[PATCH v2] xfstests: Introduce test case 285 to check statfs(2) will not cause ASSERT(XFS_IS_QUOTA_RUNNING(mp)) failed.

[Jeff Liu]

Hi Ben,

On 04/17/2012 01:49 AM, Ben Myers wrote:

> Jeff,
> 
> On Fri, Apr 13, 2012 at 05:09:17PM +0800, Jeff Liu wrote:
>> Call statfs(2) against a project directory cause ASSERT(XFS_IS_QUOTA_RUNNING(mp)) at xfs_qm_dqget() failed pre-vanilla kernel-3.4
>> if the "pquota" or "prjquota" mount option was not enabled on the underlying partition.  This test case can help checking it.
> 
> A nice find.
> 
>> +echo "Silence is golden."
>> +
>> +# Modify as appropriate.
>> +_supported_fs xfs
>> +_supported_os Linux
>> +_require_xfs_quota
>> +_require_scratch
>> +
>> +proj_dir="${SCRATCH_MNT}/test_project"
>> +test_file="${proj_dir}/test_file"
>> +
>> +_scratch_mkfs_xfs			>/dev/null 2>&1
>> +_qmount
>> +
>> +mkdir $proj_dir
>> +touch $test_file
>> +
>> +$XFS_QUOTA_PROG -x -c "project -s -p ${proj_dir} 2012" $SCRATCH_DEV \
>> +					>/dev/null 2>&1
>> +du -sh $proj_dir			>/dev/null 2>&1
> 
> I was unable to reproduce this until I changed 'du -sh' to 'df'.  Now it does
> crash consistently without your fix for xfs_fs_statfs().

Thanks for help review.
Strange, I can always reproducing that with du(1).  How about call `stat -f /xxx` to be a bit more explicit from the testing point of view?

> 
> With that change you can consider this
> Reviewed-by: Ben Myers <bpm at sgi.com>

Sure, I will add that.

Thanks,
-Jeff

> 
> -Ben
> 
> _______________________________________________
> xfs mailing list
> xfs at oss.sgi.com
> http://oss.sgi.com/mailman/listinfo/xfs