Adding upstream version 241.upstream/241upstream

Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>

1 files changed, 458 insertions, 0 deletions

diff --git a/src/libsystemd/sd-netlink/netlink-socket.c b/src/libsystemd/sd-netlink/netlink-socket.c
new file mode 100644
index 0000000..432e8e8
--- /dev/null
+++ b/src/libsystemd/sd-netlink/netlink-socket.c
@@ -0,0 +1,458 @@
+/* SPDX-License-Identifier: LGPL-2.1+ */
+
+#include <netinet/in.h>
+#include <stdbool.h>
+#include <unistd.h>
+
+#include "sd-netlink.h"
+
+#include "alloc-util.h"
+#include "fd-util.h"
+#include "format-util.h"
+#include "io-util.h"
+#include "missing.h"
+#include "netlink-internal.h"
+#include "netlink-types.h"
+#include "netlink-util.h"
+#include "refcnt.h"
+#include "socket-util.h"
+#include "util.h"
+
+int socket_open(int family) {
+        int fd;
+
+        fd = socket(PF_NETLINK, SOCK_RAW|SOCK_CLOEXEC|SOCK_NONBLOCK, family);
+        if (fd < 0)
+                return -errno;
+
+        return fd_move_above_stdio(fd);
+}
+
+static int broadcast_groups_get(sd_netlink *nl) {
+        _cleanup_free_ uint32_t *groups = NULL;
+        socklen_t len = 0, old_len;
+        unsigned i, j;
+        int r;
+
+        assert(nl);
+        assert(nl->fd >= 0);
+
+        r = getsockopt(nl->fd, SOL_NETLINK, NETLINK_LIST_MEMBERSHIPS, NULL, &len);
+        if (r < 0) {
+                if (errno == ENOPROTOOPT) {
+                        nl->broadcast_group_dont_leave = true;
+                        return 0;
+                } else
+                        return -errno;
+        }
+
+        if (len == 0)
+                return 0;
+
+        groups = new0(uint32_t, len);
+        if (!groups)
+                return -ENOMEM;
+
+        old_len = len;
+
+        r = getsockopt(nl->fd, SOL_NETLINK, NETLINK_LIST_MEMBERSHIPS, groups, &len);
+        if (r < 0)
+                return -errno;
+
+        if (old_len != len)
+                return -EIO;
+
+        r = hashmap_ensure_allocated(&nl->broadcast_group_refs, NULL);
+        if (r < 0)
+                return r;
+
+        for (i = 0; i < len; i++) {
+                for (j = 0; j < sizeof(uint32_t) * 8; j++) {
+                        uint32_t offset;
+                        unsigned group;
+
+                        offset = 1U << j;
+
+                        if (!(groups[i] & offset))
+                                continue;
+
+                        group = i * sizeof(uint32_t) * 8 + j + 1;
+
+                        r = hashmap_put(nl->broadcast_group_refs, UINT_TO_PTR(group), UINT_TO_PTR(1));
+                        if (r < 0)
+                                return r;
+                }
+        }
+
+        return 0;
+}
+
+int socket_bind(sd_netlink *nl) {
+        socklen_t addrlen;
+        int r;
+
+        r = setsockopt_int(nl->fd, SOL_NETLINK, NETLINK_PKTINFO, true);
+        if (r < 0)
+                return r;
+
+        addrlen = sizeof(nl->sockaddr);
+
+        r = bind(nl->fd, &nl->sockaddr.sa, addrlen);
+        /* ignore EINVAL to allow opening an already bound socket */
+        if (r < 0 && errno != EINVAL)
+                return -errno;
+
+        r = getsockname(nl->fd, &nl->sockaddr.sa, &addrlen);
+        if (r < 0)
+                return -errno;
+
+        r = broadcast_groups_get(nl);
+        if (r < 0)
+                return r;
+
+        return 0;
+}
+
+static unsigned broadcast_group_get_ref(sd_netlink *nl, unsigned group) {
+        assert(nl);
+
+        return PTR_TO_UINT(hashmap_get(nl->broadcast_group_refs, UINT_TO_PTR(group)));
+}
+
+static int broadcast_group_set_ref(sd_netlink *nl, unsigned group, unsigned n_ref) {
+        int r;
+
+        assert(nl);
+
+        r = hashmap_replace(nl->broadcast_group_refs, UINT_TO_PTR(group), UINT_TO_PTR(n_ref));
+        if (r < 0)
+                return r;
+
+        return 0;
+}
+
+static int broadcast_group_join(sd_netlink *nl, unsigned group) {
+        int r;
+
+        assert(nl);
+        assert(nl->fd >= 0);
+        assert(group > 0);
+
+        r = setsockopt(nl->fd, SOL_NETLINK, NETLINK_ADD_MEMBERSHIP, &group, sizeof(group));
+        if (r < 0)
+                return -errno;
+
+        return 0;
+}
+
+int socket_broadcast_group_ref(sd_netlink *nl, unsigned group) {
+        unsigned n_ref;
+        int r;
+
+        assert(nl);
+
+        n_ref = broadcast_group_get_ref(nl, group);
+
+        n_ref++;
+
+        r = hashmap_ensure_allocated(&nl->broadcast_group_refs, NULL);
+        if (r < 0)
+                return r;
+
+        r = broadcast_group_set_ref(nl, group, n_ref);
+        if (r < 0)
+                return r;
+
+        if (n_ref > 1)
+                /* not yet in the group */
+                return 0;
+
+        r = broadcast_group_join(nl, group);
+        if (r < 0)
+                return r;
+
+        return 0;
+}
+
+static int broadcast_group_leave(sd_netlink *nl, unsigned group) {
+        int r;
+
+        assert(nl);
+        assert(nl->fd >= 0);
+        assert(group > 0);
+
+        if (nl->broadcast_group_dont_leave)
+                return 0;
+
+        r = setsockopt(nl->fd, SOL_NETLINK, NETLINK_DROP_MEMBERSHIP, &group, sizeof(group));
+        if (r < 0)
+                return -errno;
+
+        return 0;
+}
+
+int socket_broadcast_group_unref(sd_netlink *nl, unsigned group) {
+        unsigned n_ref;
+        int r;
+
+        assert(nl);
+
+        n_ref = broadcast_group_get_ref(nl, group);
+
+        assert(n_ref > 0);
+
+        n_ref--;
+
+        r = broadcast_group_set_ref(nl, group, n_ref);
+        if (r < 0)
+                return r;
+
+        if (n_ref > 0)
+                /* still refs left */
+                return 0;
+
+        r = broadcast_group_leave(nl, group);
+        if (r < 0)
+                return r;
+
+        return 0;
+}
+
+/* returns the number of bytes sent, or a negative error code */
+int socket_write_message(sd_netlink *nl, sd_netlink_message *m) {
+        union {
+                struct sockaddr sa;
+                struct sockaddr_nl nl;
+        } addr = {
+                .nl.nl_family = AF_NETLINK,
+        };
+        ssize_t k;
+
+        assert(nl);
+        assert(m);
+        assert(m->hdr);
+
+        k = sendto(nl->fd, m->hdr, m->hdr->nlmsg_len,
+                        0, &addr.sa, sizeof(addr));
+        if (k < 0)
+                return -errno;
+
+        return k;
+}
+
+static int socket_recv_message(int fd, struct iovec *iov, uint32_t *_group, bool peek) {
+        union sockaddr_union sender;
+        uint8_t cmsg_buffer[CMSG_SPACE(sizeof(struct nl_pktinfo))];
+        struct msghdr msg = {
+                .msg_iov = iov,
+                .msg_iovlen = 1,
+                .msg_name = &sender,
+                .msg_namelen = sizeof(sender),
+                .msg_control = cmsg_buffer,
+                .msg_controllen = sizeof(cmsg_buffer),
+        };
+        struct cmsghdr *cmsg;
+        uint32_t group = 0;
+        ssize_t n;
+
+        assert(fd >= 0);
+        assert(iov);
+
+        n = recvmsg(fd, &msg, MSG_TRUNC | (peek ? MSG_PEEK : 0));
+        if (n < 0) {
+                /* no data */
+                if (errno == ENOBUFS)
+                        log_debug("rtnl: kernel receive buffer overrun");
+                else if (errno == EAGAIN)
+                        log_debug("rtnl: no data in socket");
+
+                return IN_SET(errno, EAGAIN, EINTR) ? 0 : -errno;
+        }
+
+        if (sender.nl.nl_pid != 0) {
+                /* not from the kernel, ignore */
+                log_debug("rtnl: ignoring message from portid %"PRIu32, sender.nl.nl_pid);
+
+                if (peek) {
+                        /* drop the message */
+                        n = recvmsg(fd, &msg, 0);
+                        if (n < 0)
+                                return IN_SET(errno, EAGAIN, EINTR) ? 0 : -errno;
+                }
+
+                return 0;
+        }
+
+        CMSG_FOREACH(cmsg, &msg) {
+                if (cmsg->cmsg_level == SOL_NETLINK &&
+                    cmsg->cmsg_type == NETLINK_PKTINFO &&
+                    cmsg->cmsg_len == CMSG_LEN(sizeof(struct nl_pktinfo))) {
+                        struct nl_pktinfo *pktinfo = (void *)CMSG_DATA(cmsg);
+
+                        /* multi-cast group */
+                        group = pktinfo->group;
+                }
+        }
+
+        if (_group)
+                *_group = group;
+
+        return (int) n;
+}
+
+/* On success, the number of bytes received is returned and *ret points to the received message
+ * which has a valid header and the correct size.
+ * If nothing useful was received 0 is returned.
+ * On failure, a negative error code is returned.
+ */
+int socket_read_message(sd_netlink *rtnl) {
+        _cleanup_(sd_netlink_message_unrefp) sd_netlink_message *first = NULL;
+        struct iovec iov = {};
+        uint32_t group = 0;
+        bool multi_part = false, done = false;
+        struct nlmsghdr *new_msg;
+        size_t len;
+        int r;
+        unsigned i = 0;
+        const NLTypeSystem *type_system_root;
+
+        assert(rtnl);
+        assert(rtnl->rbuffer);
+        assert(rtnl->rbuffer_allocated >= sizeof(struct nlmsghdr));
+
+        type_system_root = type_system_get_root(rtnl->protocol);
+
+        /* read nothing, just get the pending message size */
+        r = socket_recv_message(rtnl->fd, &iov, NULL, true);
+        if (r <= 0)
+                return r;
+        else
+                len = (size_t) r;
+
+        /* make room for the pending message */
+        if (!greedy_realloc((void **)&rtnl->rbuffer,
+                            &rtnl->rbuffer_allocated,
+                            len, sizeof(uint8_t)))
+                return -ENOMEM;
+
+        iov = IOVEC_MAKE(rtnl->rbuffer, rtnl->rbuffer_allocated);
+
+        /* read the pending message */
+        r = socket_recv_message(rtnl->fd, &iov, &group, false);
+        if (r <= 0)
+                return r;
+        else
+                len = (size_t) r;
+
+        if (len > rtnl->rbuffer_allocated)
+                /* message did not fit in read buffer */
+                return -EIO;
+
+        if (NLMSG_OK(rtnl->rbuffer, len) && rtnl->rbuffer->nlmsg_flags & NLM_F_MULTI) {
+                multi_part = true;
+
+                for (i = 0; i < rtnl->rqueue_partial_size; i++) {
+                        if (rtnl_message_get_serial(rtnl->rqueue_partial[i]) ==
+                            rtnl->rbuffer->nlmsg_seq) {
+                                first = rtnl->rqueue_partial[i];
+                                break;
+                        }
+                }
+        }
+
+        for (new_msg = rtnl->rbuffer; NLMSG_OK(new_msg, len) && !done; new_msg = NLMSG_NEXT(new_msg, len)) {
+                _cleanup_(sd_netlink_message_unrefp) sd_netlink_message *m = NULL;
+                const NLType *nl_type;
+
+                if (!group && new_msg->nlmsg_pid != rtnl->sockaddr.nl.nl_pid)
+                        /* not broadcast and not for us */
+                        continue;
+
+                if (new_msg->nlmsg_type == NLMSG_NOOP)
+                        /* silently drop noop messages */
+                        continue;
+
+                if (new_msg->nlmsg_type == NLMSG_DONE) {
+                        /* finished reading multi-part message */
+                        done = true;
+
+                        /* if first is not defined, put NLMSG_DONE into the receive queue. */
+                        if (first)
+                                continue;
+                }
+
+                /* check that we support this message type */
+                r = type_system_get_type(type_system_root, &nl_type, new_msg->nlmsg_type);
+                if (r < 0) {
+                        if (r == -EOPNOTSUPP)
+                                log_debug("sd-netlink: ignored message with unknown type: %i",
+                                          new_msg->nlmsg_type);
+
+                        continue;
+                }
+
+                /* check that the size matches the message type */
+                if (new_msg->nlmsg_len < NLMSG_LENGTH(type_get_size(nl_type))) {
+                        log_debug("sd-netlink: message larger than expected, dropping");
+                        continue;
+                }
+
+                r = message_new_empty(rtnl, &m);
+                if (r < 0)
+                        return r;
+
+                m->broadcast = !!group;
+
+                m->hdr = memdup(new_msg, new_msg->nlmsg_len);
+                if (!m->hdr)
+                        return -ENOMEM;
+
+                /* seal and parse the top-level message */
+                r = sd_netlink_message_rewind(m);
+                if (r < 0)
+                        return r;
+
+                /* push the message onto the multi-part message stack */
+                if (first)
+                        m->next = first;
+                first = TAKE_PTR(m);
+        }
+
+        if (len > 0)
+                log_debug("sd-netlink: discarding %zu bytes of incoming message", len);
+
+        if (!first)
+                return 0;
+
+        if (!multi_part || done) {
+                /* we got a complete message, push it on the read queue */
+                r = rtnl_rqueue_make_room(rtnl);
+                if (r < 0)
+                        return r;
+
+                rtnl->rqueue[rtnl->rqueue_size++] = TAKE_PTR(first);
+
+                if (multi_part && (i < rtnl->rqueue_partial_size)) {
+                        /* remove the message form the partial read queue */
+                        memmove(rtnl->rqueue_partial + i,rtnl->rqueue_partial + i + 1,
+                                sizeof(sd_netlink_message*) * (rtnl->rqueue_partial_size - i - 1));
+                        rtnl->rqueue_partial_size--;
+                }
+
+                return 1;
+        } else {
+                /* we only got a partial multi-part message, push it on the
+                   partial read queue */
+                if (i < rtnl->rqueue_partial_size)
+                        rtnl->rqueue_partial[i] = TAKE_PTR(first);
+                else {
+                        r = rtnl_rqueue_partial_make_room(rtnl);
+                        if (r < 0)
+                                return r;
+
+                        rtnl->rqueue_partial[rtnl->rqueue_partial_size++] = TAKE_PTR(first);
+                }
+
+                return 0;
+        }
+}