On Tue, Jul 30, 2013 at 9:27 PM, Jeff Liu <jeff.liu@xxxxxxxxxx> wrote:
> On 07/30/2013 07:06 PM, Zhi Yong Wu wrote:
>> On Mon, Jul 29, 2013 at 4:33 PM, Zhi Yong Wu <zwu.kernel@xxxxxxxxx> wrote:
>>> On Fri, Jul 26, 2013 at 7:37 PM, Dave Chinner <david@xxxxxxxxxxxxx> wrote:
>>>> On Fri, Jul 26, 2013 at 02:01:23PM +0800, Zhi Yong Wu wrote:
>>>>> No, it still raised the same assertion as below:
>>>>> [ 521.715103] XFS: Assertion failed: !(bip->bli_item.li_flags &
>>>>> XFS_LI_IN_AIL), file: fs/xfs/xfs_buf_item.c, line: 940
>>>> How are you reproducing it?
>>>> Can you take an event trace with trace-cmd and attach the output of
>>>> the report?
>>> After kernel base is switched to Linux 3.11-rc3, when i try to take
>>> event trace with trace-cmd, that issue doesn't appear any more in my
>>> environment. :)
>> When your patch is not applied, i hit this issue sometime today, and
>> take some event trace with trace-cmd, and attach it in this mail. But
>> after your patch is applied, i tried some times, but haven't hit so
>> far, so i want to see if i can hit it in next days.
> I have ran compilebench multiple times combine with different options against
> xfs-next in the past few days, but I can not reproduce this issue.
I ran it against Linus' tree, and hit this assertion sometimes, not
every time regardless that Dave's patch is applied.
> Could you please show us your test case as Dave has asked? Sometimes, the
> steps is more useful for someone who has interest in your post.
I just modified the test steps post by Dave.
root@debian-i386:/home/zwu# cat recovery_time_compilebench.sh
while(($i > 0))
mkfs.xfs -f /dev/sda;
mount /dev/sda /mnt/scratch
chmod 777 /mnt/scratch;
trace-cmd record -e xfs/* &
./compilebench --no-sync -D /mnt/scratch &
xfs_logprint -t /dev/sda |head -20
time mount /dev/sda /mnt/scratch
rm -rf ~/trace.dat*
Zhi Yong Wu