GFS2: Pre-pull patch posting (fixes)

GFS2: Pre-pull patch posting (fixes)

[Steven Whitehouse]

Hi,

Here are a couple of small fixes which just missed the previous pull
request. Both fairly short and self-explanatory,

Steve.

[Cluster-devel] GFS2: Pre-pull patch posting (fixes)

[Steven Whitehouse]

Hi,

Here are a couple of small fixes which just missed the previous pull
request. Both fairly short and self-explanatory,

Steve.

[PATCH 1/2] GFS2: Wipe directory hash table metadata when deallocating a directory

[Steven Whitehouse]

The deallocation code for directories in GFS2 is largely divided into
two parts. The first part deallocates any directory leaf blocks and
marks the directory as being a regular file when that is complete. The
second stage was identical to deallocating regular files.

Regular files have their data blocks in a different
address space to directories, and thus what would have been normal data
blocks in a regular file (the hash table in a GFS2 directory) were
deallocated correctly. However, a reference to these blocks was left in the
journal (assuming of course that some previous activity had resulted in
those blocks being in the journal or ail list).

This patch uses the i_depth as a test of whether the inode is an
exhash directory (we cannot test the inode type as that has already
been changed to a regular file at this stage in deallocation)

The original issue was reported by Chris Hertel as an issue he encountered
running bonnie++

Reported-by: Christopher R. Hertel <crh@samba.org>
Cc: Abhijith Das <adas@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index 74add2d..e65493a 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -780,6 +780,8 @@ static int do_strip(struct gfs2_inode *ip, struct buffer_head *dibh,
 	metadata = (height != ip->i_height - 1);
 	if (metadata)
 		revokes = (height) ? sdp->sd_inptrs : sdp->sd_diptrs;
+	else if (ip->i_depth)
+		revokes = sdp->sd_inptrs;
 
 	if (ip != GFS2_I(sdp->sd_rindex))
 		error = gfs2_rindex_hold(sdp, &ip->i_alloc->al_ri_gh);
diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c
index 7273ad3..9b780df 100644
--- a/fs/gfs2/rgrp.c
+++ b/fs/gfs2/rgrp.c
@@ -1629,6 +1629,10 @@ void __gfs2_free_data(struct gfs2_inode *ip, u64 bstart, u32 blen)
 	gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data);
 
 	gfs2_trans_add_rg(rgd);
+
+	/* Directories keep their data in the metadata address space */
+	if (ip->i_depth)
+		gfs2_meta_wipe(ip, bstart, blen);
 }
 
 /**
-- 
1.7.4

[Cluster-devel] [PATCH 1/2] GFS2: Wipe directory hash table metadata when deallocating a directory

[Steven Whitehouse]

The deallocation code for directories in GFS2 is largely divided into
two parts. The first part deallocates any directory leaf blocks and
marks the directory as being a regular file when that is complete. The
second stage was identical to deallocating regular files.

Regular files have their data blocks in a different
address space to directories, and thus what would have been normal data
blocks in a regular file (the hash table in a GFS2 directory) were
deallocated correctly. However, a reference to these blocks was left in the
journal (assuming of course that some previous activity had resulted in
those blocks being in the journal or ail list).

This patch uses the i_depth as a test of whether the inode is an
exhash directory (we cannot test the inode type as that has already
been changed to a regular file at this stage in deallocation)

The original issue was reported by Chris Hertel as an issue he encountered
running bonnie++

Reported-by: Christopher R. Hertel <crh@samba.org>
Cc: Abhijith Das <adas@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index 74add2d..e65493a 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -780,6 +780,8 @@ static int do_strip(struct gfs2_inode *ip, struct buffer_head *dibh,
 	metadata = (height != ip->i_height - 1);
 	if (metadata)
 		revokes = (height) ? sdp->sd_inptrs : sdp->sd_diptrs;
+	else if (ip->i_depth)
+		revokes = sdp->sd_inptrs;
 
 	if (ip != GFS2_I(sdp->sd_rindex))
 		error = gfs2_rindex_hold(sdp, &ip->i_alloc->al_ri_gh);
diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c
index 7273ad3..9b780df 100644
--- a/fs/gfs2/rgrp.c
+++ b/fs/gfs2/rgrp.c
@@ -1629,6 +1629,10 @@ void __gfs2_free_data(struct gfs2_inode *ip, u64 bstart, u32 blen)
 	gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data);
 
 	gfs2_trans_add_rg(rgd);
+
+	/* Directories keep their data in the metadata address space */
+	if (ip->i_depth)
+		gfs2_meta_wipe(ip, bstart, blen);
 }
 
 /**
-- 
1.7.4

[PATCH 2/2] GFS2: Wait properly when flushing the ail list

[Steven Whitehouse]

The ail flush code has always relied upon log flushing to prevent
it from spinning needlessly. This fixes it to wait on the last
I/O request submitted (we don't need to wait for all of it)
instead of either spinning with io_schedule or sleeping.

As a result cpu usage of gfs2_logd is much reduced with certain
workloads.

Reported-by: Abhijith Das <adas@redhat.com>
Tested-by: Abhijith Das <adas@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/log.c b/fs/gfs2/log.c
index cec26c0..903115f 100644
--- a/fs/gfs2/log.c
+++ b/fs/gfs2/log.c
@@ -228,6 +228,27 @@ static int gfs2_ail1_empty(struct gfs2_sbd *sdp)
 	return ret;
 }
 
+static void gfs2_ail1_wait(struct gfs2_sbd *sdp)
+{
+	struct gfs2_ail *ai;
+	struct gfs2_bufdata *bd;
+	struct buffer_head *bh;
+
+	spin_lock(&sdp->sd_ail_lock);
+	list_for_each_entry_reverse(ai, &sdp->sd_ail1_list, ai_list) {
+		list_for_each_entry(bd, &ai->ai_ail1_list, bd_ail_st_list) {
+			bh = bd->bd_bh;
+			if (!buffer_locked(bh))
+				continue;
+			get_bh(bh);
+			spin_unlock(&sdp->sd_ail_lock);
+			wait_on_buffer(bh);
+			brelse(bh);
+			return;
+		}
+	}
+	spin_unlock(&sdp->sd_ail_lock);
+}
 
 /**
  * gfs2_ail2_empty_one - Check whether or not a trans in the AIL has been synced
@@ -878,9 +899,9 @@ void gfs2_meta_syncfs(struct gfs2_sbd *sdp)
 	gfs2_log_flush(sdp, NULL);
 	for (;;) {
 		gfs2_ail1_start(sdp);
+		gfs2_ail1_wait(sdp);
 		if (gfs2_ail1_empty(sdp))
 			break;
-		msleep(10);
 	}
 }
 
@@ -920,12 +941,14 @@ int gfs2_logd(void *data)
 
 		if (gfs2_ail_flush_reqd(sdp)) {
 			gfs2_ail1_start(sdp);
-			io_schedule();
+			gfs2_ail1_wait(sdp);
 			gfs2_ail1_empty(sdp);
 			gfs2_log_flush(sdp, NULL);
 		}
 
-		wake_up(&sdp->sd_log_waitq);
+		if (!gfs2_ail_flush_reqd(sdp))
+			wake_up(&sdp->sd_log_waitq);
+
 		t = gfs2_tune_get(sdp, gt_logd_secs) * HZ;
 		if (freezing(current))
 			refrigerator();
-- 
1.7.4

[Cluster-devel] [PATCH 2/2] GFS2: Wait properly when flushing the ail list

[Steven Whitehouse]

The ail flush code has always relied upon log flushing to prevent
it from spinning needlessly. This fixes it to wait on the last
I/O request submitted (we don't need to wait for all of it)
instead of either spinning with io_schedule or sleeping.

As a result cpu usage of gfs2_logd is much reduced with certain
workloads.

Reported-by: Abhijith Das <adas@redhat.com>
Tested-by: Abhijith Das <adas@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/log.c b/fs/gfs2/log.c
index cec26c0..903115f 100644
--- a/fs/gfs2/log.c
+++ b/fs/gfs2/log.c
@@ -228,6 +228,27 @@ static int gfs2_ail1_empty(struct gfs2_sbd *sdp)
 	return ret;
 }
 
+static void gfs2_ail1_wait(struct gfs2_sbd *sdp)
+{
+	struct gfs2_ail *ai;
+	struct gfs2_bufdata *bd;
+	struct buffer_head *bh;
+
+	spin_lock(&sdp->sd_ail_lock);
+	list_for_each_entry_reverse(ai, &sdp->sd_ail1_list, ai_list) {
+		list_for_each_entry(bd, &ai->ai_ail1_list, bd_ail_st_list) {
+			bh = bd->bd_bh;
+			if (!buffer_locked(bh))
+				continue;
+			get_bh(bh);
+			spin_unlock(&sdp->sd_ail_lock);
+			wait_on_buffer(bh);
+			brelse(bh);
+			return;
+		}
+	}
+	spin_unlock(&sdp->sd_ail_lock);
+}
 
 /**
  * gfs2_ail2_empty_one - Check whether or not a trans in the AIL has been synced
@@ -878,9 +899,9 @@ void gfs2_meta_syncfs(struct gfs2_sbd *sdp)
 	gfs2_log_flush(sdp, NULL);
 	for (;;) {
 		gfs2_ail1_start(sdp);
+		gfs2_ail1_wait(sdp);
 		if (gfs2_ail1_empty(sdp))
 			break;
-		msleep(10);
 	}
 }
 
@@ -920,12 +941,14 @@ int gfs2_logd(void *data)
 
 		if (gfs2_ail_flush_reqd(sdp)) {
 			gfs2_ail1_start(sdp);
-			io_schedule();
+			gfs2_ail1_wait(sdp);
 			gfs2_ail1_empty(sdp);
 			gfs2_log_flush(sdp, NULL);
 		}
 
-		wake_up(&sdp->sd_log_waitq);
+		if (!gfs2_ail_flush_reqd(sdp))
+			wake_up(&sdp->sd_log_waitq);
+
 		t = gfs2_tune_get(sdp, gt_logd_secs) * HZ;
 		if (freezing(current))
 			refrigerator();
-- 
1.7.4

GFS2: Pre-pull patch posting (fixes)

[Steven Whitehouse]

Hi,

Here are a number of small fixes for GFS2. There is a fix for FIEMAP
on large sparse files, a negative dentry hashing fix, a fix for
flock, and a bug fix relating to d_splice_alias usage. There are
also (patches 1 and 5) a couple of updates which are less
critical, but small and low risk.

Steve.

GFS2: Pre-pull patch posting (fixes)

[Steven Whitehouse]

Hi,

Here are the current set of small fixes relating to GFS2.

This patch set contains two minor docs/spelling fixes, some fixes for
flock, a change to use GFP_NOFS to avoid recursion on a rarely used code
path and a fix for a race relating to the glock lru,

Steve.

GFS2: Pre-pull patch posting (fixes)

[Steven Whitehouse]

Hi,

Here is a set of small fixes for GFS2. There is a fix to drop
s_umount which is copied in from the core vfs, two patches
relate to a hard to hit "use after free" and memory leak.
Two patches related to using DIO and buffered I/O on the same
file to ensure correct operation in relation to glock state
changes. The final patch adds an RCU read lock to ensure
correct locking on an error path,

Steve.

GFS2: Pre-pull patch posting (fixes)

[Steven Whitehouse]

Hi,

Here are a couple of very small, but important, fixes,

Steve.

GFS2: Pre-pull patch posting (fixes)

[Steven Whitehouse]

Hi,

Out of these fives patches, the one for ensuring that the number of
revokes is not exceeded, and the one for checking the glock is not
already held in gfs2_getxattr are the two most important. The latter
can be triggered by selinux.

The other three patches are very small and fix mostly fairly
trivial issues,

Steve.

GFS2: Pre-pull patch posting (fixes)

[Steven Whitehouse]

There are four patches this time. The first fixes a problem where the
wrong descriptor type was being written into the log for journaled data
blocks. The second fixes a race relating to the deallocation of allocator
data. The third provides a fallback if kmalloc is unable to satisfy a
request to allocate a directory hash table. The fourth fixes the iopen
glock caching so that inodes are deleted in a more timely manner after
rmdir/unlink,

Steve.

GFS2: Pre-pull patch posting (fixes)

[Steven Whitehouse]

This time there are just four fixes. There are a couple of minor updates
to the quota code, a fix for KConfig to ensure that only valid combinations
including GFS2 can be built, and a fix for a typo affecting end i/o
processing when writing the journal.

Also, there is a temporary fix for a performance regression relating to
block reservations and directories. A longer fix will be applied in
due course, but this deals with the most immediate problem for now,

Steve.

GFS2: Pre-pull patch posting (fixes)

[Steven Whitehouse]

Hi,

Here are a number of GFS2 bug fixes. There are three from Andy Price
which fix various issues spotted by automated code analysis. There are two
from Lukas Czerner fixing my mistaken assumptions as to how FITRIM
should work. Finally Ben Marzinski has fixed a bug relating to mmap and
atime and also a bug relating to a locking issue in the transaction code.

Steve.

GFS2: Pre-pull patch posting (fixes)

[Steven Whitehouse]

Hi,

Here are three GFS2 fixes for the current kernel tree. These are all
related to the block reservation code which was added at the merge
window. That code will be getting an update at the forthcoming merge
window too. In the mean time though there are a few smaller issues
which should be fixed.

The first patch resolves an issue with write sizes of greater than
32 bits with the size hinting code. The second ensures that the
allocation data structure is initialised when using xattrs and the
third takes into account allocations which may have been made by
other nodes which affect a reservation on the local node,

Steve.

GFS2: Pre-pull patch posting (fixes)

[Steven Whitehouse]

Hi,

Here are four fixes for issues that have come up since the last
merge window,

Steve.

GFS2: Pre-pull patch posting (fixes)

[Steven Whitehouse]

Hi,

Here are four patches which provided fixes for bugs found in the
current upstream code. Please see individual patches for
descriptions,

Steve.

GFS2: Pre-pull patch posting (fixes)

[Steven Whitehouse]

Hi,

The following three fixes should help ensure that 3.0 is the best
release yet, GFS2-wise at least. I've had the first two queued for
some time waiting for the final one of this set. All three are relatively
short, although the third is a bit longer than the others, mainly
due to the extra comments added describing the inode eviction process,

Steve.

GFS2: Pre-pull patch posting (fixes)

[Steven Whitehouse]

Hi,

Here are four fixes for GFS2. See the individual patches for the detailed
descriptions,

Steve.

GFS2: Pre-pull patch posting (fixes)

[Steven Whitehouse]

Hi,

Here are a few small fixes for GFS2,

Steve.

GFS2: Pre-pull patch posting (fixes)

[Steven Whitehouse]

Hi,

These are three important, but relatively small, bug fixes for GFS2.
The first prevents a kernel BUG triggering in a relatively unlikely
(but possible) scenario when a log flush caused by glock demotion
races with a log flush from some other initiator (e.g. fsync).

The second and third patches add extra conditions to two areas
of code. This makes their behaviour match ext3 in those cases,

Steve.

GFS2: Pre-pull patch posting (fixes)

[Steven Whitehouse]

Hi,

Here are a couple of patches which between them fix a problem where
occasionally it was possible for the GFS2 module to be unloaded
before all the glocks were deallocated, which, needless to say, made
the slab allocator unhappy,

Steve.

GFS2: Pre-pull patch posting (fixes)

[Steven Whitehouse]

Here are four small fixes for GFS2. Assuming that nobody spots
any errors, I'll be sending a pull request for these shortly,

Steve.

GFS2: Pre-pull patch posting (fixes)

[Steven Whitehouse]

Hi,

Here is the current content of the GFS2 -fixes git tree. Nothing
very exciting this time... some fixes for issues we've
had relating to flushing glocks/memory usage, plus a couple of
other fixes relating to statfs and the timely removal of inodes
which have been unlinked on a remote node,

Steve.