btrfs-progs: allow zoned RAID

Allow for RAID levels 0, 1 and 10 on zoned devices if the RAID stripe tree
is used.

Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: David Sterba <dsterba@suse.com>

common/fsfeatures.c

@@ -190,6 +190,14 @@ static const struct btrfs_feature mkfs_features[] = {
 		VERSION_NULL(safe),
 		VERSION_NULL(default),
 		.desc		= "new extent tree format"
+	} , {
+		.name		= "raid-stripe-tree",
+		.incompat_flag	= BTRFS_FEATURE_INCOMPAT_RAID_STRIPE_TREE,
+		.sysfs_name	= NULL,
+		VERSION_TO_STRING2(compat, 6,7),
+		VERSION_NULL(safe),
+		VERSION_NULL(default),
+		.desc		= "raid stripe tree"
 	},
 #endif
 	/* Keep this one last */

kernel-shared/ctree.h

@@ -103,7 +103,8 @@ static inline u32 __BTRFS_LEAF_DATA_SIZE(u32 nodesize)
 	 BTRFS_FEATURE_INCOMPAT_RAID1C34 |		\
 	 BTRFS_FEATURE_INCOMPAT_METADATA_UUID |		\
 	 BTRFS_FEATURE_INCOMPAT_ZONED |			\
-	 BTRFS_FEATURE_INCOMPAT_EXTENT_TREE_V2)
+	 BTRFS_FEATURE_INCOMPAT_EXTENT_TREE_V2 |	\
+	 BTRFS_FEATURE_INCOMPAT_RAID_STRIPE_TREE)
 #else
 #define BTRFS_FEATURE_INCOMPAT_SUPP			\
 	(BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF |		\

kernel-shared/zoned.c

@@ -737,7 +737,7 @@ out:
 	return ret;
 }
 
-bool zoned_profile_supported(u64 map_type)
+bool zoned_profile_supported(u64 map_type, bool rst)
 {
 	bool data = (map_type & BTRFS_BLOCK_GROUP_DATA);
 	u64 flags = (map_type & BTRFS_BLOCK_GROUP_PROFILE_MASK);
@@ -746,9 +746,37 @@ bool zoned_profile_supported(u64 map_type)
 	if (flags == 0)
 		return true;
 
-	/* We can support DUP on metadata */
+#if EXPERIMENTAL
+	if (data) {
+		if ((flags & BTRFS_BLOCK_GROUP_DUP) && rst)
+			return true;
+		/* Data RAID1 needs a raid-stripe-tree. */
+		if ((flags & BTRFS_BLOCK_GROUP_RAID1_MASK) && rst)
+			return true;
+		/* Data RAID0 needs a raid-stripe-tree. */
+		if ((flags & BTRFS_BLOCK_GROUP_RAID0) && rst)
+			return true;
+		/* Data RAID10 needs a raid-stripe-tree. */
+		if ((flags & BTRFS_BLOCK_GROUP_RAID10) && rst)
+			return true;
+	} else {
+		/* We can support DUP on metadata/system. */
+		if (flags & BTRFS_BLOCK_GROUP_DUP)
+			return true;
+		/* We can support RAID1 on metadata/system. */
+		if (flags & BTRFS_BLOCK_GROUP_RAID1_MASK)
+			return true;
+		/* We can support RAID0 on metadata/system. */
+		if (flags & BTRFS_BLOCK_GROUP_RAID0)
+			return true;
+		/* We can support RAID10 on metadata/system. */
+		if (flags & BTRFS_BLOCK_GROUP_RAID10)
+			return true;
+	}
+#else
 	if (!data && (flags & BTRFS_BLOCK_GROUP_DUP))
 		return true;
+#endif
 
 	/* All other profiles are not supported yet */
 	return false;
@@ -863,7 +891,7 @@ int btrfs_load_block_group_zone_info(struct btrfs_fs_info *fs_info,
 		}
 	}
 
-	if (!zoned_profile_supported(map->type)) {
+	if (!zoned_profile_supported(map->type, !!fs_info->stripe_root)) {
 		error("zoned: profile %s not yet supported",
 		      btrfs_group_profile_str(map->type));
 		ret = -EINVAL;

kernel-shared/zoned.h

@@ -133,7 +133,7 @@ static inline bool btrfs_dev_is_empty_zone(struct btrfs_device *device, u64 pos)
 	return zinfo->zones[zno].cond == BLK_ZONE_COND_EMPTY;
 }
 
-bool zoned_profile_supported(u64 map_type);
+bool zoned_profile_supported(u64 map_type, bool rst);
 int btrfs_reset_dev_zone(int fd, struct blk_zone *zone);
 u64 btrfs_find_allocatable_zones(struct btrfs_device *device, u64 hole_start,
 				 u64 hole_end, u64 num_bytes);
@@ -214,7 +214,7 @@ static inline int btrfs_wipe_temporary_sb(struct btrfs_fs_devices *fs_devices)
 	return 0;
 }
 
-static inline bool zoned_profile_supported(u64 map_type)
+static inline bool zoned_profile_supported(u64 map_type, bool rst)
 {
 	return false;
 }

mkfs/main.c

@@ -962,6 +962,36 @@ fail:
 	return ret;
 }
 
+static int setup_raid_stripe_tree_root(struct btrfs_fs_info *fs_info)
+{
+	struct btrfs_trans_handle *trans;
+	struct btrfs_root *stripe_root;
+	struct btrfs_key key = {
+		.objectid = BTRFS_RAID_STRIPE_TREE_OBJECTID,
+		.type = BTRFS_ROOT_ITEM_KEY,
+	};
+	int ret;
+
+	trans = btrfs_start_transaction(fs_info->tree_root, 0);
+	if (IS_ERR(trans))
+		return PTR_ERR(trans);
+
+	stripe_root = btrfs_create_tree(trans, fs_info, &key);
+	if (IS_ERR(stripe_root))  {
+		ret = PTR_ERR(stripe_root);
+		btrfs_abort_transaction(trans, ret);
+		return ret;
+	}
+	fs_info->stripe_root = stripe_root;
+	add_root_to_dirty_list(stripe_root);
+
+	ret = btrfs_commit_transaction(trans, fs_info->tree_root);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
 /* Thread callback for device preparation */
 static void *prepare_one_device(void *ctx)
 {
@@ -1472,11 +1502,39 @@ int BOX_MAIN(mkfs)(int argc, char **argv)
 	if (ret)
 		goto error;
 
-	if (opt_zoned && (!zoned_profile_supported(BTRFS_BLOCK_GROUP_METADATA | metadata_profile) ||
+#if EXPERIMENTAL
-		      !zoned_profile_supported(BTRFS_BLOCK_GROUP_DATA | data_profile))) {
+	if (opt_zoned && device_count) {
+		switch (data_profile & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
+		case BTRFS_BLOCK_GROUP_DUP:
+		case BTRFS_BLOCK_GROUP_RAID1:
+		case BTRFS_BLOCK_GROUP_RAID1C3:
+		case BTRFS_BLOCK_GROUP_RAID1C4:
+		case BTRFS_BLOCK_GROUP_RAID0:
+		case BTRFS_BLOCK_GROUP_RAID10:
+			features.incompat_flags |= BTRFS_FEATURE_INCOMPAT_RAID_STRIPE_TREE;
+			break;
+		default:
+			break;
+		}
+	}
+#endif
+
+	if (opt_zoned) {
+		u64 metadata = BTRFS_BLOCK_GROUP_METADATA | metadata_profile;
+		u64 data = BTRFS_BLOCK_GROUP_DATA | data_profile;
+		bool rst = false;
+
+#if EXPERIMENTAL
+		if (features.incompat_flags & BTRFS_FEATURE_INCOMPAT_RAID_STRIPE_TREE)
+			rst = true;
+#endif
+
+		if (!zoned_profile_supported(metadata, rst) ||
+		    !zoned_profile_supported(data, rst)) {
 			error("zoned mode does not yet support RAID/DUP profiles, please specify '-d single -m single' manually");
 			goto error;
 		}
+	}
 
 	t_prepare = calloc(device_count, sizeof(*t_prepare));
 	prepare_ctx = calloc(device_count, sizeof(*prepare_ctx));
@@ -1585,6 +1643,14 @@ int BOX_MAIN(mkfs)(int argc, char **argv)
 		goto error;
 	}
 
+	if (features.incompat_flags & BTRFS_FEATURE_INCOMPAT_RAID_STRIPE_TREE) {
+		ret = setup_raid_stripe_tree_root(fs_info);
+		if (ret < 0) {
+			error("failed to initialize raid-stripe-tree: %d (%m)", ret);
+			goto out;
+		}
+	}
+
 	trans = btrfs_start_transaction(root, 1);
 	if (IS_ERR(trans)) {
 		errno = -PTR_ERR(trans);