Re: [PATCH v3] generic/430: add copy in same file test

[Zorro Lang <zlang@redhat.com>]

On Wed, Sep 04, 2019 at 09:01:34PM +0800, Jianhong.Yin wrote:
> now xfstests has not cover the senario that copy in same file
> related bug:
>  copy_file_range return "Invalid argument" when copy in the same file
>  https://bugzilla.kernel.org/show_bug.cgi?id=202935
> 
> step1:
>  copy whole file to itself, and compare with original file.
> step2:
>  copy whole file to end of itself, and compare the two halves with the original file
> step3:
>  recover from original file.
> 
> Signed-off-by: Jianhong Yin <yin-jianhong@163.com>
> ---
> 
> this is a base function test, I don't think we should add a new
> test for it like 553 554, so add it in generic/430
> what do you think?
> 
>  tests/generic/430     | 12 ++++++++++++
>  tests/generic/430.out |  4 ++++
>  2 files changed, 16 insertions(+)
> 
> diff --git a/tests/generic/430 b/tests/generic/430
> index 1b11f60d..4ce0ee26 100755
> --- a/tests/generic/430
> +++ b/tests/generic/430
> @@ -6,6 +6,7 @@
>  #
>  # Tests vfs_copy_file_range():
>  #   - Copy a file
> +#   - Copy in same file
>  #   - Copy beginning of original to new file
>  #   - Copy middle of original to a new file
>  #   - Copy end of original to new file
> @@ -52,6 +53,17 @@ cmp $testdir/file $testdir/copy
>  echo "Original md5sums:"
>  md5sum $testdir/{file,copy} | _filter_test_dir
>  
> +fsize=$(stat -c %s $testdir/copy)

I don't think you need to do `stat` to get the file size, due to this case
has definite fixed file size (5000 bytes). If it fails to write 5000 bytes,
the case should be failed for something wrong.

> +echo "Copy in same file"
> +$XFS_IO_PROG    -c "copy_range -l $fsize $testdir/copy" "$testdir/copy"
> +cmp $testdir/file $testdir/copy
> +$XFS_IO_PROG    -c "copy_range -l $fsize -d $fsize $testdir/copy" "$testdir/copy"
> +cmp -n $fsize $testdir/file $testdir/copy
> +cmp -i 0:$fsize $testdir/file $testdir/copy

I think test once on same file is enough, but I don't have objection if you and
maintainer preper twice.

> +$XFS_IO_PROG -f -c "copy_range $testdir/file" "$testdir/copy"
> +echo "md5sums after copying in same file and recover:"
> +md5sum $testdir/{file,copy} | _filter_test_dir

This test base on an implicit operation that xfs_io copy_range command always
truncates target file to 0 before doing copy_file_range, if there's not
-l or -d or -s specified.

So I think if you'd like to do this test, I prefer truncating $testdir/copy
by an explicit operation (e.g: use -t option, or -c 'truncate 0', or even
rm -f $testdir/copy).

And as above, I don't think this test is necessary, but I don't have objection
if you and maintainer preper that.

Thanks,
Zorro

> +
>  echo "Copy beginning of original file"
>  $XFS_IO_PROG -f -c "copy_range -l 1000 $testdir/file" "$testdir/beginning"
>  cmp -n 1000 $testdir/file $testdir/beginning
> diff --git a/tests/generic/430.out b/tests/generic/430.out
> index 4b4ca75d..57584957 100644
> --- a/tests/generic/430.out
> +++ b/tests/generic/430.out
> @@ -3,6 +3,10 @@ Create the original file and then copy
>  Original md5sums:
>  e11fbace556cba26bf0076e74cab90a3  TEST_DIR/test-430/file
>  e11fbace556cba26bf0076e74cab90a3  TEST_DIR/test-430/copy
> +Copy in same file
> +md5sums after copying in same file and recover:
> +e11fbace556cba26bf0076e74cab90a3  TEST_DIR/test-430/file
> +e11fbace556cba26bf0076e74cab90a3  TEST_DIR/test-430/copy
>  Copy beginning of original file
>  md5sums after copying beginning:
>  e11fbace556cba26bf0076e74cab90a3  TEST_DIR/test-430/file
> -- 
> 2.21.0
>