[Cluster-devel] [GFS2] Patches to add extent allocation to GFS2

[Cluster-devel] [GFS2] Patches to add extent allocation to GFS2

[swhiteho]

This is a set of five patches which together add extent allocation to
GFS2. Also there is a bit of preliminary clean-up which results in
reducing the size of a gfs2_inode and merging some functions which
were almost identical.

The plan is to change gfs2's bmap functions next to take advantage
of these changes to improve performance, particularly with large
(i.e. multi-page) writes.

Steve.

[Cluster-devel] [PATCH 1/5] [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 9365221..a9aec62 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 268ecca..a5153c5 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 4be6bf5..1879350 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 2/5] [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 a5153c5..5612c53 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 3/5] [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 a9aec62..5f7d0e3 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 8651e09..d2f7fa7 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 4/5] [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 5f7d0e3..572f3f0 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 d2f7fa7..c19cde0 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 5/5] [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 572f3f0..ccc51ac 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 c19cde0..2b49669 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