Re: [PATCH] btrfs: add a test for defrag of contiguous file extents

[Anand Jain <anand.jain@oracle.com>]

On 30/10/24 08:48, Anand Jain wrote:
> On 30/10/24 01:23, fdmanana@kernel.org wrote:
>> From: Filipe Manana <fdmanana@suse.com>
>>
>> Test that defrag merges adjacent extents that are contiguous.
>> This exercises a regression fixed by a patchset for the kernel that is
>> comprissed of the following patches:
>>
>>    btrfs: fix extent map merging not happening for adjacent extents
>>    btrfs: fix defrag not merging contiguous extents due to merged 
>> extent maps
>>
>> Signed-off-by: Filipe Manana <fdmanana@suse.com>
>> ---
>>   tests/btrfs/325     | 80 +++++++++++++++++++++++++++++++++++++++++++++
>>   tests/btrfs/325.out | 22 +++++++++++++
>>   2 files changed, 102 insertions(+)
>>   create mode 100755 tests/btrfs/325
>>   create mode 100644 tests/btrfs/325.out
>>
>> diff --git a/tests/btrfs/325 b/tests/btrfs/325
>> new file mode 100755
>> index 00000000..0b1ab3c2
>> --- /dev/null
>> +++ b/tests/btrfs/325
>> @@ -0,0 +1,80 @@
>> +#! /bin/bash
>> +# SPDX-License-Identifier: GPL-2.0
>> +# Copyright (C) 2024 SUSE Linux Products GmbH. All Rights Reserved.
>> +#
>> +# FS QA Test 325
>> +#
>> +# Test that defrag merges adjacent extents that are contiguous.
>> +#
>> +. ./common/preamble
>> +_begin_fstest auto quick preallocrw defrag
>> +
>> +. ./common/filter
>> +
>> +_require_scratch
>> +_require_btrfs_command inspect-internal dump-tree
>> +_require_xfs_io_command "falloc"
>> +
>> +_fixed_by_kernel_commit xxxxxxxxxxxx \
>> +    "btrfs: fix extent map merging not happening for adjacent extents"
>> +_fixed_by_kernel_commit xxxxxxxxxxxx \
>> +    "btrfs: fix defrag not merging contiguous extents due to merged 
>> extent maps"
>> +
>> +count_file_extent_items()
>> +{
>> +    # We count file extent extent items through dump-tree instead of 
>> using
>> +    # fiemap because fiemap merges adjacent extent items when they are
>> +    # contiguous.
>> +    # We unmount because all metadata must be ondisk for dump-tree to 
>> see
>> +    # it and work correctly.
>> +    _scratch_unmount
>> +    $BTRFS_UTIL_PROG inspect-internal dump-tree -t 5 $SCRATCH_DEV | \
>> +        grep EXTENT_DATA | wc -l
>> +    _scratch_mount
>> +}
>> +
>> +_scratch_mkfs >>$seqres.full 2>&1 || _fail "mkfs failed"
>> +_scratch_mount
>> +
>> +# Create a file with a size of 256K and 4 written extents of 64K each.
>> +# We fallocate to guarantee exact extent size, even if compression mount
>> +# option is give, and write to them because defrag skips prealloc 
>> extents.


>> +$XFS_IO_PROG -f -c "falloc 0 64K" \
>> +         -c "pwrite -S 0xab 0 64K" \
>> +         -c "falloc 64K 64K" \
>> +         -c "pwrite -S 0xcd 64K 64K" \
>> +         -c "falloc 128K 64K" \
>> +         -c "pwrite -S 0xef 128K 64K" \
>> +         -c "falloc 192K 64K" \
>> +         -c "pwrite -S 0x73 192K 64K" \
>> +         $SCRATCH_MNT/foo | _filter_xfs_io
>> +

With the fix applied, this test case fails when the compress option is used.

--------------
FSTYP         -- btrfs
PLATFORM      -- Linux/aarch64 dev2 6.12.0-rc4+ #36 SMP PREEMPT_DYNAMIC 
Tue Oct 29 00:25:40 +08 2024
MKFS_OPTIONS  -- /dev/sdb
MOUNT_OPTIONS -- -o compress -o context=system_u:object_r:root_t:s0 
/dev/sdb /mnt/scratch

btrfs/325 1s ... - output mismatch (see 
/Volumes/work/ws/fstests/results//btrfs/325.out.bad)
     --- tests/btrfs/325.out	2024-10-30 08:07:48.507531048 +0800
     +++ /Volumes/work/ws/fstests/results//btrfs/325.out.bad	2024-10-30 
08:57:35.097186024 +0800
     @@ -8,8 +8,8 @@
      wrote 65536/65536 bytes at offset 196608
      XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
      Initial number of file extent items: 4
     -Number of file extent items after defrag with 128K treshold: 1
     -Number of file extent items after defrag with 256K treshold: 1
     +Number of file extent items after defrag with 128K treshold: 2
     +Number of file extent items after defrag with 256K treshold: 2
     ...
     (Run 'diff -u /Volumes/work/ws/fstests/tests/btrfs/325.out 
/Volumes/work/ws/fstests/results//btrfs/325.out.bad'  to see the entire 
diff)
--------------

One approach is to use random data, ensuring that compression remains
ineffective.


>> +echo -n "Initial number of file extent items: "
>> +count_file_extent_items
>> +
>> +# Read the whole file in order to load extent maps and merge them.
>> +cat $SCRATCH_MNT/foo > /dev/null
>> +
>> +# Now defragment with a threshold of 128K. After this we expect to 
>> get a file
> 
>> +# with 1 file extent item - the treshold is 128K but since all the 
>> extents are
> 
>> +# contiguous, they should be merged into a single one of 256K.
>> +$BTRFS_UTIL_PROG filesystem defragment -t 128K $SCRATCH_MNT/foo
> 
>> +echo -n "Number of file extent items after defrag with 128K treshold: "
> 
> Nit: s/treshold/threshold/g
> 
>> +count_file_extent_items
>> +
>> +# Read the whole file in order to load extent maps and merge them.
>> +cat $SCRATCH_MNT/foo > /dev/null
>> +
>> +# Now defragment with a threshold of 256K. After this we expect to 
>> get a file
>> +# with only 1 file extent item.
>> +$BTRFS_UTIL_PROG filesystem defragment -t 256K $SCRATCH_MNT/foo
>> +echo -n "Number of file extent items after defrag with 256K treshold: "
>> +count_file_extent_items
>> +
>> +# Check that the file has the expected data, that defrag didn't cause 
>> any data
>> +# loss or corruption.
>> +echo "File data after defrag:"
>> +_hexdump $SCRATCH_MNT/foo
>> +
> 
> Nice.
> 
> Nit: This can be a generic test-case.
> 

> Reviewed-by: Anand Jain <anand.jain@oracle.com>

Please ignore the rb for now.

Thanks, Anand


> 
> Thx, Anand
> 
> 
>> +status=0
>> +exit
>> diff --git a/tests/btrfs/325.out b/tests/btrfs/325.out
>> new file mode 100644
>> index 00000000..96053925
>> --- /dev/null
>> +++ b/tests/btrfs/325.out
>> @@ -0,0 +1,22 @@
>> +QA output created by 325
>> +wrote 65536/65536 bytes at offset 0
>> +XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
>> +wrote 65536/65536 bytes at offset 65536
>> +XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
>> +wrote 65536/65536 bytes at offset 131072
>> +XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
>> +wrote 65536/65536 bytes at offset 196608
>> +XXX Bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
>> +Initial number of file extent items: 4
>> +Number of file extent items after defrag with 128K treshold: 1
>> +Number of file extent items after defrag with 256K treshold: 1
>> +File data after defrag:
>> +000000 ab ab ab ab ab ab ab ab ab ab ab ab ab ab ab ab  
>> >................<
>> +*
>> +010000 cd cd cd cd cd cd cd cd cd cd cd cd cd cd cd cd  
>> >................<
>> +*
>> +020000 ef ef ef ef ef ef ef ef ef ef ef ef ef ef ef ef  
>> >................<
>> +*
>> +030000 73 73 73 73 73 73 73 73 73 73 73 73 73 73 73 73  
>> >ssssssssssssssss<
>> +*
>> +040000
>