From 77ed2c5745d93416992857d124f35834b62b3e70 Mon Sep 17 00:00:00 2001 From: Tetsuo Handa Date: Tue, 8 Mar 2016 20:01:32 +0900 Subject: [PATCH] android,lowmemorykiller: Don't abuse TIF_MEMDIE. Currently, lowmemorykiller (LMK) is using TIF_MEMDIE for two purposes. One is to remember processes killed by LMK, and the other is to accelerate termination of processes killed by LMK. But since LMK is invoked as a memory shrinker function, there still should be some memory available. It is very likely that memory allocations by processes killed by LMK will succeed without using ALLOC_NO_WATERMARKS via TIF_MEMDIE. Even if their allocations cannot escape from memory allocation loop unless they use ALLOC_NO_WATERMARKS, lowmem_deathpending_timeout can guarantee forward progress by choosing next victim process. On the other hand, mark_oom_victim() assumes that it must be called with oom_lock held and it must not be called after oom_killer_disable() was called. But LMK is calling it without holding oom_lock and checking oom_killer_disabled. It is possible that LMK calls mark_oom_victim() due to allocation requests by kernel threads after current thread returned from oom_killer_disabled(). This will break synchronization for PM/suspend. This patch introduces per a task_struct flag for remembering processes killed by LMK, and replaces TIF_MEMDIE with that flag. By applying this patch, assumption by mark_oom_victim() becomes true. Signed-off-by: Tetsuo Handa Acked-by: Michal Hocko Cc: Arve Hjonnevag Cc: Riley Andrews Signed-off-by: Greg Kroah-Hartman --- drivers/staging/android/lowmemorykiller.c | 9 ++------- include/linux/sched.h | 4 ++++ 2 files changed, 6 insertions(+), 7 deletions(-) diff --git a/drivers/staging/android/lowmemorykiller.c b/drivers/staging/android/lowmemorykiller.c index 2509e5df7244..c79f22425fa8 100644 --- a/drivers/staging/android/lowmemorykiller.c +++ b/drivers/staging/android/lowmemorykiller.c @@ -131,7 +131,7 @@ static unsigned long lowmem_scan(struct shrinker *s, struct shrink_control *sc) if (!p) continue; - if (test_tsk_thread_flag(p, TIF_MEMDIE) && + if (task_lmk_waiting(p) && p->mm && time_before_eq(jiffies, lowmem_deathpending_timeout)) { task_unlock(p); rcu_read_unlock(); @@ -162,13 +162,8 @@ static unsigned long lowmem_scan(struct shrinker *s, struct shrink_control *sc) if (selected) { task_lock(selected); send_sig(SIGKILL, selected, 0); - /* - * FIXME: lowmemorykiller shouldn't abuse global OOM killer - * infrastructure. There is no real reason why the selected - * task should have access to the memory reserves. - */ if (selected->mm) - mark_oom_victim(selected); + task_set_lmk_waiting(selected); task_unlock(selected); lowmem_print(1, "Killing '%s' (%d), adj %hd,\n" " to free %ldkB on behalf of '%s' (%d) because\n" diff --git a/include/linux/sched.h b/include/linux/sched.h index 60bba7e032dc..9dff190e6a0a 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -2184,6 +2184,7 @@ static inline void memalloc_noio_restore(unsigned int flags) #define PFA_NO_NEW_PRIVS 0 /* May not gain new privileges. */ #define PFA_SPREAD_PAGE 1 /* Spread page cache over cpuset */ #define PFA_SPREAD_SLAB 2 /* Spread some slab caches over cpuset */ +#define PFA_LMK_WAITING 3 /* Lowmemorykiller is waiting */ #define TASK_PFA_TEST(name, func) \ @@ -2207,6 +2208,9 @@ TASK_PFA_TEST(SPREAD_SLAB, spread_slab) TASK_PFA_SET(SPREAD_SLAB, spread_slab) TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab) +TASK_PFA_TEST(LMK_WAITING, lmk_waiting) +TASK_PFA_SET(LMK_WAITING, lmk_waiting) + /* * task->jobctl flags */ -- 2.20.1