From mboxrd@z Thu Jan 1 00:00:00 1970 From: Michal Hocko Subject: Re: [patch 1/2] mm: memcontrol: handle potential crash when rmap races with task exit Date: Thu, 4 Oct 2012 20:49:58 +0200 Message-ID: <20121004184958.GG27536@dhcp22.suse.cz> References: <1349374157-20604-1-git-send-email-hannes@cmpxchg.org> <1349374157-20604-2-git-send-email-hannes@cmpxchg.org> Mime-Version: 1.0 Return-path: Content-Disposition: inline In-Reply-To: <1349374157-20604-2-git-send-email-hannes@cmpxchg.org> Sender: owner-linux-mm@kvack.org List-ID: Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: Johannes Weiner Cc: Andrew Morton , Konstantin Khlebnikov , linux-mm@kvack.org, cgroups@vger.kernel.org, linux-kernel@vger.kernel.org On Thu 04-10-12 14:09:16, Johannes Weiner wrote: > page_referenced() counts only references of mm's that are associated > with the memcg hierarchy that is being reclaimed. However, if it > races with the owner of the mm exiting, mm->owner may be NULL. Don't > crash, just ignore the reference. This seems to be fixed by Hugh's patch 3a981f48 "memcg: fix use_hierarchy css_is_ancestor oops regression" which seems to be merged already. > > Signed-off-by: Johannes Weiner > Cc: stable@kernel.org [3.5] > --- > include/linux/memcontrol.h | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h > index 8d9489f..8686294 100644 > --- a/include/linux/memcontrol.h > +++ b/include/linux/memcontrol.h > @@ -91,7 +91,7 @@ int mm_match_cgroup(const struct mm_struct *mm, const struct mem_cgroup *cgroup) > > rcu_read_lock(); > memcg = mem_cgroup_from_task(rcu_dereference((mm)->owner)); > - match = __mem_cgroup_same_or_subtree(cgroup, memcg); > + match = memcg && __mem_cgroup_same_or_subtree(cgroup, memcg); > rcu_read_unlock(); > return match; > } > -- > 1.7.11.4 > -- 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: email@kvack.org From mboxrd@z Thu Jan 1 00:00:00 1970 From: Johannes Weiner Subject: Re: [patch 1/2] mm: memcontrol: handle potential crash when rmap races with task exit Date: Thu, 4 Oct 2012 16:19:08 -0400 Message-ID: <20121004201908.GA2625@cmpxchg.org> References: <1349374157-20604-1-git-send-email-hannes@cmpxchg.org> <1349374157-20604-2-git-send-email-hannes@cmpxchg.org> <20121004184958.GG27536@dhcp22.suse.cz> Mime-Version: 1.0 Return-path: Content-Disposition: inline In-Reply-To: <20121004184958.GG27536-2MMpYkNvuYDjFM9bn6wA6Q@public.gmane.org> Sender: cgroups-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org List-ID: Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: Michal Hocko Cc: Andrew Morton , Konstantin Khlebnikov , linux-mm-Bw31MaZKKs3YtjvyW6yDsg@public.gmane.org, cgroups-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org On Thu, Oct 04, 2012 at 08:49:58PM +0200, Michal Hocko wrote: > On Thu 04-10-12 14:09:16, Johannes Weiner wrote: > > page_referenced() counts only references of mm's that are associated > > with the memcg hierarchy that is being reclaimed. However, if it > > races with the owner of the mm exiting, mm->owner may be NULL. Don't > > crash, just ignore the reference. > > This seems to be fixed by Hugh's patch 3a981f48 "memcg: fix use_hierarchy > css_is_ancestor oops regression" which seems to be merged already. And look who acked the patch. I'll show myself out... From mboxrd@z Thu Jan 1 00:00:00 1970 From: Johannes Weiner Subject: [patch 0/2] memcg fixups that fell through cracks in 3.5 Date: Thu, 4 Oct 2012 14:09:15 -0400 Message-ID: <1349374157-20604-1-git-send-email-hannes@cmpxchg.org> Return-path: Sender: cgroups-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org List-ID: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: Andrew Morton Cc: Michal Hocko , Konstantin Khlebnikov , linux-mm-Bw31MaZKKs3YtjvyW6yDsg@public.gmane.org, cgroups-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org Hi Andrew, an older version of "mm: memcg: count pte references from every member of the reclaimed hierarchy" patch made it into the 3.5 (c3ac9a8ade65ccbfd145fbff895ae8d8d62d09b0) by accident and I didn't notice until just now. #1 is a fixup on top, tagged for 3.5-stable. #2 is a cleanup you requested but probably missed as it was attached to a mail buried in the thread. For 3.6. Patches against mm.git. From mboxrd@z Thu Jan 1 00:00:00 1970 From: Johannes Weiner Subject: [patch 2/2] mm: memcg: clean up mm_match_cgroup() signature Date: Thu, 4 Oct 2012 14:09:17 -0400 Message-ID: <1349374157-20604-3-git-send-email-hannes@cmpxchg.org> References: <1349374157-20604-1-git-send-email-hannes@cmpxchg.org> Return-path: In-Reply-To: <1349374157-20604-1-git-send-email-hannes@cmpxchg.org> Sender: linux-kernel-owner@vger.kernel.org List-ID: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: Andrew Morton Cc: Michal Hocko , Konstantin Khlebnikov , linux-mm@kvack.org, cgroups@vger.kernel.org, linux-kernel@vger.kernel.org It really should return a boolean for match/no match. And since it takes a memcg, not a cgroup, fix that parameter name as well. Signed-off-by: Johannes Weiner Acked-by: Michal Hocko --- include/linux/memcontrol.h | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h index 8686294..7698182 100644 --- a/include/linux/memcontrol.h +++ b/include/linux/memcontrol.h @@ -84,14 +84,14 @@ extern struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg); extern struct mem_cgroup *mem_cgroup_from_cont(struct cgroup *cont); static inline -int mm_match_cgroup(const struct mm_struct *mm, const struct mem_cgroup *cgroup) +bool mm_match_cgroup(const struct mm_struct *mm, const struct mem_cgroup *memcg) { - struct mem_cgroup *memcg; - int match; + struct mem_cgroup *task_memcg; + bool match; rcu_read_lock(); - memcg = mem_cgroup_from_task(rcu_dereference((mm)->owner)); - match = memcg && __mem_cgroup_same_or_subtree(cgroup, memcg); + task_memcg = mem_cgroup_from_task(rcu_dereference((mm)->owner)); + match = task_memcg && __mem_cgroup_same_or_subtree(memcg, task_memcg); rcu_read_unlock(); return match; } @@ -258,10 +258,10 @@ static inline struct mem_cgroup *try_get_mem_cgroup_from_mm(struct mm_struct *mm return NULL; } -static inline int mm_match_cgroup(struct mm_struct *mm, +static inline bool mm_match_cgroup(struct mm_struct *mm, struct mem_cgroup *memcg) { - return 1; + return true; } static inline int task_in_mem_cgroup(struct task_struct *task, -- 1.7.11.4 From mboxrd@z Thu Jan 1 00:00:00 1970 From: Johannes Weiner Subject: [patch 1/2] mm: memcontrol: handle potential crash when rmap races with task exit Date: Thu, 4 Oct 2012 14:09:16 -0400 Message-ID: <1349374157-20604-2-git-send-email-hannes@cmpxchg.org> References: <1349374157-20604-1-git-send-email-hannes@cmpxchg.org> Return-path: In-Reply-To: <1349374157-20604-1-git-send-email-hannes@cmpxchg.org> Sender: linux-kernel-owner@vger.kernel.org List-ID: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: Andrew Morton Cc: Michal Hocko , Konstantin Khlebnikov , linux-mm@kvack.org, cgroups@vger.kernel.org, linux-kernel@vger.kernel.org page_referenced() counts only references of mm's that are associated with the memcg hierarchy that is being reclaimed. However, if it races with the owner of the mm exiting, mm->owner may be NULL. Don't crash, just ignore the reference. Signed-off-by: Johannes Weiner Cc: stable@kernel.org [3.5] --- include/linux/memcontrol.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h index 8d9489f..8686294 100644 --- a/include/linux/memcontrol.h +++ b/include/linux/memcontrol.h @@ -91,7 +91,7 @@ int mm_match_cgroup(const struct mm_struct *mm, const struct mem_cgroup *cgroup) rcu_read_lock(); memcg = mem_cgroup_from_task(rcu_dereference((mm)->owner)); - match = __mem_cgroup_same_or_subtree(cgroup, memcg); + match = memcg && __mem_cgroup_same_or_subtree(cgroup, memcg); rcu_read_unlock(); return match; } -- 1.7.11.4 From mboxrd@z Thu Jan 1 00:00:00 1970 From: Michal Hocko Subject: Re: [patch 1/2] mm: memcontrol: handle potential crash when rmap races with task exit Date: Fri, 5 Oct 2012 09:05:21 +0200 Message-ID: <20121005070521.GB27757@dhcp22.suse.cz> References: <1349374157-20604-1-git-send-email-hannes@cmpxchg.org> <1349374157-20604-2-git-send-email-hannes@cmpxchg.org> <20121004184958.GG27536@dhcp22.suse.cz> <20121004201908.GA2625@cmpxchg.org> Mime-Version: 1.0 Return-path: Content-Disposition: inline In-Reply-To: <20121004201908.GA2625-druUgvl0LCNAfugRpC6u6w@public.gmane.org> Sender: cgroups-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org List-ID: Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: Johannes Weiner Cc: Andrew Morton , Konstantin Khlebnikov , linux-mm-Bw31MaZKKs3YtjvyW6yDsg@public.gmane.org, cgroups-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org On Thu 04-10-12 16:19:08, Johannes Weiner wrote: > On Thu, Oct 04, 2012 at 08:49:58PM +0200, Michal Hocko wrote: > > On Thu 04-10-12 14:09:16, Johannes Weiner wrote: > > > page_referenced() counts only references of mm's that are associated > > > with the memcg hierarchy that is being reclaimed. However, if it > > > races with the owner of the mm exiting, mm->owner may be NULL. Don't > > > crash, just ignore the reference. > > > > This seems to be fixed by Hugh's patch 3a981f48 "memcg: fix use_hierarchy > > css_is_ancestor oops regression" which seems to be merged already. > > And look who acked the patch. I'll show myself out... My memory is a bit fuzzy but I remember we had two alternatives and Hugh's won. -- Michal Hocko SUSE Labs From mboxrd@z Thu Jan 1 00:00:00 1970 From: Andrew Morton Subject: Re: [patch 2/2] mm: memcg: clean up mm_match_cgroup() signature Date: Fri, 5 Oct 2012 13:53:36 -0700 Message-ID: <20121005135336.1ee7082f.akpm@linux-foundation.org> References: <1349374157-20604-1-git-send-email-hannes@cmpxchg.org> <1349374157-20604-3-git-send-email-hannes@cmpxchg.org> Mime-Version: 1.0 Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <1349374157-20604-3-git-send-email-hannes-druUgvl0LCNAfugRpC6u6w@public.gmane.org> Sender: cgroups-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org List-ID: Content-Type: text/plain; charset="us-ascii" To: Johannes Weiner Cc: Michal Hocko , Konstantin Khlebnikov , linux-mm-Bw31MaZKKs3YtjvyW6yDsg@public.gmane.org, cgroups-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org On Thu, 4 Oct 2012 14:09:17 -0400 Johannes Weiner wrote: > It really should return a boolean for match/no match. And since it > takes a memcg, not a cgroup, fix that parameter name as well. > > --- a/include/linux/memcontrol.h > +++ b/include/linux/memcontrol.h > @@ -84,14 +84,14 @@ extern struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg); > extern struct mem_cgroup *mem_cgroup_from_cont(struct cgroup *cont); > > static inline > -int mm_match_cgroup(const struct mm_struct *mm, const struct mem_cgroup *cgroup) > +bool mm_match_cgroup(const struct mm_struct *mm, const struct mem_cgroup *memcg) > { > - struct mem_cgroup *memcg; > - int match; > + struct mem_cgroup *task_memcg; > + bool match; > > rcu_read_lock(); > - memcg = mem_cgroup_from_task(rcu_dereference((mm)->owner)); > - match = memcg && __mem_cgroup_same_or_subtree(cgroup, memcg); > + task_memcg = mem_cgroup_from_task(rcu_dereference((mm)->owner)); > + match = task_memcg && __mem_cgroup_same_or_subtree(memcg, task_memcg); > rcu_read_unlock(); > return match; > } This needed massaging after droppage of your [1/2]: From: Johannes Weiner Subject: mm: memcg: clean up mm_match_cgroup() signature It really should return a boolean for match/no match. And since it takes a memcg, not a cgroup, fix that parameter name as well. Signed-off-by: Johannes Weiner Acked-by: Michal Hocko --- 1 file changed, 7 insertions(+), 7 deletions(-) index 8686294..7698182 100644 Signed-off-by: Andrew Morton --- include/linux/memcontrol.h | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff -puN include/linux/memcontrol.h~mm-memcg-clean-up-mm_match_cgroup-signature include/linux/memcontrol.h --- a/include/linux/memcontrol.h~mm-memcg-clean-up-mm_match_cgroup-signature +++ a/include/linux/memcontrol.h @@ -84,13 +84,13 @@ extern struct mem_cgroup *parent_mem_cgr extern struct mem_cgroup *mem_cgroup_from_cont(struct cgroup *cont); static inline -int mm_match_cgroup(const struct mm_struct *mm, const struct mem_cgroup *cgroup) +bool mm_match_cgroup(const struct mm_struct *mm, const struct mem_cgroup *memcg) { - struct mem_cgroup *memcg; - int match; + struct mem_cgroup *task_memcg; + bool match; rcu_read_lock(); - memcg = mem_cgroup_from_task(rcu_dereference((mm)->owner)); + task_memcg = mem_cgroup_from_task(rcu_dereference((mm)->owner)); match = __mem_cgroup_same_or_subtree(cgroup, task_memcg); rcu_read_unlock(); return match; @@ -258,10 +258,10 @@ static inline struct mem_cgroup *try_get return NULL; } -static inline int mm_match_cgroup(struct mm_struct *mm, +static inline bool mm_match_cgroup(struct mm_struct *mm, struct mem_cgroup *memcg) { - return 1; + return true; } static inline int task_in_mem_cgroup(struct task_struct *task, _ Also, From: Andrew Morton Subject: mm-memcg-clean-up-mm_match_cgroup-signature-fix mm_match_cgroup is not a macro Cc: Johannes Weiner Cc: Michal Hocko Signed-off-by: Andrew Morton --- include/linux/memcontrol.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff -puN include/linux/memcontrol.h~mm-memcg-clean-up-mm_match_cgroup-signature-fix include/linux/memcontrol.h --- a/include/linux/memcontrol.h~mm-memcg-clean-up-mm_match_cgroup-signature-fix +++ a/include/linux/memcontrol.h @@ -90,7 +90,7 @@ bool mm_match_cgroup(const struct mm_str bool match; rcu_read_lock(); - task_memcg = mem_cgroup_from_task(rcu_dereference((mm)->owner)); + task_memcg = mem_cgroup_from_task(rcu_dereference(mm->owner)); match = __mem_cgroup_same_or_subtree(cgroup, task_memcg); rcu_read_unlock(); return match; _