Merge branch 'linus' into core/rcu

2 files changed, 73 insertions, 22 deletions

diff --git a/net/unix/af_unix.c b/net/unix/af_unix.c
index 4d3c6071b9a4..66d5ac4773ab 100644
--- a/net/unix/af_unix.c
+++ b/net/unix/af_unix.c
@@ -1302,14 +1302,23 @@ static void unix_destruct_fds(struct sk_buff *skb)
         sock_wfree(skb);
 }
 
-static void unix_attach_fds(struct scm_cookie *scm, struct sk_buff *skb)
+static int unix_attach_fds(struct scm_cookie *scm, struct sk_buff *skb)
 {
         int i;
+
+        /*
+         * Need to duplicate file references for the sake of garbage
+         * collection.  Otherwise a socket in the fps might become a
+         * candidate for GC while the skb is not yet queued.
+         */
+        UNIXCB(skb).fp = scm_fp_dup(scm->fp);
+        if (!UNIXCB(skb).fp)
+                return -ENOMEM;
+
         for (i=scm->fp->count-1; i>=0; i--)
                 unix_inflight(scm->fp->fp[i]);
-        UNIXCB(skb).fp = scm->fp;
         skb->destructor = unix_destruct_fds;
-        scm->fp = NULL;
+        return 0;
 }
 
 /*
@@ -1334,6 +1343,7 @@ static int unix_dgram_sendmsg(struct kiocb *kiocb, struct socket *sock,
 
         if (NULL == siocb->scm)
                 siocb->scm = &tmp_scm;
+        wait_for_unix_gc();
         err = scm_send(sock, msg, siocb->scm);
         if (err < 0)
                 return err;
@@ -1368,8 +1378,11 @@ static int unix_dgram_sendmsg(struct kiocb *kiocb, struct socket *sock,
                 goto out;
 
         memcpy(UNIXCREDS(skb), &siocb->scm->creds, sizeof(struct ucred));
-        if (siocb->scm->fp)
-                unix_attach_fds(siocb->scm, skb);
+        if (siocb->scm->fp) {
+                err = unix_attach_fds(siocb->scm, skb);
+                if (err)
+                        goto out_free;
+        }
         unix_get_secdata(siocb->scm, skb);
 
         skb_reset_transport_header(skb);
@@ -1481,6 +1494,7 @@ static int unix_stream_sendmsg(struct kiocb *kiocb, struct socket *sock,
 
         if (NULL == siocb->scm)
                 siocb->scm = &tmp_scm;
+        wait_for_unix_gc();
         err = scm_send(sock, msg, siocb->scm);
         if (err < 0)
                 return err;
@@ -1538,8 +1552,13 @@ static int unix_stream_sendmsg(struct kiocb *kiocb, struct socket *sock,
                 size = min_t(int, size, skb_tailroom(skb));
 
                 memcpy(UNIXCREDS(skb), &siocb->scm->creds, sizeof(struct ucred));
-                if (siocb->scm->fp)
-                        unix_attach_fds(siocb->scm, skb);
+                if (siocb->scm->fp) {
+                        err = unix_attach_fds(siocb->scm, skb);
+                        if (err) {
+                                kfree_skb(skb);
+                                goto out_err;
+                        }
+                }
 
                 if ((err = memcpy_fromiovec(skb_put(skb,size), msg->msg_iov, size)) != 0) {
                         kfree_skb(skb);
diff --git a/net/unix/garbage.c b/net/unix/garbage.c
index 2a27b84f740b..abb3ab34cb1e 100644
--- a/net/unix/garbage.c
+++ b/net/unix/garbage.c
@@ -80,6 +80,7 @@
 #include <linux/file.h>
 #include <linux/proc_fs.h>
 #include <linux/mutex.h>
+#include <linux/wait.h>
 
 #include <net/sock.h>
 #include <net/af_unix.h>
@@ -91,6 +92,7 @@
 static LIST_HEAD(gc_inflight_list);
 static LIST_HEAD(gc_candidates);
 static DEFINE_SPINLOCK(unix_gc_lock);
+static DECLARE_WAIT_QUEUE_HEAD(unix_gc_wait);
 
 unsigned int unix_tot_inflight;
 
@@ -186,8 +188,17 @@ static void scan_inflight(struct sock *x, void (*func)(struct unix_sock *),
                                  */
                                 struct sock *sk = unix_get_socket(*fp++);
                                 if (sk) {
-                                        hit = true;
-                                        func(unix_sk(sk));
+                                        struct unix_sock *u = unix_sk(sk);
+
+                                        /*
+                                         * Ignore non-candidates, they could
+                                         * have been added to the queues after
+                                         * starting the garbage collection
+                                         */
+                                        if (u->gc_candidate) {
+                                                hit = true;
+                                                func(u);
+                                        }
                                 }
                         }
                         if (hit && hitlist != NULL) {
@@ -249,24 +260,29 @@ static void inc_inflight_move_tail(struct unix_sock *u)
 {
         atomic_long_inc(&u->inflight);
         /*
-         * If this is still a candidate, move it to the end of the
-         * list, so that it's checked even if it was already passed
-         * over
+         * If this still might be part of a cycle, move it to the end
+         * of the list, so that it's checked even if it was already
+         * passed over
          */
-        if (u->gc_candidate)
+        if (u->gc_maybe_cycle)
                 list_move_tail(&u->link, &gc_candidates);
 }
 
-/* The external entry point: unix_gc() */
+static bool gc_in_progress = false;
 
-void unix_gc(void)
+void wait_for_unix_gc(void)
 {
-        static bool gc_in_progress = false;
+        wait_event(unix_gc_wait, gc_in_progress == false);
+}
 
+/* The external entry point: unix_gc() */
+void unix_gc(void)
+{
         struct unix_sock *u;
         struct unix_sock *next;
         struct sk_buff_head hitlist;
         struct list_head cursor;
+        LIST_HEAD(not_cycle_list);
 
         spin_lock(&unix_gc_lock);
 
@@ -282,10 +298,14 @@ void unix_gc(void)
          *
          * Holding unix_gc_lock will protect these candidates from
          * being detached, and hence from gaining an external
-         * reference.  This also means, that since there are no
-         * possible receivers, the receive queues of these sockets are
-         * static during the GC, even though the dequeue is done
-         * before the detach without atomicity guarantees.
+         * reference.  Since there are no possible receivers, all
+         * buffers currently on the candidates' queues stay there
+         * during the garbage collection.
+         *
+         * We also know that no new candidate can be added onto the
+         * receive queues.  Other, non candidate sockets _can_ be
+         * added to queue, so we must make sure only to touch
+         * candidates.
          */
         list_for_each_entry_safe(u, next, &gc_inflight_list, link) {
                 long total_refs;
@@ -299,6 +319,7 @@ void unix_gc(void)
                 if (total_refs == inflight_refs) {
                         list_move_tail(&u->link, &gc_candidates);
                         u->gc_candidate = 1;
+                        u->gc_maybe_cycle = 1;
                 }
         }
 
@@ -325,14 +346,24 @@ void unix_gc(void)
                 list_move(&cursor, &u->link);
 
                 if (atomic_long_read(&u->inflight) > 0) {
-                        list_move_tail(&u->link, &gc_inflight_list);
-                        u->gc_candidate = 0;
+                        list_move_tail(&u->link, &not_cycle_list);
+                        u->gc_maybe_cycle = 0;
                         scan_children(&u->sk, inc_inflight_move_tail, NULL);
                 }
         }
         list_del(&cursor);
 
         /*
+         * not_cycle_list contains those sockets which do not make up a
+         * cycle.  Restore these to the inflight list.
+         */
+        while (!list_empty(&not_cycle_list)) {
+                u = list_entry(not_cycle_list.next, struct unix_sock, link);
+                u->gc_candidate = 0;
+                list_move_tail(&u->link, &gc_inflight_list);
+        }
+
+        /*
          * Now gc_candidates contains only garbage.  Restore original
          * inflight counters for these as well, and remove the skbuffs
          * which are creating the cycle(s).
@@ -351,6 +382,7 @@ void unix_gc(void)
         /* All candidates should have been detached by now. */
         BUG_ON(!list_empty(&gc_candidates));
         gc_in_progress = false;
+        wake_up(&unix_gc_wait);
 
  out:
         spin_unlock(&unix_gc_lock);