[Cluster-devel] GFS2 Pre-pull patch posting (merge window)

[Cluster-devel] GFS2 Pre-pull patch posting (merge window)

[Steven Whitehouse]

Hi,

This is the smallest merge window patch set for GFS2 for quite
some time. Only one of the patches (moving gfs2_sync_meta) is
a non-bug fix patch, although the merge ordered and writeback
writepage patch is also a nice clean up.

A couple of the patches are quite recently added, due to my only
having recently returned from holiday, so I'll give them a couple
of extra days in -next before sending the pull request.

Steve.

[Cluster-devel] [PATCH 1/6] GFS2: Move gfs2_sync_meta to lops.c

[Steven Whitehouse]

Since gfs2_sync_meta() is only called from a single file, lets move
it to lops.c where it is used, and mark it static. At the same
time, we can clean up the meta_io.h header too.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/lops.c b/fs/gfs2/lops.c
index 17c5b5d..010b9fb 100644
--- a/fs/gfs2/lops.c
+++ b/fs/gfs2/lops.c
@@ -579,6 +579,24 @@ static int buf_lo_scan_elements(struct gfs2_jdesc *jd, unsigned int start,
 	return error;
 }
 
+/**
+ * gfs2_meta_sync - Sync all buffers associated with a glock
+ * @gl: The glock
+ *
+ */
+
+static void gfs2_meta_sync(struct gfs2_glock *gl)
+{
+	struct address_space *mapping = gfs2_glock2aspace(gl);
+	int error;
+
+	filemap_fdatawrite(mapping);
+	error = filemap_fdatawait(mapping);
+
+	if (error)
+		gfs2_io_error(gl->gl_sbd);
+}
+
 static void buf_lo_after_scan(struct gfs2_jdesc *jd, int error, int pass)
 {
 	struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
diff --git a/fs/gfs2/meta_io.c b/fs/gfs2/meta_io.c
index 0da3906..9324150 100644
--- a/fs/gfs2/meta_io.c
+++ b/fs/gfs2/meta_io.c
@@ -98,24 +98,6 @@ const struct address_space_operations gfs2_meta_aops = {
 };
 
 /**
- * gfs2_meta_sync - Sync all buffers associated with a glock
- * @gl: The glock
- *
- */
-
-void gfs2_meta_sync(struct gfs2_glock *gl)
-{
-	struct address_space *mapping = gfs2_glock2aspace(gl);
-	int error;
-
-	filemap_fdatawrite(mapping);
-	error = filemap_fdatawait(mapping);
-
-	if (error)
-		gfs2_io_error(gl->gl_sbd);
-}
-
-/**
  * gfs2_getbuf - Get a buffer with a given address space
  * @gl: the glock
  * @blkno: the block number (filesystem scope)
diff --git a/fs/gfs2/meta_io.h b/fs/gfs2/meta_io.h
index 0d4c843..4823b93 100644
--- a/fs/gfs2/meta_io.h
+++ b/fs/gfs2/meta_io.h
@@ -48,21 +48,17 @@ static inline struct gfs2_sbd *gfs2_mapping2sbd(struct address_space *mapping)
 		return inode->i_sb->s_fs_info;
 }
 
-void gfs2_meta_sync(struct gfs2_glock *gl);
-
-struct buffer_head *gfs2_meta_new(struct gfs2_glock *gl, u64 blkno);
-int gfs2_meta_read(struct gfs2_glock *gl, u64 blkno,
-		   int flags, struct buffer_head **bhp);
-int gfs2_meta_wait(struct gfs2_sbd *sdp, struct buffer_head *bh);
-struct buffer_head *gfs2_getbuf(struct gfs2_glock *gl, u64 blkno, int create);
-
-void gfs2_remove_from_journal(struct buffer_head *bh, struct gfs2_trans *tr,
-			      int meta);
-
-void gfs2_meta_wipe(struct gfs2_inode *ip, u64 bstart, u32 blen);
-
-int gfs2_meta_indirect_buffer(struct gfs2_inode *ip, int height, u64 num,
-			      struct buffer_head **bhp);
+extern struct buffer_head *gfs2_meta_new(struct gfs2_glock *gl, u64 blkno);
+extern int gfs2_meta_read(struct gfs2_glock *gl, u64 blkno, int flags,
+			  struct buffer_head **bhp);
+extern int gfs2_meta_wait(struct gfs2_sbd *sdp, struct buffer_head *bh);
+extern struct buffer_head *gfs2_getbuf(struct gfs2_glock *gl, u64 blkno,
+				       int create);
+extern void gfs2_remove_from_journal(struct buffer_head *bh,
+				     struct gfs2_trans *tr, int meta);
+extern void gfs2_meta_wipe(struct gfs2_inode *ip, u64 bstart, u32 blen);
+extern int gfs2_meta_indirect_buffer(struct gfs2_inode *ip, int height, u64 num,
+				     struct buffer_head **bhp);
 
 static inline int gfs2_meta_inode_buffer(struct gfs2_inode *ip,
 					 struct buffer_head **bhp)
-- 
1.7.4

[Cluster-devel] [PATCH 2/6] GFS2: Take glock reference in examine_bucket()

[Steven Whitehouse]

We need to check the glock ref counter in a race free way
in order to ensure that the gfs2_glock_hold() call will
succeed. The easiest way to do that is to simply take the
reference count early in the common code of examine_bucket,
skipping any glocks with zero ref count.

That means that the examiner functions all need to put their
reference on the glock once they've performed their function.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Reported-by: David Teigland <teigland@redhat.com>
Tested-by: David Teigland <teigland@redhat.com>

diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c
index 544a809..ce7078d 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -1488,7 +1488,7 @@ static void examine_bucket(glock_examiner examiner, const struct gfs2_sbd *sdp,
 
 	rcu_read_lock();
 	hlist_bl_for_each_entry_rcu(gl, pos, head, gl_list) {
-		if ((gl->gl_sbd == sdp) && atomic_read(&gl->gl_ref))
+		if ((gl->gl_sbd == sdp) && atomic_inc_not_zero(&gl->gl_ref))
 			examiner(gl);
 	}
 	rcu_read_unlock();
@@ -1508,18 +1508,17 @@ static void glock_hash_walk(glock_examiner examiner, const struct gfs2_sbd *sdp)
  * thaw_glock - thaw out a glock which has an unprocessed reply waiting
  * @gl: The glock to thaw
  *
- * N.B. When we freeze a glock, we leave a ref to the glock outstanding,
- * so this has to result in the ref count being dropped by one.
  */
 
 static void thaw_glock(struct gfs2_glock *gl)
 {
 	if (!test_and_clear_bit(GLF_FROZEN, &gl->gl_flags))
-		return;
+		goto out;
 	set_bit(GLF_REPLY_PENDING, &gl->gl_flags);
-	gfs2_glock_hold(gl);
-	if (queue_delayed_work(glock_workqueue, &gl->gl_work, 0) == 0)
+	if (queue_delayed_work(glock_workqueue, &gl->gl_work, 0) == 0) {
+out:
 		gfs2_glock_put(gl);
+	}
 }
 
 /**
@@ -1536,7 +1535,6 @@ static void clear_glock(struct gfs2_glock *gl)
 	if (gl->gl_state != LM_ST_UNLOCKED)
 		handle_callback(gl, LM_ST_UNLOCKED, 0, false);
 	spin_unlock(&gl->gl_spin);
-	gfs2_glock_hold(gl);
 	if (queue_delayed_work(glock_workqueue, &gl->gl_work, 0) == 0)
 		gfs2_glock_put(gl);
 }
-- 
1.7.4

[Cluster-devel] [PATCH 3/6] GFS2: Merge ordered and writeback writepage

[Steven Whitehouse]

The writepages function was recently merged between writeback
and ordered mode. This completes the change by doing the same
with writepage. The remaining differences in writepage were
left over from some earlier time and not actually doing anything
useful.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/aops.c b/fs/gfs2/aops.c
index ee48ad3..a9ea6f0 100644
--- a/fs/gfs2/aops.c
+++ b/fs/gfs2/aops.c
@@ -122,14 +122,13 @@ out:
 }
 
 /**
- * gfs2_writeback_writepage - Write page for writeback mappings
+ * gfs2_writepage - Write page for writeback mappings
  * @page: The page
  * @wbc: The writeback control
  *
  */
 
-static int gfs2_writeback_writepage(struct page *page,
-				    struct writeback_control *wbc)
+static int gfs2_writepage(struct page *page, struct writeback_control *wbc)
 {
 	int ret;
 
@@ -141,32 +140,6 @@ static int gfs2_writeback_writepage(struct page *page,
 }
 
 /**
- * gfs2_ordered_writepage - Write page for ordered data files
- * @page: The page to write
- * @wbc: The writeback control
- *
- */
-
-static int gfs2_ordered_writepage(struct page *page,
-				  struct writeback_control *wbc)
-{
-	struct inode *inode = page->mapping->host;
-	struct gfs2_inode *ip = GFS2_I(inode);
-	int ret;
-
-	ret = gfs2_writepage_common(page, wbc);
-	if (ret <= 0)
-		return ret;
-
-	if (!page_has_buffers(page)) {
-		create_empty_buffers(page, inode->i_sb->s_blocksize,
-				     (1 << BH_Dirty)|(1 << BH_Uptodate));
-	}
-	gfs2_page_add_databufs(ip, page, 0, inode->i_sb->s_blocksize-1);
-	return block_write_full_page(page, gfs2_get_block_noalloc, wbc);
-}
-
-/**
  * __gfs2_jdata_writepage - The core of jdata writepage
  * @page: The page to write
  * @wbc: The writeback control
@@ -1107,7 +1080,7 @@ cannot_release:
 }
 
 static const struct address_space_operations gfs2_writeback_aops = {
-	.writepage = gfs2_writeback_writepage,
+	.writepage = gfs2_writepage,
 	.writepages = gfs2_writepages,
 	.readpage = gfs2_readpage,
 	.readpages = gfs2_readpages,
@@ -1123,7 +1096,7 @@ static const struct address_space_operations gfs2_writeback_aops = {
 };
 
 static const struct address_space_operations gfs2_ordered_aops = {
-	.writepage = gfs2_ordered_writepage,
+	.writepage = gfs2_writepage,
 	.writepages = gfs2_writepages,
 	.readpage = gfs2_readpage,
 	.readpages = gfs2_readpages,
-- 
1.7.4

[Cluster-devel] [PATCH 4/6] GFS2: Remove unnecessary memory barrier

[Steven Whitehouse]

From: Bob Peterson <rpeterso@redhat.com>

Function test_and_clear_bit implies a memory barrier, so subsequent
memory barriers are unnecessary.

Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c
index ce7078d..722329c 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -1411,7 +1411,6 @@ __acquires(&lru_lock)
 		if (demote_ok(gl))
 			handle_callback(gl, LM_ST_UNLOCKED, 0, false);
 		WARN_ON(!test_and_clear_bit(GLF_LOCK, &gl->gl_flags));
-		smp_mb__after_clear_bit();
 		if (queue_delayed_work(glock_workqueue, &gl->gl_work, 0) == 0)
 			gfs2_glock_put_nolock(gl);
 		spin_unlock(&gl->gl_spin);
-- 
1.7.4

[Cluster-devel] [PATCH 5/6] GFS2: Don't flag consistency error if first mounter is a spectator

[Steven Whitehouse]

From: Bob Peterson <rpeterso@redhat.com>

This patch checks for the first mounter being a specator. If so, it
makes sure all the journals are clean. If there's a dirty journal,
the mount fails.

Testing results:

# insmod gfs2.ko
# mount -tgfs2 -o spectator /dev/sasdrives/scratch /mnt/gfs2
mount: permission denied
# dmesg | tail -2
[ 3390.655996] GFS2: fsid=MUSKETEER:home: Now mounting FS...
[ 3390.841336] GFS2: fsid=MUSKETEER:home.s: jid=0: Journal is dirty, so the first mounter must not be a spectator.
# mount -tgfs2 /dev/sasdrives/scratch /mnt/gfs2
# umount /mnt/gfs2
# mount -tgfs2 -o spectator /dev/sasdrives/scratch /mnt/gfs2
# ls /mnt/gfs2|wc -l
352
# umount /mnt/gfs2

Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c
index 0262c19..19ff5e8 100644
--- a/fs/gfs2/ops_fstype.c
+++ b/fs/gfs2/ops_fstype.c
@@ -646,6 +646,48 @@ static int gfs2_jindex_hold(struct gfs2_sbd *sdp, struct gfs2_holder *ji_gh)
 	return error;
 }
 
+/**
+ * check_journal_clean - Make sure a journal is clean for a spectator mount
+ * @sdp: The GFS2 superblock
+ * @jd: The journal descriptor
+ *
+ * Returns: 0 if the journal is clean or locked, else an error
+ */
+static int check_journal_clean(struct gfs2_sbd *sdp, struct gfs2_jdesc *jd)
+{
+	int error;
+	struct gfs2_holder j_gh;
+	struct gfs2_log_header_host head;
+	struct gfs2_inode *ip;
+
+	ip = GFS2_I(jd->jd_inode);
+	error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_NOEXP |
+				   GL_EXACT | GL_NOCACHE, &j_gh);
+	if (error) {
+		fs_err(sdp, "Error locking journal for spectator mount.\n");
+		return -EPERM;
+	}
+	error = gfs2_jdesc_check(jd);
+	if (error) {
+		fs_err(sdp, "Error checking journal for spectator mount.\n");
+		goto out_unlock;
+	}
+	error = gfs2_find_jhead(jd, &head);
+	if (error) {
+		fs_err(sdp, "Error parsing journal for spectator mount.\n");
+		goto out_unlock;
+	}
+	if (!(head.lh_flags & GFS2_LOG_HEAD_UNMOUNT)) {
+		error = -EPERM;
+		fs_err(sdp, "jid=%u: Journal is dirty, so the first mounter "
+		       "must not be a spectator.\n", jd->jd_jid);
+	}
+
+out_unlock:
+	gfs2_glock_dq_uninit(&j_gh);
+	return error;
+}
+
 static int init_journal(struct gfs2_sbd *sdp, int undo)
 {
 	struct inode *master = sdp->sd_master_dir->d_inode;
@@ -732,8 +774,15 @@ static int init_journal(struct gfs2_sbd *sdp, int undo)
 	if (sdp->sd_lockstruct.ls_first) {
 		unsigned int x;
 		for (x = 0; x < sdp->sd_journals; x++) {
-			error = gfs2_recover_journal(gfs2_jdesc_find(sdp, x),
-						     true);
+			struct gfs2_jdesc *jd = gfs2_jdesc_find(sdp, x);
+
+			if (sdp->sd_args.ar_spectator) {
+				error = check_journal_clean(sdp, jd);
+				if (error)
+					goto fail_jinode_gh;
+				continue;
+			}
+			error = gfs2_recover_journal(jd, true);
 			if (error) {
 				fs_err(sdp, "error recovering journal %u: %d\n",
 				       x, error);
-- 
1.7.4

[Cluster-devel] [PATCH 6/6] GFS2: dirty inode correctly in gfs2_write_end

[Steven Whitehouse]

From: Benjamin Marzinski <bmarzins@redhat.com>

GFS2 was only setting I_DIRTY_DATASYNC on files that it wrote to, when
it actually increased the file size.  If gfs2_fsync was called without
I_DIRTY_DATASYNC set, it didn't flush the incore data to the log before
returning, so any metadata or journaled data changes were not getting
fsynced. This meant that writes to the middle of files were not always
getting fsynced properly.

This patch makes gfs2 set I_DIRTY_DATASYNC whenever metadata has been
updated during a write. It also make gfs2_sync flush the incore log
if I_DIRTY_PAGES is set, and the file is using data journalling. This
will make sure that all incore logged data gets written to disk before
returning from a fsync.

Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/aops.c b/fs/gfs2/aops.c
index a9ea6f0..1f7d805 100644
--- a/fs/gfs2/aops.c
+++ b/fs/gfs2/aops.c
@@ -815,6 +815,8 @@ static int gfs2_write_end(struct file *file, struct address_space *mapping,
 	unsigned int from = pos & (PAGE_CACHE_SIZE - 1);
 	unsigned int to = from + len;
 	int ret;
+	struct gfs2_trans *tr = current->journal_info;
+	BUG_ON(!tr);
 
 	BUG_ON(gfs2_glock_is_locked_by_me(ip->i_gl) == NULL);
 
@@ -825,8 +827,6 @@ static int gfs2_write_end(struct file *file, struct address_space *mapping,
 		goto failed;
 	}
 
-	gfs2_trans_add_meta(ip->i_gl, dibh);
-
 	if (gfs2_is_stuffed(ip))
 		return gfs2_stuffed_write_end(inode, dibh, pos, len, copied, page);
 
@@ -834,6 +834,11 @@ static int gfs2_write_end(struct file *file, struct address_space *mapping,
 		gfs2_page_add_databufs(ip, page, from, to);
 
 	ret = generic_write_end(file, mapping, pos, len, copied, page, fsdata);
+	if (tr->tr_num_buf_new)
+		__mark_inode_dirty(inode, I_DIRTY_DATASYNC);
+	else
+		gfs2_trans_add_meta(ip->i_gl, dibh);
+
 
 	if (inode == sdp->sd_rindex) {
 		adjust_fs_space(inode);
diff --git a/fs/gfs2/file.c b/fs/gfs2/file.c
index 72c3866..0621b46 100644
--- a/fs/gfs2/file.c
+++ b/fs/gfs2/file.c
@@ -650,7 +650,7 @@ static int gfs2_fsync(struct file *file, loff_t start, loff_t end,
 {
 	struct address_space *mapping = file->f_mapping;
 	struct inode *inode = mapping->host;
-	int sync_state = inode->i_state & (I_DIRTY_SYNC|I_DIRTY_DATASYNC);
+	int sync_state = inode->i_state & I_DIRTY;
 	struct gfs2_inode *ip = GFS2_I(inode);
 	int ret = 0, ret1 = 0;
 
@@ -660,6 +660,8 @@ static int gfs2_fsync(struct file *file, loff_t start, loff_t end,
 			return ret1;
 	}
 
+	if (!gfs2_is_jdata(ip))
+		sync_state &= ~I_DIRTY_PAGES;
 	if (datasync)
 		sync_state &= ~I_DIRTY_SYNC;
 
-- 
1.7.4