[PATCH] Avoid a NULL pointer in btrfs

[PATCH] Avoid a NULL pointer in btrfs

[Liuwenyi]

In Yang Ruirui's mail, the btrfs will create a oops. This is caused by a
null pointer in test_range_bit() while lock the spinlock.

So, It is necessary to add a pointer check into test_range_bit()

---
Signed-off-by: Liu Wenyi <qingshenlwy@gmail.com>

---
fs/btrfs/extent_io.c | 3 +++
1 files changed, 3 insertions(+), 0 deletions(-)

diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index 5e7a94d..0da42b5 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -1553,6 +1553,9 @@ int test_range_bit(struct extent_io_tree *tree,
u64 start, u64 end,
struct rb_node *node;
int bitset = 0;

+ if (tree == NULL)
+ return bitset;
+
spin_lock(&tree->lock);
if (cached && cached->tree && cached->start == start)
node = &cached->rb_node;
-- 
1.7.3.3



---
Best Regards,
Liu Wenyi

Re: [PATCH] Avoid a NULL pointer in btrfs

[Josef Bacik]

On Wed, Jan 19, 2011 at 10:08:13PM +0800, Liuwenyi wrote:
> In Yang Ruirui's mail, the btrfs will create a oops. This is caused by a
> null pointer in test_range_bit() while lock the spinlock.
> 
> So, It is necessary to add a pointer check into test_range_bit()
> 


NAK, the tree shouldn't be null coming into this function, something else is
going wrong.  What oops is this?  Thanks,

Josef

Re: [PATCH] Avoid a NULL pointer in btrfs

[Chris Mason]

Excerpts from Josef Bacik's message of 2011-01-19 09:14:02 -0500:
> On Wed, Jan 19, 2011 at 10:08:13PM +0800, Liuwenyi wrote:
> > In Yang Ruirui's mail, the btrfs will create a oops. This is caused by a
> > null pointer in test_range_bit() while lock the spinlock.
> > 
> > So, It is necessary to add a pointer check into test_range_bit()
> > 
> 
> 
> NAK, the tree shouldn't be null coming into this function, something else is
> going wrong.  What oops is this?  Thanks,

What was your metadata blocksize for this oops?  This call should never
happen.

I think there is a larger problem, probably in the IO error handling
code since the trace had io errors beforehand.

-chris

Re: [PATCH] Avoid a NULL pointer in btrfs

[Liuwenyi]

=E4=BA=8E 2011-1-19 23:52, Chris Mason =E5=86=99=E9=81=93:
> Excerpts from Josef Bacik's message of 2011-01-19 09:14:02 -0500:
>> On Wed, Jan 19, 2011 at 10:08:13PM +0800, Liuwenyi wrote:
>>> In Yang Ruirui's mail, the btrfs will create a oops. This is caused=
 by a
>>> null pointer in test_range_bit() while lock the spinlock.
>>>
>>> So, It is necessary to add a pointer check into test_range_bit()
>>>
>>
>> NAK, the tree shouldn't be null coming into this function, something=
 else is
>> going wrong.  What oops is this?  Thanks,
The mail is here http://lkml.org/lkml/2011/1/19/24

another mesg is here http://www.aei.mpg.de/~crmafra/dmesg-2.6.35.3.txt
> What was your metadata blocksize for this oops?  This call should nev=
er
> happen.
It is rare and hard to reproduced. So, I update this patch, just avoid =
a=20
null pointer calling.
> I think there is a larger problem, probably in the IO error handling
> code since the trace had io errors beforehand.
Yes, I agree. This situation is too strange to happen.
> -chris
---
Best Regards,
Liu Wenyi