(stats->pcpu_stats, cpu); + memset(&s->cpu_migr, 0, sizeof(s->cpu_migr)); + } + + return 0; +} + +int rtrs_clt_reset_reconnects_stat(struct rtrs_clt_stats *stats, bool enable) +{ + if (!enable) + return -EINVAL; + + memset(&stats->reconnects, 0, sizeof(stats->reconnects)); + + return 0; +} + +int rtrs_clt_reset_all_stats(struct rtrs_clt_stats *s, bool enable) +{ + if (enable) { + rtrs_clt_reset_rdma_stats(s, enable); + rtrs_clt_reset_cpu_migr_stats(s, enable); + rtrs_clt_reset_reconnects_stat(s, enable); + atomic_set(&s->inflight, 0); + return 0; + } + + return -EINVAL; +} + +static inline void rtrs_clt_update_rdma_stats(struct rtrs_clt_stats *stats, + size_t size, int d) +{ + struct rtrs_clt_stats_pcpu *s; + + s = this_cpu_ptr(stats->pcpu_stats); + s->rdma.dir[d].cnt++; + s->rdma.dir[d].size_total += size; +} + +void rtrs_clt_update_all_stats(struct rtrs_clt_io_req *req, int dir) +{ + struct rtrs_clt_con *con = req->con; + struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess); + struct rtrs_clt_stats *stats = sess->stats; + unsigned int len; + + len = req->usr_len + req->data_len; + rtrs_clt_update_rdma_stats(stats, len, dir); + if (sess->clt->mp_policy == MP_POLICY_MIN_INFLIGHT) + atomic_inc(&stats->inflight); +} + +int rtrs_clt_init_stats(struct rtrs_clt_stats *stats) +{ + stats->pcpu_stats = alloc_percpu(typeof(*stats->pcpu_stats)); + if (!stats->pcpu_stats) + return -ENOMEM; + + /* + * successful_cnt will be set to 0 after session + * is established for the first time + */ + stats->reconnects.successful_cnt = -1; + + return 0; +} diff --git a/drivers/infiniband/ulp/rtrs/rtrs-clt-sysfs.c b/drivers/infiniband/ulp/rtrs/rtrs-clt-sysfs.c new file mode 100644 index 000000000..2ee3806f2 --- /dev/null +++ b/drivers/infiniband/ulp/rtrs/rtrs-clt-sysfs.c @@ -0,0 +1,485 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * RDMA Transport Layer + * + * Copyright (c) 2014 - 2018 ProfitBricks GmbH. All rights reserved. + * Copyright (c) 2018 - 2019 1&1 IONOS Cloud GmbH. All rights reserved. + * Copyright (c) 2019 - 2020 1&1 IONOS SE. All rights reserved. + */ +#undef pr_fmt +#define pr_fmt(fmt) KBUILD_MODNAME " L" __stringify(__LINE__) ": " fmt + +#include "rtrs-pri.h" +#include "rtrs-clt.h" +#include "rtrs-log.h" + +#define MIN_MAX_RECONN_ATT -1 +#define MAX_MAX_RECONN_ATT 9999 + +static void rtrs_clt_sess_release(struct kobject *kobj) +{ + struct rtrs_clt_sess *sess; + + sess = container_of(kobj, struct rtrs_clt_sess, kobj); + + free_sess(sess); +} + +static struct kobj_type ktype_sess = { + .sysfs_ops = &kobj_sysfs_ops, + .release = rtrs_clt_sess_release +}; + +static void rtrs_clt_sess_stats_release(struct kobject *kobj) +{ + struct rtrs_clt_stats *stats; + + stats = container_of(kobj, struct rtrs_clt_stats, kobj_stats); + + free_percpu(stats->pcpu_stats); + + kfree(stats); +} + +static struct kobj_type ktype_stats = { + .sysfs_ops = &kobj_sysfs_ops, + .release = rtrs_clt_sess_stats_release, +}; + +static ssize_t max_reconnect_attempts_show(struct device *dev, + struct device_attribute *attr, + char *page) +{ + struct rtrs_clt *clt = container_of(dev, struct rtrs_clt, dev); + + return sprintf(page, "%d\n", rtrs_clt_get_max_reconnect_attempts(clt)); +} + +static ssize_t max_reconnect_attempts_store(struct device *dev, + struct device_attribute *attr, + const char *buf, + size_t count) +{ + int value; + int ret; + struct rtrs_clt *clt = container_of(dev, struct rtrs_clt, dev); + + ret = kstrtoint(buf, 10, &value); + if (ret) { + rtrs_err(clt, "%s: failed to convert string '%s' to int\n", + attr->attr.name, buf); + return ret; + } + if (value > MAX_MAX_RECONN_ATT || + value < MIN_MAX_RECONN_ATT) { + rtrs_err(clt, + "%s: invalid range (provided: '%s', accepted: min: %d, max: %d)\n", + attr->attr.name, buf, MIN_MAX_RECONN_ATT, + MAX_MAX_RECONN_ATT); + return -EINVAL; + } + rtrs_clt_set_max_reconnect_attempts(clt, value); + + return count; +} + +static DEVICE_ATTR_RW(max_reconnect_attempts); + +static ssize_t mpath_policy_show(struct device *dev, + struct device_attribute *attr, + char *page) +{ + struct rtrs_clt *clt; + + clt = container_of(dev, struct rtrs_clt, dev); + + switch (clt->mp_policy) { + case MP_POLICY_RR: + return sprintf(page, "round-robin (RR: %d)\n", clt->mp_policy); + case MP_POLICY_MIN_INFLIGHT: + return sprintf(page, "min-inflight (MI: %d)\n", clt->mp_policy); + default: + return sprintf(page, "Unknown (%d)\n", clt->mp_policy); + } +} + +static ssize_t mpath_policy_store(struct device *dev, + struct device_attribute *attr, + const char *buf, + size_t count) +{ + struct rtrs_clt *clt; + int value; + int ret; + + clt = container_of(dev, struct rtrs_clt, dev); + + ret = kstrtoint(buf, 10, &value); + if (!ret && (value == MP_POLICY_RR || + value == MP_POLICY_MIN_INFLIGHT)) { + clt->mp_policy = value; + return count; + } + + if (!strncasecmp(buf, "round-robin", 11) || + !strncasecmp(buf, "rr", 2)) + clt->mp_policy = MP_POLICY_RR; + else if (!strncasecmp(buf, "min-inflight", 12) || + !strncasecmp(buf, "mi", 2)) + clt->mp_policy = MP_POLICY_MIN_INFLIGHT; + else + return -EINVAL; + + return count; +} + +static DEVICE_ATTR_RW(mpath_policy); + +static ssize_t add_path_show(struct device *dev, + struct device_attribute *attr, char *page) +{ + return scnprintf(page, PAGE_SIZE, + "Usage: echo [<source addr>@]<destination addr> > %s\n\n*addr ::= [ ip:<ipv4|ipv6> | gid:<gid> ]\n", + attr->attr.name); +} + +static ssize_t add_path_store(struct device *dev, + struct device_attribute *attr, + const char *buf, size_t count) +{ + struct sockaddr_storage srcaddr, dstaddr; + struct rtrs_addr addr = { + .src = &srcaddr, + .dst = &dstaddr + }; + struct rtrs_clt *clt; + const char *nl; + size_t len; + int err; + + clt = container_of(dev, struct rtrs_clt, dev); + + nl = strchr(buf, '\n'); + if (nl) + len = nl - buf; + else + len = count; + err = rtrs_addr_to_sockaddr(buf, len, clt->port, &addr); + if (err) + return -EINVAL; + + err = rtrs_clt_create_path_from_sysfs(clt, &addr); + if (err) + return err; + + return count; +} + +static DEVICE_ATTR_RW(add_path); + +static ssize_t rtrs_clt_state_show(struct kobject *kobj, + struct kobj_attribute *attr, char *page) +{ + struct rtrs_clt_sess *sess; + + sess = container_of(kobj, struct rtrs_clt_sess, kobj); + if (sess->state == RTRS_CLT_CONNECTED) + return sprintf(page, "connected\n"); + + return sprintf(page, "disconnected\n"); +} + +static struct kobj_attribute rtrs_clt_state_attr = + __ATTR(state, 0444, rtrs_clt_state_show, NULL); + +static ssize_t rtrs_clt_reconnect_show(struct kobject *kobj, + struct kobj_attribute *attr, + char *page) +{ + return scnprintf(page, PAGE_SIZE, "Usage: echo 1 > %s\n", + attr->attr.name); +} + +static ssize_t rtrs_clt_reconnect_store(struct kobject *kobj, + struct kobj_attribute *attr, + const char *buf, size_t count) +{ + struct rtrs_clt_sess *sess; + int ret; + + sess = container_of(kobj, struct rtrs_clt_sess, kobj); + if (!sysfs_streq(buf, "1")) { + rtrs_err(sess->clt, "%s: unknown value: '%s'\n", + attr->attr.name, buf); + return -EINVAL; + } + ret = rtrs_clt_reconnect_from_sysfs(sess); + if (ret) + return ret; + + return count; +} + +static struct kobj_attribute rtrs_clt_reconnect_attr = + __ATTR(reconnect, 0644, rtrs_clt_reconnect_show, + rtrs_clt_reconnect_store); + +static ssize_t rtrs_clt_disconnect_show(struct kobject *kobj, + struct kobj_attribute *attr, + char *page) +{ + return scnprintf(page, PAGE_SIZE, "Usage: echo 1 > %s\n", + attr->attr.name); +} + +static ssize_t rtrs_clt_disconnect_store(struct kobject *kobj, + struct kobj_attribute *attr, + const char *buf, size_t count) +{ + struct rtrs_clt_sess *sess; + int ret; + + sess = container_of(kobj, struct rtrs_clt_sess, kobj); + if (!sysfs_streq(buf, "1")) { + rtrs_err(sess->clt, "%s: unknown value: '%s'\n", + attr->attr.name, buf); + return -EINVAL; + } + ret = rtrs_clt_disconnect_from_sysfs(sess); + if (ret) + return ret; + + return count; +} + +static struct kobj_attribute rtrs_clt_disconnect_attr = + __ATTR(disconnect, 0644, rtrs_clt_disconnect_show, + rtrs_clt_disconnect_store); + +static ssize_t rtrs_clt_remove_path_show(struct kobject *kobj, + struct kobj_attribute *attr, + char *page) +{ + return scnprintf(page, PAGE_SIZE, "Usage: echo 1 > %s\n", + attr->attr.name); +} + +static ssize_t rtrs_clt_remove_path_store(struct kobject *kobj, + struct kobj_attribute *attr, + const char *buf, size_t count) +{ + struct rtrs_clt_sess *sess; + int ret; + + sess = container_of(kobj, struct rtrs_clt_sess, kobj); + if (!sysfs_streq(buf, "1")) { + rtrs_err(sess->clt, "%s: unknown value: '%s'\n", + attr->attr.name, buf); + return -EINVAL; + } + ret = rtrs_clt_remove_path_from_sysfs(sess, &attr->attr); + if (ret) + return ret; + + return count; +} + +static struct kobj_attribute rtrs_clt_remove_path_attr = + __ATTR(remove_path, 0644, rtrs_clt_remove_path_show, + rtrs_clt_remove_path_store); + +STAT_ATTR(struct rtrs_clt_stats, cpu_migration, + rtrs_clt_stats_migration_cnt_to_str, + rtrs_clt_reset_cpu_migr_stats); + +STAT_ATTR(struct rtrs_clt_stats, reconnects, + rtrs_clt_stats_reconnects_to_str, + rtrs_clt_reset_reconnects_stat); + +STAT_ATTR(struct rtrs_clt_stats, rdma, + rtrs_clt_stats_rdma_to_str, + rtrs_clt_reset_rdma_stats); + +STAT_ATTR(struct rtrs_clt_stats, reset_all, + rtrs_clt_reset_all_help, + rtrs_clt_reset_all_stats); + +static struct attribute *rtrs_clt_stats_attrs[] = { + &cpu_migration_attr.attr, + &reconnects_attr.attr, + &rdma_attr.attr, + &reset_all_attr.attr, + NULL +}; + +static const struct attribute_group rtrs_clt_stats_attr_group = { + .attrs = rtrs_clt_stats_attrs, +}; + +static ssize_t rtrs_clt_hca_port_show(struct kobject *kobj, + struct kobj_attribute *attr, + char *page) +{ + struct rtrs_clt_sess *sess; + + sess = container_of(kobj, typeof(*sess), kobj); + + return scnprintf(page, PAGE_SIZE, "%u\n", sess->hca_port); +} + +static struct kobj_attribute rtrs_clt_hca_port_attr = + __ATTR(hca_port, 0444, rtrs_clt_hca_port_show, NULL); + +static ssize_t rtrs_clt_hca_name_show(struct kobject *kobj, + struct kobj_attribute *attr, + char *page) +{ + struct rtrs_clt_sess *sess; + + sess = container_of(kobj, struct rtrs_clt_sess, kobj); + + return scnprintf(page, PAGE_SIZE, "%s\n", sess->hca_name); +} + +static struct kobj_attribute rtrs_clt_hca_name_attr = + __ATTR(hca_name, 0444, rtrs_clt_hca_name_show, NULL); + +static ssize_t rtrs_clt_src_addr_show(struct kobject *kobj, + struct kobj_attribute *attr, + char *page) +{ + struct rtrs_clt_sess *sess; + int cnt; + + sess = container_of(kobj, struct rtrs_clt_sess, kobj); + cnt = sockaddr_to_str((struct sockaddr *)&sess->s.src_addr, + page, PAGE_SIZE); + return cnt + scnprintf(page + cnt, PAGE_SIZE - cnt, "\n"); +} + +static struct kobj_attribute rtrs_clt_src_addr_attr = + __ATTR(src_addr, 0444, rtrs_clt_src_addr_show, NULL); + +static ssize_t rtrs_clt_dst_addr_show(struct kobject *kobj, + struct kobj_attribute *attr, + char *page) +{ + struct rtrs_clt_sess *sess; + int cnt; + + sess = container_of(kobj, struct rtrs_clt_sess, kobj); + cnt = sockaddr_to_str((struct sockaddr *)&sess->s.dst_addr, + page, PAGE_SIZE); + return cnt + scnprintf(page + cnt, PAGE_SIZE - cnt, "\n"); +} + +static struct kobj_attribute rtrs_clt_dst_addr_attr = + __ATTR(dst_addr, 0444, rtrs_clt_dst_addr_show, NULL); + +static struct attribute *rtrs_clt_sess_attrs[] = { + &rtrs_clt_hca_name_attr.attr, + &rtrs_clt_hca_port_attr.attr, + &rtrs_clt_src_addr_attr.attr, + &rtrs_clt_dst_addr_attr.attr, + &rtrs_clt_state_attr.attr, + &rtrs_clt_reconnect_attr.attr, + &rtrs_clt_disconnect_attr.attr, + &rtrs_clt_remove_path_attr.attr, + NULL, +}; + +static const struct attribute_group rtrs_clt_sess_attr_group = { + .attrs = rtrs_clt_sess_attrs, +}; + +int rtrs_clt_create_sess_files(struct rtrs_clt_sess *sess) +{ + struct rtrs_clt *clt = sess->clt; + char str[NAME_MAX]; + int err, cnt; + + cnt = sockaddr_to_str((struct sockaddr *)&sess->s.src_addr, + str, sizeof(str)); + cnt += scnprintf(str + cnt, sizeof(str) - cnt, "@"); + sockaddr_to_str((struct sockaddr *)&sess->s.dst_addr, + str + cnt, sizeof(str) - cnt); + + err = kobject_init_and_add(&sess->kobj, &ktype_sess, clt->kobj_paths, + "%s", str); + if (err) { + pr_err("kobject_init_and_add: %d\n", err); + kobject_put(&sess->kobj); + return err; + } + err = sysfs_create_group(&sess->kobj, &rtrs_clt_sess_attr_group); + if (err) { + pr_err("sysfs_create_group(): %d\n", err); + goto put_kobj; + } + err = kobject_init_and_add(&sess->stats->kobj_stats, &ktype_stats, + &sess->kobj, "stats"); + if (err) { + pr_err("kobject_init_and_add: %d\n", err); + kobject_put(&sess->stats->kobj_stats); + goto remove_group; + } + + err = sysfs_create_group(&sess->stats->kobj_stats, + &rtrs_clt_stats_attr_group); + if (err) { + pr_err("failed to create stats sysfs group, err: %d\n", err); + goto put_kobj_stats; + } + + return 0; + +put_kobj_stats: + kobject_del(&sess->stats->kobj_stats); + kobject_put(&sess->stats->kobj_stats); +remove_group: + sysfs_remove_group(&sess->kobj, &rtrs_clt_sess_attr_group); +put_kobj: + kobject_del(&sess->kobj); + kobject_put(&sess->kobj); + + return err; +} + +void rtrs_clt_destroy_sess_files(struct rtrs_clt_sess *sess, + const struct attribute *sysfs_self) +{ + kobject_del(&sess->stats->kobj_stats); + kobject_put(&sess->stats->kobj_stats); + if (sysfs_self) + sysfs_remove_file_self(&sess->kobj, sysfs_self); + kobject_del(&sess->kobj); +} + +static struct attribute *rtrs_clt_attrs[] = { + &dev_attr_max_reconnect_attempts.attr, + &dev_attr_mpath_policy.attr, + &dev_attr_add_path.attr, + NULL, +}; + +static const struct attribute_group rtrs_clt_attr_group = { + .attrs = rtrs_clt_attrs, +}; + +int rtrs_clt_create_sysfs_root_files(struct rtrs_clt *clt) +{ + return sysfs_create_group(&clt->dev.kobj, &rtrs_clt_attr_group); +} + +void rtrs_clt_destroy_sysfs_root_folders(struct rtrs_clt *clt) +{ + if (clt->kobj_paths) { + kobject_del(clt->kobj_paths); + kobject_put(clt->kobj_paths); + } +} + +void rtrs_clt_destroy_sysfs_root_files(struct rtrs_clt *clt) +{ + sysfs_remove_group(&clt->dev.kobj, &rtrs_clt_attr_group); +} diff --git a/drivers/infiniband/ulp/rtrs/rtrs-clt.c b/drivers/infiniband/ulp/rtrs/rtrs-clt.c new file mode 100644 index 000000000..e8b2b58cc --- /dev/null +++ b/drivers/infiniband/ulp/rtrs/rtrs-clt.c @@ -0,0 +1,3021 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * RDMA Transport Layer + * + * Copyright (c) 2014 - 2018 ProfitBricks GmbH. All rights reserved. + * Copyright (c) 2018 - 2019 1&1 IONOS Cloud GmbH. All rights reserved. + * Copyright (c) 2019 - 2020 1&1 IONOS SE. All rights reserved. + */ + +#undef pr_fmt +#define pr_fmt(fmt) KBUILD_MODNAME " L" __stringify(__LINE__) ": " fmt + +#include <linux/module.h> +#include <linux/rculist.h> +#include <linux/random.h> + +#include "rtrs-clt.h" +#include "rtrs-log.h" + +#define RTRS_CONNECT_TIMEOUT_MS 30000 +/* + * Wait a bit before trying to reconnect after a failure + * in order to give server time to finish clean up which + * leads to "false positives" failed reconnect attempts + */ +#define RTRS_RECONNECT_BACKOFF 1000 +/* + * Wait for additional random time between 0 and 8 seconds + * before starting to reconnect to avoid clients reconnecting + * all at once in case of a major network outage + */ +#define RTRS_RECONNECT_SEED 8 + +#define FIRST_CONN 0x01 + +MODULE_DESCRIPTION("RDMA Transport Client"); +MODULE_LICENSE("GPL"); + +static const struct rtrs_rdma_dev_pd_ops dev_pd_ops; +static struct rtrs_rdma_dev_pd dev_pd = { + .ops = &dev_pd_ops +}; + +static struct workqueue_struct *rtrs_wq; +static struct class *rtrs_clt_dev_class; + +static inline bool rtrs_clt_is_connected(const struct rtrs_clt *clt) +{ + struct rtrs_clt_sess *sess; + bool connected = false; + + rcu_read_lock(); + list_for_each_entry_rcu(sess, &clt->paths_list, s.entry) + connected |= READ_ONCE(sess->state) == RTRS_CLT_CONNECTED; + rcu_read_unlock(); + + return connected; +} + +static struct rtrs_permit * +__rtrs_get_permit(struct rtrs_clt *clt, enum rtrs_clt_con_type con_type) +{ + size_t max_depth = clt->queue_depth; + struct rtrs_permit *permit; + int bit; + + /* + * Adapted from null_blk get_tag(). Callers from different cpus may + * grab the same bit, since find_first_zero_bit is not atomic. + * But then the test_and_set_bit_lock will fail for all the + * callers but one, so that they will loop again. + * This way an explicit spinlock is not required. + */ + do { + bit = find_first_zero_bit(clt->permits_map, max_depth); + if (unlikely(bit >= max_depth)) + return NULL; + } while (unlikely(test_and_set_bit_lock(bit, clt->permits_map))); + + permit = get_permit(clt, bit); + WARN_ON(permit->mem_id != bit); + permit->cpu_id = raw_smp_processor_id(); + permit->con_type = con_type; + + return permit; +} + +static inline void __rtrs_put_permit(struct rtrs_clt *clt, + struct rtrs_permit *permit) +{ + clear_bit_unlock(permit->mem_id, clt->permits_map); +} + +/** + * rtrs_clt_get_permit() - allocates permit for future RDMA operation + * @clt: Current session + * @con_type: Type of connection to use with the permit + * @can_wait: Wait type + * + * Description: + * Allocates permit for the following RDMA operation. Permit is used + * to preallocate all resources and to propagate memory pressure + * up earlier. + * + * Context: + * Can sleep if @wait == RTRS_TAG_WAIT + */ +struct rtrs_permit *rtrs_clt_get_permit(struct rtrs_clt *clt, + enum rtrs_clt_con_type con_type, + int can_wait) +{ + struct rtrs_permit *permit; + DEFINE_WAIT(wait); + + permit = __rtrs_get_permit(clt, con_type); + if (likely(permit) || !can_wait) + return permit; + + do { + prepare_to_wait(&clt->permits_wait, &wait, + TASK_UNINTERRUPTIBLE); + permit = __rtrs_get_permit(clt, con_type); + if (likely(permit)) + break; + + io_schedule(); + } while (1); + + finish_wait(&clt->permits_wait, &wait); + + return permit; +} +EXPORT_SYMBOL(rtrs_clt_get_permit); + +/** + * rtrs_clt_put_permit() - puts allocated permit + * @clt: Current session + * @permit: Permit to be freed + * + * Context: + * Does not matter + */ +void rtrs_clt_put_permit(struct rtrs_clt *clt, struct rtrs_permit *permit) +{ + if (WARN_ON(!test_bit(permit->mem_id, clt->permits_map))) + return; + + __rtrs_put_permit(clt, permit); + + /* + * rtrs_clt_get_permit() adds itself to the &clt->permits_wait list + * before calling schedule(). So if rtrs_clt_get_permit() is sleeping + * it must have added itself to &clt->permits_wait before + * __rtrs_put_permit() finished. + * Hence it is safe to guard wake_up() with a waitqueue_active() test. + */ + if (waitqueue_active(&clt->permits_wait)) + wake_up(&clt->permits_wait); +} +EXPORT_SYMBOL(rtrs_clt_put_permit); + +void *rtrs_permit_to_pdu(struct rtrs_permit *permit) +{ + return permit + 1; +} +EXPORT_SYMBOL(rtrs_permit_to_pdu); + +/** + * rtrs_permit_to_clt_con() - returns RDMA connection pointer by the permit + * @sess: client session pointer + * @permit: permit for the allocation of the RDMA buffer + * Note: + * IO connection starts from 1. + * 0 connection is for user messages. + */ +static +struct rtrs_clt_con *rtrs_permit_to_clt_con(struct rtrs_clt_sess *sess, + struct rtrs_permit *permit) +{ + int id = 0; + + if (likely(permit->con_type == RTRS_IO_CON)) + id = (permit->cpu_id % (sess->s.con_num - 1)) + 1; + + return to_clt_con(sess->s.con[id]); +} + +/** + * __rtrs_clt_change_state() - change the session state through session state + * machine. + * + * @sess: client session to change the state of. + * @new_state: state to change to. + * + * returns true if successful, false if the requested state can not be set. + * + * Locks: + * state_wq lock must be hold. + */ +static bool __rtrs_clt_change_state(struct rtrs_clt_sess *sess, + enum rtrs_clt_state new_state) +{ + enum rtrs_clt_state old_state; + bool changed = false; + + lockdep_assert_held(&sess->state_wq.lock); + + old_state = sess->state; + switch (new_state) { + case RTRS_CLT_CONNECTING: + switch (old_state) { + case RTRS_CLT_RECONNECTING: + changed = true; + fallthrough; + default: + break; + } + break; + case RTRS_CLT_RECONNECTING: + switch (old_state) { + case RTRS_CLT_CONNECTED: + case RTRS_CLT_CONNECTING_ERR: + case RTRS_CLT_CLOSED: + changed = true; + fallthrough; + default: + break; + } + break; + case RTRS_CLT_CONNECTED: + switch (old_state) { + case RTRS_CLT_CONNECTING: + changed = true; + fallthrough; + default: + break; + } + break; + case RTRS_CLT_CONNECTING_ERR: + switch (old_state) { + case RTRS_CLT_CONNECTING: + changed = true; + fallthrough; + default: + break; + } + break; + case RTRS_CLT_CLOSING: + switch (old_state) { + case RTRS_CLT_CONNECTING: + case RTRS_CLT_CONNECTING_ERR: + case RTRS_CLT_RECONNECTING: + case RTRS_CLT_CONNECTED: + changed = true; + fallthrough; + default: + break; + } + break; + case RTRS_CLT_CLOSED: + switch (old_state) { + case RTRS_CLT_CLOSING: + changed = true; + fallthrough; + default: + break; + } + break; + case RTRS_CLT_DEAD: + switch (old_state) { + case RTRS_CLT_CLOSED: + changed = true; + fallthrough; + default: + break; + } + break; + default: + break; + } + if (changed) { + sess->state = new_state; + wake_up_locked(&sess->state_wq); + } + + return changed; +} + +static bool rtrs_clt_change_state_from_to(struct rtrs_clt_sess *sess, + enum rtrs_clt_state old_state, + enum rtrs_clt_state new_state) +{ + bool changed = false; + + spin_lock_irq(&sess->state_wq.lock); + if (sess->state == old_state) + changed = __rtrs_clt_change_state(sess, new_state); + spin_unlock_irq(&sess->state_wq.lock); + + return changed; +} + +static void rtrs_rdma_error_recovery(struct rtrs_clt_con *con) +{ + struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess); + + if (rtrs_clt_change_state_from_to(sess, + RTRS_CLT_CONNECTED, + RTRS_CLT_RECONNECTING)) { + struct rtrs_clt *clt = sess->clt; + unsigned int delay_ms; + + /* + * Normal scenario, reconnect if we were successfully connected + */ + delay_ms = clt->reconnect_delay_sec * 1000; + queue_delayed_work(rtrs_wq, &sess->reconnect_dwork, + msecs_to_jiffies(delay_ms + + prandom_u32() % RTRS_RECONNECT_SEED)); + } else { + /* + * Error can happen just on establishing new connection, + * so notify waiter with error state, waiter is responsible + * for cleaning the rest and reconnect if needed. + */ + rtrs_clt_change_state_from_to(sess, + RTRS_CLT_CONNECTING, + RTRS_CLT_CONNECTING_ERR); + } +} + +static void rtrs_clt_fast_reg_done(struct ib_cq *cq, struct ib_wc *wc) +{ + struct rtrs_clt_con *con = cq->cq_context; + + if (unlikely(wc->status != IB_WC_SUCCESS)) { + rtrs_err(con->c.sess, "Failed IB_WR_REG_MR: %s\n", + ib_wc_status_msg(wc->status)); + rtrs_rdma_error_recovery(con); + } +} + +static struct ib_cqe fast_reg_cqe = { + .done = rtrs_clt_fast_reg_done +}; + +static void complete_rdma_req(struct rtrs_clt_io_req *req, int errno, + bool notify, bool can_wait); + +static void rtrs_clt_inv_rkey_done(struct ib_cq *cq, struct ib_wc *wc) +{ + struct rtrs_clt_io_req *req = + container_of(wc->wr_cqe, typeof(*req), inv_cqe); + struct rtrs_clt_con *con = cq->cq_context; + + if (unlikely(wc->status != IB_WC_SUCCESS)) { + rtrs_err(con->c.sess, "Failed IB_WR_LOCAL_INV: %s\n", + ib_wc_status_msg(wc->status)); + rtrs_rdma_error_recovery(con); + } + req->need_inv = false; + if (likely(req->need_inv_comp)) + complete(&req->inv_comp); + else + /* Complete request from INV callback */ + complete_rdma_req(req, req->inv_errno, true, false); +} + +static int rtrs_inv_rkey(struct rtrs_clt_io_req *req) +{ + struct rtrs_clt_con *con = req->con; + struct ib_send_wr wr = { + .opcode = IB_WR_LOCAL_INV, + .wr_cqe = &req->inv_cqe, + .send_flags = IB_SEND_SIGNALED, + .ex.invalidate_rkey = req->mr->rkey, + }; + req->inv_cqe.done = rtrs_clt_inv_rkey_done; + + return ib_post_send(con->c.qp, &wr, NULL); +} + +static void complete_rdma_req(struct rtrs_clt_io_req *req, int errno, + bool notify, bool can_wait) +{ + struct rtrs_clt_con *con = req->con; + struct rtrs_clt_sess *sess; + int err; + + if (!req->in_use) + return; + if (WARN_ON(!req->con)) + return; + sess = to_clt_sess(con->c.sess); + + if (req->sg_cnt) { + if (unlikely(req->dir == DMA_FROM_DEVICE && req->need_inv)) { + /* + * We are here to invalidate read requests + * ourselves. In normal scenario server should + * send INV for all read requests, but + * we are here, thus two things could happen: + * + * 1. this is failover, when errno != 0 + * and can_wait == 1, + * + * 2. something totally bad happened and + * server forgot to send INV, so we + * should do that ourselves. + */ + + if (likely(can_wait)) { + req->need_inv_comp = true; + } else { + /* This should be IO path, so always notify */ + WARN_ON(!notify); + /* Save errno for INV callback */ + req->inv_errno = errno; + } + + err = rtrs_inv_rkey(req); + if (unlikely(err)) { + rtrs_err(con->c.sess, "Send INV WR key=%#x: %d\n", + req->mr->rkey, err); + } else if (likely(can_wait)) { + wait_for_completion(&req->inv_comp); + } else { + /* + * Something went wrong, so request will be + * completed from INV callback. + */ + WARN_ON_ONCE(1); + + return; + } + } + ib_dma_unmap_sg(sess->s.dev->ib_dev, req->sglist, + req->sg_cnt, req->dir); + } + if (sess->clt->mp_policy == MP_POLICY_MIN_INFLIGHT) + atomic_dec(&sess->stats->inflight); + + req->in_use = false; + req->con = NULL; + + if (notify) + req->conf(req->priv, errno); +} + +static int rtrs_post_send_rdma(struct rtrs_clt_con *con, + struct rtrs_clt_io_req *req, + struct rtrs_rbuf *rbuf, u32 off, + u32 imm, struct ib_send_wr *wr) +{ + struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess); + enum ib_send_flags flags; + struct ib_sge sge; + + if (unlikely(!req->sg_size)) { + rtrs_wrn(con->c.sess, + "Doing RDMA Write failed, no data supplied\n"); + return -EINVAL; + } + + /* user data and user message in the first list element */ + sge.addr = req->iu->dma_addr; + sge.length = req->sg_size; + sge.lkey = sess->s.dev->ib_pd->local_dma_lkey; + + /* + * From time to time we have to post signalled sends, + * or send queue will fill up and only QP reset can help. + */ + flags = atomic_inc_return(&con->io_cnt) % sess->queue_depth ? + 0 : IB_SEND_SIGNALED; + + ib_dma_sync_single_for_device(sess->s.dev->ib_dev, req->iu->dma_addr, + req->sg_size, DMA_TO_DEVICE); + + return rtrs_iu_post_rdma_write_imm(&con->c, req->iu, &sge, 1, + rbuf->rkey, rbuf->addr + off, + imm, flags, wr); +} + +static void process_io_rsp(struct rtrs_clt_sess *sess, u32 msg_id, + s16 errno, bool w_inval) +{ + struct rtrs_clt_io_req *req; + + if (WARN_ON(msg_id >= sess->queue_depth)) + return; + + req = &sess->reqs[msg_id]; + /* Drop need_inv if server responded with send with invalidation */ + req->need_inv &= !w_inval; + complete_rdma_req(req, errno, true, false); +} + +static void rtrs_clt_recv_done(struct rtrs_clt_con *con, struct ib_wc *wc) +{ + struct rtrs_iu *iu; + int err; + struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess); + + WARN_ON((sess->flags & RTRS_MSG_NEW_RKEY_F) == 0); + iu = container_of(wc->wr_cqe, struct rtrs_iu, + cqe); + err = rtrs_iu_post_recv(&con->c, iu); + if (unlikely(err)) { + rtrs_err(con->c.sess, "post iu failed %d\n", err); + rtrs_rdma_error_recovery(con); + } +} + +static void rtrs_clt_rkey_rsp_done(struct rtrs_clt_con *con, struct ib_wc *wc) +{ + struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess); + struct rtrs_msg_rkey_rsp *msg; + u32 imm_type, imm_payload; + bool w_inval = false; + struct rtrs_iu *iu; + u32 buf_id; + int err; + + WARN_ON((sess->flags & RTRS_MSG_NEW_RKEY_F) == 0); + + iu = container_of(wc->wr_cqe, struct rtrs_iu, cqe); + + if (unlikely(wc->byte_len < sizeof(*msg))) { + rtrs_err(con->c.sess, "rkey response is malformed: size %d\n", + wc->byte_len); + goto out; + } + ib_dma_sync_single_for_cpu(sess->s.dev->ib_dev, iu->dma_addr, + iu->size, DMA_FROM_DEVICE); + msg = iu->buf; + if (unlikely(le16_to_cpu(msg->type) != RTRS_MSG_RKEY_RSP)) { + rtrs_err(sess->clt, "rkey response is malformed: type %d\n", + le16_to_cpu(msg->type)); + goto out; + } + buf_id = le16_to_cpu(msg->buf_id); + if (WARN_ON(buf_id >= sess->queue_depth)) + goto out; + + rtrs_from_imm(be32_to_cpu(wc->ex.imm_data), &imm_type, &imm_payload); + if (likely(imm_type == RTRS_IO_RSP_IMM || + imm_type == RTRS_IO_RSP_W_INV_IMM)) { + u32 msg_id; + + w_inval = (imm_type == RTRS_IO_RSP_W_INV_IMM); + rtrs_from_io_rsp_imm(imm_payload, &msg_id, &err); + + if (WARN_ON(buf_id != msg_id)) + goto out; + sess->rbufs[buf_id].rkey = le32_to_cpu(msg->rkey); + process_io_rsp(sess, msg_id, err, w_inval); + } + ib_dma_sync_single_for_device(sess->s.dev->ib_dev, iu->dma_addr, + iu->size, DMA_FROM_DEVICE); + return rtrs_clt_recv_done(con, wc); +out: + rtrs_rdma_error_recovery(con); +} + +static void rtrs_clt_rdma_done(struct ib_cq *cq, struct ib_wc *wc); + +static struct ib_cqe io_comp_cqe = { + .done = rtrs_clt_rdma_done +}; + +/* + * Post x2 empty WRs: first is for this RDMA with IMM, + * second is for RECV with INV, which happened earlier. + */ +static int rtrs_post_recv_empty_x2(struct rtrs_con *con, struct ib_cqe *cqe) +{ + struct ib_recv_wr wr_arr[2], *wr; + int i; + + memset(wr_arr, 0, sizeof(wr_arr)); + for (i = 0; i < ARRAY_SIZE(wr_arr); i++) { + wr = &wr_arr[i]; + wr->wr_cqe = cqe; + if (i) + /* Chain backwards */ + wr->next = &wr_arr[i - 1]; + } + + return ib_post_recv(con->qp, wr, NULL); +} + +static void rtrs_clt_rdma_done(struct ib_cq *cq, struct ib_wc *wc) +{ + struct rtrs_clt_con *con = cq->cq_context; + struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess); + u32 imm_type, imm_payload; + bool w_inval = false; + int err; + + if (unlikely(wc->status != IB_WC_SUCCESS)) { + if (wc->status != IB_WC_WR_FLUSH_ERR) { + rtrs_err(sess->clt, "RDMA failed: %s\n", + ib_wc_status_msg(wc->status)); + rtrs_rdma_error_recovery(con); + } + return; + } + rtrs_clt_update_wc_stats(con); + + switch (wc->opcode) { + case IB_WC_RECV_RDMA_WITH_IMM: + /* + * post_recv() RDMA write completions of IO reqs (read/write) + * and hb + */ + if (WARN_ON(wc->wr_cqe->done != rtrs_clt_rdma_done)) + return; + rtrs_from_imm(be32_to_cpu(wc->ex.imm_data), + &imm_type, &imm_payload); + if (likely(imm_type == RTRS_IO_RSP_IMM || + imm_type == RTRS_IO_RSP_W_INV_IMM)) { + u32 msg_id; + + w_inval = (imm_type == RTRS_IO_RSP_W_INV_IMM); + rtrs_from_io_rsp_imm(imm_payload, &msg_id, &err); + + process_io_rsp(sess, msg_id, err, w_inval); + } else if (imm_type == RTRS_HB_MSG_IMM) { + WARN_ON(con->c.cid); + rtrs_send_hb_ack(&sess->s); + if (sess->flags & RTRS_MSG_NEW_RKEY_F) + return rtrs_clt_recv_done(con, wc); + } else if (imm_type == RTRS_HB_ACK_IMM) { + WARN_ON(con->c.cid); + sess->s.hb_missed_cnt = 0; + if (sess->flags & RTRS_MSG_NEW_RKEY_F) + return rtrs_clt_recv_done(con, wc); + } else { + rtrs_wrn(con->c.sess, "Unknown IMM type %u\n", + imm_type); + } + if (w_inval) + /* + * Post x2 empty WRs: first is for this RDMA with IMM, + * second is for RECV with INV, which happened earlier. + */ + err = rtrs_post_recv_empty_x2(&con->c, &io_comp_cqe); + else + err = rtrs_post_recv_empty(&con->c, &io_comp_cqe); + if (unlikely(err)) { + rtrs_err(con->c.sess, "rtrs_post_recv_empty(): %d\n", + err); + rtrs_rdma_error_recovery(con); + break; + } + break; + case IB_WC_RECV: + /* + * Key invalidations from server side + */ + WARN_ON(!(wc->wc_flags & IB_WC_WITH_INVALIDATE || + wc->wc_flags & IB_WC_WITH_IMM)); + WARN_ON(wc->wr_cqe->done != rtrs_clt_rdma_done); + if (sess->flags & RTRS_MSG_NEW_RKEY_F) { + if (wc->wc_flags & IB_WC_WITH_INVALIDATE) + return rtrs_clt_recv_done(con, wc); + + return rtrs_clt_rkey_rsp_done(con, wc); + } + break; + case IB_WC_RDMA_WRITE: + /* + * post_send() RDMA write completions of IO reqs (read/write) + */ + break; + + default: + rtrs_wrn(sess->clt, "Unexpected WC type: %d\n", wc->opcode); + return; + } +} + +static int post_recv_io(struct rtrs_clt_con *con, size_t q_size) +{ + int err, i; + struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess); + + for (i = 0; i < q_size; i++) { + if (sess->flags & RTRS_MSG_NEW_RKEY_F) { + struct rtrs_iu *iu = &con->rsp_ius[i]; + + err = rtrs_iu_post_recv(&con->c, iu); + } else { + err = rtrs_post_recv_empty(&con->c, &io_comp_cqe); + } + if (unlikely(err)) + return err; + } + + return 0; +} + +static int post_recv_sess(struct rtrs_clt_sess *sess) +{ + size_t q_size = 0; + int err, cid; + + for (cid = 0; cid < sess->s.con_num; cid++) { + if (cid == 0) + q_size = SERVICE_CON_QUEUE_DEPTH; + else + q_size = sess->queue_depth; + + /* + * x2 for RDMA read responses + FR key invalidations, + * RDMA writes do not require any FR registrations. + */ + q_size *= 2; + + err = post_recv_io(to_clt_con(sess->s.con[cid]), q_size); + if (unlikely(err)) { + rtrs_err(sess->clt, "post_recv_io(), err: %d\n", err); + return err; + } + } + + return 0; +} + +struct path_it { + int i; + struct list_head skip_list; + struct rtrs_clt *clt; + struct rtrs_clt_sess *(*next_path)(struct path_it *it); +}; + +/** + * list_next_or_null_rr_rcu - get next list element in round-robin fashion. + * @head: the head for the list. + * @ptr: the list head to take the next element from. + * @type: the type of the struct this is embedded in. + * @memb: the name of the list_head within the struct. + * + * Next element returned in round-robin fashion, i.e. head will be skipped, + * but if list is observed as empty, NULL will be returned. + * + * This primitive may safely run concurrently with the _rcu list-mutation + * primitives such as list_add_rcu() as long as it's guarded by rcu_read_lock(). + */ +#define list_next_or_null_rr_rcu(head, ptr, type, memb) \ +({ \ + list_next_or_null_rcu(head, ptr, type, memb) ?: \ + list_next_or_null_rcu(head, READ_ONCE((ptr)->next), \ + type, memb); \ +}) + +/** + * get_next_path_rr() - Returns path in round-robin fashion. + * @it: the path pointer + * + * Related to @MP_POLICY_RR + * + * Locks: + * rcu_read_lock() must be hold. + */ +static struct rtrs_clt_sess *get_next_path_rr(struct path_it *it) +{ + struct rtrs_clt_sess __rcu **ppcpu_path; + struct rtrs_clt_sess *path; + struct rtrs_clt *clt; + + clt = it->clt; + + /* + * Here we use two RCU objects: @paths_list and @pcpu_path + * pointer. See rtrs_clt_remove_path_from_arr() for details + * how that is handled. + */ + + ppcpu_path = this_cpu_ptr(clt->pcpu_path); + path = rcu_dereference(*ppcpu_path); + if (unlikely(!path)) + path = list_first_or_null_rcu(&clt->paths_list, + typeof(*path), s.entry); + else + path = list_next_or_null_rr_rcu(&clt->paths_list, + &path->s.entry, + typeof(*path), + s.entry); + rcu_assign_pointer(*ppcpu_path, path); + + return path; +} + +/** + * get_next_path_min_inflight() - Returns path with minimal inflight count. + * @it: the path pointer + * + * Related to @MP_POLICY_MIN_INFLIGHT + * + * Locks: + * rcu_read_lock() must be hold. + */ +static struct rtrs_clt_sess *get_next_path_min_inflight(struct path_it *it) +{ + struct rtrs_clt_sess *min_path = NULL; + struct rtrs_clt *clt = it->clt; + struct rtrs_clt_sess *sess; + int min_inflight = INT_MAX; + int inflight; + + list_for_each_entry_rcu(sess, &clt->paths_list, s.entry) { + if (unlikely(READ_ONCE(sess->state) != RTRS_CLT_CONNECTED)) + continue; + + if (unlikely(!list_empty(raw_cpu_ptr(sess->mp_skip_entry)))) + continue; + + inflight = atomic_read(&sess->stats->inflight); + + if (inflight < min_inflight) { + min_inflight = inflight; + min_path = sess; + } + } + + /* + * add the path to the skip list, so that next time we can get + * a different one + */ + if (min_path) + list_add(raw_cpu_ptr(min_path->mp_skip_entry), &it->skip_list); + + return min_path; +} + +static inline void path_it_init(struct path_it *it, struct rtrs_clt *clt) +{ + INIT_LIST_HEAD(&it->skip_list); + it->clt = clt; + it->i = 0; + + if (clt->mp_policy == MP_POLICY_RR) + it->next_path = get_next_path_rr; + else + it->next_path = get_next_path_min_inflight; +} + +static inline void path_it_deinit(struct path_it *it) +{ + struct list_head *skip, *tmp; + /* + * The skip_list is used only for the MIN_INFLIGHT policy. + * We need to remove paths from it, so that next IO can insert + * paths (->mp_skip_entry) into a skip_list again. + */ + list_for_each_safe(skip, tmp, &it->skip_list) + list_del_init(skip); +} + +/** + * rtrs_clt_init_req() Initialize an rtrs_clt_io_req holding information + * about an inflight IO. + * The user buffer holding user control message (not data) is copied into + * the corresponding buffer of rtrs_iu (req->iu->buf), which later on will + * also hold the control message of rtrs. + * @req: an io request holding information about IO. + * @sess: client session + * @conf: conformation callback function to notify upper layer. + * @permit: permit for allocation of RDMA remote buffer + * @priv: private pointer + * @vec: kernel vector containing control message + * @usr_len: length of the user message + * @sg: scater list for IO data + * @sg_cnt: number of scater list entries + * @data_len: length of the IO data + * @dir: direction of the IO. + */ +static void rtrs_clt_init_req(struct rtrs_clt_io_req *req, + struct rtrs_clt_sess *sess, + void (*conf)(void *priv, int errno), + struct rtrs_permit *permit, void *priv, + const struct kvec *vec, size_t usr_len, + struct scatterlist *sg, size_t sg_cnt, + size_t data_len, int dir) +{ + struct iov_iter iter; + size_t len; + + req->permit = permit; + req->in_use = true; + req->usr_len = usr_len; + req->data_len = data_len; + req->sglist = sg; + req->sg_cnt = sg_cnt; + req->priv = priv; + req->dir = dir; + req->con = rtrs_permit_to_clt_con(sess, permit); + req->conf = conf; + req->need_inv = false; + req->need_inv_comp = false; + req->inv_errno = 0; + + iov_iter_kvec(&iter, WRITE, vec, 1, usr_len); + len = _copy_from_iter(req->iu->buf, usr_len, &iter); + WARN_ON(len != usr_len); + + reinit_completion(&req->inv_comp); +} + +static struct rtrs_clt_io_req * +rtrs_clt_get_req(struct rtrs_clt_sess *sess, + void (*conf)(void *priv, int errno), + struct rtrs_permit *permit, void *priv, + const struct kvec *vec, size_t usr_len, + struct scatterlist *sg, size_t sg_cnt, + size_t data_len, int dir) +{ + struct rtrs_clt_io_req *req; + + req = &sess->reqs[permit->mem_id]; + rtrs_clt_init_req(req, sess, conf, permit, priv, vec, usr_len, + sg, sg_cnt, data_len, dir); + return req; +} + +static struct rtrs_clt_io_req * +rtrs_clt_get_copy_req(struct rtrs_clt_sess *alive_sess, + struct rtrs_clt_io_req *fail_req) +{ + struct rtrs_clt_io_req *req; + struct kvec vec = { + .iov_base = fail_req->iu->buf, + .iov_len = fail_req->usr_len + }; + + req = &alive_sess->reqs[fail_req->permit->mem_id]; + rtrs_clt_init_req(req, alive_sess, fail_req->conf, fail_req->permit, + fail_req->priv, &vec, fail_req->usr_len, + fail_req->sglist, fail_req->sg_cnt, + fail_req->data_len, fail_req->dir); + return req; +} + +static int rtrs_post_rdma_write_sg(struct rtrs_clt_con *con, + struct rtrs_clt_io_req *req, + struct rtrs_rbuf *rbuf, + u32 size, u32 imm) +{ + struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess); + struct ib_sge *sge = req->sge; + enum ib_send_flags flags; + struct scatterlist *sg; + size_t num_sge; + int i; + + for_each_sg(req->sglist, sg, req->sg_cnt, i) { + sge[i].addr = sg_dma_address(sg); + sge[i].length = sg_dma_len(sg); + sge[i].lkey = sess->s.dev->ib_pd->local_dma_lkey; + } + sge[i].addr = req->iu->dma_addr; + sge[i].length = size; + sge[i].lkey = sess->s.dev->ib_pd->local_dma_lkey; + + num_sge = 1 + req->sg_cnt; + + /* + * From time to time we have to post signalled sends, + * or send queue will fill up and only QP reset can help. + */ + flags = atomic_inc_return(&con->io_cnt) % sess->queue_depth ? + 0 : IB_SEND_SIGNALED; + + ib_dma_sync_single_for_device(sess->s.dev->ib_dev, req->iu->dma_addr, + size, DMA_TO_DEVICE); + + return rtrs_iu_post_rdma_write_imm(&con->c, req->iu, sge, num_sge, + rbuf->rkey, rbuf->addr, imm, + flags, NULL); +} + +static int rtrs_clt_write_req(struct rtrs_clt_io_req *req) +{ + struct rtrs_clt_con *con = req->con; + struct rtrs_sess *s = con->c.sess; + struct rtrs_clt_sess *sess = to_clt_sess(s); + struct rtrs_msg_rdma_write *msg; + + struct rtrs_rbuf *rbuf; + int ret, count = 0; + u32 imm, buf_id; + + const size_t tsize = sizeof(*msg) + req->data_len + req->usr_len; + + if (unlikely(tsize > sess->chunk_size)) { + rtrs_wrn(s, "Write request failed, size too big %zu > %d\n", + tsize, sess->chunk_size); + return -EMSGSIZE; + } + if (req->sg_cnt) { + count = ib_dma_map_sg(sess->s.dev->ib_dev, req->sglist, + req->sg_cnt, req->dir); + if (unlikely(!count)) { + rtrs_wrn(s, "Write request failed, map failed\n"); + return -EINVAL; + } + } + /* put rtrs msg after sg and user message */ + msg = req->iu->buf + req->usr_len; + msg->type = cpu_to_le16(RTRS_MSG_WRITE); + msg->usr_len = cpu_to_le16(req->usr_len); + + /* rtrs message on server side will be after user data and message */ + imm = req->permit->mem_off + req->data_len + req->usr_len; + imm = rtrs_to_io_req_imm(imm); + buf_id = req->permit->mem_id; + req->sg_size = tsize; + rbuf = &sess->rbufs[buf_id]; + + /* + * Update stats now, after request is successfully sent it is not + * safe anymore to touch it. + */ + rtrs_clt_update_all_stats(req, WRITE); + + ret = rtrs_post_rdma_write_sg(req->con, req, rbuf, + req->usr_len + sizeof(*msg), + imm); + if (unlikely(ret)) { + rtrs_err(s, "Write request failed: %d\n", ret); + if (sess->clt->mp_policy == MP_POLICY_MIN_INFLIGHT) + atomic_dec(&sess->stats->inflight); + if (req->sg_cnt) + ib_dma_unmap_sg(sess->s.dev->ib_dev, req->sglist, + req->sg_cnt, req->dir); + } + + return ret; +} + +static int rtrs_map_sg_fr(struct rtrs_clt_io_req *req, size_t count) +{ + int nr; + + /* Align the MR to a 4K page size to match the block virt boundary */ + nr = ib_map_mr_sg(req->mr, req->sglist, count, NULL, SZ_4K); + if (nr < 0) + return nr; + if (unlikely(nr < req->sg_cnt)) + return -EINVAL; + ib_update_fast_reg_key(req->mr, ib_inc_rkey(req->mr->rkey)); + + return nr; +} + +static int rtrs_clt_read_req(struct rtrs_clt_io_req *req) +{ + struct rtrs_clt_con *con = req->con; + struct rtrs_sess *s = con->c.sess; + struct rtrs_clt_sess *sess = to_clt_sess(s); + struct rtrs_msg_rdma_read *msg; + struct rtrs_ib_dev *dev; + + struct ib_reg_wr rwr; + struct ib_send_wr *wr = NULL; + + int ret, count = 0; + u32 imm, buf_id; + + const size_t tsize = sizeof(*msg) + req->data_len + req->usr_len; + + s = &sess->s; + dev = sess->s.dev; + + if (unlikely(tsize > sess->chunk_size)) { + rtrs_wrn(s, + "Read request failed, message size is %zu, bigger than CHUNK_SIZE %d\n", + tsize, sess->chunk_size); + return -EMSGSIZE; + } + + if (req->sg_cnt) { + count = ib_dma_map_sg(dev->ib_dev, req->sglist, req->sg_cnt, + req->dir); + if (unlikely(!count)) { + rtrs_wrn(s, + "Read request failed, dma map failed\n"); + return -EINVAL; + } + } + /* put our message into req->buf after user message*/ + msg = req->iu->buf + req->usr_len; + msg->type = cpu_to_le16(RTRS_MSG_READ); + msg->usr_len = cpu_to_le16(req->usr_len); + + if (count) { + ret = rtrs_map_sg_fr(req, count); + if (ret < 0) { + rtrs_err_rl(s, + "Read request failed, failed to map fast reg. data, err: %d\n", + ret); + ib_dma_unmap_sg(dev->ib_dev, req->sglist, req->sg_cnt, + req->dir); + return ret; + } + rwr = (struct ib_reg_wr) { + .wr.opcode = IB_WR_REG_MR, + .wr.wr_cqe = &fast_reg_cqe, + .mr = req->mr, + .key = req->mr->rkey, + .access = (IB_ACCESS_LOCAL_WRITE | + IB_ACCESS_REMOTE_WRITE), + }; + wr = &rwr.wr; + + msg->sg_cnt = cpu_to_le16(1); + msg->flags = cpu_to_le16(RTRS_MSG_NEED_INVAL_F); + + msg->desc[0].addr = cpu_to_le64(req->mr->iova); + msg->desc[0].key = cpu_to_le32(req->mr->rkey); + msg->desc[0].len = cpu_to_le32(req->mr->length); + + /* Further invalidation is required */ + req->need_inv = !!RTRS_MSG_NEED_INVAL_F; + + } else { + msg->sg_cnt = 0; + msg->flags = 0; + } + /* + * rtrs message will be after the space reserved for disk data and + * user message + */ + imm = req->permit->mem_off + req->data_len + req->usr_len; + imm = rtrs_to_io_req_imm(imm); + buf_id = req->permit->mem_id; + + req->sg_size = sizeof(*msg); + req->sg_size += le16_to_cpu(msg->sg_cnt) * sizeof(struct rtrs_sg_desc); + req->sg_size += req->usr_len; + + /* + * Update stats now, after request is successfully sent it is not + * safe anymore to touch it. + */ + rtrs_clt_update_all_stats(req, READ); + + ret = rtrs_post_send_rdma(req->con, req, &sess->rbufs[buf_id], + req->data_len, imm, wr); + if (unlikely(ret)) { + rtrs_err(s, "Read request failed: %d\n", ret); + if (sess->clt->mp_policy == MP_POLICY_MIN_INFLIGHT) + atomic_dec(&sess->stats->inflight); + req->need_inv = false; + if (req->sg_cnt) + ib_dma_unmap_sg(dev->ib_dev, req->sglist, + req->sg_cnt, req->dir); + } + + return ret; +} + +/** + * rtrs_clt_failover_req() Try to find an active path for a failed request + * @clt: clt context + * @fail_req: a failed io request. + */ +static int rtrs_clt_failover_req(struct rtrs_clt *clt, + struct rtrs_clt_io_req *fail_req) +{ + struct rtrs_clt_sess *alive_sess; + struct rtrs_clt_io_req *req; + int err = -ECONNABORTED; + struct path_it it; + + rcu_read_lock(); + for (path_it_init(&it, clt); + (alive_sess = it.next_path(&it)) && it.i < it.clt->paths_num; + it.i++) { + if (unlikely(READ_ONCE(alive_sess->state) != + RTRS_CLT_CONNECTED)) + continue; + req = rtrs_clt_get_copy_req(alive_sess, fail_req); + if (req->dir == DMA_TO_DEVICE) + err = rtrs_clt_write_req(req); + else + err = rtrs_clt_read_req(req); + if (unlikely(err)) { + req->in_use = false; + continue; + } + /* Success path */ + rtrs_clt_inc_failover_cnt(alive_sess->stats); + break; + } + path_it_deinit(&it); + rcu_read_unlock(); + + return err; +} + +static void fail_all_outstanding_reqs(struct rtrs_clt_sess *sess) +{ + struct rtrs_clt *clt = sess->clt; + struct rtrs_clt_io_req *req; + int i, err; + + if (!sess->reqs) + return; + for (i = 0; i < sess->queue_depth; ++i) { + req = &sess->reqs[i]; + if (!req->in_use) + continue; + + /* + * Safely (without notification) complete failed request. + * After completion this request is still useble and can + * be failovered to another path. + */ + complete_rdma_req(req, -ECONNABORTED, false, true); + + err = rtrs_clt_failover_req(clt, req); + if (unlikely(err)) + /* Failover failed, notify anyway */ + req->conf(req->priv, err); + } +} + +static void free_sess_reqs(struct rtrs_clt_sess *sess) +{ + struct rtrs_clt_io_req *req; + int i; + + if (!sess->reqs) + return; + for (i = 0; i < sess->queue_depth; ++i) { + req = &sess->reqs[i]; + if (req->mr) + ib_dereg_mr(req->mr); + kfree(req->sge); + rtrs_iu_free(req->iu, sess->s.dev->ib_dev, 1); + } + kfree(sess->reqs); + sess->reqs = NULL; +} + +static int alloc_sess_reqs(struct rtrs_clt_sess *sess) +{ + struct rtrs_clt_io_req *req; + struct rtrs_clt *clt = sess->clt; + int i, err = -ENOMEM; + + sess->reqs = kcalloc(sess->queue_depth, sizeof(*sess->reqs), + GFP_KERNEL); + if (!sess->reqs) + return -ENOMEM; + + for (i = 0; i < sess->queue_depth; ++i) { + req = &sess->reqs[i]; + req->iu = rtrs_iu_alloc(1, sess->max_hdr_size, GFP_KERNEL, + sess->s.dev->ib_dev, + DMA_TO_DEVICE, + rtrs_clt_rdma_done); + if (!req->iu) + goto out; + + req->sge = kmalloc_array(clt->max_segments + 1, + sizeof(*req->sge), GFP_KERNEL); + if (!req->sge) + goto out; + + req->mr = ib_alloc_mr(sess->s.dev->ib_pd, IB_MR_TYPE_MEM_REG, + sess->max_pages_per_mr); + if (IS_ERR(req->mr)) { + err = PTR_ERR(req->mr); + req->mr = NULL; + pr_err("Failed to alloc sess->max_pages_per_mr %d\n", + sess->max_pages_per_mr); + goto out; + } + + init_completion(&req->inv_comp); + } + + return 0; + +out: + free_sess_reqs(sess); + + return err; +} + +static int alloc_permits(struct rtrs_clt *clt) +{ + unsigned int chunk_bits; + int err, i; + + clt->permits_map = kcalloc(BITS_TO_LONGS(clt->queue_depth), + sizeof(long), GFP_KERNEL); + if (!clt->permits_map) { + err = -ENOMEM; + goto out_err; + } + clt->permits = kcalloc(clt->queue_depth, permit_size(clt), GFP_KERNEL); + if (!clt->permits) { + err = -ENOMEM; + goto err_map; + } + chunk_bits = ilog2(clt->queue_depth - 1) + 1; + for (i = 0; i < clt->queue_depth; i++) { + struct rtrs_permit *permit; + + permit = get_permit(clt, i); + permit->mem_id = i; + permit->mem_off = i << (MAX_IMM_PAYL_BITS - chunk_bits); + } + + return 0; + +err_map: + kfree(clt->permits_map); + clt->permits_map = NULL; +out_err: + return err; +} + +static void free_permits(struct rtrs_clt *clt) +{ + if (clt->permits_map) { + size_t sz = clt->queue_depth; + + wait_event(clt->permits_wait, + find_first_bit(clt->permits_map, sz) >= sz); + } + kfree(clt->permits_map); + clt->permits_map = NULL; + kfree(clt->permits); + clt->permits = NULL; +} + +static void query_fast_reg_mode(struct rtrs_clt_sess *sess) +{ + struct ib_device *ib_dev; + u64 max_pages_per_mr; + int mr_page_shift; + + ib_dev = sess->s.dev->ib_dev; + + /* + * Use the smallest page size supported by the HCA, down to a + * minimum of 4096 bytes. We're unlikely to build large sglists + * out of smaller entries. + */ + mr_page_shift = max(12, ffs(ib_dev->attrs.page_size_cap) - 1); + max_pages_per_mr = ib_dev->attrs.max_mr_size; + do_div(max_pages_per_mr, (1ull << mr_page_shift)); + sess->max_pages_per_mr = + min3(sess->max_pages_per_mr, (u32)max_pages_per_mr, + ib_dev->attrs.max_fast_reg_page_list_len); + sess->max_send_sge = ib_dev->attrs.max_send_sge; +} + +static bool rtrs_clt_change_state_get_old(struct rtrs_clt_sess *sess, + enum rtrs_clt_state new_state, + enum rtrs_clt_state *old_state) +{ + bool changed; + + spin_lock_irq(&sess->state_wq.lock); + *old_state = sess->state; + changed = __rtrs_clt_change_state(sess, new_state); + spin_unlock_irq(&sess->state_wq.lock); + + return changed; +} + +static bool rtrs_clt_change_state(struct rtrs_clt_sess *sess, + enum rtrs_clt_state new_state) +{ + enum rtrs_clt_state old_state; + + return rtrs_clt_change_state_get_old(sess, new_state, &old_state); +} + +static void rtrs_clt_hb_err_handler(struct rtrs_con *c) +{ + struct rtrs_clt_con *con = container_of(c, typeof(*con), c); + + rtrs_rdma_error_recovery(con); +} + +static void rtrs_clt_init_hb(struct rtrs_clt_sess *sess) +{ + rtrs_init_hb(&sess->s, &io_comp_cqe, + RTRS_HB_INTERVAL_MS, + RTRS_HB_MISSED_MAX, + rtrs_clt_hb_err_handler, + rtrs_wq); +} + +static void rtrs_clt_start_hb(struct rtrs_clt_sess *sess) +{ + rtrs_start_hb(&sess->s); +} + +static void rtrs_clt_stop_hb(struct rtrs_clt_sess *sess) +{ + rtrs_stop_hb(&sess->s); +} + +static void rtrs_clt_reconnect_work(struct work_struct *work); +static void rtrs_clt_close_work(struct work_struct *work); + +static struct rtrs_clt_sess *alloc_sess(struct rtrs_clt *clt, + const struct rtrs_addr *path, + size_t con_num, u16 max_segments, + size_t max_segment_size) +{ + struct rtrs_clt_sess *sess; + int err = -ENOMEM; + int cpu; + + sess = kzalloc(sizeof(*sess), GFP_KERNEL); + if (!sess) + goto err; + + /* Extra connection for user messages */ + con_num += 1; + + sess->s.con = kcalloc(con_num, sizeof(*sess->s.con), GFP_KERNEL); + if (!sess->s.con) + goto err_free_sess; + + sess->stats = kzalloc(sizeof(*sess->stats), GFP_KERNEL); + if (!sess->stats) + goto err_free_con; + + mutex_init(&sess->init_mutex); + uuid_gen(&sess->s.uuid); + memcpy(&sess->s.dst_addr, path->dst, + rdma_addr_size((struct sockaddr *)path->dst)); + + /* + * rdma_resolve_addr() passes src_addr to cma_bind_addr, which + * checks the sa_family to be non-zero. If user passed src_addr=NULL + * the sess->src_addr will contain only zeros, which is then fine. + */ + if (path->src) + memcpy(&sess->s.src_addr, path->src, + rdma_addr_size((struct sockaddr *)path->src)); + strlcpy(sess->s.sessname, clt->sessname, sizeof(sess->s.sessname)); + sess->s.con_num = con_num; + sess->clt = clt; + sess->max_pages_per_mr = max_segments * max_segment_size >> 12; + init_waitqueue_head(&sess->state_wq); + sess->state = RTRS_CLT_CONNECTING; + atomic_set(&sess->connected_cnt, 0); + INIT_WORK(&sess->close_work, rtrs_clt_close_work); + INIT_DELAYED_WORK(&sess->reconnect_dwork, rtrs_clt_reconnect_work); + rtrs_clt_init_hb(sess); + + sess->mp_skip_entry = alloc_percpu(typeof(*sess->mp_skip_entry)); + if (!sess->mp_skip_entry) + goto err_free_stats; + + for_each_possible_cpu(cpu) + INIT_LIST_HEAD(per_cpu_ptr(sess->mp_skip_entry, cpu)); + + err = rtrs_clt_init_stats(sess->stats); + if (err) + goto err_free_percpu; + + return sess; + +err_free_percpu: + free_percpu(sess->mp_skip_entry); +err_free_stats: + kfree(sess->stats); +err_free_con: + kfree(sess->s.con); +err_free_sess: + kfree(sess); +err: + return ERR_PTR(err); +} + +void free_sess(struct rtrs_clt_sess *sess) +{ + free_percpu(sess->mp_skip_entry); + mutex_destroy(&sess->init_mutex); + kfree(sess->s.con); + kfree(sess->rbufs); + kfree(sess); +} + +static int create_con(struct rtrs_clt_sess *sess, unsigned int cid) +{ + struct rtrs_clt_con *con; + + con = kzalloc(sizeof(*con), GFP_KERNEL); + if (!con) + return -ENOMEM; + + /* Map first two connections to the first CPU */ + con->cpu = (cid ? cid - 1 : 0) % nr_cpu_ids; + con->c.cid = cid; + con->c.sess = &sess->s; + atomic_set(&con->io_cnt, 0); + + sess->s.con[cid] = &con->c; + + return 0; +} + +static void destroy_con(struct rtrs_clt_con *con) +{ + struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess); + + sess->s.con[con->c.cid] = NULL; + kfree(con); +} + +static int create_con_cq_qp(struct rtrs_clt_con *con) +{ + struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess); + u32 max_send_wr, max_recv_wr, cq_size; + int err, cq_vector; + struct rtrs_msg_rkey_rsp *rsp; + + /* + * This function can fail, but still destroy_con_cq_qp() should + * be called, this is because create_con_cq_qp() is called on cm + * event path, thus caller/waiter never knows: have we failed before + * create_con_cq_qp() or after. To solve this dilemma without + * creating any additional flags just allow destroy_con_cq_qp() be + * called many times. + */ + + if (con->c.cid == 0) { + /* + * One completion for each receive and two for each send + * (send request + registration) + * + 2 for drain and heartbeat + * in case qp gets into error state + */ + max_send_wr = SERVICE_CON_QUEUE_DEPTH * 2 + 2; + max_recv_wr = SERVICE_CON_QUEUE_DEPTH * 2 + 2; + /* We must be the first here */ + if (WARN_ON(sess->s.dev)) + return -EINVAL; + + /* + * The whole session uses device from user connection. + * Be careful not to close user connection before ib dev + * is gracefully put. + */ + sess->s.dev = rtrs_ib_dev_find_or_add(con->c.cm_id->device, + &dev_pd); + if (!sess->s.dev) { + rtrs_wrn(sess->clt, + "rtrs_ib_dev_find_get_or_add(): no memory\n"); + return -ENOMEM; + } + sess->s.dev_ref = 1; + query_fast_reg_mode(sess); + } else { + /* + * Here we assume that session members are correctly set. + * This is always true if user connection (cid == 0) is + * established first. + */ + if (WARN_ON(!sess->s.dev)) + return -EINVAL; + if (WARN_ON(!sess->queue_depth)) + return -EINVAL; + + /* Shared between connections */ + sess->s.dev_ref++; + max_send_wr = + min_t(int, sess->s.dev->ib_dev->attrs.max_qp_wr, + /* QD * (REQ + RSP + FR REGS or INVS) + drain */ + sess->queue_depth * 3 + 1); + max_recv_wr = + min_t(int, sess->s.dev->ib_dev->attrs.max_qp_wr, + sess->queue_depth * 3 + 1); + } + /* alloc iu to recv new rkey reply when server reports flags set */ + if (sess->flags & RTRS_MSG_NEW_RKEY_F || con->c.cid == 0) { + con->rsp_ius = rtrs_iu_alloc(max_recv_wr, sizeof(*rsp), + GFP_KERNEL, sess->s.dev->ib_dev, + DMA_FROM_DEVICE, + rtrs_clt_rdma_done); + if (!con->rsp_ius) + return -ENOMEM; + con->queue_size = max_recv_wr; + } + cq_size = max_send_wr + max_recv_wr; + cq_vector = con->cpu % sess->s.dev->ib_dev->num_comp_vectors; + err = rtrs_cq_qp_create(&sess->s, &con->c, sess->max_send_sge, + cq_vector, cq_size, max_send_wr, + max_recv_wr, IB_POLL_SOFTIRQ); + /* + * In case of error we do not bother to clean previous allocations, + * since destroy_con_cq_qp() must be called. + */ + return err; +} + +static void destroy_con_cq_qp(struct rtrs_clt_con *con) +{ + struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess); + + /* + * Be careful here: destroy_con_cq_qp() can be called even + * create_con_cq_qp() failed, see comments there. + */ + + rtrs_cq_qp_destroy(&con->c); + if (con->rsp_ius) { + rtrs_iu_free(con->rsp_ius, sess->s.dev->ib_dev, con->queue_size); + con->rsp_ius = NULL; + con->queue_size = 0; + } + if (sess->s.dev_ref && !--sess->s.dev_ref) { + rtrs_ib_dev_put(sess->s.dev); + sess->s.dev = NULL; + } +} + +static void stop_cm(struct rtrs_clt_con *con) +{ + rdma_disconnect(con->c.cm_id); + if (con->c.qp) + ib_drain_qp(con->c.qp); +} + +static void destroy_cm(struct rtrs_clt_con *con) +{ + rdma_destroy_id(con->c.cm_id); + con->c.cm_id = NULL; +} + +static int rtrs_rdma_addr_resolved(struct rtrs_clt_con *con) +{ + struct rtrs_sess *s = con->c.sess; + int err; + + err = create_con_cq_qp(con); + if (err) { + rtrs_err(s, "create_con_cq_qp(), err: %d\n", err); + return err; + } + err = rdma_resolve_route(con->c.cm_id, RTRS_CONNECT_TIMEOUT_MS); + if (err) + rtrs_err(s, "Resolving route failed, err: %d\n", err); + + return err; +} + +static int rtrs_rdma_route_resolved(struct rtrs_clt_con *con) +{ + struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess); + struct rtrs_clt *clt = sess->clt; + struct rtrs_msg_conn_req msg; + struct rdma_conn_param param; + + int err; + + param = (struct rdma_conn_param) { + .retry_count = 7, + .rnr_retry_count = 7, + .private_data = &msg, + .private_data_len = sizeof(msg), + }; + + msg = (struct rtrs_msg_conn_req) { + .magic = cpu_to_le16(RTRS_MAGIC), + .version = cpu_to_le16(RTRS_PROTO_VER), + .cid = cpu_to_le16(con->c.cid), + .cid_num = cpu_to_le16(sess->s.con_num), + .recon_cnt = cpu_to_le16(sess->s.recon_cnt), + }; + msg.first_conn = sess->for_new_clt ? FIRST_CONN : 0; + uuid_copy(&msg.sess_uuid, &sess->s.uuid); + uuid_copy(&msg.paths_uuid, &clt->paths_uuid); + + err = rdma_connect_locked(con->c.cm_id, ¶m); + if (err) + rtrs_err(clt, "rdma_connect_locked(): %d\n", err); + + return err; +} + +static int rtrs_rdma_conn_established(struct rtrs_clt_con *con, + struct rdma_cm_event *ev) +{ + struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess); + struct rtrs_clt *clt = sess->clt; + const struct rtrs_msg_conn_rsp *msg; + u16 version, queue_depth; + int errno; + u8 len; + + msg = ev->param.conn.private_data; + len = ev->param.conn.private_data_len; + if (len < sizeof(*msg)) { + rtrs_err(clt, "Invalid RTRS connection response\n"); + return -ECONNRESET; + } + if (le16_to_cpu(msg->magic) != RTRS_MAGIC) { + rtrs_err(clt, "Invalid RTRS magic\n"); + return -ECONNRESET; + } + version = le16_to_cpu(msg->version); + if (version >> 8 != RTRS_PROTO_VER_MAJOR) { + rtrs_err(clt, "Unsupported major RTRS version: %d, expected %d\n", + version >> 8, RTRS_PROTO_VER_MAJOR); + return -ECONNRESET; + } + errno = le16_to_cpu(msg->errno); + if (errno) { + rtrs_err(clt, "Invalid RTRS message: errno %d\n", + errno); + return -ECONNRESET; + } + if (con->c.cid == 0) { + queue_depth = le16_to_cpu(msg->queue_depth); + + if (sess->queue_depth > 0 && queue_depth != sess->queue_depth) { + rtrs_err(clt, "Error: queue depth changed\n"); + + /* + * Stop any more reconnection attempts + */ + sess->reconnect_attempts = -1; + rtrs_err(clt, + "Disabling auto-reconnect. Trigger a manual reconnect after issue is resolved\n"); + return -ECONNRESET; + } + + if (!sess->rbufs) { + kfree(sess->rbufs); + sess->rbufs = kcalloc(queue_depth, sizeof(*sess->rbufs), + GFP_KERNEL); + if (!sess->rbufs) + return -ENOMEM; + } + sess->queue_depth = queue_depth; + sess->max_hdr_size = le32_to_cpu(msg->max_hdr_size); + sess->max_io_size = le32_to_cpu(msg->max_io_size); + sess->flags = le32_to_cpu(msg->flags); + sess->chunk_size = sess->max_io_size + sess->max_hdr_size; + + /* + * Global IO size is always a minimum. + * If while a reconnection server sends us a value a bit + * higher - client does not care and uses cached minimum. + * + * Since we can have several sessions (paths) restablishing + * connections in parallel, use lock. + */ + mutex_lock(&clt->paths_mutex); + clt->queue_depth = sess->queue_depth; + clt->max_io_size = min_not_zero(sess->max_io_size, + clt->max_io_size); + mutex_unlock(&clt->paths_mutex); + + /* + * Cache the hca_port and hca_name for sysfs + */ + sess->hca_port = con->c.cm_id->port_num; + scnprintf(sess->hca_name, sizeof(sess->hca_name), + sess->s.dev->ib_dev->name); + sess->s.src_addr = con->c.cm_id->route.addr.src_addr; + /* set for_new_clt, to allow future reconnect on any path */ + sess->for_new_clt = 1; + } + + return 0; +} + +static inline void flag_success_on_conn(struct rtrs_clt_con *con) +{ + struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess); + + atomic_inc(&sess->connected_cnt); + con->cm_err = 1; +} + +static int rtrs_rdma_conn_rejected(struct rtrs_clt_con *con, + struct rdma_cm_event *ev) +{ + struct rtrs_sess *s = con->c.sess; + const struct rtrs_msg_conn_rsp *msg; + const char *rej_msg; + int status, errno; + u8 data_len; + + status = ev->status; + rej_msg = rdma_reject_msg(con->c.cm_id, status); + msg = rdma_consumer_reject_data(con->c.cm_id, ev, &data_len); + + if (msg && data_len >= sizeof(*msg)) { + errno = (int16_t)le16_to_cpu(msg->errno); + if (errno == -EBUSY) + rtrs_err(s, + "Previous session is still exists on the server, please reconnect later\n"); + else + rtrs_err(s, + "Connect rejected: status %d (%s), rtrs errno %d\n", + status, rej_msg, errno); + } else { + rtrs_err(s, + "Connect rejected but with malformed message: status %d (%s)\n", + status, rej_msg); + } + + return -ECONNRESET; +} + +static void rtrs_clt_close_conns(struct rtrs_clt_sess *sess, bool wait) +{ + if (rtrs_clt_change_state(sess, RTRS_CLT_CLOSING)) + queue_work(rtrs_wq, &sess->close_work); + if (wait) + flush_work(&sess->close_work); +} + +static inline void flag_error_on_conn(struct rtrs_clt_con *con, int cm_err) +{ + if (con->cm_err == 1) { + struct rtrs_clt_sess *sess; + + sess = to_clt_sess(con->c.sess); + if (atomic_dec_and_test(&sess->connected_cnt)) + + wake_up(&sess->state_wq); + } + con->cm_err = cm_err; +} + +static int rtrs_clt_rdma_cm_handler(struct rdma_cm_id *cm_id, + struct rdma_cm_event *ev) +{ + struct rtrs_clt_con *con = cm_id->context; + struct rtrs_sess *s = con->c.sess; + struct rtrs_clt_sess *sess = to_clt_sess(s); + int cm_err = 0; + + switch (ev->event) { + case RDMA_CM_EVENT_ADDR_RESOLVED: + cm_err = rtrs_rdma_addr_resolved(con); + break; + case RDMA_CM_EVENT_ROUTE_RESOLVED: + cm_err = rtrs_rdma_route_resolved(con); + break; + case RDMA_CM_EVENT_ESTABLISHED: + cm_err = rtrs_rdma_conn_established(con, ev); + if (likely(!cm_err)) { + /* + * Report success and wake up. Here we abuse state_wq, + * i.e. wake up without state change, but we set cm_err. + */ + flag_success_on_conn(con); + wake_up(&sess->state_wq); + return 0; + } + break; + case RDMA_CM_EVENT_REJECTED: + cm_err = rtrs_rdma_conn_rejected(con, ev); + break; + case RDMA_CM_EVENT_CONNECT_ERROR: + case RDMA_CM_EVENT_UNREACHABLE: + rtrs_wrn(s, "CM error event %d\n", ev->event); + cm_err = -ECONNRESET; + break; + case RDMA_CM_EVENT_ADDR_ERROR: + case RDMA_CM_EVENT_ROUTE_ERROR: + cm_err = -EHOSTUNREACH; + break; + case RDMA_CM_EVENT_DISCONNECTED: + case RDMA_CM_EVENT_ADDR_CHANGE: + case RDMA_CM_EVENT_TIMEWAIT_EXIT: + cm_err = -ECONNRESET; + break; + case RDMA_CM_EVENT_DEVICE_REMOVAL: + /* + * Device removal is a special case. Queue close and return 0. + */ + rtrs_clt_close_conns(sess, false); + return 0; + default: + rtrs_err(s, "Unexpected RDMA CM event (%d)\n", ev->event); + cm_err = -ECONNRESET; + break; + } + + if (cm_err) { + /* + * cm error makes sense only on connection establishing, + * in other cases we rely on normal procedure of reconnecting. + */ + flag_error_on_conn(con, cm_err); + rtrs_rdma_error_recovery(con); + } + + return 0; +} + +static int create_cm(struct rtrs_clt_con *con) +{ + struct rtrs_sess *s = con->c.sess; + struct rtrs_clt_sess *sess = to_clt_sess(s); + struct rdma_cm_id *cm_id; + int err; + + cm_id = rdma_create_id(&init_net, rtrs_clt_rdma_cm_handler, con, + sess->s.dst_addr.ss_family == AF_IB ? + RDMA_PS_IB : RDMA_PS_TCP, IB_QPT_RC); + if (IS_ERR(cm_id)) { + err = PTR_ERR(cm_id); + rtrs_err(s, "Failed to create CM ID, err: %d\n", err); + + return err; + } + con->c.cm_id = cm_id; + con->cm_err = 0; + /* allow the port to be reused */ + err = rdma_set_reuseaddr(cm_id, 1); + if (err != 0) { + rtrs_err(s, "Set address reuse failed, err: %d\n", err); + goto destroy_cm; + } + err = rdma_resolve_addr(cm_id, (struct sockaddr *)&sess->s.src_addr, + (struct sockaddr *)&sess->s.dst_addr, + RTRS_CONNECT_TIMEOUT_MS); + if (err) { + rtrs_err(s, "Failed to resolve address, err: %d\n", err); + goto destroy_cm; + } + /* + * Combine connection status and session events. This is needed + * for waiting two possible cases: cm_err has something meaningful + * or session state was really changed to error by device removal. + */ + err = wait_event_interruptible_timeout( + sess->state_wq, + con->cm_err || sess->state != RTRS_CLT_CONNECTING, + msecs_to_jiffies(RTRS_CONNECT_TIMEOUT_MS)); + if (err == 0 || err == -ERESTARTSYS) { + if (err == 0) + err = -ETIMEDOUT; + /* Timedout or interrupted */ + goto errr; + } + if (con->cm_err < 0) { + err = con->cm_err; + goto errr; + } + if (READ_ONCE(sess->state) != RTRS_CLT_CONNECTING) { + /* Device removal */ + err = -ECONNABORTED; + goto errr; + } + + return 0; + +errr: + stop_cm(con); + /* Is safe to call destroy if cq_qp is not inited */ + destroy_con_cq_qp(con); +destroy_cm: + destroy_cm(con); + + return err; +} + +static void rtrs_clt_sess_up(struct rtrs_clt_sess *sess) +{ + struct rtrs_clt *clt = sess->clt; + int up; + + /* + * We can fire RECONNECTED event only when all paths were + * connected on rtrs_clt_open(), then each was disconnected + * and the first one connected again. That's why this nasty + * game with counter value. + */ + + mutex_lock(&clt->paths_ev_mutex); + up = ++clt->paths_up; + /* + * Here it is safe to access paths num directly since up counter + * is greater than MAX_PATHS_NUM only while rtrs_clt_open() is + * in progress, thus paths removals are impossible. + */ + if (up > MAX_PATHS_NUM && up == MAX_PATHS_NUM + clt->paths_num) + clt->paths_up = clt->paths_num; + else if (up == 1) + clt->link_ev(clt->priv, RTRS_CLT_LINK_EV_RECONNECTED); + mutex_unlock(&clt->paths_ev_mutex); + + /* Mark session as established */ + sess->established = true; + sess->reconnect_attempts = 0; + sess->stats->reconnects.successful_cnt++; +} + +static void rtrs_clt_sess_down(struct rtrs_clt_sess *sess) +{ + struct rtrs_clt *clt = sess->clt; + + if (!sess->established) + return; + + sess->established = false; + mutex_lock(&clt->paths_ev_mutex); + WARN_ON(!clt->paths_up); + if (--clt->paths_up == 0) + clt->link_ev(clt->priv, RTRS_CLT_LINK_EV_DISCONNECTED); + mutex_unlock(&clt->paths_ev_mutex); +} + +static void rtrs_clt_stop_and_destroy_conns(struct rtrs_clt_sess *sess) +{ + struct rtrs_clt_con *con; + unsigned int cid; + + WARN_ON(READ_ONCE(sess->state) == RTRS_CLT_CONNECTED); + + /* + * Possible race with rtrs_clt_open(), when DEVICE_REMOVAL comes + * exactly in between. Start destroying after it finishes. + */ + mutex_lock(&sess->init_mutex); + mutex_unlock(&sess->init_mutex); + + /* + * All IO paths must observe !CONNECTED state before we + * free everything. + */ + synchronize_rcu(); + + rtrs_clt_stop_hb(sess); + + /* + * The order it utterly crucial: firstly disconnect and complete all + * rdma requests with error (thus set in_use=false for requests), + * then fail outstanding requests checking in_use for each, and + * eventually notify upper layer about session disconnection. + */ + + for (cid = 0; cid < sess->s.con_num; cid++) { + if (!sess->s.con[cid]) + break; + con = to_clt_con(sess->s.con[cid]); + stop_cm(con); + } + fail_all_outstanding_reqs(sess); + free_sess_reqs(sess); + rtrs_clt_sess_down(sess); + + /* + * Wait for graceful shutdown, namely when peer side invokes + * rdma_disconnect(). 'connected_cnt' is decremented only on + * CM events, thus if other side had crashed and hb has detected + * something is wrong, here we will stuck for exactly timeout ms, + * since CM does not fire anything. That is fine, we are not in + * hurry. + */ + wait_event_timeout(sess->state_wq, !atomic_read(&sess->connected_cnt), + msecs_to_jiffies(RTRS_CONNECT_TIMEOUT_MS)); + + for (cid = 0; cid < sess->s.con_num; cid++) { + if (!sess->s.con[cid]) + break; + con = to_clt_con(sess->s.con[cid]); + destroy_con_cq_qp(con); + destroy_cm(con); + destroy_con(con); + } +} + +static inline bool xchg_sessions(struct rtrs_clt_sess __rcu **rcu_ppcpu_path, + struct rtrs_clt_sess *sess, + struct rtrs_clt_sess *next) +{ + struct rtrs_clt_sess **ppcpu_path; + + /* Call cmpxchg() without sparse warnings */ + ppcpu_path = (typeof(ppcpu_path))rcu_ppcpu_path; + return sess == cmpxchg(ppcpu_path, sess, next); +} + +static void rtrs_clt_remove_path_from_arr(struct rtrs_clt_sess *sess) +{ + struct rtrs_clt *clt = sess->clt; + struct rtrs_clt_sess *next; + bool wait_for_grace = false; + int cpu; + + mutex_lock(&clt->paths_mutex); + list_del_rcu(&sess->s.entry); + + /* Make sure everybody observes path removal. */ + synchronize_rcu(); + + /* + * At this point nobody sees @sess in the list, but still we have + * dangling pointer @pcpu_path which _can_ point to @sess. Since + * nobody can observe @sess in the list, we guarantee that IO path + * will not assign @sess to @pcpu_path, i.e. @pcpu_path can be equal + * to @sess, but can never again become @sess. + */ + + /* + * Decrement paths number only after grace period, because + * caller of do_each_path() must firstly observe list without + * path and only then decremented paths number. + * + * Otherwise there can be the following situation: + * o Two paths exist and IO is coming. + * o One path is removed: + * CPU#0 CPU#1 + * do_each_path(): rtrs_clt_remove_path_from_arr(): + * path = get_next_path() + * ^^^ list_del_rcu(path) + * [!CONNECTED path] clt->paths_num-- + * ^^^^^^^^^ + * load clt->paths_num from 2 to 1 + * ^^^^^^^^^ + * sees 1 + * + * path is observed as !CONNECTED, but do_each_path() loop + * ends, because expression i < clt->paths_num is false. + */ + clt->paths_num--; + + /* + * Get @next connection from current @sess which is going to be + * removed. If @sess is the last element, then @next is NULL. + */ + rcu_read_lock(); + next = list_next_or_null_rr_rcu(&clt->paths_list, &sess->s.entry, + typeof(*next), s.entry); + rcu_read_unlock(); + + /* + * @pcpu paths can still point to the path which is going to be + * removed, so change the pointer manually. + */ + for_each_possible_cpu(cpu) { + struct rtrs_clt_sess __rcu **ppcpu_path; + + ppcpu_path = per_cpu_ptr(clt->pcpu_path, cpu); + if (rcu_dereference_protected(*ppcpu_path, + lockdep_is_held(&clt->paths_mutex)) != sess) + /* + * synchronize_rcu() was called just after deleting + * entry from the list, thus IO code path cannot + * change pointer back to the pointer which is going + * to be removed, we are safe here. + */ + continue; + + /* + * We race with IO code path, which also changes pointer, + * thus we have to be careful not to overwrite it. + */ + if (xchg_sessions(ppcpu_path, sess, next)) + /* + * @ppcpu_path was successfully replaced with @next, + * that means that someone could also pick up the + * @sess and dereferencing it right now, so wait for + * a grace period is required. + */ + wait_for_grace = true; + } + if (wait_for_grace) + synchronize_rcu(); + + mutex_unlock(&clt->paths_mutex); +} + +static void rtrs_clt_add_path_to_arr(struct rtrs_clt_sess *sess, + struct rtrs_addr *addr) +{ + struct rtrs_clt *clt = sess->clt; + + mutex_lock(&clt->paths_mutex); + clt->paths_num++; + + list_add_tail_rcu(&sess->s.entry, &clt->paths_list); + mutex_unlock(&clt->paths_mutex); +} + +static void rtrs_clt_close_work(struct work_struct *work) +{ + struct rtrs_clt_sess *sess; + + sess = container_of(work, struct rtrs_clt_sess, close_work); + + cancel_delayed_work_sync(&sess->reconnect_dwork); + rtrs_clt_stop_and_destroy_conns(sess); + rtrs_clt_change_state(sess, RTRS_CLT_CLOSED); +} + +static int init_conns(struct rtrs_clt_sess *sess) +{ + unsigned int cid; + int err; + + /* + * On every new session connections increase reconnect counter + * to avoid clashes with previous sessions not yet closed + * sessions on a server side. + */ + sess->s.recon_cnt++; + + /* Establish all RDMA connections */ + for (cid = 0; cid < sess->s.con_num; cid++) { + err = create_con(sess, cid); + if (err) + goto destroy; + + err = create_cm(to_clt_con(sess->s.con[cid])); + if (err) { + destroy_con(to_clt_con(sess->s.con[cid])); + goto destroy; + } + } + err = alloc_sess_reqs(sess); + if (err) + goto destroy; + + rtrs_clt_start_hb(sess); + + return 0; + +destroy: + while (cid--) { + struct rtrs_clt_con *con = to_clt_con(sess->s.con[cid]); + + stop_cm(con); + destroy_con_cq_qp(con); + destroy_cm(con); + destroy_con(con); + } + /* + * If we've never taken async path and got an error, say, + * doing rdma_resolve_addr(), switch to CONNECTION_ERR state + * manually to keep reconnecting. + */ + rtrs_clt_change_state(sess, RTRS_CLT_CONNECTING_ERR); + + return err; +} + +static void rtrs_clt_info_req_done(struct ib_cq *cq, struct ib_wc *wc) +{ + struct rtrs_clt_con *con = cq->cq_context; + struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess); + struct rtrs_iu *iu; + + iu = container_of(wc->wr_cqe, struct rtrs_iu, cqe); + rtrs_iu_free(iu, sess->s.dev->ib_dev, 1); + + if (unlikely(wc->status != IB_WC_SUCCESS)) { + rtrs_err(sess->clt, "Sess info request send failed: %s\n", + ib_wc_status_msg(wc->status)); + rtrs_clt_change_state(sess, RTRS_CLT_CONNECTING_ERR); + return; + } + + rtrs_clt_update_wc_stats(con); +} + +static int process_info_rsp(struct rtrs_clt_sess *sess, + const struct rtrs_msg_info_rsp *msg) +{ + unsigned int sg_cnt, total_len; + int i, sgi; + + sg_cnt = le16_to_cpu(msg->sg_cnt); + if (unlikely(!sg_cnt)) + return -EINVAL; + /* + * Check if IB immediate data size is enough to hold the mem_id and + * the offset inside the memory chunk. + */ + if (unlikely((ilog2(sg_cnt - 1) + 1) + + (ilog2(sess->chunk_size - 1) + 1) > + MAX_IMM_PAYL_BITS)) { + rtrs_err(sess->clt, + "RDMA immediate size (%db) not enough to encode %d buffers of size %dB\n", + MAX_IMM_PAYL_BITS, sg_cnt, sess->chunk_size); + return -EINVAL; + } + if (unlikely(!sg_cnt || (sess->queue_depth % sg_cnt))) { + rtrs_err(sess->clt, "Incorrect sg_cnt %d, is not multiple\n", + sg_cnt); + return -EINVAL; + } + total_len = 0; + for (sgi = 0, i = 0; sgi < sg_cnt && i < sess->queue_depth; sgi++) { + const struct rtrs_sg_desc *desc = &msg->desc[sgi]; + u32 len, rkey; + u64 addr; + + addr = le64_to_cpu(desc->addr); + rkey = le32_to_cpu(desc->key); + len = le32_to_cpu(desc->len); + + total_len += len; + + if (unlikely(!len || (len % sess->chunk_size))) { + rtrs_err(sess->clt, "Incorrect [%d].len %d\n", sgi, + len); + return -EINVAL; + } + for ( ; len && i < sess->queue_depth; i++) { + sess->rbufs[i].addr = addr; + sess->rbufs[i].rkey = rkey; + + len -= sess->chunk_size; + addr += sess->chunk_size; + } + } + /* Sanity check */ + if (unlikely(sgi != sg_cnt || i != sess->queue_depth)) { + rtrs_err(sess->clt, "Incorrect sg vector, not fully mapped\n"); + return -EINVAL; + } + if (unlikely(total_len != sess->chunk_size * sess->queue_depth)) { + rtrs_err(sess->clt, "Incorrect total_len %d\n", total_len); + return -EINVAL; + } + + return 0; +} + +static void rtrs_clt_info_rsp_done(struct ib_cq *cq, struct ib_wc *wc) +{ + struct rtrs_clt_con *con = cq->cq_context; + struct rtrs_clt_sess *sess = to_clt_sess(con->c.sess); + struct rtrs_msg_info_rsp *msg; + enum rtrs_clt_state state; + struct rtrs_iu *iu; + size_t rx_sz; + int err; + + state = RTRS_CLT_CONNECTING_ERR; + + WARN_ON(con->c.cid); + iu = container_of(wc->wr_cqe, struct rtrs_iu, cqe); + if (unlikely(wc->status != IB_WC_SUCCESS)) { + rtrs_err(sess->clt, "Sess info response recv failed: %s\n", + ib_wc_status_msg(wc->status)); + goto out; + } + WARN_ON(wc->opcode != IB_WC_RECV); + + if (unlikely(wc->byte_len < sizeof(*msg))) { + rtrs_err(sess->clt, "Sess info response is malformed: size %d\n", + wc->byte_len); + goto out; + } + ib_dma_sync_single_for_cpu(sess->s.dev->ib_dev, iu->dma_addr, + iu->size, DMA_FROM_DEVICE); + msg = iu->buf; + if (unlikely(le16_to_cpu(msg->type) != RTRS_MSG_INFO_RSP)) { + rtrs_err(sess->clt, "Sess info response is malformed: type %d\n", + le16_to_cpu(msg->type)); + goto out; + } + rx_sz = sizeof(*msg); + rx_sz += sizeof(msg->desc[0]) * le16_to_cpu(msg->sg_cnt); + if (unlikely(wc->byte_len < rx_sz)) { + rtrs_err(sess->clt, "Sess info response is malformed: size %d\n", + wc->byte_len); + goto out; + } + err = process_info_rsp(sess, msg); + if (unlikely(err)) + goto out; + + err = post_recv_sess(sess); + if (unlikely(err)) + goto out; + + state = RTRS_CLT_CONNECTED; + +out: + rtrs_clt_update_wc_stats(con); + rtrs_iu_free(iu, sess->s.dev->ib_dev, 1); + rtrs_clt_change_state(sess, state); +} + +static int rtrs_send_sess_info(struct rtrs_clt_sess *sess) +{ + struct rtrs_clt_con *usr_con = to_clt_con(sess->s.con[0]); + struct rtrs_msg_info_req *msg; + struct rtrs_iu *tx_iu, *rx_iu; + size_t rx_sz; + int err; + + rx_sz = sizeof(struct rtrs_msg_info_rsp); + rx_sz += sizeof(u64) * MAX_SESS_QUEUE_DEPTH; + + tx_iu = rtrs_iu_alloc(1, sizeof(struct rtrs_msg_info_req), GFP_KERNEL, + sess->s.dev->ib_dev, DMA_TO_DEVICE, + rtrs_clt_info_req_done); + rx_iu = rtrs_iu_alloc(1, rx_sz, GFP_KERNEL, sess->s.dev->ib_dev, + DMA_FROM_DEVICE, rtrs_clt_info_rsp_done); + if (unlikely(!tx_iu || !rx_iu)) { + err = -ENOMEM; + goto out; + } + /* Prepare for getting info response */ + err = rtrs_iu_post_recv(&usr_con->c, rx_iu); + if (unlikely(err)) { + rtrs_err(sess->clt, "rtrs_iu_post_recv(), err: %d\n", err); + goto out; + } + rx_iu = NULL; + + msg = tx_iu->buf; + msg->type = cpu_to_le16(RTRS_MSG_INFO_REQ); + memcpy(msg->sessname, sess->s.sessname, sizeof(msg->sessname)); + + ib_dma_sync_single_for_device(sess->s.dev->ib_dev, tx_iu->dma_addr, + tx_iu->size, DMA_TO_DEVICE); + + /* Send info request */ + err = rtrs_iu_post_send(&usr_con->c, tx_iu, sizeof(*msg), NULL); + if (unlikely(err)) { + rtrs_err(sess->clt, "rtrs_iu_post_send(), err: %d\n", err); + goto out; + } + tx_iu = NULL; + + /* Wait for state change */ + wait_event_interruptible_timeout(sess->state_wq, + sess->state != RTRS_CLT_CONNECTING, + msecs_to_jiffies( + RTRS_CONNECT_TIMEOUT_MS)); + if (unlikely(READ_ONCE(sess->state) != RTRS_CLT_CONNECTED)) { + if (READ_ONCE(sess->state) == RTRS_CLT_CONNECTING_ERR) + err = -ECONNRESET; + else + err = -ETIMEDOUT; + goto out; + } + +out: + if (tx_iu) + rtrs_iu_free(tx_iu, sess->s.dev->ib_dev, 1); + if (rx_iu) + rtrs_iu_free(rx_iu, sess->s.dev->ib_dev, 1); + if (unlikely(err)) + /* If we've never taken async path because of malloc problems */ + rtrs_clt_change_state(sess, RTRS_CLT_CONNECTING_ERR); + + return err; +} + +/** + * init_sess() - establishes all session connections and does handshake + * @sess: client session. + * In case of error full close or reconnect procedure should be taken, + * because reconnect or close async works can be started. + */ +static int init_sess(struct rtrs_clt_sess *sess) +{ + int err; + + mutex_lock(&sess->init_mutex); + err = init_conns(sess); + if (err) { + rtrs_err(sess->clt, "init_conns(), err: %d\n", err); + goto out; + } + err = rtrs_send_sess_info(sess); + if (err) { + rtrs_err(sess->clt, "rtrs_send_sess_info(), err: %d\n", err); + goto out; + } + rtrs_clt_sess_up(sess); +out: + mutex_unlock(&sess->init_mutex); + + return err; +} + +static void rtrs_clt_reconnect_work(struct work_struct *work) +{ + struct rtrs_clt_sess *sess; + struct rtrs_clt *clt; + unsigned int delay_ms; + int err; + + sess = container_of(to_delayed_work(work), struct rtrs_clt_sess, + reconnect_dwork); + clt = sess->clt; + + if (READ_ONCE(sess->state) != RTRS_CLT_RECONNECTING) + return; + + if (sess->reconnect_attempts >= clt->max_reconnect_attempts) { + /* Close a session completely if max attempts is reached */ + rtrs_clt_close_conns(sess, false); + return; + } + sess->reconnect_attempts++; + + /* Stop everything */ + rtrs_clt_stop_and_destroy_conns(sess); + msleep(RTRS_RECONNECT_BACKOFF); + if (rtrs_clt_change_state(sess, RTRS_CLT_CONNECTING)) { + err = init_sess(sess); + if (err) + goto reconnect_again; + } + + return; + +reconnect_again: + if (rtrs_clt_change_state(sess, RTRS_CLT_RECONNECTING)) { + sess->stats->reconnects.fail_cnt++; + delay_ms = clt->reconnect_delay_sec * 1000; + queue_delayed_work(rtrs_wq, &sess->reconnect_dwork, + msecs_to_jiffies(delay_ms + + prandom_u32() % + RTRS_RECONNECT_SEED)); + } +} + +static void rtrs_clt_dev_release(struct device *dev) +{ + struct rtrs_clt *clt = container_of(dev, struct rtrs_clt, dev); + + mutex_destroy(&clt->paths_ev_mutex); + mutex_destroy(&clt->paths_mutex); + kfree(clt); +} + +static struct rtrs_clt *alloc_clt(const char *sessname, size_t paths_num, + u16 port, size_t pdu_sz, void *priv, + void (*link_ev)(void *priv, + enum rtrs_clt_link_ev ev), + unsigned int max_segments, + size_t max_segment_size, + unsigned int reconnect_delay_sec, + unsigned int max_reconnect_attempts) +{ + struct rtrs_clt *clt; + int err; + + if (!paths_num || paths_num > MAX_PATHS_NUM) + return ERR_PTR(-EINVAL); + + if (strlen(sessname) >= sizeof(clt->sessname)) + return ERR_PTR(-EINVAL); + + clt = kzalloc(sizeof(*clt), GFP_KERNEL); + if (!clt) + return ERR_PTR(-ENOMEM); + + clt->pcpu_path = alloc_percpu(typeof(*clt->pcpu_path)); + if (!clt->pcpu_path) { + kfree(clt); + return ERR_PTR(-ENOMEM); + } + + clt->dev.class = rtrs_clt_dev_class; + clt->dev.release = rtrs_clt_dev_release; + uuid_gen(&clt->paths_uuid); + INIT_LIST_HEAD_RCU(&clt->paths_list); + clt->paths_num = paths_num; + clt->paths_up = MAX_PATHS_NUM; + clt->port = port; + clt->pdu_sz = pdu_sz; + clt->max_segments = max_segments; + clt->max_segment_size = max_segment_size; + clt->reconnect_delay_sec = reconnect_delay_sec; + clt->max_reconnect_attempts = max_reconnect_attempts; + clt->priv = priv; + clt->link_ev = link_ev; + clt->mp_policy = MP_POLICY_MIN_INFLIGHT; + strlcpy(clt->sessname, sessname, sizeof(clt->sessname)); + init_waitqueue_head(&clt->permits_wait); + mutex_init(&clt->paths_ev_mutex); + mutex_init(&clt->paths_mutex); + device_initialize(&clt->dev); + + err = dev_set_name(&clt->dev, "%s", sessname); + if (err) + goto err_put; + + /* + * Suppress user space notification until + * sysfs files are created + */ + dev_set_uevent_suppress(&clt->dev, true); + err = device_add(&clt->dev); + if (err) + goto err_put; + + clt->kobj_paths = kobject_create_and_add("paths", &clt->dev.kobj); + if (!clt->kobj_paths) { + err = -ENOMEM; + goto err_del; + } + err = rtrs_clt_create_sysfs_root_files(clt); + if (err) { + kobject_del(clt->kobj_paths); + kobject_put(clt->kobj_paths); + goto err_del; + } + dev_set_uevent_suppress(&clt->dev, false); + kobject_uevent(&clt->dev.kobj, KOBJ_ADD); + + return clt; +err_del: + device_del(&clt->dev); +err_put: + free_percpu(clt->pcpu_path); + put_device(&clt->dev); + return ERR_PTR(err); +} + +static void free_clt(struct rtrs_clt *clt) +{ + free_percpu(clt->pcpu_path); + + /* + * release callback will free clt and destroy mutexes in last put + */ + device_unregister(&clt->dev); +} + +/** + * rtrs_clt_open() - Open a session to an RTRS server + * @ops: holds the link event callback and the private pointer. + * @sessname: name of the session + * @paths: Paths to be established defined by their src and dst addresses + * @paths_num: Number of elements in the @paths array + * @port: port to be used by the RTRS session + * @pdu_sz: Size of extra payload which can be accessed after permit allocation. + * @reconnect_delay_sec: time between reconnect tries + * @max_segments: Max. number of segments per IO request + * @max_segment_size: Max. size of one segment + * @max_reconnect_attempts: Number of times to reconnect on error before giving + * up, 0 for * disabled, -1 for forever + * + * Starts session establishment with the rtrs_server. The function can block + * up to ~2000ms before it returns. + * + * Return a valid pointer on success otherwise PTR_ERR. + */ +struct rtrs_clt *rtrs_clt_open(struct rtrs_clt_ops *ops, + const char *sessname, + const struct rtrs_addr *paths, + size_t paths_num, u16 port, + size_t pdu_sz, u8 reconnect_delay_sec, + u16 max_segments, + size_t max_segment_size, + s16 max_reconnect_attempts) +{ + struct rtrs_clt_sess *sess, *tmp; + struct rtrs_clt *clt; + int err, i; + + clt = alloc_clt(sessname, paths_num, port, pdu_sz, ops->priv, + ops->link_ev, + max_segments, max_segment_size, reconnect_delay_sec, + max_reconnect_attempts); + if (IS_ERR(clt)) { + err = PTR_ERR(clt); + goto out; + } + for (i = 0; i < paths_num; i++) { + struct rtrs_clt_sess *sess; + + sess = alloc_sess(clt, &paths[i], nr_cpu_ids, + max_segments, max_segment_size); + if (IS_ERR(sess)) { + err = PTR_ERR(sess); + goto close_all_sess; + } + if (!i) + sess->for_new_clt = 1; + list_add_tail_rcu(&sess->s.entry, &clt->paths_list); + + err = init_sess(sess); + if (err) { + list_del_rcu(&sess->s.entry); + rtrs_clt_close_conns(sess, true); + free_percpu(sess->stats->pcpu_stats); + kfree(sess->stats); + free_sess(sess); + goto close_all_sess; + } + + err = rtrs_clt_create_sess_files(sess); + if (err) { + list_del_rcu(&sess->s.entry); + rtrs_clt_close_conns(sess, true); + free_percpu(sess->stats->pcpu_stats); + kfree(sess->stats); + free_sess(sess); + goto close_all_sess; + } + } + err = alloc_permits(clt); + if (err) + goto close_all_sess; + + return clt; + +close_all_sess: + list_for_each_entry_safe(sess, tmp, &clt->paths_list, s.entry) { + rtrs_clt_destroy_sess_files(sess, NULL); + rtrs_clt_close_conns(sess, true); + kobject_put(&sess->kobj); + } + rtrs_clt_destroy_sysfs_root_files(clt); + rtrs_clt_destroy_sysfs_root_folders(clt); + free_clt(clt); + +out: + return ERR_PTR(err); +} +EXPORT_SYMBOL(rtrs_clt_open); + +/** + * rtrs_clt_close() - Close a session + * @clt: Session handle. Session is freed upon return. + */ +void rtrs_clt_close(struct rtrs_clt *clt) +{ + struct rtrs_clt_sess *sess, *tmp; + + /* Firstly forbid sysfs access */ + rtrs_clt_destroy_sysfs_root_files(clt); + rtrs_clt_destroy_sysfs_root_folders(clt); + + /* Now it is safe to iterate over all paths without locks */ + list_for_each_entry_safe(sess, tmp, &clt->paths_list, s.entry) { + rtrs_clt_close_conns(sess, true); + rtrs_clt_destroy_sess_files(sess, NULL); + kobject_put(&sess->kobj); + } + free_permits(clt); + free_clt(clt); +} +EXPORT_SYMBOL(rtrs_clt_close); + +int rtrs_clt_reconnect_from_sysfs(struct rtrs_clt_sess *sess) +{ + enum rtrs_clt_state old_state; + int err = -EBUSY; + bool changed; + + changed = rtrs_clt_change_state_get_old(sess, RTRS_CLT_RECONNECTING, + &old_state); + if (changed) { + sess->reconnect_attempts = 0; + queue_delayed_work(rtrs_wq, &sess->reconnect_dwork, 0); + } + if (changed || old_state == RTRS_CLT_RECONNECTING) { + /* + * flush_delayed_work() queues pending work for immediate + * execution, so do the flush if we have queued something + * right now or work is pending. + */ + flush_delayed_work(&sess->reconnect_dwork); + err = (READ_ONCE(sess->state) == + RTRS_CLT_CONNECTED ? 0 : -ENOTCONN); + } + + return err; +} + +int rtrs_clt_disconnect_from_sysfs(struct rtrs_clt_sess *sess) +{ + rtrs_clt_close_conns(sess, true); + + return 0; +} + +int rtrs_clt_remove_path_from_sysfs(struct rtrs_clt_sess *sess, + const struct attribute *sysfs_self) +{ + enum rtrs_clt_state old_state; + bool changed; + + /* + * Continue stopping path till state was changed to DEAD or + * state was observed as DEAD: + * 1. State was changed to DEAD - we were fast and nobody + * invoked rtrs_clt_reconnect(), which can again start + * reconnecting. + * 2. State was observed as DEAD - we have someone in parallel + * removing the path. + */ + do { + rtrs_clt_close_conns(sess, true); + changed = rtrs_clt_change_state_get_old(sess, + RTRS_CLT_DEAD, + &old_state); + } while (!changed && old_state != RTRS_CLT_DEAD); + + if (likely(changed)) { + rtrs_clt_remove_path_from_arr(sess); + rtrs_clt_destroy_sess_files(sess, sysfs_self); + kobject_put(&sess->kobj); + } + + return 0; +} + +void rtrs_clt_set_max_reconnect_attempts(struct rtrs_clt *clt, int value) +{ + clt->max_reconnect_attempts = (unsigned int)value; +} + +int rtrs_clt_get_max_reconnect_attempts(const struct rtrs_clt *clt) +{ + return (int)clt->max_reconnect_attempts; +} + +/** + * rtrs_clt_request() - Request data transfer to/from server via RDMA. + * + * @dir: READ/WRITE + * @ops: callback function to be called as confirmation, and the pointer. + * @clt: Session + * @permit: Preallocated permit + * @vec: Message that is sent to server together with the request. + * Sum of len of all @vec elements limited to <= IO_MSG_SIZE. + * Since the msg is copied internally it can be allocated on stack. + * @nr: Number of elements in @vec. + * @data_len: length of data sent to/from server + * @sg: Pages to be sent/received to/from server. + * @sg_cnt: Number of elements in the @sg + * + * Return: + * 0: Success + * <0: Error + * + * On dir=READ rtrs client will request a data transfer from Server to client. + * The data that the server will respond with will be stored in @sg when + * the user receives an %RTRS_CLT_RDMA_EV_RDMA_REQUEST_WRITE_COMPL event. + * On dir=WRITE rtrs client will rdma write data in sg to server side. + */ +int rtrs_clt_request(int dir, struct rtrs_clt_req_ops *ops, + struct rtrs_clt *clt, struct rtrs_permit *permit, + const struct kvec *vec, size_t nr, size_t data_len, + struct scatterlist *sg, unsigned int sg_cnt) +{ + struct rtrs_clt_io_req *req; + struct rtrs_clt_sess *sess; + + enum dma_data_direction dma_dir; + int err = -ECONNABORTED, i; + size_t usr_len, hdr_len; + struct path_it it; + + /* Get kvec length */ + for (i = 0, usr_len = 0; i < nr; i++) + usr_len += vec[i].iov_len; + + if (dir == READ) { + hdr_len = sizeof(struct rtrs_msg_rdma_read) + + sg_cnt * sizeof(struct rtrs_sg_desc); + dma_dir = DMA_FROM_DEVICE; + } else { + hdr_len = sizeof(struct rtrs_msg_rdma_write); + dma_dir = DMA_TO_DEVICE; + } + + rcu_read_lock(); + for (path_it_init(&it, clt); + (sess = it.next_path(&it)) && it.i < it.clt->paths_num; it.i++) { + if (unlikely(READ_ONCE(sess->state) != RTRS_CLT_CONNECTED)) + continue; + + if (unlikely(usr_len + hdr_len > sess->max_hdr_size)) { + rtrs_wrn_rl(sess->clt, + "%s request failed, user message size is %zu and header length %zu, but max size is %u\n", + dir == READ ? "Read" : "Write", + usr_len, hdr_len, sess->max_hdr_size); + err = -EMSGSIZE; + break; + } + req = rtrs_clt_get_req(sess, ops->conf_fn, permit, ops->priv, + vec, usr_len, sg, sg_cnt, data_len, + dma_dir); + if (dir == READ) + err = rtrs_clt_read_req(req); + else + err = rtrs_clt_write_req(req); + if (unlikely(err)) { + req->in_use = false; + continue; + } + /* Success path */ + break; + } + path_it_deinit(&it); + rcu_read_unlock(); + + return err; +} +EXPORT_SYMBOL(rtrs_clt_request); + +/** + * rtrs_clt_query() - queries RTRS session attributes + *@clt: session pointer + *@attr: query results for session attributes. + * Returns: + * 0 on success + * -ECOMM no connection to the server + */ +int rtrs_clt_query(struct rtrs_clt *clt, struct rtrs_attrs *attr) +{ + if (!rtrs_clt_is_connected(clt)) + return -ECOMM; + + attr->queue_depth = clt->queue_depth; + attr->max_io_size = clt->max_io_size; + attr->sess_kobj = &clt->dev.kobj; + strlcpy(attr->sessname, clt->sessname, sizeof(attr->sessname)); + + return 0; +} +EXPORT_SYMBOL(rtrs_clt_query); + +int rtrs_clt_create_path_from_sysfs(struct rtrs_clt *clt, + struct rtrs_addr *addr) +{ + struct rtrs_clt_sess *sess; + int err; + + sess = alloc_sess(clt, addr, nr_cpu_ids, clt->max_segments, + clt->max_segment_size); + if (IS_ERR(sess)) + return PTR_ERR(sess); + + /* + * It is totally safe to add path in CONNECTING state: coming + * IO will never grab it. Also it is very important to add + * path before init, since init fires LINK_CONNECTED event. + */ + rtrs_clt_add_path_to_arr(sess, addr); + + err = init_sess(sess); + if (err) + goto close_sess; + + err = rtrs_clt_create_sess_files(sess); + if (err) + goto close_sess; + + return 0; + +close_sess: + rtrs_clt_remove_path_from_arr(sess); + rtrs_clt_close_conns(sess, true); + free_percpu(sess->stats->pcpu_stats); + kfree(sess->stats); + free_sess(sess); + + return err; +} + +static int rtrs_clt_ib_dev_init(struct rtrs_ib_dev *dev) +{ + if (!(dev->ib_dev->attrs.device_cap_flags & + IB_DEVICE_MEM_MGT_EXTENSIONS)) { + pr_err("Memory registrations not supported.\n"); + return -ENOTSUPP; + } + + return 0; +} + +static const struct rtrs_rdma_dev_pd_ops dev_pd_ops = { + .init = rtrs_clt_ib_dev_init +}; + +static int __init rtrs_client_init(void) +{ + rtrs_rdma_dev_pd_init(0, &dev_pd); + + rtrs_clt_dev_class = class_create(THIS_MODULE, "rtrs-client"); + if (IS_ERR(rtrs_clt_dev_class)) { + pr_err("Failed to create rtrs-client dev class\n"); + return PTR_ERR(rtrs_clt_dev_class); + } + rtrs_wq = alloc_workqueue("rtrs_client_wq", 0, 0); + if (!rtrs_wq) { + class_destroy(rtrs_clt_dev_class); + return -ENOMEM; + } + + return 0; +} + +static void __exit rtrs_client_exit(void) +{ + destroy_workqueue(rtrs_wq); + class_destroy(rtrs_clt_dev_class); + rtrs_rdma_dev_pd_deinit(&dev_pd); +} + +module_init(rtrs_client_init); +module_exit(rtrs_client_exit); diff --git a/drivers/infiniband/ulp/rtrs/rtrs-clt.h b/drivers/infiniband/ulp/rtrs/rtrs-clt.h new file mode 100644 index 000000000..22da5d50c --- /dev/null +++ b/drivers/infiniband/ulp/rtrs/rtrs-clt.h @@ -0,0 +1,253 @@ +/* SPDX-License-Identifier: GPL-2.0-or-later */ +/* + * RDMA Transport Layer + * + * Copyright (c) 2014 - 2018 ProfitBricks GmbH. All rights reserved. + * Copyright (c) 2018 - 2019 1&1 IONOS Cloud GmbH. All rights reserved. + * Copyright (c) 2019 - 2020 1&1 IONOS SE. All rights reserved. + */ + +#ifndef RTRS_CLT_H +#define RTRS_CLT_H + +#include <linux/device.h> +#include "rtrs-pri.h" + +/** + * enum rtrs_clt_state - Client states. + */ +enum rtrs_clt_state { + RTRS_CLT_CONNECTING, + RTRS_CLT_CONNECTING_ERR, + RTRS_CLT_RECONNECTING, + RTRS_CLT_CONNECTED, + RTRS_CLT_CLOSING, + RTRS_CLT_CLOSED, + RTRS_CLT_DEAD, +}; + +enum rtrs_mp_policy { + MP_POLICY_RR, + MP_POLICY_MIN_INFLIGHT, +}; + +/* see Documentation/ABI/testing/sysfs-class-rtrs-client for details */ +struct rtrs_clt_stats_reconnects { + int successful_cnt; + int fail_cnt; +}; + +/* see Documentation/ABI/testing/sysfs-class-rtrs-client for details */ +struct rtrs_clt_stats_cpu_migr { + atomic_t from; + int to; +}; + +/* stats for Read and write operation. + * see Documentation/ABI/testing/sysfs-class-rtrs-client for details + */ +struct rtrs_clt_stats_rdma { + struct { + u64 cnt; + u64 size_total; + } dir[2]; + + u64 failover_cnt; +}; + +struct rtrs_clt_stats_pcpu { + struct rtrs_clt_stats_cpu_migr cpu_migr; + struct rtrs_clt_stats_rdma rdma; +}; + +struct rtrs_clt_stats { + struct kobject kobj_stats; + struct rtrs_clt_stats_pcpu __percpu *pcpu_stats; + struct rtrs_clt_stats_reconnects reconnects; + atomic_t inflight; +}; + +struct rtrs_clt_con { + struct rtrs_con c; + struct rtrs_iu *rsp_ius; + u32 queue_size; + unsigned int cpu; + atomic_t io_cnt; + int cm_err; +}; + +/** + * rtrs_permit - permits the memory allocation for future RDMA operation. + * Combine with irq pinning to keep IO on same CPU. + */ +struct rtrs_permit { + enum rtrs_clt_con_type con_type; + unsigned int cpu_id; + unsigned int mem_id; + unsigned int mem_off; +}; + +/** + * rtrs_clt_io_req - describes one inflight IO request + */ +struct rtrs_clt_io_req { + struct list_head list; + struct rtrs_iu *iu; + struct scatterlist *sglist; /* list holding user data */ + unsigned int sg_cnt; + unsigned int sg_size; + unsigned int data_len; + unsigned int usr_len; + void *priv; + bool in_use; + struct rtrs_clt_con *con; + struct rtrs_sg_desc *desc; + struct ib_sge *sge; + struct rtrs_permit *permit; + enum dma_data_direction dir; + void (*conf)(void *priv, int errno); + unsigned long start_jiffies; + + struct ib_mr *mr; + struct ib_cqe inv_cqe; + struct completion inv_comp; + int inv_errno; + bool need_inv_comp; + bool need_inv; +}; + +struct rtrs_rbuf { + u64 addr; + u32 rkey; +}; + +struct rtrs_clt_sess { + struct rtrs_sess s; + struct rtrs_clt *clt; + wait_queue_head_t state_wq; + enum rtrs_clt_state state; + atomic_t connected_cnt; + struct mutex init_mutex; + struct rtrs_clt_io_req *reqs; + struct delayed_work reconnect_dwork; + struct work_struct close_work; + unsigned int reconnect_attempts; + bool established; + struct rtrs_rbuf *rbufs; + size_t max_io_size; + u32 max_hdr_size; + u32 chunk_size; + size_t queue_depth; + u32 max_pages_per_mr; + int max_send_sge; + u32 flags; + struct kobject kobj; + u8 for_new_clt; + struct rtrs_clt_stats *stats; + /* cache hca_port and hca_name to display in sysfs */ + u8 hca_port; + char hca_name[IB_DEVICE_NAME_MAX]; + struct list_head __percpu + *mp_skip_entry; +}; + +struct rtrs_clt { + struct list_head paths_list; /* rcu protected list */ + size_t paths_num; + struct rtrs_clt_sess + __rcu * __percpu *pcpu_path; + uuid_t paths_uuid; + int paths_up; + struct mutex paths_mutex; + struct mutex paths_ev_mutex; + char sessname[NAME_MAX]; + u16 port; + unsigned int max_reconnect_attempts; + unsigned int reconnect_delay_sec; + unsigned int max_segments; + size_t max_segment_size; + void *permits; + unsigned long *permits_map; + size_t queue_depth; + size_t max_io_size; + wait_queue_head_t permits_wait; + size_t pdu_sz; + void *priv; + void (*link_ev)(void *priv, + enum rtrs_clt_link_ev ev); + struct device dev; + struct kobject *kobj_paths; + enum rtrs_mp_policy mp_policy; +}; + +static inline struct rtrs_clt_con *to_clt_con(struct rtrs_con *c) +{ + return container_of(c, struct rtrs_clt_con, c); +} + +static inline struct rtrs_clt_sess *to_clt_sess(struct rtrs_sess *s) +{ + return container_of(s, struct rtrs_clt_sess, s); +} + +static inline int permit_size(struct rtrs_clt *clt) +{ + return sizeof(struct rtrs_permit) + clt->pdu_sz; +} + +static inline struct rtrs_permit *get_permit(struct rtrs_clt *clt, int idx) +{ + return (struct rtrs_permit *)(clt->permits + permit_size(clt) * idx); +} + +int rtrs_clt_reconnect_from_sysfs(struct rtrs_clt_sess *sess); +int rtrs_clt_disconnect_from_sysfs(struct rtrs_clt_sess *sess); +int rtrs_clt_create_path_from_sysfs(struct rtrs_clt *clt, + struct rtrs_addr *addr); +int rtrs_clt_remove_path_from_sysfs(struct rtrs_clt_sess *sess, + const struct attribute *sysfs_self); + +void rtrs_clt_set_max_reconnect_attempts(struct rtrs_clt *clt, int value); +int rtrs_clt_get_max_reconnect_attempts(const struct rtrs_clt *clt); +void free_sess(struct rtrs_clt_sess *sess); + +/* rtrs-clt-stats.c */ + +int rtrs_clt_init_stats(struct rtrs_clt_stats *stats); + +void rtrs_clt_inc_failover_cnt(struct rtrs_clt_stats *s); + +void rtrs_clt_update_wc_stats(struct rtrs_clt_con *con); +void rtrs_clt_update_all_stats(struct rtrs_clt_io_req *req, int dir); + +int rtrs_clt_reset_rdma_lat_distr_stats(struct rtrs_clt_stats *stats, + bool enable); +ssize_t rtrs_clt_stats_rdma_lat_distr_to_str(struct rtrs_clt_stats *stats, + char *page, size_t len); +int rtrs_clt_reset_cpu_migr_stats(struct rtrs_clt_stats *stats, bool enable); +int rtrs_clt_stats_migration_cnt_to_str(struct rtrs_clt_stats *stats, char *buf, + size_t len); +int rtrs_clt_reset_reconnects_stat(struct rtrs_clt_stats *stats, bool enable); +int rtrs_clt_stats_reconnects_to_str(struct rtrs_clt_stats *stats, char *buf, + size_t len); +int rtrs_clt_reset_wc_comp_stats(struct rtrs_clt_stats *stats, bool enable); +int rtrs_clt_stats_wc_completion_to_str(struct rtrs_clt_stats *stats, char *buf, + size_t len); +int rtrs_clt_reset_rdma_stats(struct rtrs_clt_stats *stats, bool enable); +ssize_t rtrs_clt_stats_rdma_to_str(struct rtrs_clt_stats *stats, + char *page, size_t len); +int rtrs_clt_reset_all_stats(struct rtrs_clt_stats *stats, bool enable); +ssize_t rtrs_clt_reset_all_help(struct rtrs_clt_stats *stats, + char *page, size_t len); + +/* rtrs-clt-sysfs.c */ + +int rtrs_clt_create_sysfs_root_files(struct rtrs_clt *clt); +void rtrs_clt_destroy_sysfs_root_folders(struct rtrs_clt *clt); +void rtrs_clt_destroy_sysfs_root_files(struct rtrs_clt *clt); + +int rtrs_clt_create_sess_files(struct rtrs_clt_sess *sess); +void rtrs_clt_destroy_sess_files(struct rtrs_clt_sess *sess, + const struct attribute *sysfs_self); + +#endif /* RTRS_CLT_H */ diff --git a/drivers/infiniband/ulp/rtrs/rtrs-log.h b/drivers/infiniband/ulp/rtrs/rtrs-log.h new file mode 100644 index 000000000..53c785b99 --- /dev/null +++ b/drivers/infiniband/ulp/rtrs/rtrs-log.h @@ -0,0 +1,28 @@ +/* SPDX-License-Identifier: GPL-2.0-or-later */ +/* + * RDMA Transport Layer + * + * Copyright (c) 2014 - 2018 ProfitBricks GmbH. All rights reserved. + * Copyright (c) 2018 - 2019 1&1 IONOS Cloud GmbH. All rights reserved. + * Copyright (c) 2019 - 2020 1&1 IONOS SE. All rights reserved. + */ +#ifndef RTRS_LOG_H +#define RTRS_LOG_H + +#define rtrs_log(fn, obj, fmt, ...) \ + fn("<%s>: " fmt, obj->sessname, ##__VA_ARGS__) + +#define rtrs_err(obj, fmt, ...) \ + rtrs_log(pr_err, obj, fmt, ##__VA_ARGS__) +#define rtrs_err_rl(obj, fmt, ...) \ + rtrs_log(pr_err_ratelimited, obj, fmt, ##__VA_ARGS__) +#define rtrs_wrn(obj, fmt, ...) \ + rtrs_log(pr_warn, obj, fmt, ##__VA_ARGS__) +#define rtrs_wrn_rl(obj, fmt, ...) \ + rtrs_log(pr_warn_ratelimited, obj, fmt, ##__VA_ARGS__) +#define rtrs_info(obj, fmt, ...) \ + rtrs_log(pr_info, obj, fmt, ##__VA_ARGS__) +#define rtrs_info_rl(obj, fmt, ...) \ + rtrs_log(pr_info_ratelimited, obj, fmt, ##__VA_ARGS__) + +#endif /* RTRS_LOG_H */ diff --git a/drivers/infiniband/ulp/rtrs/rtrs-pri.h b/drivers/infiniband/ulp/rtrs/rtrs-pri.h new file mode 100644 index 000000000..c5ca123d5 --- /dev/null +++ b/drivers/infiniband/ulp/rtrs/rtrs-pri.h @@ -0,0 +1,405 @@ +/* SPDX-License-Identifier: GPL-2.0-or-later */ +/* + * RDMA Transport Layer + * + * Copyright (c) 2014 - 2018 ProfitBricks GmbH. All rights reserved. + * Copyright (c) 2018 - 2019 1&1 IONOS Cloud GmbH. All rights reserved. + * Copyright (c) 2019 - 2020 1&1 IONOS SE. All rights reserved. + */ + +#ifndef RTRS_PRI_H +#define RTRS_PRI_H + +#include <linux/uuid.h> +#include <rdma/rdma_cm.h> +#include <rdma/ib_verbs.h> +#include <rdma/ib.h> + +#include "rtrs.h" + +#define RTRS_PROTO_VER_MAJOR 2 +#define RTRS_PROTO_VER_MINOR 0 + +#define RTRS_PROTO_VER_STRING __stringify(RTRS_PROTO_VER_MAJOR) "." \ + __stringify(RTRS_PROTO_VER_MINOR) + +/* + * Max IB immediate data size is 2^28 (MAX_IMM_PAYL_BITS) + * and the minimum chunk size is 4096 (2^12). + * So the maximum sess_queue_depth is 65536 (2^16) in theory. + * But mempool_create, create_qp and ib_post_send fail with + * "cannot allocate memory" error if sess_queue_depth is too big. + * Therefore the pratical max value of sess_queue_depth is + * somewhere between 1 and 65534 and it depends on the system. + */ +#define MAX_SESS_QUEUE_DEPTH 65535 + +enum rtrs_imm_const { + MAX_IMM_TYPE_BITS = 4, + MAX_IMM_TYPE_MASK = ((1 << MAX_IMM_TYPE_BITS) - 1), + MAX_IMM_PAYL_BITS = 28, + MAX_IMM_PAYL_MASK = ((1 << MAX_IMM_PAYL_BITS) - 1), +}; + +enum rtrs_imm_type { + RTRS_IO_REQ_IMM = 0, /* client to server */ + RTRS_IO_RSP_IMM = 1, /* server to client */ + RTRS_IO_RSP_W_INV_IMM = 2, /* server to client */ + + RTRS_HB_MSG_IMM = 8, /* HB: HeartBeat */ + RTRS_HB_ACK_IMM = 9, + + RTRS_LAST_IMM, +}; + +enum { + SERVICE_CON_QUEUE_DEPTH = 512, + + MAX_PATHS_NUM = 128, + + MIN_CHUNK_SIZE = 8192, + + RTRS_HB_INTERVAL_MS = 5000, + RTRS_HB_MISSED_MAX = 5, + + RTRS_MAGIC = 0x1BBD, + RTRS_PROTO_VER = (RTRS_PROTO_VER_MAJOR << 8) | RTRS_PROTO_VER_MINOR, +}; + +struct rtrs_ib_dev; + +struct rtrs_rdma_dev_pd_ops { + struct rtrs_ib_dev *(*alloc)(void); + void (*free)(struct rtrs_ib_dev *dev); + int (*init)(struct rtrs_ib_dev *dev); + void (*deinit)(struct rtrs_ib_dev *dev); +}; + +struct rtrs_rdma_dev_pd { + struct mutex mutex; + struct list_head list; + enum ib_pd_flags pd_flags; + const struct rtrs_rdma_dev_pd_ops *ops; +}; + +struct rtrs_ib_dev { + struct ib_device *ib_dev; + struct ib_pd *ib_pd; + struct kref ref; + struct list_head entry; + struct rtrs_rdma_dev_pd *pool; +}; + +struct rtrs_con { + struct rtrs_sess *sess; + struct ib_qp *qp; + struct ib_cq *cq; + struct rdma_cm_id *cm_id; + unsigned int cid; +}; + +struct rtrs_sess { + struct list_head entry; + struct sockaddr_storage dst_addr; + struct sockaddr_storage src_addr; + char sessname[NAME_MAX]; + uuid_t uuid; + struct rtrs_con **con; + unsigned int con_num; + unsigned int recon_cnt; + struct rtrs_ib_dev *dev; + int dev_ref; + struct ib_cqe *hb_cqe; + void (*hb_err_handler)(struct rtrs_con *con); + struct workqueue_struct *hb_wq; + struct delayed_work hb_dwork; + unsigned int hb_interval_ms; + unsigned int hb_missed_cnt; + unsigned int hb_missed_max; +}; + +/* rtrs information unit */ +struct rtrs_iu { + struct ib_cqe cqe; + dma_addr_t dma_addr; + void *buf; + size_t size; + enum dma_data_direction direction; +}; + +/** + * enum rtrs_msg_types - RTRS message types, see also rtrs/README + * @RTRS_MSG_INFO_REQ: Client additional info request to the server + * @RTRS_MSG_INFO_RSP: Server additional info response to the client + * @RTRS_MSG_WRITE: Client writes data per RDMA to server + * @RTRS_MSG_READ: Client requests data transfer from server + * @RTRS_MSG_RKEY_RSP: Server refreshed rkey for rbuf + */ +enum rtrs_msg_types { + RTRS_MSG_INFO_REQ, + RTRS_MSG_INFO_RSP, + RTRS_MSG_WRITE, + RTRS_MSG_READ, + RTRS_MSG_RKEY_RSP, +}; + +/** + * enum rtrs_msg_flags - RTRS message flags. + * @RTRS_NEED_INVAL: Send invalidation in response. + * @RTRS_MSG_NEW_RKEY_F: Send refreshed rkey in response. + */ +enum rtrs_msg_flags { + RTRS_MSG_NEED_INVAL_F = 1 << 0, + RTRS_MSG_NEW_RKEY_F = 1 << 1, +}; + +/** + * struct rtrs_sg_desc - RDMA-Buffer entry description + * @addr: Address of RDMA destination buffer + * @key: Authorization rkey to write to the buffer + * @len: Size of the buffer + */ +struct rtrs_sg_desc { + __le64 addr; + __le32 key; + __le32 len; +}; + +/** + * struct rtrs_msg_conn_req - Client connection request to the server + * @magic: RTRS magic + * @version: RTRS protocol version + * @cid: Current connection id + * @cid_num: Number of connections per session + * @recon_cnt: Reconnections counter + * @sess_uuid: UUID of a session (path) + * @paths_uuid: UUID of a group of sessions (paths) + * + * NOTE: max size 56 bytes, see man rdma_connect(). + */ +struct rtrs_msg_conn_req { + /* Is set to 0 by cma.c in case of AF_IB, do not touch that. + * see https://www.spinics.net/lists/linux-rdma/msg22397.html + */ + u8 __cma_version; + /* On sender side that should be set to 0, or cma_save_ip_info() + * extract garbage and will fail. + */ + u8 __ip_version; + __le16 magic; + __le16 version; + __le16 cid; + __le16 cid_num; + __le16 recon_cnt; + uuid_t sess_uuid; + uuid_t paths_uuid; + u8 first_conn : 1; + u8 reserved_bits : 7; + u8 reserved[11]; +}; + +/** + * struct rtrs_msg_conn_rsp - Server connection response to the client + * @magic: RTRS magic + * @version: RTRS protocol version + * @errno: If rdma_accept() then 0, if rdma_reject() indicates error + * @queue_depth: max inflight messages (queue-depth) in this session + * @max_io_size: max io size server supports + * @max_hdr_size: max msg header size server supports + * + * NOTE: size is 56 bytes, max possible is 136 bytes, see man rdma_accept(). + */ +struct rtrs_msg_conn_rsp { + __le16 magic; + __le16 version; + __le16 errno; + __le16 queue_depth; + __le32 max_io_size; + __le32 max_hdr_size; + __le32 flags; + u8 reserved[36]; +}; + +/** + * struct rtrs_msg_info_req + * @type: @RTRS_MSG_INFO_REQ + * @sessname: Session name chosen by client + */ +struct rtrs_msg_info_req { + __le16 type; + u8 sessname[NAME_MAX]; + u8 reserved[15]; +}; + +/** + * struct rtrs_msg_info_rsp + * @type: @RTRS_MSG_INFO_RSP + * @sg_cnt: Number of @desc entries + * @desc: RDMA buffers where the client can write to server + */ +struct rtrs_msg_info_rsp { + __le16 type; + __le16 sg_cnt; + u8 reserved[4]; + struct rtrs_sg_desc desc[]; +}; + +/** + * struct rtrs_msg_rkey_rsp + * @type: @RTRS_MSG_RKEY_RSP + * @buf_id: RDMA buf_id of the new rkey + * @rkey: new remote key for RDMA buffers id from server + */ +struct rtrs_msg_rkey_rsp { + __le16 type; + __le16 buf_id; + __le32 rkey; +}; + +/** + * struct rtrs_msg_rdma_read - RDMA data transfer request from client + * @type: always @RTRS_MSG_READ + * @usr_len: length of user payload + * @sg_cnt: number of @desc entries + * @desc: RDMA buffers where the server can write the result to + */ +struct rtrs_msg_rdma_read { + __le16 type; + __le16 usr_len; + __le16 flags; + __le16 sg_cnt; + struct rtrs_sg_desc desc[]; +}; + +/** + * struct_msg_rdma_write - Message transferred to server with RDMA-Write + * @type: always @RTRS_MSG_WRITE + * @usr_len: length of user payload + */ +struct rtrs_msg_rdma_write { + __le16 type; + __le16 usr_len; +}; + +/** + * struct_msg_rdma_hdr - header for read or write request + * @type: @RTRS_MSG_WRITE | @RTRS_MSG_READ + */ +struct rtrs_msg_rdma_hdr { + __le16 type; +}; + +/* rtrs.c */ + +struct rtrs_iu *rtrs_iu_alloc(u32 queue_size, size_t size, gfp_t t, + struct ib_device *dev, enum dma_data_direction, + void (*done)(struct ib_cq *cq, struct ib_wc *wc)); +void rtrs_iu_free(struct rtrs_iu *iu, struct ib_device *dev, u32 queue_size); +int rtrs_iu_post_recv(struct rtrs_con *con, struct rtrs_iu *iu); +int rtrs_iu_post_send(struct rtrs_con *con, struct rtrs_iu *iu, size_t size, + struct ib_send_wr *head); +int rtrs_iu_post_rdma_write_imm(struct rtrs_con *con, struct rtrs_iu *iu, + struct ib_sge *sge, unsigned int num_sge, + u32 rkey, u64 rdma_addr, u32 imm_data, + enum ib_send_flags flags, + struct ib_send_wr *head); + +int rtrs_post_recv_empty(struct rtrs_con *con, struct ib_cqe *cqe); +int rtrs_post_rdma_write_imm_empty(struct rtrs_con *con, struct ib_cqe *cqe, + u32 imm_data, enum ib_send_flags flags, + struct ib_send_wr *head); + +int rtrs_cq_qp_create(struct rtrs_sess *rtrs_sess, struct rtrs_con *con, + u32 max_send_sge, int cq_vector, int cq_size, + u32 max_send_wr, u32 max_recv_wr, + enum ib_poll_context poll_ctx); +void rtrs_cq_qp_destroy(struct rtrs_con *con); + +void rtrs_init_hb(struct rtrs_sess *sess, struct ib_cqe *cqe, + unsigned int interval_ms, unsigned int missed_max, + void (*err_handler)(struct rtrs_con *con), + struct workqueue_struct *wq); +void rtrs_start_hb(struct rtrs_sess *sess); +void rtrs_stop_hb(struct rtrs_sess *sess); +void rtrs_send_hb_ack(struct rtrs_sess *sess); + +void rtrs_rdma_dev_pd_init(enum ib_pd_flags pd_flags, + struct rtrs_rdma_dev_pd *pool); +void rtrs_rdma_dev_pd_deinit(struct rtrs_rdma_dev_pd *pool); + +struct rtrs_ib_dev *rtrs_ib_dev_find_or_add(struct ib_device *ib_dev, + struct rtrs_rdma_dev_pd *pool); +int rtrs_ib_dev_put(struct rtrs_ib_dev *dev); + +static inline u32 rtrs_to_imm(u32 type, u32 payload) +{ + BUILD_BUG_ON(MAX_IMM_PAYL_BITS + MAX_IMM_TYPE_BITS != 32); + BUILD_BUG_ON(RTRS_LAST_IMM > (1<<MAX_IMM_TYPE_BITS)); + return ((type & MAX_IMM_TYPE_MASK) << MAX_IMM_PAYL_BITS) | + (payload & MAX_IMM_PAYL_MASK); +} + +static inline void rtrs_from_imm(u32 imm, u32 *type, u32 *payload) +{ + *payload = imm & MAX_IMM_PAYL_MASK; + *type = imm >> MAX_IMM_PAYL_BITS; +} + +static inline u32 rtrs_to_io_req_imm(u32 addr) +{ + return rtrs_to_imm(RTRS_IO_REQ_IMM, addr); +} + +static inline u32 rtrs_to_io_rsp_imm(u32 msg_id, int errno, bool w_inval) +{ + enum rtrs_imm_type type; + u32 payload; + + /* 9 bits for errno, 19 bits for msg_id */ + payload = (abs(errno) & 0x1ff) << 19 | (msg_id & 0x7ffff); + type = w_inval ? RTRS_IO_RSP_W_INV_IMM : RTRS_IO_RSP_IMM; + + return rtrs_to_imm(type, payload); +} + +static inline void rtrs_from_io_rsp_imm(u32 payload, u32 *msg_id, int *errno) +{ + /* 9 bits for errno, 19 bits for msg_id */ + *msg_id = payload & 0x7ffff; + *errno = -(int)((payload >> 19) & 0x1ff); +} + +#define STAT_STORE_FUNC(type, set_value, reset) \ +static ssize_t set_value##_store(struct kobject *kobj, \ + struct kobj_attribute *attr, \ + const char *buf, size_t count) \ +{ \ + int ret = -EINVAL; \ + type *stats = container_of(kobj, type, kobj_stats); \ + \ + if (sysfs_streq(buf, "1")) \ + ret = reset(stats, true); \ + else if (sysfs_streq(buf, "0")) \ + ret = reset(stats, false); \ + if (ret) \ + return ret; \ + \ + return count; \ +} + +#define STAT_SHOW_FUNC(type, get_value, print) \ +static ssize_t get_value##_show(struct kobject *kobj, \ + struct kobj_attribute *attr, \ + char *page) \ +{ \ + type *stats = container_of(kobj, type, kobj_stats); \ + \ + return print(stats, page, PAGE_SIZE); \ +} + +#define STAT_ATTR(type, stat, print, reset) \ +STAT_STORE_FUNC(type, stat, reset) \ +STAT_SHOW_FUNC(type, stat, print) \ +static struct kobj_attribute stat##_attr = __ATTR_RW(stat) + +#endif /* RTRS_PRI_H */ diff --git a/drivers/infiniband/ulp/rtrs/rtrs-srv-stats.c b/drivers/infiniband/ulp/rtrs/rtrs-srv-stats.c new file mode 100644 index 000000000..e102b1368 --- /dev/null +++ b/drivers/infiniband/ulp/rtrs/rtrs-srv-stats.c @@ -0,0 +1,38 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * RDMA Transport Layer + * + * Copyright (c) 2014 - 2018 ProfitBricks GmbH. All rights reserved. + * Copyright (c) 2018 - 2019 1&1 IONOS Cloud GmbH. All rights reserved. + * Copyright (c) 2019 - 2020 1&1 IONOS SE. All rights reserved. + */ +#undef pr_fmt +#define pr_fmt(fmt) KBUILD_MODNAME " L" __stringify(__LINE__) ": " fmt + +#include "rtrs-srv.h" + +int rtrs_srv_reset_rdma_stats(struct rtrs_srv_stats *stats, bool enable) +{ + if (enable) { + struct rtrs_srv_stats_rdma_stats *r = &stats->rdma_stats; + + memset(r, 0, sizeof(*r)); + return 0; + } + + return -EINVAL; +} + +ssize_t rtrs_srv_stats_rdma_to_str(struct rtrs_srv_stats *stats, + char *page, size_t len) +{ + struct rtrs_srv_stats_rdma_stats *r = &stats->rdma_stats; + struct rtrs_srv_sess *sess = stats->sess; + + return scnprintf(page, len, "%lld %lld %lld %lld %u\n", + (s64)atomic64_read(&r->dir[READ].cnt), + (s64)atomic64_read(&r->dir[READ].size_total), + (s64)atomic64_read(&r->dir[WRITE].cnt), + (s64)atomic64_read(&r->dir[WRITE].size_total), + atomic_read(&sess->ids_inflight)); +} diff --git a/drivers/infiniband/ulp/rtrs/rtrs-srv-sysfs.c b/drivers/infiniband/ulp/rtrs/rtrs-srv-sysfs.c new file mode 100644 index 000000000..7c75e1459 --- /dev/null +++ b/drivers/infiniband/ulp/rtrs/rtrs-srv-sysfs.c @@ -0,0 +1,319 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * RDMA Transport Layer + * + * Copyright (c) 2014 - 2018 ProfitBricks GmbH. All rights reserved. + * Copyright (c) 2018 - 2019 1&1 IONOS Cloud GmbH. All rights reserved. + * Copyright (c) 2019 - 2020 1&1 IONOS SE. All rights reserved. + */ +#undef pr_fmt +#define pr_fmt(fmt) KBUILD_MODNAME " L" __stringify(__LINE__) ": " fmt + +#include "rtrs-pri.h" +#include "rtrs-srv.h" +#include "rtrs-log.h" + +static void rtrs_srv_release(struct kobject *kobj) +{ + struct rtrs_srv_sess *sess; + + sess = container_of(kobj, struct rtrs_srv_sess, kobj); + kfree(sess); +} + +static struct kobj_type ktype = { + .sysfs_ops = &kobj_sysfs_ops, + .release = rtrs_srv_release, +}; + +static ssize_t rtrs_srv_disconnect_show(struct kobject *kobj, + struct kobj_attribute *attr, + char *page) +{ + return scnprintf(page, PAGE_SIZE, "Usage: echo 1 > %s\n", + attr->attr.name); +} + +static ssize_t rtrs_srv_disconnect_store(struct kobject *kobj, + struct kobj_attribute *attr, + const char *buf, size_t count) +{ + struct rtrs_srv_sess *sess; + struct rtrs_sess *s; + char str[MAXHOSTNAMELEN]; + + sess = container_of(kobj, struct rtrs_srv_sess, kobj); + s = &sess->s; + if (!sysfs_streq(buf, "1")) { + rtrs_err(s, "%s: invalid value: '%s'\n", + attr->attr.name, buf); + return -EINVAL; + } + + sockaddr_to_str((struct sockaddr *)&sess->s.dst_addr, str, sizeof(str)); + + rtrs_info(s, "disconnect for path %s requested\n", str); + /* first remove sysfs itself to avoid deadlock */ + sysfs_remove_file_self(&sess->kobj, &attr->attr); + close_sess(sess); + + return count; +} + +static struct kobj_attribute rtrs_srv_disconnect_attr = + __ATTR(disconnect, 0644, + rtrs_srv_disconnect_show, rtrs_srv_disconnect_store); + +static ssize_t rtrs_srv_hca_port_show(struct kobject *kobj, + struct kobj_attribute *attr, + char *page) +{ + struct rtrs_srv_sess *sess; + struct rtrs_con *usr_con; + + sess = container_of(kobj, typeof(*sess), kobj); + usr_con = sess->s.con[0]; + + return scnprintf(page, PAGE_SIZE, "%u\n", + usr_con->cm_id->port_num); +} + +static struct kobj_attribute rtrs_srv_hca_port_attr = + __ATTR(hca_port, 0444, rtrs_srv_hca_port_show, NULL); + +static ssize_t rtrs_srv_hca_name_show(struct kobject *kobj, + struct kobj_attribute *attr, + char *page) +{ + struct rtrs_srv_sess *sess; + + sess = container_of(kobj, struct rtrs_srv_sess, kobj); + + return scnprintf(page, PAGE_SIZE, "%s\n", + sess->s.dev->ib_dev->name); +} + +static struct kobj_attribute rtrs_srv_hca_name_attr = + __ATTR(hca_name, 0444, rtrs_srv_hca_name_show, NULL); + +static ssize_t rtrs_srv_src_addr_show(struct kobject *kobj, + struct kobj_attribute *attr, + char *page) +{ + struct rtrs_srv_sess *sess; + int cnt; + + sess = container_of(kobj, struct rtrs_srv_sess, kobj); + cnt = sockaddr_to_str((struct sockaddr *)&sess->s.dst_addr, + page, PAGE_SIZE); + return cnt + scnprintf(page + cnt, PAGE_SIZE - cnt, "\n"); +} + +static struct kobj_attribute rtrs_srv_src_addr_attr = + __ATTR(src_addr, 0444, rtrs_srv_src_addr_show, NULL); + +static ssize_t rtrs_srv_dst_addr_show(struct kobject *kobj, + struct kobj_attribute *attr, + char *page) +{ + struct rtrs_srv_sess *sess; + int cnt; + + sess = container_of(kobj, struct rtrs_srv_sess, kobj); + cnt = sockaddr_to_str((struct sockaddr *)&sess->s.src_addr, + page, PAGE_SIZE); + return cnt + scnprintf(page + cnt, PAGE_SIZE - cnt, "\n"); +} + +static struct kobj_attribute rtrs_srv_dst_addr_attr = + __ATTR(dst_addr, 0444, rtrs_srv_dst_addr_show, NULL); + +static struct attribute *rtrs_srv_sess_attrs[] = { + &rtrs_srv_hca_name_attr.attr, + &rtrs_srv_hca_port_attr.attr, + &rtrs_srv_src_addr_attr.attr, + &rtrs_srv_dst_addr_attr.attr, + &rtrs_srv_disconnect_attr.attr, + NULL, +}; + +static const struct attribute_group rtrs_srv_sess_attr_group = { + .attrs = rtrs_srv_sess_attrs, +}; + +STAT_ATTR(struct rtrs_srv_stats, rdma, + rtrs_srv_stats_rdma_to_str, + rtrs_srv_reset_rdma_stats); + +static struct attribute *rtrs_srv_stats_attrs[] = { + &rdma_attr.attr, + NULL, +}; + +static const struct attribute_group rtrs_srv_stats_attr_group = { + .attrs = rtrs_srv_stats_attrs, +}; + +static int rtrs_srv_create_once_sysfs_root_folders(struct rtrs_srv_sess *sess) +{ + struct rtrs_srv *srv = sess->srv; + int err = 0; + + mutex_lock(&srv->paths_mutex); + if (srv->dev_ref++) { + /* + * Device needs to be registered only on the first session + */ + goto unlock; + } + srv->dev.class = rtrs_dev_class; + err = dev_set_name(&srv->dev, "%s", sess->s.sessname); + if (err) + goto unlock; + + /* + * Suppress user space notification until + * sysfs files are created + */ + dev_set_uevent_suppress(&srv->dev, true); + err = device_add(&srv->dev); + if (err) { + pr_err("device_add(): %d\n", err); + goto put; + } + srv->kobj_paths = kobject_create_and_add("paths", &srv->dev.kobj); + if (!srv->kobj_paths) { + err = -ENOMEM; + pr_err("kobject_create_and_add(): %d\n", err); + device_del(&srv->dev); + put_device(&srv->dev); + goto unlock; + } + dev_set_uevent_suppress(&srv->dev, false); + kobject_uevent(&srv->dev.kobj, KOBJ_ADD); + goto unlock; + +put: + put_device(&srv->dev); +unlock: + mutex_unlock(&srv->paths_mutex); + + return err; +} + +static void +rtrs_srv_destroy_once_sysfs_root_folders(struct rtrs_srv_sess *sess) +{ + struct rtrs_srv *srv = sess->srv; + + mutex_lock(&srv->paths_mutex); + if (!--srv->dev_ref) { + kobject_del(srv->kobj_paths); + kobject_put(srv->kobj_paths); + mutex_unlock(&srv->paths_mutex); + device_del(&srv->dev); + put_device(&srv->dev); + } else { + put_device(&srv->dev); + mutex_unlock(&srv->paths_mutex); + } +} + +static void rtrs_srv_sess_stats_release(struct kobject *kobj) +{ + struct rtrs_srv_stats *stats; + + stats = container_of(kobj, struct rtrs_srv_stats, kobj_stats); + + kfree(stats); +} + +static struct kobj_type ktype_stats = { + .sysfs_ops = &kobj_sysfs_ops, + .release = rtrs_srv_sess_stats_release, +}; + +static int rtrs_srv_create_stats_files(struct rtrs_srv_sess *sess) +{ + int err; + struct rtrs_sess *s = &sess->s; + + err = kobject_init_and_add(&sess->stats->kobj_stats, &ktype_stats, + &sess->kobj, "stats"); + if (err) { + rtrs_err(s, "kobject_init_and_add(): %d\n", err); + kobject_put(&sess->stats->kobj_stats); + return err; + } + err = sysfs_create_group(&sess->stats->kobj_stats, + &rtrs_srv_stats_attr_group); + if (err) { + rtrs_err(s, "sysfs_create_group(): %d\n", err); + goto err; + } + + return 0; + +err: + kobject_del(&sess->stats->kobj_stats); + kobject_put(&sess->stats->kobj_stats); + + return err; +} + +int rtrs_srv_create_sess_files(struct rtrs_srv_sess *sess) +{ + struct rtrs_srv *srv = sess->srv; + struct rtrs_sess *s = &sess->s; + char str[NAME_MAX]; + int err, cnt; + + cnt = sockaddr_to_str((struct sockaddr *)&sess->s.dst_addr, + str, sizeof(str)); + cnt += scnprintf(str + cnt, sizeof(str) - cnt, "@"); + sockaddr_to_str((struct sockaddr *)&sess->s.src_addr, + str + cnt, sizeof(str) - cnt); + + err = rtrs_srv_create_once_sysfs_root_folders(sess); + if (err) + return err; + + err = kobject_init_and_add(&sess->kobj, &ktype, srv->kobj_paths, + "%s", str); + if (err) { + rtrs_err(s, "kobject_init_and_add(): %d\n", err); + goto destroy_root; + } + err = sysfs_create_group(&sess->kobj, &rtrs_srv_sess_attr_group); + if (err) { + rtrs_err(s, "sysfs_create_group(): %d\n", err); + goto put_kobj; + } + err = rtrs_srv_create_stats_files(sess); + if (err) + goto remove_group; + + return 0; + +remove_group: + sysfs_remove_group(&sess->kobj, &rtrs_srv_sess_attr_group); +put_kobj: + kobject_del(&sess->kobj); +destroy_root: + kobject_put(&sess->kobj); + rtrs_srv_destroy_once_sysfs_root_folders(sess); + + return err; +} + +void rtrs_srv_destroy_sess_files(struct rtrs_srv_sess *sess) +{ + if (sess->kobj.state_in_sysfs) { + kobject_del(&sess->stats->kobj_stats); + kobject_put(&sess->stats->kobj_stats); + sysfs_remove_group(&sess->kobj, &rtrs_srv_sess_attr_group); + kobject_put(&sess->kobj); + + rtrs_srv_destroy_once_sysfs_root_folders(sess); + } +} diff --git a/drivers/infiniband/ulp/rtrs/rtrs-srv.c b/drivers/infiniband/ulp/rtrs/rtrs-srv.c new file mode 100644 index 000000000..b152a742c --- /dev/null +++ b/drivers/infiniband/ulp/rtrs/rtrs-srv.c @@ -0,0 +1,2267 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * RDMA Transport Layer + * + * Copyright (c) 2014 - 2018 ProfitBricks GmbH. All rights reserved. + * Copyright (c) 2018 - 2019 1&1 IONOS Cloud GmbH. All rights reserved. + * Copyright (c) 2019 - 2020 1&1 IONOS SE. All rights reserved. + */ + +#undef pr_fmt +#define pr_fmt(fmt) KBUILD_MODNAME " L" __stringify(__LINE__) ": " fmt + +#include <linux/module.h> +#include <linux/mempool.h> + +#include "rtrs-srv.h" +#include "rtrs-log.h" +#include <rdma/ib_cm.h> +#include <rdma/ib_verbs.h> + +MODULE_DESCRIPTION("RDMA Transport Server"); +MODULE_LICENSE("GPL"); + +/* Must be power of 2, see mask from mr->page_size in ib_sg_to_pages() */ +#define DEFAULT_MAX_CHUNK_SIZE (128 << 10) +#define DEFAULT_SESS_QUEUE_DEPTH 512 +#define MAX_HDR_SIZE PAGE_SIZE + +/* We guarantee to serve 10 paths at least */ +#define CHUNK_POOL_SZ 10 + +static struct rtrs_rdma_dev_pd dev_pd; +static mempool_t *chunk_pool; +struct class *rtrs_dev_class; +static struct rtrs_srv_ib_ctx ib_ctx; + +static int __read_mostly max_chunk_size = DEFAULT_MAX_CHUNK_SIZE; +static int __read_mostly sess_queue_depth = DEFAULT_SESS_QUEUE_DEPTH; + +static bool always_invalidate = true; +module_param(always_invalidate, bool, 0444); +MODULE_PARM_DESC(always_invalidate, + "Invalidate memory registration for contiguous memory regions before accessing."); + +module_param_named(max_chunk_size, max_chunk_size, int, 0444); +MODULE_PARM_DESC(max_chunk_size, + "Max size for each IO request, when change the unit is in byte (default: " + __stringify(DEFAULT_MAX_CHUNK_SIZE) "KB)"); + +module_param_named(sess_queue_depth, sess_queue_depth, int, 0444); +MODULE_PARM_DESC(sess_queue_depth, + "Number of buffers for pending I/O requests to allocate per session. Maximum: " + __stringify(MAX_SESS_QUEUE_DEPTH) " (default: " + __stringify(DEFAULT_SESS_QUEUE_DEPTH) ")"); + +static cpumask_t cq_affinity_mask = { CPU_BITS_ALL }; + +static struct workqueue_struct *rtrs_wq; + +static inline struct rtrs_srv_con *to_srv_con(struct rtrs_con *c) +{ + return container_of(c, struct rtrs_srv_con, c); +} + +static inline struct rtrs_srv_sess *to_srv_sess(struct rtrs_sess *s) +{ + return container_of(s, struct rtrs_srv_sess, s); +} + +static bool __rtrs_srv_change_state(struct rtrs_srv_sess *sess, + enum rtrs_srv_state new_state) +{ + enum rtrs_srv_state old_state; + bool changed = false; + + lockdep_assert_held(&sess->state_lock); + old_state = sess->state; + switch (new_state) { + case RTRS_SRV_CONNECTED: + switch (old_state) { + case RTRS_SRV_CONNECTING: + changed = true; + fallthrough; + default: + break; + } + break; + case RTRS_SRV_CLOSING: + switch (old_state) { + case RTRS_SRV_CONNECTING: + case RTRS_SRV_CONNECTED: + changed = true; + fallthrough; + default: + break; + } + break; + case RTRS_SRV_CLOSED: + switch (old_state) { + case RTRS_SRV_CLOSING: + changed = true; + fallthrough; + default: + break; + } + break; + default: + break; + } + if (changed) + sess->state = new_state; + + return changed; +} + +static bool rtrs_srv_change_state_get_old(struct rtrs_srv_sess *sess, + enum rtrs_srv_state new_state, + enum rtrs_srv_state *old_state) +{ + bool changed; + + spin_lock_irq(&sess->state_lock); + *old_state = sess->state; + changed = __rtrs_srv_change_state(sess, new_state); + spin_unlock_irq(&sess->state_lock); + + return changed; +} + +static bool rtrs_srv_change_state(struct rtrs_srv_sess *sess, + enum rtrs_srv_state new_state) +{ + enum rtrs_srv_state old_state; + + return rtrs_srv_change_state_get_old(sess, new_state, &old_state); +} + +static void free_id(struct rtrs_srv_op *id) +{ + if (!id) + return; + kfree(id); +} + +static void rtrs_srv_free_ops_ids(struct rtrs_srv_sess *sess) +{ + struct rtrs_srv *srv = sess->srv; + int i; + + WARN_ON(atomic_read(&sess->ids_inflight)); + if (sess->ops_ids) { + for (i = 0; i < srv->queue_depth; i++) + free_id(sess->ops_ids[i]); + kfree(sess->ops_ids); + sess->ops_ids = NULL; + } +} + +static void rtrs_srv_rdma_done(struct ib_cq *cq, struct ib_wc *wc); + +static struct ib_cqe io_comp_cqe = { + .done = rtrs_srv_rdma_done +}; + +static int rtrs_srv_alloc_ops_ids(struct rtrs_srv_sess *sess) +{ + struct rtrs_srv *srv = sess->srv; + struct rtrs_srv_op *id; + int i; + + sess->ops_ids = kcalloc(srv->queue_depth, sizeof(*sess->ops_ids), + GFP_KERNEL); + if (!sess->ops_ids) + goto err; + + for (i = 0; i < srv->queue_depth; ++i) { + id = kzalloc(sizeof(*id), GFP_KERNEL); + if (!id) + goto err; + + sess->ops_ids[i] = id; + } + init_waitqueue_head(&sess->ids_waitq); + atomic_set(&sess->ids_inflight, 0); + + return 0; + +err: + rtrs_srv_free_ops_ids(sess); + return -ENOMEM; +} + +static inline void rtrs_srv_get_ops_ids(struct rtrs_srv_sess *sess) +{ + atomic_inc(&sess->ids_inflight); +} + +static inline void rtrs_srv_put_ops_ids(struct rtrs_srv_sess *sess) +{ + if (atomic_dec_and_test(&sess->ids_inflight)) + wake_up(&sess->ids_waitq); +} + +static void rtrs_srv_wait_ops_ids(struct rtrs_srv_sess *sess) +{ + wait_event(sess->ids_waitq, !atomic_read(&sess->ids_inflight)); +} + + +static void rtrs_srv_reg_mr_done(struct ib_cq *cq, struct ib_wc *wc) +{ + struct rtrs_srv_con *con = cq->cq_context; + struct rtrs_sess *s = con->c.sess; + struct rtrs_srv_sess *sess = to_srv_sess(s); + + if (unlikely(wc->status != IB_WC_SUCCESS)) { + rtrs_err(s, "REG MR failed: %s\n", + ib_wc_status_msg(wc->status)); + close_sess(sess); + return; + } +} + +static struct ib_cqe local_reg_cqe = { + .done = rtrs_srv_reg_mr_done +}; + +static int rdma_write_sg(struct rtrs_srv_op *id) +{ + struct rtrs_sess *s = id->con->c.sess; + struct rtrs_srv_sess *sess = to_srv_sess(s); + dma_addr_t dma_addr = sess->dma_addr[id->msg_id]; + struct rtrs_srv_mr *srv_mr; + struct rtrs_srv *srv = sess->srv; + struct ib_send_wr inv_wr; + struct ib_rdma_wr imm_wr; + struct ib_rdma_wr *wr = NULL; + enum ib_send_flags flags; + size_t sg_cnt; + int err, offset; + bool need_inval; + u32 rkey = 0; + struct ib_reg_wr rwr; + struct ib_sge *plist; + struct ib_sge list; + + sg_cnt = le16_to_cpu(id->rd_msg->sg_cnt); + need_inval = le16_to_cpu(id->rd_msg->flags) & RTRS_MSG_NEED_INVAL_F; + if (unlikely(sg_cnt != 1)) + return -EINVAL; + + offset = 0; + + wr = &id->tx_wr; + plist = &id->tx_sg; + plist->addr = dma_addr + offset; + plist->length = le32_to_cpu(id->rd_msg->desc[0].len); + + /* WR will fail with length error + * if this is 0 + */ + if (unlikely(plist->length == 0)) { + rtrs_err(s, "Invalid RDMA-Write sg list length 0\n"); + return -EINVAL; + } + + plist->lkey = sess->s.dev->ib_pd->local_dma_lkey; + offset += plist->length; + + wr->wr.sg_list = plist; + wr->wr.num_sge = 1; + wr->remote_addr = le64_to_cpu(id->rd_msg->desc[0].addr); + wr->rkey = le32_to_cpu(id->rd_msg->desc[0].key); + if (rkey == 0) + rkey = wr->rkey; + else + /* Only one key is actually used */ + WARN_ON_ONCE(rkey != wr->rkey); + + wr->wr.opcode = IB_WR_RDMA_WRITE; + wr->wr.wr_cqe = &io_comp_cqe; + wr->wr.ex.imm_data = 0; + wr->wr.send_flags = 0; + + if (need_inval && always_invalidate) { + wr->wr.next = &rwr.wr; + rwr.wr.next = &inv_wr; + inv_wr.next = &imm_wr.wr; + } else if (always_invalidate) { + wr->wr.next = &rwr.wr; + rwr.wr.next = &imm_wr.wr; + } else if (need_inval) { + wr->wr.next = &inv_wr; + inv_wr.next = &imm_wr.wr; + } else { + wr->wr.next = &imm_wr.wr; + } + /* + * From time to time we have to post signaled sends, + * or send queue will fill up and only QP reset can help. + */ + flags = (atomic_inc_return(&id->con->wr_cnt) % srv->queue_depth) ? + 0 : IB_SEND_SIGNALED; + + if (need_inval) { + inv_wr.sg_list = NULL; + inv_wr.num_sge = 0; + inv_wr.opcode = IB_WR_SEND_WITH_INV; + inv_wr.wr_cqe = &io_comp_cqe; + inv_wr.send_flags = 0; + inv_wr.ex.invalidate_rkey = rkey; + } + + imm_wr.wr.next = NULL; + if (always_invalidate) { + struct rtrs_msg_rkey_rsp *msg; + + srv_mr = &sess->mrs[id->msg_id]; + rwr.wr.opcode = IB_WR_REG_MR; + rwr.wr.wr_cqe = &local_reg_cqe; + rwr.wr.num_sge = 0; + rwr.mr = srv_mr->mr; + rwr.wr.send_flags = 0; + rwr.key = srv_mr->mr->rkey; + rwr.access = (IB_ACCESS_LOCAL_WRITE | + IB_ACCESS_REMOTE_WRITE); + msg = srv_mr->iu->buf; + msg->buf_id = cpu_to_le16(id->msg_id); + msg->type = cpu_to_le16(RTRS_MSG_RKEY_RSP); + msg->rkey = cpu_to_le32(srv_mr->mr->rkey); + + list.addr = srv_mr->iu->dma_addr; + list.length = sizeof(*msg); + list.lkey = sess->s.dev->ib_pd->local_dma_lkey; + imm_wr.wr.sg_list = &list; + imm_wr.wr.num_sge = 1; + imm_wr.wr.opcode = IB_WR_SEND_WITH_IMM; + ib_dma_sync_single_for_device(sess->s.dev->ib_dev, + srv_mr->iu->dma_addr, + srv_mr->iu->size, DMA_TO_DEVICE); + } else { + imm_wr.wr.sg_list = NULL; + imm_wr.wr.num_sge = 0; + imm_wr.wr.opcode = IB_WR_RDMA_WRITE_WITH_IMM; + } + imm_wr.wr.send_flags = flags; + imm_wr.wr.ex.imm_data = cpu_to_be32(rtrs_to_io_rsp_imm(id->msg_id, + 0, need_inval)); + + imm_wr.wr.wr_cqe = &io_comp_cqe; + ib_dma_sync_single_for_device(sess->s.dev->ib_dev, dma_addr, + offset, DMA_BIDIRECTIONAL); + + err = ib_post_send(id->con->c.qp, &id->tx_wr.wr, NULL); + if (unlikely(err)) + rtrs_err(s, + "Posting RDMA-Write-Request to QP failed, err: %d\n", + err); + + return err; +} + +/** + * send_io_resp_imm() - respond to client with empty IMM on failed READ/WRITE + * requests or on successful WRITE request. + * @con: the connection to send back result + * @id: the id associated with the IO + * @errno: the error number of the IO. + * + * Return 0 on success, errno otherwise. + */ +static int send_io_resp_imm(struct rtrs_srv_con *con, struct rtrs_srv_op *id, + int errno) +{ + struct rtrs_sess *s = con->c.sess; + struct rtrs_srv_sess *sess = to_srv_sess(s); + struct ib_send_wr inv_wr, *wr = NULL; + struct ib_rdma_wr imm_wr; + struct ib_reg_wr rwr; + struct rtrs_srv *srv = sess->srv; + struct rtrs_srv_mr *srv_mr; + bool need_inval = false; + enum ib_send_flags flags; + u32 imm; + int err; + + if (id->dir == READ) { + struct rtrs_msg_rdma_read *rd_msg = id->rd_msg; + size_t sg_cnt; + + need_inval = le16_to_cpu(rd_msg->flags) & + RTRS_MSG_NEED_INVAL_F; + sg_cnt = le16_to_cpu(rd_msg->sg_cnt); + + if (need_inval) { + if (likely(sg_cnt)) { + inv_wr.wr_cqe = &io_comp_cqe; + inv_wr.sg_list = NULL; + inv_wr.num_sge = 0; + inv_wr.opcode = IB_WR_SEND_WITH_INV; + inv_wr.send_flags = 0; + /* Only one key is actually used */ + inv_wr.ex.invalidate_rkey = + le32_to_cpu(rd_msg->desc[0].key); + } else { + WARN_ON_ONCE(1); + need_inval = false; + } + } + } + + if (need_inval && always_invalidate) { + wr = &inv_wr; + inv_wr.next = &rwr.wr; + rwr.wr.next = &imm_wr.wr; + } else if (always_invalidate) { + wr = &rwr.wr; + rwr.wr.next = &imm_wr.wr; + } else if (need_inval) { + wr = &inv_wr; + inv_wr.next = &imm_wr.wr; + } else { + wr = &imm_wr.wr; + } + /* + * From time to time we have to post signalled sends, + * or send queue will fill up and only QP reset can help. + */ + flags = (atomic_inc_return(&con->wr_cnt) % srv->queue_depth) ? + 0 : IB_SEND_SIGNALED; + imm = rtrs_to_io_rsp_imm(id->msg_id, errno, need_inval); + imm_wr.wr.next = NULL; + if (always_invalidate) { + struct ib_sge list; + struct rtrs_msg_rkey_rsp *msg; + + srv_mr = &sess->mrs[id->msg_id]; + rwr.wr.next = &imm_wr.wr; + rwr.wr.opcode = IB_WR_REG_MR; + rwr.wr.wr_cqe = &local_reg_cqe; + rwr.wr.num_sge = 0; + rwr.wr.send_flags = 0; + rwr.mr = srv_mr->mr; + rwr.key = srv_mr->mr->rkey; + rwr.access = (IB_ACCESS_LOCAL_WRITE | + IB_ACCESS_REMOTE_WRITE); + msg = srv_mr->iu->buf; + msg->buf_id = cpu_to_le16(id->msg_id); + msg->type = cpu_to_le16(RTRS_MSG_RKEY_RSP); + msg->rkey = cpu_to_le32(srv_mr->mr->rkey); + + list.addr = srv_mr->iu->dma_addr; + list.length = sizeof(*msg); + list.lkey = sess->s.dev->ib_pd->local_dma_lkey; + imm_wr.wr.sg_list = &list; + imm_wr.wr.num_sge = 1; + imm_wr.wr.opcode = IB_WR_SEND_WITH_IMM; + ib_dma_sync_single_for_device(sess->s.dev->ib_dev, + srv_mr->iu->dma_addr, + srv_mr->iu->size, DMA_TO_DEVICE); + } else { + imm_wr.wr.sg_list = NULL; + imm_wr.wr.num_sge = 0; + imm_wr.wr.opcode = IB_WR_RDMA_WRITE_WITH_IMM; + } + imm_wr.wr.send_flags = flags; + imm_wr.wr.wr_cqe = &io_comp_cqe; + + imm_wr.wr.ex.imm_data = cpu_to_be32(imm); + + err = ib_post_send(id->con->c.qp, wr, NULL); + if (unlikely(err)) + rtrs_err_rl(s, "Posting RDMA-Reply to QP failed, err: %d\n", + err); + + return err; +} + +void close_sess(struct rtrs_srv_sess *sess) +{ + enum rtrs_srv_state old_state; + + if (rtrs_srv_change_state_get_old(sess, RTRS_SRV_CLOSING, + &old_state)) + queue_work(rtrs_wq, &sess->close_work); + WARN_ON(sess->state != RTRS_SRV_CLOSING); +} + +static inline const char *rtrs_srv_state_str(enum rtrs_srv_state state) +{ + switch (state) { + case RTRS_SRV_CONNECTING: + return "RTRS_SRV_CONNECTING"; + case RTRS_SRV_CONNECTED: + return "RTRS_SRV_CONNECTED"; + case RTRS_SRV_CLOSING: + return "RTRS_SRV_CLOSING"; + case RTRS_SRV_CLOSED: + return "RTRS_SRV_CLOSED"; + default: + return "UNKNOWN"; + } +} + +/** + * rtrs_srv_resp_rdma() - Finish an RDMA request + * + * @id: Internal RTRS operation identifier + * @status: Response Code sent to the other side for this operation. + * 0 = success, <=0 error + * Context: any + * + * Finish a RDMA operation. A message is sent to the client and the + * corresponding memory areas will be released. + */ +bool rtrs_srv_resp_rdma(struct rtrs_srv_op *id, int status) +{ + struct rtrs_srv_sess *sess; + struct rtrs_srv_con *con; + struct rtrs_sess *s; + int err; + + if (WARN_ON(!id)) + return true; + + con = id->con; + s = con->c.sess; + sess = to_srv_sess(s); + + id->status = status; + + if (unlikely(sess->state != RTRS_SRV_CONNECTED)) { + rtrs_err_rl(s, + "Sending I/O response failed, session is disconnected, sess state %s\n", + rtrs_srv_state_str(sess->state)); + goto out; + } + if (always_invalidate) { + struct rtrs_srv_mr *mr = &sess->mrs[id->msg_id]; + + ib_update_fast_reg_key(mr->mr, ib_inc_rkey(mr->mr->rkey)); + } + if (unlikely(atomic_sub_return(1, + &con->sq_wr_avail) < 0)) { + pr_err("IB send queue full\n"); + atomic_add(1, &con->sq_wr_avail); + spin_lock(&con->rsp_wr_wait_lock); + list_add_tail(&id->wait_list, &con->rsp_wr_wait_list); + spin_unlock(&con->rsp_wr_wait_lock); + return false; + } + + if (status || id->dir == WRITE || !id->rd_msg->sg_cnt) + err = send_io_resp_imm(con, id, status); + else + err = rdma_write_sg(id); + + if (unlikely(err)) { + rtrs_err_rl(s, "IO response failed: %d\n", err); + close_sess(sess); + } +out: + rtrs_srv_put_ops_ids(sess); + return true; +} +EXPORT_SYMBOL(rtrs_srv_resp_rdma); + +/** + * rtrs_srv_set_sess_priv() - Set private pointer in rtrs_srv. + * @srv: Session pointer + * @priv: The private pointer that is associated with the session. + */ +void rtrs_srv_set_sess_priv(struct rtrs_srv *srv, void *priv) +{ + srv->priv = priv; +} +EXPORT_SYMBOL(rtrs_srv_set_sess_priv); + +static void unmap_cont_bufs(struct rtrs_srv_sess *sess) +{ + int i; + + for (i = 0; i < sess->mrs_num; i++) { + struct rtrs_srv_mr *srv_mr; + + srv_mr = &sess->mrs[i]; + rtrs_iu_free(srv_mr->iu, sess->s.dev->ib_dev, 1); + ib_dereg_mr(srv_mr->mr); + ib_dma_unmap_sg(sess->s.dev->ib_dev, srv_mr->sgt.sgl, + srv_mr->sgt.nents, DMA_BIDIRECTIONAL); + sg_free_table(&srv_mr->sgt); + } + kfree(sess->mrs); +} + +static int map_cont_bufs(struct rtrs_srv_sess *sess) +{ + struct rtrs_srv *srv = sess->srv; + struct rtrs_sess *ss = &sess->s; + int i, mri, err, mrs_num; + unsigned int chunk_bits; + int chunks_per_mr = 1; + + /* + * Here we map queue_depth chunks to MR. Firstly we have to + * figure out how many chunks can we map per MR. + */ + if (always_invalidate) { + /* + * in order to do invalidate for each chunks of memory, we needs + * more memory regions. + */ + mrs_num = srv->queue_depth; + } else { + chunks_per_mr = + sess->s.dev->ib_dev->attrs.max_fast_reg_page_list_len; + mrs_num = DIV_ROUND_UP(srv->queue_depth, chunks_per_mr); + chunks_per_mr = DIV_ROUND_UP(srv->queue_depth, mrs_num); + } + + sess->mrs = kcalloc(mrs_num, sizeof(*sess->mrs), GFP_KERNEL); + if (!sess->mrs) + return -ENOMEM; + + sess->mrs_num = mrs_num; + + for (mri = 0; mri < mrs_num; mri++) { + struct rtrs_srv_mr *srv_mr = &sess->mrs[mri]; + struct sg_table *sgt = &srv_mr->sgt; + struct scatterlist *s; + struct ib_mr *mr; + int nr, chunks; + + chunks = chunks_per_mr * mri; + if (!always_invalidate) + chunks_per_mr = min_t(int, chunks_per_mr, + srv->queue_depth - chunks); + + err = sg_alloc_table(sgt, chunks_per_mr, GFP_KERNEL); + if (err) + goto err; + + for_each_sg(sgt->sgl, s, chunks_per_mr, i) + sg_set_page(s, srv->chunks[chunks + i], + max_chunk_size, 0); + + nr = ib_dma_map_sg(sess->s.dev->ib_dev, sgt->sgl, + sgt->nents, DMA_BIDIRECTIONAL); + if (nr < sgt->nents) { + err = nr < 0 ? nr : -EINVAL; + goto free_sg; + } + mr = ib_alloc_mr(sess->s.dev->ib_pd, IB_MR_TYPE_MEM_REG, + sgt->nents); + if (IS_ERR(mr)) { + err = PTR_ERR(mr); + goto unmap_sg; + } + nr = ib_map_mr_sg(mr, sgt->sgl, sgt->nents, + NULL, max_chunk_size); + if (nr < 0 || nr < sgt->nents) { + err = nr < 0 ? nr : -EINVAL; + goto dereg_mr; + } + + if (always_invalidate) { + srv_mr->iu = rtrs_iu_alloc(1, + sizeof(struct rtrs_msg_rkey_rsp), + GFP_KERNEL, sess->s.dev->ib_dev, + DMA_TO_DEVICE, rtrs_srv_rdma_done); + if (!srv_mr->iu) { + err = -ENOMEM; + rtrs_err(ss, "rtrs_iu_alloc(), err: %d\n", err); + goto dereg_mr; + } + } + /* Eventually dma addr for each chunk can be cached */ + for_each_sg(sgt->sgl, s, sgt->orig_nents, i) + sess->dma_addr[chunks + i] = sg_dma_address(s); + + ib_update_fast_reg_key(mr, ib_inc_rkey(mr->rkey)); + srv_mr->mr = mr; + + continue; +err: + while (mri--) { + srv_mr = &sess->mrs[mri]; + sgt = &srv_mr->sgt; + mr = srv_mr->mr; + rtrs_iu_free(srv_mr->iu, sess->s.dev->ib_dev, 1); +dereg_mr: + ib_dereg_mr(mr); +unmap_sg: + ib_dma_unmap_sg(sess->s.dev->ib_dev, sgt->sgl, + sgt->nents, DMA_BIDIRECTIONAL); +free_sg: + sg_free_table(sgt); + } + kfree(sess->mrs); + + return err; + } + + chunk_bits = ilog2(srv->queue_depth - 1) + 1; + sess->mem_bits = (MAX_IMM_PAYL_BITS - chunk_bits); + + return 0; +} + +static void rtrs_srv_hb_err_handler(struct rtrs_con *c) +{ + close_sess(to_srv_sess(c->sess)); +} + +static void rtrs_srv_init_hb(struct rtrs_srv_sess *sess) +{ + rtrs_init_hb(&sess->s, &io_comp_cqe, + RTRS_HB_INTERVAL_MS, + RTRS_HB_MISSED_MAX, + rtrs_srv_hb_err_handler, + rtrs_wq); +} + +static void rtrs_srv_start_hb(struct rtrs_srv_sess *sess) +{ + rtrs_start_hb(&sess->s); +} + +static void rtrs_srv_stop_hb(struct rtrs_srv_sess *sess) +{ + rtrs_stop_hb(&sess->s); +} + +static void rtrs_srv_info_rsp_done(struct ib_cq *cq, struct ib_wc *wc) +{ + struct rtrs_srv_con *con = cq->cq_context; + struct rtrs_sess *s = con->c.sess; + struct rtrs_srv_sess *sess = to_srv_sess(s); + struct rtrs_iu *iu; + + iu = container_of(wc->wr_cqe, struct rtrs_iu, cqe); + rtrs_iu_free(iu, sess->s.dev->ib_dev, 1); + + if (unlikely(wc->status != IB_WC_SUCCESS)) { + rtrs_err(s, "Sess info response send failed: %s\n", + ib_wc_status_msg(wc->status)); + close_sess(sess); + return; + } + WARN_ON(wc->opcode != IB_WC_SEND); +} + +static void rtrs_srv_sess_up(struct rtrs_srv_sess *sess) +{ + struct rtrs_srv *srv = sess->srv; + struct rtrs_srv_ctx *ctx = srv->ctx; + int up; + + mutex_lock(&srv->paths_ev_mutex); + up = ++srv->paths_up; + if (up == 1) + ctx->ops.link_ev(srv, RTRS_SRV_LINK_EV_CONNECTED, NULL); + mutex_unlock(&srv->paths_ev_mutex); + + /* Mark session as established */ + sess->established = true; +} + +static void rtrs_srv_sess_down(struct rtrs_srv_sess *sess) +{ + struct rtrs_srv *srv = sess->srv; + struct rtrs_srv_ctx *ctx = srv->ctx; + + if (!sess->established) + return; + + sess->established = false; + mutex_lock(&srv->paths_ev_mutex); + WARN_ON(!srv->paths_up); + if (--srv->paths_up == 0) + ctx->ops.link_ev(srv, RTRS_SRV_LINK_EV_DISCONNECTED, srv->priv); + mutex_unlock(&srv->paths_ev_mutex); +} + +static int post_recv_sess(struct rtrs_srv_sess *sess); + +static int process_info_req(struct rtrs_srv_con *con, + struct rtrs_msg_info_req *msg) +{ + struct rtrs_sess *s = con->c.sess; + struct rtrs_srv_sess *sess = to_srv_sess(s); + struct ib_send_wr *reg_wr = NULL; + struct rtrs_msg_info_rsp *rsp; + struct rtrs_iu *tx_iu; + struct ib_reg_wr *rwr; + int mri, err; + size_t tx_sz; + + err = post_recv_sess(sess); + if (unlikely(err)) { + rtrs_err(s, "post_recv_sess(), err: %d\n", err); + return err; + } + rwr = kcalloc(sess->mrs_num, sizeof(*rwr), GFP_KERNEL); + if (unlikely(!rwr)) + return -ENOMEM; + strlcpy(sess->s.sessname, msg->sessname, sizeof(sess->s.sessname)); + + tx_sz = sizeof(*rsp); + tx_sz += sizeof(rsp->desc[0]) * sess->mrs_num; + tx_iu = rtrs_iu_alloc(1, tx_sz, GFP_KERNEL, sess->s.dev->ib_dev, + DMA_TO_DEVICE, rtrs_srv_info_rsp_done); + if (unlikely(!tx_iu)) { + err = -ENOMEM; + goto rwr_free; + } + + rsp = tx_iu->buf; + rsp->type = cpu_to_le16(RTRS_MSG_INFO_RSP); + rsp->sg_cnt = cpu_to_le16(sess->mrs_num); + + for (mri = 0; mri < sess->mrs_num; mri++) { + struct ib_mr *mr = sess->mrs[mri].mr; + + rsp->desc[mri].addr = cpu_to_le64(mr->iova); + rsp->desc[mri].key = cpu_to_le32(mr->rkey); + rsp->desc[mri].len = cpu_to_le32(mr->length); + + /* + * Fill in reg MR request and chain them *backwards* + */ + rwr[mri].wr.next = mri ? &rwr[mri - 1].wr : NULL; + rwr[mri].wr.opcode = IB_WR_REG_MR; + rwr[mri].wr.wr_cqe = &local_reg_cqe; + rwr[mri].wr.num_sge = 0; + rwr[mri].wr.send_flags = 0; + rwr[mri].mr = mr; + rwr[mri].key = mr->rkey; + rwr[mri].access = (IB_ACCESS_LOCAL_WRITE | + IB_ACCESS_REMOTE_WRITE); + reg_wr = &rwr[mri].wr; + } + + err = rtrs_srv_create_sess_files(sess); + if (unlikely(err)) + goto iu_free; + kobject_get(&sess->kobj); + get_device(&sess->srv->dev); + rtrs_srv_change_state(sess, RTRS_SRV_CONNECTED); + rtrs_srv_start_hb(sess); + + /* + * We do not account number of established connections at the current + * moment, we rely on the client, which should send info request when + * all connections are successfully established. Thus, simply notify + * listener with a proper event if we are the first path. + */ + rtrs_srv_sess_up(sess); + + ib_dma_sync_single_for_device(sess->s.dev->ib_dev, tx_iu->dma_addr, + tx_iu->size, DMA_TO_DEVICE); + + /* Send info response */ + err = rtrs_iu_post_send(&con->c, tx_iu, tx_sz, reg_wr); + if (unlikely(err)) { + rtrs_err(s, "rtrs_iu_post_send(), err: %d\n", err); +iu_free: + rtrs_iu_free(tx_iu, sess->s.dev->ib_dev, 1); + } +rwr_free: + kfree(rwr); + + return err; +} + +static void rtrs_srv_info_req_done(struct ib_cq *cq, struct ib_wc *wc) +{ + struct rtrs_srv_con *con = cq->cq_context; + struct rtrs_sess *s = con->c.sess; + struct rtrs_srv_sess *sess = to_srv_sess(s); + struct rtrs_msg_info_req *msg; + struct rtrs_iu *iu; + int err; + + WARN_ON(con->c.cid); + + iu = container_of(wc->wr_cqe, struct rtrs_iu, cqe); + if (unlikely(wc->status != IB_WC_SUCCESS)) { + rtrs_err(s, "Sess info request receive failed: %s\n", + ib_wc_status_msg(wc->status)); + goto close; + } + WARN_ON(wc->opcode != IB_WC_RECV); + + if (unlikely(wc->byte_len < sizeof(*msg))) { + rtrs_err(s, "Sess info request is malformed: size %d\n", + wc->byte_len); + goto close; + } + ib_dma_sync_single_for_cpu(sess->s.dev->ib_dev, iu->dma_addr, + iu->size, DMA_FROM_DEVICE); + msg = iu->buf; + if (unlikely(le16_to_cpu(msg->type) != RTRS_MSG_INFO_REQ)) { + rtrs_err(s, "Sess info request is malformed: type %d\n", + le16_to_cpu(msg->type)); + goto close; + } + err = process_info_req(con, msg); + if (unlikely(err)) + goto close; + +out: + rtrs_iu_free(iu, sess->s.dev->ib_dev, 1); + return; +close: + close_sess(sess); + goto out; +} + +static int post_recv_info_req(struct rtrs_srv_con *con) +{ + struct rtrs_sess *s = con->c.sess; + struct rtrs_srv_sess *sess = to_srv_sess(s); + struct rtrs_iu *rx_iu; + int err; + + rx_iu = rtrs_iu_alloc(1, sizeof(struct rtrs_msg_info_req), + GFP_KERNEL, sess->s.dev->ib_dev, + DMA_FROM_DEVICE, rtrs_srv_info_req_done); + if (unlikely(!rx_iu)) + return -ENOMEM; + /* Prepare for getting info response */ + err = rtrs_iu_post_recv(&con->c, rx_iu); + if (unlikely(err)) { + rtrs_err(s, "rtrs_iu_post_recv(), err: %d\n", err); + rtrs_iu_free(rx_iu, sess->s.dev->ib_dev, 1); + return err; + } + + return 0; +} + +static int post_recv_io(struct rtrs_srv_con *con, size_t q_size) +{ + int i, err; + + for (i = 0; i < q_size; i++) { + err = rtrs_post_recv_empty(&con->c, &io_comp_cqe); + if (unlikely(err)) + return err; + } + + return 0; +} + +static int post_recv_sess(struct rtrs_srv_sess *sess) +{ + struct rtrs_srv *srv = sess->srv; + struct rtrs_sess *s = &sess->s; + size_t q_size; + int err, cid; + + for (cid = 0; cid < sess->s.con_num; cid++) { + if (cid == 0) + q_size = SERVICE_CON_QUEUE_DEPTH; + else + q_size = srv->queue_depth; + + err = post_recv_io(to_srv_con(sess->s.con[cid]), q_size); + if (unlikely(err)) { + rtrs_err(s, "post_recv_io(), err: %d\n", err); + return err; + } + } + + return 0; +} + +static void process_read(struct rtrs_srv_con *con, + struct rtrs_msg_rdma_read *msg, + u32 buf_id, u32 off) +{ + struct rtrs_sess *s = con->c.sess; + struct rtrs_srv_sess *sess = to_srv_sess(s); + struct rtrs_srv *srv = sess->srv; + struct rtrs_srv_ctx *ctx = srv->ctx; + struct rtrs_srv_op *id; + + size_t usr_len, data_len; + void *data; + int ret; + + if (unlikely(sess->state != RTRS_SRV_CONNECTED)) { + rtrs_err_rl(s, + "Processing read request failed, session is disconnected, sess state %s\n", + rtrs_srv_state_str(sess->state)); + return; + } + if (unlikely(msg->sg_cnt != 1 && msg->sg_cnt != 0)) { + rtrs_err_rl(s, + "Processing read request failed, invalid message\n"); + return; + } + rtrs_srv_get_ops_ids(sess); + rtrs_srv_update_rdma_stats(sess->stats, off, READ); + id = sess->ops_ids[buf_id]; + id->con = con; + id->dir = READ; + id->msg_id = buf_id; + id->rd_msg = msg; + usr_len = le16_to_cpu(msg->usr_len); + data_len = off - usr_len; + data = page_address(srv->chunks[buf_id]); + ret = ctx->ops.rdma_ev(srv, srv->priv, id, READ, data, data_len, + data + data_len, usr_len); + + if (unlikely(ret)) { + rtrs_err_rl(s, + "Processing read request failed, user module cb reported for msg_id %d, err: %d\n", + buf_id, ret); + goto send_err_msg; + } + + return; + +send_err_msg: + ret = send_io_resp_imm(con, id, ret); + if (ret < 0) { + rtrs_err_rl(s, + "Sending err msg for failed RDMA-Write-Req failed, msg_id %d, err: %d\n", + buf_id, ret); + close_sess(sess); + } + rtrs_srv_put_ops_ids(sess); +} + +static void process_write(struct rtrs_srv_con *con, + struct rtrs_msg_rdma_write *req, + u32 buf_id, u32 off) +{ + struct rtrs_sess *s = con->c.sess; + struct rtrs_srv_sess *sess = to_srv_sess(s); + struct rtrs_srv *srv = sess->srv; + struct rtrs_srv_ctx *ctx = srv->ctx; + struct rtrs_srv_op *id; + + size_t data_len, usr_len; + void *data; + int ret; + + if (unlikely(sess->state != RTRS_SRV_CONNECTED)) { + rtrs_err_rl(s, + "Processing write request failed, session is disconnected, sess state %s\n", + rtrs_srv_state_str(sess->state)); + return; + } + rtrs_srv_get_ops_ids(sess); + rtrs_srv_update_rdma_stats(sess->stats, off, WRITE); + id = sess->ops_ids[buf_id]; + id->con = con; + id->dir = WRITE; + id->msg_id = buf_id; + + usr_len = le16_to_cpu(req->usr_len); + data_len = off - usr_len; + data = page_address(srv->chunks[buf_id]); + ret = ctx->ops.rdma_ev(srv, srv->priv, id, WRITE, data, data_len, + data + data_len, usr_len); + if (unlikely(ret)) { + rtrs_err_rl(s, + "Processing write request failed, user module callback reports err: %d\n", + ret); + goto send_err_msg; + } + + return; + +send_err_msg: + ret = send_io_resp_imm(con, id, ret); + if (ret < 0) { + rtrs_err_rl(s, + "Processing write request failed, sending I/O response failed, msg_id %d, err: %d\n", + buf_id, ret); + close_sess(sess); + } + rtrs_srv_put_ops_ids(sess); +} + +static void process_io_req(struct rtrs_srv_con *con, void *msg, + u32 id, u32 off) +{ + struct rtrs_sess *s = con->c.sess; + struct rtrs_srv_sess *sess = to_srv_sess(s); + struct rtrs_msg_rdma_hdr *hdr; + unsigned int type; + + ib_dma_sync_single_for_cpu(sess->s.dev->ib_dev, sess->dma_addr[id], + max_chunk_size, DMA_BIDIRECTIONAL); + hdr = msg; + type = le16_to_cpu(hdr->type); + + switch (type) { + case RTRS_MSG_WRITE: + process_write(con, msg, id, off); + break; + case RTRS_MSG_READ: + process_read(con, msg, id, off); + break; + default: + rtrs_err(s, + "Processing I/O request failed, unknown message type received: 0x%02x\n", + type); + goto err; + } + + return; + +err: + close_sess(sess); +} + +static void rtrs_srv_inv_rkey_done(struct ib_cq *cq, struct ib_wc *wc) +{ + struct rtrs_srv_mr *mr = + container_of(wc->wr_cqe, typeof(*mr), inv_cqe); + struct rtrs_srv_con *con = cq->cq_context; + struct rtrs_sess *s = con->c.sess; + struct rtrs_srv_sess *sess = to_srv_sess(s); + struct rtrs_srv *srv = sess->srv; + u32 msg_id, off; + void *data; + + if (unlikely(wc->status != IB_WC_SUCCESS)) { + rtrs_err(s, "Failed IB_WR_LOCAL_INV: %s\n", + ib_wc_status_msg(wc->status)); + close_sess(sess); + } + msg_id = mr->msg_id; + off = mr->msg_off; + data = page_address(srv->chunks[msg_id]) + off; + process_io_req(con, data, msg_id, off); +} + +static int rtrs_srv_inv_rkey(struct rtrs_srv_con *con, + struct rtrs_srv_mr *mr) +{ + struct ib_send_wr wr = { + .opcode = IB_WR_LOCAL_INV, + .wr_cqe = &mr->inv_cqe, + .send_flags = IB_SEND_SIGNALED, + .ex.invalidate_rkey = mr->mr->rkey, + }; + mr->inv_cqe.done = rtrs_srv_inv_rkey_done; + + return ib_post_send(con->c.qp, &wr, NULL); +} + +static void rtrs_rdma_process_wr_wait_list(struct rtrs_srv_con *con) +{ + spin_lock(&con->rsp_wr_wait_lock); + while (!list_empty(&con->rsp_wr_wait_list)) { + struct rtrs_srv_op *id; + int ret; + + id = list_entry(con->rsp_wr_wait_list.next, + struct rtrs_srv_op, wait_list); + list_del(&id->wait_list); + + spin_unlock(&con->rsp_wr_wait_lock); + ret = rtrs_srv_resp_rdma(id, id->status); + spin_lock(&con->rsp_wr_wait_lock); + + if (!ret) { + list_add(&id->wait_list, &con->rsp_wr_wait_list); + break; + } + } + spin_unlock(&con->rsp_wr_wait_lock); +} + +static void rtrs_srv_rdma_done(struct ib_cq *cq, struct ib_wc *wc) +{ + struct rtrs_srv_con *con = cq->cq_context; + struct rtrs_sess *s = con->c.sess; + struct rtrs_srv_sess *sess = to_srv_sess(s); + struct rtrs_srv *srv = sess->srv; + u32 imm_type, imm_payload; + int err; + + if (unlikely(wc->status != IB_WC_SUCCESS)) { + if (wc->status != IB_WC_WR_FLUSH_ERR) { + rtrs_err(s, + "%s (wr_cqe: %p, type: %d, vendor_err: 0x%x, len: %u)\n", + ib_wc_status_msg(wc->status), wc->wr_cqe, + wc->opcode, wc->vendor_err, wc->byte_len); + close_sess(sess); + } + return; + } + + switch (wc->opcode) { + case IB_WC_RECV_RDMA_WITH_IMM: + /* + * post_recv() RDMA write completions of IO reqs (read/write) + * and hb + */ + if (WARN_ON(wc->wr_cqe != &io_comp_cqe)) + return; + err = rtrs_post_recv_empty(&con->c, &io_comp_cqe); + if (unlikely(err)) { + rtrs_err(s, "rtrs_post_recv(), err: %d\n", err); + close_sess(sess); + break; + } + rtrs_from_imm(be32_to_cpu(wc->ex.imm_data), + &imm_type, &imm_payload); + if (likely(imm_type == RTRS_IO_REQ_IMM)) { + u32 msg_id, off; + void *data; + + msg_id = imm_payload >> sess->mem_bits; + off = imm_payload & ((1 << sess->mem_bits) - 1); + if (unlikely(msg_id >= srv->queue_depth || + off >= max_chunk_size)) { + rtrs_err(s, "Wrong msg_id %u, off %u\n", + msg_id, off); + close_sess(sess); + return; + } + if (always_invalidate) { + struct rtrs_srv_mr *mr = &sess->mrs[msg_id]; + + mr->msg_off = off; + mr->msg_id = msg_id; + err = rtrs_srv_inv_rkey(con, mr); + if (unlikely(err)) { + rtrs_err(s, "rtrs_post_recv(), err: %d\n", + err); + close_sess(sess); + break; + } + } else { + data = page_address(srv->chunks[msg_id]) + off; + process_io_req(con, data, msg_id, off); + } + } else if (imm_type == RTRS_HB_MSG_IMM) { + WARN_ON(con->c.cid); + rtrs_send_hb_ack(&sess->s); + } else if (imm_type == RTRS_HB_ACK_IMM) { + WARN_ON(con->c.cid); + sess->s.hb_missed_cnt = 0; + } else { + rtrs_wrn(s, "Unknown IMM type %u\n", imm_type); + } + break; + case IB_WC_RDMA_WRITE: + case IB_WC_SEND: + /* + * post_send() RDMA write completions of IO reqs (read/write) + */ + atomic_add(srv->queue_depth, &con->sq_wr_avail); + + if (unlikely(!list_empty_careful(&con->rsp_wr_wait_list))) + rtrs_rdma_process_wr_wait_list(con); + + break; + default: + rtrs_wrn(s, "Unexpected WC type: %d\n", wc->opcode); + return; + } +} + +/** + * rtrs_srv_get_sess_name() - Get rtrs_srv peer hostname. + * @srv: Session + * @sessname: Sessname buffer + * @len: Length of sessname buffer + */ +int rtrs_srv_get_sess_name(struct rtrs_srv *srv, char *sessname, size_t len) +{ + struct rtrs_srv_sess *sess; + int err = -ENOTCONN; + + mutex_lock(&srv->paths_mutex); + list_for_each_entry(sess, &srv->paths_list, s.entry) { + if (sess->state != RTRS_SRV_CONNECTED) + continue; + strlcpy(sessname, sess->s.sessname, + min_t(size_t, sizeof(sess->s.sessname), len)); + err = 0; + break; + } + mutex_unlock(&srv->paths_mutex); + + return err; +} +EXPORT_SYMBOL(rtrs_srv_get_sess_name); + +/** + * rtrs_srv_get_sess_qdepth() - Get rtrs_srv qdepth. + * @srv: Session + */ +int rtrs_srv_get_queue_depth(struct rtrs_srv *srv) +{ + return srv->queue_depth; +} +EXPORT_SYMBOL(rtrs_srv_get_queue_depth); + +static int find_next_bit_ring(struct rtrs_srv_sess *sess) +{ + struct ib_device *ib_dev = sess->s.dev->ib_dev; + int v; + + v = cpumask_next(sess->cur_cq_vector, &cq_affinity_mask); + if (v >= nr_cpu_ids || v >= ib_dev->num_comp_vectors) + v = cpumask_first(&cq_affinity_mask); + return v; +} + +static int rtrs_srv_get_next_cq_vector(struct rtrs_srv_sess *sess) +{ + sess->cur_cq_vector = find_next_bit_ring(sess); + + return sess->cur_cq_vector; +} + +static void rtrs_srv_dev_release(struct device *dev) +{ + struct rtrs_srv *srv = container_of(dev, struct rtrs_srv, dev); + + kfree(srv); +} + +static void free_srv(struct rtrs_srv *srv) +{ + int i; + + WARN_ON(refcount_read(&srv->refcount)); + for (i = 0; i < srv->queue_depth; i++) + mempool_free(srv->chunks[i], chunk_pool); + kfree(srv->chunks); + mutex_destroy(&srv->paths_mutex); + mutex_destroy(&srv->paths_ev_mutex); + /* last put to release the srv structure */ + put_device(&srv->dev); +} + +static struct rtrs_srv *get_or_create_srv(struct rtrs_srv_ctx *ctx, + const uuid_t *paths_uuid, + bool first_conn) +{ + struct rtrs_srv *srv; + int i; + + mutex_lock(&ctx->srv_mutex); + list_for_each_entry(srv, &ctx->srv_list, ctx_list) { + if (uuid_equal(&srv->paths_uuid, paths_uuid) && + refcount_inc_not_zero(&srv->refcount)) { + mutex_unlock(&ctx->srv_mutex); + return srv; + } + } + mutex_unlock(&ctx->srv_mutex); + /* + * If this request is not the first connection request from the + * client for this session then fail and return error. + */ + if (!first_conn) + return ERR_PTR(-ENXIO); + + /* need to allocate a new srv */ + srv = kzalloc(sizeof(*srv), GFP_KERNEL); + if (!srv) + return ERR_PTR(-ENOMEM); + + INIT_LIST_HEAD(&srv->paths_list); + mutex_init(&srv->paths_mutex); + mutex_init(&srv->paths_ev_mutex); + uuid_copy(&srv->paths_uuid, paths_uuid); + srv->queue_depth = sess_queue_depth; + srv->ctx = ctx; + device_initialize(&srv->dev); + srv->dev.release = rtrs_srv_dev_release; + + srv->chunks = kcalloc(srv->queue_depth, sizeof(*srv->chunks), + GFP_KERNEL); + if (!srv->chunks) + goto err_free_srv; + + for (i = 0; i < srv->queue_depth; i++) { + srv->chunks[i] = mempool_alloc(chunk_pool, GFP_KERNEL); + if (!srv->chunks[i]) + goto err_free_chunks; + } + refcount_set(&srv->refcount, 1); + mutex_lock(&ctx->srv_mutex); + list_add(&srv->ctx_list, &ctx->srv_list); + mutex_unlock(&ctx->srv_mutex); + + return srv; + +err_free_chunks: + while (i--) + mempool_free(srv->chunks[i], chunk_pool); + kfree(srv->chunks); + +err_free_srv: + kfree(srv); + return ERR_PTR(-ENOMEM); +} + +static void put_srv(struct rtrs_srv *srv) +{ + if (refcount_dec_and_test(&srv->refcount)) { + struct rtrs_srv_ctx *ctx = srv->ctx; + + WARN_ON(srv->dev.kobj.state_in_sysfs); + + mutex_lock(&ctx->srv_mutex); + list_del(&srv->ctx_list); + mutex_unlock(&ctx->srv_mutex); + free_srv(srv); + } +} + +static void __add_path_to_srv(struct rtrs_srv *srv, + struct rtrs_srv_sess *sess) +{ + list_add_tail(&sess->s.entry, &srv->paths_list); + srv->paths_num++; + WARN_ON(srv->paths_num >= MAX_PATHS_NUM); +} + +static void del_path_from_srv(struct rtrs_srv_sess *sess) +{ + struct rtrs_srv *srv = sess->srv; + + if (WARN_ON(!srv)) + return; + + mutex_lock(&srv->paths_mutex); + list_del(&sess->s.entry); + WARN_ON(!srv->paths_num); + srv->paths_num--; + mutex_unlock(&srv->paths_mutex); +} + +/* return true if addresses are the same, error other wise */ +static int sockaddr_cmp(const struct sockaddr *a, const struct sockaddr *b) +{ + switch (a->sa_family) { + case AF_IB: + return memcmp(&((struct sockaddr_ib *)a)->sib_addr, + &((struct sockaddr_ib *)b)->sib_addr, + sizeof(struct ib_addr)) && + (b->sa_family == AF_IB); + case AF_INET: + return memcmp(&((struct sockaddr_in *)a)->sin_addr, + &((struct sockaddr_in *)b)->sin_addr, + sizeof(struct in_addr)) && + (b->sa_family == AF_INET); + case AF_INET6: + return memcmp(&((struct sockaddr_in6 *)a)->sin6_addr, + &((struct sockaddr_in6 *)b)->sin6_addr, + sizeof(struct in6_addr)) && + (b->sa_family == AF_INET6); + default: + return -ENOENT; + } +} + +static bool __is_path_w_addr_exists(struct rtrs_srv *srv, + struct rdma_addr *addr) +{ + struct rtrs_srv_sess *sess; + + list_for_each_entry(sess, &srv->paths_list, s.entry) + if (!sockaddr_cmp((struct sockaddr *)&sess->s.dst_addr, + (struct sockaddr *)&addr->dst_addr) && + !sockaddr_cmp((struct sockaddr *)&sess->s.src_addr, + (struct sockaddr *)&addr->src_addr)) + return true; + + return false; +} + +static void free_sess(struct rtrs_srv_sess *sess) +{ + if (sess->kobj.state_in_sysfs) { + kobject_del(&sess->kobj); + kobject_put(&sess->kobj); + } else { + kfree(sess->stats); + kfree(sess); + } +} + +static void rtrs_srv_close_work(struct work_struct *work) +{ + struct rtrs_srv_sess *sess; + struct rtrs_srv_con *con; + int i; + + sess = container_of(work, typeof(*sess), close_work); + + rtrs_srv_destroy_sess_files(sess); + rtrs_srv_stop_hb(sess); + + for (i = 0; i < sess->s.con_num; i++) { + if (!sess->s.con[i]) + continue; + con = to_srv_con(sess->s.con[i]); + rdma_disconnect(con->c.cm_id); + ib_drain_qp(con->c.qp); + } + /* Wait for all inflights */ + rtrs_srv_wait_ops_ids(sess); + + /* Notify upper layer if we are the last path */ + rtrs_srv_sess_down(sess); + + unmap_cont_bufs(sess); + rtrs_srv_free_ops_ids(sess); + + for (i = 0; i < sess->s.con_num; i++) { + if (!sess->s.con[i]) + continue; + con = to_srv_con(sess->s.con[i]); + rtrs_cq_qp_destroy(&con->c); + rdma_destroy_id(con->c.cm_id); + kfree(con); + } + rtrs_ib_dev_put(sess->s.dev); + + del_path_from_srv(sess); + put_srv(sess->srv); + sess->srv = NULL; + rtrs_srv_change_state(sess, RTRS_SRV_CLOSED); + + kfree(sess->dma_addr); + kfree(sess->s.con); + free_sess(sess); +} + +static int rtrs_rdma_do_accept(struct rtrs_srv_sess *sess, + struct rdma_cm_id *cm_id) +{ + struct rtrs_srv *srv = sess->srv; + struct rtrs_msg_conn_rsp msg; + struct rdma_conn_param param; + int err; + + param = (struct rdma_conn_param) { + .rnr_retry_count = 7, + .private_data = &msg, + .private_data_len = sizeof(msg), + }; + + msg = (struct rtrs_msg_conn_rsp) { + .magic = cpu_to_le16(RTRS_MAGIC), + .version = cpu_to_le16(RTRS_PROTO_VER), + .queue_depth = cpu_to_le16(srv->queue_depth), + .max_io_size = cpu_to_le32(max_chunk_size - MAX_HDR_SIZE), + .max_hdr_size = cpu_to_le32(MAX_HDR_SIZE), + }; + + if (always_invalidate) + msg.flags = cpu_to_le32(RTRS_MSG_NEW_RKEY_F); + + err = rdma_accept(cm_id, ¶m); + if (err) + pr_err("rdma_accept(), err: %d\n", err); + + return err; +} + +static int rtrs_rdma_do_reject(struct rdma_cm_id *cm_id, int errno) +{ + struct rtrs_msg_conn_rsp msg; + int err; + + msg = (struct rtrs_msg_conn_rsp) { + .magic = cpu_to_le16(RTRS_MAGIC), + .version = cpu_to_le16(RTRS_PROTO_VER), + .errno = cpu_to_le16(errno), + }; + + err = rdma_reject(cm_id, &msg, sizeof(msg), IB_CM_REJ_CONSUMER_DEFINED); + if (err) + pr_err("rdma_reject(), err: %d\n", err); + + /* Bounce errno back */ + return errno; +} + +static struct rtrs_srv_sess * +__find_sess(struct rtrs_srv *srv, const uuid_t *sess_uuid) +{ + struct rtrs_srv_sess *sess; + + list_for_each_entry(sess, &srv->paths_list, s.entry) { + if (uuid_equal(&sess->s.uuid, sess_uuid)) + return sess; + } + + return NULL; +} + +static int create_con(struct rtrs_srv_sess *sess, + struct rdma_cm_id *cm_id, + unsigned int cid) +{ + struct rtrs_srv *srv = sess->srv; + struct rtrs_sess *s = &sess->s; + struct rtrs_srv_con *con; + + u32 cq_size, max_send_wr, max_recv_wr, wr_limit; + int err, cq_vector; + + con = kzalloc(sizeof(*con), GFP_KERNEL); + if (!con) { + err = -ENOMEM; + goto err; + } + + spin_lock_init(&con->rsp_wr_wait_lock); + INIT_LIST_HEAD(&con->rsp_wr_wait_list); + con->c.cm_id = cm_id; + con->c.sess = &sess->s; + con->c.cid = cid; + atomic_set(&con->wr_cnt, 1); + + if (con->c.cid == 0) { + /* + * All receive and all send (each requiring invalidate) + * + 2 for drain and heartbeat + */ + max_send_wr = SERVICE_CON_QUEUE_DEPTH * 2 + 2; + max_recv_wr = SERVICE_CON_QUEUE_DEPTH + 2; + cq_size = max_send_wr + max_recv_wr; + } else { + /* + * In theory we might have queue_depth * 32 + * outstanding requests if an unsafe global key is used + * and we have queue_depth read requests each consisting + * of 32 different addresses. div 3 for mlx5. + */ + wr_limit = sess->s.dev->ib_dev->attrs.max_qp_wr / 3; + /* when always_invlaidate enalbed, we need linv+rinv+mr+imm */ + if (always_invalidate) + max_send_wr = + min_t(int, wr_limit, + srv->queue_depth * (1 + 4) + 1); + else + max_send_wr = + min_t(int, wr_limit, + srv->queue_depth * (1 + 2) + 1); + + max_recv_wr = srv->queue_depth + 1; + /* + * If we have all receive requests posted and + * all write requests posted and each read request + * requires an invalidate request + drain + * and qp gets into error state. + */ + cq_size = max_send_wr + max_recv_wr; + } + atomic_set(&con->sq_wr_avail, max_send_wr); + cq_vector = rtrs_srv_get_next_cq_vector(sess); + + /* TODO: SOFTIRQ can be faster, but be careful with softirq context */ + err = rtrs_cq_qp_create(&sess->s, &con->c, 1, cq_vector, cq_size, + max_send_wr, max_recv_wr, + IB_POLL_WORKQUEUE); + if (err) { + rtrs_err(s, "rtrs_cq_qp_create(), err: %d\n", err); + goto free_con; + } + if (con->c.cid == 0) { + err = post_recv_info_req(con); + if (err) + goto free_cqqp; + } + WARN_ON(sess->s.con[cid]); + sess->s.con[cid] = &con->c; + + /* + * Change context from server to current connection. The other + * way is to use cm_id->qp->qp_context, which does not work on OFED. + */ + cm_id->context = &con->c; + + return 0; + +free_cqqp: + rtrs_cq_qp_destroy(&con->c); +free_con: + kfree(con); + +err: + return err; +} + +static struct rtrs_srv_sess *__alloc_sess(struct rtrs_srv *srv, + struct rdma_cm_id *cm_id, + unsigned int con_num, + unsigned int recon_cnt, + const uuid_t *uuid) +{ + struct rtrs_srv_sess *sess; + int err = -ENOMEM; + + if (srv->paths_num >= MAX_PATHS_NUM) { + err = -ECONNRESET; + goto err; + } + if (__is_path_w_addr_exists(srv, &cm_id->route.addr)) { + err = -EEXIST; + pr_err("Path with same addr exists\n"); + goto err; + } + sess = kzalloc(sizeof(*sess), GFP_KERNEL); + if (!sess) + goto err; + + sess->stats = kzalloc(sizeof(*sess->stats), GFP_KERNEL); + if (!sess->stats) + goto err_free_sess; + + sess->stats->sess = sess; + + sess->dma_addr = kcalloc(srv->queue_depth, sizeof(*sess->dma_addr), + GFP_KERNEL); + if (!sess->dma_addr) + goto err_free_stats; + + sess->s.con = kcalloc(con_num, sizeof(*sess->s.con), GFP_KERNEL); + if (!sess->s.con) + goto err_free_dma_addr; + + sess->state = RTRS_SRV_CONNECTING; + sess->srv = srv; + sess->cur_cq_vector = -1; + sess->s.dst_addr = cm_id->route.addr.dst_addr; + sess->s.src_addr = cm_id->route.addr.src_addr; + sess->s.con_num = con_num; + sess->s.recon_cnt = recon_cnt; + uuid_copy(&sess->s.uuid, uuid); + spin_lock_init(&sess->state_lock); + INIT_WORK(&sess->close_work, rtrs_srv_close_work); + rtrs_srv_init_hb(sess); + + sess->s.dev = rtrs_ib_dev_find_or_add(cm_id->device, &dev_pd); + if (!sess->s.dev) { + err = -ENOMEM; + goto err_free_con; + } + err = map_cont_bufs(sess); + if (err) + goto err_put_dev; + + err = rtrs_srv_alloc_ops_ids(sess); + if (err) + goto err_unmap_bufs; + + __add_path_to_srv(srv, sess); + + return sess; + +err_unmap_bufs: + unmap_cont_bufs(sess); +err_put_dev: + rtrs_ib_dev_put(sess->s.dev); +err_free_con: + kfree(sess->s.con); +err_free_dma_addr: + kfree(sess->dma_addr); +err_free_stats: + kfree(sess->stats); +err_free_sess: + kfree(sess); +err: + return ERR_PTR(err); +} + +static int rtrs_rdma_connect(struct rdma_cm_id *cm_id, + const struct rtrs_msg_conn_req *msg, + size_t len) +{ + struct rtrs_srv_ctx *ctx = cm_id->context; + struct rtrs_srv_sess *sess; + struct rtrs_srv *srv; + + u16 version, con_num, cid; + u16 recon_cnt; + int err; + + if (len < sizeof(*msg)) { + pr_err("Invalid RTRS connection request\n"); + goto reject_w_econnreset; + } + if (le16_to_cpu(msg->magic) != RTRS_MAGIC) { + pr_err("Invalid RTRS magic\n"); + goto reject_w_econnreset; + } + version = le16_to_cpu(msg->version); + if (version >> 8 != RTRS_PROTO_VER_MAJOR) { + pr_err("Unsupported major RTRS version: %d, expected %d\n", + version >> 8, RTRS_PROTO_VER_MAJOR); + goto reject_w_econnreset; + } + con_num = le16_to_cpu(msg->cid_num); + if (con_num > 4096) { + /* Sanity check */ + pr_err("Too many connections requested: %d\n", con_num); + goto reject_w_econnreset; + } + cid = le16_to_cpu(msg->cid); + if (cid >= con_num) { + /* Sanity check */ + pr_err("Incorrect cid: %d >= %d\n", cid, con_num); + goto reject_w_econnreset; + } + recon_cnt = le16_to_cpu(msg->recon_cnt); + srv = get_or_create_srv(ctx, &msg->paths_uuid, msg->first_conn); + if (IS_ERR(srv)) { + err = PTR_ERR(srv); + goto reject_w_err; + } + mutex_lock(&srv->paths_mutex); + sess = __find_sess(srv, &msg->sess_uuid); + if (sess) { + struct rtrs_sess *s = &sess->s; + + /* Session already holds a reference */ + put_srv(srv); + + if (sess->state != RTRS_SRV_CONNECTING) { + rtrs_err(s, "Session in wrong state: %s\n", + rtrs_srv_state_str(sess->state)); + mutex_unlock(&srv->paths_mutex); + goto reject_w_econnreset; + } + /* + * Sanity checks + */ + if (con_num != s->con_num || cid >= s->con_num) { + rtrs_err(s, "Incorrect request: %d, %d\n", + cid, con_num); + mutex_unlock(&srv->paths_mutex); + goto reject_w_econnreset; + } + if (s->con[cid]) { + rtrs_err(s, "Connection already exists: %d\n", + cid); + mutex_unlock(&srv->paths_mutex); + goto reject_w_econnreset; + } + } else { + sess = __alloc_sess(srv, cm_id, con_num, recon_cnt, + &msg->sess_uuid); + if (IS_ERR(sess)) { + mutex_unlock(&srv->paths_mutex); + put_srv(srv); + err = PTR_ERR(sess); + goto reject_w_err; + } + } + err = create_con(sess, cm_id, cid); + if (err) { + (void)rtrs_rdma_do_reject(cm_id, err); + /* + * Since session has other connections we follow normal way + * through workqueue, but still return an error to tell cma.c + * to call rdma_destroy_id() for current connection. + */ + goto close_and_return_err; + } + err = rtrs_rdma_do_accept(sess, cm_id); + if (err) { + (void)rtrs_rdma_do_reject(cm_id, err); + /* + * Since current connection was successfully added to the + * session we follow normal way through workqueue to close the + * session, thus return 0 to tell cma.c we call + * rdma_destroy_id() ourselves. + */ + err = 0; + goto close_and_return_err; + } + mutex_unlock(&srv->paths_mutex); + + return 0; + +reject_w_err: + return rtrs_rdma_do_reject(cm_id, err); + +reject_w_econnreset: + return rtrs_rdma_do_reject(cm_id, -ECONNRESET); + +close_and_return_err: + mutex_unlock(&srv->paths_mutex); + close_sess(sess); + + return err; +} + +static int rtrs_srv_rdma_cm_handler(struct rdma_cm_id *cm_id, + struct rdma_cm_event *ev) +{ + struct rtrs_srv_sess *sess = NULL; + struct rtrs_sess *s = NULL; + + if (ev->event != RDMA_CM_EVENT_CONNECT_REQUEST) { + struct rtrs_con *c = cm_id->context; + + s = c->sess; + sess = to_srv_sess(s); + } + + switch (ev->event) { + case RDMA_CM_EVENT_CONNECT_REQUEST: + /* + * In case of error cma.c will destroy cm_id, + * see cma_process_remove() + */ + return rtrs_rdma_connect(cm_id, ev->param.conn.private_data, + ev->param.conn.private_data_len); + case RDMA_CM_EVENT_ESTABLISHED: + /* Nothing here */ + break; + case RDMA_CM_EVENT_REJECTED: + case RDMA_CM_EVENT_CONNECT_ERROR: + case RDMA_CM_EVENT_UNREACHABLE: + rtrs_err(s, "CM error (CM event: %s, err: %d)\n", + rdma_event_msg(ev->event), ev->status); + close_sess(sess); + break; + case RDMA_CM_EVENT_DISCONNECTED: + case RDMA_CM_EVENT_ADDR_CHANGE: + case RDMA_CM_EVENT_TIMEWAIT_EXIT: + close_sess(sess); + break; + case RDMA_CM_EVENT_DEVICE_REMOVAL: + close_sess(sess); + break; + default: + pr_err("Ignoring unexpected CM event %s, err %d\n", + rdma_event_msg(ev->event), ev->status); + break; + } + + return 0; +} + +static struct rdma_cm_id *rtrs_srv_cm_init(struct rtrs_srv_ctx *ctx, + struct sockaddr *addr, + enum rdma_ucm_port_space ps) +{ + struct rdma_cm_id *cm_id; + int ret; + + cm_id = rdma_create_id(&init_net, rtrs_srv_rdma_cm_handler, + ctx, ps, IB_QPT_RC); + if (IS_ERR(cm_id)) { + ret = PTR_ERR(cm_id); + pr_err("Creating id for RDMA connection failed, err: %d\n", + ret); + goto err_out; + } + ret = rdma_bind_addr(cm_id, addr); + if (ret) { + pr_err("Binding RDMA address failed, err: %d\n", ret); + goto err_cm; + } + ret = rdma_listen(cm_id, 64); + if (ret) { + pr_err("Listening on RDMA connection failed, err: %d\n", + ret); + goto err_cm; + } + + return cm_id; + +err_cm: + rdma_destroy_id(cm_id); +err_out: + + return ERR_PTR(ret); +} + +static int rtrs_srv_rdma_init(struct rtrs_srv_ctx *ctx, u16 port) +{ + struct sockaddr_in6 sin = { + .sin6_family = AF_INET6, + .sin6_addr = IN6ADDR_ANY_INIT, + .sin6_port = htons(port), + }; + struct sockaddr_ib sib = { + .sib_family = AF_IB, + .sib_sid = cpu_to_be64(RDMA_IB_IP_PS_IB | port), + .sib_sid_mask = cpu_to_be64(0xffffffffffffffffULL), + .sib_pkey = cpu_to_be16(0xffff), + }; + struct rdma_cm_id *cm_ip, *cm_ib; + int ret; + + /* + * We accept both IPoIB and IB connections, so we need to keep + * two cm id's, one for each socket type and port space. + * If the cm initialization of one of the id's fails, we abort + * everything. + */ + cm_ip = rtrs_srv_cm_init(ctx, (struct sockaddr *)&sin, RDMA_PS_TCP); + if (IS_ERR(cm_ip)) + return PTR_ERR(cm_ip); + + cm_ib = rtrs_srv_cm_init(ctx, (struct sockaddr *)&sib, RDMA_PS_IB); + if (IS_ERR(cm_ib)) { + ret = PTR_ERR(cm_ib); + goto free_cm_ip; + } + + ctx->cm_id_ip = cm_ip; + ctx->cm_id_ib = cm_ib; + + return 0; + +free_cm_ip: + rdma_destroy_id(cm_ip); + + return ret; +} + +static struct rtrs_srv_ctx *alloc_srv_ctx(struct rtrs_srv_ops *ops) +{ + struct rtrs_srv_ctx *ctx; + + ctx = kzalloc(sizeof(*ctx), GFP_KERNEL); + if (!ctx) + return NULL; + + ctx->ops = *ops; + mutex_init(&ctx->srv_mutex); + INIT_LIST_HEAD(&ctx->srv_list); + + return ctx; +} + +static void free_srv_ctx(struct rtrs_srv_ctx *ctx) +{ + WARN_ON(!list_empty(&ctx->srv_list)); + mutex_destroy(&ctx->srv_mutex); + kfree(ctx); +} + +static int rtrs_srv_add_one(struct ib_device *device) +{ + struct rtrs_srv_ctx *ctx; + int ret = 0; + + mutex_lock(&ib_ctx.ib_dev_mutex); + if (ib_ctx.ib_dev_count) + goto out; + + /* + * Since our CM IDs are NOT bound to any ib device we will create them + * only once + */ + ctx = ib_ctx.srv_ctx; + ret = rtrs_srv_rdma_init(ctx, ib_ctx.port); + if (ret) { + /* + * We errored out here. + * According to the ib code, if we encounter an error here then the + * error code is ignored, and no more calls to our ops are made. + */ + pr_err("Failed to initialize RDMA connection"); + goto err_out; + } + +out: + /* + * Keep a track on the number of ib devices added + */ + ib_ctx.ib_dev_count++; + +err_out: + mutex_unlock(&ib_ctx.ib_dev_mutex); + return ret; +} + +static void rtrs_srv_remove_one(struct ib_device *device, void *client_data) +{ + struct rtrs_srv_ctx *ctx; + + mutex_lock(&ib_ctx.ib_dev_mutex); + ib_ctx.ib_dev_count--; + + if (ib_ctx.ib_dev_count) + goto out; + + /* + * Since our CM IDs are NOT bound to any ib device we will remove them + * only once, when the last device is removed + */ + ctx = ib_ctx.srv_ctx; + rdma_destroy_id(ctx->cm_id_ip); + rdma_destroy_id(ctx->cm_id_ib); + +out: + mutex_unlock(&ib_ctx.ib_dev_mutex); +} + +static struct ib_client rtrs_srv_client = { + .name = "rtrs_server", + .add = rtrs_srv_add_one, + .remove = rtrs_srv_remove_one +}; + +/** + * rtrs_srv_open() - open RTRS server context + * @ops: callback functions + * @port: port to listen on + * + * Creates server context with specified callbacks. + * + * Return a valid pointer on success otherwise PTR_ERR. + */ +struct rtrs_srv_ctx *rtrs_srv_open(struct rtrs_srv_ops *ops, u16 port) +{ + struct rtrs_srv_ctx *ctx; + int err; + + ctx = alloc_srv_ctx(ops); + if (!ctx) + return ERR_PTR(-ENOMEM); + + mutex_init(&ib_ctx.ib_dev_mutex); + ib_ctx.srv_ctx = ctx; + ib_ctx.port = port; + + err = ib_register_client(&rtrs_srv_client); + if (err) { + free_srv_ctx(ctx); + return ERR_PTR(err); + } + + return ctx; +} +EXPORT_SYMBOL(rtrs_srv_open); + +static void close_sessions(struct rtrs_srv *srv) +{ + struct rtrs_srv_sess *sess; + + mutex_lock(&srv->paths_mutex); + list_for_each_entry(sess, &srv->paths_list, s.entry) + close_sess(sess); + mutex_unlock(&srv->paths_mutex); +} + +static void close_ctx(struct rtrs_srv_ctx *ctx) +{ + struct rtrs_srv *srv; + + mutex_lock(&ctx->srv_mutex); + list_for_each_entry(srv, &ctx->srv_list, ctx_list) + close_sessions(srv); + mutex_unlock(&ctx->srv_mutex); + flush_workqueue(rtrs_wq); +} + +/** + * rtrs_srv_close() - close RTRS server context + * @ctx: pointer to server context + * + * Closes RTRS server context with all client sessions. + */ +void rtrs_srv_close(struct rtrs_srv_ctx *ctx) +{ + ib_unregister_client(&rtrs_srv_client); + mutex_destroy(&ib_ctx.ib_dev_mutex); + close_ctx(ctx); + free_srv_ctx(ctx); +} +EXPORT_SYMBOL(rtrs_srv_close); + +static int check_module_params(void) +{ + if (sess_queue_depth < 1 || sess_queue_depth > MAX_SESS_QUEUE_DEPTH) { + pr_err("Invalid sess_queue_depth value %d, has to be >= %d, <= %d.\n", + sess_queue_depth, 1, MAX_SESS_QUEUE_DEPTH); + return -EINVAL; + } + if (max_chunk_size < MIN_CHUNK_SIZE || !is_power_of_2(max_chunk_size)) { + pr_err("Invalid max_chunk_size value %d, has to be >= %d and should be power of two.\n", + max_chunk_size, MIN_CHUNK_SIZE); + return -EINVAL; + } + + /* + * Check if IB immediate data size is enough to hold the mem_id and the + * offset inside the memory chunk + */ + if ((ilog2(sess_queue_depth - 1) + 1) + + (ilog2(max_chunk_size - 1) + 1) > MAX_IMM_PAYL_BITS) { + pr_err("RDMA immediate size (%db) not enough to encode %d buffers of size %dB. Reduce 'sess_queue_depth' or 'max_chunk_size' parameters.\n", + MAX_IMM_PAYL_BITS, sess_queue_depth, max_chunk_size); + return -EINVAL; + } + + return 0; +} + +static int __init rtrs_server_init(void) +{ + int err; + + pr_info("Loading module %s, proto %s: (max_chunk_size: %d (pure IO %ld, headers %ld) , sess_queue_depth: %d, always_invalidate: %d)\n", + KBUILD_MODNAME, RTRS_PROTO_VER_STRING, + max_chunk_size, max_chunk_size - MAX_HDR_SIZE, MAX_HDR_SIZE, + sess_queue_depth, always_invalidate); + + rtrs_rdma_dev_pd_init(0, &dev_pd); + + err = check_module_params(); + if (err) { + pr_err("Failed to load module, invalid module parameters, err: %d\n", + err); + return err; + } + chunk_pool = mempool_create_page_pool(sess_queue_depth * CHUNK_POOL_SZ, + get_order(max_chunk_size)); + if (!chunk_pool) + return -ENOMEM; + rtrs_dev_class = class_create(THIS_MODULE, "rtrs-server"); + if (IS_ERR(rtrs_dev_class)) { + err = PTR_ERR(rtrs_dev_class); + goto out_chunk_pool; + } + rtrs_wq = alloc_workqueue("rtrs_server_wq", 0, 0); + if (!rtrs_wq) { + err = -ENOMEM; + goto out_dev_class; + } + + return 0; + +out_dev_class: + class_destroy(rtrs_dev_class); +out_chunk_pool: + mempool_destroy(chunk_pool); + + return err; +} + +static void __exit rtrs_server_exit(void) +{ + destroy_workqueue(rtrs_wq); + class_destroy(rtrs_dev_class); + mempool_destroy(chunk_pool); + rtrs_rdma_dev_pd_deinit(&dev_pd); +} + +module_init(rtrs_server_init); +module_exit(rtrs_server_exit); diff --git a/drivers/infiniband/ulp/rtrs/rtrs-srv.h b/drivers/infiniband/ulp/rtrs/rtrs-srv.h new file mode 100644 index 000000000..08b0b8a6e --- /dev/null +++ b/drivers/infiniband/ulp/rtrs/rtrs-srv.h @@ -0,0 +1,155 @@ +/* SPDX-License-Identifier: GPL-2.0-or-later */ +/* + * RDMA Transport Layer + * + * Copyright (c) 2014 - 2018 ProfitBricks GmbH. All rights reserved. + * Copyright (c) 2018 - 2019 1&1 IONOS Cloud GmbH. All rights reserved. + * Copyright (c) 2019 - 2020 1&1 IONOS SE. All rights reserved. + */ + +#ifndef RTRS_SRV_H +#define RTRS_SRV_H + +#include <linux/device.h> +#include <linux/refcount.h> +#include "rtrs-pri.h" + +/* + * enum rtrs_srv_state - Server states. + */ +enum rtrs_srv_state { + RTRS_SRV_CONNECTING, + RTRS_SRV_CONNECTED, + RTRS_SRV_CLOSING, + RTRS_SRV_CLOSED, +}; + +/* stats for Read and write operation. + * see Documentation/ABI/testing/sysfs-class-rtrs-server for details + */ +struct rtrs_srv_stats_rdma_stats { + struct { + atomic64_t cnt; + atomic64_t size_total; + } dir[2]; +}; + +struct rtrs_srv_stats { + struct kobject kobj_stats; + struct rtrs_srv_stats_rdma_stats rdma_stats; + struct rtrs_srv_sess *sess; +}; + +struct rtrs_srv_con { + struct rtrs_con c; + atomic_t wr_cnt; + atomic_t sq_wr_avail; + struct list_head rsp_wr_wait_list; + spinlock_t rsp_wr_wait_lock; +}; + +/* IO context in rtrs_srv, each io has one */ +struct rtrs_srv_op { + struct rtrs_srv_con *con; + u32 msg_id; + u8 dir; + struct rtrs_msg_rdma_read *rd_msg; + struct ib_rdma_wr tx_wr; + struct ib_sge tx_sg; + struct list_head wait_list; + int status; +}; + +/* + * server side memory region context, when always_invalidate=Y, we need + * queue_depth of memory regrion to invalidate each memory region. + */ +struct rtrs_srv_mr { + struct ib_mr *mr; + struct sg_table sgt; + struct ib_cqe inv_cqe; /* only for always_invalidate=true */ + u32 msg_id; /* only for always_invalidate=true */ + u32 msg_off; /* only for always_invalidate=true */ + struct rtrs_iu *iu; /* send buffer for new rkey msg */ +}; + +struct rtrs_srv_sess { + struct rtrs_sess s; + struct rtrs_srv *srv; + struct work_struct close_work; + enum rtrs_srv_state state; + spinlock_t state_lock; + int cur_cq_vector; + struct rtrs_srv_op **ops_ids; + atomic_t ids_inflight; + wait_queue_head_t ids_waitq; + struct rtrs_srv_mr *mrs; + unsigned int mrs_num; + dma_addr_t *dma_addr; + bool established; + unsigned int mem_bits; + struct kobject kobj; + struct rtrs_srv_stats *stats; +}; + +struct rtrs_srv { + struct list_head paths_list; + int paths_up; + struct mutex paths_ev_mutex; + size_t paths_num; + struct mutex paths_mutex; + uuid_t paths_uuid; + refcount_t refcount; + struct rtrs_srv_ctx *ctx; + struct list_head ctx_list; + void *priv; + size_t queue_depth; + struct page **chunks; + struct device dev; + unsigned int dev_ref; + struct kobject *kobj_paths; +}; + +struct rtrs_srv_ctx { + struct rtrs_srv_ops ops; + struct rdma_cm_id *cm_id_ip; + struct rdma_cm_id *cm_id_ib; + struct mutex srv_mutex; + struct list_head srv_list; +}; + +struct rtrs_srv_ib_ctx { + struct rtrs_srv_ctx *srv_ctx; + u16 port; + struct mutex ib_dev_mutex; + int ib_dev_count; +}; + +extern struct class *rtrs_dev_class; + +void close_sess(struct rtrs_srv_sess *sess); + +static inline void rtrs_srv_update_rdma_stats(struct rtrs_srv_stats *s, + size_t size, int d) +{ + atomic64_inc(&s->rdma_stats.dir[d].cnt); + atomic64_add(size, &s->rdma_stats.dir[d].size_total); +} + +/* functions which are implemented in rtrs-srv-stats.c */ +int rtrs_srv_reset_rdma_stats(struct rtrs_srv_stats *stats, bool enable); +ssize_t rtrs_srv_stats_rdma_to_str(struct rtrs_srv_stats *stats, + char *page, size_t len); +int rtrs_srv_reset_wc_completion_stats(struct rtrs_srv_stats *stats, + bool enable); +int rtrs_srv_stats_wc_completion_to_str(struct rtrs_srv_stats *stats, char *buf, + size_t len); +int rtrs_srv_reset_all_stats(struct rtrs_srv_stats *stats, bool enable); +ssize_t rtrs_srv_reset_all_help(struct rtrs_srv_stats *stats, + char *page, size_t len); + +/* functions which are implemented in rtrs-srv-sysfs.c */ +int rtrs_srv_create_sess_files(struct rtrs_srv_sess *sess); +void rtrs_srv_destroy_sess_files(struct rtrs_srv_sess *sess); + +#endif /* RTRS_SRV_H */ diff --git a/drivers/infiniband/ulp/rtrs/rtrs.c b/drivers/infiniband/ulp/rtrs/rtrs.c new file mode 100644 index 000000000..76b993e8d --- /dev/null +++ b/drivers/infiniband/ulp/rtrs/rtrs.c @@ -0,0 +1,600 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * RDMA Transport Layer + * + * Copyright (c) 2014 - 2018 ProfitBricks GmbH. All rights reserved. + * Copyright (c) 2018 - 2019 1&1 IONOS Cloud GmbH. All rights reserved. + * Copyright (c) 2019 - 2020 1&1 IONOS SE. All rights reserved. + */ +#undef pr_fmt +#define pr_fmt(fmt) KBUILD_MODNAME " L" __stringify(__LINE__) ": " fmt + +#include <linux/module.h> +#include <linux/inet.h> + +#include "rtrs-pri.h" +#include "rtrs-log.h" + +MODULE_DESCRIPTION("RDMA Transport Core"); +MODULE_LICENSE("GPL"); + +struct rtrs_iu *rtrs_iu_alloc(u32 queue_size, size_t size, gfp_t gfp_mask, + struct ib_device *dma_dev, + enum dma_data_direction dir, + void (*done)(struct ib_cq *cq, struct ib_wc *wc)) +{ + struct rtrs_iu *ius, *iu; + int i; + + ius = kcalloc(queue_size, sizeof(*ius), gfp_mask); + if (!ius) + return NULL; + for (i = 0; i < queue_size; i++) { + iu = &ius[i]; + iu->direction = dir; + iu->buf = kzalloc(size, gfp_mask); + if (!iu->buf) + goto err; + + iu->dma_addr = ib_dma_map_single(dma_dev, iu->buf, size, dir); + if (ib_dma_mapping_error(dma_dev, iu->dma_addr)) { + kfree(iu->buf); + goto err; + } + + iu->cqe.done = done; + iu->size = size; + } + return ius; +err: + rtrs_iu_free(ius, dma_dev, i); + return NULL; +} +EXPORT_SYMBOL_GPL(rtrs_iu_alloc); + +void rtrs_iu_free(struct rtrs_iu *ius, struct ib_device *ibdev, u32 queue_size) +{ + struct rtrs_iu *iu; + int i; + + if (!ius) + return; + + for (i = 0; i < queue_size; i++) { + iu = &ius[i]; + ib_dma_unmap_single(ibdev, iu->dma_addr, iu->size, iu->direction); + kfree(iu->buf); + } + kfree(ius); +} +EXPORT_SYMBOL_GPL(rtrs_iu_free); + +int rtrs_iu_post_recv(struct rtrs_con *con, struct rtrs_iu *iu) +{ + struct rtrs_sess *sess = con->sess; + struct ib_recv_wr wr; + struct ib_sge list; + + list.addr = iu->dma_addr; + list.length = iu->size; + list.lkey = sess->dev->ib_pd->local_dma_lkey; + + if (list.length == 0) { + rtrs_wrn(con->sess, + "Posting receive work request failed, sg list is empty\n"); + return -EINVAL; + } + wr = (struct ib_recv_wr) { + .wr_cqe = &iu->cqe, + .sg_list = &list, + .num_sge = 1, + }; + + return ib_post_recv(con->qp, &wr, NULL); +} +EXPORT_SYMBOL_GPL(rtrs_iu_post_recv); + +int rtrs_post_recv_empty(struct rtrs_con *con, struct ib_cqe *cqe) +{ + struct ib_recv_wr wr; + + wr = (struct ib_recv_wr) { + .wr_cqe = cqe, + }; + + return ib_post_recv(con->qp, &wr, NULL); +} +EXPORT_SYMBOL_GPL(rtrs_post_recv_empty); + +static int rtrs_post_send(struct ib_qp *qp, struct ib_send_wr *head, + struct ib_send_wr *wr) +{ + if (head) { + struct ib_send_wr *tail = head; + + while (tail->next) + tail = tail->next; + tail->next = wr; + } else { + head = wr; + } + + return ib_post_send(qp, head, NULL); +} + +int rtrs_iu_post_send(struct rtrs_con *con, struct rtrs_iu *iu, size_t size, + struct ib_send_wr *head) +{ + struct rtrs_sess *sess = con->sess; + struct ib_send_wr wr; + struct ib_sge list; + + if (WARN_ON(size == 0)) + return -EINVAL; + + list.addr = iu->dma_addr; + list.length = size; + list.lkey = sess->dev->ib_pd->local_dma_lkey; + + wr = (struct ib_send_wr) { + .wr_cqe = &iu->cqe, + .sg_list = &list, + .num_sge = 1, + .opcode = IB_WR_SEND, + .send_flags = IB_SEND_SIGNALED, + }; + + return rtrs_post_send(con->qp, head, &wr); +} +EXPORT_SYMBOL_GPL(rtrs_iu_post_send); + +int rtrs_iu_post_rdma_write_imm(struct rtrs_con *con, struct rtrs_iu *iu, + struct ib_sge *sge, unsigned int num_sge, + u32 rkey, u64 rdma_addr, u32 imm_data, + enum ib_send_flags flags, + struct ib_send_wr *head) +{ + struct ib_rdma_wr wr; + int i; + + wr = (struct ib_rdma_wr) { + .wr.wr_cqe = &iu->cqe, + .wr.sg_list = sge, + .wr.num_sge = num_sge, + .rkey = rkey, + .remote_addr = rdma_addr, + .wr.opcode = IB_WR_RDMA_WRITE_WITH_IMM, + .wr.ex.imm_data = cpu_to_be32(imm_data), + .wr.send_flags = flags, + }; + + /* + * If one of the sges has 0 size, the operation will fail with a + * length error + */ + for (i = 0; i < num_sge; i++) + if (WARN_ON(sge[i].length == 0)) + return -EINVAL; + + return rtrs_post_send(con->qp, head, &wr.wr); +} +EXPORT_SYMBOL_GPL(rtrs_iu_post_rdma_write_imm); + +int rtrs_post_rdma_write_imm_empty(struct rtrs_con *con, struct ib_cqe *cqe, + u32 imm_data, enum ib_send_flags flags, + struct ib_send_wr *head) +{ + struct ib_rdma_wr wr; + + wr = (struct ib_rdma_wr) { + .wr.wr_cqe = cqe, + .wr.send_flags = flags, + .wr.opcode = IB_WR_RDMA_WRITE_WITH_IMM, + .wr.ex.imm_data = cpu_to_be32(imm_data), + }; + + return rtrs_post_send(con->qp, head, &wr.wr); +} +EXPORT_SYMBOL_GPL(rtrs_post_rdma_write_imm_empty); + +static void qp_event_handler(struct ib_event *ev, void *ctx) +{ + struct rtrs_con *con = ctx; + + switch (ev->event) { + case IB_EVENT_COMM_EST: + rtrs_info(con->sess, "QP event %s (%d) received\n", + ib_event_msg(ev->event), ev->event); + rdma_notify(con->cm_id, IB_EVENT_COMM_EST); + break; + default: + rtrs_info(con->sess, "Unhandled QP event %s (%d) received\n", + ib_event_msg(ev->event), ev->event); + break; + } +} + +static int create_cq(struct rtrs_con *con, int cq_vector, u16 cq_size, + enum ib_poll_context poll_ctx) +{ + struct rdma_cm_id *cm_id = con->cm_id; + struct ib_cq *cq; + + cq = ib_alloc_cq(cm_id->device, con, cq_size, + cq_vector, poll_ctx); + if (IS_ERR(cq)) { + rtrs_err(con->sess, "Creating completion queue failed, errno: %ld\n", + PTR_ERR(cq)); + return PTR_ERR(cq); + } + con->cq = cq; + + return 0; +} + +static int create_qp(struct rtrs_con *con, struct ib_pd *pd, + u32 max_send_wr, u32 max_recv_wr, u32 max_sge) +{ + struct ib_qp_init_attr init_attr = {NULL}; + struct rdma_cm_id *cm_id = con->cm_id; + int ret; + + init_attr.cap.max_send_wr = max_send_wr; + init_attr.cap.max_recv_wr = max_recv_wr; + init_attr.cap.max_recv_sge = 1; + init_attr.event_handler = qp_event_handler; + init_attr.qp_context = con; + init_attr.cap.max_send_sge = max_sge; + + init_attr.qp_type = IB_QPT_RC; + init_attr.send_cq = con->cq; + init_attr.recv_cq = con->cq; + init_attr.sq_sig_type = IB_SIGNAL_REQ_WR; + + ret = rdma_create_qp(cm_id, pd, &init_attr); + if (ret) { + rtrs_err(con->sess, "Creating QP failed, err: %d\n", ret); + return ret; + } + con->qp = cm_id->qp; + + return ret; +} + +int rtrs_cq_qp_create(struct rtrs_sess *sess, struct rtrs_con *con, + u32 max_send_sge, int cq_vector, int cq_size, + u32 max_send_wr, u32 max_recv_wr, + enum ib_poll_context poll_ctx) +{ + int err; + + err = create_cq(con, cq_vector, cq_size, poll_ctx); + if (err) + return err; + + err = create_qp(con, sess->dev->ib_pd, max_send_wr, max_recv_wr, + max_send_sge); + if (err) { + ib_free_cq(con->cq); + con->cq = NULL; + return err; + } + con->sess = sess; + + return 0; +} +EXPORT_SYMBOL_GPL(rtrs_cq_qp_create); + +void rtrs_cq_qp_destroy(struct rtrs_con *con) +{ + if (con->qp) { + rdma_destroy_qp(con->cm_id); + con->qp = NULL; + } + if (con->cq) { + ib_free_cq(con->cq); + con->cq = NULL; + } +} +EXPORT_SYMBOL_GPL(rtrs_cq_qp_destroy); + +static void schedule_hb(struct rtrs_sess *sess) +{ + queue_delayed_work(sess->hb_wq, &sess->hb_dwork, + msecs_to_jiffies(sess->hb_interval_ms)); +} + +void rtrs_send_hb_ack(struct rtrs_sess *sess) +{ + struct rtrs_con *usr_con = sess->con[0]; + u32 imm; + int err; + + imm = rtrs_to_imm(RTRS_HB_ACK_IMM, 0); + err = rtrs_post_rdma_write_imm_empty(usr_con, sess->hb_cqe, imm, + 0, NULL); + if (err) { + sess->hb_err_handler(usr_con); + return; + } +} +EXPORT_SYMBOL_GPL(rtrs_send_hb_ack); + +static void hb_work(struct work_struct *work) +{ + struct rtrs_con *usr_con; + struct rtrs_sess *sess; + u32 imm; + int err; + + sess = container_of(to_delayed_work(work), typeof(*sess), hb_dwork); + usr_con = sess->con[0]; + + if (sess->hb_missed_cnt > sess->hb_missed_max) { + sess->hb_err_handler(usr_con); + return; + } + if (sess->hb_missed_cnt++) { + /* Reschedule work without sending hb */ + schedule_hb(sess); + return; + } + imm = rtrs_to_imm(RTRS_HB_MSG_IMM, 0); + err = rtrs_post_rdma_write_imm_empty(usr_con, sess->hb_cqe, imm, + 0, NULL); + if (err) { + sess->hb_err_handler(usr_con); + return; + } + + schedule_hb(sess); +} + +void rtrs_init_hb(struct rtrs_sess *sess, struct ib_cqe *cqe, + unsigned int interval_ms, unsigned int missed_max, + void (*err_handler)(struct rtrs_con *con), + struct workqueue_struct *wq) +{ + sess->hb_cqe = cqe; + sess->hb_interval_ms = interval_ms; + sess->hb_err_handler = err_handler; + sess->hb_wq = wq; + sess->hb_missed_max = missed_max; + sess->hb_missed_cnt = 0; + INIT_DELAYED_WORK(&sess->hb_dwork, hb_work); +} +EXPORT_SYMBOL_GPL(rtrs_init_hb); + +void rtrs_start_hb(struct rtrs_sess *sess) +{ + schedule_hb(sess); +} +EXPORT_SYMBOL_GPL(rtrs_start_hb); + +void rtrs_stop_hb(struct rtrs_sess *sess) +{ + cancel_delayed_work_sync(&sess->hb_dwork); + sess->hb_missed_cnt = 0; +} +EXPORT_SYMBOL_GPL(rtrs_stop_hb); + +static int rtrs_str_gid_to_sockaddr(const char *addr, size_t len, + short port, struct sockaddr_storage *dst) +{ + struct sockaddr_ib *dst_ib = (struct sockaddr_ib *)dst; + int ret; + + /* + * We can use some of the IPv6 functions since GID is a valid + * IPv6 address format + */ + ret = in6_pton(addr, len, dst_ib->sib_addr.sib_raw, '\0', NULL); + if (ret == 0) + return -EINVAL; + + dst_ib->sib_family = AF_IB; + /* + * Use the same TCP server port number as the IB service ID + * on the IB port space range + */ + dst_ib->sib_sid = cpu_to_be64(RDMA_IB_IP_PS_IB | port); + dst_ib->sib_sid_mask = cpu_to_be64(0xffffffffffffffffULL); + dst_ib->sib_pkey = cpu_to_be16(0xffff); + + return 0; +} + +/** + * rtrs_str_to_sockaddr() - Convert rtrs address string to sockaddr + * @addr: String representation of an addr (IPv4, IPv6 or IB GID): + * - "ip:192.168.1.1" + * - "ip:fe80::200:5aee:feaa:20a2" + * - "gid:fe80::200:5aee:feaa:20a2" + * @len: String address length + * @port: Destination port + * @dst: Destination sockaddr structure + * + * Returns 0 if conversion successful. Non-zero on error. + */ +static int rtrs_str_to_sockaddr(const char *addr, size_t len, + u16 port, struct sockaddr_storage *dst) +{ + if (strncmp(addr, "gid:", 4) == 0) { + return rtrs_str_gid_to_sockaddr(addr + 4, len - 4, port, dst); + } else if (strncmp(addr, "ip:", 3) == 0) { + char port_str[8]; + char *cpy; + int err; + + snprintf(port_str, sizeof(port_str), "%u", port); + cpy = kstrndup(addr + 3, len - 3, GFP_KERNEL); + err = cpy ? inet_pton_with_scope(&init_net, AF_UNSPEC, + cpy, port_str, dst) : -ENOMEM; + kfree(cpy); + + return err; + } + return -EPROTONOSUPPORT; +} + +/** + * sockaddr_to_str() - convert sockaddr to a string. + * @addr: the sockadddr structure to be converted. + * @buf: string containing socket addr. + * @len: string length. + * + * The return value is the number of characters written into buf not + * including the trailing '\0'. If len is == 0 the function returns 0.. + */ +int sockaddr_to_str(const struct sockaddr *addr, char *buf, size_t len) +{ + + switch (addr->sa_family) { + case AF_IB: + return scnprintf(buf, len, "gid:%pI6", + &((struct sockaddr_ib *)addr)->sib_addr.sib_raw); + case AF_INET: + return scnprintf(buf, len, "ip:%pI4", + &((struct sockaddr_in *)addr)->sin_addr); + case AF_INET6: + return scnprintf(buf, len, "ip:%pI6c", + &((struct sockaddr_in6 *)addr)->sin6_addr); + } + return scnprintf(buf, len, "<invalid address family>"); +} +EXPORT_SYMBOL(sockaddr_to_str); + +/** + * rtrs_addr_to_sockaddr() - convert path string "src,dst" or "src@dst" + * to sockaddreses + * @str: string containing source and destination addr of a path + * separated by ',' or '@' I.e. "ip:1.1.1.1,ip:1.1.1.2" or + * "ip:1.1.1.1@ip:1.1.1.2". If str contains only one address it's + * considered to be destination. + * @len: string length + * @port: Destination port number. + * @addr: will be set to the source/destination address or to NULL + * if str doesn't contain any source address. + * + * Returns zero if conversion successful. Non-zero otherwise. + */ +int rtrs_addr_to_sockaddr(const char *str, size_t len, u16 port, + struct rtrs_addr *addr) +{ + const char *d; + + d = strchr(str, ','); + if (!d) + d = strchr(str, '@'); + if (d) { + if (rtrs_str_to_sockaddr(str, d - str, 0, addr->src)) + return -EINVAL; + d += 1; + len -= d - str; + str = d; + + } else { + addr->src = NULL; + } + return rtrs_str_to_sockaddr(str, len, port, addr->dst); +} +EXPORT_SYMBOL(rtrs_addr_to_sockaddr); + +void rtrs_rdma_dev_pd_init(enum ib_pd_flags pd_flags, + struct rtrs_rdma_dev_pd *pool) +{ + WARN_ON(pool->ops && (!pool->ops->alloc ^ !pool->ops->free)); + INIT_LIST_HEAD(&pool->list); + mutex_init(&pool->mutex); + pool->pd_flags = pd_flags; +} +EXPORT_SYMBOL(rtrs_rdma_dev_pd_init); + +void rtrs_rdma_dev_pd_deinit(struct rtrs_rdma_dev_pd *pool) +{ + mutex_destroy(&pool->mutex); + WARN_ON(!list_empty(&pool->list)); +} +EXPORT_SYMBOL(rtrs_rdma_dev_pd_deinit); + +static void dev_free(struct kref *ref) +{ + struct rtrs_rdma_dev_pd *pool; + struct rtrs_ib_dev *dev; + + dev = container_of(ref, typeof(*dev), ref); + pool = dev->pool; + + mutex_lock(&pool->mutex); + list_del(&dev->entry); + mutex_unlock(&pool->mutex); + + if (pool->ops && pool->ops->deinit) + pool->ops->deinit(dev); + + ib_dealloc_pd(dev->ib_pd); + + if (pool->ops && pool->ops->free) + pool->ops->free(dev); + else + kfree(dev); +} + +int rtrs_ib_dev_put(struct rtrs_ib_dev *dev) +{ + return kref_put(&dev->ref, dev_free); +} +EXPORT_SYMBOL(rtrs_ib_dev_put); + +static int rtrs_ib_dev_get(struct rtrs_ib_dev *dev) +{ + return kref_get_unless_zero(&dev->ref); +} + +struct rtrs_ib_dev * +rtrs_ib_dev_find_or_add(struct ib_device *ib_dev, + struct rtrs_rdma_dev_pd *pool) +{ + struct rtrs_ib_dev *dev; + + mutex_lock(&pool->mutex); + list_for_each_entry(dev, &pool->list, entry) { + if (dev->ib_dev->node_guid == ib_dev->node_guid && + rtrs_ib_dev_get(dev)) + goto out_unlock; + } + mutex_unlock(&pool->mutex); + if (pool->ops && pool->ops->alloc) + dev = pool->ops->alloc(); + else + dev = kzalloc(sizeof(*dev), GFP_KERNEL); + if (IS_ERR_OR_NULL(dev)) + goto out_err; + + kref_init(&dev->ref); + dev->pool = pool; + dev->ib_dev = ib_dev; + dev->ib_pd = ib_alloc_pd(ib_dev, pool->pd_flags); + if (IS_ERR(dev->ib_pd)) + goto out_free_dev; + + if (pool->ops && pool->ops->init && pool->ops->init(dev)) + goto out_free_pd; + + mutex_lock(&pool->mutex); + list_add(&dev->entry, &pool->list); +out_unlock: + mutex_unlock(&pool->mutex); + return dev; + +out_free_pd: + ib_dealloc_pd(dev->ib_pd); +out_free_dev: + if (pool->ops && pool->ops->free) + pool->ops->free(dev); + else + kfree(dev); +out_err: + return NULL; +} +EXPORT_SYMBOL(rtrs_ib_dev_find_or_add); diff --git a/drivers/infiniband/ulp/rtrs/rtrs.h b/drivers/infiniband/ulp/rtrs/rtrs.h new file mode 100644 index 000000000..9af750f4d --- /dev/null +++ b/drivers/infiniband/ulp/rtrs/rtrs.h @@ -0,0 +1,196 @@ +/* SPDX-License-Identifier: GPL-2.0-or-later */ +/* + * RDMA Transport Layer + * + * Copyright (c) 2014 - 2018 ProfitBricks GmbH. All rights reserved. + * Copyright (c) 2018 - 2019 1&1 IONOS Cloud GmbH. All rights reserved. + * Copyright (c) 2019 - 2020 1&1 IONOS SE. All rights reserved. + */ +#ifndef RTRS_H +#define RTRS_H + +#include <linux/socket.h> +#include <linux/scatterlist.h> + +struct rtrs_permit; +struct rtrs_clt; +struct rtrs_srv_ctx; +struct rtrs_srv; +struct rtrs_srv_op; + +/* + * RDMA transport (RTRS) client API + */ + +/** + * enum rtrs_clt_link_ev - Events about connectivity state of a client + * @RTRS_CLT_LINK_EV_RECONNECTED Client was reconnected. + * @RTRS_CLT_LINK_EV_DISCONNECTED Client was disconnected. + */ +enum rtrs_clt_link_ev { + RTRS_CLT_LINK_EV_RECONNECTED, + RTRS_CLT_LINK_EV_DISCONNECTED, +}; + +/** + * Source and destination address of a path to be established + */ +struct rtrs_addr { + struct sockaddr_storage *src; + struct sockaddr_storage *dst; +}; + +/** + * rtrs_clt_ops - it holds the link event callback and private pointer. + * @priv: User supplied private data. + * @link_ev: Event notification callback function for connection state changes + * @priv: User supplied data that was passed to rtrs_clt_open() + * @ev: Occurred event + */ +struct rtrs_clt_ops { + void *priv; + void (*link_ev)(void *priv, enum rtrs_clt_link_ev ev); +}; + +struct rtrs_clt *rtrs_clt_open(struct rtrs_clt_ops *ops, + const char *sessname, + const struct rtrs_addr *paths, + size_t path_cnt, u16 port, + size_t pdu_sz, u8 reconnect_delay_sec, + u16 max_segments, + size_t max_segment_size, + s16 max_reconnect_attempts); + +void rtrs_clt_close(struct rtrs_clt *sess); + +/** + * rtrs_permit_to_pdu() - converts rtrs_permit to opaque pdu pointer + * @permit: RTRS permit pointer, it associates the memory allocation for future + * RDMA operation. + */ +void *rtrs_permit_to_pdu(struct rtrs_permit *permit); + +enum { + RTRS_PERMIT_NOWAIT = 0, + RTRS_PERMIT_WAIT = 1, +}; + +/** + * enum rtrs_clt_con_type() type of ib connection to use with a given + * rtrs_permit + * @ADMIN_CON - use connection reserved for "service" messages + * @IO_CON - use a connection reserved for IO + */ +enum rtrs_clt_con_type { + RTRS_ADMIN_CON, + RTRS_IO_CON +}; + +struct rtrs_permit *rtrs_clt_get_permit(struct rtrs_clt *sess, + enum rtrs_clt_con_type con_type, + int wait); + +void rtrs_clt_put_permit(struct rtrs_clt *sess, struct rtrs_permit *permit); + +/** + * rtrs_clt_req_ops - it holds the request confirmation callback + * and a private pointer. + * @priv: User supplied private data. + * @conf_fn: callback function to be called as confirmation + * @priv: User provided data, passed back with corresponding + * @(conf) confirmation. + * @errno: error number. + */ +struct rtrs_clt_req_ops { + void *priv; + void (*conf_fn)(void *priv, int errno); +}; + +int rtrs_clt_request(int dir, struct rtrs_clt_req_ops *ops, + struct rtrs_clt *sess, struct rtrs_permit *permit, + const struct kvec *vec, size_t nr, size_t len, + struct scatterlist *sg, unsigned int sg_cnt); + +/** + * rtrs_attrs - RTRS session attributes + */ +struct rtrs_attrs { + u32 queue_depth; + u32 max_io_size; + u8 sessname[NAME_MAX]; + struct kobject *sess_kobj; +}; + +int rtrs_clt_query(struct rtrs_clt *sess, struct rtrs_attrs *attr); + +/* + * Here goes RTRS server API + */ + +/** + * enum rtrs_srv_link_ev - Server link events + * @RTRS_SRV_LINK_EV_CONNECTED: Connection from client established + * @RTRS_SRV_LINK_EV_DISCONNECTED: Connection was disconnected, all + * connection RTRS resources were freed. + */ +enum rtrs_srv_link_ev { + RTRS_SRV_LINK_EV_CONNECTED, + RTRS_SRV_LINK_EV_DISCONNECTED, +}; + +struct rtrs_srv_ops { + /** + * rdma_ev(): Event notification for RDMA operations + * If the callback returns a value != 0, an error + * message for the data transfer will be sent to + * the client. + + * @sess: Session + * @priv: Private data set by rtrs_srv_set_sess_priv() + * @id: internal RTRS operation id + * @dir: READ/WRITE + * @data: Pointer to (bidirectional) rdma memory area: + * - in case of %RTRS_SRV_RDMA_EV_RECV contains + * data sent by the client + * - in case of %RTRS_SRV_RDMA_EV_WRITE_REQ points + * to the memory area where the response is to be + * written to + * @datalen: Size of the memory area in @data + * @usr: The extra user message sent by the client (%vec) + * @usrlen: Size of the user message + */ + int (*rdma_ev)(struct rtrs_srv *sess, void *priv, + struct rtrs_srv_op *id, int dir, + void *data, size_t datalen, const void *usr, + size_t usrlen); + /** + * link_ev(): Events about connectivity state changes + * If the callback returns != 0 and the event + * %RTRS_SRV_LINK_EV_CONNECTED the corresponding + * session will be destroyed. + * @sess: Session + * @ev: event + * @priv: Private data from user if previously set with + * rtrs_srv_set_sess_priv() + */ + int (*link_ev)(struct rtrs_srv *sess, enum rtrs_srv_link_ev ev, + void *priv); +}; + +struct rtrs_srv_ctx *rtrs_srv_open(struct rtrs_srv_ops *ops, u16 port); + +void rtrs_srv_close(struct rtrs_srv_ctx *ctx); + +bool rtrs_srv_resp_rdma(struct rtrs_srv_op *id, int errno); + +void rtrs_srv_set_sess_priv(struct rtrs_srv *sess, void *priv); + +int rtrs_srv_get_sess_name(struct rtrs_srv *sess, char *sessname, size_t len); + +int rtrs_srv_get_queue_depth(struct rtrs_srv *sess); + +int rtrs_addr_to_sockaddr(const char *str, size_t len, u16 port, + struct rtrs_addr *addr); + +int sockaddr_to_str(const struct sockaddr *addr, char *buf, size_t len); +#endif |