Btrfs: add mount -o ssd_spread to spread allocations out

Some SSDs perform best when reusing block numbers often, while
others perform much better when clustering strictly allocates
big chunks of unused space.

The default mount -o ssd will find rough groupings of blocks
where there are a bunch of free blocks that might have some
allocated blocks mixed in.

mount -o ssd_spread will make sure there are no allocated blocks
mixed in.  It should perform better on lower end SSDs.

Signed-off-by: Chris Mason <chris.mason@oracle.com>

5 files changed, 22 insertions, 6 deletions

diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index ce3ab4e13064..b9d8788b299e 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -1100,6 +1100,7 @@ struct btrfs_root {
 #define BTRFS_MOUNT_COMPRESS		(1 << 5)
 #define BTRFS_MOUNT_NOTREELOG           (1 << 6)
 #define BTRFS_MOUNT_FLUSHONCOMMIT       (1 << 7)
+#define BTRFS_MOUNT_SSD_SPREAD		(1 << 8)
 
 #define btrfs_clear_opt(o, opt)		((o) &= ~BTRFS_MOUNT_##opt)
 #define btrfs_set_opt(o, opt)		((o) |= BTRFS_MOUNT_##opt)
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index a42419c276e2..3355d7ea8308 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -3607,7 +3607,7 @@ refill_cluster:
 			last_ptr_loop = 0;
 
 			/* allocate a cluster in this block group */
-			ret = btrfs_find_space_cluster(trans,
+			ret = btrfs_find_space_cluster(trans, root,
 					       block_group, last_ptr,
 					       offset, num_bytes,
 					       empty_cluster + empty_size);
diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c
index ac23476beb6e..4538e48581a5 100644
--- a/fs/btrfs/free-space-cache.c
+++ b/fs/btrfs/free-space-cache.c
@@ -579,6 +579,7 @@ out:
  * it returns -enospc
  */
 int btrfs_find_space_cluster(struct btrfs_trans_handle *trans,
+			     struct btrfs_root *root,
 			     struct btrfs_block_group_cache *block_group,
 			     struct btrfs_free_cluster *cluster,
 			     u64 offset, u64 bytes, u64 empty_size)
@@ -595,7 +596,9 @@ int btrfs_find_space_cluster(struct btrfs_trans_handle *trans,
 	int ret;
 
 	/* for metadata, allow allocates with more holes */
-	if (block_group->flags & BTRFS_BLOCK_GROUP_METADATA) {
+	if (btrfs_test_opt(root, SSD_SPREAD)) {
+		min_bytes = bytes + empty_size;
+	} else if (block_group->flags & BTRFS_BLOCK_GROUP_METADATA) {
 		/*
 		 * we want to do larger allocations when we are
 		 * flushing out the delayed refs, it helps prevent
diff --git a/fs/btrfs/free-space-cache.h b/fs/btrfs/free-space-cache.h
index ab0bdc0a63ce..266fb8764054 100644
--- a/fs/btrfs/free-space-cache.h
+++ b/fs/btrfs/free-space-cache.h
@@ -31,6 +31,7 @@ void btrfs_dump_free_space(struct btrfs_block_group_cache *block_group,
 			   u64 bytes);
 u64 btrfs_block_group_free_space(struct btrfs_block_group_cache *block_group);
 int btrfs_find_space_cluster(struct btrfs_trans_handle *trans,
+			     struct btrfs_root *root,
 			     struct btrfs_block_group_cache *block_group,
 			     struct btrfs_free_cluster *cluster,
 			     u64 offset, u64 bytes, u64 empty_size);
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 22855a18eb48..7f5b2889949a 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -66,8 +66,8 @@ static void btrfs_put_super(struct super_block *sb)
 enum {
 	Opt_degraded, Opt_subvol, Opt_device, Opt_nodatasum, Opt_nodatacow,
 	Opt_max_extent, Opt_max_inline, Opt_alloc_start, Opt_nobarrier,
-	Opt_ssd, Opt_nossd, Opt_thread_pool, Opt_noacl,  Opt_compress,
-	Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_err,
+	Opt_ssd, Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl,
+	Opt_compress, Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_err,
 };
 
 static match_table_t tokens = {
@@ -83,6 +83,7 @@ static match_table_t tokens = {
 	{Opt_thread_pool, "thread_pool=%d"},
 	{Opt_compress, "compress"},
 	{Opt_ssd, "ssd"},
+	{Opt_ssd_spread, "ssd_spread"},
 	{Opt_nossd, "nossd"},
 	{Opt_noacl, "noacl"},
 	{Opt_notreelog, "notreelog"},
@@ -174,9 +175,17 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
 			printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
 			btrfs_set_opt(info->mount_opt, SSD);
 			break;
+		case Opt_ssd_spread:
+			printk(KERN_INFO "btrfs: use spread ssd "
+			       "allocation scheme\n");
+			btrfs_set_opt(info->mount_opt, SSD);
+			btrfs_set_opt(info->mount_opt, SSD_SPREAD);
+			break;
 		case Opt_nossd:
-			printk(KERN_INFO "btrfs: not using ssd allocation scheme\n");
+			printk(KERN_INFO "btrfs: not using ssd allocation "
+			       "scheme\n");
 			btrfs_clear_opt(info->mount_opt, SSD);
+			btrfs_clear_opt(info->mount_opt, SSD_SPREAD);
 			break;
 		case Opt_nobarrier:
 			printk(KERN_INFO "btrfs: turning off barriers\n");
@@ -429,7 +438,9 @@ static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
 		seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
 	if (btrfs_test_opt(root, COMPRESS))
 		seq_puts(seq, ",compress");
-	if (btrfs_test_opt(root, SSD))
+	if (btrfs_test_opt(root, SSD_SPREAD))
+		seq_puts(seq, ",ssd_spread");
+	else if (btrfs_test_opt(root, SSD))
 		seq_puts(seq, ",ssd");
 	if (btrfs_test_opt(root, NOTREELOG))
 		seq_puts(seq, ",notreelog");