Re: [PATCH] generic: test COW writeback failure when overlapping non-shared blocks

["Darrick J. Wong" <djwong@kernel.org>]

On Thu, Oct 21, 2021 at 12:39:59PM -0400, Brian Foster wrote:
> Test that COW writeback that overlaps non-shared delalloc blocks
> does not leave around stale delalloc blocks on I/O failure. This
> triggers assert failures and free space accounting corruption on
> XFS.
> 
> Signed-off-by: Brian Foster <bfoster@redhat.com>
> ---
> 
> This test targets the problem addressed by the following patch in XFS:
> 
> https://lore.kernel.org/linux-xfs/20211021163330.1886516-1-bfoster@redhat.com/
> 
> Brian
> 
>  tests/generic/651     | 53 +++++++++++++++++++++++++++++++++++++++++++
>  tests/generic/651.out |  2 ++
>  2 files changed, 55 insertions(+)
>  create mode 100755 tests/generic/651
>  create mode 100644 tests/generic/651.out
> 
> diff --git a/tests/generic/651 b/tests/generic/651
> new file mode 100755
> index 00000000..8d4e6728
> --- /dev/null
> +++ b/tests/generic/651
> @@ -0,0 +1,53 @@
> +#! /bin/bash
> +# SPDX-License-Identifier: GPL-2.0
> +# Copyright (c) 2021 Red Hat, Inc.  All Rights Reserved.
> +#
> +# FS QA Test 651
> +#
> +# Test that COW writeback that overlaps non-shared delalloc blocks does not
> +# leave around stale delalloc blocks on I/O failure. This triggers assert
> +# failures and free space accounting corruption on XFS.
> +#
> +. ./common/preamble
> +_begin_fstest auto quick clone
> +
> +_cleanup()
> +{
> +	_cleanup_flakey
> +	cd /
> +	rm -r -f $tmp.*
> +}
> +
> +# Import common functions.
> +. ./common/reflink
> +. ./common/dmflakey
> +
> +# real QA test starts here
> +_supported_fs generic
> +_require_scratch_reflink
> +_require_flakey_with_error_writes

_require_cp_reflink

> +
> +_scratch_mkfs >> $seqres.full
> +_init_flakey
> +_mount_flakey
> +
> +# create two files that share a single block
> +$XFS_IO_PROG -fc "pwrite 4k 4k" $SCRATCH_MNT/file1 >> $seqres.full

Please use:

blksz=$(_get_file_block_size $SCRATCH_MNT)
$XFS_IO_PROG -fc "pwrite $blksz $blksz" $SCRATCH_MNT/file1 >> $seqres.full

So that this test will work properly on filesystems with bs > 4k.

> +cp --reflink $SCRATCH_MNT/file1 $SCRATCH_MNT/file2

Nit: This could be shortened to use the _cp_reflink helper, though it
doesn't really matter to me if you do.

> +# Perform a buffered write across the shared and non-shared blocks. On XFS, this
> +# creates a COW fork extent that covers the shared block as well as the just

Ah, the reason why there's a cow fork extent covering the delalloc
reservation is due to the default cow extent size hint, right?  In that
case, you need:

_require_xfs_io_command "cowextsize"
$XFS_IO_PROG -c "cowextsize 0" $SCRATCH_MNT >> $seqres.full

to ensure that the speculative cow preallocation actually gets set up.
Otherwise, I think test won't reproduce the bug if the test config has
-d cowextsize=1 in the mkfs options.

> +# created non-shared delalloc block. Fail the writeback to verify that all
> +# delayed allocation is cleaned up properly.
> +_load_flakey_table $FLAKEY_ERROR_WRITES
> +$XFS_IO_PROG -c "pwrite 0 8k" -c fsync $SCRATCH_MNT/file2 >> $seqres.full

$((2 * blksz)), not 8k

Other than that, this looks reasonable to me.  I'll go look at the fix
patch now. :)

--D

> +_load_flakey_table $FLAKEY_ALLOW_WRITES
> +
> +# Try a post-fail reflink and then unmount. Both of these are known to produce
> +# errors and/or assert failures on XFS if we trip over a stale delalloc block.
> +cp --reflink $SCRATCH_MNT/file2 $SCRATCH_MNT/file3
> +_unmount_flakey
> +
> +# success, all done
> +status=0
> +exit
> diff --git a/tests/generic/651.out b/tests/generic/651.out
> new file mode 100644
> index 00000000..bd44c80c
> --- /dev/null
> +++ b/tests/generic/651.out
> @@ -0,0 +1,2 @@
> +QA output created by 651
> +fsync: Input/output error
> -- 
> 2.31.1
>