On Fri, May 26, 2023 at 2:13 AM Mariusz Tkaczyk <mariusz.tkaczyk@xxxxxxxxxxxxxxx> wrote: > > On Fri, 26 May 2023 15:23:58 +0800 > Xiao Ni <xni@xxxxxxxxxx> wrote: > > > On Fri, May 26, 2023 at 3:12 PM Guoqing Jiang <guoqing.jiang@xxxxxxxxx> wrote: > > > > > > > > > > > > On 5/26/23 14:45, Xiao Ni wrote: > > > > On Fri, May 26, 2023 at 11:09 AM Guoqing Jiang <guoqing.jiang@xxxxxxxxx> > > > > wrote: > > > >> > > > >> > > > >> On 5/26/23 09:49, Xiao Ni wrote: > > > >>> Hi all > > > >>> > > > >>> We found a problem recently. The read data is wrong when recovery > > > >>> happens. Now we've found it's introduced by patch 10764815f (md: add io > > > >>> accounting for raid0 and raid5). I can reproduce this 100%. This > > > >>> problem exists in upstream. The test steps are like this: > > > >>> > > > >>> 1. mdadm -CR $devname -l5 -n4 /dev/sd[b-e] --force --assume-clean > > > >>> 2. mkfs.ext4 -F $devname > > > >>> 3. mount $devname $mount_point > > > >>> 4. mdadm --incremental --fail sdd > > > >>> 5. dd if=/dev/zero of=/tmp/pythontest/file1 bs=1M count=100000 > > > >>> status=progress > > > > > > I suppose /tmp is the mount point. > > > > /tmp/pythontest is the mount point > > > > > > > > >>> 6. mdadm /dev/md126 --add /dev/sdd > > > >>> 7. create 31 processes that writes and reads. It compares the content > > > >>> with md5sum. The test will go on until the recovery stops > > > > > > Could you share the test code/script for step 7? Will try it from my side. > > > > The test scripts are written by people from intel. > > Hi, Mariusz. Can I share the test scripts here? > > Yes. Let us know if there is something else we can do to help here. I tried to reproduce this with fio, but didn't get much luck. Please share the test scripts. Thanks, Song
