[PATCH] xfs/107: fix formatting failures

[PATCH] xfs/107: fix formatting failures

[Darrick J. Wong]

From: Darrick J. Wong <djwong@kernel.org>

Zorro Lang reported that the _scratch_mkfs_sized call in the new xfs/107
fstest sometimes fails on more exotic storage due to insufficient log
size on account of raid stripes, etc.   These are side effects of the
filesystem being too small.

Change the filesystem size to 256M to avoid these problems, and change
the allocstale parameters to use the same file size (16M) as before.
Given that ALLOCSP produces stale disk contents pretty quickly this
shouldn't affect the test runtime too much.

Reported-by: Zorro Lang <zlang@redhat.com>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
---
 tests/xfs/107 |   10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)

diff --git a/tests/xfs/107 b/tests/xfs/107
index 6034dbc2..577094b2 100755
--- a/tests/xfs/107
+++ b/tests/xfs/107
@@ -22,7 +22,10 @@ _require_test
 _require_scratch
 _require_test_program allocstale
 
-size_mb=32
+# Create a 256MB filesystem to avoid running into mkfs problems with too-small
+# filesystems.
+size_mb=256
+
 # Write a known pattern to the disk so that we can detect stale disk blocks
 # being mapped into the file.  In the test author's experience, the bug will
 # reproduce within the first 500KB's worth of ALLOCSP calls, so running up
@@ -39,9 +42,10 @@ _scratch_mount
 _xfs_force_bdev data $SCRATCH_MNT
 testfile=$SCRATCH_MNT/a
 
-# Allow the test program to expand the file to consume half the free space.
+# Allow the test program to expand the file to 32MB.  If we can't find any
+# stale blocks at that point, the kernel has probably been patched.
 blksz=$(_get_file_block_size $SCRATCH_MNT)
-iterations=$(( (size_mb / 2) * 1048576 / blksz))
+iterations=$(( (size_mb / 16) * 1048576 / blksz))
 echo "Setting up $iterations runs for block size $blksz" >> $seqres.full
 
 # Run reproducer program and dump file contents if we see stale data.  Full

Re: [PATCH] xfs/107: fix formatting failures

[Zorro Lang]

On Wed, Jan 19, 2022 at 04:49:44PM -0800, Darrick J. Wong wrote:
> From: Darrick J. Wong <djwong@kernel.org>
> 
> Zorro Lang reported that the _scratch_mkfs_sized call in the new xfs/107
> fstest sometimes fails on more exotic storage due to insufficient log
> size on account of raid stripes, etc.   These are side effects of the
> filesystem being too small.
> 
> Change the filesystem size to 256M to avoid these problems, and change
> the allocstale parameters to use the same file size (16M) as before.
> Given that ALLOCSP produces stale disk contents pretty quickly this
> shouldn't affect the test runtime too much.
> 
> Reported-by: Zorro Lang <zlang@redhat.com>
> Signed-off-by: Darrick J. Wong <djwong@kernel.org>
> ---

Oh, I didn't notice that the xfs/107 has been merged :-P
This patch is good to me, and it fixes two "data/log space too small"
issues on my side.

Reviewed-by: Zorro Lang <zlang@redhat.com>

>  tests/xfs/107 |   10 +++++++---
>  1 file changed, 7 insertions(+), 3 deletions(-)
> 
> diff --git a/tests/xfs/107 b/tests/xfs/107
> index 6034dbc2..577094b2 100755
> --- a/tests/xfs/107
> +++ b/tests/xfs/107
> @@ -22,7 +22,10 @@ _require_test
>  _require_scratch
>  _require_test_program allocstale
>  
> -size_mb=32
> +# Create a 256MB filesystem to avoid running into mkfs problems with too-small
> +# filesystems.
> +size_mb=256
> +
>  # Write a known pattern to the disk so that we can detect stale disk blocks
>  # being mapped into the file.  In the test author's experience, the bug will
>  # reproduce within the first 500KB's worth of ALLOCSP calls, so running up
> @@ -39,9 +42,10 @@ _scratch_mount
>  _xfs_force_bdev data $SCRATCH_MNT
>  testfile=$SCRATCH_MNT/a
>  
> -# Allow the test program to expand the file to consume half the free space.
> +# Allow the test program to expand the file to 32MB.  If we can't find any
> +# stale blocks at that point, the kernel has probably been patched.
>  blksz=$(_get_file_block_size $SCRATCH_MNT)
> -iterations=$(( (size_mb / 2) * 1048576 / blksz))
> +iterations=$(( (size_mb / 16) * 1048576 / blksz))
>  echo "Setting up $iterations runs for block size $blksz" >> $seqres.full
>  
>  # Run reproducer program and dump file contents if we see stale data.  Full
>