[PATCH 3/3] xfs: cancel COW in xfs_cancel_ioend

[Darrick J. Wong]

On Fri, Jan 08, 2016 at 02:47:04PM +0100, Christoph Hellwig wrote:
> Thanks Darrick!
> 
> I've not seen the inode eviction asserts anymore, but I now hit a
> corruption warnings in generic/168 reliably.  I did hit before as
> well, but not very reliably.

I'll see if I can repro the 168 error; it's been running in a loop all
night and hasn't bombed yet.

In the meantime, I added some more tests and fixed a CoW corruption when an
xfs_io_overwrite extent has cow reservations in the middle of the extent.

I also restarted testing on arm64, ppc64{,el}, and i686; it seems
stable enough right now to pass all ~130 reflink xfstests here.

> generic/168 17s ...[  296.988867] XFS (vdc): Metadata corruption detected at
> xfs_refcountbt_write_verify+0x3e/0x90, block 0xb0
> [  296.990412] XFS (vdc): Unmount and run xfs_repair
> [  296.991300] XFS (vdc): First 64 bytes of corrupted metadata buffer:
> [  296.992349] ffff88007b755000: 52 33 46 43 00 01 00 01 ff ff ff ff ff ff ff
> ff  R3FC............
> [  296.993893] ffff88007b755010: 00 00 00 00 00 00 00 b0 00 00 00 00 00 00 00
> 00  ................

No LSN?  That's ... odd.

> [  297.003293] ffff88007b755020: fe 94 c7 69 fd bd 4a 77 85 3d 9d 6d 7b 45 38
> 07  ...i..Jw.=.m{E8.

Does this match the FS UUID?

> [  297.004442] ffff88007b755030: 00 00 00 00 00 00 00 00 00 00 0b 9c 00 00 7f
> 07  ................

Also no owner or CRC.  Hmmmm.

> [  297.005867] XFS (vdc): xfs_do_force_shutdown(0x8) called from line 1247 of
> file fs/xfs/xfs_buf.c.  Return address = 0xffffffff814cce52
> 
> I'll investigate it in a little more detail.

--D