[PATCH] Btrfs-progs: add support for mixed data+metadata block groups V3

[PATCH] Btrfs-progs: add support for mixed data+metadata block groups V3

[Josef Bacik]

So alot of crazy people (I'm looking at you Meego) want to use btrfs on phones
and such with small devices.  Unfortunately the way we split out metadata/data
chunks it makes space usage inefficient for volumes that are smaller than
1gigabyte.  So add a -M option for mixing metadata+data, and default to this
mixed mode if the filesystem is less than or equal to 1 gigabyte.  I've tested
this with xfstests on a 100mb filesystem and everything is a-ok.

Signed-off-by: Josef Bacik <josef@redhat.com>
---
V2->V3: Add an incompat flag since the kernel will panic with mixed block groups
unless its had a few key fixes.

V1->V2: My original patch was screwing up if you specified the data or metadata
profile.  This update fixes that problem.

 btrfs-vol.c  |    4 +-
 btrfs_cmds.c |   13 +++++-
 ctree.h      |    4 +-
 mkfs.c       |  122 +++++++++++++++++++++++++++++++++++++++++-----------------
 utils.c      |   10 ++--
 utils.h      |    2 +-
 6 files changed, 109 insertions(+), 46 deletions(-)

diff --git a/btrfs-vol.c b/btrfs-vol.c
index 8069778..7200bbc 100644
--- a/btrfs-vol.c
+++ b/btrfs-vol.c
@@ -129,7 +129,9 @@ int main(int ac, char **av)
 		exit(1);
 	}
 	if (cmd == BTRFS_IOC_ADD_DEV) {
-		ret = btrfs_prepare_device(devfd, device, 1, &dev_block_count);
+		int mixed = 0;
+
+		ret = btrfs_prepare_device(devfd, device, 1, &dev_block_count, &mixed);
 		if (ret) {
 			fprintf(stderr, "Unable to init %s\n", device);
 			exit(1);
diff --git a/btrfs_cmds.c b/btrfs_cmds.c
index 8031c58..683aec0 100644
--- a/btrfs_cmds.c
+++ b/btrfs_cmds.c
@@ -705,6 +705,7 @@ int do_add_volume(int nargs, char **args)
 		int	devfd, res;
 		u64 dev_block_count = 0;
 		struct stat st;
+		int mixed = 0;
 
 		devfd = open(args[i], O_RDWR);
 		if (!devfd) {
@@ -727,7 +728,7 @@ int do_add_volume(int nargs, char **args)
 			continue;
 		}
 
-		res = btrfs_prepare_device(devfd, args[i], 1, &dev_block_count);
+		res = btrfs_prepare_device(devfd, args[i], 1, &dev_block_count, &mixed);
 		if (res) {
 			fprintf(stderr, "ERROR: Unable to init '%s'\n", args[i]);
 			close(devfd);
@@ -889,8 +890,14 @@ int do_df_filesystem(int nargs, char **argv)
 		memset(description, 0, 80);
 
 		if (flags & BTRFS_BLOCK_GROUP_DATA) {
-			snprintf(description, 5, "%s", "Data");
-			written += 4;
+			if (flags & BTRFS_BLOCK_GROUP_METADATA) {
+				snprintf(description, 15, "%s",
+					 "Data+Metadata");
+				written += 14;
+			} else {
+				snprintf(description, 5, "%s", "Data");
+				written += 4;
+			}
 		} else if (flags & BTRFS_BLOCK_GROUP_SYSTEM) {
 			snprintf(description, 7, "%s", "System");
 			written += 6;
diff --git a/ctree.h b/ctree.h
index 4d0e1e2..8276e5e 100644
--- a/ctree.h
+++ b/ctree.h
@@ -352,13 +352,15 @@ struct btrfs_super_block {
 #define BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF	(1ULL << 0)
 #define BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL	(1ULL << 1)
 #define BTRFS_FEATURE_INCOMPAT_SPACE_CACHE	(1ULL << 2)
+#define BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS	(1ULL << 3)
 
 #define BTRFS_FEATURE_COMPAT_SUPP		0ULL
 #define BTRFS_FEATURE_COMPAT_RO_SUPP		0ULL
 #define BTRFS_FEATURE_INCOMPAT_SUPP			\
 	(BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF |		\
 	 BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL |	\
-	 BTRFS_FEATURE_INCOMPAT_SPACE_CACHE)
+	 BTRFS_FEATURE_INCOMPAT_SPACE_CACHE |		\
+	 BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS)
 
 /*
  * A leaf is full of items. offset and size tell us where to find
diff --git a/mkfs.c b/mkfs.c
index 2e99b95..04de93a 100644
--- a/mkfs.c
+++ b/mkfs.c
@@ -69,7 +69,7 @@ static u64 parse_size(char *s)
 	return atol(s) * mult;
 }
 
-static int make_root_dir(struct btrfs_root *root)
+static int make_root_dir(struct btrfs_root *root, int mixed)
 {
 	struct btrfs_trans_handle *trans;
 	struct btrfs_key location;
@@ -88,30 +88,47 @@ static int make_root_dir(struct btrfs_root *root)
 				     0, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
 	BUG_ON(ret);
 
-	ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
-				&chunk_start, &chunk_size,
-				BTRFS_BLOCK_GROUP_METADATA);
-	BUG_ON(ret);
-	ret = btrfs_make_block_group(trans, root, 0,
-				     BTRFS_BLOCK_GROUP_METADATA,
-				     BTRFS_FIRST_CHUNK_TREE_OBJECTID,
-				     chunk_start, chunk_size);
-	BUG_ON(ret);
+	if (mixed) {
+		ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
+					&chunk_start, &chunk_size,
+					BTRFS_BLOCK_GROUP_METADATA |
+					BTRFS_BLOCK_GROUP_DATA);
+		BUG_ON(ret);
+		ret = btrfs_make_block_group(trans, root, 0,
+					     BTRFS_BLOCK_GROUP_METADATA |
+					     BTRFS_BLOCK_GROUP_DATA,
+					     BTRFS_FIRST_CHUNK_TREE_OBJECTID,
+					     chunk_start, chunk_size);
+		BUG_ON(ret);
+		printf("Created a data/metadata chunk of size %llu\n", chunk_size);
+	} else {
+		ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
+					&chunk_start, &chunk_size,
+					BTRFS_BLOCK_GROUP_METADATA);
+		BUG_ON(ret);
+		ret = btrfs_make_block_group(trans, root, 0,
+					     BTRFS_BLOCK_GROUP_METADATA,
+					     BTRFS_FIRST_CHUNK_TREE_OBJECTID,
+					     chunk_start, chunk_size);
+		BUG_ON(ret);
+	}
 
 	root->fs_info->system_allocs = 0;
 	btrfs_commit_transaction(trans, root);
 	trans = btrfs_start_transaction(root, 1);
 	BUG_ON(!trans);
 
-	ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
-				&chunk_start, &chunk_size,
-				BTRFS_BLOCK_GROUP_DATA);
-	BUG_ON(ret);
-	ret = btrfs_make_block_group(trans, root, 0,
-				     BTRFS_BLOCK_GROUP_DATA,
-				     BTRFS_FIRST_CHUNK_TREE_OBJECTID,
-				     chunk_start, chunk_size);
-	BUG_ON(ret);
+	if (!mixed) {
+		ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
+					&chunk_start, &chunk_size,
+					BTRFS_BLOCK_GROUP_DATA);
+		BUG_ON(ret);
+		ret = btrfs_make_block_group(trans, root, 0,
+					     BTRFS_BLOCK_GROUP_DATA,
+					     BTRFS_FIRST_CHUNK_TREE_OBJECTID,
+					     chunk_start, chunk_size);
+		BUG_ON(ret);
+	}
 
 	ret = btrfs_make_root_dir(trans, root->fs_info->tree_root,
 			      BTRFS_ROOT_TREE_DIR_OBJECTID);
@@ -200,7 +217,7 @@ static int create_one_raid_group(struct btrfs_trans_handle *trans,
 
 static int create_raid_groups(struct btrfs_trans_handle *trans,
 			      struct btrfs_root *root, u64 data_profile,
-			      u64 metadata_profile)
+			      u64 metadata_profile, int mixed)
 {
 	u64 num_devices = btrfs_super_num_devices(&root->fs_info->super_copy);
 	u64 allowed;
@@ -215,20 +232,24 @@ static int create_raid_groups(struct btrfs_trans_handle *trans,
 		allowed = BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1;
 
 	if (allowed & metadata_profile) {
+		u64 meta_flags = BTRFS_BLOCK_GROUP_METADATA;
+
 		ret = create_one_raid_group(trans, root,
 					    BTRFS_BLOCK_GROUP_SYSTEM |
 					    (allowed & metadata_profile));
 		BUG_ON(ret);
 
-		ret = create_one_raid_group(trans, root,
-					    BTRFS_BLOCK_GROUP_METADATA |
+		if (mixed)
+			meta_flags |= BTRFS_BLOCK_GROUP_DATA;
+
+		ret = create_one_raid_group(trans, root, meta_flags |
 					    (allowed & metadata_profile));
 		BUG_ON(ret);
 
 		ret = recow_roots(trans, root);
 		BUG_ON(ret);
 	}
-	if (num_devices > 1 && (allowed & data_profile)) {
+	if (!mixed && num_devices > 1 && (allowed & data_profile)) {
 		ret = create_one_raid_group(trans, root,
 					    BTRFS_BLOCK_GROUP_DATA |
 					    (allowed & data_profile));
@@ -274,6 +295,7 @@ static void print_usage(void)
 	fprintf(stderr, "\t -l --leafsize size of btree leaves\n");
 	fprintf(stderr, "\t -L --label set a label\n");
 	fprintf(stderr, "\t -m --metadata metadata profile, values like data profile\n");
+	fprintf(stderr, "\t -M --mixed mix metadata and data together\n");
 	fprintf(stderr, "\t -n --nodesize size of btree nodes\n");
 	fprintf(stderr, "\t -s --sectorsize min block allocation\n");
 	fprintf(stderr, "%s\n", BTRFS_BUILD_VERSION);
@@ -328,6 +350,7 @@ static struct option long_options[] = {
 	{ "leafsize", 1, NULL, 'l' },
 	{ "label", 1, NULL, 'L'},
 	{ "metadata", 1, NULL, 'm' },
+	{ "mixed", 0, NULL, 'M' },
 	{ "nodesize", 1, NULL, 'n' },
 	{ "sectorsize", 1, NULL, 's' },
 	{ "data", 1, NULL, 'd' },
@@ -358,10 +381,13 @@ int main(int ac, char **av)
 	int first_fd;
 	int ret;
 	int i;
+	int mixed = 0;
+	int data_profile_opt = 0;
+	int metadata_profile_opt = 0;
 
 	while(1) {
 		int c;
-		c = getopt_long(ac, av, "A:b:l:n:s:m:d:L:V", long_options,
+		c = getopt_long(ac, av, "A:b:l:n:s:m:d:L:VM", long_options,
 				&option_index);
 		if (c < 0)
 			break;
@@ -371,6 +397,7 @@ int main(int ac, char **av)
 				break;
 			case 'd':
 				data_profile = parse_profile(optarg);
+				data_profile_opt = 1;
 				break;
 			case 'l':
 				leafsize = parse_size(optarg);
@@ -380,6 +407,10 @@ int main(int ac, char **av)
 				break;
 			case 'm':
 				metadata_profile = parse_profile(optarg);
+				metadata_profile_opt = 1;
+				break;
+			case 'M':
+				mixed = 1;
 				break;
 			case 'n':
 				nodesize = parse_size(optarg);
@@ -389,12 +420,10 @@ int main(int ac, char **av)
 				break;
 			case 'b':
 				block_count = parse_size(optarg);
-				if (block_count < 256*1024*1024) {
-					fprintf(stderr, "File system size "
-						"%llu bytes is too small, "
-						"256M is required at least\n",
-						(unsigned long long)block_count);
-					exit(1);
+				if (block_count <= 1024*1024*1024) {
+					printf("SMALL VOLUME: forcing mixed "
+					       "metadata/data groups\n");
+					mixed = 1;
 				}
 				zero_end = 0;
 				break;
@@ -439,9 +468,22 @@ int main(int ac, char **av)
 	}
 	first_fd = fd;
 	first_file = file;
-	ret = btrfs_prepare_device(fd, file, zero_end, &dev_block_count);
+	ret = btrfs_prepare_device(fd, file, zero_end, &dev_block_count,
+				   &mixed);
 	if (block_count == 0)
 		block_count = dev_block_count;
+	if (mixed) {
+		if (!metadata_profile_opt)
+			metadata_profile = 0;
+		if (!data_profile_opt)
+			data_profile = 0;
+
+		if (metadata_profile != data_profile) {
+			fprintf(stderr, "With mixed block groups data and metadata "
+				"profiles must be the same\n");
+			exit(1);
+		}
+	}
 
 	blocks[0] = BTRFS_SUPER_INFO_OFFSET;
 	for (i = 1; i < 7; i++) {
@@ -459,7 +501,7 @@ int main(int ac, char **av)
 	root = open_ctree(file, 0, O_RDWR);
 	root->fs_info->alloc_start = alloc_start;
 
-	ret = make_root_dir(root);
+	ret = make_root_dir(root, mixed);
 	if (ret) {
 		fprintf(stderr, "failed to setup the root directory\n");
 		exit(1);
@@ -478,6 +520,8 @@ int main(int ac, char **av)
 
 	zero_end = 1;
 	while(ac-- > 0) {
+		int old_mixed = mixed;
+
 		file = av[optind++];
 		ret = check_mounted(file);
 		if (ret < 0) {
@@ -503,8 +547,8 @@ int main(int ac, char **av)
 			continue;
 		}
 		ret = btrfs_prepare_device(fd, file, zero_end,
-					   &dev_block_count);
-
+					   &dev_block_count, &mixed);
+		mixed = old_mixed;
 		BUG_ON(ret);
 
 		ret = btrfs_add_to_fsid(trans, root, fd, file, dev_block_count,
@@ -515,12 +559,20 @@ int main(int ac, char **av)
 
 raid_groups:
 	ret = create_raid_groups(trans, root, data_profile,
-				 metadata_profile);
+				 metadata_profile, mixed);
 	BUG_ON(ret);
 
 	ret = create_data_reloc_tree(trans, root);
 	BUG_ON(ret);
 
+	if (mixed) {
+		struct btrfs_super_block *super = &root->fs_info->super_copy;
+		u64 flags = btrfs_super_incompat_flags(super);
+
+		flags |= BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS;
+		btrfs_set_super_incompat_flags(super, flags);
+	}
+
 	printf("fs created label %s on %s\n\tnodesize %u leafsize %u "
 	    "sectorsize %u size %s\n",
 	    label, first_file, nodesize, leafsize, sectorsize,
diff --git a/utils.c b/utils.c
index 2f4c6e1..0de43df 100644
--- a/utils.c
+++ b/utils.c
@@ -507,7 +507,8 @@ int btrfs_add_to_fsid(struct btrfs_trans_handle *trans,
 	return 0;
 }
 
-int btrfs_prepare_device(int fd, char *file, int zero_end, u64 *block_count_ret)
+int btrfs_prepare_device(int fd, char *file, int zero_end, u64 *block_count_ret,
+			 int *mixed)
 {
 	u64 block_count;
 	u64 bytenr;
@@ -527,10 +528,9 @@ int btrfs_prepare_device(int fd, char *file, int zero_end, u64 *block_count_ret)
 	}
 	zero_end = 1;
 
-	if (block_count < 256 * 1024 * 1024) {
-		fprintf(stderr, "device %s is too small "
-		        "(must be at least 256 MB)\n", file);
-		exit(1);
+	if (block_count < 1024 * 1024 * 1024 && !(*mixed)) {
+		printf("SMALL VOLUME: forcing mixed metadata/data groups\n");
+		*mixed = 1;
 	}
 	ret = zero_dev_start(fd);
 	if (ret) {
diff --git a/utils.h b/utils.h
index 7ff542b..b91140e 100644
--- a/utils.h
+++ b/utils.h
@@ -27,7 +27,7 @@ int make_btrfs(int fd, const char *device, const char *label,
 int btrfs_make_root_dir(struct btrfs_trans_handle *trans,
 			struct btrfs_root *root, u64 objectid);
 int btrfs_prepare_device(int fd, char *file, int zero_end,
-			 u64 *block_count_ret);
+			 u64 *block_count_ret, int *mixed);
 int btrfs_add_to_fsid(struct btrfs_trans_handle *trans,
 		      struct btrfs_root *root, int fd, char *path,
 		      u64 block_count, u32 io_width, u32 io_align,
-- 
1.6.6.1

Re: [PATCH] Btrfs-progs: add support for mixed data+metadata block groups V3

[Cláudio Martins]

On Wed, 27 Oct 2010 14:37:34 -0400 Josef Bacik <josef@redhat.com> wrote=
:
> So alot of crazy people (I'm looking at you Meego) want to use btrfs =
on phones
> and such with small devices.  Unfortunately the way we split out meta=
data/data
> chunks it makes space usage inefficient for volumes that are smaller =
than
> 1gigabyte.  So add a -M option for mixing metadata+data, and default =
to this
> mixed mode if the filesystem is less than or equal to 1 gigabyte.  I'=
ve tested
> this with xfstests on a 100mb filesystem and everything is a-ok.
>=20

 Hi,

 Could you provide some rationale as to why btrfs should support these
two modes of data/metadata (mixed and separate) allocation instead of
just always doing mixed data+metadata?

 The reason I'm asking this is that having two separate modes implies
having more codepaths in the kernel, which naturally adds complexity to
the filesystem and also more code to test, two things that should be
avoided, if possible. I'd love if you could add a small explanation to
your patch description (or a comment in the code) about why it is
useful to keep the default mode of allocation, instead of simply
droping it and doing mixed allocation in all cases.

 Thanks in advance.

Best regards

Cl=C3=A1udio

--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" =
in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] Btrfs-progs: add support for mixed data+metadata block groups V3

[Josef Bacik]

On Fri, Oct 29, 2010 at 01:20:13AM +0100, Cl=E1udio Martins wrote:
>=20
> On Wed, 27 Oct 2010 14:37:34 -0400 Josef Bacik <josef@redhat.com> wro=
te:
> > So alot of crazy people (I'm looking at you Meego) want to use btrf=
s on phones
> > and such with small devices.  Unfortunately the way we split out me=
tadata/data
> > chunks it makes space usage inefficient for volumes that are smalle=
r than
> > 1gigabyte.  So add a -M option for mixing metadata+data, and defaul=
t to this
> > mixed mode if the filesystem is less than or equal to 1 gigabyte.  =
I've tested
> > this with xfstests on a 100mb filesystem and everything is a-ok.
> >=20
>=20
>  Hi,
>=20
>  Could you provide some rationale as to why btrfs should support thes=
e
> two modes of data/metadata (mixed and separate) allocation instead of
> just always doing mixed data+metadata?
>

It makes the volume management stuff easier and gives us more flexibili=
ty for
how we lay out data and metadata.  Thanks,

Josef=20
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" =
in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] Btrfs-progs: add support for mixed data+metadata block groups V3

[Mitch Harder]

My apologies if I missed it, but this looks like the most recent
version of this patch ([PATCH] Btrfs-progs: add support for mixed
data+metadata block groups V3), and this version needs updating to
take into account that BTRFS_FEATURE_INCOMPAT_SPACE_CACHE is no longer
needed.

The kernel btrfs code expects:

#define BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS    (1ULL << 2)

But this version of the patch gives:

#define BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS    (1ULL << 3)

with (1ULL << 2) being occupied by the space_cache incompat flag.