[PATCH 1/1] perf_event: add check for group_list if the parent isn't counted

[PATCH 1/1] perf_event: add check for group_list if the parent isn't counted

[Liming Wang]

If the parent has no entry in group_list, child_ctx will not be
allocated, which will lead dereference of a NULL child_ctx.

Signed-off-by: Liming Wang <liming.wang@windriver.com>
---
 kernel/perf_event.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)

diff --git a/kernel/perf_event.c b/kernel/perf_event.c
index 5b987b4..3664c4b 100644
--- a/kernel/perf_event.c
+++ b/kernel/perf_event.c
@@ -5126,6 +5126,8 @@ int perf_event_init_task(struct task_struct *child)
 	 */
 	mutex_lock(&parent_ctx->mutex);
 
+	if (list_empty(&parent_ctx->group_list))
+		goto exit;
 	/*
 	 * We dont have to disable NMIs - we are only looking at
 	 * the list, not manipulating it:
-- 
1.6.0.3

Re: [PATCH 1/1] perf_event: add check for group_list if the parent isn't counted

[Peter Zijlstra]

On Wed, 2009-12-30 at 19:28 +0800, Liming Wang wrote:
> If the parent has no entry in group_list, child_ctx will not be
> allocated, which will lead dereference of a NULL child_ctx.

That changelog sucks.

Best I can make of it is that there is a race where the parent gets his
context instantiated and we manage to get the mutex before the other
thread manages to add the first event.

Then we observe parent_event_ctx but have an empty list.

Is that it?

In that case, would it not be better to change the if (inherited_all)
condition to if (inherited_all && child_ctx) ?

> Signed-off-by: Liming Wang <liming.wang@windriver.com>
> ---
>  kernel/perf_event.c |    2 ++
>  1 files changed, 2 insertions(+), 0 deletions(-)
> 
> diff --git a/kernel/perf_event.c b/kernel/perf_event.c
> index 5b987b4..3664c4b 100644
> --- a/kernel/perf_event.c
> +++ b/kernel/perf_event.c
> @@ -5126,6 +5126,8 @@ int perf_event_init_task(struct task_struct *child)
>  	 */
>  	mutex_lock(&parent_ctx->mutex);
>  
> +	if (list_empty(&parent_ctx->group_list))
> +		goto exit;
>  	/*
>  	 * We dont have to disable NMIs - we are only looking at
>  	 * the list, not manipulating it:

Re: [PATCH 1/1] perf_event: add check for group_list if the parent isn't counted

[Wang Liming]

Peter Zijlstra wrote:
> On Wed, 2009-12-30 at 19:28 +0800, Liming Wang wrote:
>> If the parent has no entry in group_list, child_ctx will not be
>> allocated, which will lead dereference of a NULL child_ctx.
> 
> That changelog sucks.
Sorry the changelog is too simple.

> 
> Best I can make of it is that there is a race where the parent gets his
> context instantiated and we manage to get the mutex before the other
> thread manages to add the first event.
> 
> Then we observe parent_event_ctx but have an empty list.
> 
> Is that it?
I didn't find this case.
In my case, if I perf record a existing process with "--pid" and finish record,
and if later the recorded process forks a process, the condition will occur.

Liming Wang
> 
> In that case, would it not be better to change the if (inherited_all)
> condition to if (inherited_all && child_ctx) ?
> 
>> Signed-off-by: Liming Wang <liming.wang@windriver.com>
>> ---
>>  kernel/perf_event.c |    2 ++
>>  1 files changed, 2 insertions(+), 0 deletions(-)
>>
>> diff --git a/kernel/perf_event.c b/kernel/perf_event.c
>> index 5b987b4..3664c4b 100644
>> --- a/kernel/perf_event.c
>> +++ b/kernel/perf_event.c
>> @@ -5126,6 +5126,8 @@ int perf_event_init_task(struct task_struct *child)
>>  	 */
>>  	mutex_lock(&parent_ctx->mutex);
>>  
>> +	if (list_empty(&parent_ctx->group_list))
>> +		goto exit;
>>  	/*
>>  	 * We dont have to disable NMIs - we are only looking at
>>  	 * the list, not manipulating it:
> 
> 
> 

Re: [PATCH 1/1] perf_event: add check for group_list if the parent isn't counted

[Wang Liming]

Peter Zijlstra wrote:
> On Wed, 2009-12-30 at 22:36 +0800, Wang Liming wrote:
>>> Best I can make of it is that there is a race where the parent gets his
>>> context instantiated and we manage to get the mutex before the other
>>> thread manages to add the first event.
>>>
>>> Then we observe parent_event_ctx but have an empty list.
>>>
>>> Is that it?
>> I didn't find this case.
>> In my case, if I perf record a existing process with "--pid" and finish record,
>> and if later the recorded process forks a process, the condition will occur.
> 
> Ah, right, that will lead to the same state, since closing the last
> counter will not remove the context.
> 
> Does the below also fix your issue?
Yes, it's OK to me.
Thanks a lot!

Liming Wang
> 
> ---
> Subject: perf: Fix NULL deref in inheritance code
> From: Peter Zijlstra <a.p.zijlstra@chello.nl>
> Date: Wed Dec 30 16:00:35 CET 2009
> 
> Liming found a NULL deref when a task has a perf context but no counters 
> when it forks.
> 
> This can occur in two cases, a race during construction where the fork hits
> after installing the context but before the first counter gets inserted, or
> more reproducably, a fork after the last counter is closed (which leaves the
> context around).
> 
> CC: stable@kernel.org
> Reported-by: Wang Liming <liming.wang@windriver.com>
> Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
> ---
>  kernel/perf_event.c |    5 ++---
>  1 file changed, 2 insertions(+), 3 deletions(-)
> 
> Index: linux-2.6/kernel/perf_event.c
> ===================================================================
> --- linux-2.6.orig/kernel/perf_event.c
> +++ linux-2.6/kernel/perf_event.c
> @@ -5149,7 +5149,7 @@ int perf_event_init_task(struct task_str
>  					    GFP_KERNEL);
>  			if (!child_ctx) {
>  				ret = -ENOMEM;
> -				goto exit;
> +				break;
>  			}
>  
>  			__perf_event_init_context(child_ctx, child);
> @@ -5165,7 +5165,7 @@ int perf_event_init_task(struct task_str
>  		}
>  	}
>  
> -	if (inherited_all) {
> +	if (child_ctx && inherited_all) {
>  		/*
>  		 * Mark the child context as a clone of the parent
>  		 * context, or of whatever the parent is a clone of.
> @@ -5185,7 +5185,6 @@ int perf_event_init_task(struct task_str
>  		get_ctx(child_ctx->parent_ctx);
>  	}
>  
> -exit:
>  	mutex_unlock(&parent_ctx->mutex);
>  
>  	perf_unpin_context(parent_ctx);
> 
> 
> 

Re: [PATCH 1/1] perf_event: add check for group_list if the parent isn't counted

[Peter Zijlstra]

On Wed, 2009-12-30 at 22:36 +0800, Wang Liming wrote:
> > Best I can make of it is that there is a race where the parent gets his
> > context instantiated and we manage to get the mutex before the other
> > thread manages to add the first event.
> > 
> > Then we observe parent_event_ctx but have an empty list.
> > 
> > Is that it?
> I didn't find this case.
> In my case, if I perf record a existing process with "--pid" and finish record,
> and if later the recorded process forks a process, the condition will occur.

Ah, right, that will lead to the same state, since closing the last
counter will not remove the context.

Does the below also fix your issue?

---
Subject: perf: Fix NULL deref in inheritance code
From: Peter Zijlstra <a.p.zijlstra@chello.nl>
Date: Wed Dec 30 16:00:35 CET 2009

Liming found a NULL deref when a task has a perf context but no counters 
when it forks.

This can occur in two cases, a race during construction where the fork hits
after installing the context but before the first counter gets inserted, or
more reproducably, a fork after the last counter is closed (which leaves the
context around).

CC: stable@kernel.org
Reported-by: Wang Liming <liming.wang@windriver.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
---
 kernel/perf_event.c |    5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

Index: linux-2.6/kernel/perf_event.c
===================================================================
--- linux-2.6.orig/kernel/perf_event.c
+++ linux-2.6/kernel/perf_event.c
@@ -5149,7 +5149,7 @@ int perf_event_init_task(struct task_str
 					    GFP_KERNEL);
 			if (!child_ctx) {
 				ret = -ENOMEM;
-				goto exit;
+				break;
 			}
 
 			__perf_event_init_context(child_ctx, child);
@@ -5165,7 +5165,7 @@ int perf_event_init_task(struct task_str
 		}
 	}
 
-	if (inherited_all) {
+	if (child_ctx && inherited_all) {
 		/*
 		 * Mark the child context as a clone of the parent
 		 * context, or of whatever the parent is a clone of.
@@ -5185,7 +5185,6 @@ int perf_event_init_task(struct task_str
 		get_ctx(child_ctx->parent_ctx);
 	}
 
-exit:
 	mutex_unlock(&parent_ctx->mutex);
 
 	perf_unpin_context(parent_ctx);

[tip:perf/urgent] perf: Fix NULL deref in inheritance code

[tip-bot for Peter Zijlstra]

Commit-ID:  05cbaa2853cdfc255fdd04e65a82bfe9208c4e52
Gitweb:     http://git.kernel.org/tip/05cbaa2853cdfc255fdd04e65a82bfe9208c4e52
Author:     Peter Zijlstra <a.p.zijlstra@chello.nl>
AuthorDate: Wed, 30 Dec 2009 16:00:35 +0100
Committer:  Ingo Molnar <mingo@elte.hu>
CommitDate: Thu, 31 Dec 2009 13:11:31 +0100

perf: Fix NULL deref in inheritance code

Liming found a NULL deref when a task has a perf context but no
counters  when it forks.

This can occur in two cases, a race during construction where
the fork hits after installing the context but before the first
counter gets inserted, or more reproducably, a fork after the
last counter is closed (which leaves the context around).

Reported-by: Wang Liming <liming.wang@windriver.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
CC: <stable@kernel.org>
LKML-Reference: <1262185684.7135.222.camel@laptop>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
---
 kernel/perf_event.c |    5 ++---
 1 files changed, 2 insertions(+), 3 deletions(-)

diff --git a/kernel/perf_event.c b/kernel/perf_event.c
index 03cc061..58ed1da 100644
--- a/kernel/perf_event.c
+++ b/kernel/perf_event.c
@@ -5148,7 +5148,7 @@ int perf_event_init_task(struct task_struct *child)
 					    GFP_KERNEL);
 			if (!child_ctx) {
 				ret = -ENOMEM;
-				goto exit;
+				break;
 			}
 
 			__perf_event_init_context(child_ctx, child);
@@ -5164,7 +5164,7 @@ int perf_event_init_task(struct task_struct *child)
 		}
 	}
 
-	if (inherited_all) {
+	if (child_ctx && inherited_all) {
 		/*
 		 * Mark the child context as a clone of the parent
 		 * context, or of whatever the parent is a clone of.
@@ -5184,7 +5184,6 @@ int perf_event_init_task(struct task_struct *child)
 		get_ctx(child_ctx->parent_ctx);
 	}
 
-exit:
 	mutex_unlock(&parent_ctx->mutex);
 
 	perf_unpin_context(parent_ctx);