[PATCH] xfs: Keep sb_bad_features2 consistent with sb_features2

[PATCH] xfs: Keep sb_bad_features2 consistent with sb_features2

[Jan Kara]

Currently when we modify sb_features2, we store the same value also in
sb_bad_features2. However in most places we forget to mark field
sb_bad_features2 for logging and thus it can happen that a change to it
is lost. This results in an inconsistent sb_features2 and
sb_bad_features2 fields e.g. after xfstests test xfs/187.

Fix the problem by changing XFS_SB_FEATURES2 to actually mean both
sb_features2 and sb_bad_features2 fields since this is always what we
want to log. This isn't ideal because the fact that XFS_SB_FEATURES2
means two fields could cause some problem in future however the code is
hopefully less error prone that it is now.

Signed-off-by: Jan Kara <jack@suse.cz>
---
 fs/xfs/libxfs/xfs_format.h | 10 +++++-----
 fs/xfs/xfs_mount.c         |  5 ++---
 2 files changed, 7 insertions(+), 8 deletions(-)

diff --git a/fs/xfs/libxfs/xfs_format.h b/fs/xfs/libxfs/xfs_format.h
index fbd6da263571..476273287aaf 100644
--- a/fs/xfs/libxfs/xfs_format.h
+++ b/fs/xfs/libxfs/xfs_format.h
@@ -304,8 +304,8 @@ typedef enum {
 #define XFS_SB_ICOUNT		XFS_SB_MVAL(ICOUNT)
 #define XFS_SB_IFREE		XFS_SB_MVAL(IFREE)
 #define XFS_SB_FDBLOCKS		XFS_SB_MVAL(FDBLOCKS)
-#define XFS_SB_FEATURES2	XFS_SB_MVAL(FEATURES2)
-#define XFS_SB_BAD_FEATURES2	XFS_SB_MVAL(BAD_FEATURES2)
+#define XFS_SB_FEATURES2	(XFS_SB_MVAL(FEATURES2) | \
+				 XFS_SB_MVAL(BAD_FEATURES2))
 #define XFS_SB_FEATURES_COMPAT	XFS_SB_MVAL(FEATURES_COMPAT)
 #define XFS_SB_FEATURES_RO_COMPAT XFS_SB_MVAL(FEATURES_RO_COMPAT)
 #define XFS_SB_FEATURES_INCOMPAT XFS_SB_MVAL(FEATURES_INCOMPAT)
@@ -319,9 +319,9 @@ typedef enum {
 	 XFS_SB_VERSIONNUM | XFS_SB_UQUOTINO | XFS_SB_GQUOTINO | \
 	 XFS_SB_QFLAGS | XFS_SB_SHARED_VN | XFS_SB_UNIT | XFS_SB_WIDTH | \
 	 XFS_SB_ICOUNT | XFS_SB_IFREE | XFS_SB_FDBLOCKS | XFS_SB_FEATURES2 | \
-	 XFS_SB_BAD_FEATURES2 | XFS_SB_FEATURES_COMPAT | \
-	 XFS_SB_FEATURES_RO_COMPAT | XFS_SB_FEATURES_INCOMPAT | \
-	 XFS_SB_FEATURES_LOG_INCOMPAT | XFS_SB_PQUOTINO)
+	 XFS_SB_FEATURES_COMPAT | XFS_SB_FEATURES_RO_COMPAT | \
+	 XFS_SB_FEATURES_INCOMPAT | XFS_SB_FEATURES_LOG_INCOMPAT | \
+	 XFS_SB_PQUOTINO)
 
 
 /*
diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c
index d3d38836f87f..71d2c97cf58a 100644
--- a/fs/xfs/xfs_mount.c
+++ b/fs/xfs/xfs_mount.c
@@ -678,7 +678,7 @@ xfs_mountfs(
 		xfs_warn(mp, "correcting sb_features alignment problem");
 		sbp->sb_features2 |= sbp->sb_bad_features2;
 		sbp->sb_bad_features2 = sbp->sb_features2;
-		mp->m_update_flags |= XFS_SB_FEATURES2 | XFS_SB_BAD_FEATURES2;
+		mp->m_update_flags |= XFS_SB_FEATURES2;
 
 		/*
 		 * Re-check for ATTR2 in case it was found in bad_features2
@@ -1436,8 +1436,7 @@ xfs_mount_log_sb(
 	int		error;
 
 	ASSERT(fields & (XFS_SB_UNIT | XFS_SB_WIDTH | XFS_SB_UUID |
-			 XFS_SB_FEATURES2 | XFS_SB_BAD_FEATURES2 |
-			 XFS_SB_VERSIONNUM));
+			 XFS_SB_FEATURES2 | XFS_SB_VERSIONNUM));
 
 	tp = xfs_trans_alloc(mp, XFS_TRANS_SB_UNIT);
 	error = xfs_trans_reserve(tp, &M_RES(mp)->tr_sb, 0, 0);
-- 
1.8.4.5

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

Re: [PATCH] xfs: Keep sb_bad_features2 consistent with sb_features2

[Dave Chinner]

On Wed, Dec 17, 2014 at 09:27:26PM +0100, Jan Kara wrote:
> Currently when we modify sb_features2, we store the same value also in
> sb_bad_features2. However in most places we forget to mark field
> sb_bad_features2 for logging and thus it can happen that a change to it
> is lost. This results in an inconsistent sb_features2 and
> sb_bad_features2 fields e.g. after xfstests test xfs/187.
> 
> Fix the problem by changing XFS_SB_FEATURES2 to actually mean both
> sb_features2 and sb_bad_features2 fields since this is always what we
> want to log. This isn't ideal because the fact that XFS_SB_FEATURES2
> means two fields could cause some problem in future however the code is
> hopefully less error prone that it is now.

Actually, I have patches that fix this differently that I'm planning
to push for the 3.20 cycle.  They get rid of the "update random SB
fields" problem altogether simply by logging and updating the entire
SB every time.

http://oss.sgi.com/archives/xfs/2014-09/msg00448.html

> Signed-off-by: Jan Kara <jack@suse.cz>
> ---
>  fs/xfs/libxfs/xfs_format.h | 10 +++++-----
>  fs/xfs/xfs_mount.c         |  5 ++---
>  2 files changed, 7 insertions(+), 8 deletions(-)
> 
> diff --git a/fs/xfs/libxfs/xfs_format.h b/fs/xfs/libxfs/xfs_format.h
> index fbd6da263571..476273287aaf 100644
> --- a/fs/xfs/libxfs/xfs_format.h
> +++ b/fs/xfs/libxfs/xfs_format.h
> @@ -304,8 +304,8 @@ typedef enum {
>  #define XFS_SB_ICOUNT		XFS_SB_MVAL(ICOUNT)
>  #define XFS_SB_IFREE		XFS_SB_MVAL(IFREE)
>  #define XFS_SB_FDBLOCKS		XFS_SB_MVAL(FDBLOCKS)
> -#define XFS_SB_FEATURES2	XFS_SB_MVAL(FEATURES2)
> -#define XFS_SB_BAD_FEATURES2	XFS_SB_MVAL(BAD_FEATURES2)
> +#define XFS_SB_FEATURES2	(XFS_SB_MVAL(FEATURES2) | \
> +				 XFS_SB_MVAL(BAD_FEATURES2))

That doesn't guarantee that both fields are updated together, just
that the'll get logged and written together. The code still has to
ensure both fields are updated beforehand. I have a patch to address
this as well, and will post them soon.

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com

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

Re: [PATCH] xfs: Keep sb_bad_features2 consistent with sb_features2

[Jan Kara]

On Thu 18-12-14 08:22:55, Dave Chinner wrote:
> On Wed, Dec 17, 2014 at 09:27:26PM +0100, Jan Kara wrote:
> > Currently when we modify sb_features2, we store the same value also in
> > sb_bad_features2. However in most places we forget to mark field
> > sb_bad_features2 for logging and thus it can happen that a change to it
> > is lost. This results in an inconsistent sb_features2 and
> > sb_bad_features2 fields e.g. after xfstests test xfs/187.
> > 
> > Fix the problem by changing XFS_SB_FEATURES2 to actually mean both
> > sb_features2 and sb_bad_features2 fields since this is always what we
> > want to log. This isn't ideal because the fact that XFS_SB_FEATURES2
> > means two fields could cause some problem in future however the code is
> > hopefully less error prone that it is now.
> 
> Actually, I have patches that fix this differently that I'm planning
> to push for the 3.20 cycle.  They get rid of the "update random SB
> fields" problem altogether simply by logging and updating the entire
> SB every time.
> 
> http://oss.sgi.com/archives/xfs/2014-09/msg00448.html
  OK, thanks for info.

								Honza
> 
> > Signed-off-by: Jan Kara <jack@suse.cz>
> > ---
> >  fs/xfs/libxfs/xfs_format.h | 10 +++++-----
> >  fs/xfs/xfs_mount.c         |  5 ++---
> >  2 files changed, 7 insertions(+), 8 deletions(-)
> > 
> > diff --git a/fs/xfs/libxfs/xfs_format.h b/fs/xfs/libxfs/xfs_format.h
> > index fbd6da263571..476273287aaf 100644
> > --- a/fs/xfs/libxfs/xfs_format.h
> > +++ b/fs/xfs/libxfs/xfs_format.h
> > @@ -304,8 +304,8 @@ typedef enum {
> >  #define XFS_SB_ICOUNT		XFS_SB_MVAL(ICOUNT)
> >  #define XFS_SB_IFREE		XFS_SB_MVAL(IFREE)
> >  #define XFS_SB_FDBLOCKS		XFS_SB_MVAL(FDBLOCKS)
> > -#define XFS_SB_FEATURES2	XFS_SB_MVAL(FEATURES2)
> > -#define XFS_SB_BAD_FEATURES2	XFS_SB_MVAL(BAD_FEATURES2)
> > +#define XFS_SB_FEATURES2	(XFS_SB_MVAL(FEATURES2) | \
> > +				 XFS_SB_MVAL(BAD_FEATURES2))
> 
> That doesn't guarantee that both fields are updated together, just
> that the'll get logged and written together. The code still has to
> ensure both fields are updated beforehand. I have a patch to address
> this as well, and will post them soon.
> 
> Cheers,
> 
> Dave.
> -- 
> Dave Chinner
> david@fromorbit.com
-- 
Jan Kara <jack@suse.cz>
SUSE Labs, CR

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

Re: [PATCH] xfs: Keep sb_bad_features2 consistent with sb_features2

[Dave Chinner]

On Thu, Dec 18, 2014 at 10:59:19AM +0100, Jan Kara wrote:
> On Thu 18-12-14 08:22:55, Dave Chinner wrote:
> > On Wed, Dec 17, 2014 at 09:27:26PM +0100, Jan Kara wrote:
> > > Currently when we modify sb_features2, we store the same value also in
> > > sb_bad_features2. However in most places we forget to mark field
> > > sb_bad_features2 for logging and thus it can happen that a change to it
> > > is lost. This results in an inconsistent sb_features2 and
> > > sb_bad_features2 fields e.g. after xfstests test xfs/187.
> > > 
> > > Fix the problem by changing XFS_SB_FEATURES2 to actually mean both
> > > sb_features2 and sb_bad_features2 fields since this is always what we
> > > want to log. This isn't ideal because the fact that XFS_SB_FEATURES2
> > > means two fields could cause some problem in future however the code is
> > > hopefully less error prone that it is now.
> > 
> > Actually, I have patches that fix this differently that I'm planning
> > to push for the 3.20 cycle.  They get rid of the "update random SB
> > fields" problem altogether simply by logging and updating the entire
> > SB every time.
> > 
> > http://oss.sgi.com/archives/xfs/2014-09/msg00448.html
>   OK, thanks for info.

FWIW, I'll pull the fixes you have posted and rebase the rework I
have on top of them. That way we have fixes that can be backported
easily without pulling the entire rework in the commit history.

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com

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