Re: [PATCH] fstests: generic/362: remove the old file to reflect new mount options

On Tue, May 26, 2026 at 8:02 AM Qu Wenruo <[email protected]> wrote:
>
> [HIDDEN BUG]
> There is a btrfs bug that will only trigger on newly formated TEST_DEV,
> with mount option "nodatasum":
>
>  FSTYP         -- btrfs
>  PLATFORM      -- Linux/x86_64 btrfs-vm 7.1.0-rc4-custom+ #381 SMP PREEMPT_DYNAMIC Tue May 26 10:47:14 ACST 2026
>  MKFS_OPTIONS  -- -O bgt -K /dev/mapper/test-scratch1
>  MOUNT_OPTIONS -- -o nodatasum /dev/mapper/test-scratch1 /mnt/scratch
>
>  generic/362  0s ... - output mismatch (see /home/adam/xfstests/results//generic/362.out.bad)
>      --- tests/generic/362.out  2024-08-24 15:31:37.200000000 +0930
>      +++ /home/adam/xfstests/results//generic/362.out.bad       2026-05-26 16:24:40.201201849 +0930
>      @@ -1,2 +1,3 @@
>       QA output created by 362
>      +First write failed: Input/output error
>       Silence is golden
>      ...
>      (Run 'diff -u /home/adam/xfstests/tests/generic/362.out /home/adam/xfstests/results//generic/362.out.bad'  to see the entire diff)
>
> But if one has formated TEST_DEV, run test with default mount option,
> then change the mount option to "nodatasum", the test will not fail
> anymore:
>
>  FSTYP         -- btrfs
>  PLATFORM      -- Linux/x86_64 btrfs-vm 7.1.0-rc4-custom+ #381 SMP PREEMPT_DYNAMIC Tue May 26 10:47:14 ACST 2026
>  MKFS_OPTIONS  -- -O bgt -K /dev/mapper/test-scratch1
>  MOUNT_OPTIONS -- /dev/mapper/test-scratch1 /mnt/scratch
>
>  generic/362  0s ...  0s
>  Ran: generic/362
>  Passed all 1 tests
>
>  FSTYP         -- btrfs
>  PLATFORM      -- Linux/x86_64 btrfs-vm 7.1.0-rc4-custom+ #381 SMP PREEMPT_DYNAMIC Tue May 26 10:47:14 ACST 2026
>  MKFS_OPTIONS  -- -O bgt -K /dev/mapper/test-scratch1
>  MOUNT_OPTIONS -- -o nodatasum /dev/mapper/test-scratch1 /mnt/scratch
>
>  generic/362  0s ...  0s
>  Ran: generic/362
>  Passed all 1 tests
>
> [CAUSE]
> Btrfs' nodatasum mount option only affect new files, but the test case
> itself is using TEST_DEV, and never delete the file
> "$TEST_DIR/dio-append-buf-fault"
>
> So if the file is created with default mount option, then all later
> "nodatasum" mount option will not affect that file, thus hide the test
> failure.
>
> [FIX]
> Always delete the target file "$TEST_DIR/dio-append-buf-fault" before
> running dio-append-buf-fault command.
>
> So that the new target file is always newly created and will follow
> btrfs' new mount option and expose the failure for nodatasum.
>
> Signed-off-by: Qu Wenruo <[email protected]>
> ---
> I'm already looking into the test failure, because it's mostly hidden by
> the btrfs' falling back to buffered IO behavior.
>
> With the incoming IOMAP_DIO_BOUNCE usage inside btrfs, it will expose
> the failure unconditionally.
> ---
>  tests/generic/362 | 4 ++++
>  1 file changed, 4 insertions(+)
>
> diff --git a/tests/generic/362 b/tests/generic/362
> index 0cfaa726..86191e57 100755
> --- a/tests/generic/362
> +++ b/tests/generic/362
> @@ -20,6 +20,10 @@ _require_test_program dio-append-buf-fault
>  _fixed_by_fs_commit btrfs 939b656bc8ab \
>         "btrfs: fix corruption after buffer fault in during direct IO append write"
>
> +# Remove the existing file, so a new inode can be created, and will be
> +# affected by changed mount options.
> +rm -rf $TEST_DIR/dio-append-buf-fault

We can just remove the file in a _cleanup function - that's what most
other tests that exercise faults during direct IO do, like
generic/647.
Some tests like this one are missing that, and there's at least one
more: generic/708. We had several similar tests, and many were copied
from another test.

Thanks.

> +
>  # On error the test program writes messages to stderr, causing a golden output
>  # mismatch and making the test fail.
>  $here/src/dio-append-buf-fault $TEST_DIR/dio-append-buf-fault
> --
> 2.51.2
>
>