On Wed 20-10-21 17:21:33, Vasily Averin wrote:

On 20.10.2021 15:41, Michal Hocko wrote:

quoted

On Wed 20-10-21 15:13:46, Vasily Averin wrote:

quoted

ToDo: should we keep task_is_dying() in mem_cgroup_out_of_memory() ?

Signed-off-by: Vasily Averin <redacted>
---
 mm/memcontrol.c | 20 +++++++-------------
 1 file changed, 7 insertions(+), 13 deletions(-)

diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index 6da5020a8656..74a7379dbac1 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c

@@ -239,7 +239,7 @@ enum res_type {
 	     iter != NULL;				\
 	     iter = mem_cgroup_iter(NULL, iter, NULL))
 
-static inline bool should_force_charge(void)
+static inline bool task_is_dying(void)
 {
 	return tsk_is_oom_victim(current) || fatal_signal_pending(current) ||
 		(current->flags & PF_EXITING);

@@ -1575,7 +1575,7 @@ static bool mem_cgroup_out_of_memory(struct mem_cgroup *memcg, gfp_t gfp_mask,
 	 * A few threads which were not waiting at mutex_lock_killable() can
 	 * fail to bail out. Therefore, check again after holding oom_lock.
 	 */
-	ret = should_force_charge() || out_of_memory(&oc);
+	ret = task_is_dying() || out_of_memory(&oc);

Why are you keeping the task_is_dying check here? IIRC I have already
pointed out that out_of_memory already has some means to do a bypass
when needed.

It was a misunderstanding.

Sorry if I made you confused.

I've been waiting for your final decision.

I have no good arguments "pro" or strong objection "contra". 
However, I prefer to keep task_is_dying() so as not to touch other tasks unnecessarily.

One argument for removing it from here is the maintainability. Now you
have a memcg specific check which is not in sync with the oom. E.g.
out_of_memory does task_will_free_mem as the very first thing. You are
also automatically excluding oom killer for cases where that might make
a sense.
-- 
Michal Hocko
SUSE Labs