[PATCH] mm,oom: Set ->signal->oom_mm to all thread groups sharing the victim's mm.

[PATCH] mm,oom: Set ->signal->oom_mm to all thread groups sharing the victim's mm.

[Tetsuo Handa]

Due to commit 696453e66630ad45 ("mm, oom: task_will_free_mem should skip
oom_reaped tasks") and patch "mm,oom: Use ALLOC_OOM for OOM victim's last
second allocation.", thread groups sharing the OOM victim's mm without
setting ->signal->oom_mm before task_will_free_mem(current) is called
might fail to try ALLOC_OOM allocation attempt.

Therefore, make sure that all thread groups sharing the OOM victim's mm can
try ALLOC_OOM allocation attempt by calling mark_oom_victim() on all thread
groups sharing the OOM victim's mm, by calling oom_kill_process() even if
task_will_free_mem(current) is true.

Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Fixes: 696453e66630ad45 ("mm, oom: task_will_free_mem should skip oom_reaped tasks")
---
 mm/oom_kill.c | 49 ++++++++++++++++++++++++++-----------------------
 1 file changed, 26 insertions(+), 23 deletions(-)

diff --git a/mm/oom_kill.c b/mm/oom_kill.c
index 3b0d0fe..399ae36 100644
--- a/mm/oom_kill.c
+++ b/mm/oom_kill.c
@@ -826,6 +826,7 @@ static void oom_kill_process(struct oom_control *oc, const char *message)
 	static DEFINE_RATELIMIT_STATE(oom_rs, DEFAULT_RATELIMIT_INTERVAL,
 					      DEFAULT_RATELIMIT_BURST);
 	bool can_oom_reap = true;
+	bool verbose = false;
 
 	/*
 	 * If the task is already exiting, don't alarm the sysadmin or kill
@@ -833,13 +834,8 @@ static void oom_kill_process(struct oom_control *oc, const char *message)
 	 * so it can die quickly
 	 */
 	task_lock(p);
-	if (task_will_free_mem(p)) {
-		mark_oom_victim(p);
-		wake_oom_reaper(p);
-		task_unlock(p);
-		put_task_struct(p);
-		return;
-	}
+	if (task_will_free_mem(p))
+		goto mark_victims;
 	task_unlock(p);
 
 	if (__ratelimit(&oom_rs))
@@ -876,7 +872,9 @@ static void oom_kill_process(struct oom_control *oc, const char *message)
 	}
 	read_unlock(&tasklist_lock);
 
+	verbose = true;
 	p = find_lock_task_mm(victim);
+ mark_victims:
 	if (!p) {
 		put_task_struct(victim);
 		return;
@@ -891,8 +889,10 @@ static void oom_kill_process(struct oom_control *oc, const char *message)
 	mmgrab(mm);
 
 	/* Raise event before sending signal: task reaper must see this */
-	count_vm_event(OOM_KILL);
-	count_memcg_event_mm(mm, OOM_KILL);
+	if (verbose) {
+		count_vm_event(OOM_KILL);
+		count_memcg_event_mm(mm, OOM_KILL);
+	}
 
 	/*
 	 * We should send SIGKILL before granting access to memory reserves
@@ -901,21 +901,18 @@ static void oom_kill_process(struct oom_control *oc, const char *message)
 	 */
 	do_send_sig_info(SIGKILL, SEND_SIG_FORCED, victim, true);
 	mark_oom_victim(victim);
-	pr_err("Killed process %d (%s) total-vm:%lukB, anon-rss:%lukB, file-rss:%lukB, shmem-rss:%lukB\n",
-		task_pid_nr(victim), victim->comm, K(victim->mm->total_vm),
-		K(get_mm_counter(victim->mm, MM_ANONPAGES)),
-		K(get_mm_counter(victim->mm, MM_FILEPAGES)),
-		K(get_mm_counter(victim->mm, MM_SHMEMPAGES)));
+	if (verbose)
+		pr_err("Killed process %d (%s) total-vm:%lukB, anon-rss:%lukB, file-rss:%lukB, shmem-rss:%lukB\n",
+		       task_pid_nr(victim), victim->comm, K(mm->total_vm),
+		       K(get_mm_counter(mm, MM_ANONPAGES)),
+		       K(get_mm_counter(mm, MM_FILEPAGES)),
+		       K(get_mm_counter(mm, MM_SHMEMPAGES)));
 	task_unlock(victim);
 
 	/*
-	 * Kill all user processes sharing victim->mm in other thread groups, if
-	 * any.  They don't get access to memory reserves, though, to avoid
-	 * depletion of all memory.  This prevents mm->mmap_sem livelock when an
-	 * oom killed thread cannot exit because it requires the semaphore and
-	 * its contended by another thread trying to allocate memory itself.
-	 * That thread will now get access to memory reserves since it has a
-	 * pending fatal signal.
+	 * Kill all user processes, if any, sharing victim->mm in other thread
+	 * groups and grant access to memory reserves. This helps the OOM
+	 * reaper to reclaim memory.
 	 */
 	rcu_read_lock();
 	for_each_process(p) {
@@ -938,6 +935,11 @@ static void oom_kill_process(struct oom_control *oc, const char *message)
 		if (unlikely(p->flags & PF_KTHREAD))
 			continue;
 		do_send_sig_info(SIGKILL, SEND_SIG_FORCED, p, true);
+		t = find_lock_task_mm(p);
+		if (!t)
+			continue;
+		mark_oom_victim(t);
+		task_unlock(t);
 	}
 	rcu_read_unlock();
 
@@ -1018,8 +1020,9 @@ bool out_of_memory(struct oom_control *oc)
 	 * quickly exit and free its memory.
 	 */
 	if (task_will_free_mem(current)) {
-		mark_oom_victim(current);
-		wake_oom_reaper(current);
+		get_task_struct(current);
+		oc->chosen = current;
+		oom_kill_process(oc, "");
 		return true;
 	}
 
-- 
1.8.3.1

--
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>

Re: [PATCH] mm,oom: Set ->signal->oom_mm to all thread groups sharing the victim's mm.

[Michal Hocko]

On Tue 28-11-17 21:41:28, Tetsuo Handa wrote:
> Due to commit 696453e66630ad45 ("mm, oom: task_will_free_mem should skip
> oom_reaped tasks") and patch "mm,oom: Use ALLOC_OOM for OOM victim's last
> second allocation.", thread groups sharing the OOM victim's mm without
> setting ->signal->oom_mm before task_will_free_mem(current) is called
> might fail to try ALLOC_OOM allocation attempt.

Look, this is getting insane. The code complexity grows without any
real users asking for this. While this might look like an interesting
excercise to you I really hate the direction you are heading. This code
will always be just a heuristic and the more complicated it will be the
bigger chances of other side effects there will be as well.

So NACK to this unless I you can show a _real_ usecase that would
_suffer_ by this corner case.

-- 
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>

Re: [PATCH] mm,oom: Set ->signal->oom_mm to all thread groups sharing the victim's mm.

[Tetsuo Handa]

Michal Hocko wrote:
> On Tue 28-11-17 21:41:28, Tetsuo Handa wrote:
> > Due to commit 696453e66630ad45 ("mm, oom: task_will_free_mem should skip
> > oom_reaped tasks") and patch "mm,oom: Use ALLOC_OOM for OOM victim's last
> > second allocation.", thread groups sharing the OOM victim's mm without
> > setting ->signal->oom_mm before task_will_free_mem(current) is called
> > might fail to try ALLOC_OOM allocation attempt.
> 
> Look, this is getting insane. The code complexity grows without any
> real users asking for this.

This is the result of applying "mm,oom: Use ALLOC_OOM for OOM victim's
last second allocation." instead of "mm, oom: task_will_free_mem(current)
should ignore MMF_OOM_SKIP for once." More you go per-mm oriented rather
than per-signal oriented or per-thread oriented, more atomicity will be
needed.

>                             While this might look like an interesting
> excercise to you I really hate the direction you are heading. This code
> will always be just a heuristic and the more complicated it will be the
> bigger chances of other side effects there will be as well.
> 
> So NACK to this unless I you can show a _real_ usecase that would
> _suffer_ by this corner case.

But we send SIGKILL to all thread groups sharing the OOM victim's memory.
This means that (though it might be artificial/malicious) there can be
programs which hit this corner case.

This resembles setting TIF_MEMDIE to all threads at mark_oom_victim(),
and (if I understand correctly) cgroup-aware OOM killer discussion is
after all trying to split oom_kill_process() into "printk()" part and
"non-printk()" part.

Also, please don't preserve outdated comments.

--
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>

[PATCH v2] mm,oom: Set ->signal->oom_mm to all thread groups sharing the victim's mm.

[Tetsuo Handa]

Due to commit 696453e66630ad45 ("mm, oom: task_will_free_mem should skip
oom_reaped tasks") and patch "mm,oom: Use ALLOC_OOM for OOM victim's last
second allocation.", thread groups sharing the OOM victim's mm without
setting ->signal->oom_mm before task_will_free_mem(current) is called
might fail to try ALLOC_OOM allocation attempt.

Therefore, make sure that all thread groups sharing the OOM victim's mm can
try ALLOC_OOM allocation attempt by calling mark_oom_victim() on all thread
groups sharing the OOM victim's mm, by splitting oom_kill_process() into
"select final victim and print message" part and "kill final victim" part.

Roman is proposing similar change for implementing cgroup-aware OOM killer
at http://lkml.kernel.org/r/20171019185218.12663-2-guro@fb.com , and this
patch can be reused for that purpose.

Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Roman Gushchin <guro@fb.com>
Fixes: 696453e66630ad45 ("mm, oom: task_will_free_mem should skip oom_reaped tasks")
Cc: Michal Hocko <mhocko@suse.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Tejun Heo <tj@kernel.org>
---
 mm/oom_kill.c | 129 ++++++++++++++++++++++++++++++----------------------------
 1 file changed, 67 insertions(+), 62 deletions(-)

diff --git a/mm/oom_kill.c b/mm/oom_kill.c
index 3b0d0fe..f859144 100644
--- a/mm/oom_kill.c
+++ b/mm/oom_kill.c
@@ -814,6 +814,64 @@ static bool task_will_free_mem(struct task_struct *task)
 	return ret;
 }
 
+static void __oom_kill_process(struct task_struct *victim)
+{
+	bool can_oom_reap = true;
+	struct task_struct *p;
+	struct task_struct *t;
+	struct mm_struct *mm;
+
+	victim = find_lock_task_mm(victim);
+	if (!victim)
+		return;
+	get_task_struct(victim);
+	/* Get a reference to safely compare mm after task_unlock(victim) */
+	mm = victim->mm;
+	mmgrab(mm);
+	task_unlock(victim);
+
+	/*
+	 * Kill all user processes sharing victim's mm and then grant them
+	 * access to memory reserves.
+	 */
+	rcu_read_lock();
+	for_each_process(p) {
+		if (!process_shares_mm(p, mm))
+			continue;
+		if (is_global_init(p)) {
+			can_oom_reap = false;
+			set_bit(MMF_OOM_SKIP, &mm->flags);
+			pr_info("oom killer %d (%s) has mm pinned by %d (%s)\n",
+				task_pid_nr(victim), victim->comm,
+				task_pid_nr(p), p->comm);
+			continue;
+		}
+		/*
+		 * No use_mm() user needs to read from the userspace so we are
+		 * ok to reap it.
+		 */
+		if (unlikely(p->flags & PF_KTHREAD))
+			continue;
+		/*
+		 * We should send SIGKILL before granting access to memory
+		 * reserves in order to prevent the OOM victim from depleting
+		 * the memory reserves from the user space under its control.
+		 */
+		do_send_sig_info(SIGKILL, SEND_SIG_FORCED, p, true);
+		t = find_lock_task_mm(p);
+		if (!t)
+			continue;
+		mark_oom_victim(t);
+		task_unlock(t);
+	}
+	rcu_read_unlock();
+
+	if (can_oom_reap)
+		wake_oom_reaper(victim);
+	mmdrop(mm);
+	put_task_struct(victim);
+}
+
 static void oom_kill_process(struct oom_control *oc, const char *message)
 {
 	struct task_struct *p = oc->chosen;
@@ -825,7 +883,6 @@ static void oom_kill_process(struct oom_control *oc, const char *message)
 	unsigned int victim_points = 0;
 	static DEFINE_RATELIMIT_STATE(oom_rs, DEFAULT_RATELIMIT_INTERVAL,
 					      DEFAULT_RATELIMIT_BURST);
-	bool can_oom_reap = true;
 
 	/*
 	 * If the task is already exiting, don't alarm the sysadmin or kill
@@ -833,13 +890,8 @@ static void oom_kill_process(struct oom_control *oc, const char *message)
 	 * so it can die quickly
 	 */
 	task_lock(p);
-	if (task_will_free_mem(p)) {
-		mark_oom_victim(p);
-		wake_oom_reaper(p);
-		task_unlock(p);
-		put_task_struct(p);
-		return;
-	}
+	if (task_will_free_mem(p))
+		goto kill_victims;
 	task_unlock(p);
 
 	if (__ratelimit(&oom_rs))
@@ -885,66 +937,20 @@ static void oom_kill_process(struct oom_control *oc, const char *message)
 		put_task_struct(victim);
 		victim = p;
 	}
-
-	/* Get a reference to safely compare mm after task_unlock(victim) */
 	mm = victim->mm;
-	mmgrab(mm);
 
 	/* Raise event before sending signal: task reaper must see this */
 	count_vm_event(OOM_KILL);
 	count_memcg_event_mm(mm, OOM_KILL);
 
-	/*
-	 * We should send SIGKILL before granting access to memory reserves
-	 * in order to prevent the OOM victim from depleting the memory
-	 * reserves from the user space under its control.
-	 */
-	do_send_sig_info(SIGKILL, SEND_SIG_FORCED, victim, true);
-	mark_oom_victim(victim);
 	pr_err("Killed process %d (%s) total-vm:%lukB, anon-rss:%lukB, file-rss:%lukB, shmem-rss:%lukB\n",
-		task_pid_nr(victim), victim->comm, K(victim->mm->total_vm),
-		K(get_mm_counter(victim->mm, MM_ANONPAGES)),
-		K(get_mm_counter(victim->mm, MM_FILEPAGES)),
-		K(get_mm_counter(victim->mm, MM_SHMEMPAGES)));
+	       task_pid_nr(victim), victim->comm, K(mm->total_vm),
+	       K(get_mm_counter(mm, MM_ANONPAGES)),
+	       K(get_mm_counter(mm, MM_FILEPAGES)),
+	       K(get_mm_counter(mm, MM_SHMEMPAGES)));
+kill_victims:
 	task_unlock(victim);
-
-	/*
-	 * Kill all user processes sharing victim->mm in other thread groups, if
-	 * any.  They don't get access to memory reserves, though, to avoid
-	 * depletion of all memory.  This prevents mm->mmap_sem livelock when an
-	 * oom killed thread cannot exit because it requires the semaphore and
-	 * its contended by another thread trying to allocate memory itself.
-	 * That thread will now get access to memory reserves since it has a
-	 * pending fatal signal.
-	 */
-	rcu_read_lock();
-	for_each_process(p) {
-		if (!process_shares_mm(p, mm))
-			continue;
-		if (same_thread_group(p, victim))
-			continue;
-		if (is_global_init(p)) {
-			can_oom_reap = false;
-			set_bit(MMF_OOM_SKIP, &mm->flags);
-			pr_info("oom killer %d (%s) has mm pinned by %d (%s)\n",
-					task_pid_nr(victim), victim->comm,
-					task_pid_nr(p), p->comm);
-			continue;
-		}
-		/*
-		 * No use_mm() user needs to read from the userspace so we are
-		 * ok to reap it.
-		 */
-		if (unlikely(p->flags & PF_KTHREAD))
-			continue;
-		do_send_sig_info(SIGKILL, SEND_SIG_FORCED, p, true);
-	}
-	rcu_read_unlock();
-
-	if (can_oom_reap)
-		wake_oom_reaper(victim);
-
-	mmdrop(mm);
+	__oom_kill_process(victim);
 	put_task_struct(victim);
 }
 #undef K
@@ -1018,8 +1024,7 @@ bool out_of_memory(struct oom_control *oc)
 	 * quickly exit and free its memory.
 	 */
 	if (task_will_free_mem(current)) {
-		mark_oom_victim(current);
-		wake_oom_reaper(current);
+		__oom_kill_process(current);
 		return true;
 	}
 
-- 
1.8.3.1

--
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>

Re: [PATCH v2] mm,oom: Set ->signal->oom_mm to all thread groups sharing the victim's mm.

[Michal Hocko]

On Wed 29-11-17 01:17:15, Tetsuo Handa wrote:
> Due to commit 696453e66630ad45 ("mm, oom: task_will_free_mem should skip
> oom_reaped tasks") and patch "mm,oom: Use ALLOC_OOM for OOM victim's last
> second allocation.", thread groups sharing the OOM victim's mm without
> setting ->signal->oom_mm before task_will_free_mem(current) is called
> might fail to try ALLOC_OOM allocation attempt.

I've already NACKed your previous attempt. Now you are interfering with
the patchset Roman plans to resubmit and cause further potential clashes
for him.

Stop this already!
-- 
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>