On 06/11/2012 06:53 PM, Dave Chinner wrote:
> On Mon, Jun 11, 2012 at 01:42:31PM -0700, Chris J Arges wrote:
>> Mark Tinguely-3 wrote:
>>> The perl script to recreate this problem is very similar to xfstest 273.
>>> I use that because it avoids all the filesystem mount/unmount that
>>> happen between the test 273 loops. You can build the log size that you
>>> want to test, create the directories and let it run until it hangs.
>> I tested xfstest 273, and it looks like it exhibits a different issue than
>> the OP in this thread.
>> For example, if I run this test in a while  loop, I get the following
>> [16413.073946] XFS (sda5): Invalid block length (0xfffff48b) for buffer
> That looks bad. How big is the log on this filesystem?
The test and scratch partitions were created with the following commands:
mkfs.xfs -b size=1024 -l size=576b /dev/sda5
mkfs.xfs -b size=1024 -l size=576b /dev/sda6
>> [16413.073963] BUG: unable to handle kernel NULL pointer dereference at
>> [16413.074274] IP: [<ffffffffa02bb870>] uuid_is_nil+0x10/0x50 [xfs]
> I can't really see how this function can get a null pointer
> dereference. It checks the pointer passed in for being null before
> doing anything, and otherwise it just increments and dereferences
> the char pointer 16 times. I can't see how that results in a NULL
> being dereferenced - I might just be blind though.
Yea, this is odd. What I'm really trying to accomplish is seeing if this
xfstest 273 produces a similar hang to Juerg's original bug. If you've
already run this test and it produced a hang can you post the backtrace
or let me know if it is similar?