[PATCH 0/3] xfsprogs: xfs_quota: a few fixes

[PATCH 0/3] xfsprogs: xfs_quota: a few fixes

[Alex Elder]

The main problem solved in this series is that for project
quotas, the numbers of blocks reported by the xfs_quota "df"
subcommand is twice what it should be.  While looking at
this I also noticed two other problems that I corrected.

Below is a script that demonstrates the problem.

					-Alex

#!/bin/bash

############################################################
# Note that the following clobbers whatever is on /dev/sdk
# It also clobbers files /etc/project and /etc/projid.
# And it uses and then removes /qmnt.
############################################################

# Initialize the filesystem and mount it with project quotas
mkfs.xfs -f /dev/sdk
mkdir -p /qmnt
mount -t xfs -o prjquota /dev/sdk /qmnt

# Set up the project quota configuration files
echo test:1 > /etc/projid
echo 1:/qmnt/test > /etc/projects

# Make the top-level directory, set up quotas on it
mkdir /qmnt/test
xfs_quota -x -c 'project -s test' /qmnt

# Assign block quota limits 
xfs_quota -x -c 'limit -p bsoft=500m bhard=500m test' /qmnt

# The "report" command will show the right numbers
xfs_quota -x -c report /qmnt
# But the "df" command will show doubled block counts
xfs_quota -x -c df /qmnt

# This time using "human readable" output
xfs_quota -x -c 'report -h' /qmnt
xfs_quota -x -c 'df -h' /qmnt

# Clean up
umount /qmnt
rmdir /qmnt
rm /etc/projid /etc/projects

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

[PATCH 1/3] xfsprogs: xfs_quota: return real-time used data as intended

[Alex Elder]

In projects_free_space_data() the real-time used space consumption
is never set.  Instead, that value is returned in the field that
should hold the quota limit.

Found by inspection.  Never seen/noticed because we currently don't
support quotas when a filesystem has a realtime volume.

Signed-off-by: Alex Elder <aelder@sgi.com>
---
 quota/free.c |    4 +++-
 1 files changed, 3 insertions(+), 1 deletions(-)

diff --git a/quota/free.c b/quota/free.c
index 825ce34..26ec293 100644
--- a/quota/free.c
+++ b/quota/free.c
@@ -177,16 +177,18 @@ projects_free_space_data(
 		*bfree = (d.d_blk_softlimit - d.d_bcount) << 1;
 	}
 	*bused = d.d_bcount << 1;
+
 	if (d.d_ino_softlimit) {
 		*icount = d.d_ino_softlimit;
 		*ifree = (d.d_ino_softlimit - d.d_icount);
 	}
 	*iused = d.d_icount;
+
 	if (d.d_rtb_softlimit) {
 		*rcount = d.d_rtb_softlimit << 1;
 		*rfree = (d.d_rtb_softlimit - d.d_rtbcount) << 1;
 	}
-	*rcount = d.d_rtbcount << 1;
+	*rused = d.d_rtbcount << 1;
 
 	close(fd);
 	return 1;
-- 
1.7.6

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

[PATCH 2/3] xfsprogs: xfs_quota: don't double project block counts

[Alex Elder]

In projects_free_space_data() all of the block counts returned are
doubled.  This was probably a mistaken attempt to convert to or from
512-byte basic block units.  The caller expects the value returned
to be in 512-byte units, which is exactly what the fs_disk_quota
structure holds, so there should be no doubling.

The effect of this bug is that the disk space used by the "df"
xfs_quota command shows block counts twice what they should be.

SGI PV 1015651

Signed-off-by: Alex Elder <aelder@sgi.com>
---
 quota/free.c |   20 ++++++++++++++------
 1 files changed, 14 insertions(+), 6 deletions(-)

diff --git a/quota/free.c b/quota/free.c
index 26ec293..71a7cb2 100644
--- a/quota/free.c
+++ b/quota/free.c
@@ -41,6 +41,10 @@ free_help(void)
 "\n"));
 }
 
+/*
+ * The data and realtime block counts returned (count, used, and
+ * free) are all in basic block units.
+ */
 static int
 mount_free_space_data(
 	struct fs_path		*mount,
@@ -104,6 +108,10 @@ mount_free_space_data(
 	return 1;
 }
 
+/*
+ * The data and realtime block counts returned (count, used, and
+ * free) are all in basic block units.
+ */
 static int
 projects_free_space_data(
 	struct fs_path		*path,
@@ -173,10 +181,10 @@ projects_free_space_data(
 	}
 
 	if (d.d_blk_softlimit) {
-		*bcount = d.d_blk_softlimit << 1;
-		*bfree = (d.d_blk_softlimit - d.d_bcount) << 1;
+		*bcount = d.d_blk_softlimit;
+		*bfree = (d.d_blk_softlimit - d.d_bcount);
 	}
-	*bused = d.d_bcount << 1;
+	*bused = d.d_bcount;
 
 	if (d.d_ino_softlimit) {
 		*icount = d.d_ino_softlimit;
@@ -185,10 +193,10 @@ projects_free_space_data(
 	*iused = d.d_icount;
 
 	if (d.d_rtb_softlimit) {
-		*rcount = d.d_rtb_softlimit << 1;
-		*rfree = (d.d_rtb_softlimit - d.d_rtbcount) << 1;
+		*rcount = d.d_rtb_softlimit;
+		*rfree = (d.d_rtb_softlimit - d.d_rtbcount);
 	}
-	*rused = d.d_rtbcount << 1;
+	*rused = d.d_rtbcount;
 
 	close(fd);
 	return 1;
-- 
1.7.6

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

[PATCH 3/3] xfsprogs: xfs_quota: improve calculation for percentage display

[Alex Elder]

The pct_to_string() function determines the percentage it produces
in a strange way.  Simplify the function, and make it return the
simple rounded percentage value.  Handle the case of an error
return from snprintf() as well.

Signed-off-by: Alex Elder <aelder@sgi.com>
---
 quota/util.c |   20 ++++++++++----------
 1 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/quota/util.c b/quota/util.c
index 18ccae2..179aafd 100644
--- a/quota/util.c
+++ b/quota/util.c
@@ -172,18 +172,18 @@ num_to_string(
 
 char *
 pct_to_string(
-	__uint64_t	v,
-	__uint64_t	t,
-	char		*sp,
+	__uint64_t	portion,
+	__uint64_t	whole,
+	char		*buf,
 	uint		size)
 {
-	if (t == 0 || v == 0)
-		snprintf(sp, size, "%3u", (uint)0);
-	else if (t == v)
-		snprintf(sp, size, "%3u", (uint)100);
-	else
-		snprintf(sp, size, "%3u", (uint)(((double)v / t) * 100 + 1));
-	return sp;
+	uint		percent;
+
+	percent = whole ? (uint) (100.0 * portion / whole + 0.5) : 0;
+	if (snprintf(buf, size, "%3u", percent) < 0)
+		return "???";
+
+	return buf;
 }
 
 char *
-- 
1.7.6

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

Re: [PATCH 0/3] xfsprogs: xfs_quota: a few fixes

[Christoph Hellwig]

On Wed, Aug 24, 2011 at 04:53:40PM -0500, Alex Elder wrote:
> The main problem solved in this series is that for project
> quotas, the numbers of blocks reported by the xfs_quota "df"
> subcommand is twice what it should be.  While looking at
> this I also noticed two other problems that I corrected.
> 
> Below is a script that demonstrates the problem.

This script should be turned into an xfsqa testcase.

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

Re: [PATCH 1/3] xfsprogs: xfs_quota: return real-time used data as intended

[Christoph Hellwig]

On Wed, Aug 24, 2011 at 04:53:41PM -0500, Alex Elder wrote:
> In projects_free_space_data() the real-time used space consumption
> is never set.  Instead, that value is returned in the field that
> should hold the quota limit.
> 
> Found by inspection.  Never seen/noticed because we currently don't
> support quotas when a filesystem has a realtime volume.

Looks good,

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

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

Re: [PATCH 2/3] xfsprogs: xfs_quota: don't double project block counts

[Christoph Hellwig]

On Wed, Aug 24, 2011 at 04:53:42PM -0500, Alex Elder wrote:
> In projects_free_space_data() all of the block counts returned are
> doubled.  This was probably a mistaken attempt to convert to or from
> 512-byte basic block units.  The caller expects the value returned
> to be in 512-byte units, which is exactly what the fs_disk_quota
> structure holds, so there should be no doubling.
> 
> The effect of this bug is that the disk space used by the "df"
> xfs_quota command shows block counts twice what they should be.
> 
> SGI PV 1015651
> 
> Signed-off-by: Alex Elder <aelder@sgi.com>

Looks good,

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

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

Re: [PATCH 3/3] xfsprogs: xfs_quota: improve calculation for percentage display

[Christoph Hellwig]

On Wed, Aug 24, 2011 at 04:53:43PM -0500, Alex Elder wrote:
> The pct_to_string() function determines the percentage it produces
> in a strange way.  Simplify the function, and make it return the
> simple rounded percentage value.  Handle the case of an error
> return from snprintf() as well.

Looks fine,

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

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

Re: [PATCH 0/3] xfsprogs: xfs_quota: a few fixes

[Alex Elder]

On Thu, 2011-08-25 at 00:58 -0400, Christoph Hellwig wrote:
> On Wed, Aug 24, 2011 at 04:53:40PM -0500, Alex Elder wrote:
> > The main problem solved in this series is that for project
> > quotas, the numbers of blocks reported by the xfs_quota "df"
> > subcommand is twice what it should be.  While looking at
> > this I also noticed two other problems that I corrected.
> > 
> > Below is a script that demonstrates the problem.
> 
> This script should be turned into an xfsqa testcase.
> 

I knew someone would say that (but was hoping
I would not hear it)...

I will do it.

					-Alex

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