oom: fix tasklist_lock leak
authorKOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Thu, 19 Aug 2010 21:13:39 +0000 (14:13 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 20 Aug 2010 16:34:55 +0000 (09:34 -0700)
Commit 0aad4b3124 ("oom: fold __out_of_memory into out_of_memory")
introduced a tasklist_lock leak.  Then it caused following obvious
danger warnings and panic.

    ================================================
    [ BUG: lock held when returning to user space! ]
    ------------------------------------------------
    rsyslogd/1422 is leaving the kernel with locks still held!
    1 lock held by rsyslogd/1422:
     #0:  (tasklist_lock){.+.+.+}, at: [<ffffffff810faf64>] out_of_memory+0x164/0x3f0
    BUG: scheduling while atomic: rsyslogd/1422/0x00000002
    INFO: lockdep is turned off.

This patch fixes it.

Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Reviewed-by: Minchan Kim <minchan.kim@gmail.com>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/oom_kill.c

index 17d48a6..c48c5ef 100644 (file)
@@ -646,6 +646,7 @@ void out_of_memory(struct zonelist *zonelist, gfp_t gfp_mask,
        unsigned long freed = 0;
        unsigned int points;
        enum oom_constraint constraint = CONSTRAINT_NONE;
+       int killed = 0;
 
        blocking_notifier_call_chain(&oom_notify_list, 0, &freed);
        if (freed > 0)
@@ -683,7 +684,7 @@ void out_of_memory(struct zonelist *zonelist, gfp_t gfp_mask,
                if (!oom_kill_process(current, gfp_mask, order, 0, totalpages,
                                NULL, nodemask,
                                "Out of memory (oom_kill_allocating_task)"))
-                       return;
+                       goto out;
        }
 
 retry:
@@ -691,7 +692,7 @@ retry:
                        constraint == CONSTRAINT_MEMORY_POLICY ? nodemask :
                                                                 NULL);
        if (PTR_ERR(p) == -1UL)
-               return;
+               goto out;
 
        /* Found nothing?!?! Either we hang forever, or we panic. */
        if (!p) {
@@ -703,13 +704,15 @@ retry:
        if (oom_kill_process(p, gfp_mask, order, points, totalpages, NULL,
                                nodemask, "Out of memory"))
                goto retry;
+       killed = 1;
+out:
        read_unlock(&tasklist_lock);
 
        /*
         * Give "p" a good chance of killing itself before we
         * retry to allocate memory unless "p" is current
         */
-       if (!test_thread_flag(TIF_MEMDIE))
+       if (killed && !test_thread_flag(TIF_MEMDIE))
                schedule_timeout_uninterruptible(1);
 }