based on oom-introduce-find_lock_task_mm-to-fix-mm-false-positives.patch
tested on mm-of-the-moment snapshot 2010-06-11-16-40.
==
From: KAMEZAWA Hiroyuki <[email protected]>
When the OOM killer scans task, it check a task is under memcg or
not when it's called via memcg's context.
But, as Oleg pointed out, a thread group leader may have NULL ->mm
and task_in_mem_cgroup() may do wrong decision. We have to use
find_lock_task_mm() in memcg as generic OOM-Killer does.
Cc: Oleg Nesterov <[email protected]>
Cc: Daisuke Nishimura <[email protected]>
Cc: Balbir Singh <[email protected]>
Signed-off-by: KAMEZAWA Hiroyuki <[email protected]>
---
include/linux/oom.h | 2 ++
mm/memcontrol.c | 10 +++++++---
mm/oom_kill.c | 8 ++++++--
3 files changed, 15 insertions(+), 5 deletions(-)
Index: mmotm-2.6.35-0611/include/linux/oom.h
===================================================================
--- mmotm-2.6.35-0611.orig/include/linux/oom.h
+++ mmotm-2.6.35-0611/include/linux/oom.h
@@ -45,6 +45,8 @@ static inline void oom_killer_enable(voi
oom_killer_disabled = false;
}
+extern struct task_struct *find_lock_task_mm(struct task_struct *p);
+
/* sysctls */
extern int sysctl_oom_dump_tasks;
extern int sysctl_oom_kill_allocating_task;
Index: mmotm-2.6.35-0611/mm/memcontrol.c
===================================================================
--- mmotm-2.6.35-0611.orig/mm/memcontrol.c
+++ mmotm-2.6.35-0611/mm/memcontrol.c
@@ -47,6 +47,7 @@
#include <linux/mm_inline.h>
#include <linux/page_cgroup.h>
#include <linux/cpu.h>
+#include <linux/oom.h>
#include "internal.h"
#include <asm/uaccess.h>
@@ -838,10 +839,13 @@ int task_in_mem_cgroup(struct task_struc
{
int ret;
struct mem_cgroup *curr = NULL;
+ struct task_struct *p;
- task_lock(task);
- curr = try_get_mem_cgroup_from_mm(task->mm);
- task_unlock(task);
+ p = find_lock_task_mm(task);
+ if (!p)
+ return 0;
+ curr = try_get_mem_cgroup_from_mm(p->mm);
+ task_unlock(p);
if (!curr)
return 0;
/*
Index: mmotm-2.6.35-0611/mm/oom_kill.c
===================================================================
--- mmotm-2.6.35-0611.orig/mm/oom_kill.c
+++ mmotm-2.6.35-0611/mm/oom_kill.c
@@ -81,13 +81,17 @@ static bool has_intersects_mems_allowed(
}
#endif /* CONFIG_NUMA */
-/*
+/**
+ * find_lock_task_mm - Checking a process which a task belongs to has valid mm
+ * and return a locked task which has a valid pointer to mm.
+ *
+ * @p: the task of a process to be checked.
* The process p may have detached its own ->mm while exiting or through
* use_mm(), but one or more of its subthreads may still have a valid
* pointer. Return p, or any of its subthreads with a valid ->mm, with
* task_lock() held.
*/
-static struct task_struct *find_lock_task_mm(struct task_struct *p)
+struct task_struct *find_lock_task_mm(struct task_struct *p)
{
struct task_struct *t = p;
Hi, Kame.
On Tue, Jun 15, 2010 at 3:24 PM, KAMEZAWA Hiroyuki
<[email protected]> wrote:
>
> based on oom-introduce-find_lock_task_mm-to-fix-mm-false-positives.patch
> tested on mm-of-the-moment snapshot 2010-06-11-16-40.
>
> ==
> From: KAMEZAWA Hiroyuki <[email protected]>
>
> When the OOM killer scans task, it check a task is under memcg or
> not when it's called via memcg's context.
>
> But, as Oleg pointed out, a thread group leader may have NULL ->mm
> and task_in_mem_cgroup() may do wrong decision. We have to use
> find_lock_task_mm() in memcg as generic OOM-Killer does.
>
> Cc: Oleg Nesterov <[email protected]>
> Cc: Daisuke Nishimura <[email protected]>
> Cc: Balbir Singh <[email protected]>
> Signed-off-by: KAMEZAWA Hiroyuki <[email protected]>
Reviewed-by: Minchan Kim <[email protected]>
I have a trivial comment below.
> ---
> include/linux/oom.h | 2 ++
> mm/memcontrol.c | 10 +++++++---
> mm/oom_kill.c | 8 ++++++--
> 3 files changed, 15 insertions(+), 5 deletions(-)
>
> Index: mmotm-2.6.35-0611/include/linux/oom.h
> ===================================================================
> --- mmotm-2.6.35-0611.orig/include/linux/oom.h
> +++ mmotm-2.6.35-0611/include/linux/oom.h
> @@ -45,6 +45,8 @@ static inline void oom_killer_enable(voi
> oom_killer_disabled = false;
> }
>
> +extern struct task_struct *find_lock_task_mm(struct task_struct *p);
> +
> /* sysctls */
> extern int sysctl_oom_dump_tasks;
> extern int sysctl_oom_kill_allocating_task;
> Index: mmotm-2.6.35-0611/mm/memcontrol.c
> ===================================================================
> --- mmotm-2.6.35-0611.orig/mm/memcontrol.c
> +++ mmotm-2.6.35-0611/mm/memcontrol.c
> @@ -47,6 +47,7 @@
> #include <linux/mm_inline.h>
> #include <linux/page_cgroup.h>
> #include <linux/cpu.h>
> +#include <linux/oom.h>
> #include "internal.h"
>
> #include <asm/uaccess.h>
> @@ -838,10 +839,13 @@ int task_in_mem_cgroup(struct task_struc
> {
> int ret;
> struct mem_cgroup *curr = NULL;
> + struct task_struct *p;
>
> - task_lock(task);
> - curr = try_get_mem_cgroup_from_mm(task->mm);
> - task_unlock(task);
> + p = find_lock_task_mm(task);
> + if (!p)
> + return 0;
> + curr = try_get_mem_cgroup_from_mm(p->mm);
> + task_unlock(p);
> if (!curr)
> return 0;
> /*
> Index: mmotm-2.6.35-0611/mm/oom_kill.c
> ===================================================================
> --- mmotm-2.6.35-0611.orig/mm/oom_kill.c
> +++ mmotm-2.6.35-0611/mm/oom_kill.c
> @@ -81,13 +81,17 @@ static bool has_intersects_mems_allowed(
> }
> #endif /* CONFIG_NUMA */
>
> -/*
> +/**
> + * find_lock_task_mm - Checking a process which a task belongs to has valid mm
> + * and return a locked task which has a valid pointer to mm.
> + *
This comment should have been another patch.
BTW, below comment uses "subthread" word.
Personally it's easy to understand function's goal to me. :)
How about following as?
Checking a process which has any subthread with vaild mm
....
> + * @p: the task of a process to be checked.
> * The process p may have detached its own ->mm while exiting or through
> * use_mm(), but one or more of its subthreads may still have a valid
> * pointer. Return p, or any of its subthreads with a valid ->mm, with
> * task_lock() held.
> */
> -static struct task_struct *find_lock_task_mm(struct task_struct *p)
> +struct task_struct *find_lock_task_mm(struct task_struct *p)
> {
> struct task_struct *t = p;
>
>
> --
> To unsubscribe, send a message with 'unsubscribe linux-mm' in
> the body to [email protected]. For more info on Linux MM,
> see: http://www.linux-mm.org/ .
> Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
>
--
Kind regards,
Minchan Kim
On Tue, 15 Jun 2010 18:59:25 +0900
Minchan Kim <[email protected]> wrote:
> > -/*
> > +/**
> > + * find_lock_task_mm - Checking a process which a task belongs to has valid mm
> > + * and return a locked task which has a valid pointer to mm.
> > + *
>
> This comment should have been another patch.
> BTW, below comment uses "subthread" word.
> Personally it's easy to understand function's goal to me. :)
>
> How about following as?
> Checking a process which has any subthread with vaild mm
> ....
>
Sure. thank you. v2 is here. I removed unnecessary parts.
==
From: KAMEZAWA Hiroyuki <[email protected]>
When the OOM killer scans task, it check a task is under memcg or
not when it's called via memcg's context.
But, as Oleg pointed out, a thread group leader may have NULL ->mm
and task_in_mem_cgroup() may do wrong decision. We have to use
find_lock_task_mm() in memcg as generic OOM-Killer does.
Changelog:
- removed unnecessary changes in comments.
Cc: Oleg Nesterov <[email protected]>
Cc: Daisuke Nishimura <[email protected]>
Cc: Balbir Singh <[email protected]>
Reviewed-by: Minchan Kim <[email protected]>
Signed-off-by: KAMEZAWA Hiroyuki <[email protected]>
---
include/linux/oom.h | 2 ++
mm/memcontrol.c | 10 +++++++---
mm/oom_kill.c | 2 +-
3 files changed, 10 insertions(+), 4 deletions(-)
Index: mmotm-2.6.35-0611/include/linux/oom.h
===================================================================
--- mmotm-2.6.35-0611.orig/include/linux/oom.h
+++ mmotm-2.6.35-0611/include/linux/oom.h
@@ -45,6 +45,8 @@ static inline void oom_killer_enable(voi
oom_killer_disabled = false;
}
+extern struct task_struct *find_lock_task_mm(struct task_struct *p);
+
/* sysctls */
extern int sysctl_oom_dump_tasks;
extern int sysctl_oom_kill_allocating_task;
Index: mmotm-2.6.35-0611/mm/memcontrol.c
===================================================================
--- mmotm-2.6.35-0611.orig/mm/memcontrol.c
+++ mmotm-2.6.35-0611/mm/memcontrol.c
@@ -47,6 +47,7 @@
#include <linux/mm_inline.h>
#include <linux/page_cgroup.h>
#include <linux/cpu.h>
+#include <linux/oom.h>
#include "internal.h"
#include <asm/uaccess.h>
@@ -838,10 +839,13 @@ int task_in_mem_cgroup(struct task_struc
{
int ret;
struct mem_cgroup *curr = NULL;
+ struct task_struct *p;
- task_lock(task);
- curr = try_get_mem_cgroup_from_mm(task->mm);
- task_unlock(task);
+ p = find_lock_task_mm(task);
+ if (!p)
+ return 0;
+ curr = try_get_mem_cgroup_from_mm(p->mm);
+ task_unlock(p);
if (!curr)
return 0;
/*
Index: mmotm-2.6.35-0611/mm/oom_kill.c
===================================================================
--- mmotm-2.6.35-0611.orig/mm/oom_kill.c
+++ mmotm-2.6.35-0611/mm/oom_kill.c
@@ -87,7 +87,7 @@ static bool has_intersects_mems_allowed(
* pointer. Return p, or any of its subthreads with a valid ->mm, with
* task_lock() held.
*/
-static struct task_struct *find_lock_task_mm(struct task_struct *p)
+struct task_struct *find_lock_task_mm(struct task_struct *p)
{
struct task_struct *t = p;
On Wed, Jun 16, 2010 at 5:33 AM, KAMEZAWA Hiroyuki
<[email protected]> wrote:
> On Tue, 15 Jun 2010 18:59:25 +0900
> Minchan Kim <[email protected]> wrote:
>
>> > -/*
>> > +/**
>> > + * find_lock_task_mm - Checking a process which a task belongs to has valid mm
>> > + * and return a locked task which has a valid pointer to mm.
>> > + *
>>
>> This comment should have been another patch.
>> BTW, below comment uses "subthread" word.
>> Personally it's easy to understand function's goal to me. :)
>>
>> How about following as?
>> Checking a process which has any subthread with vaild mm
>> ....
>>
> Sure. thank you. v2 is here. I removed unnecessary parts.
>
> ==
> From: KAMEZAWA Hiroyuki <[email protected]>
>
> When the OOM killer scans task, it check a task is under memcg or
> not when it's called via memcg's context.
>
> But, as Oleg pointed out, a thread group leader may have NULL ->mm
> and task_in_mem_cgroup() may do wrong decision. We have to use
> find_lock_task_mm() in memcg as generic OOM-Killer does.
>
> Changelog:
> ?- removed unnecessary changes in comments.
>
mm->owner solves the same problem, but unfortunately we have task
based selection in OOM killer, so we need this patch. It is quite
ironic that we find the mm from the task and then eventually the task
back from mm->owner and then the mem cgroup. If we already know the mm
from oom_kill.c, I think we can change the function to work off of
that. mm->owner->cgroup..no?
Balbir
On Wed, 16 Jun 2010 10:12:20 +0530
Balbir Singh <[email protected]> wrote:
> On Wed, Jun 16, 2010 at 5:33 AM, KAMEZAWA Hiroyuki
> <[email protected]> wrote:
> > On Tue, 15 Jun 2010 18:59:25 +0900
> > Minchan Kim <[email protected]> wrote:
> >
> >> > -/*
> >> > +/**
> >> > + * find_lock_task_mm - Checking a process which a task belongs to has valid mm
> >> > + * and return a locked task which has a valid pointer to mm.
> >> > + *
> >>
> >> This comment should have been another patch.
> >> BTW, below comment uses "subthread" word.
> >> Personally it's easy to understand function's goal to me. :)
> >>
> >> How about following as?
> >> Checking a process which has any subthread with vaild mm
> >> ....
> >>
> > Sure. thank you. v2 is here. I removed unnecessary parts.
> >
> > ==
> > From: KAMEZAWA Hiroyuki <[email protected]>
> >
> > When the OOM killer scans task, it check a task is under memcg or
> > not when it's called via memcg's context.
> >
> > But, as Oleg pointed out, a thread group leader may have NULL ->mm
> > and task_in_mem_cgroup() may do wrong decision. We have to use
> > find_lock_task_mm() in memcg as generic OOM-Killer does.
> >
> > Changelog:
> > - removed unnecessary changes in comments.
> >
>
> mm->owner solves the same problem, but unfortunately we have task
> based selection in OOM killer, so we need this patch. It is quite
> ironic that we find the mm from the task and then eventually the task
> back from mm->owner and then the mem cgroup. If we already know the mm
> from oom_kill.c, I think we can change the function to work off of
> that. mm->owner->cgroup..no?
>
There is no function as for_each_mm(). There is only for_each_process().
And, generally, there is no way to get a task via mm_struct.
To send signal etc. we need task.
(mm_owner is an option but not always configured and there will be
complicated problem as vfork() etc.)
I think oom_kill.c has to depend on thread, not mm, unfortunately.
Thanks,
-Kame