Re: [PATCH] common/xfs: filter mkfs.xfs stripe geometry warning

[Eric Sandeen <sandeen@sandeen.net>]

On 8/24/26 11:33 PM, Manognya Singuru wrote:
> On Mon, Aug 24, 2026 at 9:41 PM Darrick J. Wong <djwong@kernel.org> wrote:
>>
>> On Mon, Aug 24, 2026 at 12:26:04PM +0530, Manognya Singuru wrote:
>>> mkfs.xfs prints a warning when creating small filesystems on devices
>>> that advertise stripe geometry:
>>>
>>> mkfs.xfs: small data volume, ignoring data volume stripe unit XXX and
>>> stripe width XXX
>>>
>>> This warning causes several XFS tests to fail their golden output
>>> comparison when SCRATCH_DEV has striped device geometry such as
>>> scsi_debug devices.
>>>
>>> Filter the warning in _try_scratch_mkfs_xfs() instead of
>>> updating individual test expected outputs, as the filesystem is created
>>> successfully and the warning does not indicate a failure.
>>>
>>> Signed-off-by: Manognya Singuru <msinguru@redhat.com>
>>> ---
>>> Tested with xfs/015, xfs/030, xfs/041, xfs/042, xfs/075, xfs/104,
>>> xfs/109, xfs/168 and xfs/177.
>>>
>>> Output for xfs/030 before the filter:
>>> QA output created by 030
>>> ...
>>> +mkfs.xfs: small data volume, ignoring data volume stripe unit 64 and
>>
>> How big is this volume?  I'm guessing less than a gigabyte or two?
>>
> 
> Yes, It is 1 gigabyte.


If you're curious, you can demonstrate it with a device like:

modprobe scsi_debug dev_size_mb=1024 opt_xferlen_exp=6 opt_blks=64

- that makes a 1G device with the stripe unit 8 and stripe width 8 from mkfs.xfs's perspective

I think I saw this in "real" testing in some environment but I'm afraid I don't
remember how big (small) those devices actually were. The code looks like it
only emits it for < GIGABLOCKS(1 ...) but I don't know why I would have been
testing with a 1G device.

Worth filtering out anyway, I think.

-Eric

> 
>>> stripe width 64
>>> ...
>>> Ran: xfs/030
>>> Failures: xfs/030
>>> Failed 1 of 1 tests
>>>
>>> Output for xfs/030 after adding the filter:
>>> xfs/030             15s
>>> Ran: xfs/030
>>> Passed all 1 tests
>>>
>>>  common/xfs | 3 ++-
>>>  1 file changed, 2 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/common/xfs b/common/xfs
>>> index 81add208..98981e62 100644
>>> --- a/common/xfs
>>> +++ b/common/xfs
>>> @@ -157,7 +157,8 @@ _try_scratch_mkfs_xfs()
>>>       local mkfs_cmd="`_scratch_mkfs_xfs_opts`"
>>>       local mkfs_filter="sed -e '/less than device physical sector/d' \
>>>                              -e '/switching to logical sector/d' \
>>> -                            -e '/Default configuration/d'"
>>> +                            -e '/Default configuration/d' \
>>> +                            -e '/small data volume, ignoring data volume stripe unit/d'"
>>
>> But yeah, that's not output that we care about.
> 
> Yes, many tests fail due to these extra lines in the output.
> 
>> Reviewed-by: "Darrick J. Wong" <djwong@kernel.org>
>>
>> --D
>>
>>>       local tmp=`mktemp -u`
>>>       local mkfs_status
>>>
>>> --
>>> 2.54.0
>>>
>>>
>>
> 
>