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

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

[Steven Whitehouse]

Hi,

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

Steve.

[Cluster-devel] [PATCH 1/4] GFS2: write_end error path fails to unlock transaction lock

[Steven Whitehouse]

From: Bob Peterson <rpeterso@redhat.com>

I did an audit of gfs2's transaction glock for bugzilla bug
658619 and ran across this:

In function gfs2_write_end, in the unlikely event that
gfs2_meta_inode_buffer returns an error, the code may forget
to unlock the transaction lock because the "failed" label
appears after the call to function gfs2_trans_end.

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

diff --git a/fs/gfs2/aops.c b/fs/gfs2/aops.c
index c71995b..0f5c4f9 100644
--- a/fs/gfs2/aops.c
+++ b/fs/gfs2/aops.c
@@ -884,8 +884,8 @@ static int gfs2_write_end(struct file *file, struct address_space *mapping,
 	}
 
 	brelse(dibh);
-	gfs2_trans_end(sdp);
 failed:
+	gfs2_trans_end(sdp);
 	if (al) {
 		gfs2_inplace_release(ip);
 		gfs2_quota_unlock(ip);
-- 
1.7.4

[Cluster-devel] [PATCH 2/4] GFS2: directly write blocks past i_size

[Steven Whitehouse]

From: Benjamin Marzinski <bmarzins@redhat.com>

GFS2 was relying on the writepage code to write out the zeroed data for
fallocate.  However, with FALLOC_FL_KEEP_SIZE set, this may be past i_size.
If it is, it will be ignored.  To work around this, gfs2 now calls
write_dirty_buffer directly on the buffer_heads when FALLOC_FL_KEEP_SIZE
is set, and it's writing past i_size.

This version is just a cleanup of my last version

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

diff --git a/fs/gfs2/file.c b/fs/gfs2/file.c
index b2682e0..e483108 100644
--- a/fs/gfs2/file.c
+++ b/fs/gfs2/file.c
@@ -617,18 +617,51 @@ static ssize_t gfs2_file_aio_write(struct kiocb *iocb, const struct iovec *iov,
 	return generic_file_aio_write(iocb, iov, nr_segs, pos);
 }
 
-static void empty_write_end(struct page *page, unsigned from,
-			   unsigned to)
+static int empty_write_end(struct page *page, unsigned from,
+			   unsigned to, int mode)
 {
-	struct gfs2_inode *ip = GFS2_I(page->mapping->host);
+	struct inode *inode = page->mapping->host;
+	struct gfs2_inode *ip = GFS2_I(inode);
+	struct buffer_head *bh;
+	unsigned offset, blksize = 1 << inode->i_blkbits;
+	pgoff_t end_index = i_size_read(inode) >> PAGE_CACHE_SHIFT;
 
 	zero_user(page, from, to-from);
 	mark_page_accessed(page);
 
-	if (!gfs2_is_writeback(ip))
-		gfs2_page_add_databufs(ip, page, from, to);
+	if (page->index < end_index || !(mode & FALLOC_FL_KEEP_SIZE)) {
+		if (!gfs2_is_writeback(ip))
+			gfs2_page_add_databufs(ip, page, from, to);
+
+		block_commit_write(page, from, to);
+		return 0;
+	}
+
+	offset = 0;
+	bh = page_buffers(page);
+	while (offset < to) {
+		if (offset >= from) {
+			set_buffer_uptodate(bh);
+			mark_buffer_dirty(bh);
+			clear_buffer_new(bh);
+			write_dirty_buffer(bh, WRITE);
+		}
+		offset += blksize;
+		bh = bh->b_this_page;
+	}
 
-	block_commit_write(page, from, to);
+	offset = 0;
+	bh = page_buffers(page);
+	while (offset < to) {
+		if (offset >= from) {
+			wait_on_buffer(bh);
+			if (!buffer_uptodate(bh))
+				return -EIO;
+		}
+		offset += blksize;
+		bh = bh->b_this_page;
+	}
+	return 0;
 }
 
 static int needs_empty_write(sector_t block, struct inode *inode)
@@ -643,7 +676,8 @@ static int needs_empty_write(sector_t block, struct inode *inode)
 	return !buffer_mapped(&bh_map);
 }
 
-static int write_empty_blocks(struct page *page, unsigned from, unsigned to)
+static int write_empty_blocks(struct page *page, unsigned from, unsigned to,
+			      int mode)
 {
 	struct inode *inode = page->mapping->host;
 	unsigned start, end, next, blksize;
@@ -668,7 +702,9 @@ static int write_empty_blocks(struct page *page, unsigned from, unsigned to)
 							  gfs2_block_map);
 				if (unlikely(ret))
 					return ret;
-				empty_write_end(page, start, end);
+				ret = empty_write_end(page, start, end, mode);
+				if (unlikely(ret))
+					return ret;
 				end = 0;
 			}
 			start = next;
@@ -682,7 +718,9 @@ static int write_empty_blocks(struct page *page, unsigned from, unsigned to)
 		ret = __block_write_begin(page, start, end - start, gfs2_block_map);
 		if (unlikely(ret))
 			return ret;
-		empty_write_end(page, start, end);
+		ret = empty_write_end(page, start, end, mode);
+		if (unlikely(ret))
+			return ret;
 	}
 
 	return 0;
@@ -731,7 +769,7 @@ static int fallocate_chunk(struct inode *inode, loff_t offset, loff_t len,
 
 		if (curr == end)
 			to = end_offset;
-		error = write_empty_blocks(page, from, to);
+		error = write_empty_blocks(page, from, to, mode);
 		if (!error && offset + to > inode->i_size &&
 		    !(mode & FALLOC_FL_KEEP_SIZE)) {
 			i_size_write(inode, offset + to);
-- 
1.7.4

[Cluster-devel] [PATCH 3/4] GFS2: Don't try to deallocate unlinked inodes when mounted ro

[Steven Whitehouse]

This adds a couple of missing tests to avoid read-only nodes
from attempting to deallocate unlinked inodes.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Reported-by: Michel Andre de la Porte <madelaporte@ubi.com>

diff --git a/fs/gfs2/glops.c b/fs/gfs2/glops.c
index 3754e3c..25eeb2b 100644
--- a/fs/gfs2/glops.c
+++ b/fs/gfs2/glops.c
@@ -385,6 +385,10 @@ static int trans_go_demote_ok(const struct gfs2_glock *gl)
 static void iopen_go_callback(struct gfs2_glock *gl)
 {
 	struct gfs2_inode *ip = (struct gfs2_inode *)gl->gl_object;
+	struct gfs2_sbd *sdp = gl->gl_sbd;
+
+	if (sdp->sd_vfs->s_flags & MS_RDONLY)
+		return;
 
 	if (gl->gl_demote_state == LM_ST_UNLOCKED &&
 	    gl->gl_state == LM_ST_SHARED && ip) {
diff --git a/fs/gfs2/super.c b/fs/gfs2/super.c
index a4e23d6..5b2cb81 100644
--- a/fs/gfs2/super.c
+++ b/fs/gfs2/super.c
@@ -1318,12 +1318,13 @@ static int gfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
 
 static void gfs2_evict_inode(struct inode *inode)
 {
-	struct gfs2_sbd *sdp = inode->i_sb->s_fs_info;
+	struct super_block *sb = inode->i_sb;
+	struct gfs2_sbd *sdp = sb->s_fs_info;
 	struct gfs2_inode *ip = GFS2_I(inode);
 	struct gfs2_holder gh;
 	int error;
 
-	if (inode->i_nlink)
+	if (inode->i_nlink || (sb->s_flags & MS_RDONLY))
 		goto out;
 
 	error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
-- 
1.7.4

[Cluster-devel] [PATCH 4/4] GFS2: filesystem hang caused by incorrect lock order

[Steven Whitehouse]

From: Bob Peterson <rpeterso@redhat.com>

This patch fixes a deadlock in GFS2 where two processes are trying
to reclaim an unlinked dinode:
One holds the inode glock and calls gfs2_lookup_by_inum trying to look
up the inode, which it can't, due to I_FREEING.  The other has set
I_FREEING from vfs and is at the beginning of gfs2_delete_inode
waiting for the glock, which is held by the first.  The solution is to
add a new non_block parameter to the gfs2_iget function that causes it
to return -ENOENT if the inode is being freed.

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

diff --git a/fs/gfs2/dir.c b/fs/gfs2/dir.c
index 5c356d0..f789c57 100644
--- a/fs/gfs2/dir.c
+++ b/fs/gfs2/dir.c
@@ -1506,7 +1506,7 @@ struct inode *gfs2_dir_search(struct inode *dir, const struct qstr *name)
 		inode = gfs2_inode_lookup(dir->i_sb, 
 				be16_to_cpu(dent->de_type),
 				be64_to_cpu(dent->de_inum.no_addr),
-				be64_to_cpu(dent->de_inum.no_formal_ino));
+				be64_to_cpu(dent->de_inum.no_formal_ino), 0);
 		brelse(bh);
 		return inode;
 	}
diff --git a/fs/gfs2/inode.c b/fs/gfs2/inode.c
index 97d54a2..9134dcb 100644
--- a/fs/gfs2/inode.c
+++ b/fs/gfs2/inode.c
@@ -40,37 +40,61 @@ struct gfs2_inum_range_host {
 	u64 ir_length;
 };
 
+struct gfs2_skip_data {
+	u64 no_addr;
+	int skipped;
+	int non_block;
+};
+
 static int iget_test(struct inode *inode, void *opaque)
 {
 	struct gfs2_inode *ip = GFS2_I(inode);
-	u64 *no_addr = opaque;
+	struct gfs2_skip_data *data = opaque;
 
-	if (ip->i_no_addr == *no_addr)
+	if (ip->i_no_addr == data->no_addr) {
+		if (data->non_block &&
+		    inode->i_state & (I_FREEING|I_CLEAR|I_WILL_FREE)) {
+			data->skipped = 1;
+			return 0;
+		}
 		return 1;
-
+	}
 	return 0;
 }
 
 static int iget_set(struct inode *inode, void *opaque)
 {
 	struct gfs2_inode *ip = GFS2_I(inode);
-	u64 *no_addr = opaque;
+	struct gfs2_skip_data *data = opaque;
 
-	inode->i_ino = (unsigned long)*no_addr;
-	ip->i_no_addr = *no_addr;
+	if (data->skipped)
+		return -ENOENT;
+	inode->i_ino = (unsigned long)(data->no_addr);
+	ip->i_no_addr = data->no_addr;
 	return 0;
 }
 
 struct inode *gfs2_ilookup(struct super_block *sb, u64 no_addr)
 {
 	unsigned long hash = (unsigned long)no_addr;
-	return ilookup5(sb, hash, iget_test, &no_addr);
+	struct gfs2_skip_data data;
+
+	data.no_addr = no_addr;
+	data.skipped = 0;
+	data.non_block = 0;
+	return ilookup5(sb, hash, iget_test, &data);
 }
 
-static struct inode *gfs2_iget(struct super_block *sb, u64 no_addr)
+static struct inode *gfs2_iget(struct super_block *sb, u64 no_addr,
+			       int non_block)
 {
+	struct gfs2_skip_data data;
 	unsigned long hash = (unsigned long)no_addr;
-	return iget5_locked(sb, hash, iget_test, iget_set, &no_addr);
+
+	data.no_addr = no_addr;
+	data.skipped = 0;
+	data.non_block = non_block;
+	return iget5_locked(sb, hash, iget_test, iget_set, &data);
 }
 
 /**
@@ -111,19 +135,20 @@ static void gfs2_set_iop(struct inode *inode)
  * @sb: The super block
  * @no_addr: The inode number
  * @type: The type of the inode
+ * non_block: Can we block on inodes that are being freed?
  *
  * Returns: A VFS inode, or an error
  */
 
 struct inode *gfs2_inode_lookup(struct super_block *sb, unsigned int type,
-				u64 no_addr, u64 no_formal_ino)
+				u64 no_addr, u64 no_formal_ino, int non_block)
 {
 	struct inode *inode;
 	struct gfs2_inode *ip;
 	struct gfs2_glock *io_gl = NULL;
 	int error;
 
-	inode = gfs2_iget(sb, no_addr);
+	inode = gfs2_iget(sb, no_addr, non_block);
 	ip = GFS2_I(inode);
 
 	if (!inode)
@@ -185,11 +210,12 @@ struct inode *gfs2_lookup_by_inum(struct gfs2_sbd *sdp, u64 no_addr,
 {
 	struct super_block *sb = sdp->sd_vfs;
 	struct gfs2_holder i_gh;
-	struct inode *inode;
+	struct inode *inode = NULL;
 	int error;
 
+	/* Must not read in block until block type is verified */
 	error = gfs2_glock_nq_num(sdp, no_addr, &gfs2_inode_glops,
-				  LM_ST_SHARED, LM_FLAG_ANY, &i_gh);
+				  LM_ST_EXCLUSIVE, GL_SKIP, &i_gh);
 	if (error)
 		return ERR_PTR(error);
 
@@ -197,7 +223,7 @@ struct inode *gfs2_lookup_by_inum(struct gfs2_sbd *sdp, u64 no_addr,
 	if (error)
 		goto fail;
 
-	inode = gfs2_inode_lookup(sb, DT_UNKNOWN, no_addr, 0);
+	inode = gfs2_inode_lookup(sb, DT_UNKNOWN, no_addr, 0, 1);
 	if (IS_ERR(inode))
 		goto fail;
 
@@ -843,7 +869,7 @@ struct inode *gfs2_createi(struct gfs2_holder *ghs, const struct qstr *name,
 		goto fail_gunlock2;
 
 	inode = gfs2_inode_lookup(dir->i_sb, IF2DT(mode), inum.no_addr,
-				  inum.no_formal_ino);
+				  inum.no_formal_ino, 0);
 	if (IS_ERR(inode))
 		goto fail_gunlock2;
 
diff --git a/fs/gfs2/inode.h b/fs/gfs2/inode.h
index 3e00a66..099ca30 100644
--- a/fs/gfs2/inode.h
+++ b/fs/gfs2/inode.h
@@ -97,7 +97,8 @@ err:
 }
 
 extern struct inode *gfs2_inode_lookup(struct super_block *sb, unsigned type, 
-				       u64 no_addr, u64 no_formal_ino);
+				       u64 no_addr, u64 no_formal_ino,
+				       int non_block);
 extern struct inode *gfs2_lookup_by_inum(struct gfs2_sbd *sdp, u64 no_addr,
 					 u64 *no_formal_ino,
 					 unsigned int blktype);
diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c
index 42ef243..d3c69eb 100644
--- a/fs/gfs2/ops_fstype.c
+++ b/fs/gfs2/ops_fstype.c
@@ -430,7 +430,7 @@ static int gfs2_lookup_root(struct super_block *sb, struct dentry **dptr,
 	struct dentry *dentry;
 	struct inode *inode;
 
-	inode = gfs2_inode_lookup(sb, DT_DIR, no_addr, 0);
+	inode = gfs2_inode_lookup(sb, DT_DIR, no_addr, 0, 0);
 	if (IS_ERR(inode)) {
 		fs_err(sdp, "can't read in %s inode: %ld\n", name, PTR_ERR(inode));
 		return PTR_ERR(inode);
diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c
index cf930cd..6fcae84 100644
--- a/fs/gfs2/rgrp.c
+++ b/fs/gfs2/rgrp.c
@@ -945,7 +945,7 @@ static void try_rgrp_unlink(struct gfs2_rgrpd *rgd, u64 *last_unlinked, u64 skip
 		/* rgblk_search can return a block < goal, so we need to
 		   keep it marching forward. */
 		no_addr = block + rgd->rd_data0;
-		goal++;
+		goal = max(block + 1, goal + 1);
 		if (*last_unlinked != NO_BLOCK && no_addr <= *last_unlinked)
 			continue;
 		if (no_addr == skip)
@@ -971,7 +971,7 @@ static void try_rgrp_unlink(struct gfs2_rgrpd *rgd, u64 *last_unlinked, u64 skip
 			found++;
 
 		/* Limit reclaim to sensible number of tasks */
-		if (found > 2*NR_CPUS)
+		if (found > NR_CPUS)
 			return;
 	}
 
diff --git a/fs/gfs2/super.c b/fs/gfs2/super.c
index 5b2cb81..b9f28e6 100644
--- a/fs/gfs2/super.c
+++ b/fs/gfs2/super.c
@@ -1327,7 +1327,8 @@ static void gfs2_evict_inode(struct inode *inode)
 	if (inode->i_nlink || (sb->s_flags & MS_RDONLY))
 		goto out;
 
-	error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
+	/* Must not read inode block until block type has been verified */
+	error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, GL_SKIP, &gh);
 	if (unlikely(error)) {
 		gfs2_glock_dq_uninit(&ip->i_iopen_gh);
 		goto out;
@@ -1337,6 +1338,12 @@ static void gfs2_evict_inode(struct inode *inode)
 	if (error)
 		goto out_truncate;
 
+	if (test_bit(GIF_INVALID, &ip->i_flags)) {
+		error = gfs2_inode_refresh(ip);
+		if (error)
+			goto out_truncate;
+	}
+
 	ip->i_iopen_gh.gh_flags |= GL_NOCACHE;
 	gfs2_glock_dq_wait(&ip->i_iopen_gh);
 	gfs2_holder_reinit(LM_ST_EXCLUSIVE, LM_FLAG_TRY_1CB | GL_NOCACHE, &ip->i_iopen_gh);
-- 
1.7.4