Btrfs: superblock duplication
This patch implements superblock duplication. Superblocks
are stored at offset 16K, 64M and 256G on every devices.
Spaces used by superblocks are preserved by the allocator,
which uses a reverse mapping function to find the logical
addresses that correspond to superblocks. Thank you,
Signed-off-by: Yan Zheng <zheng.yan@oracle.com>
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 61dc3b2..c72f4f3 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -1595,8 +1595,7 @@
fs_info, BTRFS_ROOT_TREE_OBJECTID);
- bh = __bread(fs_devices->latest_bdev,
- BTRFS_SUPER_INFO_OFFSET / 4096, 4096);
+ bh = btrfs_read_dev_super(fs_devices->latest_bdev);
if (!bh)
goto fail_iput;
@@ -1710,7 +1709,7 @@
}
mutex_lock(&fs_info->chunk_mutex);
- ret = btrfs_read_sys_array(tree_root);
+ ret = btrfs_read_sys_array(tree_root, btrfs_super_bytenr(disk_super));
mutex_unlock(&fs_info->chunk_mutex);
if (ret) {
printk("btrfs: failed to read the system array on %s\n",
@@ -1905,19 +1904,147 @@
put_bh(bh);
}
-static int write_all_supers(struct btrfs_root *root)
+struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
+{
+ struct buffer_head *bh;
+ struct buffer_head *latest = NULL;
+ struct btrfs_super_block *super;
+ int i;
+ u64 transid = 0;
+ u64 bytenr;
+
+ /* we would like to check all the supers, but that would make
+ * a btrfs mount succeed after a mkfs from a different FS.
+ * So, we need to add a special mount option to scan for
+ * later supers, using BTRFS_SUPER_MIRROR_MAX instead
+ */
+ for (i = 0; i < 1; i++) {
+ bytenr = btrfs_sb_offset(i);
+ if (bytenr + 4096 >= i_size_read(bdev->bd_inode))
+ break;
+ bh = __bread(bdev, bytenr / 4096, 4096);
+ if (!bh)
+ continue;
+
+ super = (struct btrfs_super_block *)bh->b_data;
+ if (btrfs_super_bytenr(super) != bytenr ||
+ strncmp((char *)(&super->magic), BTRFS_MAGIC,
+ sizeof(super->magic))) {
+ brelse(bh);
+ continue;
+ }
+
+ if (!latest || btrfs_super_generation(super) > transid) {
+ brelse(latest);
+ latest = bh;
+ transid = btrfs_super_generation(super);
+ } else {
+ brelse(bh);
+ }
+ }
+ return latest;
+}
+
+static int write_dev_supers(struct btrfs_device *device,
+ struct btrfs_super_block *sb,
+ int do_barriers, int wait, int max_mirrors)
+{
+ struct buffer_head *bh;
+ int i;
+ int ret;
+ int errors = 0;
+ u32 crc;
+ u64 bytenr;
+ int last_barrier = 0;
+
+ if (max_mirrors == 0)
+ max_mirrors = BTRFS_SUPER_MIRROR_MAX;
+
+ /* make sure only the last submit_bh does a barrier */
+ if (do_barriers) {
+ for (i = 0; i < max_mirrors; i++) {
+ bytenr = btrfs_sb_offset(i);
+ if (bytenr + BTRFS_SUPER_INFO_SIZE >=
+ device->total_bytes)
+ break;
+ last_barrier = i;
+ }
+ }
+
+ for (i = 0; i < max_mirrors; i++) {
+ bytenr = btrfs_sb_offset(i);
+ if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
+ break;
+
+ if (wait) {
+ bh = __find_get_block(device->bdev, bytenr / 4096,
+ BTRFS_SUPER_INFO_SIZE);
+ BUG_ON(!bh);
+ brelse(bh);
+ wait_on_buffer(bh);
+ if (buffer_uptodate(bh)) {
+ brelse(bh);
+ continue;
+ }
+ } else {
+ btrfs_set_super_bytenr(sb, bytenr);
+
+ crc = ~(u32)0;
+ crc = btrfs_csum_data(NULL, (char *)sb +
+ BTRFS_CSUM_SIZE, crc,
+ BTRFS_SUPER_INFO_SIZE -
+ BTRFS_CSUM_SIZE);
+ btrfs_csum_final(crc, sb->csum);
+
+ bh = __getblk(device->bdev, bytenr / 4096,
+ BTRFS_SUPER_INFO_SIZE);
+ memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
+
+ set_buffer_uptodate(bh);
+ get_bh(bh);
+ lock_buffer(bh);
+ bh->b_end_io = btrfs_end_buffer_write_sync;
+ }
+
+ if (i == last_barrier && do_barriers && device->barriers) {
+ ret = submit_bh(WRITE_BARRIER, bh);
+ if (ret == -EOPNOTSUPP) {
+ printk("btrfs: disabling barriers on dev %s\n",
+ device->name);
+ set_buffer_uptodate(bh);
+ device->barriers = 0;
+ get_bh(bh);
+ lock_buffer(bh);
+ ret = submit_bh(WRITE, bh);
+ }
+ } else {
+ ret = submit_bh(WRITE, bh);
+ }
+
+ if (!ret && wait) {
+ wait_on_buffer(bh);
+ if (!buffer_uptodate(bh))
+ errors++;
+ } else if (ret) {
+ errors++;
+ }
+ if (wait)
+ brelse(bh);
+ }
+ return errors < i ? 0 : -1;
+}
+
+int write_all_supers(struct btrfs_root *root, int max_mirrors)
{
struct list_head *cur;
struct list_head *head = &root->fs_info->fs_devices->devices;
struct btrfs_device *dev;
struct btrfs_super_block *sb;
struct btrfs_dev_item *dev_item;
- struct buffer_head *bh;
int ret;
int do_barriers;
int max_errors;
int total_errors = 0;
- u32 crc;
u64 flags;
max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1;
@@ -1944,40 +2071,11 @@
btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
+
flags = btrfs_super_flags(sb);
btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
-
- crc = ~(u32)0;
- crc = btrfs_csum_data(root, (char *)sb + BTRFS_CSUM_SIZE, crc,
- BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
- btrfs_csum_final(crc, sb->csum);
-
- bh = __getblk(dev->bdev, BTRFS_SUPER_INFO_OFFSET / 4096,
- BTRFS_SUPER_INFO_SIZE);
-
- memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
- dev->pending_io = bh;
-
- get_bh(bh);
- set_buffer_uptodate(bh);
- lock_buffer(bh);
- bh->b_end_io = btrfs_end_buffer_write_sync;
-
- if (do_barriers && dev->barriers) {
- ret = submit_bh(WRITE_BARRIER, bh);
- if (ret == -EOPNOTSUPP) {
- printk("btrfs: disabling barriers on dev %s\n",
- dev->name);
- set_buffer_uptodate(bh);
- dev->barriers = 0;
- get_bh(bh);
- lock_buffer(bh);
- ret = submit_bh(WRITE, bh);
- }
- } else {
- ret = submit_bh(WRITE, bh);
- }
+ ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
if (ret)
total_errors++;
}
@@ -1985,8 +2083,8 @@
printk("btrfs: %d errors while writing supers\n", total_errors);
BUG();
}
- total_errors = 0;
+ total_errors = 0;
list_for_each(cur, head) {
dev = list_entry(cur, struct btrfs_device, dev_list);
if (!dev->bdev)
@@ -1994,29 +2092,9 @@
if (!dev->in_fs_metadata || !dev->writeable)
continue;
- BUG_ON(!dev->pending_io);
- bh = dev->pending_io;
- wait_on_buffer(bh);
- if (!buffer_uptodate(dev->pending_io)) {
- if (do_barriers && dev->barriers) {
- printk("btrfs: disabling barriers on dev %s\n",
- dev->name);
- set_buffer_uptodate(bh);
- get_bh(bh);
- lock_buffer(bh);
- dev->barriers = 0;
- ret = submit_bh(WRITE, bh);
- BUG_ON(ret);
- wait_on_buffer(bh);
- if (!buffer_uptodate(bh))
- total_errors++;
- } else {
- total_errors++;
- }
-
- }
- dev->pending_io = NULL;
- brelse(bh);
+ ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
+ if (ret)
+ total_errors++;
}
if (total_errors > max_errors) {
printk("btrfs: %d errors while writing supers\n", total_errors);
@@ -2025,12 +2103,12 @@
return 0;
}
-int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
- *root)
+int write_ctree_super(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root, int max_mirrors)
{
int ret;
- ret = write_all_supers(root);
+ ret = write_all_supers(root, max_mirrors);
return ret;
}
@@ -2116,7 +2194,7 @@
ret = btrfs_write_and_wait_transaction(NULL, root);
BUG_ON(ret);
- ret = write_ctree_super(NULL, root);
+ ret = write_ctree_super(NULL, root, 0);
return ret;
}
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 717e948..c0ff404 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -19,8 +19,20 @@
#ifndef __DISKIO__
#define __DISKIO__
-#define BTRFS_SUPER_INFO_OFFSET (16 * 1024)
+#define BTRFS_SUPER_INFO_OFFSET (64 * 1024)
#define BTRFS_SUPER_INFO_SIZE 4096
+
+#define BTRFS_SUPER_MIRROR_MAX 3
+#define BTRFS_SUPER_MIRROR_SHIFT 12
+
+static inline u64 btrfs_sb_offset(int mirror)
+{
+ u64 start = 16 * 1024;
+ if (mirror)
+ return start << (BTRFS_SUPER_MIRROR_SHIFT * mirror);
+ return BTRFS_SUPER_INFO_OFFSET;
+}
+
struct btrfs_device;
struct btrfs_fs_devices;
@@ -37,7 +49,8 @@
char *options);
int close_ctree(struct btrfs_root *root);
int write_ctree_super(struct btrfs_trans_handle *trans,
- struct btrfs_root *root);
+ struct btrfs_root *root, int max_mirrors);
+struct buffer_head *btrfs_read_dev_super(struct block_device *bdev);
int btrfs_commit_super(struct btrfs_root *root);
struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
u64 bytenr, u32 blocksize);
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index d156385..803647b 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -189,6 +189,29 @@
return 0;
}
+static int remove_sb_from_cache(struct btrfs_root *root,
+ struct btrfs_block_group_cache *cache)
+{
+ u64 bytenr;
+ u64 *logical;
+ int stripe_len;
+ int i, nr, ret;
+
+ for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
+ bytenr = btrfs_sb_offset(i);
+ ret = btrfs_rmap_block(&root->fs_info->mapping_tree,
+ cache->key.objectid, bytenr, 0,
+ &logical, &nr, &stripe_len);
+ BUG_ON(ret);
+ while (nr--) {
+ btrfs_remove_free_space(cache, logical[nr],
+ stripe_len);
+ }
+ kfree(logical);
+ }
+ return 0;
+}
+
static int cache_block_group(struct btrfs_root *root,
struct btrfs_block_group_cache *block_group)
{
@@ -197,9 +220,7 @@
struct btrfs_key key;
struct extent_buffer *leaf;
int slot;
- u64 last = 0;
- u64 first_free;
- int found = 0;
+ u64 last = block_group->key.objectid;
if (!block_group)
return 0;
@@ -220,23 +241,13 @@
* skip the locking here
*/
path->skip_locking = 1;
- first_free = max_t(u64, block_group->key.objectid,
- BTRFS_SUPER_INFO_OFFSET + BTRFS_SUPER_INFO_SIZE);
- key.objectid = block_group->key.objectid;
+ key.objectid = last;
key.offset = 0;
btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
if (ret < 0)
goto err;
- ret = btrfs_previous_item(root, path, 0, BTRFS_EXTENT_ITEM_KEY);
- if (ret < 0)
- goto err;
- if (ret == 0) {
- leaf = path->nodes[0];
- btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
- if (key.objectid + key.offset > first_free)
- first_free = key.objectid + key.offset;
- }
+
while(1) {
leaf = path->nodes[0];
slot = path->slots[0];
@@ -258,11 +269,6 @@
break;
if (btrfs_key_type(&key) == BTRFS_EXTENT_ITEM_KEY) {
- if (!found) {
- last = first_free;
- found = 1;
- }
-
add_new_free_space(block_group, root->fs_info, last,
key.objectid);
@@ -272,13 +278,11 @@
path->slots[0]++;
}
- if (!found)
- last = first_free;
-
add_new_free_space(block_group, root->fs_info, last,
block_group->key.objectid +
block_group->key.offset);
+ remove_sb_from_cache(root, block_group);
block_group->cached = 1;
ret = 0;
err:
@@ -1974,10 +1978,8 @@
if (alloc) {
old_val += num_bytes;
cache->space_info->bytes_used += num_bytes;
- if (cache->ro) {
+ if (cache->ro)
cache->space_info->bytes_readonly -= num_bytes;
- WARN_ON(1);
- }
btrfs_set_block_group_used(&cache->item, old_val);
spin_unlock(&cache->lock);
spin_unlock(&cache->space_info->lock);
diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c
index 09462ad..2e69b9c 100644
--- a/fs/btrfs/free-space-cache.c
+++ b/fs/btrfs/free-space-cache.c
@@ -290,7 +290,6 @@
ret = -EINVAL;
goto out;
}
-
unlink_free_space(block_group, info);
if (info->bytes == bytes) {
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index c38f6a0..47cd5fc 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -1038,7 +1038,7 @@
mutex_unlock(&root->fs_info->trans_mutex);
ret = btrfs_write_and_wait_transaction(trans, root);
BUG_ON(ret);
- write_ctree_super(trans, root);
+ write_ctree_super(trans, root, 0);
/*
* the super is written, we can safely allow the tree-loggers
diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c
index 08469ec0..d3f9c2c 100644
--- a/fs/btrfs/tree-log.c
+++ b/fs/btrfs/tree-log.c
@@ -1996,7 +1996,7 @@
btrfs_set_super_log_root_level(&root->fs_info->super_for_commit,
btrfs_header_level(log->fs_info->log_root_tree->node));
- write_ctree_super(trans, log->fs_info->tree_root);
+ write_ctree_super(trans, log->fs_info->tree_root, 2);
log->fs_info->tree_log_transid++;
log->fs_info->tree_log_batch = 0;
atomic_set(&log->fs_info->tree_log_commit, 0);
@@ -2006,7 +2006,6 @@
out:
mutex_unlock(&log->fs_info->tree_log_mutex);
return 0;
-
}
/* * free all the extents used by the tree log. This should be called
diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index 2049d17..a79b3cc 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -423,15 +423,11 @@
}
set_blocksize(bdev, 4096);
- bh = __bread(bdev, BTRFS_SUPER_INFO_OFFSET / 4096, 4096);
+ bh = btrfs_read_dev_super(bdev);
if (!bh)
goto error_close;
disk_super = (struct btrfs_super_block *)bh->b_data;
- if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
- sizeof(disk_super->magic)))
- goto error_brelse;
-
devid = le64_to_cpu(disk_super->dev_item.devid);
if (devid != device->devid)
goto error_brelse;
@@ -529,17 +525,12 @@
ret = set_blocksize(bdev, 4096);
if (ret)
goto error_close;
- bh = __bread(bdev, BTRFS_SUPER_INFO_OFFSET / 4096, 4096);
+ bh = btrfs_read_dev_super(bdev);
if (!bh) {
ret = -EIO;
goto error_close;
}
disk_super = (struct btrfs_super_block *)bh->b_data;
- if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
- sizeof(disk_super->magic))) {
- ret = -EINVAL;
- goto error_brelse;
- }
devid = le64_to_cpu(disk_super->dev_item.devid);
transid = btrfs_super_generation(disk_super);
if (disk_super->label[0])
@@ -553,7 +544,6 @@
printk("devid %Lu transid %Lu %s\n", devid, transid, path);
ret = device_list_add(path, disk_super, devid, fs_devices_ret);
-error_brelse:
brelse(bh);
error_close:
close_bdev_exclusive(bdev, flags);
@@ -1016,17 +1006,12 @@
}
set_blocksize(bdev, 4096);
- bh = __bread(bdev, BTRFS_SUPER_INFO_OFFSET / 4096, 4096);
+ bh = btrfs_read_dev_super(bdev);
if (!bh) {
ret = -EIO;
goto error_close;
}
disk_super = (struct btrfs_super_block *)bh->b_data;
- if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
- sizeof(disk_super->magic))) {
- ret = -ENOENT;
- goto error_brelse;
- }
devid = le64_to_cpu(disk_super->dev_item.devid);
dev_uuid = disk_super->dev_item.uuid;
device = btrfs_find_device(root, devid, dev_uuid,
@@ -2563,6 +2548,88 @@
mirror_num, NULL);
}
+int btrfs_rmap_block(struct btrfs_mapping_tree *map_tree,
+ u64 chunk_start, u64 physical, u64 devid,
+ u64 **logical, int *naddrs, int *stripe_len)
+{
+ struct extent_map_tree *em_tree = &map_tree->map_tree;
+ struct extent_map *em;
+ struct map_lookup *map;
+ u64 *buf;
+ u64 bytenr;
+ u64 length;
+ u64 stripe_nr;
+ int i, j, nr = 0;
+
+ spin_lock(&em_tree->lock);
+ em = lookup_extent_mapping(em_tree, chunk_start, 1);
+ spin_unlock(&em_tree->lock);
+
+ BUG_ON(!em || em->start != chunk_start);
+ map = (struct map_lookup *)em->bdev;
+
+ length = em->len;
+ if (map->type & BTRFS_BLOCK_GROUP_RAID10)
+ do_div(length, map->num_stripes / map->sub_stripes);
+ else if (map->type & BTRFS_BLOCK_GROUP_RAID0)
+ do_div(length, map->num_stripes);
+
+ buf = kzalloc(sizeof(u64) * map->num_stripes, GFP_NOFS);
+ BUG_ON(!buf);
+
+ for (i = 0; i < map->num_stripes; i++) {
+ if (devid && map->stripes[i].dev->devid != devid)
+ continue;
+ if (map->stripes[i].physical > physical ||
+ map->stripes[i].physical + length <= physical)
+ continue;
+
+ stripe_nr = physical - map->stripes[i].physical;
+ do_div(stripe_nr, map->stripe_len);
+
+ if (map->type & BTRFS_BLOCK_GROUP_RAID10) {
+ stripe_nr = stripe_nr * map->num_stripes + i;
+ do_div(stripe_nr, map->sub_stripes);
+ } else if (map->type & BTRFS_BLOCK_GROUP_RAID0) {
+ stripe_nr = stripe_nr * map->num_stripes + i;
+ }
+ bytenr = chunk_start + stripe_nr * map->stripe_len;
+ for (j = 0; j < nr; j++) {
+ if (buf[j] == bytenr)
+ break;
+ }
+ if (j == nr)
+ buf[nr++] = bytenr;
+ }
+
+ for (i = 0; i > nr; i++) {
+ struct btrfs_multi_bio *multi;
+ struct btrfs_bio_stripe *stripe;
+ int ret;
+
+ length = 1;
+ ret = btrfs_map_block(map_tree, WRITE, buf[i],
+ &length, &multi, 0);
+ BUG_ON(ret);
+
+ stripe = multi->stripes;
+ for (j = 0; j < multi->num_stripes; j++) {
+ if (stripe->physical >= physical &&
+ physical < stripe->physical + length)
+ break;
+ }
+ BUG_ON(j >= multi->num_stripes);
+ kfree(multi);
+ }
+
+ *logical = buf;
+ *naddrs = nr;
+ *stripe_len = map->stripe_len;
+
+ free_extent_map(em);
+ return 0;
+}
+
int btrfs_unplug_page(struct btrfs_mapping_tree *map_tree,
u64 logical, struct page *page)
{
@@ -3003,7 +3070,7 @@
return read_one_dev(root, buf, dev_item);
}
-int btrfs_read_sys_array(struct btrfs_root *root)
+int btrfs_read_sys_array(struct btrfs_root *root, u64 sb_bytenr)
{
struct btrfs_super_block *super_copy = &root->fs_info->super_copy;
struct extent_buffer *sb;
@@ -3018,7 +3085,7 @@
u32 cur;
struct btrfs_key key;
- sb = btrfs_find_create_tree_block(root, BTRFS_SUPER_INFO_OFFSET,
+ sb = btrfs_find_create_tree_block(root, sb_bytenr,
BTRFS_SUPER_INFO_SIZE);
if (!sb)
return -ENOMEM;
diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h
index fcbdcb3..bdebe83 100644
--- a/fs/btrfs/volumes.h
+++ b/fs/btrfs/volumes.h
@@ -28,7 +28,6 @@
struct list_head dev_alloc_list;
struct btrfs_fs_devices *fs_devices;
struct btrfs_root *dev_root;
- struct buffer_head *pending_io;
struct bio *pending_bios;
struct bio *pending_bio_tail;
int running_pending;
@@ -125,7 +124,10 @@
int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw,
u64 logical, u64 *length,
struct btrfs_multi_bio **multi_ret, int mirror_num);
-int btrfs_read_sys_array(struct btrfs_root *root);
+int btrfs_rmap_block(struct btrfs_mapping_tree *map_tree,
+ u64 chunk_start, u64 physical, u64 devid,
+ u64 **logical, int *naddrs, int *stripe_len);
+int btrfs_read_sys_array(struct btrfs_root *root, u64 sb_bytenr);
int btrfs_read_chunk_tree(struct btrfs_root *root);
int btrfs_alloc_chunk(struct btrfs_trans_handle *trans,
struct btrfs_root *extent_root, u64 type);