Thread (3 messages) 3 messages, 2 authors, 2016-02-25

Re: [PATCH v2] mm,oom: don't abort on exiting processes when selecting a victim.

From: Michal Hocko <mhocko@kernel.org>
Date: 2016-02-17 14:39:21
Also in: lkml

On Wed 17-02-16 23:31:25, Tetsuo Handa wrote:
Currently, oom_scan_process_thread() returns OOM_SCAN_ABORT when there is
a thread which is exiting. But it is possible that that thread is blocked
at down_read(&mm->mmap_sem) in exit_mm() called from do_exit() whereas
one of threads sharing that memory is doing a GFP_KERNEL allocation
between down_write(&mm->mmap_sem) and up_write(&mm->mmap_sem)
(e.g. mmap()).

----------
T1                  T2
                    Calls mmap()
Calls _exit(0)
                    Arrives at vm_mmap_pgoff()
Arrives at do_exit()
Gets PF_EXITING via exit_signals()
                    Calls down_write(&mm->mmap_sem)
                    Calls do_mmap_pgoff()
Calls down_read(&mm->mmap_sem) from exit_mm()
                    Calls out of memory via a GFP_KERNEL allocation but
                    oom_scan_process_thread(T1) returns OOM_SCAN_ABORT
----------

down_read(&mm->mmap_sem) by T1 is waiting for up_write(&mm->mmap_sem) by
T2 while oom_scan_process_thread() by T2 is waiting for T1 to set
T1->mm = NULL. Under such situation, the OOM killer does not choose
a victim, which results in silent OOM livelock problem.

This patch changes oom_scan_process_thread() not to return OOM_SCAN_ABORT
when there is a thread which is exiting.
Thank you for the updated changelog. This makes much more sense now.
This problem exists for quite some time but I would be hesitant to
mark it for stable because the side effects are quite hard to evaluate.
We could e.g. see a premature OOM killer invocation while the currently
exiting task just didn't get to finish and release its mm.
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Michal Hocko <mhocko@suse.com>
quoted hunk ↗ jump to hunk
---
 mm/oom_kill.c | 3 ---
 1 file changed, 3 deletions(-)
diff --git a/mm/oom_kill.c b/mm/oom_kill.c
index cf87153..6e6abaf 100644
--- a/mm/oom_kill.c
+++ b/mm/oom_kill.c
@@ -292,9 +292,6 @@ enum oom_scan_t oom_scan_process_thread(struct oom_control *oc,
 	if (oom_task_origin(task))
 		return OOM_SCAN_SELECT;
 
-	if (task_will_free_mem(task) && !is_sysrq_oom(oc))
-		return OOM_SCAN_ABORT;
-
 	return OOM_SCAN_OK;
 }
 
-- 
1.8.3.1
-- 
Michal Hocko
SUSE Labs

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help