[Cluster-devel] [GFS2] Pre-pull patch posting

[Cluster-devel] [GFS2] Pre-pull patch posting

[swhiteho]

This is the current content of the GFS2 -nmw git tree. Mostly bug
fixes, there are some changes relating to block mapping which are
working towards cleaning up this code and allowing more efficient
block mapping. There is a second part to that work which is not
included in this patch set - the plan is that it will be in the
next patch set and its currently undergoing testing.

There are a number of clean up patches in the series too. We have
been continuing the work of gradually reducing the fields in the
various ..._host structures with a view to eventually eliminating
them completely. They were introduced as a stop-gap measure to fix
the endianess annotation and the fields are now gradually being moved
to other structures (or eliminated).

Bob Peterson's new improved "bitfit" algorithm provides a nice
speed up when we are allocating blocks as well as cleaning up
that area of the code.

Steve.

[Cluster-devel] [PATCH 01/48] [GFS2] Speed up gfs2_write_alloc_required, deprecate gfs2_extent_map

[swhiteho]

From: Steven Whitehouse <swhiteho@redhat.com>

This patch removes the call to gfs2_extent_map from gfs2_write_alloc_required,
instead we call gfs2_block_map directly. This results in fewer overall calls
to gfs2_block_map in the multi-block case.

Also, gfs2_extent_map is marked as deprecated so that people know that its
going away as soon as all the callers have been converted.

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

diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index e9456eb..a25444a 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -548,6 +548,9 @@ out_fail:
 	return error;
 }
 
+/*
+ * Deprecated: do not use in new code
+ */
 int gfs2_extent_map(struct inode *inode, u64 lblock, int *new, u64 *dblock, unsigned *extlen)
 {
 	struct buffer_head bh = { .b_state = 0, .b_blocknr = 0 };
@@ -1197,10 +1200,9 @@ int gfs2_write_alloc_required(struct gfs2_inode *ip, u64 offset,
 			      unsigned int len, int *alloc_required)
 {
 	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
-	u64 lblock, lblock_stop, dblock;
-	u32 extlen;
-	int new = 0;
-	int error = 0;
+	struct buffer_head bh;
+	unsigned int shift;
+	u64 lblock, lblock_stop, size;
 
 	*alloc_required = 0;
 
@@ -1214,6 +1216,8 @@ int gfs2_write_alloc_required(struct gfs2_inode *ip, u64 offset,
 		return 0;
 	}
 
+	*alloc_required = 1;
+	shift = sdp->sd_sb.sb_bsize_shift;
 	if (gfs2_is_dir(ip)) {
 		unsigned int bsize = sdp->sd_jbsize;
 		lblock = offset;
@@ -1221,27 +1225,25 @@ int gfs2_write_alloc_required(struct gfs2_inode *ip, u64 offset,
 		lblock_stop = offset + len + bsize - 1;
 		do_div(lblock_stop, bsize);
 	} else {
-		unsigned int shift = sdp->sd_sb.sb_bsize_shift;
 		u64 end_of_file = (ip->i_di.di_size + sdp->sd_sb.sb_bsize - 1) >> shift;
 		lblock = offset >> shift;
 		lblock_stop = (offset + len + sdp->sd_sb.sb_bsize - 1) >> shift;
-		if (lblock_stop > end_of_file) {
-			*alloc_required = 1;
+		if (lblock_stop > end_of_file)
 			return 0;
-		}
 	}
 
-	for (; lblock < lblock_stop; lblock += extlen) {
-		error = gfs2_extent_map(&ip->i_inode, lblock, &new, &dblock, &extlen);
-		if (error)
-			return error;
-
-		if (!dblock) {
-			*alloc_required = 1;
+	size = (lblock_stop - lblock) << shift;
+	do {
+		bh.b_state = 0;
+		bh.b_size = size;
+		gfs2_block_map(&ip->i_inode, lblock, &bh, 0);
+		if (!buffer_mapped(&bh))
 			return 0;
-		}
-	}
+		size -= bh.b_size;
+		lblock += (bh.b_size >> ip->i_inode.i_blkbits);
+	} while(size > 0);
 
+	*alloc_required = 0;
 	return 0;
 }
 
-- 
1.5.1.2

[Cluster-devel] [PATCH 02/48] [GFS2] Streamline indirect pointer tree height calculation

[swhiteho]

From: Steven Whitehouse <swhiteho@redhat.com>

This patch improves the calculation of the tree height in order to reduce
the number of operations which are carried out on each call to gfs2_block_map.
In the common case, we now make a single comparison, rather than calculating
the required tree height from scratch each time. Also in the case that the
tree does need some extra height, we start from the current height rather from
zero when we work out what the new height ought to be.

In addition the di_height field is moved into the inode proper and reduced
in size to a u8 since the value must be between 0 and GFS2_MAX_META_HEIGHT (10).

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

diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index a25444a..5a31870 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -166,7 +166,7 @@ int gfs2_unstuff_dinode(struct gfs2_inode *ip, struct page *page)
 		di->di_blocks = cpu_to_be64(ip->i_di.di_blocks);
 	}
 
-	ip->i_di.di_height = 1;
+	ip->i_height = 1;
 	di->di_height = cpu_to_be16(1);
 
 out_brelse:
@@ -177,43 +177,6 @@ out:
 }
 
 /**
- * calc_tree_height - Calculate the height of a metadata tree
- * @ip: The GFS2 inode
- * @size: The proposed size of the file
- *
- * Work out how tall a metadata tree needs to be in order to accommodate a
- * file of a particular size. If size is less than the current size of
- * the inode, then the current size of the inode is used instead of the
- * supplied one.
- *
- * Returns: the height the tree should be
- */
-
-static unsigned int calc_tree_height(struct gfs2_inode *ip, u64 size)
-{
-	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
-	u64 *arr;
-	unsigned int max, height;
-
-	if (ip->i_di.di_size > size)
-		size = ip->i_di.di_size;
-
-	if (gfs2_is_dir(ip)) {
-		arr = sdp->sd_jheightsize;
-		max = sdp->sd_max_jheight;
-	} else {
-		arr = sdp->sd_heightsize;
-		max = sdp->sd_max_height;
-	}
-
-	for (height = 0; height < max; height++)
-		if (arr[height] >= size)
-			break;
-
-	return height;
-}
-
-/**
  * build_height - Build a metadata tree of the requested height
  * @ip: The GFS2 inode
  * @height: The height to build to
@@ -225,7 +188,7 @@ static unsigned int calc_tree_height(struct gfs2_inode *ip, u64 size)
 static int build_height(struct inode *inode, unsigned height)
 {
 	struct gfs2_inode *ip = GFS2_I(inode);
-	unsigned new_height = height - ip->i_di.di_height;
+	unsigned new_height = height - ip->i_height;
 	struct buffer_head *dibh;
 	struct buffer_head *blocks[GFS2_MAX_META_HEIGHT];
 	struct gfs2_dinode *di;
@@ -234,7 +197,7 @@ static int build_height(struct inode *inode, unsigned height)
 	u64 bn;
 	unsigned n;
 
-	if (height <= ip->i_di.di_height)
+	if (height <= ip->i_height)
 		return 0;
 
 	error = gfs2_meta_inode_buffer(ip, &dibh);
@@ -270,10 +233,10 @@ static int build_height(struct inode *inode, unsigned height)
 	di = (struct gfs2_dinode *)dibh->b_data;
 	gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode));
 	*(__be64 *)(di + 1) = cpu_to_be64(bn);
-	ip->i_di.di_height += new_height;
+	ip->i_height += new_height;
 	ip->i_di.di_blocks += new_height;
 	gfs2_set_inode_blocks(&ip->i_inode);
-	di->di_height = cpu_to_be16(ip->i_di.di_height);
+	di->di_height = cpu_to_be16(ip->i_height);
 	di->di_blocks = cpu_to_be64(ip->i_di.di_blocks);
 	brelse(dibh);
 	return error;
@@ -345,7 +308,7 @@ static void find_metapath(struct gfs2_inode *ip, u64 block,
 	u64 b = block;
 	unsigned int i;
 
-	for (i = ip->i_di.di_height; i--;)
+	for (i = ip->i_height; i--;)
 		mp->mp_list[i] = do_div(b, sdp->sd_inptrs);
 
 }
@@ -407,7 +370,7 @@ static int lookup_block(struct gfs2_inode *ip, struct buffer_head *bh,
 	if (!create)
 		return 0;
 
-	if (height == ip->i_di.di_height - 1 && !gfs2_is_dir(ip))
+	if (height == ip->i_height - 1 && !gfs2_is_dir(ip))
 		*block = gfs2_alloc_data(ip);
 	else
 		*block = gfs2_alloc_meta(ip);
@@ -458,8 +421,7 @@ int gfs2_block_map(struct inode *inode, sector_t lblock,
 	struct gfs2_inode *ip = GFS2_I(inode);
 	struct gfs2_sbd *sdp = GFS2_SB(inode);
 	struct buffer_head *bh;
-	unsigned int bsize;
-	unsigned int height;
+	unsigned int bsize = sdp->sd_sb.sb_bsize;
 	unsigned int end_of_metadata;
 	unsigned int x;
 	int error = 0;
@@ -470,7 +432,7 @@ int gfs2_block_map(struct inode *inode, sector_t lblock,
 	struct metapath mp;
 	u64 size;
 	struct buffer_head *dibh = NULL;
-
+	const u64 *arr = sdp->sd_heightsize;
 	BUG_ON(maxlen == 0);
 
 	if (gfs2_assert_warn(sdp, !gfs2_is_stuffed(ip)))
@@ -480,23 +442,25 @@ int gfs2_block_map(struct inode *inode, sector_t lblock,
 	clear_buffer_mapped(bh_map);
 	clear_buffer_new(bh_map);
 	clear_buffer_boundary(bh_map);
-	bsize = gfs2_is_dir(ip) ? sdp->sd_jbsize : sdp->sd_sb.sb_bsize;
+	if (gfs2_is_dir(ip)) {
+		bsize = sdp->sd_jbsize;
+		arr = sdp->sd_jheightsize;
+	}
 	size = (lblock + 1) * bsize;
 
-	if (size > ip->i_di.di_size) {
-		height = calc_tree_height(ip, size);
-		if (ip->i_di.di_height < height) {
-			if (!create)
-				goto out_ok;
-	
-			error = build_height(inode, height);
-			if (error)
-				goto out_fail;
-		}
+	if (size > arr[ip->i_height]) {
+		u8 height = ip->i_height;
+		if (!create)
+			goto out_ok;
+		while (size > arr[height])
+			height++;
+		error = build_height(inode, height);
+		if (error)
+			goto out_fail;
 	}
 
 	find_metapath(ip, lblock, &mp);
-	end_of_metadata = ip->i_di.di_height - 1;
+	end_of_metadata = ip->i_height - 1;
 	error = gfs2_meta_inode_buffer(ip, &bh);
 	if (error)
 		goto out_fail;
@@ -624,7 +588,7 @@ static int recursive_scan(struct gfs2_inode *ip, struct buffer_head *dibh,
 	if (error)
 		goto out;
 
-	if (height < ip->i_di.di_height - 1)
+	if (height < ip->i_height - 1)
 		for (; top < bottom; top++, first = 0) {
 			if (!*top)
 				continue;
@@ -682,7 +646,7 @@ static int do_strip(struct gfs2_inode *ip, struct buffer_head *dibh,
 		sm->sm_first = 0;
 	}
 
-	metadata = (height != ip->i_di.di_height - 1);
+	metadata = (height != ip->i_height - 1);
 	if (metadata)
 		revokes = (height) ? sdp->sd_inptrs : sdp->sd_diptrs;
 
@@ -807,7 +771,6 @@ static int do_grow(struct gfs2_inode *ip, u64 size)
 	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
 	struct gfs2_alloc *al;
 	struct buffer_head *dibh;
-	unsigned int h;
 	int error;
 
 	al = gfs2_alloc_get(ip);
@@ -833,20 +796,23 @@ static int do_grow(struct gfs2_inode *ip, u64 size)
 		goto out_ipres;
 
 	if (size > sdp->sd_sb.sb_bsize - sizeof(struct gfs2_dinode)) {
+		const u64 *arr = sdp->sd_heightsize;
 		if (gfs2_is_stuffed(ip)) {
 			error = gfs2_unstuff_dinode(ip, NULL);
 			if (error)
 				goto out_end_trans;
 		}
 
-		h = calc_tree_height(ip, size);
-		if (ip->i_di.di_height < h) {
-			down_write(&ip->i_rw_mutex);
-			error = build_height(&ip->i_inode, h);
-			up_write(&ip->i_rw_mutex);
-			if (error)
-				goto out_end_trans;
+		down_write(&ip->i_rw_mutex);
+		if (size > arr[ip->i_height]) {
+			u8 height = ip->i_height;
+			while(size > arr[height])
+				height++;
+			error = build_height(&ip->i_inode, height);
 		}
+		up_write(&ip->i_rw_mutex);
+		if (error)
+			goto out_end_trans;
 	}
 
 	ip->i_di.di_size = size;
@@ -989,7 +955,7 @@ out:
 
 static int trunc_dealloc(struct gfs2_inode *ip, u64 size)
 {
-	unsigned int height = ip->i_di.di_height;
+	unsigned int height = ip->i_height;
 	u64 lblock;
 	struct metapath mp;
 	int error;
@@ -1040,7 +1006,7 @@ static int trunc_end(struct gfs2_inode *ip)
 		goto out;
 
 	if (!ip->i_di.di_size) {
-		ip->i_di.di_height = 0;
+		ip->i_height = 0;
 		ip->i_di.di_goal_meta =
 			ip->i_di.di_goal_data =
 			ip->i_no_addr;
diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h
index 525dcae..43472ba 100644
--- a/fs/gfs2/incore.h
+++ b/fs/gfs2/incore.h
@@ -246,7 +246,6 @@ struct gfs2_dinode_host {
 	u64 di_goal_data;	/* data block goal */
 	u64 di_generation;	/* generation number for NFS */
 	u32 di_flags;		/* GFS2_DIF_... */
-	u16 di_height;		/* height of metadata */
 	/* These only apply to directories  */
 	u16 di_depth;		/* Number of bits in the table */
 	u32 di_entries;		/* The number of entries in the directory */
@@ -268,6 +267,7 @@ struct gfs2_inode {
 	u64 i_last_rg_alloc;
 
 	struct rw_semaphore i_rw_mutex;
+	u8 i_height;
 };
 
 /*
@@ -490,9 +490,9 @@ struct gfs2_sbd {
 	u32 sd_qc_per_block;
 	u32 sd_max_dirres;	/* Max blocks needed to add a directory entry */
 	u32 sd_max_height;	/* Max height of a file's metadata tree */
-	u64 sd_heightsize[GFS2_MAX_META_HEIGHT];
+	u64 sd_heightsize[GFS2_MAX_META_HEIGHT + 1];
 	u32 sd_max_jheight; /* Max height of journaled file's meta tree */
-	u64 sd_jheightsize[GFS2_MAX_META_HEIGHT];
+	u64 sd_jheightsize[GFS2_MAX_META_HEIGHT + 1];
 
 	struct gfs2_args sd_args;	/* Mount arguments */
 	struct gfs2_tune sd_tune;	/* Filesystem tuning structure */
diff --git a/fs/gfs2/inode.c b/fs/gfs2/inode.c
index 37725ad..ff66ab7 100644
--- a/fs/gfs2/inode.c
+++ b/fs/gfs2/inode.c
@@ -248,12 +248,10 @@ static int gfs2_dinode_in(struct gfs2_inode *ip, const void *buf)
 {
 	struct gfs2_dinode_host *di = &ip->i_di;
 	const struct gfs2_dinode *str = buf;
+	u16 height;
 
-	if (ip->i_no_addr != be64_to_cpu(str->di_num.no_addr)) {
-		if (gfs2_consist_inode(ip))
-			gfs2_dinode_print(ip);
-		return -EIO;
-	}
+	if (unlikely(ip->i_no_addr != be64_to_cpu(str->di_num.no_addr)))
+		goto corrupt;
 	ip->i_no_formal_ino = be64_to_cpu(str->di_num.no_formal_ino);
 	ip->i_inode.i_mode = be32_to_cpu(str->di_mode);
 	ip->i_inode.i_rdev = 0;
@@ -290,7 +288,10 @@ static int gfs2_dinode_in(struct gfs2_inode *ip, const void *buf)
 
 	di->di_flags = be32_to_cpu(str->di_flags);
 	gfs2_set_inode_flags(&ip->i_inode);
-	di->di_height = be16_to_cpu(str->di_height);
+	height = be16_to_cpu(str->di_height);
+	if (unlikely(height > GFS2_MAX_META_HEIGHT))
+		goto corrupt;
+	ip->i_height = (u8)height;
 
 	di->di_depth = be16_to_cpu(str->di_depth);
 	di->di_entries = be32_to_cpu(str->di_entries);
@@ -300,6 +301,10 @@ static int gfs2_dinode_in(struct gfs2_inode *ip, const void *buf)
 		gfs2_set_aops(&ip->i_inode);
 
 	return 0;
+corrupt:
+	if (gfs2_consist_inode(ip))
+		gfs2_dinode_print(ip);
+	return -EIO;
 }
 
 /**
@@ -1401,7 +1406,7 @@ void gfs2_dinode_out(const struct gfs2_inode *ip, void *buf)
 	str->di_generation = cpu_to_be64(di->di_generation);
 
 	str->di_flags = cpu_to_be32(di->di_flags);
-	str->di_height = cpu_to_be16(di->di_height);
+	str->di_height = cpu_to_be16(ip->i_height);
 	str->di_payload_format = cpu_to_be32(S_ISDIR(ip->i_inode.i_mode) &&
 					     !(ip->i_di.di_flags & GFS2_DIF_EXHASH) ?
 					     GFS2_FORMAT_DE : 0);
@@ -1430,7 +1435,6 @@ void gfs2_dinode_print(const struct gfs2_inode *ip)
 	printk(KERN_INFO "  di_goal_data = %llu\n",
 	       (unsigned long long)di->di_goal_data);
 	printk(KERN_INFO "  di_flags = 0x%.8X\n", di->di_flags);
-	printk(KERN_INFO "  di_height = %u\n", di->di_height);
 	printk(KERN_INFO "  di_depth = %u\n", di->di_depth);
 	printk(KERN_INFO "  di_entries = %u\n", di->di_entries);
 	printk(KERN_INFO "  di_eattr = %llu\n",
diff --git a/fs/gfs2/inode.h b/fs/gfs2/inode.h
index d446506..db73868 100644
--- a/fs/gfs2/inode.h
+++ b/fs/gfs2/inode.h
@@ -12,7 +12,7 @@
 
 static inline int gfs2_is_stuffed(const struct gfs2_inode *ip)
 {
-	return !ip->i_di.di_height;
+	return !ip->i_height;
 }
 
 static inline int gfs2_is_jdata(const struct gfs2_inode *ip)
diff --git a/fs/gfs2/super.c b/fs/gfs2/super.c
index ef0562c..88497b0 100644
--- a/fs/gfs2/super.c
+++ b/fs/gfs2/super.c
@@ -316,6 +316,7 @@ int gfs2_read_sb(struct gfs2_sbd *sdp, struct gfs2_glock *gl, int silent)
 		sdp->sd_heightsize[x] = space;
 	}
 	sdp->sd_max_height = x;
+	sdp->sd_heightsize[x] = ~0;
 	gfs2_assert(sdp, sdp->sd_max_height <= GFS2_MAX_META_HEIGHT);
 
 	sdp->sd_jheightsize[0] = sdp->sd_sb.sb_bsize -
@@ -334,6 +335,7 @@ int gfs2_read_sb(struct gfs2_sbd *sdp, struct gfs2_glock *gl, int silent)
 		sdp->sd_jheightsize[x] = space;
 	}
 	sdp->sd_max_jheight = x;
+	sdp->sd_jheightsize[x] = ~0;
 	gfs2_assert(sdp, sdp->sd_max_jheight <= GFS2_MAX_META_HEIGHT);
 
 	return 0;
-- 
1.5.1.2

[Cluster-devel] [PATCH 03/48] [GFS2] Get rid of unneeded parameter in gfs2_rlist_alloc

[swhiteho]

From: Bob Peterson <rpeterso@redhat.com>

This patch removed the unnecessary parameter from function
gfs2_rlist_alloc.  The parameter was always passed in as 0.

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

diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index 5a31870..e8e48b6 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -680,7 +680,7 @@ static int do_strip(struct gfs2_inode *ip, struct buffer_head *dibh,
 	else
 		goto out; /* Nothing to do */
 
-	gfs2_rlist_alloc(&rlist, LM_ST_EXCLUSIVE, 0);
+	gfs2_rlist_alloc(&rlist, LM_ST_EXCLUSIVE);
 
 	for (x = 0; x < rlist.rl_rgrps; x++) {
 		struct gfs2_rgrpd *rgd;
diff --git a/fs/gfs2/dir.c b/fs/gfs2/dir.c
index c347095..78c236f 100644
--- a/fs/gfs2/dir.c
+++ b/fs/gfs2/dir.c
@@ -1894,7 +1894,7 @@ static int leaf_dealloc(struct gfs2_inode *dip, u32 index, u32 len,
 		l_blocks++;
 	}
 
-	gfs2_rlist_alloc(&rlist, LM_ST_EXCLUSIVE, 0);
+	gfs2_rlist_alloc(&rlist, LM_ST_EXCLUSIVE);
 
 	for (x = 0; x < rlist.rl_rgrps; x++) {
 		struct gfs2_rgrpd *rgd;
diff --git a/fs/gfs2/eattr.c b/fs/gfs2/eattr.c
index bee9970..04febbc 100644
--- a/fs/gfs2/eattr.c
+++ b/fs/gfs2/eattr.c
@@ -1347,7 +1347,7 @@ static int ea_dealloc_indirect(struct gfs2_inode *ip)
 	else
 		goto out;
 
-	gfs2_rlist_alloc(&rlist, LM_ST_EXCLUSIVE, 0);
+	gfs2_rlist_alloc(&rlist, LM_ST_EXCLUSIVE);
 
 	for (x = 0; x < rlist.rl_rgrps; x++) {
 		struct gfs2_rgrpd *rgd;
diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c
index 3552110..7b9d6f1 100644
--- a/fs/gfs2/rgrp.c
+++ b/fs/gfs2/rgrp.c
@@ -1,6 +1,6 @@
 /*
  * Copyright (C) Sistina Software, Inc.  1997-2003 All rights reserved.
- * Copyright (C) 2004-2007 Red Hat, Inc.  All rights reserved.
+ * Copyright (C) 2004-2008 Red Hat, Inc.  All rights reserved.
  *
  * This copyrighted material is made available to anyone wishing to use,
  * modify, copy, or redistribute it subject to the terms and conditions
@@ -1699,8 +1699,7 @@ void gfs2_rlist_add(struct gfs2_sbd *sdp, struct gfs2_rgrp_list *rlist,
  *
  */
 
-void gfs2_rlist_alloc(struct gfs2_rgrp_list *rlist, unsigned int state,
-		      int flags)
+void gfs2_rlist_alloc(struct gfs2_rgrp_list *rlist, unsigned int state)
 {
 	unsigned int x;
 
@@ -1708,7 +1707,7 @@ void gfs2_rlist_alloc(struct gfs2_rgrp_list *rlist, unsigned int state,
 				GFP_NOFS | __GFP_NOFAIL);
 	for (x = 0; x < rlist->rl_rgrps; x++)
 		gfs2_holder_init(rlist->rl_rgd[x]->rd_gl,
-				state, flags,
+				state, 0,
 				&rlist->rl_ghs[x]);
 }
 
diff --git a/fs/gfs2/rgrp.h b/fs/gfs2/rgrp.h
index 149bb16..5683605 100644
--- a/fs/gfs2/rgrp.h
+++ b/fs/gfs2/rgrp.h
@@ -1,6 +1,6 @@
 /*
  * Copyright (C) Sistina Software, Inc.  1997-2003 All rights reserved.
- * Copyright (C) 2004-2006 Red Hat, Inc.  All rights reserved.
+ * Copyright (C) 2004-2008 Red Hat, Inc.  All rights reserved.
  *
  * This copyrighted material is made available to anyone wishing to use,
  * modify, copy, or redistribute it subject to the terms and conditions
@@ -64,8 +64,7 @@ struct gfs2_rgrp_list {
 
 void gfs2_rlist_add(struct gfs2_sbd *sdp, struct gfs2_rgrp_list *rlist,
 		    u64 block);
-void gfs2_rlist_alloc(struct gfs2_rgrp_list *rlist, unsigned int state,
-		      int flags);
+void gfs2_rlist_alloc(struct gfs2_rgrp_list *rlist, unsigned int state);
 void gfs2_rlist_free(struct gfs2_rgrp_list *rlist);
 u64 gfs2_ri_total(struct gfs2_sbd *sdp);
 
-- 
1.5.1.2

[Cluster-devel] [PATCH 04/48] [GFS2] Fix debug inode printing

[swhiteho]

From: Bob Peterson <rpeterso@redhat.com>

I noticed that the latest change to i_height got rid of the
value from the inode dump.  This patch adds it back.

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

diff --git a/fs/gfs2/inode.c b/fs/gfs2/inode.c
index ff66ab7..db5961a 100644
--- a/fs/gfs2/inode.c
+++ b/fs/gfs2/inode.c
@@ -1,6 +1,6 @@
 /*
  * Copyright (C) Sistina Software, Inc.  1997-2003 All rights reserved.
- * Copyright (C) 2004-2006 Red Hat, Inc.  All rights reserved.
+ * Copyright (C) 2004-2008 Red Hat, Inc.  All rights reserved.
  *
  * This copyrighted material is made available to anyone wishing to use,
  * modify, copy, or redistribute it subject to the terms and conditions
@@ -1435,6 +1435,7 @@ void gfs2_dinode_print(const struct gfs2_inode *ip)
 	printk(KERN_INFO "  di_goal_data = %llu\n",
 	       (unsigned long long)di->di_goal_data);
 	printk(KERN_INFO "  di_flags = 0x%.8X\n", di->di_flags);
+	printk(KERN_INFO "  i_height = %u\n", ip->i_height);
 	printk(KERN_INFO "  di_depth = %u\n", di->di_depth);
 	printk(KERN_INFO "  di_entries = %u\n", di->di_entries);
 	printk(KERN_INFO "  di_eattr = %llu\n",
-- 
1.5.1.2

[Cluster-devel] [PATCH 05/48] [GFS2] Only do lo_incore_commit once

[swhiteho]

From: Bob Peterson <rpeterso@redhat.com>

This patch is performance related.  When we're doing a log flush,
I noticed we were calling buf_lo_incore_commit twice: once for
data bufs and once for metadata bufs.  Since this is the same
function and does the same thing in both cases, there should be
no reason to call it twice.  Since we only need to call it once,
we can also make it faster by removing it from the generic "lops"
code and making it a stand-along static function.

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

diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h
index 43472ba..7ae12d2 100644
--- a/fs/gfs2/incore.h
+++ b/fs/gfs2/incore.h
@@ -44,7 +44,6 @@ struct gfs2_log_header_host {
 
 struct gfs2_log_operations {
 	void (*lo_add) (struct gfs2_sbd *sdp, struct gfs2_log_element *le);
-	void (*lo_incore_commit) (struct gfs2_sbd *sdp, struct gfs2_trans *tr);
 	void (*lo_before_commit) (struct gfs2_sbd *sdp);
 	void (*lo_after_commit) (struct gfs2_sbd *sdp, struct gfs2_ail *ai);
 	void (*lo_before_scan) (struct gfs2_jdesc *jd,
diff --git a/fs/gfs2/log.c b/fs/gfs2/log.c
index 161ab6f..b335304 100644
--- a/fs/gfs2/log.c
+++ b/fs/gfs2/log.c
@@ -779,6 +779,21 @@ static void log_refund(struct gfs2_sbd *sdp, struct gfs2_trans *tr)
 	gfs2_log_unlock(sdp);
 }
 
+static void buf_lo_incore_commit(struct gfs2_sbd *sdp, struct gfs2_trans *tr)
+{
+	struct list_head *head = &tr->tr_list_buf;
+	struct gfs2_bufdata *bd;
+
+	gfs2_log_lock(sdp);
+	while (!list_empty(head)) {
+		bd = list_entry(head->next, struct gfs2_bufdata, bd_list_tr);
+		list_del_init(&bd->bd_list_tr);
+		tr->tr_num_buf--;
+	}
+	gfs2_log_unlock(sdp);
+	gfs2_assert_warn(sdp, !tr->tr_num_buf);
+}
+
 /**
  * gfs2_log_commit - Commit a transaction to the log
  * @sdp: the filesystem
@@ -790,7 +805,7 @@ static void log_refund(struct gfs2_sbd *sdp, struct gfs2_trans *tr)
 void gfs2_log_commit(struct gfs2_sbd *sdp, struct gfs2_trans *tr)
 {
 	log_refund(sdp, tr);
-	lops_incore_commit(sdp, tr);
+	buf_lo_incore_commit(sdp, tr);
 
 	sdp->sd_vfs->s_dirt = 1;
 	up_read(&sdp->sd_log_flush_lock);
diff --git a/fs/gfs2/lops.c b/fs/gfs2/lops.c
index fae59d6..7138737 100644
--- a/fs/gfs2/lops.c
+++ b/fs/gfs2/lops.c
@@ -152,21 +152,6 @@ out:
 	unlock_buffer(bd->bd_bh);
 }
 
-static void buf_lo_incore_commit(struct gfs2_sbd *sdp, struct gfs2_trans *tr)
-{
-	struct list_head *head = &tr->tr_list_buf;
-	struct gfs2_bufdata *bd;
-
-	gfs2_log_lock(sdp);
-	while (!list_empty(head)) {
-		bd = list_entry(head->next, struct gfs2_bufdata, bd_list_tr);
-		list_del_init(&bd->bd_list_tr);
-		tr->tr_num_buf--;
-	}
-	gfs2_log_unlock(sdp);
-	gfs2_assert_warn(sdp, !tr->tr_num_buf);
-}
-
 static void buf_lo_before_commit(struct gfs2_sbd *sdp)
 {
 	struct buffer_head *bh;
@@ -737,7 +722,6 @@ static void databuf_lo_after_commit(struct gfs2_sbd *sdp, struct gfs2_ail *ai)
 
 const struct gfs2_log_operations gfs2_buf_lops = {
 	.lo_add = buf_lo_add,
-	.lo_incore_commit = buf_lo_incore_commit,
 	.lo_before_commit = buf_lo_before_commit,
 	.lo_after_commit = buf_lo_after_commit,
 	.lo_before_scan = buf_lo_before_scan,
@@ -763,7 +747,6 @@ const struct gfs2_log_operations gfs2_rg_lops = {
 
 const struct gfs2_log_operations gfs2_databuf_lops = {
 	.lo_add = databuf_lo_add,
-	.lo_incore_commit = buf_lo_incore_commit,
 	.lo_before_commit = databuf_lo_before_commit,
 	.lo_after_commit = databuf_lo_after_commit,
 	.lo_scan_elements = databuf_lo_scan_elements,
diff --git a/fs/gfs2/lops.h b/fs/gfs2/lops.h
index 41a00df..3c0b273 100644
--- a/fs/gfs2/lops.h
+++ b/fs/gfs2/lops.h
@@ -1,6 +1,6 @@
 /*
  * Copyright (C) Sistina Software, Inc.  1997-2003 All rights reserved.
- * Copyright (C) 2004-2006 Red Hat, Inc.  All rights reserved.
+ * Copyright (C) 2004-2008 Red Hat, Inc.  All rights reserved.
  *
  * This copyrighted material is made available to anyone wishing to use,
  * modify, copy, or redistribute it subject to the terms and conditions
@@ -57,15 +57,6 @@ static inline void lops_add(struct gfs2_sbd *sdp, struct gfs2_log_element *le)
 		le->le_ops->lo_add(sdp, le);
 }
 
-static inline void lops_incore_commit(struct gfs2_sbd *sdp,
-				      struct gfs2_trans *tr)
-{
-	int x;
-	for (x = 0; gfs2_log_ops[x]; x++)
-		if (gfs2_log_ops[x]->lo_incore_commit)
-			gfs2_log_ops[x]->lo_incore_commit(sdp, tr);
-}
-
 static inline void lops_before_commit(struct gfs2_sbd *sdp)
 {
 	int x;
-- 
1.5.1.2

[Cluster-devel] [PATCH 06/48] [GFS2] Misc fixups

[swhiteho]

From: Bob Peterson <rpeterso@redhat.com>

This patch contains two small fixups that didn't fit elsewhere.
They are: (1) get rid of temp variable in find_metapath.
(2) Remove vestigial "ret" variable from gfs2_writepage_common.

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

diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index e8e48b6..e84e384 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -305,11 +305,10 @@ static void find_metapath(struct gfs2_inode *ip, u64 block,
 			  struct metapath *mp)
 {
 	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
-	u64 b = block;
 	unsigned int i;
 
 	for (i = ip->i_height; i--;)
-		mp->mp_list[i] = do_div(b, sdp->sd_inptrs);
+		mp->mp_list[i] = do_div(block, sdp->sd_inptrs);
 
 }
 
diff --git a/fs/gfs2/ops_address.c b/fs/gfs2/ops_address.c
index ac772b6..7523999 100644
--- a/fs/gfs2/ops_address.c
+++ b/fs/gfs2/ops_address.c
@@ -1,6 +1,6 @@
 /*
  * Copyright (C) Sistina Software, Inc.  1997-2003 All rights reserved.
- * Copyright (C) 2004-2007 Red Hat, Inc.  All rights reserved.
+ * Copyright (C) 2004-2008 Red Hat, Inc.  All rights reserved.
  *
  * This copyrighted material is made available to anyone wishing to use,
  * modify, copy, or redistribute it subject to the terms and conditions
@@ -104,11 +104,9 @@ static int gfs2_writepage_common(struct page *page,
 	loff_t i_size = i_size_read(inode);
 	pgoff_t end_index = i_size >> PAGE_CACHE_SHIFT;
 	unsigned offset;
-	int ret = -EIO;
 
 	if (gfs2_assert_withdraw(sdp, gfs2_glock_is_held_excl(ip->i_gl)))
 		goto out;
-	ret = 0;
 	if (current->journal_info)
 		goto redirty;
 	/* Is the page fully outside i_size? (truncate in progress) */
-- 
1.5.1.2

[Cluster-devel] [PATCH 07/48] [GFS2] Only wake the reclaim daemon if we need to

[swhiteho]

From: Bob Peterson <rpeterso@redhat.com>

This patch only wakes up the glock reclaim daemon if there is
actually something to be reclaimed.

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 7175a4d..5752dec 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -1594,10 +1594,10 @@ void gfs2_glock_schedule_for_reclaim(struct gfs2_glock *gl)
 		gfs2_glock_hold(gl);
 		list_add(&gl->gl_reclaim, &sdp->sd_reclaim_list);
 		atomic_inc(&sdp->sd_reclaim_count);
-	}
-	spin_unlock(&sdp->sd_reclaim_lock);
-
-	wake_up(&sdp->sd_reclaim_wq);
+		spin_unlock(&sdp->sd_reclaim_lock);
+		wake_up(&sdp->sd_reclaim_wq);
+	} else
+		spin_unlock(&sdp->sd_reclaim_lock);
 }
 
 /**
-- 
1.5.1.2

[Cluster-devel] [PATCH 08/48] [GFS2] make gfs2_glock_hold() static

[swhiteho]

From: Adrian Bunk <bunk@kernel.org>

gfs2_glock_hold() can now become static.

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c
index 5752dec..befcda0 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -197,7 +197,7 @@ static void glock_free(struct gfs2_glock *gl)
  *
  */
 
-void gfs2_glock_hold(struct gfs2_glock *gl)
+static void gfs2_glock_hold(struct gfs2_glock *gl)
 {
 	atomic_inc(&gl->gl_ref);
 }
diff --git a/fs/gfs2/glock.h b/fs/gfs2/glock.h
index 2f9c6d1..ace5770 100644
--- a/fs/gfs2/glock.h
+++ b/fs/gfs2/glock.h
@@ -79,7 +79,6 @@ static inline int gfs2_glock_is_blocking(struct gfs2_glock *gl)
 int gfs2_glock_get(struct gfs2_sbd *sdp,
 		   u64 number, const struct gfs2_glock_operations *glops,
 		   int create, struct gfs2_glock **glp);
-void gfs2_glock_hold(struct gfs2_glock *gl);
 int gfs2_glock_put(struct gfs2_glock *gl);
 void gfs2_holder_init(struct gfs2_glock *gl, unsigned int state, unsigned flags,
 		      struct gfs2_holder *gh);
-- 
1.5.1.2

[Cluster-devel] [PATCH 09/48] [GFS2] Plug an unlikely leak

[swhiteho]

From: Bob Peterson <rpeterso@redhat.com>

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

diff --git a/fs/gfs2/lops.c b/fs/gfs2/lops.c
index 7138737..4390f6f 100644
--- a/fs/gfs2/lops.c
+++ b/fs/gfs2/lops.c
@@ -404,8 +404,10 @@ static int revoke_lo_scan_elements(struct gfs2_jdesc *jd, unsigned int start,
 			blkno = be64_to_cpu(*(__be64 *)(bh->b_data + offset));
 
 			error = gfs2_revoke_add(sdp, blkno, start);
-			if (error < 0)
+			if (error < 0) {
+				brelse(bh);
 				return error;
+			}
 			else if (error)
 				sdp->sd_found_revokes++;
 
-- 
1.5.1.2

[Cluster-devel] [PATCH 10/48] [GFS2] Allocate gfs2_rgrpd from slab memory

[swhiteho]

From: Bob Peterson <rpeterso@redhat.com>

This patch moves the gfs2_rgrpd structure to its own slab
memory.  This makes it easier to control and monitor, and
yields less memory fragmentation.

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

diff --git a/fs/gfs2/main.c b/fs/gfs2/main.c
index 9c7765c..053e2eb 100644
--- a/fs/gfs2/main.c
+++ b/fs/gfs2/main.c
@@ -89,6 +89,12 @@ static int __init init_gfs2_fs(void)
 	if (!gfs2_bufdata_cachep)
 		goto fail;
 
+	gfs2_rgrpd_cachep = kmem_cache_create("gfs2_rgrpd",
+					      sizeof(struct gfs2_rgrpd),
+					      0, 0, NULL);
+	if (!gfs2_rgrpd_cachep)
+		goto fail;
+
 	error = register_filesystem(&gfs2_fs_type);
 	if (error)
 		goto fail;
@@ -108,6 +114,9 @@ fail_unregister:
 fail:
 	gfs2_glock_exit();
 
+	if (gfs2_rgrpd_cachep)
+		kmem_cache_destroy(gfs2_rgrpd_cachep);
+
 	if (gfs2_bufdata_cachep)
 		kmem_cache_destroy(gfs2_bufdata_cachep);
 
@@ -133,6 +142,7 @@ static void __exit exit_gfs2_fs(void)
 	unregister_filesystem(&gfs2_fs_type);
 	unregister_filesystem(&gfs2meta_fs_type);
 
+	kmem_cache_destroy(gfs2_rgrpd_cachep);
 	kmem_cache_destroy(gfs2_bufdata_cachep);
 	kmem_cache_destroy(gfs2_inode_cachep);
 	kmem_cache_destroy(gfs2_glock_cachep);
diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c
index 7b9d6f1..dc7e83e 100644
--- a/fs/gfs2/rgrp.c
+++ b/fs/gfs2/rgrp.c
@@ -353,7 +353,7 @@ static void clear_rgrpdi(struct gfs2_sbd *sdp)
 		}
 
 		kfree(rgd->rd_bits);
-		kfree(rgd);
+		kmem_cache_free(gfs2_rgrpd_cachep, rgd);
 	}
 }
 
@@ -516,7 +516,7 @@ static int read_rindex_entry(struct gfs2_inode *ip,
 		return error;
 	}
 
-	rgd = kzalloc(sizeof(struct gfs2_rgrpd), GFP_NOFS);
+	rgd = kmem_cache_zalloc(gfs2_rgrpd_cachep, GFP_NOFS);
 	error = -ENOMEM;
 	if (!rgd)
 		return error;
diff --git a/fs/gfs2/util.c b/fs/gfs2/util.c
index 424a077..fe9c28e 100644
--- a/fs/gfs2/util.c
+++ b/fs/gfs2/util.c
@@ -25,6 +25,7 @@
 struct kmem_cache *gfs2_glock_cachep __read_mostly;
 struct kmem_cache *gfs2_inode_cachep __read_mostly;
 struct kmem_cache *gfs2_bufdata_cachep __read_mostly;
+struct kmem_cache *gfs2_rgrpd_cachep __read_mostly;
 
 void gfs2_assert_i(struct gfs2_sbd *sdp)
 {
diff --git a/fs/gfs2/util.h b/fs/gfs2/util.h
index 28938a4..ac0c567 100644
--- a/fs/gfs2/util.h
+++ b/fs/gfs2/util.h
@@ -147,6 +147,7 @@ gfs2_io_error_bh_i((sdp), (bh), __FUNCTION__, __FILE__, __LINE__);
 extern struct kmem_cache *gfs2_glock_cachep;
 extern struct kmem_cache *gfs2_inode_cachep;
 extern struct kmem_cache *gfs2_bufdata_cachep;
+extern struct kmem_cache *gfs2_rgrpd_cachep;
 
 static inline unsigned int gfs2_tune_get_i(struct gfs2_tune *gt,
 					   unsigned int *p)
-- 
1.5.1.2

[Cluster-devel] [PATCH 11/48] [GFS2] Combine rg_flags and rd_flags

[swhiteho]

From: Bob Peterson <rpeterso@redhat.com>

This patch reduces the memory required by GFS2 by combining
the rd_flags and rg_flags (in core only).

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

diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h
index 7ae12d2..39bab7b 100644
--- a/fs/gfs2/incore.h
+++ b/fs/gfs2/incore.h
@@ -69,7 +69,6 @@ struct gfs2_bitmap {
 };
 
 struct gfs2_rgrp_host {
-	u32 rg_flags;
 	u32 rg_free;
 	u32 rg_dinodes;
 	u64 rg_igeneration;
@@ -95,8 +94,9 @@ struct gfs2_rgrpd {
 	u32 rd_last_alloc_data;
 	u32 rd_last_alloc_meta;
 	struct gfs2_sbd *rd_sbd;
-	unsigned long rd_flags;
-#define GFS2_RDF_CHECK        0x0001          /* Need to check for unlinked inodes */
+	unsigned char rd_flags;
+#define GFS2_RDF_CHECK        0x01      /* Need to check for unlinked inodes */
+#define GFS2_RDF_NOALLOC      0x02      /* rg prohibits allocation */
 };
 
 enum gfs2_state_bits {
diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c
index dc7e83e..da60ce8 100644
--- a/fs/gfs2/rgrp.c
+++ b/fs/gfs2/rgrp.c
@@ -655,21 +655,31 @@ int gfs2_rindex_hold(struct gfs2_sbd *sdp, struct gfs2_holder *ri_gh)
 	return error;
 }
 
-static void gfs2_rgrp_in(struct gfs2_rgrp_host *rg, const void *buf)
+static void gfs2_rgrp_in(struct gfs2_rgrpd *rgd, const void *buf)
 {
 	const struct gfs2_rgrp *str = buf;
+	struct gfs2_rgrp_host *rg = &rgd->rd_rg;
+	u32 rg_flags;
 
-	rg->rg_flags = be32_to_cpu(str->rg_flags);
+	rg_flags = be32_to_cpu(str->rg_flags);
+	if (rg_flags & GFS2_RGF_NOALLOC)
+		rgd->rd_flags |= GFS2_RDF_NOALLOC;
+	else
+		rgd->rd_flags &= ~GFS2_RDF_NOALLOC;
 	rg->rg_free = be32_to_cpu(str->rg_free);
 	rg->rg_dinodes = be32_to_cpu(str->rg_dinodes);
 	rg->rg_igeneration = be64_to_cpu(str->rg_igeneration);
 }
 
-static void gfs2_rgrp_out(const struct gfs2_rgrp_host *rg, void *buf)
+static void gfs2_rgrp_out(struct gfs2_rgrpd *rgd, void *buf)
 {
 	struct gfs2_rgrp *str = buf;
+	struct gfs2_rgrp_host *rg = &rgd->rd_rg;
+	u32 rg_flags = 0;
 
-	str->rg_flags = cpu_to_be32(rg->rg_flags);
+	if (rgd->rd_flags & GFS2_RDF_NOALLOC)
+		rg_flags |= GFS2_RGF_NOALLOC;
+	str->rg_flags = cpu_to_be32(rg_flags);
 	str->rg_free = cpu_to_be32(rg->rg_free);
 	str->rg_dinodes = cpu_to_be32(rg->rg_dinodes);
 	str->__pad = cpu_to_be32(0);
@@ -727,7 +737,7 @@ int gfs2_rgrp_bh_get(struct gfs2_rgrpd *rgd)
 	}
 
 	if (rgd->rd_rg_vn != gl->gl_vn) {
-		gfs2_rgrp_in(&rgd->rd_rg, (rgd->rd_bits[0].bi_bh)->b_data);
+		gfs2_rgrp_in(rgd, (rgd->rd_bits[0].bi_bh)->b_data);
 		rgd->rd_rg_vn = gl->gl_vn;
 	}
 
@@ -840,7 +850,7 @@ static int try_rgrp_fit(struct gfs2_rgrpd *rgd, struct gfs2_alloc *al)
 	struct gfs2_sbd *sdp = rgd->rd_sbd;
 	int ret = 0;
 
-	if (rgd->rd_rg.rg_flags & GFS2_RGF_NOALLOC)
+	if (rgd->rd_flags & GFS2_RDF_NOALLOC)
 		return 0;
 
 	spin_lock(&sdp->sd_rindex_spin);
@@ -1431,7 +1441,7 @@ u64 gfs2_alloc_data(struct gfs2_inode *ip)
 	rgd->rd_rg.rg_free--;
 
 	gfs2_trans_add_bh(rgd->rd_gl, rgd->rd_bits[0].bi_bh, 1);
-	gfs2_rgrp_out(&rgd->rd_rg, rgd->rd_bits[0].bi_bh->b_data);
+	gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data);
 
 	al->al_alloced++;
 
@@ -1476,7 +1486,7 @@ u64 gfs2_alloc_meta(struct gfs2_inode *ip)
 	rgd->rd_rg.rg_free--;
 
 	gfs2_trans_add_bh(rgd->rd_gl, rgd->rd_bits[0].bi_bh, 1);
-	gfs2_rgrp_out(&rgd->rd_rg, rgd->rd_bits[0].bi_bh->b_data);
+	gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data);
 
 	al->al_alloced++;
 
@@ -1519,7 +1529,7 @@ u64 gfs2_alloc_di(struct gfs2_inode *dip, u64 *generation)
 	rgd->rd_rg.rg_dinodes++;
 	*generation = rgd->rd_rg.rg_igeneration++;
 	gfs2_trans_add_bh(rgd->rd_gl, rgd->rd_bits[0].bi_bh, 1);
-	gfs2_rgrp_out(&rgd->rd_rg, rgd->rd_bits[0].bi_bh->b_data);
+	gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data);
 
 	al->al_alloced++;
 
@@ -1553,7 +1563,7 @@ void gfs2_free_data(struct gfs2_inode *ip, u64 bstart, u32 blen)
 	rgd->rd_rg.rg_free += blen;
 
 	gfs2_trans_add_bh(rgd->rd_gl, rgd->rd_bits[0].bi_bh, 1);
-	gfs2_rgrp_out(&rgd->rd_rg, rgd->rd_bits[0].bi_bh->b_data);
+	gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data);
 
 	gfs2_trans_add_rg(rgd);
 
@@ -1581,7 +1591,7 @@ void gfs2_free_meta(struct gfs2_inode *ip, u64 bstart, u32 blen)
 	rgd->rd_rg.rg_free += blen;
 
 	gfs2_trans_add_bh(rgd->rd_gl, rgd->rd_bits[0].bi_bh, 1);
-	gfs2_rgrp_out(&rgd->rd_rg, rgd->rd_bits[0].bi_bh->b_data);
+	gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data);
 
 	gfs2_trans_add_rg(rgd);
 
@@ -1601,7 +1611,7 @@ void gfs2_unlink_di(struct inode *inode)
 	if (!rgd)
 		return;
 	gfs2_trans_add_bh(rgd->rd_gl, rgd->rd_bits[0].bi_bh, 1);
-	gfs2_rgrp_out(&rgd->rd_rg, rgd->rd_bits[0].bi_bh->b_data);
+	gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data);
 	gfs2_trans_add_rg(rgd);
 }
 
@@ -1621,7 +1631,7 @@ static void gfs2_free_uninit_di(struct gfs2_rgrpd *rgd, u64 blkno)
 	rgd->rd_rg.rg_free++;
 
 	gfs2_trans_add_bh(rgd->rd_gl, rgd->rd_bits[0].bi_bh, 1);
-	gfs2_rgrp_out(&rgd->rd_rg, rgd->rd_bits[0].bi_bh->b_data);
+	gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data);
 
 	gfs2_statfs_change(sdp, 0, +1, -1);
 	gfs2_trans_add_rg(rgd);
-- 
1.5.1.2

[Cluster-devel] [PATCH 12/48] [GFS2] Get rid of gl_waiters2

[swhiteho]

From: Bob Peterson <rpeterso@redhat.com>

This patch reduces memory by replacing the int variable
gl_waiters2 by a single bit in the gl_flags.

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 befcda0..951cb91 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -595,11 +595,12 @@ static void run_queue(struct gfs2_glock *gl)
 			blocked = rq_mutex(gh);
 		} else if (test_bit(GLF_DEMOTE, &gl->gl_flags)) {
 			blocked = rq_demote(gl);
-			if (gl->gl_waiters2 && !blocked) {
+			if (test_bit(GLF_WAITERS2, &gl->gl_flags) &&
+				     !blocked) {
 				set_bit(GLF_DEMOTE, &gl->gl_flags);
 				gl->gl_demote_state = LM_ST_UNLOCKED;
 			}
-			gl->gl_waiters2 = 0;
+			clear_bit(GLF_WAITERS2, &gl->gl_flags);
 		} else if (!list_empty(&gl->gl_waiters3)) {
 			gh = list_entry(gl->gl_waiters3.next,
 					struct gfs2_holder, gh_list);
@@ -710,7 +711,7 @@ static void handle_callback(struct gfs2_glock *gl, unsigned int state,
 	} else if (gl->gl_demote_state != LM_ST_UNLOCKED &&
 			gl->gl_demote_state != state) {
 		if (test_bit(GLF_DEMOTE_IN_PROGRESS,  &gl->gl_flags)) 
-			gl->gl_waiters2 = 1;
+			set_bit(GLF_WAITERS2, &gl->gl_flags);
 		else 
 			gl->gl_demote_state = LM_ST_UNLOCKED;
 	}
diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h
index 39bab7b..fe14f6a 100644
--- a/fs/gfs2/incore.h
+++ b/fs/gfs2/incore.h
@@ -1,6 +1,6 @@
 /*
  * Copyright (C) Sistina Software, Inc.  1997-2003 All rights reserved.
- * Copyright (C) 2004-2007 Red Hat, Inc.  All rights reserved.
+ * Copyright (C) 2004-2008 Red Hat, Inc.  All rights reserved.
  *
  * This copyrighted material is made available to anyone wishing to use,
  * modify, copy, or redistribute it subject to the terms and conditions
@@ -167,6 +167,7 @@ enum {
 	GLF_DIRTY		= 5,
 	GLF_DEMOTE_IN_PROGRESS	= 6,
 	GLF_LFLUSH		= 7,
+	GLF_WAITERS2		= 8,
 };
 
 struct gfs2_glock {
@@ -186,7 +187,6 @@ struct gfs2_glock {
 	struct list_head gl_holders;
 	struct list_head gl_waiters1;	/* HIF_MUTEX */
 	struct list_head gl_waiters3;	/* HIF_PROMOTE */
-	int gl_waiters2;		/* GIF_DEMOTE */
 
 	const struct gfs2_glock_operations *gl_ops;
 
-- 
1.5.1.2

[Cluster-devel] [PATCH 13/48] [GFS2] Move part of gfs2_block_map into a separate function

[swhiteho]

From: Steven Whitehouse <swhiteho@redhat.com>

This is required to enable future changes to the block
mapping code.

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

diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index e84e384..08d1be4 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -384,6 +384,35 @@ static int lookup_block(struct gfs2_inode *ip, struct buffer_head *bh,
 	return 0;
 }
 
+static int lookup_metapath(struct inode *inode, struct metapath *mp,
+			   int create, int *new, u64 *dblock,
+			   struct buffer_head **dibh, struct buffer_head **bh)
+{
+	struct gfs2_inode *ip = GFS2_I(inode);
+	unsigned int end_of_metadata = ip->i_height - 1;
+	unsigned int x;
+	int ret = gfs2_meta_inode_buffer(ip, bh);
+	if (ret)
+		return ret;
+
+	*dibh = *bh;
+	get_bh(*dibh);
+
+	for (x = 0; x < end_of_metadata; x++) {
+		lookup_block(ip, *bh, x, mp, create, new, dblock);
+		brelse(*bh);
+		*bh = NULL;
+		if (!dblock)
+			return 0;
+
+		ret = gfs2_meta_indirect_buffer(ip, x+1, *dblock, *new, bh);
+		if (ret)
+			return ret;
+	}
+
+	return lookup_block(ip, *bh, end_of_metadata, mp, create, new, dblock);
+}
+
 static inline void bmap_lock(struct inode *inode, int create)
 {
 	struct gfs2_inode *ip = GFS2_I(inode);
@@ -419,10 +448,8 @@ int gfs2_block_map(struct inode *inode, sector_t lblock,
 {
 	struct gfs2_inode *ip = GFS2_I(inode);
 	struct gfs2_sbd *sdp = GFS2_SB(inode);
-	struct buffer_head *bh;
 	unsigned int bsize = sdp->sd_sb.sb_bsize;
-	unsigned int end_of_metadata;
-	unsigned int x;
+	struct buffer_head *bh = NULL;
 	int error = 0;
 	int new = 0;
 	u64 dblock = 0;
@@ -459,25 +486,11 @@ int gfs2_block_map(struct inode *inode, sector_t lblock,
 	}
 
 	find_metapath(ip, lblock, &mp);
-	end_of_metadata = ip->i_height - 1;
-	error = gfs2_meta_inode_buffer(ip, &bh);
-	if (error)
+	error = lookup_metapath(inode, &mp, create, &new, &dblock, &dibh, &bh);
+	if (error < 0)
 		goto out_fail;
-	dibh = bh;
-	get_bh(dibh);
-
-	for (x = 0; x < end_of_metadata; x++) {
-		lookup_block(ip, bh, x, &mp, create, &new, &dblock);
-		brelse(bh);
-		if (!dblock)
-			goto out_ok;
+	boundary = error;
 
-		error = gfs2_meta_indirect_buffer(ip, x+1, dblock, new, &bh);
-		if (error)
-			goto out_fail;
-	}
-
-	boundary = lookup_block(ip, bh, end_of_metadata, &mp, create, &new, &dblock);
 	if (dblock) {
 		map_bh(bh_map, inode->i_sb, dblock);
 		if (boundary)
@@ -489,6 +502,7 @@ int gfs2_block_map(struct inode *inode, sector_t lblock,
 			goto out_brelse;
 		}
 		while(--maxlen && !buffer_boundary(bh_map)) {
+			unsigned int end_of_metadata = ip->i_height - 1;
 			u64 eblock;
 
 			mp.mp_list[end_of_metadata]++;
@@ -501,7 +515,8 @@ int gfs2_block_map(struct inode *inode, sector_t lblock,
 		}
 	}
 out_brelse:
-	brelse(bh);
+	if (bh)
+		brelse(bh);
 out_ok:
 	error = 0;
 out_fail:
-- 
1.5.1.2

[Cluster-devel] [PATCH 14/48] [GFS2] Introduce array of buffers to struct metapath

[swhiteho]

From: Steven Whitehouse <swhiteho@redhat.com>

The reason for doing this is to allow all the block mapping code
to share the same array. As a result we can remove two arguments
from lookup_metapath since they are now returned via the array.

We also add a function to drop all refs to buffer heads when we
are done with the metapath. The build_height function shares the
struct metapath, but currently still frees its own buffers, and
this will change in a future patch.

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

diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index 08d1be4..2011dd2 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -33,6 +33,7 @@
  * keep it small.
  */
 struct metapath {
+	struct buffer_head *mp_bh[GFS2_MAX_META_HEIGHT];
 	__u16 mp_list[GFS2_MAX_META_HEIGHT];
 };
 
@@ -185,12 +186,11 @@ out:
  * Returns: errno
  */
 
-static int build_height(struct inode *inode, unsigned height)
+static int build_height(struct inode *inode, struct metapath *mp, unsigned height)
 {
 	struct gfs2_inode *ip = GFS2_I(inode);
 	unsigned new_height = height - ip->i_height;
 	struct buffer_head *dibh;
-	struct buffer_head *blocks[GFS2_MAX_META_HEIGHT];
 	struct gfs2_dinode *di;
 	int error;
 	__be64 *bp;
@@ -206,29 +206,29 @@ static int build_height(struct inode *inode, unsigned height)
 
 	for(n = 0; n < new_height; n++) {
 		bn = gfs2_alloc_meta(ip);
-		blocks[n] = gfs2_meta_new(ip->i_gl, bn);
-		gfs2_trans_add_bh(ip->i_gl, blocks[n], 1);
+		mp->mp_bh[n] = gfs2_meta_new(ip->i_gl, bn);
+		gfs2_trans_add_bh(ip->i_gl, mp->mp_bh[n], 1);
 	}
 
 	n = 0;
-	bn = blocks[0]->b_blocknr;
+	bn = mp->mp_bh[0]->b_blocknr;
 	if (new_height > 1) {
 		for(; n < new_height-1; n++) {
-			gfs2_metatype_set(blocks[n], GFS2_METATYPE_IN,
+			gfs2_metatype_set(mp->mp_bh[n], GFS2_METATYPE_IN,
 					  GFS2_FORMAT_IN);
-			gfs2_buffer_clear_tail(blocks[n],
+			gfs2_buffer_clear_tail(mp->mp_bh[n],
 					       sizeof(struct gfs2_meta_header));
-			bp = (__be64 *)(blocks[n]->b_data +
+			bp = (__be64 *)(mp->mp_bh[n]->b_data +
 				     sizeof(struct gfs2_meta_header));
-			*bp = cpu_to_be64(blocks[n+1]->b_blocknr);
-			brelse(blocks[n]);
-			blocks[n] = NULL;
+			*bp = cpu_to_be64(mp->mp_bh[n+1]->b_blocknr);
+			brelse(mp->mp_bh[n]);
+			mp->mp_bh[n] = NULL;
 		}
 	}
-	gfs2_metatype_set(blocks[n], GFS2_METATYPE_IN, GFS2_FORMAT_IN);
-	gfs2_buffer_copy_tail(blocks[n], sizeof(struct gfs2_meta_header),
+	gfs2_metatype_set(mp->mp_bh[n], GFS2_METATYPE_IN, GFS2_FORMAT_IN);
+	gfs2_buffer_copy_tail(mp->mp_bh[n], sizeof(struct gfs2_meta_header),
 			      dibh, sizeof(struct gfs2_dinode));
-	brelse(blocks[n]);
+	brelse(mp->mp_bh[n]);
 	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
 	di = (struct gfs2_dinode *)dibh->b_data;
 	gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode));
@@ -314,7 +314,6 @@ static void find_metapath(struct gfs2_inode *ip, u64 block,
 
 /**
  * metapointer - Return pointer to start of metadata in a buffer
- * @bh: The buffer
  * @height: The metadata height (0 = dinode)
  * @mp: The metapath
  *
@@ -323,9 +322,10 @@ static void find_metapath(struct gfs2_inode *ip, u64 block,
  * metadata tree.
  */
 
-static inline __be64 *metapointer(struct buffer_head *bh, int *boundary,
-			       unsigned int height, const struct metapath *mp)
+static inline __be64 *metapointer(int *boundary, unsigned int height,
+				  const struct metapath *mp)
 {
+	struct buffer_head *bh = mp->mp_bh[height];
 	unsigned int head_size = (height > 0) ?
 		sizeof(struct gfs2_meta_header) : sizeof(struct gfs2_dinode);
 	__be64 *ptr;
@@ -339,7 +339,6 @@ static inline __be64 *metapointer(struct buffer_head *bh, int *boundary,
 /**
  * lookup_block - Get the next metadata block in metadata tree
  * @ip: The GFS2 inode
- * @bh: Buffer containing the pointers to metadata blocks
  * @height: The height of the tree (0 = dinode)
  * @mp: The metapath
  * @create: Non-zero if we may create a new meatdata block
@@ -352,12 +351,12 @@ static inline __be64 *metapointer(struct buffer_head *bh, int *boundary,
  *
  */
 
-static int lookup_block(struct gfs2_inode *ip, struct buffer_head *bh,
-			unsigned int height, struct metapath *mp, int create,
+static int lookup_block(struct gfs2_inode *ip, unsigned int height,
+			struct metapath *mp, int create,
 			int *new, u64 *block)
 {
 	int boundary;
-	__be64 *ptr = metapointer(bh, &boundary, height, mp);
+	__be64 *ptr = metapointer(&boundary, height, mp);
 
 	if (*ptr) {
 		*block = be64_to_cpu(*ptr);
@@ -374,7 +373,7 @@ static int lookup_block(struct gfs2_inode *ip, struct buffer_head *bh,
 	else
 		*block = gfs2_alloc_meta(ip);
 
-	gfs2_trans_add_bh(ip->i_gl, bh, 1);
+	gfs2_trans_add_bh(ip->i_gl, mp->mp_bh[height], 1);
 
 	*ptr = cpu_to_be64(*block);
 	ip->i_di.di_blocks++;
@@ -385,32 +384,38 @@ static int lookup_block(struct gfs2_inode *ip, struct buffer_head *bh,
 }
 
 static int lookup_metapath(struct inode *inode, struct metapath *mp,
-			   int create, int *new, u64 *dblock,
-			   struct buffer_head **dibh, struct buffer_head **bh)
+			   int create, int *new, u64 *dblock)
 {
+	struct buffer_head *bh;
 	struct gfs2_inode *ip = GFS2_I(inode);
 	unsigned int end_of_metadata = ip->i_height - 1;
 	unsigned int x;
-	int ret = gfs2_meta_inode_buffer(ip, bh);
+	int ret = gfs2_meta_inode_buffer(ip, &bh);
 	if (ret)
 		return ret;
 
-	*dibh = *bh;
-	get_bh(*dibh);
+	mp->mp_bh[0] = bh;
 
 	for (x = 0; x < end_of_metadata; x++) {
-		lookup_block(ip, *bh, x, mp, create, new, dblock);
-		brelse(*bh);
-		*bh = NULL;
+		lookup_block(ip, x, mp, create, new, dblock);
 		if (!dblock)
 			return 0;
 
-		ret = gfs2_meta_indirect_buffer(ip, x+1, *dblock, *new, bh);
+		ret = gfs2_meta_indirect_buffer(ip, x+1, *dblock, *new, &mp->mp_bh[x+1]);
 		if (ret)
 			return ret;
 	}
 
-	return lookup_block(ip, *bh, end_of_metadata, mp, create, new, dblock);
+	return lookup_block(ip, end_of_metadata, mp, create, new, dblock);
+}
+
+static void release_metapath(struct metapath *mp)
+{
+	int i;
+
+	for (i = 0; i < GFS2_MAX_META_HEIGHT; i++)
+		if (mp->mp_bh[i])
+			brelse(mp->mp_bh[i]);
 }
 
 static inline void bmap_lock(struct inode *inode, int create)
@@ -449,7 +454,6 @@ int gfs2_block_map(struct inode *inode, sector_t lblock,
 	struct gfs2_inode *ip = GFS2_I(inode);
 	struct gfs2_sbd *sdp = GFS2_SB(inode);
 	unsigned int bsize = sdp->sd_sb.sb_bsize;
-	struct buffer_head *bh = NULL;
 	int error = 0;
 	int new = 0;
 	u64 dblock = 0;
@@ -457,13 +461,13 @@ int gfs2_block_map(struct inode *inode, sector_t lblock,
 	unsigned int maxlen = bh_map->b_size >> inode->i_blkbits;
 	struct metapath mp;
 	u64 size;
-	struct buffer_head *dibh = NULL;
 	const u64 *arr = sdp->sd_heightsize;
 	BUG_ON(maxlen == 0);
 
 	if (gfs2_assert_warn(sdp, !gfs2_is_stuffed(ip)))
 		return 0;
 
+	memset(mp.mp_bh, 0, sizeof(mp.mp_bh));
 	bmap_lock(inode, create);
 	clear_buffer_mapped(bh_map);
 	clear_buffer_new(bh_map);
@@ -480,13 +484,13 @@ int gfs2_block_map(struct inode *inode, sector_t lblock,
 			goto out_ok;
 		while (size > arr[height])
 			height++;
-		error = build_height(inode, height);
+		error = build_height(inode, &mp, height);
 		if (error)
 			goto out_fail;
 	}
 
 	find_metapath(ip, lblock, &mp);
-	error = lookup_metapath(inode, &mp, create, &new, &dblock, &dibh, &bh);
+	error = lookup_metapath(inode, &mp, create, &new, &dblock);
 	if (error < 0)
 		goto out_fail;
 	boundary = error;
@@ -496,17 +500,15 @@ int gfs2_block_map(struct inode *inode, sector_t lblock,
 		if (boundary)
 			set_buffer_boundary(bh_map);
 		if (new) {
-			gfs2_trans_add_bh(ip->i_gl, dibh, 1);
-			gfs2_dinode_out(ip, dibh->b_data);
+			gfs2_trans_add_bh(ip->i_gl, mp.mp_bh[0], 1);
+			gfs2_dinode_out(ip, mp.mp_bh[0]->b_data);
 			set_buffer_new(bh_map);
-			goto out_brelse;
+			goto out_ok;
 		}
 		while(--maxlen && !buffer_boundary(bh_map)) {
-			unsigned int end_of_metadata = ip->i_height - 1;
 			u64 eblock;
-
-			mp.mp_list[end_of_metadata]++;
-			boundary = lookup_block(ip, bh, end_of_metadata, &mp, 0, &new, &eblock);
+			mp.mp_list[ip->i_height - 1]++;
+			boundary = lookup_block(ip, ip->i_height - 1, &mp, 0, &new, &eblock);
 			if (eblock != ++dblock)
 				break;
 			bh_map->b_size += (1 << inode->i_blkbits);
@@ -514,14 +516,10 @@ int gfs2_block_map(struct inode *inode, sector_t lblock,
 				set_buffer_boundary(bh_map);
 		}
 	}
-out_brelse:
-	if (bh)
-		brelse(bh);
 out_ok:
 	error = 0;
 out_fail:
-	if (dibh)
-		brelse(dibh);
+	release_metapath(&mp);
 	bmap_unlock(inode, create);
 	return error;
 }
@@ -819,10 +817,11 @@ static int do_grow(struct gfs2_inode *ip, u64 size)
 
 		down_write(&ip->i_rw_mutex);
 		if (size > arr[ip->i_height]) {
+			struct metapath mp;
 			u8 height = ip->i_height;
 			while(size > arr[height])
 				height++;
-			error = build_height(&ip->i_inode, height);
+			error = build_height(&ip->i_inode, &mp, height);
 		}
 		up_write(&ip->i_rw_mutex);
 		if (error)
-- 
1.5.1.2

[Cluster-devel] [PATCH 15/48] [GFS2] Add consts to various bits of rgrp.c

[swhiteho]

From: Steven Whitehouse <swhiteho@redhat.com>

There are a couple of routines which scan bitmaps where we can
mark the bitmaps const, plus a couple of call sites that can
be updated too.

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

diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c
index da60ce8..5fd8710 100644
--- a/fs/gfs2/rgrp.c
+++ b/fs/gfs2/rgrp.c
@@ -126,23 +126,24 @@ static unsigned char gfs2_testbit(struct gfs2_rgrpd *rgd, unsigned char *buffer,
  * Return: the block number (bitmap buffer scope) that was found
  */
 
-static u32 gfs2_bitfit(unsigned char *buffer, unsigned int buflen, u32 goal,
-		       unsigned char old_state)
+static u32 gfs2_bitfit(const u8 *buffer, unsigned int buflen, u32 goal,
+		       u8 old_state)
 {
-	unsigned char *byte;
+	const u8 *byte;
 	u32 blk = goal;
 	unsigned int bit, bitlong;
-	unsigned long *plong, plong55;
+	const unsigned long *plong;
+#if BITS_PER_LONG == 32
+	const unsigned long plong55 = 0x55555555;
+#else
+	const unsigned long plong55 = 0x5555555555555555;
+#endif
 
 	byte = buffer + (goal / GFS2_NBBY);
-	plong = (unsigned long *)(buffer + (goal / GFS2_NBBY));
+	plong = (const unsigned long *)(buffer + (goal / GFS2_NBBY));
 	bit = (goal % GFS2_NBBY) * GFS2_BIT_SIZE;
 	bitlong = bit;
-#if BITS_PER_LONG == 32
-	plong55 = 0x55555555;
-#else
-	plong55 = 0x5555555555555555;
-#endif
+
 	while (byte < buffer + buflen) {
 
 		if (bitlong == 0 && old_state == 0 && *plong == plong55) {
@@ -179,14 +180,14 @@ static u32 gfs2_bitfit(unsigned char *buffer, unsigned int buflen, u32 goal,
  * Returns: The number of bits
  */
 
-static u32 gfs2_bitcount(struct gfs2_rgrpd *rgd, unsigned char *buffer,
-			      unsigned int buflen, unsigned char state)
+static u32 gfs2_bitcount(struct gfs2_rgrpd *rgd, const u8 *buffer,
+			 unsigned int buflen, u8 state)
 {
-	unsigned char *byte = buffer;
-	unsigned char *end = buffer + buflen;
-	unsigned char state1 = state << 2;
-	unsigned char state2 = state << 4;
-	unsigned char state3 = state << 6;
+	const u8 *byte = buffer;
+	const u8 *end = buffer + buflen;
+	const u8 state1 = state << 2;
+	const u8 state2 = state << 4;
+	const u8 state3 = state << 6;
 	u32 count = 0;
 
 	for (; byte < end; byte++) {
@@ -1327,12 +1328,11 @@ static u32 rgblk_search(struct gfs2_rgrpd *rgd, u32 goal,
 	for (x = 0; x <= length; x++) {
 		/* The GFS2_BLKST_UNLINKED state doesn't apply to the clone
 		   bitmaps, so we must search the originals for that. */
+		const u8 *buffer = bi->bi_bh->b_data + bi->bi_offset;
 		if (old_state != GFS2_BLKST_UNLINKED && bi->bi_clone)
-			blk = gfs2_bitfit(bi->bi_clone + bi->bi_offset,
-					  bi->bi_len, goal, old_state);
-		else
-			blk = gfs2_bitfit(bi->bi_bh->b_data + bi->bi_offset,
-					  bi->bi_len, goal, old_state);
+			buffer = bi->bi_clone + bi->bi_offset;
+
+		blk = gfs2_bitfit(buffer, bi->bi_len, goal, old_state);
 		if (blk != BFITNOENT)
 			break;
 
-- 
1.5.1.2

[Cluster-devel] [PATCH 16/48] [GFS2] Eliminate gl_req_bh

[swhiteho]

From: Bob Peterson <rpeterso@redhat.com>

This patch further reduces the memory needs of GFS2 by
eliminating the gl_req_bh variable from struct gfs2_glock.

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 951cb91..a8387e0 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -338,7 +338,6 @@ int gfs2_glock_get(struct gfs2_sbd *sdp, u64 number,
 	gl->gl_ip = 0;
 	gl->gl_ops = glops;
 	gl->gl_req_gh = NULL;
-	gl->gl_req_bh = NULL;
 	gl->gl_vn = 0;
 	gl->gl_stamp = jiffies;
 	gl->gl_tchange = jiffies;
@@ -744,6 +743,50 @@ static void state_change(struct gfs2_glock *gl, unsigned int new_state)
 }
 
 /**
+ * drop_bh - Called after a lock module unlock completes
+ * @gl: the glock
+ * @ret: the return status
+ *
+ * Doesn't wake up the process waiting on the struct gfs2_holder (if any)
+ * Doesn't drop the reference on the glock the top half took out
+ *
+ */
+
+static void drop_bh(struct gfs2_glock *gl, unsigned int ret)
+{
+	struct gfs2_sbd *sdp = gl->gl_sbd;
+	const struct gfs2_glock_operations *glops = gl->gl_ops;
+	struct gfs2_holder *gh = gl->gl_req_gh;
+
+	gfs2_assert_warn(sdp, test_bit(GLF_LOCK, &gl->gl_flags));
+	gfs2_assert_warn(sdp, list_empty(&gl->gl_holders));
+	gfs2_assert_warn(sdp, !ret);
+
+	state_change(gl, LM_ST_UNLOCKED);
+
+	if (glops->go_inval)
+		glops->go_inval(gl, DIO_METADATA);
+
+	if (gh) {
+		spin_lock(&gl->gl_spin);
+		list_del_init(&gh->gh_list);
+		gh->gh_error = 0;
+		spin_unlock(&gl->gl_spin);
+	}
+
+	spin_lock(&gl->gl_spin);
+	gfs2_demote_wake(gl);
+	gl->gl_req_gh = NULL;
+	clear_bit(GLF_LOCK, &gl->gl_flags);
+	spin_unlock(&gl->gl_spin);
+
+	gfs2_glock_put(gl);
+
+	if (gh)
+		gfs2_holder_wake(gh);
+}
+
+/**
  * xmote_bh - Called after the lock module is done acquiring a lock
  * @gl: The glock in question
  * @ret: the int returned from the lock module
@@ -758,6 +801,11 @@ static void xmote_bh(struct gfs2_glock *gl, unsigned int ret)
 	int prev_state = gl->gl_state;
 	int op_done = 1;
 
+	if ((ret & LM_OUT_ST_MASK) == LM_ST_UNLOCKED) {
+		drop_bh(gl, ret);
+		return;
+	}
+
 	gfs2_assert_warn(sdp, test_bit(GLF_LOCK, &gl->gl_flags));
 	gfs2_assert_warn(sdp, list_empty(&gl->gl_holders));
 	gfs2_assert_warn(sdp, !(ret & LM_OUT_ASYNC));
@@ -783,7 +831,6 @@ static void xmote_bh(struct gfs2_glock *gl, unsigned int ret)
 		} else {
 			spin_lock(&gl->gl_spin);
 			if (gl->gl_state != gl->gl_demote_state) {
-				gl->gl_req_bh = NULL;
 				spin_unlock(&gl->gl_spin);
 				gfs2_glock_drop_th(gl);
 				gfs2_glock_put(gl);
@@ -825,7 +872,6 @@ out:
 	if (op_done) {
 		spin_lock(&gl->gl_spin);
 		gl->gl_req_gh = NULL;
-		gl->gl_req_bh = NULL;
 		clear_bit(GLF_LOCK, &gl->gl_flags);
 		spin_unlock(&gl->gl_spin);
 	}
@@ -864,7 +910,6 @@ static void gfs2_glock_xmote_th(struct gfs2_glock *gl, struct gfs2_holder *gh)
 	gfs2_assert_warn(sdp, state != gl->gl_state);
 
 	gfs2_glock_hold(gl);
-	gl->gl_req_bh = xmote_bh;
 
 	lck_ret = gfs2_lm_lock(sdp, gl->gl_lock, gl->gl_state, state, lck_flags);
 
@@ -878,51 +923,6 @@ static void gfs2_glock_xmote_th(struct gfs2_glock *gl, struct gfs2_holder *gh)
 }
 
 /**
- * drop_bh - Called after a lock module unlock completes
- * @gl: the glock
- * @ret: the return status
- *
- * Doesn't wake up the process waiting on the struct gfs2_holder (if any)
- * Doesn't drop the reference on the glock the top half took out
- *
- */
-
-static void drop_bh(struct gfs2_glock *gl, unsigned int ret)
-{
-	struct gfs2_sbd *sdp = gl->gl_sbd;
-	const struct gfs2_glock_operations *glops = gl->gl_ops;
-	struct gfs2_holder *gh = gl->gl_req_gh;
-
-	gfs2_assert_warn(sdp, test_bit(GLF_LOCK, &gl->gl_flags));
-	gfs2_assert_warn(sdp, list_empty(&gl->gl_holders));
-	gfs2_assert_warn(sdp, !ret);
-
-	state_change(gl, LM_ST_UNLOCKED);
-
-	if (glops->go_inval)
-		glops->go_inval(gl, DIO_METADATA);
-
-	if (gh) {
-		spin_lock(&gl->gl_spin);
-		list_del_init(&gh->gh_list);
-		gh->gh_error = 0;
-		spin_unlock(&gl->gl_spin);
-	}
-
-	spin_lock(&gl->gl_spin);
-	gfs2_demote_wake(gl);
-	gl->gl_req_gh = NULL;
-	gl->gl_req_bh = NULL;
-	clear_bit(GLF_LOCK, &gl->gl_flags);
-	spin_unlock(&gl->gl_spin);
-
-	gfs2_glock_put(gl);
-
-	if (gh)
-		gfs2_holder_wake(gh);
-}
-
-/**
  * gfs2_glock_drop_th - call into the lock module to unlock a lock
  * @gl: the glock
  *
@@ -942,7 +942,6 @@ static void gfs2_glock_drop_th(struct gfs2_glock *gl)
 	gfs2_assert_warn(sdp, gl->gl_state != LM_ST_UNLOCKED);
 
 	gfs2_glock_hold(gl);
-	gl->gl_req_bh = drop_bh;
 
 	ret = gfs2_lm_unlock(sdp, gl->gl_lock, gl->gl_state);
 
@@ -971,8 +970,7 @@ static void do_cancels(struct gfs2_holder *gh)
 	while (gl->gl_req_gh != gh &&
 	       !test_bit(HIF_HOLDER, &gh->gh_iflags) &&
 	       !list_empty(&gh->gh_list)) {
-		if (gl->gl_req_bh && !(gl->gl_req_gh &&
-				     (gl->gl_req_gh->gh_flags & GL_NOCANCEL))) {
+		if (!(gl->gl_req_gh && (gl->gl_req_gh->gh_flags & GL_NOCANCEL))) {
 			spin_unlock(&gl->gl_spin);
 			gfs2_lm_cancel(gl->gl_sbd, gl->gl_lock);
 			msleep(100);
@@ -1042,7 +1040,6 @@ static int glock_wait_internal(struct gfs2_holder *gh)
 
 		spin_lock(&gl->gl_spin);
 		gl->gl_req_gh = NULL;
-		gl->gl_req_bh = NULL;
 		clear_bit(GLF_LOCK, &gl->gl_flags);
 		run_queue(gl);
 		spin_unlock(&gl->gl_spin);
@@ -1535,8 +1532,7 @@ void gfs2_glock_cb(void *cb_data, unsigned int type, void *data)
 		gl = gfs2_glock_find(sdp, &async->lc_name);
 		if (gfs2_assert_warn(sdp, gl))
 			return;
-		if (!gfs2_assert_warn(sdp, gl->gl_req_bh))
-			gl->gl_req_bh(gl, async->lc_ret);
+		xmote_bh(gl, async->lc_ret);
 		if (queue_delayed_work(glock_workqueue, &gl->gl_work, 0) == 0)
 			gfs2_glock_put(gl);
 		up_read(&gfs2_umount_flush_sem);
@@ -1898,7 +1894,6 @@ static int dump_glock(struct glock_iter *gi, struct gfs2_glock *gl)
 		print_dbg(gi, "  gl_owner = -1\n");
 	print_dbg(gi, "  gl_ip = %lu\n", gl->gl_ip);
 	print_dbg(gi, "  req_gh = %s\n", (gl->gl_req_gh) ? "yes" : "no");
-	print_dbg(gi, "  req_bh = %s\n", (gl->gl_req_bh) ? "yes" : "no");
 	print_dbg(gi, "  lvb_count = %d\n", atomic_read(&gl->gl_lvb_count));
 	print_dbg(gi, "  object = %s\n", (gl->gl_object) ? "yes" : "no");
 	print_dbg(gi, "  reclaim = %s\n",
diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h
index fe14f6a..65aa46a 100644
--- a/fs/gfs2/incore.h
+++ b/fs/gfs2/incore.h
@@ -191,7 +191,6 @@ struct gfs2_glock {
 	const struct gfs2_glock_operations *gl_ops;
 
 	struct gfs2_holder *gl_req_gh;
-	gfs2_glop_bh_t gl_req_bh;
 
 	void *gl_lock;
 	char *gl_lvb;
-- 
1.5.1.2

[Cluster-devel] [PATCH 17/48] [GFS2] Remove lm.[ch] and distribute content

[swhiteho]

From: Steven Whitehouse <swhiteho@redhat.com>

The functions in lm.c were just wrappers which were mostly
only used in one other file. By moving the functions to
the files where they are being used, they can be marked
static and also this will usually result in them being inlined
since they are often only used from one point in the code.

A couple of really trivial functions have been inlined by hand
into the function which called them as it makes the code clearer
to do that.

We also gain from one fewer function call in the glock lock and
unlock paths.

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

diff --git a/fs/gfs2/Makefile b/fs/gfs2/Makefile
index 8fff110..e2350df 100644
--- a/fs/gfs2/Makefile
+++ b/fs/gfs2/Makefile
@@ -1,6 +1,6 @@
 obj-$(CONFIG_GFS2_FS) += gfs2.o
 gfs2-y := acl.o bmap.o daemon.o dir.o eaops.o eattr.o glock.o \
-	glops.o inode.o lm.o log.o lops.o locking.o main.o meta_io.o \
+	glops.o inode.o log.o lops.o locking.o main.o meta_io.o \
 	mount.o ops_address.o ops_dentry.o ops_export.o ops_file.o \
 	ops_fstype.o ops_inode.o ops_super.o quota.o \
 	recovery.o rgrp.o super.o sys.o trans.o util.o
diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c
index a8387e0..611f84d 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -35,7 +35,6 @@
 #include "glock.h"
 #include "glops.h"
 #include "inode.h"
-#include "lm.h"
 #include "lops.h"
 #include "meta_io.h"
 #include "quota.h"
@@ -183,7 +182,8 @@ static void glock_free(struct gfs2_glock *gl)
 	struct gfs2_sbd *sdp = gl->gl_sbd;
 	struct inode *aspace = gl->gl_aspace;
 
-	gfs2_lm_put_lock(sdp, gl->gl_lock);
+	if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
+		sdp->sd_lockstruct.ls_ops->lm_put_lock(gl->gl_lock);
 
 	if (aspace)
 		gfs2_aspace_put(aspace);
@@ -293,6 +293,16 @@ static void glock_work_func(struct work_struct *work)
 	gfs2_glock_put(gl);
 }
 
+static int gfs2_lm_get_lock(struct gfs2_sbd *sdp, struct lm_lockname *name,
+		     void **lockp)
+{
+	int error = -EIO;
+	if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
+		error = sdp->sd_lockstruct.ls_ops->lm_get_lock(
+				sdp->sd_lockstruct.ls_lockspace, name, lockp);
+	return error;
+}
+
 /**
  * gfs2_glock_get() - Get a glock, or create one if one doesn't exist
  * @sdp: The GFS2 superblock
@@ -882,6 +892,17 @@ out:
 		gfs2_holder_wake(gh);
 }
 
+static unsigned int gfs2_lm_lock(struct gfs2_sbd *sdp, void *lock,
+				 unsigned int cur_state, unsigned int req_state,
+				 unsigned int flags)
+{
+	int ret = 0;
+	if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
+		ret = sdp->sd_lockstruct.ls_ops->lm_lock(lock, cur_state,
+							 req_state, flags);
+	return ret;
+}
+
 /**
  * gfs2_glock_xmote_th - Call into the lock module to acquire or change a glock
  * @gl: The glock in question
@@ -922,6 +943,15 @@ static void gfs2_glock_xmote_th(struct gfs2_glock *gl, struct gfs2_holder *gh)
 		xmote_bh(gl, lck_ret);
 }
 
+static unsigned int gfs2_lm_unlock(struct gfs2_sbd *sdp, void *lock,
+				   unsigned int cur_state)
+{
+	int ret = 0;
+	if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
+		ret =  sdp->sd_lockstruct.ls_ops->lm_unlock(lock, cur_state);
+	return ret;
+}
+
 /**
  * gfs2_glock_drop_th - call into the lock module to unlock a lock
  * @gl: the glock
@@ -964,6 +994,7 @@ static void gfs2_glock_drop_th(struct gfs2_glock *gl)
 static void do_cancels(struct gfs2_holder *gh)
 {
 	struct gfs2_glock *gl = gh->gh_gl;
+	struct gfs2_sbd *sdp = gl->gl_sbd;
 
 	spin_lock(&gl->gl_spin);
 
@@ -972,7 +1003,8 @@ static void do_cancels(struct gfs2_holder *gh)
 	       !list_empty(&gh->gh_list)) {
 		if (!(gl->gl_req_gh && (gl->gl_req_gh->gh_flags & GL_NOCANCEL))) {
 			spin_unlock(&gl->gl_spin);
-			gfs2_lm_cancel(gl->gl_sbd, gl->gl_lock);
+			if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
+				sdp->sd_lockstruct.ls_ops->lm_cancel(gl->gl_lock);
 			msleep(100);
 			spin_lock(&gl->gl_spin);
 		} else {
@@ -1426,6 +1458,14 @@ void gfs2_glock_dq_uninit_m(unsigned int num_gh, struct gfs2_holder *ghs)
 		gfs2_glock_dq_uninit(&ghs[x]);
 }
 
+static int gfs2_lm_hold_lvb(struct gfs2_sbd *sdp, void *lock, char **lvbp)
+{
+	int error = -EIO;
+	if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
+		error = sdp->sd_lockstruct.ls_ops->lm_hold_lvb(lock, lvbp);
+	return error;
+}
+
 /**
  * gfs2_lvb_hold - attach a LVB from a glock
  * @gl: The glock in question
@@ -1461,12 +1501,15 @@ int gfs2_lvb_hold(struct gfs2_glock *gl)
 
 void gfs2_lvb_unhold(struct gfs2_glock *gl)
 {
+	struct gfs2_sbd *sdp = gl->gl_sbd;
+
 	gfs2_glock_hold(gl);
 	gfs2_glmutex_lock(gl);
 
 	gfs2_assert(gl->gl_sbd, atomic_read(&gl->gl_lvb_count) > 0);
 	if (atomic_dec_and_test(&gl->gl_lvb_count)) {
-		gfs2_lm_unhold_lvb(gl->gl_sbd, gl->gl_lock, gl->gl_lvb);
+		if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
+			sdp->sd_lockstruct.ls_ops->lm_unhold_lvb(gl->gl_lock, gl->gl_lvb);
 		gl->gl_lvb = NULL;
 		gfs2_glock_put(gl);
 	}
diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h
index 65aa46a..8dee467 100644
--- a/fs/gfs2/incore.h
+++ b/fs/gfs2/incore.h
@@ -211,6 +211,8 @@ struct gfs2_glock {
 	struct delayed_work gl_work;
 };
 
+#define GFS2_MIN_LVB_SIZE 32	/* Min size of LVB that gfs2 supports */
+
 struct gfs2_alloc {
 	/* Quota stuff */
 
diff --git a/fs/gfs2/lm.c b/fs/gfs2/lm.c
deleted file mode 100644
index cfcc39b..0000000
--- a/fs/gfs2/lm.c
+++ /dev/null
@@ -1,210 +0,0 @@
-/*
- * Copyright (C) Sistina Software, Inc.  1997-2003 All rights reserved.
- * Copyright (C) 2004-2006 Red Hat, Inc.  All rights reserved.
- *
- * This copyrighted material is made available to anyone wishing to use,
- * modify, copy, or redistribute it subject to the terms and conditions
- * of the GNU General Public License version 2.
- */
-
-#include <linux/slab.h>
-#include <linux/spinlock.h>
-#include <linux/completion.h>
-#include <linux/buffer_head.h>
-#include <linux/delay.h>
-#include <linux/gfs2_ondisk.h>
-#include <linux/lm_interface.h>
-
-#include "gfs2.h"
-#include "incore.h"
-#include "glock.h"
-#include "lm.h"
-#include "super.h"
-#include "util.h"
-
-/**
- * gfs2_lm_mount - mount a locking protocol
- * @sdp: the filesystem
- * @args: mount arguements
- * @silent: if 1, don't complain if the FS isn't a GFS2 fs
- *
- * Returns: errno
- */
-
-int gfs2_lm_mount(struct gfs2_sbd *sdp, int silent)
-{
-	char *proto = sdp->sd_proto_name;
-	char *table = sdp->sd_table_name;
-	int flags = 0;
-	int error;
-
-	if (sdp->sd_args.ar_spectator)
-		flags |= LM_MFLAG_SPECTATOR;
-
-	fs_info(sdp, "Trying to join cluster \"%s\", \"%s\"\n", proto, table);
-
-	error = gfs2_mount_lockproto(proto, table, sdp->sd_args.ar_hostdata,
-				     gfs2_glock_cb, sdp,
-				     GFS2_MIN_LVB_SIZE, flags,
-				     &sdp->sd_lockstruct, &sdp->sd_kobj);
-	if (error) {
-		fs_info(sdp, "can't mount proto=%s, table=%s, hostdata=%s\n",
-			proto, table, sdp->sd_args.ar_hostdata);
-		goto out;
-	}
-
-	if (gfs2_assert_warn(sdp, sdp->sd_lockstruct.ls_lockspace) ||
-	    gfs2_assert_warn(sdp, sdp->sd_lockstruct.ls_ops) ||
-	    gfs2_assert_warn(sdp, sdp->sd_lockstruct.ls_lvb_size >=
-				  GFS2_MIN_LVB_SIZE)) {
-		gfs2_unmount_lockproto(&sdp->sd_lockstruct);
-		goto out;
-	}
-
-	if (sdp->sd_args.ar_spectator)
-		snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.s", table);
-	else
-		snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.%u", table,
-			 sdp->sd_lockstruct.ls_jid);
-
-	fs_info(sdp, "Joined cluster. Now mounting FS...\n");
-
-	if ((sdp->sd_lockstruct.ls_flags & LM_LSFLAG_LOCAL) &&
-	    !sdp->sd_args.ar_ignore_local_fs) {
-		sdp->sd_args.ar_localflocks = 1;
-		sdp->sd_args.ar_localcaching = 1;
-	}
-
-out:
-	return error;
-}
-
-void gfs2_lm_others_may_mount(struct gfs2_sbd *sdp)
-{
-	if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
-		sdp->sd_lockstruct.ls_ops->lm_others_may_mount(
-					sdp->sd_lockstruct.ls_lockspace);
-}
-
-void gfs2_lm_unmount(struct gfs2_sbd *sdp)
-{
-	if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
-		gfs2_unmount_lockproto(&sdp->sd_lockstruct);
-}
-
-int gfs2_lm_withdraw(struct gfs2_sbd *sdp, char *fmt, ...)
-{
-	va_list args;
-
-	if (test_and_set_bit(SDF_SHUTDOWN, &sdp->sd_flags))
-		return 0;
-
-	va_start(args, fmt);
-	vprintk(fmt, args);
-	va_end(args);
-
-	fs_err(sdp, "about to withdraw this file system\n");
-	BUG_ON(sdp->sd_args.ar_debug);
-
-	fs_err(sdp, "telling LM to withdraw\n");
-	gfs2_withdraw_lockproto(&sdp->sd_lockstruct);
-	fs_err(sdp, "withdrawn\n");
-	dump_stack();
-
-	return -1;
-}
-
-int gfs2_lm_get_lock(struct gfs2_sbd *sdp, struct lm_lockname *name,
-		     void **lockp)
-{
-	int error = -EIO;
-	if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
-		error = sdp->sd_lockstruct.ls_ops->lm_get_lock(
-				sdp->sd_lockstruct.ls_lockspace, name, lockp);
-	return error;
-}
-
-void gfs2_lm_put_lock(struct gfs2_sbd *sdp, void *lock)
-{
-	if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
-		sdp->sd_lockstruct.ls_ops->lm_put_lock(lock);
-}
-
-unsigned int gfs2_lm_lock(struct gfs2_sbd *sdp, void *lock,
-			  unsigned int cur_state, unsigned int req_state,
-			  unsigned int flags)
-{
-	int ret = 0;
-	if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
-		ret = sdp->sd_lockstruct.ls_ops->lm_lock(lock, cur_state,
-							 req_state, flags);
-	return ret;
-}
-
-unsigned int gfs2_lm_unlock(struct gfs2_sbd *sdp, void *lock,
-			    unsigned int cur_state)
-{
-	int ret = 0;
-	if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
-		ret =  sdp->sd_lockstruct.ls_ops->lm_unlock(lock, cur_state);
-	return ret;
-}
-
-void gfs2_lm_cancel(struct gfs2_sbd *sdp, void *lock)
-{
-	if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
-		sdp->sd_lockstruct.ls_ops->lm_cancel(lock);
-}
-
-int gfs2_lm_hold_lvb(struct gfs2_sbd *sdp, void *lock, char **lvbp)
-{
-	int error = -EIO;
-	if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
-		error = sdp->sd_lockstruct.ls_ops->lm_hold_lvb(lock, lvbp);
-	return error;
-}
-
-void gfs2_lm_unhold_lvb(struct gfs2_sbd *sdp, void *lock, char *lvb)
-{
-	if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
-		sdp->sd_lockstruct.ls_ops->lm_unhold_lvb(lock, lvb);
-}
-
-int gfs2_lm_plock_get(struct gfs2_sbd *sdp, struct lm_lockname *name,
-		      struct file *file, struct file_lock *fl)
-{
-	int error = -EIO;
-	if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
-		error = sdp->sd_lockstruct.ls_ops->lm_plock_get(
-				sdp->sd_lockstruct.ls_lockspace, name, file, fl);
-	return error;
-}
-
-int gfs2_lm_plock(struct gfs2_sbd *sdp, struct lm_lockname *name,
-		  struct file *file, int cmd, struct file_lock *fl)
-{
-	int error = -EIO;
-	if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
-		error = sdp->sd_lockstruct.ls_ops->lm_plock(
-				sdp->sd_lockstruct.ls_lockspace, name, file, cmd, fl);
-	return error;
-}
-
-int gfs2_lm_punlock(struct gfs2_sbd *sdp, struct lm_lockname *name,
-		    struct file *file, struct file_lock *fl)
-{
-	int error = -EIO;
-	if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
-		error = sdp->sd_lockstruct.ls_ops->lm_punlock(
-				sdp->sd_lockstruct.ls_lockspace, name, file, fl);
-	return error;
-}
-
-void gfs2_lm_recovery_done(struct gfs2_sbd *sdp, unsigned int jid,
-			   unsigned int message)
-{
-	if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
-		sdp->sd_lockstruct.ls_ops->lm_recovery_done(
-			sdp->sd_lockstruct.ls_lockspace, jid, message);
-}
-
diff --git a/fs/gfs2/lm.h b/fs/gfs2/lm.h
deleted file mode 100644
index 21cdc30..0000000
--- a/fs/gfs2/lm.h
+++ /dev/null
@@ -1,42 +0,0 @@
-/*
- * Copyright (C) Sistina Software, Inc.  1997-2003 All rights reserved.
- * Copyright (C) 2004-2006 Red Hat, Inc.  All rights reserved.
- *
- * This copyrighted material is made available to anyone wishing to use,
- * modify, copy, or redistribute it subject to the terms and conditions
- * of the GNU General Public License version 2.
- */
-
-#ifndef __LM_DOT_H__
-#define __LM_DOT_H__
-
-struct gfs2_sbd;
-
-#define GFS2_MIN_LVB_SIZE 32
-
-int gfs2_lm_mount(struct gfs2_sbd *sdp, int silent);
-void gfs2_lm_others_may_mount(struct gfs2_sbd *sdp);
-void gfs2_lm_unmount(struct gfs2_sbd *sdp);
-int gfs2_lm_withdraw(struct gfs2_sbd *sdp, char *fmt, ...)
-				__attribute__ ((format(printf, 2, 3)));
-int gfs2_lm_get_lock(struct gfs2_sbd *sdp, struct lm_lockname *name,
-		     void **lockp);
-void gfs2_lm_put_lock(struct gfs2_sbd *sdp, void *lock);
-unsigned int gfs2_lm_lock(struct gfs2_sbd *sdp, void *lock,
-			 unsigned int cur_state, unsigned int req_state,
-			 unsigned int flags);
-unsigned int gfs2_lm_unlock(struct gfs2_sbd *sdp, void *lock,
-			   unsigned int cur_state);
-void gfs2_lm_cancel(struct gfs2_sbd *sdp, void *lock);
-int gfs2_lm_hold_lvb(struct gfs2_sbd *sdp, void *lock, char **lvbp);
-void gfs2_lm_unhold_lvb(struct gfs2_sbd *sdp, void *lock, char *lvb);
-int gfs2_lm_plock_get(struct gfs2_sbd *sdp, struct lm_lockname *name,
-		      struct file *file, struct file_lock *fl);
-int gfs2_lm_plock(struct gfs2_sbd *sdp, struct lm_lockname *name,
-		  struct file *file, int cmd, struct file_lock *fl);
-int gfs2_lm_punlock(struct gfs2_sbd *sdp, struct lm_lockname *name,
-		    struct file *file, struct file_lock *fl);
-void gfs2_lm_recovery_done(struct gfs2_sbd *sdp, unsigned int jid,
-			   unsigned int message);
-
-#endif /* __LM_DOT_H__ */
diff --git a/fs/gfs2/ops_file.c b/fs/gfs2/ops_file.c
index f4842f2..f97a8b8 100644
--- a/fs/gfs2/ops_file.c
+++ b/fs/gfs2/ops_file.c
@@ -30,7 +30,6 @@
 #include "glock.h"
 #include "glops.h"
 #include "inode.h"
-#include "lm.h"
 #include "log.h"
 #include "meta_io.h"
 #include "quota.h"
@@ -596,6 +595,36 @@ static int gfs2_setlease(struct file *file, long arg, struct file_lock **fl)
 	return generic_setlease(file, arg, fl);
 }
 
+static int gfs2_lm_plock_get(struct gfs2_sbd *sdp, struct lm_lockname *name,
+		      struct file *file, struct file_lock *fl)
+{
+	int error = -EIO;
+	if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
+		error = sdp->sd_lockstruct.ls_ops->lm_plock_get(
+				sdp->sd_lockstruct.ls_lockspace, name, file, fl);
+	return error;
+}
+
+static int gfs2_lm_plock(struct gfs2_sbd *sdp, struct lm_lockname *name,
+		  struct file *file, int cmd, struct file_lock *fl)
+{
+	int error = -EIO;
+	if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
+		error = sdp->sd_lockstruct.ls_ops->lm_plock(
+				sdp->sd_lockstruct.ls_lockspace, name, file, cmd, fl);
+	return error;
+}
+
+static int gfs2_lm_punlock(struct gfs2_sbd *sdp, struct lm_lockname *name,
+		    struct file *file, struct file_lock *fl)
+{
+	int error = -EIO;
+	if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
+		error = sdp->sd_lockstruct.ls_ops->lm_punlock(
+				sdp->sd_lockstruct.ls_lockspace, name, file, fl);
+	return error;
+}
+
 /**
  * gfs2_lock - acquire/release a posix lock on a file
  * @file: the file pointer
diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c
index 4bee6aa..5b6a345 100644
--- a/fs/gfs2/ops_fstype.c
+++ b/fs/gfs2/ops_fstype.c
@@ -26,7 +26,6 @@
 #include "glock.h"
 #include "glops.h"
 #include "inode.h"
-#include "lm.h"
 #include "mount.h"
 #include "ops_fstype.h"
 #include "ops_dentry.h"
@@ -363,6 +362,13 @@ static int map_journal_extents(struct gfs2_sbd *sdp)
 	return rc;
 }
 
+static void gfs2_lm_others_may_mount(struct gfs2_sbd *sdp)
+{
+	if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
+		sdp->sd_lockstruct.ls_ops->lm_others_may_mount(
+					sdp->sd_lockstruct.ls_lockspace);
+}
+
 static int init_journal(struct gfs2_sbd *sdp, int undo)
 {
 	struct gfs2_holder ji_gh;
@@ -705,6 +711,69 @@ fail:
 }
 
 /**
+ * gfs2_lm_mount - mount a locking protocol
+ * @sdp: the filesystem
+ * @args: mount arguements
+ * @silent: if 1, don't complain if the FS isn't a GFS2 fs
+ *
+ * Returns: errno
+ */
+
+static int gfs2_lm_mount(struct gfs2_sbd *sdp, int silent)
+{
+	char *proto = sdp->sd_proto_name;
+	char *table = sdp->sd_table_name;
+	int flags = 0;
+	int error;
+
+	if (sdp->sd_args.ar_spectator)
+		flags |= LM_MFLAG_SPECTATOR;
+
+	fs_info(sdp, "Trying to join cluster \"%s\", \"%s\"\n", proto, table);
+
+	error = gfs2_mount_lockproto(proto, table, sdp->sd_args.ar_hostdata,
+				     gfs2_glock_cb, sdp,
+				     GFS2_MIN_LVB_SIZE, flags,
+				     &sdp->sd_lockstruct, &sdp->sd_kobj);
+	if (error) {
+		fs_info(sdp, "can't mount proto=%s, table=%s, hostdata=%s\n",
+			proto, table, sdp->sd_args.ar_hostdata);
+		goto out;
+	}
+
+	if (gfs2_assert_warn(sdp, sdp->sd_lockstruct.ls_lockspace) ||
+	    gfs2_assert_warn(sdp, sdp->sd_lockstruct.ls_ops) ||
+	    gfs2_assert_warn(sdp, sdp->sd_lockstruct.ls_lvb_size >=
+				  GFS2_MIN_LVB_SIZE)) {
+		gfs2_unmount_lockproto(&sdp->sd_lockstruct);
+		goto out;
+	}
+
+	if (sdp->sd_args.ar_spectator)
+		snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.s", table);
+	else
+		snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.%u", table,
+			 sdp->sd_lockstruct.ls_jid);
+
+	fs_info(sdp, "Joined cluster. Now mounting FS...\n");
+
+	if ((sdp->sd_lockstruct.ls_flags & LM_LSFLAG_LOCAL) &&
+	    !sdp->sd_args.ar_ignore_local_fs) {
+		sdp->sd_args.ar_localflocks = 1;
+		sdp->sd_args.ar_localcaching = 1;
+	}
+
+out:
+	return error;
+}
+
+void gfs2_lm_unmount(struct gfs2_sbd *sdp)
+{
+	if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
+		gfs2_unmount_lockproto(&sdp->sd_lockstruct);
+}
+
+/**
  * fill_super - Read in superblock
  * @sb: The VFS superblock
  * @data: Mount options
diff --git a/fs/gfs2/ops_super.c b/fs/gfs2/ops_super.c
index 5e52421..2278c68 100644
--- a/fs/gfs2/ops_super.c
+++ b/fs/gfs2/ops_super.c
@@ -25,7 +25,6 @@
 #include "incore.h"
 #include "glock.h"
 #include "inode.h"
-#include "lm.h"
 #include "log.h"
 #include "mount.h"
 #include "ops_super.h"
diff --git a/fs/gfs2/recovery.c b/fs/gfs2/recovery.c
index 6fb07d6..b17d3b8 100644
--- a/fs/gfs2/recovery.c
+++ b/fs/gfs2/recovery.c
@@ -20,7 +20,6 @@
 #include "bmap.h"
 #include "glock.h"
 #include "glops.h"
-#include "lm.h"
 #include "lops.h"
 #include "meta_io.h"
 #include "recovery.h"
@@ -425,6 +424,16 @@ static int clean_journal(struct gfs2_jdesc *jd, struct gfs2_log_header_host *hea
 	return error;
 }
 
+
+static void gfs2_lm_recovery_done(struct gfs2_sbd *sdp, unsigned int jid,
+				  unsigned int message)
+{
+	if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
+		sdp->sd_lockstruct.ls_ops->lm_recovery_done(
+			sdp->sd_lockstruct.ls_lockspace, jid, message);
+}
+
+
 /**
  * gfs2_recover_journal - recovery a given journal
  * @jd: the struct gfs2_jdesc describing the journal
diff --git a/fs/gfs2/super.h b/fs/gfs2/super.h
index 60a870e..44361ec 100644
--- a/fs/gfs2/super.h
+++ b/fs/gfs2/super.h
@@ -17,6 +17,7 @@ void gfs2_tune_init(struct gfs2_tune *gt);
 int gfs2_check_sb(struct gfs2_sbd *sdp, struct gfs2_sb_host *sb, int silent);
 int gfs2_read_sb(struct gfs2_sbd *sdp, struct gfs2_glock *gl, int silent);
 int gfs2_read_super(struct gfs2_sbd *sdp, sector_t sector);
+void gfs2_lm_unmount(struct gfs2_sbd *sdp);
 
 static inline unsigned int gfs2_jindex_size(struct gfs2_sbd *sdp)
 {
diff --git a/fs/gfs2/sys.c b/fs/gfs2/sys.c
index eaa3b7b..cc35ec8 100644
--- a/fs/gfs2/sys.c
+++ b/fs/gfs2/sys.c
@@ -20,7 +20,6 @@
 
 #include "gfs2.h"
 #include "incore.h"
-#include "lm.h"
 #include "sys.h"
 #include "super.h"
 #include "glock.h"
diff --git a/fs/gfs2/util.c b/fs/gfs2/util.c
index fe9c28e..d31e355 100644
--- a/fs/gfs2/util.c
+++ b/fs/gfs2/util.c
@@ -19,7 +19,6 @@
 #include "gfs2.h"
 #include "incore.h"
 #include "glock.h"
-#include "lm.h"
 #include "util.h"
 
 struct kmem_cache *gfs2_glock_cachep __read_mostly;
@@ -33,6 +32,28 @@ void gfs2_assert_i(struct gfs2_sbd *sdp)
 	       sdp->sd_fsname);
 }
 
+int gfs2_lm_withdraw(struct gfs2_sbd *sdp, char *fmt, ...)
+{
+	va_list args;
+
+	if (test_and_set_bit(SDF_SHUTDOWN, &sdp->sd_flags))
+		return 0;
+
+	va_start(args, fmt);
+	vprintk(fmt, args);
+	va_end(args);
+
+	fs_err(sdp, "about to withdraw this file system\n");
+	BUG_ON(sdp->sd_args.ar_debug);
+
+	fs_err(sdp, "telling LM to withdraw\n");
+	gfs2_withdraw_lockproto(&sdp->sd_lockstruct);
+	fs_err(sdp, "withdrawn\n");
+	dump_stack();
+
+	return -1;
+}
+
 /**
  * gfs2_assert_withdraw_i - Cause the machine to withdraw if @assertion is false
  * Returns: -1 if this call withdrew the machine,
diff --git a/fs/gfs2/util.h b/fs/gfs2/util.h
index ac0c567..509c5d6 100644
--- a/fs/gfs2/util.h
+++ b/fs/gfs2/util.h
@@ -164,6 +164,7 @@ gfs2_tune_get_i(&(sdp)->sd_tune, &(sdp)->sd_tune.field)
 
 void gfs2_icbit_munge(struct gfs2_sbd *sdp, unsigned char **bitmap,
 		      unsigned int bit, int new_value);
+int gfs2_lm_withdraw(struct gfs2_sbd *sdp, char *fmt, ...);
 
 #endif /* __UTIL_DOT_H__ */
 
-- 
1.5.1.2

[Cluster-devel] [PATCH 18/48] [GFS2] Remove rgrp and glock version numbers

[swhiteho]

From: Bob Peterson <rpeterso@redhat.com>

This patch further reduces GFS2's memory requirements by
eliminating the 64-bit version number fields in lieu of
a couple bits.

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 611f84d..d00dc37 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -1,6 +1,6 @@
 /*
  * Copyright (C) Sistina Software, Inc.  1997-2003 All rights reserved.
- * Copyright (C) 2004-2007 Red Hat, Inc.  All rights reserved.
+ * Copyright (C) 2004-2008 Red Hat, Inc.  All rights reserved.
  *
  * This copyrighted material is made available to anyone wishing to use,
  * modify, copy, or redistribute it subject to the terms and conditions
@@ -348,7 +348,6 @@ int gfs2_glock_get(struct gfs2_sbd *sdp, u64 number,
 	gl->gl_ip = 0;
 	gl->gl_ops = glops;
 	gl->gl_req_gh = NULL;
-	gl->gl_vn = 0;
 	gl->gl_stamp = jiffies;
 	gl->gl_tchange = jiffies;
 	gl->gl_object = NULL;
diff --git a/fs/gfs2/glops.c b/fs/gfs2/glops.c
index c663b7a..d31bada 100644
--- a/fs/gfs2/glops.c
+++ b/fs/gfs2/glops.c
@@ -1,6 +1,6 @@
 /*
  * Copyright (C) Sistina Software, Inc.  1997-2003 All rights reserved.
- * Copyright (C) 2004-2006 Red Hat, Inc.  All rights reserved.
+ * Copyright (C) 2004-2008 Red Hat, Inc.  All rights reserved.
  *
  * This copyrighted material is made available to anyone wishing to use,
  * modify, copy, or redistribute it subject to the terms and conditions
@@ -126,7 +126,13 @@ static void meta_go_inval(struct gfs2_glock *gl, int flags)
 		return;
 
 	gfs2_meta_inval(gl);
-	gl->gl_vn++;
+	if (gl->gl_object == GFS2_I(gl->gl_sbd->sd_rindex))
+		gl->gl_sbd->sd_rindex_uptodate = 0;
+	else if (gl->gl_ops == &gfs2_rgrp_glops && gl->gl_object) {
+		struct gfs2_rgrpd *rgd = (struct gfs2_rgrpd *)gl->gl_object;
+
+		rgd->rd_flags &= ~GFS2_RDF_UPTODATE;
+	}
 }
 
 /**
diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h
index 8dee467..e9c58dc 100644
--- a/fs/gfs2/incore.h
+++ b/fs/gfs2/incore.h
@@ -85,7 +85,6 @@ struct gfs2_rgrpd {
 	u32 rd_data;			/* num of data blocks in rgrp */
 	u32 rd_bitbytes;		/* number of bytes in data bitmaps */
 	struct gfs2_rgrp_host rd_rg;
-	u64 rd_rg_vn;
 	struct gfs2_bitmap *rd_bits;
 	unsigned int rd_bh_count;
 	struct mutex rd_mutex;
@@ -97,6 +96,7 @@ struct gfs2_rgrpd {
 	unsigned char rd_flags;
 #define GFS2_RDF_CHECK        0x01      /* Need to check for unlinked inodes */
 #define GFS2_RDF_NOALLOC      0x02      /* rg prohibits allocation */
+#define GFS2_RDF_UPTODATE     0x04      /* rg is up to date */
 };
 
 enum gfs2_state_bits {
@@ -196,7 +196,6 @@ struct gfs2_glock {
 	char *gl_lvb;
 	atomic_t gl_lvb_count;
 
-	u64 gl_vn;
 	unsigned long gl_stamp;
 	unsigned long gl_tchange;
 	void *gl_object;
@@ -533,7 +532,7 @@ struct gfs2_sbd {
 
 	/* Resource group stuff */
 
-	u64 sd_rindex_vn;
+	int sd_rindex_uptodate;
 	spinlock_t sd_rindex_spin;
 	struct mutex sd_rindex_mutex;
 	struct list_head sd_rindex_list;
diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c
index 5b6a345..c4b7a21 100644
--- a/fs/gfs2/ops_fstype.c
+++ b/fs/gfs2/ops_fstype.c
@@ -1,6 +1,6 @@
 /*
  * Copyright (C) Sistina Software, Inc.  1997-2003 All rights reserved.
- * Copyright (C) 2004-2007 Red Hat, Inc.  All rights reserved.
+ * Copyright (C) 2004-2008 Red Hat, Inc.  All rights reserved.
  *
  * This copyrighted material is made available to anyone wishing to use,
  * modify, copy, or redistribute it subject to the terms and conditions
@@ -548,7 +548,7 @@ static int init_inodes(struct gfs2_sbd *sdp, int undo)
 	}
 	ip = GFS2_I(sdp->sd_rindex);
 	set_bit(GLF_STICKY, &ip->i_gl->gl_flags);
-	sdp->sd_rindex_vn = ip->i_gl->gl_vn - 1;
+	sdp->sd_rindex_uptodate = 0;
 
 	/* Read in the quota inode */
 	sdp->sd_quota_inode = gfs2_lookup_simple(sdp->sd_master_dir, "quota");
diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c
index 5fd8710..3f10b1f 100644
--- a/fs/gfs2/rgrp.c
+++ b/fs/gfs2/rgrp.c
@@ -540,7 +540,7 @@ static int read_rindex_entry(struct gfs2_inode *ip,
 		return error;
 
 	rgd->rd_gl->gl_object = rgd;
-	rgd->rd_rg_vn = rgd->rd_gl->gl_vn - 1;
+	rgd->rd_flags &= ~GFS2_RDF_UPTODATE;
 	rgd->rd_flags |= GFS2_RDF_CHECK;
 	return error;
 }
@@ -576,7 +576,7 @@ static int gfs2_ri_update(struct gfs2_inode *ip)
 		}
 	}
 
-	sdp->sd_rindex_vn = ip->i_gl->gl_vn;
+	sdp->sd_rindex_uptodate = 1;
 	return 0;
 }
 
@@ -610,7 +610,7 @@ static int gfs2_ri_update_special(struct gfs2_inode *ip)
 		}
 	}
 
-	sdp->sd_rindex_vn = ip->i_gl->gl_vn;
+	sdp->sd_rindex_uptodate = 1;
 	return 0;
 }
 
@@ -643,9 +643,9 @@ int gfs2_rindex_hold(struct gfs2_sbd *sdp, struct gfs2_holder *ri_gh)
 		return error;
 
 	/* Read new copy from disk if we don't have the latest */
-	if (sdp->sd_rindex_vn != gl->gl_vn) {
+	if (!sdp->sd_rindex_uptodate) {
 		mutex_lock(&sdp->sd_rindex_mutex);
-		if (sdp->sd_rindex_vn != gl->gl_vn) {
+		if (!sdp->sd_rindex_uptodate) {
 			error = gfs2_ri_update(ip);
 			if (error)
 				gfs2_glock_dq_uninit(ri_gh);
@@ -737,9 +737,9 @@ int gfs2_rgrp_bh_get(struct gfs2_rgrpd *rgd)
 		}
 	}
 
-	if (rgd->rd_rg_vn != gl->gl_vn) {
+	if (!(rgd->rd_flags & GFS2_RDF_UPTODATE)) {
 		gfs2_rgrp_in(rgd, (rgd->rd_bits[0].bi_bh)->b_data);
-		rgd->rd_rg_vn = gl->gl_vn;
+		rgd->rd_flags |= GFS2_RDF_UPTODATE;
 	}
 
 	spin_lock(&sdp->sd_rindex_spin);
-- 
1.5.1.2

[Cluster-devel] [PATCH 19/48] [GFS2] Shrink & rename di_depth

[swhiteho]

From: Steven Whitehouse <swhiteho@redhat.com>

This patch forms a pair with the previous patch which shrunk
di_height. Like that patch di_depth is renamed i_depth and moved
into struct gfs2_inode directly. Also the field goes from 16 bits
to 8 bits since it is also limited to a max value which is rather
small (17 in this case). In addition we also now validate the field
against this maximum value when its read in.

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

diff --git a/fs/gfs2/dir.c b/fs/gfs2/dir.c
index 78c236f..081daa9 100644
--- a/fs/gfs2/dir.c
+++ b/fs/gfs2/dir.c
@@ -757,7 +757,7 @@ static struct gfs2_dirent *gfs2_dirent_search(struct inode *inode,
 
 	if (ip->i_di.di_flags & GFS2_DIF_EXHASH) {
 		struct gfs2_leaf *leaf;
-		unsigned hsize = 1 << ip->i_di.di_depth;
+		unsigned hsize = 1 << ip->i_depth;
 		unsigned index;
 		u64 ln;
 		if (hsize * sizeof(u64) != ip->i_di.di_size) {
@@ -765,7 +765,7 @@ static struct gfs2_dirent *gfs2_dirent_search(struct inode *inode,
 			return ERR_PTR(-EIO);
 		}
 
-		index = name->hash >> (32 - ip->i_di.di_depth);
+		index = name->hash >> (32 - ip->i_depth);
 		error = get_first_leaf(ip, index, &bh);
 		if (error)
 			return ERR_PTR(error);
@@ -910,7 +910,7 @@ static int dir_make_exhash(struct inode *inode)
 	dip->i_di.di_flags |= GFS2_DIF_EXHASH;
 
 	for (x = sdp->sd_hash_ptrs, y = -1; x; x >>= 1, y++) ;
-	dip->i_di.di_depth = y;
+	dip->i_depth = y;
 
 	gfs2_dinode_out(dip, dibh->b_data);
 
@@ -941,7 +941,7 @@ static int dir_split_leaf(struct inode *inode, const struct qstr *name)
 	int x, moved = 0;
 	int error;
 
-	index = name->hash >> (32 - dip->i_di.di_depth);
+	index = name->hash >> (32 - dip->i_depth);
 	error = get_leaf_nr(dip, index, &leaf_no);
 	if (error)
 		return error;
@@ -952,7 +952,7 @@ static int dir_split_leaf(struct inode *inode, const struct qstr *name)
 		return error;
 
 	oleaf = (struct gfs2_leaf *)obh->b_data;
-	if (dip->i_di.di_depth == be16_to_cpu(oleaf->lf_depth)) {
+	if (dip->i_depth == be16_to_cpu(oleaf->lf_depth)) {
 		brelse(obh);
 		return 1; /* can't split */
 	}
@@ -967,10 +967,10 @@ static int dir_split_leaf(struct inode *inode, const struct qstr *name)
 	bn = nbh->b_blocknr;
 
 	/*  Compute the start and len of leaf pointers in the hash table.  */
-	len = 1 << (dip->i_di.di_depth - be16_to_cpu(oleaf->lf_depth));
+	len = 1 << (dip->i_depth - be16_to_cpu(oleaf->lf_depth));
 	half_len = len >> 1;
 	if (!half_len) {
-		printk(KERN_WARNING "di_depth %u lf_depth %u index %u\n", dip->i_di.di_depth, be16_to_cpu(oleaf->lf_depth), index);
+		printk(KERN_WARNING "i_depth %u lf_depth %u index %u\n", dip->i_depth, be16_to_cpu(oleaf->lf_depth), index);
 		gfs2_consist_inode(dip);
 		error = -EIO;
 		goto fail_brelse;
@@ -997,7 +997,7 @@ static int dir_split_leaf(struct inode *inode, const struct qstr *name)
 	kfree(lp);
 
 	/*  Compute the divider  */
-	divider = (start + half_len) << (32 - dip->i_di.di_depth);
+	divider = (start + half_len) << (32 - dip->i_depth);
 
 	/*  Copy the entries  */
 	dirent_first(dip, obh, &dent);
@@ -1082,7 +1082,7 @@ static int dir_double_exhash(struct gfs2_inode *dip)
 	int x;
 	int error = 0;
 
-	hsize = 1 << dip->i_di.di_depth;
+	hsize = 1 << dip->i_depth;
 	if (hsize * sizeof(u64) != dip->i_di.di_size) {
 		gfs2_consist_inode(dip);
 		return -EIO;
@@ -1125,7 +1125,7 @@ static int dir_double_exhash(struct gfs2_inode *dip)
 
 	error = gfs2_meta_inode_buffer(dip, &dibh);
 	if (!gfs2_assert_withdraw(sdp, !error)) {
-		dip->i_di.di_depth++;
+		dip->i_depth++;
 		gfs2_dinode_out(dip, dibh->b_data);
 		brelse(dibh);
 	}
@@ -1370,14 +1370,14 @@ static int dir_e_read(struct inode *inode, u64 *offset, void *opaque,
 	int error = 0;
 	unsigned depth = 0;
 
-	hsize = 1 << dip->i_di.di_depth;
+	hsize = 1 << dip->i_depth;
 	if (hsize * sizeof(u64) != dip->i_di.di_size) {
 		gfs2_consist_inode(dip);
 		return -EIO;
 	}
 
 	hash = gfs2_dir_offset2hash(*offset);
-	index = hash >> (32 - dip->i_di.di_depth);
+	index = hash >> (32 - dip->i_depth);
 
 	lp = kmalloc(sdp->sd_hash_bsize, GFP_KERNEL);
 	if (!lp)
@@ -1405,7 +1405,7 @@ static int dir_e_read(struct inode *inode, u64 *offset, void *opaque,
 		if (error)
 			break;
 
-		len = 1 << (dip->i_di.di_depth - depth);
+		len = 1 << (dip->i_depth - depth);
 		index = (index & ~(len - 1)) + len;
 	}
 
@@ -1549,7 +1549,7 @@ static int dir_new_leaf(struct inode *inode, const struct qstr *name)
 	u32 index;
 	u64 bn;
 
-	index = name->hash >> (32 - ip->i_di.di_depth);
+	index = name->hash >> (32 - ip->i_depth);
 	error = get_first_leaf(ip, index, &obh);
 	if (error)
 		return error;
@@ -1641,7 +1641,7 @@ int gfs2_dir_add(struct inode *inode, const struct qstr *name,
 			continue;
 		if (error < 0)
 			break;
-		if (ip->i_di.di_depth < GFS2_DIR_MAX_DEPTH) {
+		if (ip->i_depth < GFS2_DIR_MAX_DEPTH) {
 			error = dir_double_exhash(ip);
 			if (error)
 				break;
@@ -1785,7 +1785,7 @@ static int foreach_leaf(struct gfs2_inode *dip, leaf_call_t lc, void *data)
 	u64 leaf_no;
 	int error = 0;
 
-	hsize = 1 << dip->i_di.di_depth;
+	hsize = 1 << dip->i_depth;
 	if (hsize * sizeof(u64) != dip->i_di.di_size) {
 		gfs2_consist_inode(dip);
 		return -EIO;
@@ -1817,7 +1817,7 @@ static int foreach_leaf(struct gfs2_inode *dip, leaf_call_t lc, void *data)
 			if (error)
 				goto out;
 			leaf = (struct gfs2_leaf *)bh->b_data;
-			len = 1 << (dip->i_di.di_depth - be16_to_cpu(leaf->lf_depth));
+			len = 1 << (dip->i_depth - be16_to_cpu(leaf->lf_depth));
 			brelse(bh);
 
 			error = lc(dip, index, len, leaf_no, data);
diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h
index e9c58dc..9dfdde3 100644
--- a/fs/gfs2/incore.h
+++ b/fs/gfs2/incore.h
@@ -246,7 +246,6 @@ struct gfs2_dinode_host {
 	u64 di_generation;	/* generation number for NFS */
 	u32 di_flags;		/* GFS2_DIF_... */
 	/* These only apply to directories  */
-	u16 di_depth;		/* Number of bits in the table */
 	u32 di_entries;		/* The number of entries in the directory */
 	u64 di_eattr;		/* extended attribute block number */
 };
@@ -267,6 +266,7 @@ struct gfs2_inode {
 
 	struct rw_semaphore i_rw_mutex;
 	u8 i_height;
+	u8 i_depth;
 };
 
 /*
diff --git a/fs/gfs2/inode.c b/fs/gfs2/inode.c
index db5961a..65fdfee 100644
--- a/fs/gfs2/inode.c
+++ b/fs/gfs2/inode.c
@@ -248,7 +248,7 @@ static int gfs2_dinode_in(struct gfs2_inode *ip, const void *buf)
 {
 	struct gfs2_dinode_host *di = &ip->i_di;
 	const struct gfs2_dinode *str = buf;
-	u16 height;
+	u16 height, depth;
 
 	if (unlikely(ip->i_no_addr != be64_to_cpu(str->di_num.no_addr)))
 		goto corrupt;
@@ -293,7 +293,10 @@ static int gfs2_dinode_in(struct gfs2_inode *ip, const void *buf)
 		goto corrupt;
 	ip->i_height = (u8)height;
 
-	di->di_depth = be16_to_cpu(str->di_depth);
+	depth = be16_to_cpu(str->di_depth);
+	if (unlikely(depth > GFS2_DIR_MAX_DEPTH))
+		goto corrupt;
+	ip->i_depth = (u8)depth;
 	di->di_entries = be32_to_cpu(str->di_entries);
 
 	di->di_eattr = be64_to_cpu(str->di_eattr);
@@ -1410,7 +1413,7 @@ void gfs2_dinode_out(const struct gfs2_inode *ip, void *buf)
 	str->di_payload_format = cpu_to_be32(S_ISDIR(ip->i_inode.i_mode) &&
 					     !(ip->i_di.di_flags & GFS2_DIF_EXHASH) ?
 					     GFS2_FORMAT_DE : 0);
-	str->di_depth = cpu_to_be16(di->di_depth);
+	str->di_depth = cpu_to_be16(ip->i_depth);
 	str->di_entries = cpu_to_be32(di->di_entries);
 
 	str->di_eattr = cpu_to_be64(di->di_eattr);
@@ -1436,7 +1439,7 @@ void gfs2_dinode_print(const struct gfs2_inode *ip)
 	       (unsigned long long)di->di_goal_data);
 	printk(KERN_INFO "  di_flags = 0x%.8X\n", di->di_flags);
 	printk(KERN_INFO "  i_height = %u\n", ip->i_height);
-	printk(KERN_INFO "  di_depth = %u\n", di->di_depth);
+	printk(KERN_INFO "  i_depth = %u\n", ip->i_depth);
 	printk(KERN_INFO "  di_entries = %u\n", di->di_entries);
 	printk(KERN_INFO "  di_eattr = %llu\n",
 	       (unsigned long long)di->di_eattr);
-- 
1.5.1.2

[Cluster-devel] [PATCH 20/48] [GFS2] Remove unused counters

[swhiteho]

From: Bob Peterson <rpeterso@redhat.com>

This is kind of trivial in the greater scheme of things, but
this removes three counters that AFAICT are never used.

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

diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h
index 9dfdde3..b67e44b 100644
--- a/fs/gfs2/incore.h
+++ b/fs/gfs2/incore.h
@@ -636,9 +636,6 @@ struct gfs2_sbd {
 
 	/* Counters */
 
-	atomic_t sd_glock_count;
-	atomic_t sd_glock_held_count;
-	atomic_t sd_inode_count;
 	atomic_t sd_reclaimed;
 
 	char sd_fsname[GFS2_FSNAME_LEN];
diff --git a/fs/gfs2/sys.c b/fs/gfs2/sys.c
index cc35ec8..9ab9fc8 100644
--- a/fs/gfs2/sys.c
+++ b/fs/gfs2/sys.c
@@ -327,15 +327,9 @@ static ssize_t name##_show(struct gfs2_sbd *sdp, char *buf)                 \
 }                                                                           \
 static struct counters_attr counters_attr_##name = __ATTR_RO(name)
 
-COUNTERS_ATTR(glock_count,      "%u\n");
-COUNTERS_ATTR(glock_held_count, "%u\n");
-COUNTERS_ATTR(inode_count,      "%u\n");
 COUNTERS_ATTR(reclaimed,        "%u\n");
 
 static struct attribute *counters_attrs[] = {
-	&counters_attr_glock_count.attr,
-	&counters_attr_glock_held_count.attr,
-	&counters_attr_inode_count.attr,
 	&counters_attr_reclaimed.attr,
 	NULL,
 };
-- 
1.5.1.2

[Cluster-devel] [PATCH 21/48] [GFS2] Reduce inode size by merging fields

[swhiteho]

From: Steven Whitehouse <swhiteho@redhat.com>

There were three fields being used to keep track of the location
of the most recently allocated block for each inode. These have
been merged into a single field in order to better keep the
data and metadata for an inode close on disk, and also to reduce
the space required for storage.

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

diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index 2011dd2..30d718b 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -1020,9 +1020,7 @@ static int trunc_end(struct gfs2_inode *ip)
 
 	if (!ip->i_di.di_size) {
 		ip->i_height = 0;
-		ip->i_di.di_goal_meta =
-			ip->i_di.di_goal_data =
-			ip->i_no_addr;
+		ip->i_goal = ip->i_no_addr;
 		gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode));
 	}
 	ip->i_inode.i_mtime = ip->i_inode.i_ctime = CURRENT_TIME;
diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h
index b67e44b..c50dcdf 100644
--- a/fs/gfs2/incore.h
+++ b/fs/gfs2/incore.h
@@ -241,8 +241,6 @@ enum {
 struct gfs2_dinode_host {
 	u64 di_size;		/* number of bytes in file */
 	u64 di_blocks;		/* number of blocks in file */
-	u64 di_goal_meta;	/* rgrp to alloc from next */
-	u64 di_goal_data;	/* data block goal */
 	u64 di_generation;	/* generation number for NFS */
 	u32 di_flags;		/* GFS2_DIF_... */
 	/* These only apply to directories  */
@@ -262,8 +260,7 @@ struct gfs2_inode {
 	struct gfs2_holder i_iopen_gh;
 	struct gfs2_holder i_gh; /* for prepare/commit_write only */
 	struct gfs2_alloc *i_alloc;
-	u64 i_last_rg_alloc;
-
+	u64 i_goal;	/* goal block for allocations */
 	struct rw_semaphore i_rw_mutex;
 	u8 i_height;
 	u8 i_depth;
diff --git a/fs/gfs2/inode.c b/fs/gfs2/inode.c
index 65fdfee..c3fe8aa 100644
--- a/fs/gfs2/inode.c
+++ b/fs/gfs2/inode.c
@@ -282,8 +282,7 @@ static int gfs2_dinode_in(struct gfs2_inode *ip, const void *buf)
 	ip->i_inode.i_ctime.tv_sec = be64_to_cpu(str->di_ctime);
 	ip->i_inode.i_ctime.tv_nsec = be32_to_cpu(str->di_ctime_nsec);
 
-	di->di_goal_meta = be64_to_cpu(str->di_goal_meta);
-	di->di_goal_data = be64_to_cpu(str->di_goal_data);
+	ip->i_goal = be64_to_cpu(str->di_goal_meta);
 	di->di_generation = be64_to_cpu(str->di_generation);
 
 	di->di_flags = be32_to_cpu(str->di_flags);
@@ -1404,8 +1403,8 @@ void gfs2_dinode_out(const struct gfs2_inode *ip, void *buf)
 	str->di_mtime = cpu_to_be64(ip->i_inode.i_mtime.tv_sec);
 	str->di_ctime = cpu_to_be64(ip->i_inode.i_ctime.tv_sec);
 
-	str->di_goal_meta = cpu_to_be64(di->di_goal_meta);
-	str->di_goal_data = cpu_to_be64(di->di_goal_data);
+	str->di_goal_meta = cpu_to_be64(ip->i_goal);
+	str->di_goal_data = cpu_to_be64(ip->i_goal);
 	str->di_generation = cpu_to_be64(di->di_generation);
 
 	str->di_flags = cpu_to_be32(di->di_flags);
@@ -1433,10 +1432,8 @@ void gfs2_dinode_print(const struct gfs2_inode *ip)
 	printk(KERN_INFO "  di_size = %llu\n", (unsigned long long)di->di_size);
 	printk(KERN_INFO "  di_blocks = %llu\n",
 	       (unsigned long long)di->di_blocks);
-	printk(KERN_INFO "  di_goal_meta = %llu\n",
-	       (unsigned long long)di->di_goal_meta);
-	printk(KERN_INFO "  di_goal_data = %llu\n",
-	       (unsigned long long)di->di_goal_data);
+	printk(KERN_INFO "  i_goal = %llu\n",
+	       (unsigned long long)ip->i_goal);
 	printk(KERN_INFO "  di_flags = 0x%.8X\n", di->di_flags);
 	printk(KERN_INFO "  i_height = %u\n", ip->i_height);
 	printk(KERN_INFO "  i_depth = %u\n", ip->i_depth);
diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c
index 3f10b1f..66193b4 100644
--- a/fs/gfs2/rgrp.c
+++ b/fs/gfs2/rgrp.c
@@ -915,24 +915,20 @@ static struct inode *try_rgrp_unlink(struct gfs2_rgrpd *rgd, u64 *last_unlinked)
 static struct gfs2_rgrpd *recent_rgrp_first(struct gfs2_sbd *sdp,
 					    u64 rglast)
 {
-	struct gfs2_rgrpd *rgd = NULL;
+	struct gfs2_rgrpd *rgd;
 
 	spin_lock(&sdp->sd_rindex_spin);
 
-	if (list_empty(&sdp->sd_rindex_recent_list))
-		goto out;
-
-	if (!rglast)
-		goto first;
-
-	list_for_each_entry(rgd, &sdp->sd_rindex_recent_list, rd_recent) {
-		if (rgd->rd_addr == rglast)
-			goto out;
+	if (rglast) {
+		list_for_each_entry(rgd, &sdp->sd_rindex_recent_list, rd_recent) {
+			if (rgrp_contains_block(rgd, rglast))
+				goto out;
+		}
 	}
-
-first:
-	rgd = list_entry(sdp->sd_rindex_recent_list.next, struct gfs2_rgrpd,
-			 rd_recent);
+	rgd = NULL;
+	if (!list_empty(&sdp->sd_rindex_recent_list))
+		rgd = list_entry(sdp->sd_rindex_recent_list.next,
+				 struct gfs2_rgrpd, rd_recent);
 out:
 	spin_unlock(&sdp->sd_rindex_spin);
 	return rgd;
@@ -1078,7 +1074,7 @@ static struct inode *get_local_rgrp(struct gfs2_inode *ip, u64 *last_unlinked)
 
 	/* Try recently successful rgrps */
 
-	rgd = recent_rgrp_first(sdp, ip->i_last_rg_alloc);
+	rgd = recent_rgrp_first(sdp, ip->i_goal);
 
 	while (rgd) {
 		rg_locked = 0;
@@ -1162,8 +1158,6 @@ static struct inode *get_local_rgrp(struct gfs2_inode *ip, u64 *last_unlinked)
 	}
 
 out:
-	ip->i_last_rg_alloc = rgd->rd_addr;
-
 	if (begin) {
 		recent_rgrp_add(rgd);
 		rgd = gfs2_rgrpd_get_next(rgd);
@@ -1425,8 +1419,8 @@ u64 gfs2_alloc_data(struct gfs2_inode *ip)
 	u32 goal, blk;
 	u64 block;
 
-	if (rgrp_contains_block(rgd, ip->i_di.di_goal_data))
-		goal = ip->i_di.di_goal_data - rgd->rd_data0;
+	if (rgrp_contains_block(rgd, ip->i_goal))
+		goal = ip->i_goal - rgd->rd_data0;
 	else
 		goal = rgd->rd_last_alloc_data;
 
@@ -1435,7 +1429,7 @@ u64 gfs2_alloc_data(struct gfs2_inode *ip)
 	rgd->rd_last_alloc_data = blk;
 
 	block = rgd->rd_data0 + blk;
-	ip->i_di.di_goal_data = block;
+	ip->i_goal = block;
 
 	gfs2_assert_withdraw(sdp, rgd->rd_rg.rg_free);
 	rgd->rd_rg.rg_free--;
@@ -1470,8 +1464,8 @@ u64 gfs2_alloc_meta(struct gfs2_inode *ip)
 	u32 goal, blk;
 	u64 block;
 
-	if (rgrp_contains_block(rgd, ip->i_di.di_goal_meta))
-		goal = ip->i_di.di_goal_meta - rgd->rd_data0;
+	if (rgrp_contains_block(rgd, ip->i_goal))
+		goal = ip->i_goal - rgd->rd_data0;
 	else
 		goal = rgd->rd_last_alloc_meta;
 
@@ -1480,7 +1474,7 @@ u64 gfs2_alloc_meta(struct gfs2_inode *ip)
 	rgd->rd_last_alloc_meta = blk;
 
 	block = rgd->rd_data0 + blk;
-	ip->i_di.di_goal_meta = block;
+	ip->i_goal = block;
 
 	gfs2_assert_withdraw(sdp, rgd->rd_rg.rg_free);
 	rgd->rd_rg.rg_free--;
-- 
1.5.1.2

[Cluster-devel] [PATCH 22/48] [GFS2] Merge the rd_last_alloc_meta and rd_last_alloc_data fields

[swhiteho]

From: Steven Whitehouse <swhiteho@redhat.com>

We don't need to keep track of when we last allocated data
and metadata separately since the only thing thats important
when searching for a free block is whether its free or not,
which is independent from what type of block it is.

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

diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h
index c50dcdf..898b456 100644
--- a/fs/gfs2/incore.h
+++ b/fs/gfs2/incore.h
@@ -90,8 +90,7 @@ struct gfs2_rgrpd {
 	struct mutex rd_mutex;
 	u32 rd_free_clone;
 	struct gfs2_log_element rd_le;
-	u32 rd_last_alloc_data;
-	u32 rd_last_alloc_meta;
+	u32 rd_last_alloc;
 	struct gfs2_sbd *rd_sbd;
 	unsigned char rd_flags;
 #define GFS2_RDF_CHECK        0x01      /* Need to check for unlinked inodes */
diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c
index 66193b4..cc28845 100644
--- a/fs/gfs2/rgrp.c
+++ b/fs/gfs2/rgrp.c
@@ -1422,11 +1422,11 @@ u64 gfs2_alloc_data(struct gfs2_inode *ip)
 	if (rgrp_contains_block(rgd, ip->i_goal))
 		goal = ip->i_goal - rgd->rd_data0;
 	else
-		goal = rgd->rd_last_alloc_data;
+		goal = rgd->rd_last_alloc;
 
 	blk = rgblk_search(rgd, goal, GFS2_BLKST_FREE, GFS2_BLKST_USED);
 	BUG_ON(blk == BFITNOENT);
-	rgd->rd_last_alloc_data = blk;
+	rgd->rd_last_alloc = blk;
 
 	block = rgd->rd_data0 + blk;
 	ip->i_goal = block;
@@ -1467,11 +1467,11 @@ u64 gfs2_alloc_meta(struct gfs2_inode *ip)
 	if (rgrp_contains_block(rgd, ip->i_goal))
 		goal = ip->i_goal - rgd->rd_data0;
 	else
-		goal = rgd->rd_last_alloc_meta;
+		goal = rgd->rd_last_alloc;
 
 	blk = rgblk_search(rgd, goal, GFS2_BLKST_FREE, GFS2_BLKST_USED);
 	BUG_ON(blk == BFITNOENT);
-	rgd->rd_last_alloc_meta = blk;
+	rgd->rd_last_alloc = blk;
 
 	block = rgd->rd_data0 + blk;
 	ip->i_goal = block;
@@ -1510,11 +1510,11 @@ u64 gfs2_alloc_di(struct gfs2_inode *dip, u64 *generation)
 	u32 blk;
 	u64 block;
 
-	blk = rgblk_search(rgd, rgd->rd_last_alloc_meta,
+	blk = rgblk_search(rgd, rgd->rd_last_alloc,
 			   GFS2_BLKST_FREE, GFS2_BLKST_DINODE);
 	BUG_ON(blk == BFITNOENT);
 
-	rgd->rd_last_alloc_meta = blk;
+	rgd->rd_last_alloc = blk;
 
 	block = rgd->rd_data0 + blk;
 
-- 
1.5.1.2

[Cluster-devel] [PATCH 23/48] [GFS2] Update gfs2_trans_add_unrevoke to accept extents

[swhiteho]

From: Steven Whitehouse <swhiteho@redhat.com>

By adding an extra argument to gfs2_trans_add_unrevoke we can now
specify an extent length of blocks to unrevoke. This means that
we only need to make one pass through the list for each extent
rather than each block. Currently the only extent length which
is used is 1, but that will change in the future.

Also gfs2_trans_add_unrevoke is removed from gfs2_alloc_meta
since its the only difference between this and gfs2_alloc_data
which is left. This will allow a future patch to merge these
two functions into one (i.e. one call to allocate both data
and metadata in a single extent in the future).

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

diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index 30d718b..651e532 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -138,7 +138,7 @@ int gfs2_unstuff_dinode(struct gfs2_inode *ip, struct page *page)
 
 		if (isdir) {
 			block = gfs2_alloc_meta(ip);
-
+			gfs2_trans_add_unrevoke(GFS2_SB(&ip->i_inode), block, 1);
 			error = gfs2_dir_get_new_buffer(ip, block, &bh);
 			if (error)
 				goto out_brelse;
@@ -206,6 +206,7 @@ static int build_height(struct inode *inode, struct metapath *mp, unsigned heigh
 
 	for(n = 0; n < new_height; n++) {
 		bn = gfs2_alloc_meta(ip);
+		gfs2_trans_add_unrevoke(GFS2_SB(inode), bn, 1);
 		mp->mp_bh[n] = gfs2_meta_new(ip->i_gl, bn);
 		gfs2_trans_add_bh(ip->i_gl, mp->mp_bh[n], 1);
 	}
@@ -370,8 +371,10 @@ static int lookup_block(struct gfs2_inode *ip, unsigned int height,
 
 	if (height == ip->i_height - 1 && !gfs2_is_dir(ip))
 		*block = gfs2_alloc_data(ip);
-	else
+	else {
 		*block = gfs2_alloc_meta(ip);
+		gfs2_trans_add_unrevoke(GFS2_SB(&ip->i_inode), *block, 1);
+	}
 
 	gfs2_trans_add_bh(ip->i_gl, mp->mp_bh[height], 1);
 
diff --git a/fs/gfs2/dir.c b/fs/gfs2/dir.c
index 081daa9..55514ee 100644
--- a/fs/gfs2/dir.c
+++ b/fs/gfs2/dir.c
@@ -810,7 +810,7 @@ static struct gfs2_leaf *new_leaf(struct inode *inode, struct buffer_head **pbh,
 	struct qstr name = { .name = "", .len = 0, .hash = 0 };
 	if (!bh)
 		return NULL;
-
+	gfs2_trans_add_unrevoke(GFS2_SB(inode), bn, 1);
 	gfs2_trans_add_bh(ip->i_gl, bh, 1);
 	gfs2_metatype_set(bh, GFS2_METATYPE_LF, GFS2_FORMAT_LF);
 	leaf = (struct gfs2_leaf *)bh->b_data;
diff --git a/fs/gfs2/eattr.c b/fs/gfs2/eattr.c
index 04febbc..c7fa0a8 100644
--- a/fs/gfs2/eattr.c
+++ b/fs/gfs2/eattr.c
@@ -585,7 +585,7 @@ static int ea_alloc_blk(struct gfs2_inode *ip, struct buffer_head **bhp)
 	u64 block;
 
 	block = gfs2_alloc_meta(ip);
-
+	gfs2_trans_add_unrevoke(sdp, block, 1);
 	*bhp = gfs2_meta_new(ip->i_gl, block);
 	gfs2_trans_add_bh(ip->i_gl, *bhp, 1);
 	gfs2_metatype_set(*bhp, GFS2_METATYPE_EA, GFS2_FORMAT_EA);
@@ -644,7 +644,7 @@ static int ea_write(struct gfs2_inode *ip, struct gfs2_ea_header *ea,
 			int mh_size = sizeof(struct gfs2_meta_header);
 
 			block = gfs2_alloc_meta(ip);
-
+			gfs2_trans_add_unrevoke(sdp, block, 1);
 			bh = gfs2_meta_new(ip->i_gl, block);
 			gfs2_trans_add_bh(ip->i_gl, bh, 1);
 			gfs2_metatype_set(bh, GFS2_METATYPE_ED, GFS2_FORMAT_ED);
@@ -968,7 +968,7 @@ static int ea_set_block(struct gfs2_inode *ip, struct gfs2_ea_request *er,
 		u64 blk;
 
 		blk = gfs2_alloc_meta(ip);
-
+		gfs2_trans_add_unrevoke(sdp, blk, 1);
 		indbh = gfs2_meta_new(ip->i_gl, blk);
 		gfs2_trans_add_bh(ip->i_gl, indbh, 1);
 		gfs2_metatype_set(indbh, GFS2_METATYPE_IN, GFS2_FORMAT_IN);
diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c
index cc28845..9f28463 100644
--- a/fs/gfs2/rgrp.c
+++ b/fs/gfs2/rgrp.c
@@ -1486,7 +1486,6 @@ u64 gfs2_alloc_meta(struct gfs2_inode *ip)
 
 	gfs2_statfs_change(sdp, 0, -1, 0);
 	gfs2_quota_change(ip, +1, ip->i_inode.i_uid, ip->i_inode.i_gid);
-	gfs2_trans_add_unrevoke(sdp, block);
 
 	spin_lock(&sdp->sd_rindex_spin);
 	rgd->rd_free_clone--;
@@ -1528,7 +1527,7 @@ u64 gfs2_alloc_di(struct gfs2_inode *dip, u64 *generation)
 	al->al_alloced++;
 
 	gfs2_statfs_change(sdp, 0, -1, +1);
-	gfs2_trans_add_unrevoke(sdp, block);
+	gfs2_trans_add_unrevoke(sdp, block, 1);
 
 	spin_lock(&sdp->sd_rindex_spin);
 	rgd->rd_free_clone--;
diff --git a/fs/gfs2/trans.c b/fs/gfs2/trans.c
index 73e5d92..f677b8a 100644
--- a/fs/gfs2/trans.c
+++ b/fs/gfs2/trans.c
@@ -146,30 +146,25 @@ void gfs2_trans_add_revoke(struct gfs2_sbd *sdp, struct gfs2_bufdata *bd)
 	lops_add(sdp, &bd->bd_le);
 }
 
-void gfs2_trans_add_unrevoke(struct gfs2_sbd *sdp, u64 blkno)
+void gfs2_trans_add_unrevoke(struct gfs2_sbd *sdp, u64 blkno, unsigned int len)
 {
-	struct gfs2_bufdata *bd;
-	int found = 0;
+	struct gfs2_bufdata *bd, *tmp;
+	struct gfs2_trans *tr = current->journal_info;
+	unsigned int n = len;
 
 	gfs2_log_lock(sdp);
-
-	list_for_each_entry(bd, &sdp->sd_log_le_revoke, bd_le.le_list) {
-		if (bd->bd_blkno == blkno) {
+	list_for_each_entry_safe(bd, tmp, &sdp->sd_log_le_revoke, bd_le.le_list) {
+		if ((bd->bd_blkno >= blkno) && (bd->bd_blkno < (blkno + len))) {
 			list_del_init(&bd->bd_le.le_list);
 			gfs2_assert_withdraw(sdp, sdp->sd_log_num_revoke);
 			sdp->sd_log_num_revoke--;
-			found = 1;
-			break;
+			kmem_cache_free(gfs2_bufdata_cachep, bd);
+			tr->tr_num_revoke_rm++;
+			if (--n == 0)
+				break;
 		}
 	}
-
 	gfs2_log_unlock(sdp);
-
-	if (found) {
-		struct gfs2_trans *tr = current->journal_info;
-		kmem_cache_free(gfs2_bufdata_cachep, bd);
-		tr->tr_num_revoke_rm++;
-	}
 }
 
 void gfs2_trans_add_rg(struct gfs2_rgrpd *rgd)
diff --git a/fs/gfs2/trans.h b/fs/gfs2/trans.h
index e826f0d..edf9d4b 100644
--- a/fs/gfs2/trans.h
+++ b/fs/gfs2/trans.h
@@ -32,7 +32,7 @@ void gfs2_trans_end(struct gfs2_sbd *sdp);
 
 void gfs2_trans_add_bh(struct gfs2_glock *gl, struct buffer_head *bh, int meta);
 void gfs2_trans_add_revoke(struct gfs2_sbd *sdp, struct gfs2_bufdata *bd);
-void gfs2_trans_add_unrevoke(struct gfs2_sbd *sdp, u64 blkno);
+void gfs2_trans_add_unrevoke(struct gfs2_sbd *sdp, u64 blkno, unsigned int len);
 void gfs2_trans_add_rg(struct gfs2_rgrpd *rgd);
 
 #endif /* __TRANS_DOT_H__ */
-- 
1.5.1.2

[Cluster-devel] [PATCH 24/48] [GFS2] Merge gfs2_alloc_meta and gfs2_alloc_data

[swhiteho]

From: Steven Whitehouse <swhiteho@redhat.com>

Thanks to the preceeding patches, the only difference between
these two functions is their name. We can thus merge them
and call the new function gfs2_alloc_block to reflect the
fact that it can allocate either kind of block.

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

diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index 651e532..e3a75a2 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -137,7 +137,7 @@ int gfs2_unstuff_dinode(struct gfs2_inode *ip, struct page *page)
 		   and write it out to disk */
 
 		if (isdir) {
-			block = gfs2_alloc_meta(ip);
+			block = gfs2_alloc_block(ip);
 			gfs2_trans_add_unrevoke(GFS2_SB(&ip->i_inode), block, 1);
 			error = gfs2_dir_get_new_buffer(ip, block, &bh);
 			if (error)
@@ -146,7 +146,7 @@ int gfs2_unstuff_dinode(struct gfs2_inode *ip, struct page *page)
 					      dibh, sizeof(struct gfs2_dinode));
 			brelse(bh);
 		} else {
-			block = gfs2_alloc_data(ip);
+			block = gfs2_alloc_block(ip);
 
 			error = gfs2_unstuffer_page(ip, dibh, block, page);
 			if (error)
@@ -205,7 +205,7 @@ static int build_height(struct inode *inode, struct metapath *mp, unsigned heigh
 		return error;
 
 	for(n = 0; n < new_height; n++) {
-		bn = gfs2_alloc_meta(ip);
+		bn = gfs2_alloc_block(ip);
 		gfs2_trans_add_unrevoke(GFS2_SB(inode), bn, 1);
 		mp->mp_bh[n] = gfs2_meta_new(ip->i_gl, bn);
 		gfs2_trans_add_bh(ip->i_gl, mp->mp_bh[n], 1);
@@ -369,12 +369,9 @@ static int lookup_block(struct gfs2_inode *ip, unsigned int height,
 	if (!create)
 		return 0;
 
-	if (height == ip->i_height - 1 && !gfs2_is_dir(ip))
-		*block = gfs2_alloc_data(ip);
-	else {
-		*block = gfs2_alloc_meta(ip);
+	*block = gfs2_alloc_block(ip);
+	if (height != ip->i_height - 1 || gfs2_is_dir(ip))
 		gfs2_trans_add_unrevoke(GFS2_SB(&ip->i_inode), *block, 1);
-	}
 
 	gfs2_trans_add_bh(ip->i_gl, mp->mp_bh[height], 1);
 
diff --git a/fs/gfs2/dir.c b/fs/gfs2/dir.c
index 55514ee..fbdf319 100644
--- a/fs/gfs2/dir.c
+++ b/fs/gfs2/dir.c
@@ -803,7 +803,7 @@ got_dent:
 static struct gfs2_leaf *new_leaf(struct inode *inode, struct buffer_head **pbh, u16 depth)
 {
 	struct gfs2_inode *ip = GFS2_I(inode);
-	u64 bn = gfs2_alloc_meta(ip);
+	u64 bn = gfs2_alloc_block(ip);
 	struct buffer_head *bh = gfs2_meta_new(ip->i_gl, bn);
 	struct gfs2_leaf *leaf;
 	struct gfs2_dirent *dent;
diff --git a/fs/gfs2/eattr.c b/fs/gfs2/eattr.c
index c7fa0a8..f9f63bc 100644
--- a/fs/gfs2/eattr.c
+++ b/fs/gfs2/eattr.c
@@ -584,7 +584,7 @@ static int ea_alloc_blk(struct gfs2_inode *ip, struct buffer_head **bhp)
 	struct gfs2_ea_header *ea;
 	u64 block;
 
-	block = gfs2_alloc_meta(ip);
+	block = gfs2_alloc_block(ip);
 	gfs2_trans_add_unrevoke(sdp, block, 1);
 	*bhp = gfs2_meta_new(ip->i_gl, block);
 	gfs2_trans_add_bh(ip->i_gl, *bhp, 1);
@@ -643,7 +643,7 @@ static int ea_write(struct gfs2_inode *ip, struct gfs2_ea_header *ea,
 			u64 block;
 			int mh_size = sizeof(struct gfs2_meta_header);
 
-			block = gfs2_alloc_meta(ip);
+			block = gfs2_alloc_block(ip);
 			gfs2_trans_add_unrevoke(sdp, block, 1);
 			bh = gfs2_meta_new(ip->i_gl, block);
 			gfs2_trans_add_bh(ip->i_gl, bh, 1);
@@ -967,7 +967,7 @@ static int ea_set_block(struct gfs2_inode *ip, struct gfs2_ea_request *er,
 	} else {
 		u64 blk;
 
-		blk = gfs2_alloc_meta(ip);
+		blk = gfs2_alloc_block(ip);
 		gfs2_trans_add_unrevoke(sdp, blk, 1);
 		indbh = gfs2_meta_new(ip->i_gl, blk);
 		gfs2_trans_add_bh(ip->i_gl, indbh, 1);
diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c
index 9f28463..274a2df 100644
--- a/fs/gfs2/rgrp.c
+++ b/fs/gfs2/rgrp.c
@@ -1405,58 +1405,13 @@ static struct gfs2_rgrpd *rgblk_free(struct gfs2_sbd *sdp, u64 bstart,
 }
 
 /**
- * gfs2_alloc_data - Allocate a data block
- * @ip: the inode to allocate the data block for
+ * gfs2_alloc_block - Allocate a block
+ * @ip: the inode to allocate the block for
  *
  * Returns: the allocated block
  */
 
-u64 gfs2_alloc_data(struct gfs2_inode *ip)
-{
-	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
-	struct gfs2_alloc *al = ip->i_alloc;
-	struct gfs2_rgrpd *rgd = al->al_rgd;
-	u32 goal, blk;
-	u64 block;
-
-	if (rgrp_contains_block(rgd, ip->i_goal))
-		goal = ip->i_goal - rgd->rd_data0;
-	else
-		goal = rgd->rd_last_alloc;
-
-	blk = rgblk_search(rgd, goal, GFS2_BLKST_FREE, GFS2_BLKST_USED);
-	BUG_ON(blk == BFITNOENT);
-	rgd->rd_last_alloc = blk;
-
-	block = rgd->rd_data0 + blk;
-	ip->i_goal = block;
-
-	gfs2_assert_withdraw(sdp, rgd->rd_rg.rg_free);
-	rgd->rd_rg.rg_free--;
-
-	gfs2_trans_add_bh(rgd->rd_gl, rgd->rd_bits[0].bi_bh, 1);
-	gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data);
-
-	al->al_alloced++;
-
-	gfs2_statfs_change(sdp, 0, -1, 0);
-	gfs2_quota_change(ip, +1, ip->i_inode.i_uid, ip->i_inode.i_gid);
-
-	spin_lock(&sdp->sd_rindex_spin);
-	rgd->rd_free_clone--;
-	spin_unlock(&sdp->sd_rindex_spin);
-
-	return block;
-}
-
-/**
- * gfs2_alloc_meta - Allocate a metadata block
- * @ip: the inode to allocate the metadata block for
- *
- * Returns: the allocated block
- */
-
-u64 gfs2_alloc_meta(struct gfs2_inode *ip)
+u64 gfs2_alloc_block(struct gfs2_inode *ip)
 {
 	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
 	struct gfs2_alloc *al = ip->i_alloc;
diff --git a/fs/gfs2/rgrp.h b/fs/gfs2/rgrp.h
index 5683605..5e66613 100644
--- a/fs/gfs2/rgrp.h
+++ b/fs/gfs2/rgrp.h
@@ -46,8 +46,7 @@ void gfs2_inplace_release(struct gfs2_inode *ip);
 
 unsigned char gfs2_get_block_type(struct gfs2_rgrpd *rgd, u64 block);
 
-u64 gfs2_alloc_data(struct gfs2_inode *ip);
-u64 gfs2_alloc_meta(struct gfs2_inode *ip);
+u64 gfs2_alloc_block(struct gfs2_inode *ip);
 u64 gfs2_alloc_di(struct gfs2_inode *ip, u64 *generation);
 
 void gfs2_free_data(struct gfs2_inode *ip, u64 bstart, u32 blen);
-- 
1.5.1.2

[Cluster-devel] [PATCH 25/48] [GFS2] Add extent allocation to block allocator

[swhiteho]

From: Steven Whitehouse <swhiteho@redhat.com>

Rather than having to allocate a single block at a time, this patch
allows the block allocator to allocate an extent. Since there is
no difference (so far as the block allocator is concerned) between
data blocks and indirect blocks, it is posible to allocate a single
extent and for the caller to unrevoke just the blocks required
for indirect blocks.

Currently the only bit of GFS2 to make use of this feature is the
build height function. The intention is that gfs2_block_map will
be changed to make use of this feature in future patches.

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

diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index e3a75a2..1fda731 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -136,8 +136,9 @@ int gfs2_unstuff_dinode(struct gfs2_inode *ip, struct page *page)
 		/* Get a free block, fill it with the stuffed data,
 		   and write it out to disk */
 
+		unsigned int n = 1;
+		block = gfs2_alloc_block(ip, &n);
 		if (isdir) {
-			block = gfs2_alloc_block(ip);
 			gfs2_trans_add_unrevoke(GFS2_SB(&ip->i_inode), block, 1);
 			error = gfs2_dir_get_new_buffer(ip, block, &bh);
 			if (error)
@@ -146,8 +147,6 @@ int gfs2_unstuff_dinode(struct gfs2_inode *ip, struct page *page)
 					      dibh, sizeof(struct gfs2_dinode));
 			brelse(bh);
 		} else {
-			block = gfs2_alloc_block(ip);
-
 			error = gfs2_unstuffer_page(ip, dibh, block, page);
 			if (error)
 				goto out_brelse;
@@ -195,7 +194,7 @@ static int build_height(struct inode *inode, struct metapath *mp, unsigned heigh
 	int error;
 	__be64 *bp;
 	u64 bn;
-	unsigned n;
+	unsigned n, i = 0;
 
 	if (height <= ip->i_height)
 		return 0;
@@ -204,12 +203,16 @@ static int build_height(struct inode *inode, struct metapath *mp, unsigned heigh
 	if (error)
 		return error;
 
-	for(n = 0; n < new_height; n++) {
-		bn = gfs2_alloc_block(ip);
-		gfs2_trans_add_unrevoke(GFS2_SB(inode), bn, 1);
-		mp->mp_bh[n] = gfs2_meta_new(ip->i_gl, bn);
-		gfs2_trans_add_bh(ip->i_gl, mp->mp_bh[n], 1);
-	}
+	do {
+		n = new_height - i;
+		bn = gfs2_alloc_block(ip, &n);
+		gfs2_trans_add_unrevoke(GFS2_SB(inode), bn, n);
+		do {
+			mp->mp_bh[i] = gfs2_meta_new(ip->i_gl, bn++);
+			gfs2_trans_add_bh(ip->i_gl, mp->mp_bh[i], 1);
+			i++;
+		} while(i < n);
+	} while(i < new_height);
 
 	n = 0;
 	bn = mp->mp_bh[0]->b_blocknr;
@@ -358,6 +361,7 @@ static int lookup_block(struct gfs2_inode *ip, unsigned int height,
 {
 	int boundary;
 	__be64 *ptr = metapointer(&boundary, height, mp);
+	unsigned int n = 1;
 
 	if (*ptr) {
 		*block = be64_to_cpu(*ptr);
@@ -369,7 +373,7 @@ static int lookup_block(struct gfs2_inode *ip, unsigned int height,
 	if (!create)
 		return 0;
 
-	*block = gfs2_alloc_block(ip);
+	*block = gfs2_alloc_block(ip, &n);
 	if (height != ip->i_height - 1 || gfs2_is_dir(ip))
 		gfs2_trans_add_unrevoke(GFS2_SB(&ip->i_inode), *block, 1);
 
diff --git a/fs/gfs2/dir.c b/fs/gfs2/dir.c
index fbdf319..93a2e6a 100644
--- a/fs/gfs2/dir.c
+++ b/fs/gfs2/dir.c
@@ -803,7 +803,8 @@ got_dent:
 static struct gfs2_leaf *new_leaf(struct inode *inode, struct buffer_head **pbh, u16 depth)
 {
 	struct gfs2_inode *ip = GFS2_I(inode);
-	u64 bn = gfs2_alloc_block(ip);
+	unsigned int n = 1;
+	u64 bn = gfs2_alloc_block(ip, &n);
 	struct buffer_head *bh = gfs2_meta_new(ip->i_gl, bn);
 	struct gfs2_leaf *leaf;
 	struct gfs2_dirent *dent;
diff --git a/fs/gfs2/eattr.c b/fs/gfs2/eattr.c
index f9f63bc..0e79cd5 100644
--- a/fs/gfs2/eattr.c
+++ b/fs/gfs2/eattr.c
@@ -582,9 +582,10 @@ static int ea_alloc_blk(struct gfs2_inode *ip, struct buffer_head **bhp)
 {
 	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
 	struct gfs2_ea_header *ea;
+	unsigned int n = 1;
 	u64 block;
 
-	block = gfs2_alloc_block(ip);
+	block = gfs2_alloc_block(ip, &n);
 	gfs2_trans_add_unrevoke(sdp, block, 1);
 	*bhp = gfs2_meta_new(ip->i_gl, block);
 	gfs2_trans_add_bh(ip->i_gl, *bhp, 1);
@@ -642,8 +643,9 @@ static int ea_write(struct gfs2_inode *ip, struct gfs2_ea_header *ea,
 			struct buffer_head *bh;
 			u64 block;
 			int mh_size = sizeof(struct gfs2_meta_header);
+			unsigned int n = 1;
 
-			block = gfs2_alloc_block(ip);
+			block = gfs2_alloc_block(ip, &n);
 			gfs2_trans_add_unrevoke(sdp, block, 1);
 			bh = gfs2_meta_new(ip->i_gl, block);
 			gfs2_trans_add_bh(ip->i_gl, bh, 1);
@@ -966,8 +968,8 @@ static int ea_set_block(struct gfs2_inode *ip, struct gfs2_ea_request *er,
 		gfs2_trans_add_bh(ip->i_gl, indbh, 1);
 	} else {
 		u64 blk;
-
-		blk = gfs2_alloc_block(ip);
+		unsigned int n = 1;
+		blk = gfs2_alloc_block(ip, &n);
 		gfs2_trans_add_unrevoke(sdp, blk, 1);
 		indbh = gfs2_meta_new(ip->i_gl, blk);
 		gfs2_trans_add_bh(ip->i_gl, indbh, 1);
diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c
index 274a2df..77eba0a 100644
--- a/fs/gfs2/rgrp.c
+++ b/fs/gfs2/rgrp.c
@@ -53,7 +53,8 @@ static const char valid_change[16] = {
 };
 
 static u32 rgblk_search(struct gfs2_rgrpd *rgd, u32 goal,
-                        unsigned char old_state, unsigned char new_state);
+                        unsigned char old_state, unsigned char new_state,
+			unsigned int *n);
 
 /**
  * gfs2_setbit - Set a bit in the bitmaps
@@ -64,26 +65,32 @@ static u32 rgblk_search(struct gfs2_rgrpd *rgd, u32 goal,
  *
  */
 
-static void gfs2_setbit(struct gfs2_rgrpd *rgd, unsigned char *buffer,
-			unsigned int buflen, u32 block,
-			unsigned char new_state)
+static inline void gfs2_setbit(struct gfs2_rgrpd *rgd, unsigned char *buf1,
+			       unsigned char *buf2, unsigned int offset,
+			       unsigned int buflen, u32 block,
+			       unsigned char new_state)
 {
-	unsigned char *byte, *end, cur_state;
-	unsigned int bit;
+	unsigned char *byte1, *byte2, *end, cur_state;
+	const unsigned int bit = (block % GFS2_NBBY) * GFS2_BIT_SIZE;
 
-	byte = buffer + (block / GFS2_NBBY);
-	bit = (block % GFS2_NBBY) * GFS2_BIT_SIZE;
-	end = buffer + buflen;
+	byte1 = buf1 + offset + (block / GFS2_NBBY);
+	end = buf1 + offset + buflen;
 
-	gfs2_assert(rgd->rd_sbd, byte < end);
+	BUG_ON(byte1 >= end);
 
-	cur_state = (*byte >> bit) & GFS2_BIT_MASK;
+	cur_state = (*byte1 >> bit) & GFS2_BIT_MASK;
 
-	if (valid_change[new_state * 4 + cur_state]) {
-		*byte ^= cur_state << bit;
-		*byte |= new_state << bit;
-	} else
+	if (unlikely(!valid_change[new_state * 4 + cur_state])) {
 		gfs2_consist_rgrpd(rgd);
+		return;
+	}
+	*byte1 ^= (cur_state ^ new_state) << bit;
+
+	if (buf2) {
+		byte2 = buf2 + offset + (block / GFS2_NBBY);
+		cur_state = (*byte2 >> bit) & GFS2_BIT_MASK;
+		*byte2 ^= (cur_state ^ new_state) << bit;
+	}
 }
 
 /**
@@ -94,10 +101,12 @@ static void gfs2_setbit(struct gfs2_rgrpd *rgd, unsigned char *buffer,
  *
  */
 
-static unsigned char gfs2_testbit(struct gfs2_rgrpd *rgd, unsigned char *buffer,
-				  unsigned int buflen, u32 block)
+static inline unsigned char gfs2_testbit(struct gfs2_rgrpd *rgd,
+					 const unsigned char *buffer,
+					 unsigned int buflen, u32 block)
 {
-	unsigned char *byte, *end, cur_state;
+	const unsigned char *byte, *end;
+	unsigned char cur_state;
 	unsigned int bit;
 
 	byte = buffer + (block / GFS2_NBBY);
@@ -877,13 +886,15 @@ static struct inode *try_rgrp_unlink(struct gfs2_rgrpd *rgd, u64 *last_unlinked)
 	u32 goal = 0, block;
 	u64 no_addr;
 	struct gfs2_sbd *sdp = rgd->rd_sbd;
+	unsigned int n;
 
 	for(;;) {
 		if (goal >= rgd->rd_data)
 			break;
 		down_write(&sdp->sd_log_flush_lock);
+		n = 1;
 		block = rgblk_search(rgd, goal, GFS2_BLKST_UNLINKED,
-				     GFS2_BLKST_UNLINKED);
+				     GFS2_BLKST_UNLINKED, &n);
 		up_write(&sdp->sd_log_flush_lock);
 		if (block == BFITNOENT)
 			break;
@@ -1280,6 +1291,7 @@ unsigned char gfs2_get_block_type(struct gfs2_rgrpd *rgd, u64 block)
  * @goal: the goal block within the RG (start here to search for avail block)
  * @old_state: GFS2_BLKST_XXX the before-allocation state to find
  * @new_state: GFS2_BLKST_XXX the after-allocation block state
+ * @n: The extent length
  *
  * Walk rgrp's bitmap to find bits that represent a block in @old_state.
  * Add the found bitmap buffer to the transaction.
@@ -1295,13 +1307,17 @@ unsigned char gfs2_get_block_type(struct gfs2_rgrpd *rgd, u64 block)
  */
 
 static u32 rgblk_search(struct gfs2_rgrpd *rgd, u32 goal,
-			unsigned char old_state, unsigned char new_state)
+			unsigned char old_state, unsigned char new_state,
+			unsigned int *n)
 {
 	struct gfs2_bitmap *bi = NULL;
-	u32 length = rgd->rd_length;
+	const u32 length = rgd->rd_length;
 	u32 blk = 0;
 	unsigned int buf, x;
+	const unsigned int elen = *n;
+	const u8 *buffer;
 
+	*n = 0;
 	/* Find bitmap block that contains bits for goal block */
 	for (buf = 0; buf < length; buf++) {
 		bi = rgd->rd_bits + buf;
@@ -1322,7 +1338,7 @@ static u32 rgblk_search(struct gfs2_rgrpd *rgd, u32 goal,
 	for (x = 0; x <= length; x++) {
 		/* The GFS2_BLKST_UNLINKED state doesn't apply to the clone
 		   bitmaps, so we must search the originals for that. */
-		const u8 *buffer = bi->bi_bh->b_data + bi->bi_offset;
+		buffer = bi->bi_bh->b_data + bi->bi_offset;
 		if (old_state != GFS2_BLKST_UNLINKED && bi->bi_clone)
 			buffer = bi->bi_clone + bi->bi_offset;
 
@@ -1337,12 +1353,21 @@ static u32 rgblk_search(struct gfs2_rgrpd *rgd, u32 goal,
 	}
 
 	if (blk != BFITNOENT && old_state != new_state) {
+		*n = 1;
 		gfs2_trans_add_bh(rgd->rd_gl, bi->bi_bh, 1);
-		gfs2_setbit(rgd, bi->bi_bh->b_data + bi->bi_offset,
+		gfs2_setbit(rgd, bi->bi_bh->b_data, bi->bi_clone, bi->bi_offset,
 			    bi->bi_len, blk, new_state);
-		if (bi->bi_clone)
-			gfs2_setbit(rgd, bi->bi_clone + bi->bi_offset,
-				    bi->bi_len, blk, new_state);
+		while(*n < elen) {
+			goal++;
+			if (goal >= (bi->bi_len / GFS2_NBBY))
+				break;
+			if (gfs2_testbit(rgd, buffer, bi->bi_len, goal) !=
+			    GFS2_BLKST_FREE)
+				break;
+			(*n)++;
+			gfs2_setbit(rgd, bi->bi_bh->b_data, bi->bi_clone,
+				    bi->bi_offset, bi->bi_len, blk, new_state);
+		}
 	}
 
 	return (blk == BFITNOENT) ? blk : (bi->bi_start * GFS2_NBBY) + blk;
@@ -1397,7 +1422,7 @@ static struct gfs2_rgrpd *rgblk_free(struct gfs2_sbd *sdp, u64 bstart,
 			       bi->bi_len);
 		}
 		gfs2_trans_add_bh(rgd->rd_gl, bi->bi_bh, 1);
-		gfs2_setbit(rgd, bi->bi_bh->b_data + bi->bi_offset,
+		gfs2_setbit(rgd, bi->bi_bh->b_data, NULL, bi->bi_offset,
 			    bi->bi_len, buf_blk, new_state);
 	}
 
@@ -1411,7 +1436,7 @@ static struct gfs2_rgrpd *rgblk_free(struct gfs2_sbd *sdp, u64 bstart,
  * Returns: the allocated block
  */
 
-u64 gfs2_alloc_block(struct gfs2_inode *ip)
+u64 gfs2_alloc_block(struct gfs2_inode *ip, unsigned int *n)
 {
 	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
 	struct gfs2_alloc *al = ip->i_alloc;
@@ -1424,26 +1449,26 @@ u64 gfs2_alloc_block(struct gfs2_inode *ip)
 	else
 		goal = rgd->rd_last_alloc;
 
-	blk = rgblk_search(rgd, goal, GFS2_BLKST_FREE, GFS2_BLKST_USED);
+	blk = rgblk_search(rgd, goal, GFS2_BLKST_FREE, GFS2_BLKST_USED, n);
 	BUG_ON(blk == BFITNOENT);
-	rgd->rd_last_alloc = blk;
 
+	rgd->rd_last_alloc = blk;
 	block = rgd->rd_data0 + blk;
 	ip->i_goal = block;
 
-	gfs2_assert_withdraw(sdp, rgd->rd_rg.rg_free);
-	rgd->rd_rg.rg_free--;
+	gfs2_assert_withdraw(sdp, rgd->rd_rg.rg_free >= *n);
+	rgd->rd_rg.rg_free -= *n;
 
 	gfs2_trans_add_bh(rgd->rd_gl, rgd->rd_bits[0].bi_bh, 1);
 	gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data);
 
-	al->al_alloced++;
+	al->al_alloced += *n;
 
-	gfs2_statfs_change(sdp, 0, -1, 0);
-	gfs2_quota_change(ip, +1, ip->i_inode.i_uid, ip->i_inode.i_gid);
+	gfs2_statfs_change(sdp, 0, -*n, 0);
+	gfs2_quota_change(ip, *n, ip->i_inode.i_uid, ip->i_inode.i_gid);
 
 	spin_lock(&sdp->sd_rindex_spin);
-	rgd->rd_free_clone--;
+	rgd->rd_free_clone -= *n;
 	spin_unlock(&sdp->sd_rindex_spin);
 
 	return block;
@@ -1463,9 +1488,10 @@ u64 gfs2_alloc_di(struct gfs2_inode *dip, u64 *generation)
 	struct gfs2_rgrpd *rgd = al->al_rgd;
 	u32 blk;
 	u64 block;
+	unsigned int n = 1;
 
 	blk = rgblk_search(rgd, rgd->rd_last_alloc,
-			   GFS2_BLKST_FREE, GFS2_BLKST_DINODE);
+			   GFS2_BLKST_FREE, GFS2_BLKST_DINODE, &n);
 	BUG_ON(blk == BFITNOENT);
 
 	rgd->rd_last_alloc = blk;
diff --git a/fs/gfs2/rgrp.h b/fs/gfs2/rgrp.h
index 5e66613..3181c7e 100644
--- a/fs/gfs2/rgrp.h
+++ b/fs/gfs2/rgrp.h
@@ -46,7 +46,7 @@ void gfs2_inplace_release(struct gfs2_inode *ip);
 
 unsigned char gfs2_get_block_type(struct gfs2_rgrpd *rgd, u64 block);
 
-u64 gfs2_alloc_block(struct gfs2_inode *ip);
+u64 gfs2_alloc_block(struct gfs2_inode *ip, unsigned int *n);
 u64 gfs2_alloc_di(struct gfs2_inode *ip, u64 *generation);
 
 void gfs2_free_data(struct gfs2_inode *ip, u64 bstart, u32 blen);
-- 
1.5.1.2

[Cluster-devel] [PATCH 26/48] [GFS2] The case of the missing asterisk

[swhiteho]

From: Steven Whitehouse <swhiteho@redhat.com>

A dereference was forgotten. This adds it back correctly.

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

diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index 1fda731..7f72564 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -402,7 +402,7 @@ static int lookup_metapath(struct inode *inode, struct metapath *mp,
 
 	for (x = 0; x < end_of_metadata; x++) {
 		lookup_block(ip, x, mp, create, new, dblock);
-		if (!dblock)
+		if (!*dblock)
 			return 0;
 
 		ret = gfs2_meta_indirect_buffer(ip, x+1, *dblock, *new, &mp->mp_bh[x+1]);
-- 
1.5.1.2

[Cluster-devel] [PATCH 27/48] [GFS2] Add a function to interate over an extent

[swhiteho]

From: Steven Whitehouse <swhiteho@redhat.com>

This adds a function (currently the only use is during mapping
of already allocated blocks, but watch this space) which iterates
over a number of pointers in a block and returns the extent length.

If the initial pointer is 0 (i.e. unallocated) it will return the
number of unallocated blocks in the extent. If the initial pointer
is allocated, then it returns the number of contiguously allocated
blocks in the extent.

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

diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index 7f72564..6780aa5 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -422,6 +422,42 @@ static void release_metapath(struct metapath *mp)
 			brelse(mp->mp_bh[i]);
 }
 
+/**
+ * gfs2_extent_length - Returns length of an extent of blocks
+ * @start: Start of the buffer
+ * @len: Length of the buffer in bytes
+ * @ptr: Current position in the buffer
+ * @limit: Max extent length to return (0 = unlimited)
+ * @eob: Set to 1 if we hit "end of block"
+ *
+ * If the first block is zero (unallocated) it will return the number of
+ * unallocated blocks in the extent, otherwise it will return the number
+ * of contiguous blocks in the extent.
+ *
+ * Returns: The length of the extent (minimum of one block)
+ */
+
+static inline unsigned int gfs2_extent_length(void *start, unsigned int len, __be64 *ptr, unsigned limit, int *eob)
+{
+	const __be64 *end = (start + len);
+	const __be64 *first = ptr;
+	u64 d = be64_to_cpu(*ptr);
+
+	*eob = 0;
+	do {
+		ptr++;
+		if (ptr >= end)
+			break;
+		if (limit && --limit == 0)
+			break;
+		if (d)
+			d++;
+	} while(be64_to_cpu(*ptr) == d);
+	if (ptr >= end)
+		*eob = 1;
+	return (ptr - first);
+}
+
 static inline void bmap_lock(struct inode *inode, int create)
 {
 	struct gfs2_inode *ip = GFS2_I(inode);
@@ -499,26 +535,26 @@ int gfs2_block_map(struct inode *inode, sector_t lblock,
 		goto out_fail;
 	boundary = error;
 
+	if (new) {
+		map_bh(bh_map, inode->i_sb, dblock);
+		if (boundary)
+			set_buffer_boundary(bh_map);
+		gfs2_trans_add_bh(ip->i_gl, mp.mp_bh[0], 1);
+		gfs2_dinode_out(ip, mp.mp_bh[0]->b_data);
+		set_buffer_new(bh_map);
+		goto out_ok;
+	}
+
 	if (dblock) {
+		unsigned int len;
+		struct buffer_head *bh = mp.mp_bh[ip->i_height - 1];
+		__be64 *ptr = metapointer(&boundary, ip->i_height - 1, &mp);
 		map_bh(bh_map, inode->i_sb, dblock);
+		len = gfs2_extent_length(bh->b_data, bh->b_size, ptr, maxlen,
+					 &boundary);
+		bh_map->b_size = (len << inode->i_blkbits);
 		if (boundary)
 			set_buffer_boundary(bh_map);
-		if (new) {
-			gfs2_trans_add_bh(ip->i_gl, mp.mp_bh[0], 1);
-			gfs2_dinode_out(ip, mp.mp_bh[0]->b_data);
-			set_buffer_new(bh_map);
-			goto out_ok;
-		}
-		while(--maxlen && !buffer_boundary(bh_map)) {
-			u64 eblock;
-			mp.mp_list[ip->i_height - 1]++;
-			boundary = lookup_block(ip, ip->i_height - 1, &mp, 0, &new, &eblock);
-			if (eblock != ++dblock)
-				break;
-			bh_map->b_size += (1 << inode->i_blkbits);
-			if (boundary)
-				set_buffer_boundary(bh_map);
-		}
 	}
 out_ok:
 	error = 0;
-- 
1.5.1.2

[Cluster-devel] [PATCH 28/48] [GFS2] Eliminate (almost) duplicate field from gfs2_inode

[swhiteho]

From: Steven Whitehouse <swhiteho@redhat.com>

The blocks counter is almost a duplicate of the i_blocks
field in the VFS inode. The only difference is that i_blocks
can be only 32bits long for 32bit arch without large single file
support. Since GFS2 doesn't handle the non-large single file
case (for 32 bit anyway) this adds a new config dependency on
64BIT || LSF. This has always been the case, however we've never
explicitly said so before.

Even if we do add support for the non-LSF case, we will still
not require this field to be duplicated since we will not be
able to access oversized files anyway.

So the net result of all this is that we shave 8 bytes from a gfs2_inode
and get our config deps correct.

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

diff --git a/fs/gfs2/Kconfig b/fs/gfs2/Kconfig
index de8e64c..d147b53 100644
--- a/fs/gfs2/Kconfig
+++ b/fs/gfs2/Kconfig
@@ -1,6 +1,6 @@
 config GFS2_FS
 	tristate "GFS2 file system support"
-	depends on EXPERIMENTAL
+	depends on EXPERIMENTAL && (64BIT || LSF)
 	select FS_POSIX_ACL
 	select CRC32
 	help
diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index 6780aa5..e27e660 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -161,9 +161,8 @@ int gfs2_unstuff_dinode(struct gfs2_inode *ip, struct page *page)
 
 	if (ip->i_di.di_size) {
 		*(__be64 *)(di + 1) = cpu_to_be64(block);
-		ip->i_di.di_blocks++;
-		gfs2_set_inode_blocks(&ip->i_inode);
-		di->di_blocks = cpu_to_be64(ip->i_di.di_blocks);
+		gfs2_add_inode_blocks(&ip->i_inode, 1);
+		di->di_blocks = cpu_to_be64(gfs2_get_inode_blocks(&ip->i_inode));
 	}
 
 	ip->i_height = 1;
@@ -238,10 +237,9 @@ static int build_height(struct inode *inode, struct metapath *mp, unsigned heigh
 	gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode));
 	*(__be64 *)(di + 1) = cpu_to_be64(bn);
 	ip->i_height += new_height;
-	ip->i_di.di_blocks += new_height;
-	gfs2_set_inode_blocks(&ip->i_inode);
+	gfs2_add_inode_blocks(&ip->i_inode, new_height);
 	di->di_height = cpu_to_be16(ip->i_height);
-	di->di_blocks = cpu_to_be64(ip->i_di.di_blocks);
+	di->di_blocks = cpu_to_be64(gfs2_get_inode_blocks(&ip->i_inode));
 	brelse(dibh);
 	return error;
 }
@@ -380,8 +378,7 @@ static int lookup_block(struct gfs2_inode *ip, unsigned int height,
 	gfs2_trans_add_bh(ip->i_gl, mp->mp_bh[height], 1);
 
 	*ptr = cpu_to_be64(*block);
-	ip->i_di.di_blocks++;
-	gfs2_set_inode_blocks(&ip->i_inode);
+	gfs2_add_inode_blocks(&ip->i_inode, 1);
 
 	*new = 1;
 	return 0;
@@ -779,10 +776,7 @@ static int do_strip(struct gfs2_inode *ip, struct buffer_head *dibh,
 		}
 
 		*p = 0;
-		if (!ip->i_di.di_blocks)
-			gfs2_consist_inode(ip);
-		ip->i_di.di_blocks--;
-		gfs2_set_inode_blocks(&ip->i_inode);
+		gfs2_add_inode_blocks(&ip->i_inode, -1);
 	}
 	if (bstart) {
 		if (metadata)
diff --git a/fs/gfs2/dir.c b/fs/gfs2/dir.c
index 93a2e6a..862aa32 100644
--- a/fs/gfs2/dir.c
+++ b/fs/gfs2/dir.c
@@ -906,8 +906,7 @@ static int dir_make_exhash(struct inode *inode)
 		*lp = cpu_to_be64(bn);
 
 	dip->i_di.di_size = sdp->sd_sb.sb_bsize / 2;
-	dip->i_di.di_blocks++;
-	gfs2_set_inode_blocks(&dip->i_inode);
+	gfs2_add_inode_blocks(&dip->i_inode, 1);
 	dip->i_di.di_flags |= GFS2_DIF_EXHASH;
 
 	for (x = sdp->sd_hash_ptrs, y = -1; x; x >>= 1, y++) ;
@@ -1045,8 +1044,7 @@ static int dir_split_leaf(struct inode *inode, const struct qstr *name)
 	error = gfs2_meta_inode_buffer(dip, &dibh);
 	if (!gfs2_assert_withdraw(GFS2_SB(&dip->i_inode), !error)) {
 		gfs2_trans_add_bh(dip->i_gl, dibh, 1);
-		dip->i_di.di_blocks++;
-		gfs2_set_inode_blocks(&dip->i_inode);
+		gfs2_add_inode_blocks(&dip->i_inode, 1);
 		gfs2_dinode_out(dip, dibh->b_data);
 		brelse(dibh);
 	}
@@ -1580,8 +1578,7 @@ static int dir_new_leaf(struct inode *inode, const struct qstr *name)
 	if (error)
 		return error;
 	gfs2_trans_add_bh(ip->i_gl, bh, 1);
-	ip->i_di.di_blocks++;
-	gfs2_set_inode_blocks(&ip->i_inode);
+	gfs2_add_inode_blocks(&ip->i_inode, 1);
 	gfs2_dinode_out(ip, bh->b_data);
 	brelse(bh);
 	return 0;
@@ -1922,11 +1919,7 @@ static int leaf_dealloc(struct gfs2_inode *dip, u32 index, u32 len,
 		brelse(bh);
 
 		gfs2_free_meta(dip, blk, 1);
-
-		if (!dip->i_di.di_blocks)
-			gfs2_consist_inode(dip);
-		dip->i_di.di_blocks--;
-		gfs2_set_inode_blocks(&dip->i_inode);
+		gfs2_add_inode_blocks(&dip->i_inode, -1);
 	}
 
 	error = gfs2_dir_write_data(dip, ht, index * sizeof(u64), size);
diff --git a/fs/gfs2/eattr.c b/fs/gfs2/eattr.c
index 0e79cd5..76ead1a 100644
--- a/fs/gfs2/eattr.c
+++ b/fs/gfs2/eattr.c
@@ -277,10 +277,7 @@ static int ea_dealloc_unstuffed(struct gfs2_inode *ip, struct buffer_head *bh,
 		}
 
 		*dataptrs = 0;
-		if (!ip->i_di.di_blocks)
-			gfs2_consist_inode(ip);
-		ip->i_di.di_blocks--;
-		gfs2_set_inode_blocks(&ip->i_inode);
+		gfs2_add_inode_blocks(&ip->i_inode, -1);
 	}
 	if (bstart)
 		gfs2_free_meta(ip, bstart, blen);
@@ -598,8 +595,7 @@ static int ea_alloc_blk(struct gfs2_inode *ip, struct buffer_head **bhp)
 	ea->ea_flags = GFS2_EAFLAG_LAST;
 	ea->ea_num_ptrs = 0;
 
-	ip->i_di.di_blocks++;
-	gfs2_set_inode_blocks(&ip->i_inode);
+	gfs2_add_inode_blocks(&ip->i_inode, 1);
 
 	return 0;
 }
@@ -651,8 +647,7 @@ static int ea_write(struct gfs2_inode *ip, struct gfs2_ea_header *ea,
 			gfs2_trans_add_bh(ip->i_gl, bh, 1);
 			gfs2_metatype_set(bh, GFS2_METATYPE_ED, GFS2_FORMAT_ED);
 
-			ip->i_di.di_blocks++;
-			gfs2_set_inode_blocks(&ip->i_inode);
+			gfs2_add_inode_blocks(&ip->i_inode, 1);
 
 			copy = data_len > sdp->sd_jbsize ? sdp->sd_jbsize :
 							   data_len;
@@ -980,8 +975,7 @@ static int ea_set_block(struct gfs2_inode *ip, struct gfs2_ea_request *er,
 		*eablk = cpu_to_be64(ip->i_di.di_eattr);
 		ip->i_di.di_eattr = blk;
 		ip->i_di.di_flags |= GFS2_DIF_EA_INDIRECT;
-		ip->i_di.di_blocks++;
-		gfs2_set_inode_blocks(&ip->i_inode);
+		gfs2_add_inode_blocks(&ip->i_inode, 1);
 
 		eablk++;
 	}
@@ -1389,10 +1383,7 @@ static int ea_dealloc_indirect(struct gfs2_inode *ip)
 		}
 
 		*eablk = 0;
-		if (!ip->i_di.di_blocks)
-			gfs2_consist_inode(ip);
-		ip->i_di.di_blocks--;
-		gfs2_set_inode_blocks(&ip->i_inode);
+		gfs2_add_inode_blocks(&ip->i_inode, -1);
 	}
 	if (bstart)
 		gfs2_free_meta(ip, bstart, blen);
@@ -1444,10 +1435,7 @@ static int ea_dealloc_block(struct gfs2_inode *ip)
 	gfs2_free_meta(ip, ip->i_di.di_eattr, 1);
 
 	ip->i_di.di_eattr = 0;
-	if (!ip->i_di.di_blocks)
-		gfs2_consist_inode(ip);
-	ip->i_di.di_blocks--;
-	gfs2_set_inode_blocks(&ip->i_inode);
+	gfs2_add_inode_blocks(&ip->i_inode, -1);
 
 	error = gfs2_meta_inode_buffer(ip, &dibh);
 	if (!error) {
diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h
index 898b456..4ba2ea6 100644
--- a/fs/gfs2/incore.h
+++ b/fs/gfs2/incore.h
@@ -239,7 +239,6 @@ enum {
 
 struct gfs2_dinode_host {
 	u64 di_size;		/* number of bytes in file */
-	u64 di_blocks;		/* number of blocks in file */
 	u64 di_generation;	/* generation number for NFS */
 	u32 di_flags;		/* GFS2_DIF_... */
 	/* These only apply to directories  */
diff --git a/fs/gfs2/inode.c b/fs/gfs2/inode.c
index c3fe8aa..5f50dd5 100644
--- a/fs/gfs2/inode.c
+++ b/fs/gfs2/inode.c
@@ -273,8 +273,7 @@ static int gfs2_dinode_in(struct gfs2_inode *ip, const void *buf)
 	ip->i_inode.i_nlink = be32_to_cpu(str->di_nlink);
 	di->di_size = be64_to_cpu(str->di_size);
 	i_size_write(&ip->i_inode, di->di_size);
-	di->di_blocks = be64_to_cpu(str->di_blocks);
-	gfs2_set_inode_blocks(&ip->i_inode);
+	gfs2_set_inode_blocks(&ip->i_inode, be64_to_cpu(str->di_blocks));
 	ip->i_inode.i_atime.tv_sec = be64_to_cpu(str->di_atime);
 	ip->i_inode.i_atime.tv_nsec = be32_to_cpu(str->di_atime_nsec);
 	ip->i_inode.i_mtime.tv_sec = be64_to_cpu(str->di_mtime);
@@ -344,7 +343,7 @@ int gfs2_dinode_dealloc(struct gfs2_inode *ip)
 	struct gfs2_rgrpd *rgd;
 	int error;
 
-	if (ip->i_di.di_blocks != 1) {
+	if (gfs2_get_inode_blocks(&ip->i_inode) != 1) {
 		if (gfs2_consist_inode(ip))
 			gfs2_dinode_print(ip);
 		return -EIO;
@@ -1398,7 +1397,7 @@ void gfs2_dinode_out(const struct gfs2_inode *ip, void *buf)
 	str->di_gid = cpu_to_be32(ip->i_inode.i_gid);
 	str->di_nlink = cpu_to_be32(ip->i_inode.i_nlink);
 	str->di_size = cpu_to_be64(di->di_size);
-	str->di_blocks = cpu_to_be64(di->di_blocks);
+	str->di_blocks = cpu_to_be64(gfs2_get_inode_blocks(&ip->i_inode));
 	str->di_atime = cpu_to_be64(ip->i_inode.i_atime.tv_sec);
 	str->di_mtime = cpu_to_be64(ip->i_inode.i_mtime.tv_sec);
 	str->di_ctime = cpu_to_be64(ip->i_inode.i_ctime.tv_sec);
@@ -1430,8 +1429,8 @@ void gfs2_dinode_print(const struct gfs2_inode *ip)
 	printk(KERN_INFO "  no_addr = %llu\n",
 	       (unsigned long long)ip->i_no_addr);
 	printk(KERN_INFO "  di_size = %llu\n", (unsigned long long)di->di_size);
-	printk(KERN_INFO "  di_blocks = %llu\n",
-	       (unsigned long long)di->di_blocks);
+	printk(KERN_INFO "  blocks = %llu\n",
+	       (unsigned long long)gfs2_get_inode_blocks(&ip->i_inode));
 	printk(KERN_INFO "  i_goal = %llu\n",
 	       (unsigned long long)ip->i_goal);
 	printk(KERN_INFO "  di_flags = 0x%.8X\n", di->di_flags);
diff --git a/fs/gfs2/inode.h b/fs/gfs2/inode.h
index db73868..580da45 100644
--- a/fs/gfs2/inode.h
+++ b/fs/gfs2/inode.h
@@ -10,6 +10,8 @@
 #ifndef __INODE_DOT_H__
 #define __INODE_DOT_H__
 
+#include "util.h"
+
 static inline int gfs2_is_stuffed(const struct gfs2_inode *ip)
 {
 	return !ip->i_height;
@@ -37,13 +39,25 @@ static inline int gfs2_is_dir(const struct gfs2_inode *ip)
 	return S_ISDIR(ip->i_inode.i_mode);
 }
 
-static inline void gfs2_set_inode_blocks(struct inode *inode)
+static inline void gfs2_set_inode_blocks(struct inode *inode, u64 blocks)
+{
+	inode->i_blocks = blocks <<
+		(GFS2_SB(inode)->sd_sb.sb_bsize_shift - GFS2_BASIC_BLOCK_SHIFT);
+}
+
+static inline u64 gfs2_get_inode_blocks(const struct inode *inode)
 {
-	struct gfs2_inode *ip = GFS2_I(inode);
-	inode->i_blocks = ip->i_di.di_blocks <<
+	return inode->i_blocks >>
 		(GFS2_SB(inode)->sd_sb.sb_bsize_shift - GFS2_BASIC_BLOCK_SHIFT);
 }
 
+static inline void gfs2_add_inode_blocks(struct inode *inode, s64 change)
+{
+	gfs2_assert(GFS2_SB(inode), (change >= 0 || inode->i_blocks > -change));
+	change *= (GFS2_SB(inode)->sd_sb.sb_bsize/GFS2_BASIC_BLOCK);
+	inode->i_blocks += change;
+}
+
 static inline int gfs2_check_inum(const struct gfs2_inode *ip, u64 no_addr,
 				  u64 no_formal_ino)
 {
diff --git a/fs/gfs2/ops_inode.c b/fs/gfs2/ops_inode.c
index e874129..301c945 100644
--- a/fs/gfs2/ops_inode.c
+++ b/fs/gfs2/ops_inode.c
@@ -981,8 +981,9 @@ static int setattr_chown(struct inode *inode, struct iattr *attr)
 	brelse(dibh);
 
 	if (ouid != NO_QUOTA_CHANGE || ogid != NO_QUOTA_CHANGE) {
-		gfs2_quota_change(ip, -ip->i_di.di_blocks, ouid, ogid);
-		gfs2_quota_change(ip, ip->i_di.di_blocks, nuid, ngid);
+		u64 blocks = gfs2_get_inode_blocks(&ip->i_inode);
+		gfs2_quota_change(ip, -blocks, ouid, ogid);
+		gfs2_quota_change(ip, blocks, nuid, ngid);
 	}
 
 out_end_trans:
-- 
1.5.1.2

[Cluster-devel] [PATCH 29/48] [GFS2] Get inode buffer only once per block map call

[swhiteho]

From: Steven Whitehouse <swhiteho@redhat.com>

In the case that we needed to grow the height of the metadata tree
we were looking up the inode buffer and then brelse()ing it despite
the fact that it is needed later in the block map process.

This patch ensures that we look up the inode's buffer once and only
once during the block map process.

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

diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index e27e660..f1f38ca 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -188,50 +188,45 @@ static int build_height(struct inode *inode, struct metapath *mp, unsigned heigh
 {
 	struct gfs2_inode *ip = GFS2_I(inode);
 	unsigned new_height = height - ip->i_height;
-	struct buffer_head *dibh;
+	struct buffer_head *dibh = mp->mp_bh[0];
 	struct gfs2_dinode *di;
-	int error;
 	__be64 *bp;
 	u64 bn;
 	unsigned n, i = 0;
 
-	if (height <= ip->i_height)
-		return 0;
-
-	error = gfs2_meta_inode_buffer(ip, &dibh);
-	if (error)
-		return error;
+	BUG_ON(height <= ip->i_height);
 
 	do {
 		n = new_height - i;
 		bn = gfs2_alloc_block(ip, &n);
 		gfs2_trans_add_unrevoke(GFS2_SB(inode), bn, n);
 		do {
-			mp->mp_bh[i] = gfs2_meta_new(ip->i_gl, bn++);
-			gfs2_trans_add_bh(ip->i_gl, mp->mp_bh[i], 1);
+			mp->mp_bh[i + 1] = gfs2_meta_new(ip->i_gl, bn++);
+			gfs2_trans_add_bh(ip->i_gl, mp->mp_bh[i + 1], 1);
 			i++;
 		} while(i < n);
 	} while(i < new_height);
 
 	n = 0;
-	bn = mp->mp_bh[0]->b_blocknr;
+	bn = mp->mp_bh[1]->b_blocknr;
 	if (new_height > 1) {
 		for(; n < new_height-1; n++) {
-			gfs2_metatype_set(mp->mp_bh[n], GFS2_METATYPE_IN,
+			gfs2_metatype_set(mp->mp_bh[n + 1], GFS2_METATYPE_IN,
 					  GFS2_FORMAT_IN);
-			gfs2_buffer_clear_tail(mp->mp_bh[n],
+			gfs2_buffer_clear_tail(mp->mp_bh[n + 1],
 					       sizeof(struct gfs2_meta_header));
-			bp = (__be64 *)(mp->mp_bh[n]->b_data +
+			bp = (__be64 *)(mp->mp_bh[n + 1]->b_data +
 				     sizeof(struct gfs2_meta_header));
-			*bp = cpu_to_be64(mp->mp_bh[n+1]->b_blocknr);
-			brelse(mp->mp_bh[n]);
-			mp->mp_bh[n] = NULL;
+			*bp = cpu_to_be64(mp->mp_bh[n+2]->b_blocknr);
+			brelse(mp->mp_bh[n+1]);
+			mp->mp_bh[n+1] = NULL;
 		}
 	}
-	gfs2_metatype_set(mp->mp_bh[n], GFS2_METATYPE_IN, GFS2_FORMAT_IN);
-	gfs2_buffer_copy_tail(mp->mp_bh[n], sizeof(struct gfs2_meta_header),
+	gfs2_metatype_set(mp->mp_bh[n+1], GFS2_METATYPE_IN, GFS2_FORMAT_IN);
+	gfs2_buffer_copy_tail(mp->mp_bh[n+1], sizeof(struct gfs2_meta_header),
 			      dibh, sizeof(struct gfs2_dinode));
-	brelse(mp->mp_bh[n]);
+	brelse(mp->mp_bh[n+1]);
+	mp->mp_bh[n+1] = NULL;
 	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
 	di = (struct gfs2_dinode *)dibh->b_data;
 	gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode));
@@ -240,8 +235,7 @@ static int build_height(struct inode *inode, struct metapath *mp, unsigned heigh
 	gfs2_add_inode_blocks(&ip->i_inode, new_height);
 	di->di_height = cpu_to_be16(ip->i_height);
 	di->di_blocks = cpu_to_be64(gfs2_get_inode_blocks(&ip->i_inode));
-	brelse(dibh);
-	return error;
+	return 0;
 }
 
 /**
@@ -391,11 +385,7 @@ static int lookup_metapath(struct inode *inode, struct metapath *mp,
 	struct gfs2_inode *ip = GFS2_I(inode);
 	unsigned int end_of_metadata = ip->i_height - 1;
 	unsigned int x;
-	int ret = gfs2_meta_inode_buffer(ip, &bh);
-	if (ret)
-		return ret;
-
-	mp->mp_bh[0] = bh;
+	int ret;
 
 	for (x = 0; x < end_of_metadata; x++) {
 		lookup_block(ip, x, mp, create, new, dblock);
@@ -515,6 +505,10 @@ int gfs2_block_map(struct inode *inode, sector_t lblock,
 	}
 	size = (lblock + 1) * bsize;
 
+	error = gfs2_meta_inode_buffer(ip, &mp.mp_bh[0]);
+	if (error)
+		goto out_fail;
+
 	if (size > arr[ip->i_height]) {
 		u8 height = ip->i_height;
 		if (!create)
-- 
1.5.1.2

[Cluster-devel] [PATCH 30/48] [GFS2] Fix bug where we called drop_bh incorrectly

[swhiteho]

From: Steven Whitehouse <swhiteho@redhat.com>

As a result of an earlier patch, drop_bh was being called in cases
when it shouldn't have been. Since we never have a gh in the drop
case and we always have a gh in the promote case, we can use that
extra information to tell which case has been seen.

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

diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c
index d00dc37..63981e2 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -765,7 +765,6 @@ static void drop_bh(struct gfs2_glock *gl, unsigned int ret)
 {
 	struct gfs2_sbd *sdp = gl->gl_sbd;
 	const struct gfs2_glock_operations *glops = gl->gl_ops;
-	struct gfs2_holder *gh = gl->gl_req_gh;
 
 	gfs2_assert_warn(sdp, test_bit(GLF_LOCK, &gl->gl_flags));
 	gfs2_assert_warn(sdp, list_empty(&gl->gl_holders));
@@ -776,23 +775,11 @@ static void drop_bh(struct gfs2_glock *gl, unsigned int ret)
 	if (glops->go_inval)
 		glops->go_inval(gl, DIO_METADATA);
 
-	if (gh) {
-		spin_lock(&gl->gl_spin);
-		list_del_init(&gh->gh_list);
-		gh->gh_error = 0;
-		spin_unlock(&gl->gl_spin);
-	}
-
 	spin_lock(&gl->gl_spin);
 	gfs2_demote_wake(gl);
-	gl->gl_req_gh = NULL;
 	clear_bit(GLF_LOCK, &gl->gl_flags);
 	spin_unlock(&gl->gl_spin);
-
 	gfs2_glock_put(gl);
-
-	if (gh)
-		gfs2_holder_wake(gh);
 }
 
 /**
@@ -810,7 +797,7 @@ static void xmote_bh(struct gfs2_glock *gl, unsigned int ret)
 	int prev_state = gl->gl_state;
 	int op_done = 1;
 
-	if ((ret & LM_OUT_ST_MASK) == LM_ST_UNLOCKED) {
+	if (!gh && (ret & LM_OUT_ST_MASK) == LM_ST_UNLOCKED) {
 		drop_bh(gl, ret);
 		return;
 	}
-- 
1.5.1.2

[Cluster-devel] [PATCH 31/48] [GFS2] be*_add_cpu conversion

[swhiteho]

From: Marcin Slusarz <marcin.slusarz@gmail.com>

replace all:
big_endian_variable = cpu_to_beX(beX_to_cpu(big_endian_variable) +
					expression_in_cpu_byteorder);
with:
	beX_add_cpu(&big_endian_variable, expression_in_cpu_byteorder);
generated with semantic patch

Signed-off-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/dir.c b/fs/gfs2/dir.c
index 862aa32..34dc8df 100644
--- a/fs/gfs2/dir.c
+++ b/fs/gfs2/dir.c
@@ -1021,13 +1021,13 @@ static int dir_split_leaf(struct inode *inode, const struct qstr *name)
 
 			new->de_inum = dent->de_inum; /* No endian worries */
 			new->de_type = dent->de_type; /* No endian worries */
-			nleaf->lf_entries = cpu_to_be16(be16_to_cpu(nleaf->lf_entries)+1);
+			be16_add_cpu(&nleaf->lf_entries, 1);
 
 			dirent_del(dip, obh, prev, dent);
 
 			if (!oleaf->lf_entries)
 				gfs2_consist_inode(dip);
-			oleaf->lf_entries = cpu_to_be16(be16_to_cpu(oleaf->lf_entries)-1);
+			be16_add_cpu(&oleaf->lf_entries, -1);
 
 			if (!prev)
 				prev = dent;
@@ -1614,7 +1614,7 @@ int gfs2_dir_add(struct inode *inode, const struct qstr *name,
 			dent->de_type = cpu_to_be16(type);
 			if (ip->i_di.di_flags & GFS2_DIF_EXHASH) {
 				leaf = (struct gfs2_leaf *)bh->b_data;
-				leaf->lf_entries = cpu_to_be16(be16_to_cpu(leaf->lf_entries) + 1);
+				be16_add_cpu(&leaf->lf_entries, 1);
 			}
 			brelse(bh);
 			error = gfs2_meta_inode_buffer(ip, &bh);
-- 
1.5.1.2

[Cluster-devel] [PATCH 32/48] [GFS2] gfs2/ops_file.c should #include "ops_inode.h"

[swhiteho]

From: Adrian Bunk <bunk@kernel.org>

Every file should include the headers containing the prototypes for
its global functions (in this case for gfs2_set_inode_flags()).

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/ops_file.c b/fs/gfs2/ops_file.c
index f97a8b8..2b25a5f 100644
--- a/fs/gfs2/ops_file.c
+++ b/fs/gfs2/ops_file.c
@@ -38,6 +38,7 @@
 #include "util.h"
 #include "eaops.h"
 #include "ops_address.h"
+#include "ops_inode.h"
 
 /**
  * gfs2_llseek - seek to a location in a file
-- 
1.5.1.2

[Cluster-devel] [PATCH 33/48] [GFS2] proper extern for gfs2/locking/dlm/mount.c:gdlm_ops

[swhiteho]

From: Adrian Bunk <bunk@kernel.org>

This patch adds a proper extern declaration for gdlm_ops in
fs/gfs2/locking/dlm/lock_dlm.h

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/locking/dlm/lock_dlm.h b/fs/gfs2/locking/dlm/lock_dlm.h
index 9e8265d..58fcf8c 100644
--- a/fs/gfs2/locking/dlm/lock_dlm.h
+++ b/fs/gfs2/locking/dlm/lock_dlm.h
@@ -183,5 +183,10 @@ int gdlm_plock_get(void *, struct lm_lockname *, struct file *,
 		struct file_lock *);
 int gdlm_punlock(void *, struct lm_lockname *, struct file *,
 		struct file_lock *);
+
+/* mount.c */
+
+extern const struct lm_lockops gdlm_ops;
+
 #endif
 
diff --git a/fs/gfs2/locking/dlm/main.c b/fs/gfs2/locking/dlm/main.c
index a0e7eda..36a2258 100644
--- a/fs/gfs2/locking/dlm/main.c
+++ b/fs/gfs2/locking/dlm/main.c
@@ -11,8 +11,6 @@
 
 #include "lock_dlm.h"
 
-extern struct lm_lockops gdlm_ops;
-
 static int __init init_lock_dlm(void)
 {
 	int error;
diff --git a/fs/gfs2/locking/dlm/sysfs.c b/fs/gfs2/locking/dlm/sysfs.c
index a87b098..8479da4 100644
--- a/fs/gfs2/locking/dlm/sysfs.c
+++ b/fs/gfs2/locking/dlm/sysfs.c
@@ -12,8 +12,6 @@
 
 #include "lock_dlm.h"
 
-extern struct lm_lockops gdlm_ops;
-
 static ssize_t proto_name_show(struct gdlm_ls *ls, char *buf)
 {
 	return sprintf(buf, "%s\n", gdlm_ops.lm_proto_name);
-- 
1.5.1.2

[Cluster-devel] [PATCH 34/48] [GFS2] Fix a page lock / glock deadlock

[swhiteho]

From: Steven Whitehouse <swhiteho@redhat.com>

We've previously been using a "try lock" in readpage on the basis that
it would prevent deadlocks due to the inverted lock ordering (our normal
lock ordering is glock first and then page lock). Unfortunately tests
have shown that this isn't enough. If the glock has a demote request
queued such that run_queue() in the glock code tries to do a demote when
its called under readpage then it will try and write out all the dirty
pages which requires locking them. This then deadlocks with the page
locked by readpage.

The solution is to always require two calls into readpage. The first
unlocks the page, gets the glock and returns AOP_TRUNCATED_PAGE, the
second does the actual readpage and unlocks the glock & page as
required.

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

diff --git a/fs/gfs2/glock.h b/fs/gfs2/glock.h
index ace5770..cdad3e6 100644
--- a/fs/gfs2/glock.h
+++ b/fs/gfs2/glock.h
@@ -32,24 +32,23 @@
 #define GLR_TRYFAILED		13
 #define GLR_CANCELED		14
 
-static inline int gfs2_glock_is_locked_by_me(struct gfs2_glock *gl)
+static inline struct gfs2_holder *gfs2_glock_is_locked_by_me(struct gfs2_glock *gl)
 {
 	struct gfs2_holder *gh;
-	int locked = 0;
 	struct pid *pid;
 
 	/* Look in glock's list of holders for one with current task as owner */
 	spin_lock(&gl->gl_spin);
 	pid = task_pid(current);
 	list_for_each_entry(gh, &gl->gl_holders, gh_list) {
-		if (gh->gh_owner_pid == pid) {
-			locked = 1;
-			break;
-		}
+		if (gh->gh_owner_pid == pid)
+			goto out;
 	}
+	gh = NULL;
+out:
 	spin_unlock(&gl->gl_spin);
 
-	return locked;
+	return gh;
 }
 
 static inline int gfs2_glock_is_held_excl(struct gfs2_glock *gl)
diff --git a/fs/gfs2/inode.c b/fs/gfs2/inode.c
index 5f50dd5..810ff02 100644
--- a/fs/gfs2/inode.c
+++ b/fs/gfs2/inode.c
@@ -493,7 +493,7 @@ struct inode *gfs2_lookupi(struct inode *dir, const struct qstr *name,
 		return dir;
 	}
 
-	if (gfs2_glock_is_locked_by_me(dip->i_gl) == 0) {
+	if (gfs2_glock_is_locked_by_me(dip->i_gl) == NULL) {
 		error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, &d_gh);
 		if (error)
 			return ERR_PTR(error);
diff --git a/fs/gfs2/ops_address.c b/fs/gfs2/ops_address.c
index 7523999..fbb4a6a 100644
--- a/fs/gfs2/ops_address.c
+++ b/fs/gfs2/ops_address.c
@@ -508,23 +508,26 @@ static int __gfs2_readpage(void *file, struct page *page)
 static int gfs2_readpage(struct file *file, struct page *page)
 {
 	struct gfs2_inode *ip = GFS2_I(page->mapping->host);
-	struct gfs2_holder gh;
+	struct gfs2_holder *gh;
 	int error;
 
-	gfs2_holder_init(ip->i_gl, LM_ST_SHARED, GL_ATIME|LM_FLAG_TRY_1CB, &gh);
-	error = gfs2_glock_nq_atime(&gh);
-	if (unlikely(error)) {
+	gh = gfs2_glock_is_locked_by_me(ip->i_gl);
+	if (!gh) {
+		gh = kmalloc(sizeof(struct gfs2_holder), GFP_NOFS);
+		if (!gh)
+			return -ENOBUFS;
+		gfs2_holder_init(ip->i_gl, LM_ST_SHARED, GL_ATIME, gh);
 		unlock_page(page);
-		goto out;
+		error = gfs2_glock_nq_atime(gh);
+		if (likely(error != 0))
+			goto out;
+		return AOP_TRUNCATED_PAGE;
 	}
 	error = __gfs2_readpage(file, page);
-	gfs2_glock_dq(&gh);
+	gfs2_glock_dq(gh);
 out:
-	gfs2_holder_uninit(&gh);
-	if (error == GLR_TRYFAILED) {
-		yield();
-		return AOP_TRUNCATED_PAGE;
-	}
+	gfs2_holder_uninit(gh);
+	kfree(gh);
 	return error;
 }
 
@@ -826,7 +829,7 @@ static int gfs2_write_end(struct file *file, struct address_space *mapping,
 	unsigned int to = from + len;
 	int ret;
 
-	BUG_ON(gfs2_glock_is_locked_by_me(ip->i_gl) == 0);
+	BUG_ON(gfs2_glock_is_locked_by_me(ip->i_gl) == NULL);
 
 	ret = gfs2_meta_inode_buffer(ip, &dibh);
 	if (unlikely(ret)) {
diff --git a/fs/gfs2/ops_dentry.c b/fs/gfs2/ops_dentry.c
index 793e334..4a5e676 100644
--- a/fs/gfs2/ops_dentry.c
+++ b/fs/gfs2/ops_dentry.c
@@ -43,7 +43,7 @@ static int gfs2_drevalidate(struct dentry *dentry, struct nameidata *nd)
 	struct gfs2_holder d_gh;
 	struct gfs2_inode *ip = NULL;
 	int error;
-	int had_lock=0;
+	int had_lock = 0;
 
 	if (inode) {
 		if (is_bad_inode(inode))
@@ -54,7 +54,7 @@ static int gfs2_drevalidate(struct dentry *dentry, struct nameidata *nd)
 	if (sdp->sd_args.ar_localcaching)
 		goto valid;
 
-	had_lock = gfs2_glock_is_locked_by_me(dip->i_gl);
+	had_lock = (gfs2_glock_is_locked_by_me(dip->i_gl) != NULL);
 	if (!had_lock) {
 		error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, &d_gh);
 		if (error)
diff --git a/fs/gfs2/ops_inode.c b/fs/gfs2/ops_inode.c
index 301c945..af7097a 100644
--- a/fs/gfs2/ops_inode.c
+++ b/fs/gfs2/ops_inode.c
@@ -898,7 +898,7 @@ static int gfs2_permission(struct inode *inode, int mask, struct nameidata *nd)
 	int error;
 	int unlock = 0;
 
-	if (gfs2_glock_is_locked_by_me(ip->i_gl) == 0) {
+	if (gfs2_glock_is_locked_by_me(ip->i_gl) == NULL) {
 		error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_ANY, &i_gh);
 		if (error)
 			return error;
@@ -1065,7 +1065,7 @@ static int gfs2_getattr(struct vfsmount *mnt, struct dentry *dentry,
 	int error;
 	int unlock = 0;
 
-	if (gfs2_glock_is_locked_by_me(ip->i_gl) == 0) {
+	if (gfs2_glock_is_locked_by_me(ip->i_gl) == NULL) {
 		error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_ANY, &gh);
 		if (error)
 			return error;
-- 
1.5.1.2

[Cluster-devel] [PATCH 35/48] [GFS2] Allow bmap to allocate extents

[swhiteho]

From: Steven Whitehouse <swhiteho@redhat.com>

We've supported mapping of extents when no block allocation is required
for some time. This patch extends that to mapping of extents when an
allocation has been requested. In that case we try to allocate as many
blocks as are requested, but we might return fewer in case there is
something preventing us from returning the complete amount (e.g. an
already allocated block is in the way).

Currently the only code path which can actually request multiple data
blocks in a single bmap call is the page_mkwrite path and even then it
only happens if there are multiple blocks per page. What this patch does
do however, is merge the allocation requests for metadata (growing the
metadata tree in either height or depth) with the allocation of the data
blocks in the case that both are needed. This results in lower overheads
even in the single block allocation case.

The one thing which we can't handle here at the moment is unstuffing. I
would like to be able to do that, but the problem which arises is that
in order to unstuff one has to get a locked page from the page cache
which results in locking problems in the (usual) case that the caller is
holding the page lock on the page it wishes to map. So that case will
have to be addressed in future patches.

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

diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index f1f38ca..c1ee635 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -175,74 +175,13 @@ out:
 	return error;
 }
 
-/**
- * build_height - Build a metadata tree of the requested height
- * @ip: The GFS2 inode
- * @height: The height to build to
- *
- *
- * Returns: errno
- */
-
-static int build_height(struct inode *inode, struct metapath *mp, unsigned height)
-{
-	struct gfs2_inode *ip = GFS2_I(inode);
-	unsigned new_height = height - ip->i_height;
-	struct buffer_head *dibh = mp->mp_bh[0];
-	struct gfs2_dinode *di;
-	__be64 *bp;
-	u64 bn;
-	unsigned n, i = 0;
-
-	BUG_ON(height <= ip->i_height);
-
-	do {
-		n = new_height - i;
-		bn = gfs2_alloc_block(ip, &n);
-		gfs2_trans_add_unrevoke(GFS2_SB(inode), bn, n);
-		do {
-			mp->mp_bh[i + 1] = gfs2_meta_new(ip->i_gl, bn++);
-			gfs2_trans_add_bh(ip->i_gl, mp->mp_bh[i + 1], 1);
-			i++;
-		} while(i < n);
-	} while(i < new_height);
-
-	n = 0;
-	bn = mp->mp_bh[1]->b_blocknr;
-	if (new_height > 1) {
-		for(; n < new_height-1; n++) {
-			gfs2_metatype_set(mp->mp_bh[n + 1], GFS2_METATYPE_IN,
-					  GFS2_FORMAT_IN);
-			gfs2_buffer_clear_tail(mp->mp_bh[n + 1],
-					       sizeof(struct gfs2_meta_header));
-			bp = (__be64 *)(mp->mp_bh[n + 1]->b_data +
-				     sizeof(struct gfs2_meta_header));
-			*bp = cpu_to_be64(mp->mp_bh[n+2]->b_blocknr);
-			brelse(mp->mp_bh[n+1]);
-			mp->mp_bh[n+1] = NULL;
-		}
-	}
-	gfs2_metatype_set(mp->mp_bh[n+1], GFS2_METATYPE_IN, GFS2_FORMAT_IN);
-	gfs2_buffer_copy_tail(mp->mp_bh[n+1], sizeof(struct gfs2_meta_header),
-			      dibh, sizeof(struct gfs2_dinode));
-	brelse(mp->mp_bh[n+1]);
-	mp->mp_bh[n+1] = NULL;
-	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
-	di = (struct gfs2_dinode *)dibh->b_data;
-	gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode));
-	*(__be64 *)(di + 1) = cpu_to_be64(bn);
-	ip->i_height += new_height;
-	gfs2_add_inode_blocks(&ip->i_inode, new_height);
-	di->di_height = cpu_to_be16(ip->i_height);
-	di->di_blocks = cpu_to_be64(gfs2_get_inode_blocks(&ip->i_inode));
-	return 0;
-}
 
 /**
  * find_metapath - Find path through the metadata tree
- * @ip: The inode pointer
+ * @sdp: The superblock
  * @mp: The metapath to return the result in
  * @block: The disk block to look up
+ * @height: The pre-calculated height of the metadata tree
  *
  *   This routine returns a struct metapath structure that defines a path
  *   through the metadata of inode "ip" to get to block "block".
@@ -297,17 +236,27 @@ static int build_height(struct inode *inode, struct metapath *mp, unsigned heigh
  *
  */
 
-static void find_metapath(struct gfs2_inode *ip, u64 block,
-			  struct metapath *mp)
+static void find_metapath(const struct gfs2_sbd *sdp, u64 block,
+			  struct metapath *mp, unsigned int height)
 {
-	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
 	unsigned int i;
 
-	for (i = ip->i_height; i--;)
+	for (i = height; i--;)
 		mp->mp_list[i] = do_div(block, sdp->sd_inptrs);
 
 }
 
+static inline unsigned int zero_metapath_length(const struct metapath *mp,
+						unsigned height)
+{
+	unsigned int i;
+	for (i = 0; i < height - 1; i++) {
+		if (mp->mp_list[i] != 0)
+			return i;
+	}
+	return height;
+}
+
 /**
  * metapointer - Return pointer to start of metadata in a buffer
  * @height: The metadata height (0 = dinode)
@@ -318,95 +267,62 @@ static void find_metapath(struct gfs2_inode *ip, u64 block,
  * metadata tree.
  */
 
-static inline __be64 *metapointer(int *boundary, unsigned int height,
-				  const struct metapath *mp)
+static inline __be64 *metapointer(unsigned int height, const struct metapath *mp)
 {
 	struct buffer_head *bh = mp->mp_bh[height];
 	unsigned int head_size = (height > 0) ?
 		sizeof(struct gfs2_meta_header) : sizeof(struct gfs2_dinode);
-	__be64 *ptr;
-	*boundary = 0;
-	ptr = ((__be64 *)(bh->b_data + head_size)) + mp->mp_list[height];
-	if (ptr + 1 == (__be64 *)(bh->b_data + bh->b_size))
-		*boundary = 1;
-	return ptr;
+	return ((__be64 *)(bh->b_data + head_size)) + mp->mp_list[height];
 }
 
 /**
- * lookup_block - Get the next metadata block in metadata tree
- * @ip: The GFS2 inode
- * @height: The height of the tree (0 = dinode)
+ * lookup_metapath - Walk the metadata tree to a specific point
+ * @ip: The inode
  * @mp: The metapath
- * @create: Non-zero if we may create a new meatdata block
- * @new: Used to indicate if we did create a new metadata block
- * @block: the returned disk block number
  *
- * Given a metatree, complete to a particular height, checks to see if the next
- * height of the tree exists. If not the next height of the tree is created.
- * The block number of the next height of the metadata tree is returned.
+ * Assumes that the inode's buffer has already been looked up and
+ * hooked onto mp->mp_bh[0] and that the metapath has been initialised
+ * by find_metapath().
  *
+ * If this function encounters part of the tree which has not been
+ * allocated, it returns the current height of the tree at the point
+ * at which it found the unallocated block. Blocks which are found are
+ * added to the mp->mp_bh[] list.
+ *
+ * Returns: error or height of metadata tree
  */
 
-static int lookup_block(struct gfs2_inode *ip, unsigned int height,
-			struct metapath *mp, int create,
-			int *new, u64 *block)
+static int lookup_metapath(struct gfs2_inode *ip, struct metapath *mp)
 {
-	int boundary;
-	__be64 *ptr = metapointer(&boundary, height, mp);
-	unsigned int n = 1;
-
-	if (*ptr) {
-		*block = be64_to_cpu(*ptr);
-		return boundary;
-	}
-
-	*block = 0;
-
-	if (!create)
-		return 0;
-
-	*block = gfs2_alloc_block(ip, &n);
-	if (height != ip->i_height - 1 || gfs2_is_dir(ip))
-		gfs2_trans_add_unrevoke(GFS2_SB(&ip->i_inode), *block, 1);
-
-	gfs2_trans_add_bh(ip->i_gl, mp->mp_bh[height], 1);
-
-	*ptr = cpu_to_be64(*block);
-	gfs2_add_inode_blocks(&ip->i_inode, 1);
-
-	*new = 1;
-	return 0;
-}
-
-static int lookup_metapath(struct inode *inode, struct metapath *mp,
-			   int create, int *new, u64 *dblock)
-{
-	struct buffer_head *bh;
-	struct gfs2_inode *ip = GFS2_I(inode);
 	unsigned int end_of_metadata = ip->i_height - 1;
 	unsigned int x;
+	__be64 *ptr;
+	u64 dblock;
 	int ret;
 
 	for (x = 0; x < end_of_metadata; x++) {
-		lookup_block(ip, x, mp, create, new, dblock);
-		if (!*dblock)
-			return 0;
+		ptr = metapointer(x, mp);
+		dblock = be64_to_cpu(*ptr);
+		if (!dblock)
+			return x + 1;
 
-		ret = gfs2_meta_indirect_buffer(ip, x+1, *dblock, *new, &mp->mp_bh[x+1]);
+		ret = gfs2_meta_indirect_buffer(ip, x+1, dblock, 0, &mp->mp_bh[x+1]);
 		if (ret)
 			return ret;
 	}
 
-	return lookup_block(ip, end_of_metadata, mp, create, new, dblock);
+	return ip->i_height;
 }
 
-static void release_metapath(struct metapath *mp)
+static inline void release_metapath(struct metapath *mp)
 {
 	int i;
 
-	for (i = 0; i < GFS2_MAX_META_HEIGHT; i++)
-		if (mp->mp_bh[i])
-			brelse(mp->mp_bh[i]);
+	for (i = 0; i < GFS2_MAX_META_HEIGHT; i++) {
+		if (mp->mp_bh[i] == NULL)
+			break;
+		brelse(mp->mp_bh[i]);
+	}
 }
 
 /**
@@ -445,32 +361,208 @@ static inline unsigned int gfs2_extent_length(void *start, unsigned int len, __b
 	return (ptr - first);
 }
 
-static inline void bmap_lock(struct inode *inode, int create)
+static inline void bmap_lock(struct gfs2_inode *ip, int create)
 {
-	struct gfs2_inode *ip = GFS2_I(inode);
 	if (create)
 		down_write(&ip->i_rw_mutex);
 	else
 		down_read(&ip->i_rw_mutex);
 }
 
-static inline void bmap_unlock(struct inode *inode, int create)
+static inline void bmap_unlock(struct gfs2_inode *ip, int create)
 {
-	struct gfs2_inode *ip = GFS2_I(inode);
 	if (create)
 		up_write(&ip->i_rw_mutex);
 	else
 		up_read(&ip->i_rw_mutex);
 }
 
+static inline __be64 *gfs2_indirect_init(struct metapath *mp,
+					 struct gfs2_glock *gl, unsigned int i,
+					 unsigned offset, u64 bn)
+{
+	__be64 *ptr = (__be64 *)(mp->mp_bh[i - 1]->b_data +
+		       ((i > 1) ? sizeof(struct gfs2_meta_header) :
+				 sizeof(struct gfs2_dinode)));
+	BUG_ON(i < 1);
+	BUG_ON(mp->mp_bh[i] != NULL);
+	mp->mp_bh[i] = gfs2_meta_new(gl, bn);
+	gfs2_trans_add_bh(gl, mp->mp_bh[i], 1);
+	gfs2_metatype_set(mp->mp_bh[i], GFS2_METATYPE_IN, GFS2_FORMAT_IN);
+	gfs2_buffer_clear_tail(mp->mp_bh[i], sizeof(struct gfs2_meta_header));
+	ptr += offset;
+	*ptr = cpu_to_be64(bn);
+	return ptr;
+}
+
+enum alloc_state {
+	ALLOC_DATA = 0,
+	ALLOC_GROW_DEPTH = 1,
+	ALLOC_GROW_HEIGHT = 2,
+	/* ALLOC_UNSTUFF = 3,   TBD and rather complicated */
+};
+
+/**
+ * gfs2_bmap_alloc - Build a metadata tree of the requested height
+ * @inode: The GFS2 inode
+ * @lblock: The logical starting block of the extent
+ * @bh_map: This is used to return the mapping details
+ * @mp: The metapath
+ * @sheight: The starting height (i.e. whats already mapped)
+ * @height: The height to build to
+ * @maxlen: The max number of data blocks to alloc
+ *
+ * In this routine we may have to alloc:
+ *   i) Indirect blocks to grow the metadata tree height
+ *  ii) Indirect blocks to fill in lower part of the metadata tree
+ * iii) Data blocks
+ *
+ * The function is in two parts. The first part works out the total
+ * number of blocks which we need. The second part does the actual
+ * allocation asking for an extent at a time (if enough contiguous free
+ * blocks are available, there will only be one request per bmap call)
+ * and uses the state machine to initialise the blocks in order.
+ *
+ * Returns: errno on error
+ */
+
+static int gfs2_bmap_alloc(struct inode *inode, const sector_t lblock,
+			   struct buffer_head *bh_map, struct metapath *mp,
+			   const unsigned int sheight,
+			   const unsigned int height,
+			   const unsigned int maxlen)
+{
+	struct gfs2_inode *ip = GFS2_I(inode);
+	struct gfs2_sbd *sdp = GFS2_SB(inode);
+	struct buffer_head *dibh = mp->mp_bh[0];
+	u64 bn, dblock = 0;
+	unsigned n, i, blks, alloced = 0, iblks = 0, zmpl = 0;
+	unsigned dblks = 0;
+	unsigned ptrs_per_blk;
+	const unsigned end_of_metadata = height - 1;
+	int eob = 0;
+	enum alloc_state state;
+	__be64 *ptr;
+	__be64 zero_bn = 0;
+
+	BUG_ON(sheight < 1);
+	BUG_ON(dibh == NULL);
+
+	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
+
+	if (height == sheight) {
+		struct buffer_head *bh;
+		/* Bottom indirect block exists, find unalloced extent size */
+		ptr = metapointer(end_of_metadata, mp);
+		bh = mp->mp_bh[end_of_metadata];
+		dblks = gfs2_extent_length(bh->b_data, bh->b_size, ptr, maxlen,
+					   &eob);
+		BUG_ON(dblks < 1);
+		state = ALLOC_DATA;
+	} else {
+		/* Need to allocate indirect blocks */
+		ptrs_per_blk = height > 1 ? sdp->sd_inptrs : sdp->sd_diptrs;
+		dblks = min(maxlen, ptrs_per_blk - mp->mp_list[end_of_metadata]);
+		if (height == ip->i_height) {
+			/* Writing into existing tree, extend tree down */
+			iblks = height - sheight;
+			state = ALLOC_GROW_DEPTH;
+		} else {
+			/* Building up tree height */
+			state = ALLOC_GROW_HEIGHT;
+			iblks = height - ip->i_height;
+			zmpl = zero_metapath_length(mp, height);
+			iblks -= zmpl;
+			iblks += height;
+		}
+	}
+
+	/* start of the second part of the function (state machine) */
+
+	blks = dblks + iblks;
+	i = sheight;
+	do {
+		n = blks - alloced;
+		bn = gfs2_alloc_block(ip, &n);
+		alloced += n;
+		if (state != ALLOC_DATA || gfs2_is_jdata(ip))
+			gfs2_trans_add_unrevoke(sdp, bn, n);
+		switch (state) {
+		/* Growing height of tree */
+		case ALLOC_GROW_HEIGHT:
+			if (i == 1) {
+				ptr = (__be64 *)(dibh->b_data +
+						 sizeof(struct gfs2_dinode));
+				zero_bn = *ptr;
+			}
+			for (; i - 1 < height - ip->i_height && n > 0; i++, n--)
+				gfs2_indirect_init(mp, ip->i_gl, i, 0, bn++);
+			if (i - 1 == height - ip->i_height) {
+				i--;
+				gfs2_buffer_copy_tail(mp->mp_bh[i],
+						sizeof(struct gfs2_meta_header),
+						dibh, sizeof(struct gfs2_dinode));
+				gfs2_buffer_clear_tail(dibh,
+						sizeof(struct gfs2_dinode) +
+						sizeof(__be64));
+				ptr = (__be64 *)(mp->mp_bh[i]->b_data +
+					sizeof(struct gfs2_meta_header));
+				*ptr = zero_bn;
+				state = ALLOC_GROW_DEPTH;
+				for(i = zmpl; i < height; i++) {
+					if (mp->mp_bh[i] == NULL)
+						break;
+					brelse(mp->mp_bh[i]);
+					mp->mp_bh[i] = NULL;
+				}
+				i = zmpl;
+			}
+			if (n == 0)
+				break;
+		/* Branching from existing tree */
+		case ALLOC_GROW_DEPTH:
+			if (i > 1 && i < height)
+				gfs2_trans_add_bh(ip->i_gl, mp->mp_bh[i-1], 1);
+			for (; i < height && n > 0; i++, n--)
+				gfs2_indirect_init(mp, ip->i_gl, i,
+						   mp->mp_list[i-1], bn++);
+			if (i == height)
+				state = ALLOC_DATA;
+			if (n == 0)
+				break;
+		/* Tree complete, adding data blocks */
+		case ALLOC_DATA:
+			BUG_ON(n > dblks);
+			BUG_ON(mp->mp_bh[end_of_metadata] == NULL);
+			gfs2_trans_add_bh(ip->i_gl, mp->mp_bh[end_of_metadata], 1);
+			dblks = n;
+			ptr = metapointer(end_of_metadata, mp);
+			dblock = bn;
+			while (n-- > 0)
+				*ptr++ = cpu_to_be64(bn++);
+			break;
+		}
+	} while (state != ALLOC_DATA);
+
+	ip->i_height = height;
+	gfs2_add_inode_blocks(&ip->i_inode, alloced);
+	gfs2_dinode_out(ip, mp->mp_bh[0]->b_data);
+	map_bh(bh_map, inode->i_sb, dblock);
+	bh_map->b_size = dblks << inode->i_blkbits;
+	set_buffer_new(bh_map);
+	return 0;
+}
+
 /**
  * gfs2_block_map - Map a block from an inode to a disk block
  * @inode: The inode
  * @lblock: The logical block number
  * @bh_map: The bh to be mapped
+ * @create: True if its ok to alloc blocks to satify the request
  *
- * Find the block number on the current device which corresponds to an
- * inode's block. If the block had to be created, "new" will be set.
+ * Sets buffer_mapped() if successful, sets buffer_boundary() if a
+ * read of metadata will be required before the next block can be
+ * mapped. Sets buffer_new() if new blocks were allocated.
  *
  * Returns: errno
  */
@@ -481,21 +573,21 @@ int gfs2_block_map(struct inode *inode, sector_t lblock,
 	struct gfs2_inode *ip = GFS2_I(inode);
 	struct gfs2_sbd *sdp = GFS2_SB(inode);
 	unsigned int bsize = sdp->sd_sb.sb_bsize;
-	int error = 0;
-	int new = 0;
-	u64 dblock = 0;
-	int boundary;
-	unsigned int maxlen = bh_map->b_size >> inode->i_blkbits;
-	struct metapath mp;
-	u64 size;
+	const unsigned int maxlen = bh_map->b_size >> inode->i_blkbits;
 	const u64 *arr = sdp->sd_heightsize;
-	BUG_ON(maxlen == 0);
+	__be64 *ptr;
+	u64 size;
+	struct metapath mp;
+	int ret;
+	int eob;
+	unsigned int len;
+	struct buffer_head *bh;
+	u8 height;
 
-	if (gfs2_assert_warn(sdp, !gfs2_is_stuffed(ip)))
-		return 0;
+	BUG_ON(maxlen == 0);
 
 	memset(mp.mp_bh, 0, sizeof(mp.mp_bh));
-	bmap_lock(inode, create);
+	bmap_lock(ip, create);
 	clear_buffer_mapped(bh_map);
 	clear_buffer_new(bh_map);
 	clear_buffer_boundary(bh_map);
@@ -503,56 +595,50 @@ int gfs2_block_map(struct inode *inode, sector_t lblock,
 		bsize = sdp->sd_jbsize;
 		arr = sdp->sd_jheightsize;
 	}
-	size = (lblock + 1) * bsize;
 
-	error = gfs2_meta_inode_buffer(ip, &mp.mp_bh[0]);
-	if (error)
-		goto out_fail;
-
-	if (size > arr[ip->i_height]) {
-		u8 height = ip->i_height;
-		if (!create)
-			goto out_ok;
-		while (size > arr[height])
-			height++;
-		error = build_height(inode, &mp, height);
-		if (error)
-			goto out_fail;
-	}
+	ret = gfs2_meta_inode_buffer(ip, &mp.mp_bh[0]);
+	if (ret)
+		goto out;
 
-	find_metapath(ip, lblock, &mp);
-	error = lookup_metapath(inode, &mp, create, &new, &dblock);
-	if (error < 0)
-		goto out_fail;
-	boundary = error;
-
-	if (new) {
-		map_bh(bh_map, inode->i_sb, dblock);
-		if (boundary)
-			set_buffer_boundary(bh_map);
-		gfs2_trans_add_bh(ip->i_gl, mp.mp_bh[0], 1);
-		gfs2_dinode_out(ip, mp.mp_bh[0]->b_data);
-		set_buffer_new(bh_map);
-		goto out_ok;
-	}
+	height = ip->i_height;
+	size = (lblock + 1) * bsize;
+	while (size > arr[height])
+		height++;
+	find_metapath(sdp, lblock, &mp, height);
+	ret = 1;
+	if (height > ip->i_height || gfs2_is_stuffed(ip))
+		goto do_alloc;
+	ret = lookup_metapath(ip, &mp);
+	if (ret < 0)
+		goto out;
+	if (ret != ip->i_height)
+		goto do_alloc;
+	ptr = metapointer(ip->i_height - 1, &mp);
+	if (*ptr == 0)
+		goto do_alloc;
+	map_bh(bh_map, inode->i_sb, be64_to_cpu(*ptr));
+	bh = mp.mp_bh[ip->i_height - 1];
+	len = gfs2_extent_length(bh->b_data, bh->b_size, ptr, maxlen, &eob);
+	bh_map->b_size = (len << inode->i_blkbits);
+	if (eob)
+		set_buffer_boundary(bh_map);
+	ret = 0;
+out:
+	release_metapath(&mp);
+	bmap_unlock(ip, create);
+	return ret;
 
-	if (dblock) {
-		unsigned int len;
-		struct buffer_head *bh = mp.mp_bh[ip->i_height - 1];
-		__be64 *ptr = metapointer(&boundary, ip->i_height - 1, &mp);
-		map_bh(bh_map, inode->i_sb, dblock);
-		len = gfs2_extent_length(bh->b_data, bh->b_size, ptr, maxlen,
-					 &boundary);
-		bh_map->b_size = (len << inode->i_blkbits);
-		if (boundary)
-			set_buffer_boundary(bh_map);
+do_alloc:
+	/* All allocations are done here, firstly check create flag */
+	if (!create) {
+		BUG_ON(gfs2_is_stuffed(ip));
+		ret = 0;
+		goto out;
 	}
-out_ok:
-	error = 0;
-out_fail:
-	release_metapath(&mp);
-	bmap_unlock(inode, create);
-	return error;
+
+	/* At this point ret is the tree depth of already allocated blocks */
+	ret = gfs2_bmap_alloc(inode, lblock, bh_map, &mp, ret, height, maxlen);
+	goto out;
 }
 
 /*
@@ -568,7 +654,7 @@ int gfs2_extent_map(struct inode *inode, u64 lblock, int *new, u64 *dblock, unsi
 	BUG_ON(!dblock);
 	BUG_ON(!new);
 
-	bh.b_size = 1 << (inode->i_blkbits + 5);
+	bh.b_size = 1 << (inode->i_blkbits + (create ? 0 : 5));
 	ret = gfs2_block_map(inode, lblock, &bh, create);
 	*extlen = bh.b_size >> inode->i_blkbits;
 	*dblock = bh.b_blocknr;
@@ -835,38 +921,25 @@ static int do_grow(struct gfs2_inode *ip, u64 size)
 	if (error)
 		goto out_ipres;
 
+	error = gfs2_meta_inode_buffer(ip, &dibh);
+	if (error)
+		goto out_end_trans;
+
 	if (size > sdp->sd_sb.sb_bsize - sizeof(struct gfs2_dinode)) {
-		const u64 *arr = sdp->sd_heightsize;
 		if (gfs2_is_stuffed(ip)) {
 			error = gfs2_unstuff_dinode(ip, NULL);
 			if (error)
-				goto out_end_trans;
-		}
-
-		down_write(&ip->i_rw_mutex);
-		if (size > arr[ip->i_height]) {
-			struct metapath mp;
-			u8 height = ip->i_height;
-			while(size > arr[height])
-				height++;
-			error = build_height(&ip->i_inode, &mp, height);
+				goto out_brelse;
 		}
-		up_write(&ip->i_rw_mutex);
-		if (error)
-			goto out_end_trans;
 	}
 
 	ip->i_di.di_size = size;
 	ip->i_inode.i_mtime = ip->i_inode.i_ctime = CURRENT_TIME;
-
-	error = gfs2_meta_inode_buffer(ip, &dibh);
-	if (error)
-		goto out_end_trans;
-
 	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
 	gfs2_dinode_out(ip, dibh->b_data);
-	brelse(dibh);
 
+out_brelse:
+	brelse(dibh);
 out_end_trans:
 	gfs2_trans_end(sdp);
 out_ipres:
@@ -996,6 +1069,7 @@ out:
 
 static int trunc_dealloc(struct gfs2_inode *ip, u64 size)
 {
+	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
 	unsigned int height = ip->i_height;
 	u64 lblock;
 	struct metapath mp;
@@ -1004,9 +1078,9 @@ static int trunc_dealloc(struct gfs2_inode *ip, u64 size)
 	if (!size)
 		lblock = 0;
 	else
-		lblock = (size - 1) >> GFS2_SB(&ip->i_inode)->sd_sb.sb_bsize_shift;
+		lblock = (size - 1) >> sdp->sd_sb.sb_bsize_shift;
 
-	find_metapath(ip, lblock, &mp);
+	find_metapath(sdp, lblock, &mp, ip->i_height);
 	gfs2_alloc_get(ip);
 
 	error = gfs2_quota_hold(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
diff --git a/fs/gfs2/dir.c b/fs/gfs2/dir.c
index 34dc8df..a3753c7 100644
--- a/fs/gfs2/dir.c
+++ b/fs/gfs2/dir.c
@@ -159,6 +159,7 @@ static int gfs2_dir_write_data(struct gfs2_inode *ip, const char *buf,
 	unsigned int o;
 	int copied = 0;
 	int error = 0;
+	int new = 0;
 
 	if (!size)
 		return 0;
@@ -183,7 +184,6 @@ static int gfs2_dir_write_data(struct gfs2_inode *ip, const char *buf,
 	while (copied < size) {
 		unsigned int amount;
 		struct buffer_head *bh;
-		int new = 0;
 
 		amount = size - copied;
 		if (amount > sdp->sd_sb.sb_bsize - o)
diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c
index 77eba0a..4291375 100644
--- a/fs/gfs2/rgrp.c
+++ b/fs/gfs2/rgrp.c
@@ -1357,16 +1357,18 @@ static u32 rgblk_search(struct gfs2_rgrpd *rgd, u32 goal,
 		gfs2_trans_add_bh(rgd->rd_gl, bi->bi_bh, 1);
 		gfs2_setbit(rgd, bi->bi_bh->b_data, bi->bi_clone, bi->bi_offset,
 			    bi->bi_len, blk, new_state);
-		while(*n < elen) {
+		goal = blk;
+		while (*n < elen) {
 			goal++;
-			if (goal >= (bi->bi_len / GFS2_NBBY))
+			if (goal >= (bi->bi_len * GFS2_NBBY))
 				break;
 			if (gfs2_testbit(rgd, buffer, bi->bi_len, goal) !=
 			    GFS2_BLKST_FREE)
 				break;
-			(*n)++;
 			gfs2_setbit(rgd, bi->bi_bh->b_data, bi->bi_clone,
-				    bi->bi_offset, bi->bi_len, blk, new_state);
+				    bi->bi_offset, bi->bi_len, goal,
+				    new_state);
+			(*n)++;
 		}
 	}
 
-- 
1.5.1.2

[Cluster-devel] [PATCH 36/48] [GFS2] fix file_system_type leak on gfs2meta mount

[swhiteho]

From: Christoph Hellwig <hch@lst.de>

get_gfs2_sb does a get_fs_type without doing a put_filesystem and
thus leaking a file_system_type reference everytime it's called.

Just use gfs2_fs_type directly instead of doing the lookup and thus
fix the problem.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c
index c4b7a21..63d5fd2 100644
--- a/fs/gfs2/ops_fstype.c
+++ b/fs/gfs2/ops_fstype.c
@@ -943,7 +943,6 @@ static struct super_block* get_gfs2_sb(const char *dev_name)
 {
 	struct kstat stat;
 	struct nameidata nd;
-	struct file_system_type *fstype;
 	struct super_block *sb = NULL, *s;
 	int error;
 
@@ -955,8 +954,7 @@ static struct super_block* get_gfs2_sb(const char *dev_name)
 	}
 	error = vfs_getattr(nd.path.mnt, nd.path.dentry, &stat);
 
-	fstype = get_fs_type("gfs2");
-	list_for_each_entry(s, &fstype->fs_supers, s_instances) {
+	list_for_each_entry(s, &gfs2_fs_type.fs_supers, s_instances) {
 		if ((S_ISBLK(stat.mode) && s->s_dev == stat.rdev) ||
 		    (S_ISDIR(stat.mode) &&
 		     s == nd.path.dentry->d_inode->i_sb)) {
-- 
1.5.1.2

[Cluster-devel] [PATCH 37/48] [GFS2] remove gfs2_dev_iops

[swhiteho]

From: Denis Cheng <crquan@gmail.com>

struct inode_operations gfs2_dev_iops is always the same as gfs2_file_iops,
since Jan 2006, when GFS2 merged into mainstream kernel.

So one of them could be removed.

Signed-off-by: Denis Cheng <crquan@gmail.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/inode.c b/fs/gfs2/inode.c
index 810ff02..7b9f31e 100644
--- a/fs/gfs2/inode.c
+++ b/fs/gfs2/inode.c
@@ -149,7 +149,7 @@ void gfs2_set_iop(struct inode *inode)
 	} else if (S_ISLNK(mode)) {
 		inode->i_op = &gfs2_symlink_iops;
 	} else {
-		inode->i_op = &gfs2_dev_iops;
+		inode->i_op = &gfs2_file_iops;
 	}
 
 	unlock_new_inode(inode);
diff --git a/fs/gfs2/ops_inode.c b/fs/gfs2/ops_inode.c
index af7097a..6cbbb5b 100644
--- a/fs/gfs2/ops_inode.c
+++ b/fs/gfs2/ops_inode.c
@@ -1149,16 +1149,6 @@ const struct inode_operations gfs2_file_iops = {
 	.removexattr = gfs2_removexattr,
 };
 
-const struct inode_operations gfs2_dev_iops = {
-	.permission = gfs2_permission,
-	.setattr = gfs2_setattr,
-	.getattr = gfs2_getattr,
-	.setxattr = gfs2_setxattr,
-	.getxattr = gfs2_getxattr,
-	.listxattr = gfs2_listxattr,
-	.removexattr = gfs2_removexattr,
-};
-
 const struct inode_operations gfs2_dir_iops = {
 	.create = gfs2_create,
 	.lookup = gfs2_lookup,
diff --git a/fs/gfs2/ops_inode.h b/fs/gfs2/ops_inode.h
index fd8cee2..14b4b79 100644
--- a/fs/gfs2/ops_inode.h
+++ b/fs/gfs2/ops_inode.h
@@ -15,7 +15,6 @@
 extern const struct inode_operations gfs2_file_iops;
 extern const struct inode_operations gfs2_dir_iops;
 extern const struct inode_operations gfs2_symlink_iops;
-extern const struct inode_operations gfs2_dev_iops;
 extern const struct file_operations gfs2_file_fops;
 extern const struct file_operations gfs2_dir_fops;
 extern const struct file_operations gfs2_file_fops_nolock;
-- 
1.5.1.2

[Cluster-devel] [PATCH 38/48] [GFS2] re-support special inode

[swhiteho]

From: Denis Cheng <crquan@gmail.com>

a previous commit removed call to
init_special_inode from inode lookuping, this cause problems as:

 # mknod /mnt/gfs2/dev/null c 1 3
 # cat /mnt/gfs2/dev/null
 cat: /mnt/gfs2/dev/null: Invalid argument

without special inode, GFS2 cannot support char device file,
block device file, fifo pipe, and socket file, lose many important
features as a common file system.

this one line patch re add special inode support.

Signed-off-by: Denis Cheng <crquan@gmail.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/inode.c b/fs/gfs2/inode.c
index 7b9f31e..92ea9af 100644
--- a/fs/gfs2/inode.c
+++ b/fs/gfs2/inode.c
@@ -150,6 +150,7 @@ void gfs2_set_iop(struct inode *inode)
 		inode->i_op = &gfs2_symlink_iops;
 	} else {
 		inode->i_op = &gfs2_file_iops;
+		init_special_inode(inode, inode->i_mode, inode->i_rdev);
 	}
 
 	unlock_new_inode(inode);
-- 
1.5.1.2

[Cluster-devel] [PATCH 39/48] [GFS2] Need to ensure that sector_t is 64bits for GFS2

[swhiteho]

From: Steven Whitehouse <swhiteho@redhat.com>

We need to ensure that sector_t is 64bits for GFS2, so that we need to
depend on LBD as well as LSF.

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

diff --git a/fs/gfs2/Kconfig b/fs/gfs2/Kconfig
index d147b53..7f7947e 100644
--- a/fs/gfs2/Kconfig
+++ b/fs/gfs2/Kconfig
@@ -1,6 +1,6 @@
 config GFS2_FS
 	tristate "GFS2 file system support"
-	depends on EXPERIMENTAL && (64BIT || LSF)
+	depends on EXPERIMENTAL && (64BIT || (LSF && LBD))
 	select FS_POSIX_ACL
 	select CRC32
 	help
-- 
1.5.1.2

[Cluster-devel] [PATCH 40/48] [GFS2] possible null pointer dereference fixup

[swhiteho]

From: Cyrill Gorcunov <gorcunov@gmail.com>

gfs2_alloc_get may fail so we have to check it to prevent
NULL pointer dereference.

Signed-off-by: Cyrill Gorcunov <gorcunov@gamil.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index c1ee635..f7093aa 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -900,6 +900,8 @@ static int do_grow(struct gfs2_inode *ip, u64 size)
 	int error;
 
 	al = gfs2_alloc_get(ip);
+	if (!al)
+		return -ENOMEM;
 
 	error = gfs2_quota_lock(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
 	if (error)
@@ -1081,7 +1083,8 @@ static int trunc_dealloc(struct gfs2_inode *ip, u64 size)
 		lblock = (size - 1) >> sdp->sd_sb.sb_bsize_shift;
 
 	find_metapath(sdp, lblock, &mp, ip->i_height);
-	gfs2_alloc_get(ip);
+	if (!gfs2_alloc_get(ip))
+		return -ENOMEM;
 
 	error = gfs2_quota_hold(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
 	if (error)
diff --git a/fs/gfs2/dir.c b/fs/gfs2/dir.c
index a3753c7..94070ad 100644
--- a/fs/gfs2/dir.c
+++ b/fs/gfs2/dir.c
@@ -1868,11 +1868,14 @@ static int leaf_dealloc(struct gfs2_inode *dip, u32 index, u32 len,
 	if (!ht)
 		return -ENOMEM;
 
-	gfs2_alloc_get(dip);
+	if (!gfs2_alloc_get(dip)) {
+		error = -ENOMEM;
+		goto out;
+	}
 
 	error = gfs2_quota_hold(dip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
 	if (error)
-		goto out;
+		goto out_put;
 
 	error = gfs2_rindex_hold(sdp, &dip->i_alloc->al_ri_gh);
 	if (error)
@@ -1946,8 +1949,9 @@ out_rlist:
 	gfs2_glock_dq_uninit(&dip->i_alloc->al_ri_gh);
 out_qs:
 	gfs2_quota_unhold(dip);
-out:
+out_put:
 	gfs2_alloc_put(dip);
+out:
 	kfree(ht);
 	return error;
 }
diff --git a/fs/gfs2/eattr.c b/fs/gfs2/eattr.c
index 76ead1a..288d5e6 100644
--- a/fs/gfs2/eattr.c
+++ b/fs/gfs2/eattr.c
@@ -318,6 +318,8 @@ static int ea_remove_unstuffed(struct gfs2_inode *ip, struct buffer_head *bh,
 	int error;
 
 	al = gfs2_alloc_get(ip);
+	if (!al)
+		return -ENOMEM;
 
 	error = gfs2_quota_hold(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
 	if (error)
@@ -681,6 +683,8 @@ static int ea_alloc_skeleton(struct gfs2_inode *ip, struct gfs2_ea_request *er,
 	int error;
 
 	al = gfs2_alloc_get(ip);
+	if (!al)
+		return -ENOMEM;
 
 	error = gfs2_quota_lock(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
 	if (error)
@@ -1464,6 +1468,8 @@ int gfs2_ea_dealloc(struct gfs2_inode *ip)
 	int error;
 
 	al = gfs2_alloc_get(ip);
+	if (!al)
+		return -ENOMEM;
 
 	error = gfs2_quota_hold(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
 	if (error)
diff --git a/fs/gfs2/inode.c b/fs/gfs2/inode.c
index 92ea9af..dcae2aa 100644
--- a/fs/gfs2/inode.c
+++ b/fs/gfs2/inode.c
@@ -351,6 +351,8 @@ int gfs2_dinode_dealloc(struct gfs2_inode *ip)
 	}
 
 	al = gfs2_alloc_get(ip);
+	if (!al)
+		return -ENOMEM;
 
 	error = gfs2_quota_hold(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
 	if (error)
@@ -825,7 +827,8 @@ static int make_dinode(struct gfs2_inode *dip, struct gfs2_glock *gl,
 	int error;
 
 	munge_mode_uid_gid(dip, &mode, &uid, &gid);
-	gfs2_alloc_get(dip);
+	if (!gfs2_alloc_get(dip))
+		return -ENOMEM;
 
 	error = gfs2_quota_lock(dip, uid, gid);
 	if (error)
@@ -860,6 +863,8 @@ static int link_dinode(struct gfs2_inode *dip, const struct qstr *name,
 	int error;
 
 	al = gfs2_alloc_get(dip);
+	if (!al)
+		return -ENOMEM;
 
 	error = gfs2_quota_lock(dip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
 	if (error)
diff --git a/fs/gfs2/ops_address.c b/fs/gfs2/ops_address.c
index fbb4a6a..2483d87 100644
--- a/fs/gfs2/ops_address.c
+++ b/fs/gfs2/ops_address.c
@@ -649,6 +649,10 @@ static int gfs2_write_begin(struct file *file, struct address_space *mapping,
 
 	if (alloc_required) {
 		al = gfs2_alloc_get(ip);
+		if (!al) {
+			error = -ENOMEM;
+			goto out_unlock;
+		}
 
 		error = gfs2_quota_lock(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
 		if (error)
diff --git a/fs/gfs2/ops_inode.c b/fs/gfs2/ops_inode.c
index 6cbbb5b..34fe571 100644
--- a/fs/gfs2/ops_inode.c
+++ b/fs/gfs2/ops_inode.c
@@ -200,6 +200,10 @@ static int gfs2_link(struct dentry *old_dentry, struct inode *dir,
 
 	if (alloc_required) {
 		struct gfs2_alloc *al = gfs2_alloc_get(dip);
+		if (!al) {
+			error = -ENOMEM;
+			goto out_gunlock;
+		}
 
 		error = gfs2_quota_lock(dip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
 		if (error)
@@ -716,6 +720,10 @@ static int gfs2_rename(struct inode *odir, struct dentry *odentry,
 
 	if (alloc_required) {
 		struct gfs2_alloc *al = gfs2_alloc_get(ndip);
+		if (!al) {
+			error = -ENOMEM;
+			goto out_gunlock;
+		}
 
 		error = gfs2_quota_lock(ndip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
 		if (error)
@@ -953,7 +961,8 @@ static int setattr_chown(struct inode *inode, struct iattr *attr)
 	if (!(attr->ia_valid & ATTR_GID) || ogid == ngid)
 		ogid = ngid = NO_QUOTA_CHANGE;
 
-	gfs2_alloc_get(ip);
+	if (!gfs2_alloc_get(ip))
+		return -ENOMEM;
 
 	error = gfs2_quota_lock(ip, nuid, ngid);
 	if (error)
diff --git a/fs/gfs2/quota.c b/fs/gfs2/quota.c
index a08dabd..636bccf 100644
--- a/fs/gfs2/quota.c
+++ b/fs/gfs2/quota.c
@@ -617,8 +617,9 @@ static int gfs2_adjust_quota(struct gfs2_inode *ip, loff_t loc,
 	int err = -EIO;
 
 	if (gfs2_is_stuffed(ip)) {
-		struct gfs2_alloc *al = NULL;
-		al = gfs2_alloc_get(ip);
+		struct gfs2_alloc *al = gfs2_alloc_get(ip);
+		if (!al)
+			return -ENOMEM;
 		/* just request 1 blk */
 		al->al_requested = 1;
 		gfs2_inplace_reserve(ip);
@@ -729,6 +730,10 @@ static int do_sync(unsigned int num_qd, struct gfs2_quota_data **qda)
 
 	if (nalloc) {
 		al = gfs2_alloc_get(ip);
+		if (!al) {
+			error = -ENOMEM;
+			goto out_gunlock;
+		}
 
 		al->al_requested = nalloc * (data_blocks + ind_blocks);
 
-- 
1.5.1.2

[Cluster-devel] [PATCH 41/48] [GFS2] gfs2_adjust_quota has broken unstuffing code

[swhiteho]

From: Abhijith Das <adas@redhat.com>

This patch combines the 2 patches in bug 434736 to correct the lock
ordering in the unstuffing of the quota inode in gfs2_adjust_quota and
adjusting the number of revokes in gfs2_write_jdata_pagevec

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

diff --git a/fs/gfs2/ops_address.c b/fs/gfs2/ops_address.c
index 2483d87..e72fd47 100644
--- a/fs/gfs2/ops_address.c
+++ b/fs/gfs2/ops_address.c
@@ -21,7 +21,6 @@
 #include <linux/gfs2_ondisk.h>
 #include <linux/lm_interface.h>
 #include <linux/backing-dev.h>
-#include <linux/pagevec.h>
 
 #include "gfs2.h"
 #include "incore.h"
@@ -278,7 +277,7 @@ static int gfs2_write_jdata_pagevec(struct address_space *mapping,
 	int i;
 	int ret;
 
-	ret = gfs2_trans_begin(sdp, nrblocks, 0);
+	ret = gfs2_trans_begin(sdp, nrblocks, nrblocks);
 	if (ret < 0)
 		return ret;
 
diff --git a/fs/gfs2/quota.c b/fs/gfs2/quota.c
index 636bccf..c71f781 100644
--- a/fs/gfs2/quota.c
+++ b/fs/gfs2/quota.c
@@ -616,17 +616,9 @@ static int gfs2_adjust_quota(struct gfs2_inode *ip, loff_t loc,
 	s64 value;
 	int err = -EIO;
 
-	if (gfs2_is_stuffed(ip)) {
-		struct gfs2_alloc *al = gfs2_alloc_get(ip);
-		if (!al)
-			return -ENOMEM;
-		/* just request 1 blk */
-		al->al_requested = 1;
-		gfs2_inplace_reserve(ip);
+	if (gfs2_is_stuffed(ip))
 		gfs2_unstuff_dinode(ip, NULL);
-		gfs2_inplace_release(ip);
-		gfs2_alloc_put(ip);
-	}
+	
 	page = grab_cache_page(mapping, index);
 	if (!page)
 		return -ENOMEM;
@@ -691,7 +683,7 @@ static int do_sync(unsigned int num_qd, struct gfs2_quota_data **qda)
 	unsigned int qx, x;
 	struct gfs2_quota_data *qd;
 	loff_t offset;
-	unsigned int nalloc = 0;
+	unsigned int nalloc = 0, blocks;
 	struct gfs2_alloc *al = NULL;
 	int error;
 
@@ -728,34 +720,33 @@ static int do_sync(unsigned int num_qd, struct gfs2_quota_data **qda)
 			nalloc++;
 	}
 
-	if (nalloc) {
-		al = gfs2_alloc_get(ip);
-		if (!al) {
-			error = -ENOMEM;
-			goto out_gunlock;
-		}
+	al = gfs2_alloc_get(ip);
+	if (!al) {
+		error = -ENOMEM;
+		goto out_gunlock;
+	}
+	/* 
+	 * 1 blk for unstuffing inode if stuffed. We add this extra
+	 * block to the reservation unconditionally. If the inode
+	 * doesn't need unstuffing, the block will be released to the 
+	 * rgrp since it won't be allocated during the transaction
+	 */
+	al->al_requested = 1;
+	/* +1 in the end for block requested above for unstuffing */
+	blocks = num_qd * data_blocks + RES_DINODE + num_qd + 1;
 
-		al->al_requested = nalloc * (data_blocks + ind_blocks);
+	if (nalloc)
+		al->al_requested += nalloc * (data_blocks + ind_blocks);		
+	error = gfs2_inplace_reserve(ip);
+	if (error)
+		goto out_alloc;
 
-		error = gfs2_inplace_reserve(ip);
-		if (error)
-			goto out_alloc;
-
-		error = gfs2_trans_begin(sdp,
-					 al->al_rgd->rd_length +
-					 num_qd * data_blocks +
-					 nalloc * ind_blocks +
-					 RES_DINODE + num_qd +
-					 RES_STATFS, 0);
-		if (error)
-			goto out_ipres;
-	} else {
-		error = gfs2_trans_begin(sdp,
-					 num_qd * data_blocks +
-					 RES_DINODE + num_qd, 0);
-		if (error)
-			goto out_gunlock;
-	}
+	if (nalloc)
+		blocks += al->al_rgd->rd_length + nalloc * ind_blocks + RES_STATFS;
+
+	error = gfs2_trans_begin(sdp, blocks, 0);
+	if (error)
+		goto out_ipres;
 
 	for (x = 0; x < num_qd; x++) {
 		qd = qda[x];
@@ -774,11 +765,9 @@ static int do_sync(unsigned int num_qd, struct gfs2_quota_data **qda)
 out_end_trans:
 	gfs2_trans_end(sdp);
 out_ipres:
-	if (nalloc)
-		gfs2_inplace_release(ip);
+	gfs2_inplace_release(ip);
 out_alloc:
-	if (nalloc)
-		gfs2_alloc_put(ip);
+	gfs2_alloc_put(ip);
 out_gunlock:
 	gfs2_glock_dq_uninit(&i_gh);
 out:
-- 
1.5.1.2

[Cluster-devel] [PATCH 42/48] [GFS2] Remove drop of module ref where not needed

[swhiteho]

From: Steven Whitehouse <swhiteho@redhat.com>

In an earlier patch "[GFS2] fix file_system_type leak on gfs2meta mount"
we removed the code to grab a ref to the module which was not needed
(since we know that the module cannot be unloaded at that time) so
this patch removes the code to drop that reference.

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

diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c
index 63d5fd2..5b518f7 100644
--- a/fs/gfs2/ops_fstype.c
+++ b/fs/gfs2/ops_fstype.c
@@ -998,7 +998,6 @@ static int gfs2_get_sb_meta(struct file_system_type *fs_type, int flags,
 		error = PTR_ERR(new);
 		goto error;
 	}
-	module_put(fs_type->owner);
 	new->s_flags = flags;
 	strlcpy(new->s_id, sb->s_id, sizeof(new->s_id));
 	sb_set_blocksize(new, sb->s_blocksize);
-- 
1.5.1.2

[Cluster-devel] [PATCH 43/48] [GFS2] Streamline quota lock/check for no-quota case

[swhiteho]

From: Steven Whitehouse <swhiteho@redhat.com>

This patch streamlines the quota checking in the "no quota" case by
making the check inline in the calling function, thus reducing the
number of function calls. Eventually we might be able to remove the
checks from the gfs2_quota_lock() and gfs2_quota_check() functions, but
currently we can't as there are a very few places in the code which need
to call these functions directly still.

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

diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index f7093aa..c19184f 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -903,14 +903,10 @@ static int do_grow(struct gfs2_inode *ip, u64 size)
 	if (!al)
 		return -ENOMEM;
 
-	error = gfs2_quota_lock(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
+	error = gfs2_quota_lock_check(ip);
 	if (error)
 		goto out;
 
-	error = gfs2_quota_check(ip, ip->i_inode.i_uid, ip->i_inode.i_gid);
-	if (error)
-		goto out_gunlock_q;
-
 	al->al_requested = sdp->sd_max_height + RES_DATA;
 
 	error = gfs2_inplace_reserve(ip);
diff --git a/fs/gfs2/eattr.c b/fs/gfs2/eattr.c
index 288d5e6..8175592 100644
--- a/fs/gfs2/eattr.c
+++ b/fs/gfs2/eattr.c
@@ -686,14 +686,10 @@ static int ea_alloc_skeleton(struct gfs2_inode *ip, struct gfs2_ea_request *er,
 	if (!al)
 		return -ENOMEM;
 
-	error = gfs2_quota_lock(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
+	error = gfs2_quota_lock_check(ip);
 	if (error)
 		goto out;
 
-	error = gfs2_quota_check(ip, ip->i_inode.i_uid, ip->i_inode.i_gid);
-	if (error)
-		goto out_gunlock_q;
-
 	al->al_requested = blks;
 
 	error = gfs2_inplace_reserve(ip);
diff --git a/fs/gfs2/ops_address.c b/fs/gfs2/ops_address.c
index e72fd47..90a04a6 100644
--- a/fs/gfs2/ops_address.c
+++ b/fs/gfs2/ops_address.c
@@ -653,14 +653,10 @@ static int gfs2_write_begin(struct file *file, struct address_space *mapping,
 			goto out_unlock;
 		}
 
-		error = gfs2_quota_lock(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
+		error = gfs2_quota_lock_check(ip);
 		if (error)
 			goto out_alloc_put;
 
-		error = gfs2_quota_check(ip, ip->i_inode.i_uid, ip->i_inode.i_gid);
-		if (error)
-			goto out_qunlock;
-
 		al->al_requested = data_blocks + ind_blocks;
 		error = gfs2_inplace_reserve(ip);
 		if (error)
diff --git a/fs/gfs2/ops_file.c b/fs/gfs2/ops_file.c
index 2b25a5f..e1b7d52 100644
--- a/fs/gfs2/ops_file.c
+++ b/fs/gfs2/ops_file.c
@@ -369,12 +369,9 @@ static int gfs2_page_mkwrite(struct vm_area_struct *vma, struct page *page)
 	if (al == NULL)
 		goto out_unlock;
 
-	ret = gfs2_quota_lock(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
+	ret = gfs2_quota_lock_check(ip);
 	if (ret)
 		goto out_alloc_put;
-	ret = gfs2_quota_check(ip, ip->i_inode.i_uid, ip->i_inode.i_gid);
-	if (ret)
-		goto out_quota_unlock;
 	al->al_requested = data_blocks + ind_blocks;
 	ret = gfs2_inplace_reserve(ip);
 	if (ret)
diff --git a/fs/gfs2/ops_inode.c b/fs/gfs2/ops_inode.c
index 34fe571..2686ad4 100644
--- a/fs/gfs2/ops_inode.c
+++ b/fs/gfs2/ops_inode.c
@@ -205,14 +205,10 @@ static int gfs2_link(struct dentry *old_dentry, struct inode *dir,
 			goto out_gunlock;
 		}
 
-		error = gfs2_quota_lock(dip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
+		error = gfs2_quota_lock_check(dip);
 		if (error)
 			goto out_alloc;
 
-		error = gfs2_quota_check(dip, dip->i_inode.i_uid, dip->i_inode.i_gid);
-		if (error)
-			goto out_gunlock_q;
-
 		al->al_requested = sdp->sd_max_dirres;
 
 		error = gfs2_inplace_reserve(dip);
@@ -725,14 +721,10 @@ static int gfs2_rename(struct inode *odir, struct dentry *odentry,
 			goto out_gunlock;
 		}
 
-		error = gfs2_quota_lock(ndip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
+		error = gfs2_quota_lock_check(ndip);
 		if (error)
 			goto out_alloc;
 
-		error = gfs2_quota_check(ndip, ndip->i_inode.i_uid, ndip->i_inode.i_gid);
-		if (error)
-			goto out_gunlock_q;
-
 		al->al_requested = sdp->sd_max_dirres;
 
 		error = gfs2_inplace_reserve(ndip);
diff --git a/fs/gfs2/quota.h b/fs/gfs2/quota.h
index a8be141..3b7f4b0 100644
--- a/fs/gfs2/quota.h
+++ b/fs/gfs2/quota.h
@@ -32,4 +32,21 @@ int gfs2_quota_init(struct gfs2_sbd *sdp);
 void gfs2_quota_scan(struct gfs2_sbd *sdp);
 void gfs2_quota_cleanup(struct gfs2_sbd *sdp);
 
+static inline int gfs2_quota_lock_check(struct gfs2_inode *ip)
+{
+	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
+	int ret;
+	if (sdp->sd_args.ar_quota == GFS2_QUOTA_OFF)
+		return 0;
+	ret = gfs2_quota_lock(ip, NO_QUOTA_CHANGE, NO_QUOTA_CHANGE);
+	if (ret)
+		return ret;
+	if (sdp->sd_args.ar_quota != GFS2_QUOTA_ON)
+		return 0;
+	ret = gfs2_quota_check(ip, ip->i_inode.i_uid, ip->i_inode.i_gid);
+	if (ret)
+		gfs2_quota_unlock(ip);
+	return ret;
+}
+
 #endif /* __QUOTA_DOT_H__ */
-- 
1.5.1.2

[Cluster-devel] [PATCH 44/48] [GFS2] Faster gfs2_bitfit algorithm

[swhiteho]

From: Bob Peterson <rpeterso@redhat.com>

This version of the gfs2_bitfit algorithm includes the latest
suggestions from Steve Whitehouse.  It is typically eight to
ten times faster than the version we're using today.  If there
is a lot of metadata mixed in (lots of small files) the
algorithm is often 15 times faster, and given the right
conditions, I've seen peaks of 20 times faster.

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

diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c
index 4291375..7e8f0b1 100644
--- a/fs/gfs2/rgrp.c
+++ b/fs/gfs2/rgrp.c
@@ -14,6 +14,7 @@
 #include <linux/fs.h>
 #include <linux/gfs2_ondisk.h>
 #include <linux/lm_interface.h>
+#include <linux/prefetch.h>
 
 #include "gfs2.h"
 #include "incore.h"
@@ -33,6 +34,16 @@
 #define BFITNOENT ((u32)~0)
 #define NO_BLOCK ((u64)~0)
 
+#if BITS_PER_LONG == 32
+#define LBITMASK   (0x55555555UL)
+#define LBITSKIP55 (0x55555555UL)
+#define LBITSKIP00 (0x00000000UL)
+#else
+#define LBITMASK   (0x5555555555555555UL)
+#define LBITSKIP55 (0x5555555555555555UL)
+#define LBITSKIP00 (0x0000000000000000UL)
+#endif
+
 /*
  * These routines are used by the resource group routines (rgrp.c)
  * to keep track of block allocation.  Each block is represented by two
@@ -138,45 +149,63 @@ static inline unsigned char gfs2_testbit(struct gfs2_rgrpd *rgd,
 static u32 gfs2_bitfit(const u8 *buffer, unsigned int buflen, u32 goal,
 		       u8 old_state)
 {
-	const u8 *byte;
-	u32 blk = goal;
-	unsigned int bit, bitlong;
-	const unsigned long *plong;
-#if BITS_PER_LONG == 32
-	const unsigned long plong55 = 0x55555555;
-#else
-	const unsigned long plong55 = 0x5555555555555555;
-#endif
-
-	byte = buffer + (goal / GFS2_NBBY);
-	plong = (const unsigned long *)(buffer + (goal / GFS2_NBBY));
-	bit = (goal % GFS2_NBBY) * GFS2_BIT_SIZE;
-	bitlong = bit;
-
-	while (byte < buffer + buflen) {
-
-		if (bitlong == 0 && old_state == 0 && *plong == plong55) {
-			plong++;
-			byte += sizeof(unsigned long);
-			blk += sizeof(unsigned long) * GFS2_NBBY;
-			continue;
+	const u8 *byte, *start, *end;
+	int bit, startbit;
+	u32 g1, g2, misaligned;
+	unsigned long *plong;
+	unsigned long lskipval;
+
+	lskipval = (old_state & GFS2_BLKST_USED) ? LBITSKIP00 : LBITSKIP55;
+	g1 = (goal / GFS2_NBBY);
+	start = buffer + g1;
+	byte = start;
+        end = buffer + buflen;
+	g2 = ALIGN(g1, sizeof(unsigned long));
+	plong = (unsigned long *)(buffer + g2);
+	startbit = bit = (goal % GFS2_NBBY) * GFS2_BIT_SIZE;
+	misaligned = g2 - g1;
+	if (!misaligned)
+		goto ulong_aligned;
+/* parse the bitmap a byte@a time */
+misaligned:
+	while (byte < end) {
+		if (((*byte >> bit) & GFS2_BIT_MASK) == old_state) {
+			return goal +
+				(((byte - start) * GFS2_NBBY) +
+				 ((bit - startbit) >> 1));
 		}
-		if (((*byte >> bit) & GFS2_BIT_MASK) == old_state)
-			return blk;
 		bit += GFS2_BIT_SIZE;
-		if (bit >= 8) {
+		if (bit >= GFS2_NBBY * GFS2_BIT_SIZE) {
 			bit = 0;
 			byte++;
+			misaligned--;
+			if (!misaligned) {
+				plong = (unsigned long *)byte;
+				goto ulong_aligned;
+			}
 		}
-		bitlong += GFS2_BIT_SIZE;
-		if (bitlong >= sizeof(unsigned long) * 8) {
-			bitlong = 0;
-			plong++;
-		}
-
-		blk++;
 	}
+	return BFITNOENT;
 
+/* parse the bitmap a unsigned long at a time */
+ulong_aligned:
+	/* Stop at "end - 1" or else prefetch can go past the end and segfault.
+	   We could "if" it but we'd lose some of the performance gained.
+	   This way will only slow down searching the very last 4/8 bytes
+	   depending on architecture.  I've experimented with several ways
+	   of writing this section such as using an else before the goto
+	   but this one seems to be the fastest. */
+	while ((unsigned char *)plong < end - 1) {
+		prefetch(plong + 1);
+		if (((*plong) & LBITMASK) != lskipval)
+			break;
+		plong++;
+	}
+	if ((unsigned char *)plong < end) {
+		byte = (const u8 *)plong;
+		misaligned += sizeof(unsigned long) - 1;
+		goto misaligned;
+	}
 	return BFITNOENT;
 }
 
-- 
1.5.1.2

[Cluster-devel] [PATCH 45/48] [GFS2] fs/gfs2/recovery.c: suppress warnings

[swhiteho]

From: akpm@linux-foundation.org <akpm@linux-foundation.org>

fs/gfs2/recovery.c: In function 'get_log_header':
fs/gfs2/recovery.c:152: warning: 'lh.lh_sequence' may be used uninitialized in this function
fs/gfs2/recovery.c:152: warning: 'lh.lh_flags' may be used uninitialized in this function
fs/gfs2/recovery.c:152: warning: 'lh.lh_tail' may be used uninitialized in this function
fs/gfs2/recovery.c:152: warning: 'lh.lh_blkno' may be used uninitialized in this function
fs/gfs2/recovery.c:152: warning: 'lh.lh_hash' may be used uninitialized in this function

Cc: David Teigland <teigland@redhat.com>
Cc: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/recovery.c b/fs/gfs2/recovery.c
index b17d3b8..06dcdc0 100644
--- a/fs/gfs2/recovery.c
+++ b/fs/gfs2/recovery.c
@@ -149,7 +149,7 @@ static int get_log_header(struct gfs2_jdesc *jd, unsigned int blk,
 			  struct gfs2_log_header_host *head)
 {
 	struct buffer_head *bh;
-	struct gfs2_log_header_host lh;
+	struct gfs2_log_header_host uninitialized_var(lh);
 	const u32 nothing = 0;
 	u32 hash;
 	int error;
-- 
1.5.1.2

[Cluster-devel] [PATCH 46/48] [GFS2] Invalidate cache at correct point

[swhiteho]

From: Benjamin Marzinski <bmarzins@redhat.com>

GFS2 wasn't invalidating its cache before it called into the lock manager
with a request that could potentially drop a lock.  This was leaving a
window where the lock could be actually be held by another node, but the
file's page cache would still appear valid, causing coherency problems.
This patch moves the cache invalidation to before the lock manager call
when dropping a lock. It also adds the option to the lock_dlm lock
manager to not use conversion mode deadlock avoidance, which, on a
conversion from shared to exclusive, could internally drop the lock, and
then reacquire in. GFS2 now asks lock_dlm to not do this.  Instead, GFS2
manually drops the lock and reacquires it.

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

diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c
index 63981e2..d636b3e 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -764,7 +764,7 @@ static void state_change(struct gfs2_glock *gl, unsigned int new_state)
 static void drop_bh(struct gfs2_glock *gl, unsigned int ret)
 {
 	struct gfs2_sbd *sdp = gl->gl_sbd;
-	const struct gfs2_glock_operations *glops = gl->gl_ops;
+	struct gfs2_holder *gh = gl->gl_req_gh;
 
 	gfs2_assert_warn(sdp, test_bit(GLF_LOCK, &gl->gl_flags));
 	gfs2_assert_warn(sdp, list_empty(&gl->gl_holders));
@@ -772,8 +772,14 @@ static void drop_bh(struct gfs2_glock *gl, unsigned int ret)
 
 	state_change(gl, LM_ST_UNLOCKED);
 
-	if (glops->go_inval)
-		glops->go_inval(gl, DIO_METADATA);
+	if (test_and_clear_bit(GLF_CONV_DEADLK, &gl->gl_flags)) {
+		spin_lock(&gl->gl_spin);
+		gh->gh_error = 0;
+		spin_unlock(&gl->gl_spin);
+		gfs2_glock_xmote_th(gl, gl->gl_req_gh);
+		gfs2_glock_put(gl);
+		return;
+	}
 
 	spin_lock(&gl->gl_spin);
 	gfs2_demote_wake(gl);
@@ -794,7 +800,6 @@ static void xmote_bh(struct gfs2_glock *gl, unsigned int ret)
 	struct gfs2_sbd *sdp = gl->gl_sbd;
 	const struct gfs2_glock_operations *glops = gl->gl_ops;
 	struct gfs2_holder *gh = gl->gl_req_gh;
-	int prev_state = gl->gl_state;
 	int op_done = 1;
 
 	if (!gh && (ret & LM_OUT_ST_MASK) == LM_ST_UNLOCKED) {
@@ -808,16 +813,6 @@ static void xmote_bh(struct gfs2_glock *gl, unsigned int ret)
 
 	state_change(gl, ret & LM_OUT_ST_MASK);
 
-	if (prev_state != LM_ST_UNLOCKED && !(ret & LM_OUT_CACHEABLE)) {
-		if (glops->go_inval)
-			glops->go_inval(gl, DIO_METADATA);
-	} else if (gl->gl_state == LM_ST_DEFERRED) {
-		/* We might not want to do this here.
-		   Look at moving to the inode glops. */
-		if (glops->go_inval)
-			glops->go_inval(gl, 0);
-	}
-
 	/*  Deal with each possible exit condition  */
 
 	if (!gh) {
@@ -837,6 +832,14 @@ static void xmote_bh(struct gfs2_glock *gl, unsigned int ret)
 		}
 	} else {
 		spin_lock(&gl->gl_spin);
+		if (ret & LM_OUT_CONV_DEADLK) {
+			gh->gh_error = 0;
+			set_bit(GLF_CONV_DEADLK, &gl->gl_flags);
+			spin_unlock(&gl->gl_spin);
+			gfs2_glock_drop_th(gl);
+			gfs2_glock_put(gl);
+			return;
+		}
 		list_del_init(&gh->gh_list);
 		gh->gh_error = -EIO;
 		if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags))) 
@@ -910,6 +913,8 @@ static void gfs2_glock_xmote_th(struct gfs2_glock *gl, struct gfs2_holder *gh)
 
 	if (glops->go_xmote_th)
 		glops->go_xmote_th(gl);
+	if (state == LM_ST_DEFERRED && glops->go_inval)
+		glops->go_inval(gl, DIO_METADATA);
 
 	gfs2_assert_warn(sdp, test_bit(GLF_LOCK, &gl->gl_flags));
 	gfs2_assert_warn(sdp, list_empty(&gl->gl_holders));
@@ -952,6 +957,8 @@ static void gfs2_glock_drop_th(struct gfs2_glock *gl)
 
 	if (glops->go_xmote_th)
 		glops->go_xmote_th(gl);
+	if (glops->go_inval)
+		glops->go_inval(gl, DIO_METADATA);
 
 	gfs2_assert_warn(sdp, test_bit(GLF_LOCK, &gl->gl_flags));
 	gfs2_assert_warn(sdp, list_empty(&gl->gl_holders));
diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h
index 4ba2ea6..9c2c0b9 100644
--- a/fs/gfs2/incore.h
+++ b/fs/gfs2/incore.h
@@ -167,6 +167,7 @@ enum {
 	GLF_DEMOTE_IN_PROGRESS	= 6,
 	GLF_LFLUSH		= 7,
 	GLF_WAITERS2		= 8,
+	GLF_CONV_DEADLK		= 9,
 };
 
 struct gfs2_glock {
diff --git a/fs/gfs2/locking/dlm/lock.c b/fs/gfs2/locking/dlm/lock.c
index 542a797..53a6ab3 100644
--- a/fs/gfs2/locking/dlm/lock.c
+++ b/fs/gfs2/locking/dlm/lock.c
@@ -137,7 +137,8 @@ static inline unsigned int make_flags(struct gdlm_lock *lp,
 
 		/* Conversion deadlock avoidance by DLM */
 
-		if (!test_bit(LFL_FORCE_PROMOTE, &lp->flags) &&
+		if (!(lp->ls->fsflags & LM_MFLAG_CONV_NODROP) &&
+		    !test_bit(LFL_FORCE_PROMOTE, &lp->flags) &&
 		    !(lkf & DLM_LKF_NOQUEUE) &&
 		    cur > DLM_LOCK_NL && req > DLM_LOCK_NL && cur != req)
 			lkf |= DLM_LKF_CONVDEADLK;
diff --git a/fs/gfs2/locking/dlm/thread.c b/fs/gfs2/locking/dlm/thread.c
index 521694f..e53db6f 100644
--- a/fs/gfs2/locking/dlm/thread.c
+++ b/fs/gfs2/locking/dlm/thread.c
@@ -135,7 +135,15 @@ static void process_complete(struct gdlm_lock *lp)
 			 lp->lksb.sb_status, lp->lockname.ln_type,
 			 (unsigned long long)lp->lockname.ln_number,
 			 lp->flags);
-		return;
+		if (lp->lksb.sb_status == -EDEADLOCK &&
+		    lp->ls->fsflags & LM_MFLAG_CONV_NODROP) {
+			lp->req = lp->cur;
+			acb.lc_ret |= LM_OUT_CONV_DEADLK;
+			if (lp->cur == DLM_LOCK_IV)
+				lp->lksb.sb_lkid = 0;
+			goto out;
+		} else
+			return;
 	}
 
 	/*
diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c
index 5b518f7..ef9c6c4 100644
--- a/fs/gfs2/ops_fstype.c
+++ b/fs/gfs2/ops_fstype.c
@@ -723,7 +723,7 @@ static int gfs2_lm_mount(struct gfs2_sbd *sdp, int silent)
 {
 	char *proto = sdp->sd_proto_name;
 	char *table = sdp->sd_table_name;
-	int flags = 0;
+	int flags = LM_MFLAG_CONV_NODROP;
 	int error;
 
 	if (sdp->sd_args.ar_spectator)
diff --git a/include/linux/lm_interface.h b/include/linux/lm_interface.h
index 1418fdc..f274997 100644
--- a/include/linux/lm_interface.h
+++ b/include/linux/lm_interface.h
@@ -21,9 +21,15 @@ typedef void (*lm_callback_t) (void *ptr, unsigned int type, void *data);
  * modify the filesystem.  The lock module shouldn't assign a journal to the FS
  * mount.  It shouldn't send recovery callbacks to the FS mount.  If the node
  * dies or withdraws, all locks can be wiped immediately.
+ *
+ * LM_MFLAG_CONV_NODROP
+ * Do not allow the dlm to internally resolve conversion deadlocks by demoting
+ * the lock to unlocked and then reacquiring it in the requested mode. Instead,
+ * it should cancel the request and return LM_OUT_CONV_DEADLK.
  */
 
 #define LM_MFLAG_SPECTATOR	0x00000001
+#define LM_MFLAG_CONV_NODROP	0x00000002
 
 /*
  * lm_lockstruct flags
@@ -110,6 +116,9 @@ typedef void (*lm_callback_t) (void *ptr, unsigned int type, void *data);
  *
  * LM_OUT_ASYNC
  * The result of the request will be returned in an LM_CB_ASYNC callback.
+ *
+ * LM_OUT_CONV_DEADLK
+ * The lock request was canceled do to a conversion deadlock.
  */
 
 #define LM_OUT_ST_MASK		0x00000003
@@ -117,6 +126,7 @@ typedef void (*lm_callback_t) (void *ptr, unsigned int type, void *data);
 #define LM_OUT_CANCELED		0x00000008
 #define LM_OUT_ASYNC		0x00000080
 #define LM_OUT_ERROR		0x00000100
+#define LM_OUT_CONV_DEADLK	0x00000200
 
 /*
  * lm_callback_t types
-- 
1.5.1.2

[Cluster-devel] [PATCH 47/48] [GFS2] test for IS_ERR rather than 0

[swhiteho]

From: Julia Lawall <julia@diku.dk>

The function gfs2_inode_lookup always returns either a valid pointer or a
value made with ERR_PTR, so its result should be tested with IS_ERR, not
with a test for 0.

The problem was found using the following semantic match.
(http://www.emn.fr/x-info/coccinelle/)

//<smpl>
@a@
expression E, E1;
statement S,S1;
position p;
@@

E = gfs2_inode_lookup(...)
... when != E = E1
if at p (E) S else S1

@n@
position a.p;
expression E,E1;
statement S,S1;
@@

E = NULL
... when != E = E1
if at p (E) S else S1

@depends on !n@
expression E;
statement S,S1;
position a.p;
@@

* if at p (E)
  S else S1
//</smpl>

Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/ops_export.c b/fs/gfs2/ops_export.c
index 334c7f8..990d9f4 100644
--- a/fs/gfs2/ops_export.c
+++ b/fs/gfs2/ops_export.c
@@ -204,8 +204,6 @@ static struct dentry *gfs2_get_dentry(struct super_block *sb,
 	inode = gfs2_inode_lookup(sb, DT_UNKNOWN,
 					inum->no_addr,
 					0, 0);
-	if (!inode)
-		goto fail;
 	if (IS_ERR(inode)) {
 		error = PTR_ERR(inode);
 		goto fail;
-- 
1.5.1.2

[Cluster-devel] [PATCH 48/48] [GFS2] fix GFP_KERNEL misuses

[swhiteho]

From: Josef Bacik <jbacik@redhat.com>

There are several places where GFP_KERNEL allocations happen under a glock,
which will result in hangs if we're under memory pressure and go to re-enter the
fs in order to flush stuff out.  This patch changes the culprits to GFS_NOFS to
keep this problem from happening.  Thank you,

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

diff --git a/fs/gfs2/acl.c b/fs/gfs2/acl.c
index 1047a8c..3e9bd46 100644
--- a/fs/gfs2/acl.c
+++ b/fs/gfs2/acl.c
@@ -116,7 +116,7 @@ static int acl_get(struct gfs2_inode *ip, int access, struct posix_acl **acl,
 		goto out;
 
 	er.er_data_len = GFS2_EA_DATA_LEN(el->el_ea);
-	er.er_data = kmalloc(er.er_data_len, GFP_KERNEL);
+	er.er_data = kmalloc(er.er_data_len, GFP_NOFS);
 	error = -ENOMEM;
 	if (!er.er_data)
 		goto out;
@@ -222,7 +222,7 @@ int gfs2_acl_create(struct gfs2_inode *dip, struct gfs2_inode *ip)
 		return error;
 	}
 
-	clone = posix_acl_clone(acl, GFP_KERNEL);
+	clone = posix_acl_clone(acl, GFP_NOFS);
 	error = -ENOMEM;
 	if (!clone)
 		goto out;
@@ -272,7 +272,7 @@ int gfs2_acl_chmod(struct gfs2_inode *ip, struct iattr *attr)
 	if (!acl)
 		return gfs2_setattr_simple(ip, attr);
 
-	clone = posix_acl_clone(acl, GFP_KERNEL);
+	clone = posix_acl_clone(acl, GFP_NOFS);
 	error = -ENOMEM;
 	if (!clone)
 		goto out;
diff --git a/fs/gfs2/dir.c b/fs/gfs2/dir.c
index 94070ad..eed040d 100644
--- a/fs/gfs2/dir.c
+++ b/fs/gfs2/dir.c
@@ -1089,7 +1089,7 @@ static int dir_double_exhash(struct gfs2_inode *dip)
 
 	/*  Allocate both the "from" and "to" buffers in one big chunk  */
 
-	buf = kcalloc(3, sdp->sd_hash_bsize, GFP_KERNEL | __GFP_NOFAIL);
+	buf = kcalloc(3, sdp->sd_hash_bsize, GFP_NOFS | __GFP_NOFAIL);
 
 	for (block = dip->i_di.di_size >> sdp->sd_hash_bsize_shift; block--;) {
 		error = gfs2_dir_read_data(dip, (char *)buf,
@@ -1378,7 +1378,7 @@ static int dir_e_read(struct inode *inode, u64 *offset, void *opaque,
 	hash = gfs2_dir_offset2hash(*offset);
 	index = hash >> (32 - dip->i_depth);
 
-	lp = kmalloc(sdp->sd_hash_bsize, GFP_KERNEL);
+	lp = kmalloc(sdp->sd_hash_bsize, GFP_NOFS);
 	if (!lp)
 		return -ENOMEM;
 
@@ -1443,7 +1443,7 @@ int gfs2_dir_read(struct inode *inode, u64 *offset, void *opaque,
 
 	error = -ENOMEM;
 	/* 96 is max number of dirents which can be stuffed into an inode */
-	darr = kmalloc(96 * sizeof(struct gfs2_dirent *), GFP_KERNEL);
+	darr = kmalloc(96 * sizeof(struct gfs2_dirent *), GFP_NOFS);
 	if (darr) {
 		g.pdent = darr;
 		g.offset = 0;
@@ -1789,7 +1789,7 @@ static int foreach_leaf(struct gfs2_inode *dip, leaf_call_t lc, void *data)
 		return -EIO;
 	}
 
-	lp = kmalloc(sdp->sd_hash_bsize, GFP_KERNEL);
+	lp = kmalloc(sdp->sd_hash_bsize, GFP_NOFS);
 	if (!lp)
 		return -ENOMEM;
 
@@ -1864,7 +1864,7 @@ static int leaf_dealloc(struct gfs2_inode *dip, u32 index, u32 len,
 
 	memset(&rlist, 0, sizeof(struct gfs2_rgrp_list));
 
-	ht = kzalloc(size, GFP_KERNEL);
+	ht = kzalloc(size, GFP_NOFS);
 	if (!ht)
 		return -ENOMEM;
 
diff --git a/fs/gfs2/eattr.c b/fs/gfs2/eattr.c
index 8175592..e3f76f4 100644
--- a/fs/gfs2/eattr.c
+++ b/fs/gfs2/eattr.c
@@ -448,7 +448,7 @@ static int ea_get_unstuffed(struct gfs2_inode *ip, struct gfs2_ea_header *ea,
 	unsigned int x;
 	int error = 0;
 
-	bh = kcalloc(nptrs, sizeof(struct buffer_head *), GFP_KERNEL);
+	bh = kcalloc(nptrs, sizeof(struct buffer_head *), GFP_NOFS);
 	if (!bh)
 		return -ENOMEM;
 
@@ -1206,7 +1206,7 @@ static int ea_acl_chmod_unstuffed(struct gfs2_inode *ip,
 	unsigned int x;
 	int error;
 
-	bh = kcalloc(nptrs, sizeof(struct buffer_head *), GFP_KERNEL);
+	bh = kcalloc(nptrs, sizeof(struct buffer_head *), GFP_NOFS);
 	if (!bh)
 		return -ENOMEM;
 
diff --git a/fs/gfs2/inode.c b/fs/gfs2/inode.c
index dcae2aa..3a9ef52 100644
--- a/fs/gfs2/inode.c
+++ b/fs/gfs2/inode.c
@@ -1231,7 +1231,7 @@ int gfs2_readlinki(struct gfs2_inode *ip, char **buf, unsigned int *len)
 
 	x = ip->i_di.di_size + 1;
 	if (x > *len) {
-		*buf = kmalloc(x, GFP_KERNEL);
+		*buf = kmalloc(x, GFP_NOFS);
 		if (!*buf) {
 			error = -ENOMEM;
 			goto out_brelse;
diff --git a/fs/gfs2/locking/dlm/lock.c b/fs/gfs2/locking/dlm/lock.c
index 53a6ab3..cf7ea8a 100644
--- a/fs/gfs2/locking/dlm/lock.c
+++ b/fs/gfs2/locking/dlm/lock.c
@@ -165,7 +165,7 @@ static int gdlm_create_lp(struct gdlm_ls *ls, struct lm_lockname *name,
 {
 	struct gdlm_lock *lp;
 
-	lp = kzalloc(sizeof(struct gdlm_lock), GFP_KERNEL);
+	lp = kzalloc(sizeof(struct gdlm_lock), GFP_NOFS);
 	if (!lp)
 		return -ENOMEM;
 
@@ -383,7 +383,7 @@ static int gdlm_add_lvb(struct gdlm_lock *lp)
 {
 	char *lvb;
 
-	lvb = kzalloc(GDLM_LVB_SIZE, GFP_KERNEL);
+	lvb = kzalloc(GDLM_LVB_SIZE, GFP_NOFS);
 	if (!lvb)
 		return -ENOMEM;
 
diff --git a/fs/gfs2/locking/nolock/main.c b/fs/gfs2/locking/nolock/main.c
index d3b8ce6..284a5ec 100644
--- a/fs/gfs2/locking/nolock/main.c
+++ b/fs/gfs2/locking/nolock/main.c
@@ -140,7 +140,7 @@ static int nolock_hold_lvb(void *lock, char **lvbp)
 	struct nolock_lockspace *nl = lock;
 	int error = 0;
 
-	*lvbp = kzalloc(nl->nl_lvb_size, GFP_KERNEL);
+	*lvbp = kzalloc(nl->nl_lvb_size, GFP_NOFS);
 	if (!*lvbp)
 		error = -ENOMEM;
 
diff --git a/fs/gfs2/quota.c b/fs/gfs2/quota.c
index c71f781..56aaf91 100644
--- a/fs/gfs2/quota.c
+++ b/fs/gfs2/quota.c
@@ -94,7 +94,7 @@ static int qd_alloc(struct gfs2_sbd *sdp, int user, u32 id,
 	struct gfs2_quota_data *qd;
 	int error;
 
-	qd = kzalloc(sizeof(struct gfs2_quota_data), GFP_KERNEL);
+	qd = kzalloc(sizeof(struct gfs2_quota_data), GFP_NOFS);
 	if (!qd)
 		return -ENOMEM;
 
@@ -690,7 +690,7 @@ static int do_sync(unsigned int num_qd, struct gfs2_quota_data **qda)
 	gfs2_write_calc_reserv(ip, sizeof(struct gfs2_quota),
 			      &data_blocks, &ind_blocks);
 
-	ghs = kcalloc(num_qd, sizeof(struct gfs2_holder), GFP_KERNEL);
+	ghs = kcalloc(num_qd, sizeof(struct gfs2_holder), GFP_NOFS);
 	if (!ghs)
 		return -ENOMEM;
 
@@ -1118,12 +1118,12 @@ int gfs2_quota_init(struct gfs2_sbd *sdp)
 	error = -ENOMEM;
 
 	sdp->sd_quota_bitmap = kcalloc(sdp->sd_quota_chunks,
-				       sizeof(unsigned char *), GFP_KERNEL);
+				       sizeof(unsigned char *), GFP_NOFS);
 	if (!sdp->sd_quota_bitmap)
 		return error;
 
 	for (x = 0; x < sdp->sd_quota_chunks; x++) {
-		sdp->sd_quota_bitmap[x] = kzalloc(PAGE_SIZE, GFP_KERNEL);
+		sdp->sd_quota_bitmap[x] = kzalloc(PAGE_SIZE, GFP_NOFS);
 		if (!sdp->sd_quota_bitmap[x])
 			goto fail;
 	}
diff --git a/fs/gfs2/recovery.c b/fs/gfs2/recovery.c
index 06dcdc0..2888e4b 100644
--- a/fs/gfs2/recovery.c
+++ b/fs/gfs2/recovery.c
@@ -68,7 +68,7 @@ int gfs2_revoke_add(struct gfs2_sbd *sdp, u64 blkno, unsigned int where)
 		return 0;
 	}
 
-	rr = kmalloc(sizeof(struct gfs2_revoke_replay), GFP_KERNEL);
+	rr = kmalloc(sizeof(struct gfs2_revoke_replay), GFP_NOFS);
 	if (!rr)
 		return -ENOMEM;
 
diff --git a/fs/gfs2/super.c b/fs/gfs2/super.c
index 88497b0..7aeacbc 100644
--- a/fs/gfs2/super.c
+++ b/fs/gfs2/super.c
@@ -210,7 +210,7 @@ int gfs2_read_super(struct gfs2_sbd *sdp, sector_t sector)
 	struct page *page;
 	struct bio *bio;
 
-	page = alloc_page(GFP_KERNEL);
+	page = alloc_page(GFP_NOFS);
 	if (unlikely(!page))
 		return -ENOBUFS;
 
@@ -218,7 +218,7 @@ int gfs2_read_super(struct gfs2_sbd *sdp, sector_t sector)
 	ClearPageDirty(page);
 	lock_page(page);
 
-	bio = bio_alloc(GFP_KERNEL, 1);
+	bio = bio_alloc(GFP_NOFS, 1);
 	if (unlikely(!bio)) {
 		__free_page(page);
 		return -ENOBUFS;
-- 
1.5.1.2

[Cluster-devel] Re: [PATCH 36/48] [GFS2] fix file_system_type leak on gfs2meta mount

[Christoph Hellwig]

On Thu, Apr 17, 2008 at 09:39:12AM +0100, swhiteho at redhat.com wrote:
> From: Christoph Hellwig <hch@lst.de>
> 
> get_gfs2_sb does a get_fs_type without doing a put_filesystem and
> thus leaking a file_system_type reference everytime it's called.
> 
> Just use gfs2_fs_type directly instead of doing the lookup and thus
> fix the problem.

Btw, after this patch we should remove the get_fs_type export.  It's
not used by modules and because put_filesystem is not exported every
modular user would fundamentally have the same leak as the one fixed in
this patch.