Re: [PATCH V2] xfs: Check if a direct write can result in a false ENOSPC error

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

On Wed, Aug 24, 2022 at 03:00:57PM +0530, Chandan Babu R wrote:
> This commit adds a test to check if a direct write on a delalloc extent
> present in CoW fork can result in a false ENOSPC error. The bug has been fixed
> by upstream commit d62113303d691 ("xfs: Fix false ENOSPC when performing
> direct write on a delalloc extent in cow fork").
> 
> Signed-off-by: Chandan Babu R <chandan.babu@oracle.com>
> ---
> Changelog:
> V1 -> V2:
>   1. Use file blocks as units instead of bytes to specify file offsets.
> 
>  tests/xfs/553     | 65 +++++++++++++++++++++++++++++++++++++++++++++++
>  tests/xfs/553.out |  9 +++++++
>  2 files changed, 74 insertions(+)
>  create mode 100755 tests/xfs/553
>  create mode 100644 tests/xfs/553.out
> 
> diff --git a/tests/xfs/553 b/tests/xfs/553
> new file mode 100755
> index 00000000..ae52e9fc
> --- /dev/null
> +++ b/tests/xfs/553
> @@ -0,0 +1,65 @@
> +#! /bin/bash
> +# SPDX-License-Identifier: GPL-2.0
> +# Copyright (c) 2022 Oracle.  All Rights Reserved.
> +#
> +# FS QA Test 553
> +#
> +# Test to check if a direct write on a delalloc extent present in CoW fork can
> +# result in an ENOSPC error.
> +#
> +. ./common/preamble
> +_begin_fstest auto quick clone
> +
> +# Import common functions.
> +. ./common/reflink
> +. ./common/inject
> +
> +# real QA test starts here
> +_supported_fs xfs
> +_fixed_by_kernel_commit d62113303d691 \
> +	"xfs: Fix false ENOSPC when performing direct write on a delalloc extent in cow fork"
> +_require_scratch_reflink
> +_require_xfs_debug
> +_require_test_program "punch-alternating"
> +_require_xfs_io_error_injection "bmap_alloc_minlen_extent"
> +_require_xfs_io_command "reflink"
> +_require_xfs_io_command "cowextsize"
> +
> +source=${SCRATCH_MNT}/source
> +destination=${SCRATCH_MNT}/destination
> +fragmented_file=${SCRATCH_MNT}/fragmented_file
> +
> +echo "Format and mount fs"
> +_scratch_mkfs >> $seqres.full
> +_scratch_mount >> $seqres.full
> +
> +blksz=$(_get_block_size $SCRATCH_MNT)

_get_file_block_size, since COW can only be done in file allocation
units.

This is (currently) a distinction that won't make a difference until
realtime reflink, where we'll have to deal with rextsize > 1FSB, but we
might as well get this correct from the start.

With that fixed,
Reviewed-by: Darrick J. Wong <djwong@kernel.org>

--D


> +
> +echo "Create source file"
> +$XFS_IO_PROG -f -c "pwrite 0 $((blksz * 8192))" $source >> $seqres.full
> +
> +echo "Reflink destination file with source file"
> +$XFS_IO_PROG -f -c "reflink $source" $destination >> $seqres.full
> +
> +echo "Set destination file's cowextsize to 4096 blocks"
> +$XFS_IO_PROG -c "cowextsize $((blksz * 4096))" $destination >> $seqres.full
> +
> +echo "Fragment FS"
> +$XFS_IO_PROG -f -c "pwrite 0 $((blksz * 16384))" $fragmented_file >> $seqres.full
> +sync
> +$here/src/punch-alternating $fragmented_file >> $seqres.full
> +
> +echo "Inject bmap_alloc_minlen_extent error tag"
> +_scratch_inject_error bmap_alloc_minlen_extent 1
> +
> +echo "Create delalloc extent of length 4096 blocks in destination file's CoW fork"
> +$XFS_IO_PROG -c "pwrite 0 $blksz" $destination >> $seqres.full
> +
> +sync
> +
> +echo "Direct I/O write at 3rd block in destination file"
> +$XFS_IO_PROG -d -c "pwrite $((blksz * 3)) $((blksz * 2))" $destination >> $seqres.full
> +
> +# success, all done
> +status=0
> +exit
> diff --git a/tests/xfs/553.out b/tests/xfs/553.out
> new file mode 100644
> index 00000000..9f5679de
> --- /dev/null
> +++ b/tests/xfs/553.out
> @@ -0,0 +1,9 @@
> +QA output created by 553
> +Format and mount fs
> +Create source file
> +Reflink destination file with source file
> +Set destination file's cowextsize to 4096 blocks
> +Fragment FS
> +Inject bmap_alloc_minlen_extent error tag
> +Create delalloc extent of length 4096 blocks in destination file's CoW fork
> +Direct I/O write at 3rd block in destination file
> -- 
> 2.35.1
>