Re: [PATCH block/for-linus] writeback: fix syncing of I_DIRTY_TIME inodes
From: Eryu Guan
Date: Thu Aug 20 2015 - 10:37:58 EST
On Thu, Aug 20, 2015 at 10:36:26PM +0800, Eryu Guan wrote:
> On Thu, Aug 20, 2015 at 02:12:24PM +0800, Eryu Guan wrote:
[snip]
>
> I installed a new vm with ext4 as root fs and got some trace info.
>
> On the new vm, only generic/048 is reproducible, generic/049 always
> passes. And I can only reproduce generic/048 when xfs tracepoints are
> enabled, if writeback tracepoints are enabled too, I can no longer
> reproduce the failure.
>
> All tests are done on 4.2-rc7 kernel.
>
> This is the trace-cmd I'm using:
>
> cd /mnt/ext4
> trace-cmd record -e xfs_file_buffered_write \
> -e xfs_file_fsync \
> -e xfs_writepage \
> -e xfs_setfilesize &
> pushd /path/to/xfstests
> ./check generic/048
> popd
> kill -s 2 $!
> trace-cmd report >trace_report.txt
>
> I attached three files:
> 1) xfs-trace-generic-048.txt.bz2[1] trace report result
> 2) xfs-trace-generic-048.diff generic/048 failure diff output, could know which files has incorrect size
> 3) xfs-trace-generic-048.metadump.bz2 metadump of SCRATCH_DEV, which contains the test files
>
> If more info is needed please let me know.
>
> Thanks,
> Eryu
>
> [1] attach this file in a following mail, to avoid xfs list 500k limit
Attached this file in this mail.
Eryu
Attachment:
xfs-trace-generic-048.txt.bz2
Description: BZip2 compressed data