Hi David, Chris,

Any comment on this patchset?

Although we don't recommend user to use RAID5/6 for now, but this 
patchset still fixes 2 quite important bugs for RAID5/6.

And current so far the test shows no regression.
(Although test cases like btrfs/011 and btrfs/069 is still causing 
problem, as v4.10-rc without this patch will also cause problems).

And the patchset is relatively small enough for an rc.

Thanks,
Qu

At 12/12/2016 05:38 PM, Qu Wenruo wrote:

Can be feteched from github:
https://github.com/adam900710/linux.git raid56_fixes

Fixes 2 scrub bugs:
1) Scrub recover correct data, but wrong parity
2) Scrub report wrong csum error number, or even unrecoverable error

The patches are still undergoing xfstests, but currect for-linus-4.10 is already
causing deadlock for btrfs/011, even without the patches.

So I'd remove btrfs/011 and continue the test, even these test cases won't trigger
real recover code.

But the current internal test cases are quite good so far.
I'll test them for several extra loop, and submit the internal test for reference.
(Since it's not suitable for xfstest, so I'd only submit the test script, which
needs manually to probe chunk layout)

Qu Wenruo (3):
  btrfs: scrub: Introduce full stripe lock for RAID56
  btrfs: scrub: Fix RAID56 recovery race conditiong
  btrfs: raid56: Use correct stolen pages to calculate P/Q

 fs/btrfs/ctree.h       |   4 ++
 fs/btrfs/extent-tree.c |   3 +
 fs/btrfs/raid56.c      |  62 ++++++++++++++--
 fs/btrfs/scrub.c       | 192 +++++++++++++++++++++++++++++++++++++++++++++++++
 4 files changed, 257 insertions(+), 4 deletions(-)