Skip to contentSkip to conversation

xfs: get new buffer for secondary superblocks from the first new AG

PATCH In review Jeff Liu 2 participants 3 messages Mar 21, 2013 XFS filesystem

lore
Share & tools
lore

Jeff Liu

author
#0 lore raw

PATCH 1/1 xfs: get new buffer for secondary superblocks from the first new AG

In growfs operation, the unused parts of superblocks in the latest old AG are zero filled, hence we can safely read the secondary superblock buffer of it rather than getting a new buffer for it again.

This is a small refinement for commits 1375cb65.

Cc Dave Chinner <[email protected]>
---
 fs/xfs/xfs_fsops.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
fs/xfs/xfs_fsops.c +1 -1
index 94eaeed..7d435dd 100644
@@ -449,7 +449,7 @@ xfs_growfs_data_private(
 		 * disk as the contents of the new area we are growing into is
 		 * completely unknown.
 		 */
-		if (agno < oagcount) {
+		if (agno <= oagcount) {
 			error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp,
 				  XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
 				  XFS_FSS_TO_BB(mp, 1), 0, &bp,
Signature · 6 lines
1.7.9.5

_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs
#1 lore raw
Quoted 5 lines · Jeff Liu

In growfs operation, the unused parts of superblocks in the latest old AG are zero filled, hence we can safely read the secondary superblock buffer of it rather than getting a new buffer for it again.

From the code:

	oagcount = mp->m_sb.sb_agcount;

And because the AG indexes are zero numbered, the index of the last original AGs is (oagcount - 1). Hence we shoul donly be reading from the AG headers if the current AG index is less than oagcount.

Quoted 13 lines

This is a small refinement for commits 1375cb65.

Cc Dave Chinner <[email protected]>
---
 fs/xfs/xfs_fsops.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
fs/xfs/xfs_fsops.c +0 -0
index 94eaeed..7d435dd 100644

fs/xfs/xfs_fsops.c xfs_growfs_data_private( · line 452

Quoted 5 lines
fs/xfs/xfs_fsops.c +1 -1
@@ -449,7 +449,7 @@ xfs_growfs_data_private(
 		 * disk as the contents of the new area we are growing into is
 		 * completely unknown.
 		 */
-		if (agno < oagcount) {
+		if (agno <= oagcount) {

With this change, when agno == oagcount we are one AG beyond the original filesystem size and so if we read that space, we are reading uninitialised data.

AFAICT, the current code is correct...

Cheers,

Dave.

Signature · 7 lines
Dave Chinner
[email protected]

_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs
author
#2 lore raw
Quoted 36 lines · Dave Chinner
Quoted 3 lines · Jeff Liu

In growfs operation, the unused parts of superblocks in the latest old AG are zero filled, hence we can safely read the secondary superblock buffer of it rather than getting a new buffer for it again.

From the code:

	oagcount = mp->m_sb.sb_agcount;

And because the AG indexes are zero numbered, the index of the last original AGs is (oagcount - 1). Hence we shoul donly be reading from the AG headers if the current AG index is less than oagcount.

Quoted 13 lines

This is a small refinement for commits 1375cb65.

Cc Dave Chinner <[email protected]>
---
 fs/xfs/xfs_fsops.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
fs/xfs/xfs_fsops.c +0 -0
index 94eaeed..7d435dd 100644

fs/xfs/xfs_fsops.c xfs_growfs_data_private( · line 452

Quoted 5 lines
fs/xfs/xfs_fsops.c +1 -1
@@ -449,7 +449,7 @@ xfs_growfs_data_private(
 		 * disk as the contents of the new area we are growing into is
 		 * completely unknown.
 		 */
-		if (agno < oagcount) {
+		if (agno <= oagcount) {

With this change, when agno == oagcount we are one AG beyond the original filesystem size and so if we read that space, we are reading uninitialised data.

Quoted 1 line

AFAICT, the current code is correct...

Oops, I made a stupid mistake. The index of the last original AGs is (oagcount - 1) rather than oagcount...

Sorry for the noise!

Thanks,
-Jeff

_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs