[PATCH] xfs: make quota metadata truncation behavior consistent to user space

[PATCH] xfs: make quota metadata truncation behavior consistent to user space

[Jeff Liu]

From: Jie Liu <jeff.liu@oracle.com>

In xfs_qm_scall_trunc_qfiles(), we ignore the error if failed to remove
the users quota metadata and proceed to remove groups and projects if
they are being there.  However, in user space, the remove operation will
break and return if failed to remove any kind of quota.
Also for v5 super block, we can enabled both group and project quota at
the same time, in this case the current error handling will cover the
group error with projects but they might failed due to different reasons.

It seems we'd better the error handling consistent to the user space and
don't trying to remove another kind of quota metadata if the previous
operation is failed.

Signed-off-by: Jie Liu <jeff.liu@oracle.com>
---
 fs/xfs/xfs_qm_syscalls.c | 18 ++++++++++++------
 1 file changed, 12 insertions(+), 6 deletions(-)

diff --git a/fs/xfs/xfs_qm_syscalls.c b/fs/xfs/xfs_qm_syscalls.c
index 437c919..3daf5ea 100644
--- a/fs/xfs/xfs_qm_syscalls.c
+++ b/fs/xfs/xfs_qm_syscalls.c
@@ -278,7 +278,7 @@ xfs_qm_scall_trunc_qfiles(
 	xfs_mount_t	*mp,
 	uint		flags)
 {
-	int		error = 0, error2 = 0;
+	int		error;
 
 	if (!xfs_sb_version_hasquota(&mp->m_sb) || flags == 0) {
 		xfs_debug(mp, "%s: flags=%x m_qflags=%x",
@@ -286,14 +286,20 @@ xfs_qm_scall_trunc_qfiles(
 		return XFS_ERROR(EINVAL);
 	}
 
-	if (flags & XFS_DQ_USER)
+	if (flags & XFS_DQ_USER) {
 		error = xfs_qm_scall_trunc_qfile(mp, mp->m_sb.sb_uquotino);
-	if (flags & XFS_DQ_GROUP)
-		error2 = xfs_qm_scall_trunc_qfile(mp, mp->m_sb.sb_gquotino);
+		if (error)
+			return error;
+	}
+	if (flags & XFS_DQ_GROUP) {
+		error = xfs_qm_scall_trunc_qfile(mp, mp->m_sb.sb_gquotino);
+		if (error)
+			return error;
+	}
 	if (flags & XFS_DQ_PROJ)
-		error2 = xfs_qm_scall_trunc_qfile(mp, mp->m_sb.sb_pquotino);
+		error = xfs_qm_scall_trunc_qfile(mp, mp->m_sb.sb_pquotino);
 
-	return error ? error : error2;
+	return error;
 }
 
 /*
-- 
1.8.3.2

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: [PATCH] xfs: make quota metadata truncation behavior consistent to user space

[Christoph Hellwig]

On Fri, Nov 22, 2013 at 02:04:00PM +0800, Jeff Liu wrote:
> From: Jie Liu <jeff.liu@oracle.com>
> 
> In xfs_qm_scall_trunc_qfiles(), we ignore the error if failed to remove
> the users quota metadata and proceed to remove groups and projects if
> they are being there.  However, in user space, the remove operation will
> break and return if failed to remove any kind of quota.
> Also for v5 super block, we can enabled both group and project quota at
> the same time, in this case the current error handling will cover the
> group error with projects but they might failed due to different reasons.
> 
> It seems we'd better the error handling consistent to the user space and
> don't trying to remove another kind of quota metadata if the previous
> operation is failed.

Looks good.  Just curious, did you manage to reproduce an error
truncating the quota files?

Reviewed-by: Christoph Hellwig <hch@lst.de>

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: [PATCH] xfs: make quota metadata truncation behavior consistent to user space

[Jeff Liu]

On 11/22 2013 17:18, Christoph Hellwig wrote:
> On Fri, Nov 22, 2013 at 02:04:00PM +0800, Jeff Liu wrote:
>> From: Jie Liu <jeff.liu@oracle.com>
>>
>> In xfs_qm_scall_trunc_qfiles(), we ignore the error if failed to remove
>> the users quota metadata and proceed to remove groups and projects if
>> they are being there.  However, in user space, the remove operation will
>> break and return if failed to remove any kind of quota.
>> Also for v5 super block, we can enabled both group and project quota at
>> the same time, in this case the current error handling will cover the
>> group error with projects but they might failed due to different reasons.
>>
>> It seems we'd better the error handling consistent to the user space and
>> don't trying to remove another kind of quota metadata if the previous
>> operation is failed.
> 
> Looks good.  Just curious, did you manage to reproduce an error
> truncating the quota files?
Yup, I thought it over how to break it while occasionally glanced over this
function for backporting upstream patches this morning, but can not figure
out a simple method.

Thanks,
-Jeff 

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs