[PATCH] VFS: call synchronize_rcu after kill_sb.

[PATCH] VFS: call synchronize_rcu after kill_sb.

[Tao Ma]

From: Tao Ma <boyu.mt@taobao.com>

In fa0d7e3, we use rcu free inode instead of freeing the inode
directly. It causes a problem when we rmmod immediately after
we umount the volume[1].

So we need to call synchronize_rcu after we kill_sb so that
the inode is freed before we do rmmod. The idea is inspired
by Chris Mason[2]. I tested with ext4 by umount+rmmod and it
doesn't show any error by now.

1. http://marc.info/?l=linux-fsdevel&m=129680863330185&w=2
2. http://marc.info/?l=linux-fsdevel&m=129684698713709&w=2 

Cc: Nick Piggin <npiggin@kernel.dk>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Chris Mason <chris.mason@oracle.com>
Cc: Boaz Harrosh <bharrosh@panasas.com>
Signed-off-by: Tao Ma <boyu.mt@taobao.com>
---
 fs/super.c |    7 +++++++
 1 files changed, 7 insertions(+), 0 deletions(-)

diff --git a/fs/super.c b/fs/super.c
index 74e149e..315bce9 100644
--- a/fs/super.c
+++ b/fs/super.c
@@ -177,6 +177,13 @@ void deactivate_locked_super(struct super_block *s)
 	struct file_system_type *fs = s->s_type;
 	if (atomic_dec_and_test(&s->s_active)) {
 		fs->kill_sb(s);
+		/*
+		 * We need to synchronize rcu here so that
+		 * the delayed rcu inode free can be executed
+		 * before we put_super.
+		 * https://bugzilla.kernel.org/show_bug.cgi?id=27652
+		 */
+		synchronize_rcu();
 		put_filesystem(fs);
 		put_super(s);
 	} else {
-- 
1.6.3.GIT

Re: [PATCH] VFS: call synchronize_rcu after kill_sb.

[Boaz Harrosh]

On 02/05/2011 11:01 AM, Tao Ma wrote:
> From: Tao Ma <boyu.mt@taobao.com>
> 
> In fa0d7e3, we use rcu free inode instead of freeing the inode
> directly. It causes a problem when we rmmod immediately after
> we umount the volume[1].
> 
> So we need to call synchronize_rcu after we kill_sb so that
> the inode is freed before we do rmmod. The idea is inspired
> by Chris Mason[2]. I tested with ext4 by umount+rmmod and it
> doesn't show any error by now.
> 
> 1. http://marc.info/?l=linux-fsdevel&m=129680863330185&w=2
> 2. http://marc.info/?l=linux-fsdevel&m=129684698713709&w=2 
> 
> Cc: Nick Piggin <npiggin@kernel.dk>
> Cc: Al Viro <viro@zeniv.linux.org.uk>
> Cc: Chris Mason <chris.mason@oracle.com>
> Cc: Boaz Harrosh <bharrosh@panasas.com>
> Signed-off-by: Tao Ma <boyu.mt@taobao.com>
> ---
>  fs/super.c |    7 +++++++
>  1 files changed, 7 insertions(+), 0 deletions(-)
> 
> diff --git a/fs/super.c b/fs/super.c
> index 74e149e..315bce9 100644
> --- a/fs/super.c
> +++ b/fs/super.c
> @@ -177,6 +177,13 @@ void deactivate_locked_super(struct super_block *s)
>  	struct file_system_type *fs = s->s_type;
>  	if (atomic_dec_and_test(&s->s_active)) {
>  		fs->kill_sb(s);
> +		/*
> +		 * We need to synchronize rcu here so that
> +		 * the delayed rcu inode free can be executed
> +		 * before we put_super.
> +		 * https://bugzilla.kernel.org/show_bug.cgi?id=27652
> +		 */
> +		synchronize_rcu();
>  		put_filesystem(fs);
>  		put_super(s);
>  	} else {


Sorry for not testing sooner.

The above does not work I still get the exact same crash!!

Looking at the code for synchronize_rcu() it looks like it might not be
enough. It looks like all it does is a memory barrier. But we need 
something that will actually pump these pending releases.
(I might be way off here)

BTW after I get the Warning from the kmem_cache_destroy:
slab error in kmem_cache_destroy(): cache `exofs_inode_cache': Can't free all objects
Call Trace: 
754efe08:  [<6007e9a6>] kmem_cache_destroy+0x82/0xca
754efe38:  [<7a9296ba>] exit_exofs+0x1a/0x1c [exofs]
754efe48:  [<60054c10>] sys_delete_module+0x1b9/0x217
754efee8:  [<60014d60>] handle_syscall+0x58/0x70
754eff08:  [<60024163>] userspace+0x2dd/0x38a
754effc8:  [<600126af>] fork_handler+0x62/0x69


I also get a Kernel crash. I suspect it's when finally these
free_rcu come and the module (and kmem_cache) are no longer there.

What to do? Nick?

Boaz

Re: [PATCH] VFS: call synchronize_rcu after kill_sb.

[Aneesh Kumar K. V]

On Mon, 07 Feb 2011 15:14:40 +0200, Boaz Harrosh <bharrosh@panasas.com> wrote:
> On 02/05/2011 11:01 AM, Tao Ma wrote:
> > From: Tao Ma <boyu.mt@taobao.com>
> > 
> > In fa0d7e3, we use rcu free inode instead of freeing the inode
> > directly. It causes a problem when we rmmod immediately after
> > we umount the volume[1].
> > 
> > So we need to call synchronize_rcu after we kill_sb so that
> > the inode is freed before we do rmmod. The idea is inspired
> > by Chris Mason[2]. I tested with ext4 by umount+rmmod and it
> > doesn't show any error by now.
> > 
> > 1. http://marc.info/?l=linux-fsdevel&m=129680863330185&w=2
> > 2. http://marc.info/?l=linux-fsdevel&m=129684698713709&w=2 
> > 
> > Cc: Nick Piggin <npiggin@kernel.dk>
> > Cc: Al Viro <viro@zeniv.linux.org.uk>
> > Cc: Chris Mason <chris.mason@oracle.com>
> > Cc: Boaz Harrosh <bharrosh@panasas.com>
> > Signed-off-by: Tao Ma <boyu.mt@taobao.com>
> > ---
> >  fs/super.c |    7 +++++++
> >  1 files changed, 7 insertions(+), 0 deletions(-)
> > 
> > diff --git a/fs/super.c b/fs/super.c
> > index 74e149e..315bce9 100644
> > --- a/fs/super.c
> > +++ b/fs/super.c
> > @@ -177,6 +177,13 @@ void deactivate_locked_super(struct super_block *s)
> >  	struct file_system_type *fs = s->s_type;
> >  	if (atomic_dec_and_test(&s->s_active)) {
> >  		fs->kill_sb(s);
> > +		/*
> > +		 * We need to synchronize rcu here so that
> > +		 * the delayed rcu inode free can be executed
> > +		 * before we put_super.
> > +		 * https://bugzilla.kernel.org/show_bug.cgi?id=27652
> > +		 */
> > +		synchronize_rcu();
> >  		put_filesystem(fs);
> >  		put_super(s);
> >  	} else {
> 
> 
> Sorry for not testing sooner.
> 
> The above does not work I still get the exact same crash!!
> 
> Looking at the code for synchronize_rcu() it looks like it might not be
> enough. It looks like all it does is a memory barrier. But we need 
> something that will actually pump these pending releases.
> (I might be way off here)
> 
> BTW after I get the Warning from the kmem_cache_destroy:
> slab error in kmem_cache_destroy(): cache `exofs_inode_cache': Can't free all objects
> Call Trace: 
> 754efe08:  [<6007e9a6>] kmem_cache_destroy+0x82/0xca
> 754efe38:  [<7a9296ba>] exit_exofs+0x1a/0x1c [exofs]
> 754efe48:  [<60054c10>] sys_delete_module+0x1b9/0x217
> 754efee8:  [<60014d60>] handle_syscall+0x58/0x70
> 754eff08:  [<60024163>] userspace+0x2dd/0x38a
> 754effc8:  [<600126af>] fork_handler+0x62/0x69
> 
> 
> I also get a Kernel crash. I suspect it's when finally these
> free_rcu come and the module (and kmem_cache) are no longer there.
> 
> What to do? Nick?


http://lwn.net/Articles/217484/ explains how to wait for rcu callback to finish

-aneesh

Re: [PATCH] VFS: call synchronize_rcu after kill_sb.

[Boaz Harrosh]

On 02/08/2011 06:57 PM, Aneesh Kumar K. V wrote:
> On Mon, 07 Feb 2011 15:14:40 +0200, Boaz Harrosh <bharrosh@panasas.com> wrote:
>> On 02/05/2011 11:01 AM, Tao Ma wrote:
>>> From: Tao Ma <boyu.mt@taobao.com>
>>>
>>> In fa0d7e3, we use rcu free inode instead of freeing the inode
>>> directly. It causes a problem when we rmmod immediately after
>>> we umount the volume[1].
>>>
>>> So we need to call synchronize_rcu after we kill_sb so that
>>> the inode is freed before we do rmmod. The idea is inspired
>>> by Chris Mason[2]. I tested with ext4 by umount+rmmod and it
>>> doesn't show any error by now.
>>>
>>> 1. http://marc.info/?l=linux-fsdevel&m=129680863330185&w=2
>>> 2. http://marc.info/?l=linux-fsdevel&m=129684698713709&w=2 
>>>
>>> Cc: Nick Piggin <npiggin@kernel.dk>
>>> Cc: Al Viro <viro@zeniv.linux.org.uk>
>>> Cc: Chris Mason <chris.mason@oracle.com>
>>> Cc: Boaz Harrosh <bharrosh@panasas.com>
>>> Signed-off-by: Tao Ma <boyu.mt@taobao.com>
>>> ---
>>>  fs/super.c |    7 +++++++
>>>  1 files changed, 7 insertions(+), 0 deletions(-)
>>>
>>> diff --git a/fs/super.c b/fs/super.c
>>> index 74e149e..315bce9 100644
>>> --- a/fs/super.c
>>> +++ b/fs/super.c
>>> @@ -177,6 +177,13 @@ void deactivate_locked_super(struct super_block *s)
>>>  	struct file_system_type *fs = s->s_type;
>>>  	if (atomic_dec_and_test(&s->s_active)) {
>>>  		fs->kill_sb(s);
>>> +		/*
>>> +		 * We need to synchronize rcu here so that
>>> +		 * the delayed rcu inode free can be executed
>>> +		 * before we put_super.
>>> +		 * https://bugzilla.kernel.org/show_bug.cgi?id=27652
>>> +		 */
>>> +		synchronize_rcu();
>>>  		put_filesystem(fs);
>>>  		put_super(s);
>>>  	} else {
>>
>>
<>
> 
> http://lwn.net/Articles/217484/ explains how to wait for rcu callback to finish
> 
> -aneesh

Yes thanks Aneesh, rcu_barrier does the trick
---
From: Boaz Harrosh <bharrosh@panasas.com>

In fa0d7e3, we use rcu free inode instead of freeing the inode
directly. It causes a problem when we rmmod immediately after
we umount the volume[1].

So we need to call rcu_barrier after we kill_sb so that
the inode is freed before we do rmmod. The idea is inspired
by Aneesh Kumar. rcu_barrier will wait for all callbacks
to end before preceding. The original patch was done by
Tao Ma, but synchronize_rcu() is not enough here.

1. http://marc.info/?l=linux-fsdevel&m=129680863330185&w=2
2. http://marc.info/?l=linux-fsdevel&m=129684698713709&w=2 

Cc: Nick Piggin <npiggin@kernel.dk>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Chris Mason <chris.mason@oracle.com>
Cc: Tao Ma <boyu.mt@taobao.com>
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
---
git diff --stat -p -M fs/super.c
 fs/super.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)

diff --git a/fs/super.c b/fs/super.c
index 74e149e..5fd4ec9 100644
--- a/fs/super.c
+++ b/fs/super.c
@@ -177,6 +177,13 @@ void deactivate_locked_super(struct super_block *s)
 	struct file_system_type *fs = s->s_type;
 	if (atomic_dec_and_test(&s->s_active)) {
 		fs->kill_sb(s);
+		/*
+		 * We need to synchronize rcu here so that
+		 * the delayed rcu inode free can be executed
+		 * before we put_super.
+		 * https://bugzilla.kernel.org/show_bug.cgi?id=27652
+		 */
+		rcu_barrier();
 		put_filesystem(fs);
 		put_super(s);
 	} else {

Re: [PATCH] VFS: call synchronize_rcu after kill_sb.

[Tao Ma]

On 02/09/2011 01:25 AM, Boaz Harrosh wrote:
> On 02/08/2011 06:57 PM, Aneesh Kumar K. V wrote:
>    
>> On Mon, 07 Feb 2011 15:14:40 +0200, Boaz Harrosh<bharrosh@panasas.com>  wrote:
>>      
>>> On 02/05/2011 11:01 AM, Tao Ma wrote:
>>>        
>>>> From: Tao Ma<boyu.mt@taobao.com>
>>>>
>>>> In fa0d7e3, we use rcu free inode instead of freeing the inode
>>>> directly. It causes a problem when we rmmod immediately after
>>>> we umount the volume[1].
>>>>
>>>> So we need to call synchronize_rcu after we kill_sb so that
>>>> the inode is freed before we do rmmod. The idea is inspired
>>>> by Chris Mason[2]. I tested with ext4 by umount+rmmod and it
>>>> doesn't show any error by now.
>>>>
>>>> 1. http://marc.info/?l=linux-fsdevel&m=129680863330185&w=2
>>>> 2. http://marc.info/?l=linux-fsdevel&m=129684698713709&w=2
>>>>
>>>> Cc: Nick Piggin<npiggin@kernel.dk>
>>>> Cc: Al Viro<viro@zeniv.linux.org.uk>
>>>> Cc: Chris Mason<chris.mason@oracle.com>
>>>> Cc: Boaz Harrosh<bharrosh@panasas.com>
>>>> Signed-off-by: Tao Ma<boyu.mt@taobao.com>
>>>> ---
>>>>   fs/super.c |    7 +++++++
>>>>   1 files changed, 7 insertions(+), 0 deletions(-)
>>>>
>>>> diff --git a/fs/super.c b/fs/super.c
>>>> index 74e149e..315bce9 100644
>>>> --- a/fs/super.c
>>>> +++ b/fs/super.c
>>>> @@ -177,6 +177,13 @@ void deactivate_locked_super(struct super_block *s)
>>>>   	struct file_system_type *fs = s->s_type;
>>>>   	if (atomic_dec_and_test(&s->s_active)) {
>>>>   		fs->kill_sb(s);
>>>> +		/*
>>>> +		 * We need to synchronize rcu here so that
>>>> +		 * the delayed rcu inode free can be executed
>>>> +		 * before we put_super.
>>>> +		 * https://bugzilla.kernel.org/show_bug.cgi?id=27652
>>>> +		 */
>>>> +		synchronize_rcu();
>>>>   		put_filesystem(fs);
>>>>   		put_super(s);
>>>>   	} else {
>>>>          
>>>
>>>        
> <>
>    
>> http://lwn.net/Articles/217484/ explains how to wait for rcu callback to finish
>>
>> -aneesh
>>      
> Yes thanks Aneesh, rcu_barrier does the trick
> ---
> From: Boaz Harrosh<bharrosh@panasas.com>
>
> In fa0d7e3, we use rcu free inode instead of freeing the inode
> directly. It causes a problem when we rmmod immediately after
> we umount the volume[1].
>
> So we need to call rcu_barrier after we kill_sb so that
> the inode is freed before we do rmmod. The idea is inspired
> by Aneesh Kumar. rcu_barrier will wait for all callbacks
> to end before preceding. The original patch was done by
> Tao Ma, but synchronize_rcu() is not enough here.
>
> 1. http://marc.info/?l=linux-fsdevel&m=129680863330185&w=2
> 2. http://marc.info/?l=linux-fsdevel&m=129684698713709&w=2
>
> Cc: Nick Piggin<npiggin@kernel.dk>
> Cc: Al Viro<viro@zeniv.linux.org.uk>
> Cc: Chris Mason<chris.mason@oracle.com>
> Cc: Tao Ma<boyu.mt@taobao.com>
> Signed-off-by: Boaz Harrosh<bharrosh@panasas.com>
>    
It works now in my ext4 test box. Thanks for your work.
Tested-by: Tao Ma <boyu.mt@taobao.com>
> ---
> git diff --stat -p -M fs/super.c
>   fs/super.c |    1 +
>   1 files changed, 1 insertions(+), 0 deletions(-)
>
> diff --git a/fs/super.c b/fs/super.c
> index 74e149e..5fd4ec9 100644
> --- a/fs/super.c
> +++ b/fs/super.c
> @@ -177,6 +177,13 @@ void deactivate_locked_super(struct super_block *s)
>   	struct file_system_type *fs = s->s_type;
>   	if (atomic_dec_and_test(&s->s_active)) {
>   		fs->kill_sb(s);
> +		/*
> +		 * We need to synchronize rcu here so that
> +		 * the delayed rcu inode free can be executed
> +		 * before we put_super.
> +		 * https://bugzilla.kernel.org/show_bug.cgi?id=27652
> +		 */
> +		rcu_barrier();
>   		put_filesystem(fs);
>   		put_super(s);
>   	} else {
> --
> To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>    

Re: [PATCH] VFS: call synchronize_rcu after kill_sb.

[Christoph Hellwig]

> +		/*
> +		 * We need to synchronize rcu here so that
> +		 * the delayed rcu inode free can be executed
> +		 * before we put_super.
> +		 * https://bugzilla.kernel.org/show_bug.cgi?id=27652
> +		 */

URLs in comments are not useful, descriptions of the issues in comments
should be complete enough to understand the issue.  I think the comment
without the url is enough, though.

[PATCH] VFS: call rcu_barrier after kill_sb.

[Boaz Harrosh]

In fa0d7e3, we use rcu free inode instead of freeing the inode
directly. It causes a crash when we rmmod immediately after
we umount the volume[1].

So we need to call rcu_barrier after we kill_sb so that
the inode is freed before we do rmmod. The idea is inspired
by Aneesh Kumar. rcu_barrier will wait for all callbacks
to end before preceding. The original patch was done by
Tao Ma, but synchronize_rcu() is not enough here.

1. http://marc.info/?l=linux-fsdevel&m=129680863330185&w=2

Cc: Nick Piggin <npiggin@kernel.dk>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Chris Mason <chris.mason@oracle.com>
Tested-by: Tao Ma <boyu.mt@taobao.com>
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
---
 fs/super.c |    5 +++++
 1 files changed, 5 insertions(+), 0 deletions(-)

diff --git a/fs/super.c b/fs/super.c
index 74e149e..7e9dd4c 100644
--- a/fs/super.c
+++ b/fs/super.c
@@ -177,6 +177,11 @@ void deactivate_locked_super(struct super_block *s)
 	struct file_system_type *fs = s->s_type;
 	if (atomic_dec_and_test(&s->s_active)) {
 		fs->kill_sb(s);
+		/*
+		 * We need to call rcu_barrier so all the delayed rcu free
+		 * inodes are flushed before we release the fs module.
+		 */
+		rcu_barrier();
 		put_filesystem(fs);
 		put_super(s);
 	} else {
-- 
1.7.2.3