mm,oom_reaper: do not attempt to reap a task twice
authorTetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Fri, 7 Oct 2016 23:58:48 +0000 (16:58 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Sat, 8 Oct 2016 01:46:27 +0000 (18:46 -0700)
"mm, oom_reaper: do not attempt to reap a task twice" tried to give the
OOM reaper one more chance to retry using MMF_OOM_NOT_REAPABLE flag.
But the usefulness of the flag is rather limited and actually never
shown in practice.  If the flag is set, it means that the holder of
mm->mmap_sem cannot call up_write() due to presumably being blocked at
unkillable wait waiting for other thread's memory allocation.  But since
one of threads sharing that mm will queue that mm immediately via
task_will_free_mem() shortcut (otherwise, oom_badness() will select the
same mm again due to oom_score_adj value unchanged), retrying
MMF_OOM_NOT_REAPABLE mm is unlikely helpful.

Let's always set MMF_OOM_REAPED.

Link: http://lkml.kernel.org/r/1472119394-11342-3-git-send-email-mhocko@kernel.org
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Michal Hocko <mhocko@suse.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Vladimir Davydov <vdavydov@parallels.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
include/linux/sched.h
mm/oom_kill.c

index 7543a47..b48cd32 100644 (file)
@@ -525,7 +525,6 @@ static inline int get_dumpable(struct mm_struct *mm)
 #define MMF_HAS_UPROBES                19      /* has uprobes */
 #define MMF_RECALC_UPROBES     20      /* MMF_HAS_UPROBES can be wrong */
 #define MMF_OOM_REAPED         21      /* mm has been already reaped */
-#define MMF_OOM_NOT_REAPABLE   22      /* mm couldn't be reaped */
 
 #define MMF_INIT_MASK          (MMF_DUMPABLE_MASK | MMF_DUMP_FILTER_MASK)
 
index 87fad95..45097f5 100644 (file)
@@ -578,20 +578,11 @@ static void oom_reap_task(struct task_struct *tsk)
        if (attempts <= MAX_OOM_REAP_RETRIES)
                goto done;
 
+       /* Ignore this mm because somebody can't call up_write(mmap_sem). */
+       set_bit(MMF_OOM_REAPED, &mm->flags);
+
        pr_info("oom_reaper: unable to reap pid:%d (%s)\n",
                task_pid_nr(tsk), tsk->comm);
-
-       /*
-        * If we've already tried to reap this task in the past and
-        * failed it probably doesn't make much sense to try yet again
-        * so hide the mm from the oom killer so that it can move on
-        * to another task with a different mm struct.
-        */
-       if (test_and_set_bit(MMF_OOM_NOT_REAPABLE, &mm->flags)) {
-               pr_info("oom_reaper: giving up pid:%d (%s)\n",
-                       task_pid_nr(tsk), tsk->comm);
-               set_bit(MMF_OOM_REAPED, &mm->flags);
-       }
        debug_show_all_locks();
 
 done: