return NULL;
}
+ /* We're loading it the fast way, so we don't have a caching_ctl. */
+ if (!cache->caching_ctl) {
+ spin_unlock(&cache->lock);
+ return NULL;
+ }
+
ctl = cache->caching_ctl;
atomic_inc(&ctl->count);
spin_unlock(&cache->lock);
return 0;
}
-static int cache_block_group(struct btrfs_block_group_cache *cache)
+static int cache_block_group(struct btrfs_block_group_cache *cache,
+ struct btrfs_trans_handle *trans,
+ int load_cache_only)
{
struct btrfs_fs_info *fs_info = cache->fs_info;
struct btrfs_caching_control *caching_ctl;
if (cache->cached != BTRFS_CACHE_NO)
return 0;
+ /*
+ * We can't do the read from on-disk cache during a commit since we need
+ * to have the normal tree locking.
+ */
+ if (!trans->transaction->in_commit) {
+ spin_lock(&cache->lock);
+ if (cache->cached != BTRFS_CACHE_NO) {
+ spin_unlock(&cache->lock);
+ return 0;
+ }
+ cache->cached = BTRFS_CACHE_STARTED;
+ spin_unlock(&cache->lock);
+
+ ret = load_free_space_cache(fs_info, cache);
+
+ spin_lock(&cache->lock);
+ if (ret == 1) {
+ cache->cached = BTRFS_CACHE_FINISHED;
+ cache->last_byte_to_unpin = (u64)-1;
+ } else {
+ cache->cached = BTRFS_CACHE_NO;
+ }
+ spin_unlock(&cache->lock);
+ if (ret == 1)
+ return 0;
+ }
+
+ if (load_cache_only)
+ return 0;
+
caching_ctl = kzalloc(sizeof(*caching_ctl), GFP_KERNEL);
BUG_ON(!caching_ctl);
rcu_read_lock();
list_for_each_entry_rcu(found, head, list) {
- if (found->flags == flags) {
+ if (found->flags & flags) {
rcu_read_unlock();
return found;
}
return cache;
}
+static int cache_save_setup(struct btrfs_block_group_cache *block_group,
+ struct btrfs_trans_handle *trans,
+ struct btrfs_path *path)
+{
+ struct btrfs_root *root = block_group->fs_info->tree_root;
+ struct inode *inode = NULL;
+ u64 alloc_hint = 0;
+ int num_pages = 0;
+ int retries = 0;
+ int ret = 0;
+
+ /*
+ * If this block group is smaller than 100 megs don't bother caching the
+ * block group.
+ */
+ if (block_group->key.offset < (100 * 1024 * 1024)) {
+ spin_lock(&block_group->lock);
+ block_group->disk_cache_state = BTRFS_DC_WRITTEN;
+ spin_unlock(&block_group->lock);
+ return 0;
+ }
+
+again:
+ inode = lookup_free_space_inode(root, block_group, path);
+ if (IS_ERR(inode) && PTR_ERR(inode) != -ENOENT) {
+ ret = PTR_ERR(inode);
+ btrfs_release_path(root, path);
+ goto out;
+ }
+
+ if (IS_ERR(inode)) {
+ BUG_ON(retries);
+ retries++;
+
+ if (block_group->ro)
+ goto out_free;
+
+ ret = create_free_space_inode(root, trans, block_group, path);
+ if (ret)
+ goto out_free;
+ goto again;
+ }
+
+ /*
+ * We want to set the generation to 0, that way if anything goes wrong
+ * from here on out we know not to trust this cache when we load up next
+ * time.
+ */
+ BTRFS_I(inode)->generation = 0;
+ ret = btrfs_update_inode(trans, root, inode);
+ WARN_ON(ret);
+
+ if (i_size_read(inode) > 0) {
+ ret = btrfs_truncate_free_space_cache(root, trans, path,
+ inode);
+ if (ret)
+ goto out_put;
+ }
+
+ spin_lock(&block_group->lock);
+ if (block_group->cached != BTRFS_CACHE_FINISHED) {
+ spin_unlock(&block_group->lock);
+ goto out_put;
+ }
+ spin_unlock(&block_group->lock);
+
+ num_pages = (int)div64_u64(block_group->key.offset, 1024 * 1024 * 1024);
+ if (!num_pages)
+ num_pages = 1;
+
+ /*
+ * Just to make absolutely sure we have enough space, we're going to
+ * preallocate 12 pages worth of space for each block group. In
+ * practice we ought to use at most 8, but we need extra space so we can
+ * add our header and have a terminator between the extents and the
+ * bitmaps.
+ */
+ num_pages *= 16;
+ num_pages *= PAGE_CACHE_SIZE;
+
+ ret = btrfs_check_data_free_space(inode, num_pages);
+ if (ret)
+ goto out_put;
+
+ ret = btrfs_prealloc_file_range_trans(inode, trans, 0, 0, num_pages,
+ num_pages, num_pages,
+ &alloc_hint);
+ btrfs_free_reserved_data_space(inode, num_pages);
+out_put:
+ iput(inode);
+out_free:
+ btrfs_release_path(root, path);
+out:
+ spin_lock(&block_group->lock);
+ if (ret)
+ block_group->disk_cache_state = BTRFS_DC_ERROR;
+ else
+ block_group->disk_cache_state = BTRFS_DC_SETUP;
+ spin_unlock(&block_group->lock);
+
+ return ret;
+}
+
int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
struct btrfs_root *root)
{
if (!path)
return -ENOMEM;
+again:
+ while (1) {
+ cache = btrfs_lookup_first_block_group(root->fs_info, last);
+ while (cache) {
+ if (cache->disk_cache_state == BTRFS_DC_CLEAR)
+ break;
+ cache = next_block_group(root, cache);
+ }
+ if (!cache) {
+ if (last == 0)
+ break;
+ last = 0;
+ continue;
+ }
+ err = cache_save_setup(cache, trans, path);
+ last = cache->key.objectid + cache->key.offset;
+ btrfs_put_block_group(cache);
+ }
+
while (1) {
if (last == 0) {
err = btrfs_run_delayed_refs(trans, root,
cache = btrfs_lookup_first_block_group(root->fs_info, last);
while (cache) {
+ if (cache->disk_cache_state == BTRFS_DC_CLEAR) {
+ btrfs_put_block_group(cache);
+ goto again;
+ }
+
if (cache->dirty)
break;
cache = next_block_group(root, cache);
continue;
}
+ if (cache->disk_cache_state == BTRFS_DC_SETUP)
+ cache->disk_cache_state = BTRFS_DC_NEED_WRITE;
cache->dirty = 0;
last = cache->key.objectid + cache->key.offset;
btrfs_put_block_group(cache);
}
+ while (1) {
+ /*
+ * I don't think this is needed since we're just marking our
+ * preallocated extent as written, but just in case it can't
+ * hurt.
+ */
+ if (last == 0) {
+ err = btrfs_run_delayed_refs(trans, root,
+ (unsigned long)-1);
+ BUG_ON(err);
+ }
+
+ cache = btrfs_lookup_first_block_group(root->fs_info, last);
+ while (cache) {
+ /*
+ * Really this shouldn't happen, but it could if we
+ * couldn't write the entire preallocated extent and
+ * splitting the extent resulted in a new block.
+ */
+ if (cache->dirty) {
+ btrfs_put_block_group(cache);
+ goto again;
+ }
+ if (cache->disk_cache_state == BTRFS_DC_NEED_WRITE)
+ break;
+ cache = next_block_group(root, cache);
+ }
+ if (!cache) {
+ if (last == 0)
+ break;
+ last = 0;
+ continue;
+ }
+
+ btrfs_write_out_cache(root, trans, cache, path);
+
+ /*
+ * If we didn't have an error then the cache state is still
+ * NEED_WRITE, so we can set it to WRITTEN.
+ */
+ if (cache->disk_cache_state == BTRFS_DC_NEED_WRITE)
+ cache->disk_cache_state = BTRFS_DC_WRITTEN;
+ last = cache->key.objectid + cache->key.offset;
+ btrfs_put_block_group(cache);
+ }
+
btrfs_free_path(path);
return 0;
}
struct btrfs_space_info *data_sinfo;
struct btrfs_root *root = BTRFS_I(inode)->root;
u64 used;
- int ret = 0, committed = 0;
+ int ret = 0, committed = 0, alloc_chunk = 1;
/* make sure bytes are sectorsize aligned */
bytes = (bytes + root->sectorsize - 1) & ~((u64)root->sectorsize - 1);
+ if (root == root->fs_info->tree_root) {
+ alloc_chunk = 0;
+ committed = 1;
+ }
+
data_sinfo = BTRFS_I(inode)->space_info;
if (!data_sinfo)
goto alloc;
* if we don't have enough free bytes in this space then we need
* to alloc a new chunk.
*/
- if (!data_sinfo->full) {
+ if (!data_sinfo->full && alloc_chunk) {
u64 alloc_target;
data_sinfo->force_alloc = 1;
}
spin_unlock(&space_info->lock);
+ /*
+ * If we have mixed data/metadata chunks we want to make sure we keep
+ * allocating mixed chunks instead of individual chunks.
+ */
+ if (btrfs_mixed_space_info(space_info))
+ flags |= (BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA);
+
/*
* if we're doing a data chunk, go ahead and make sure that
* we keep a reasonable number of metadata chunks allocated in the
struct btrfs_root *root,
u64 bytenr, u64 num_bytes, int alloc)
{
- struct btrfs_block_group_cache *cache;
+ struct btrfs_block_group_cache *cache = NULL;
struct btrfs_fs_info *info = root->fs_info;
- int factor;
u64 total = num_bytes;
u64 old_val;
u64 byte_in_group;
+ int factor;
/* block accounting for super block */
spin_lock(&info->delalloc_lock);
factor = 2;
else
factor = 1;
+ /*
+ * If this block group has free space cache written out, we
+ * need to make sure to load it if we are removing space. This
+ * is because we need the unpinning stage to actually add the
+ * space back to the block group, otherwise we will leak space.
+ */
+ if (!alloc && cache->cached == BTRFS_CACHE_NO)
+ cache_block_group(cache, trans, 1);
+
byte_in_group = bytenr - cache->key.objectid;
WARN_ON(byte_in_group > cache->key.offset);
spin_lock(&cache->space_info->lock);
spin_lock(&cache->lock);
+
+ if (btrfs_super_cache_generation(&info->super_copy) != 0 &&
+ cache->disk_cache_state < BTRFS_DC_CLEAR)
+ cache->disk_cache_state = BTRFS_DC_CLEAR;
+
cache->dirty = 1;
old_val = btrfs_block_group_used(&cache->item);
num_bytes = min(total, cache->key.offset - byte_in_group);
bool found_uncached_bg = false;
bool failed_cluster_refill = false;
bool failed_alloc = false;
+ bool use_cluster = true;
u64 ideal_cache_percent = 0;
u64 ideal_cache_offset = 0;
return -ENOSPC;
}
+ /*
+ * If the space info is for both data and metadata it means we have a
+ * small filesystem and we can't use the clustering stuff.
+ */
+ if (btrfs_mixed_space_info(space_info))
+ use_cluster = false;
+
if (orig_root->ref_cows || empty_size)
allowed_chunk_alloc = 1;
- if (data & BTRFS_BLOCK_GROUP_METADATA) {
+ if (data & BTRFS_BLOCK_GROUP_METADATA && use_cluster) {
last_ptr = &root->fs_info->meta_alloc_cluster;
if (!btrfs_test_opt(root, SSD))
empty_cluster = 64 * 1024;
}
- if ((data & BTRFS_BLOCK_GROUP_DATA) && btrfs_test_opt(root, SSD)) {
+ if ((data & BTRFS_BLOCK_GROUP_DATA) && use_cluster &&
+ btrfs_test_opt(root, SSD)) {
last_ptr = &root->fs_info->data_alloc_cluster;
}
if (unlikely(block_group->cached == BTRFS_CACHE_NO)) {
u64 free_percent;
+ ret = cache_block_group(block_group, trans, 1);
+ if (block_group->cached == BTRFS_CACHE_FINISHED)
+ goto have_block_group;
+
free_percent = btrfs_block_group_used(&block_group->item);
free_percent *= 100;
free_percent = div64_u64(free_percent,
if (loop > LOOP_CACHING_NOWAIT ||
(loop > LOOP_FIND_IDEAL &&
atomic_read(&space_info->caching_threads) < 2)) {
- ret = cache_block_group(block_group);
+ ret = cache_block_group(block_group, trans, 0);
BUG_ON(ret);
}
found_uncached_bg = true;
u64 num_bytes = ins->offset;
block_group = btrfs_lookup_block_group(root->fs_info, ins->objectid);
- cache_block_group(block_group);
+ cache_block_group(block_group, trans, 0);
caching_ctl = get_caching_control(block_group);
if (!caching_ctl) {
return ret;
}
+void btrfs_put_block_group_cache(struct btrfs_fs_info *info)
+{
+ struct btrfs_block_group_cache *block_group;
+ u64 last = 0;
+
+ while (1) {
+ struct inode *inode;
+
+ block_group = btrfs_lookup_first_block_group(info, last);
+ while (block_group) {
+ spin_lock(&block_group->lock);
+ if (block_group->iref)
+ break;
+ spin_unlock(&block_group->lock);
+ block_group = next_block_group(info->tree_root,
+ block_group);
+ }
+ if (!block_group) {
+ if (last == 0)
+ break;
+ last = 0;
+ continue;
+ }
+
+ inode = block_group->inode;
+ block_group->iref = 0;
+ block_group->inode = NULL;
+ spin_unlock(&block_group->lock);
+ iput(inode);
+ last = block_group->key.objectid + block_group->key.offset;
+ btrfs_put_block_group(block_group);
+ }
+}
+
int btrfs_free_block_groups(struct btrfs_fs_info *info)
{
struct btrfs_block_group_cache *block_group;
struct btrfs_key key;
struct btrfs_key found_key;
struct extent_buffer *leaf;
+ int need_clear = 0;
+ u64 cache_gen;
root = info->extent_root;
key.objectid = 0;
if (!path)
return -ENOMEM;
+ cache_gen = btrfs_super_cache_generation(&root->fs_info->super_copy);
+ if (cache_gen != 0 &&
+ btrfs_super_generation(&root->fs_info->super_copy) != cache_gen)
+ need_clear = 1;
+ if (btrfs_test_opt(root, CLEAR_CACHE))
+ need_clear = 1;
+ if (!btrfs_test_opt(root, SPACE_CACHE) && cache_gen)
+ printk(KERN_INFO "btrfs: disk space caching is enabled\n");
+
while (1) {
ret = find_first_block_group(root, path, &key);
if (ret > 0)
INIT_LIST_HEAD(&cache->list);
INIT_LIST_HEAD(&cache->cluster_list);
+ if (need_clear)
+ cache->disk_cache_state = BTRFS_DC_CLEAR;
+
/*
* we only want to have 32k of ram per block group for keeping
* track of free space, and if we pass 1/2 of that we want to
cache->key.offset = size;
cache->key.type = BTRFS_BLOCK_GROUP_ITEM_KEY;
cache->sectorsize = root->sectorsize;
+ cache->fs_info = root->fs_info;
/*
* we only want to have 32k of ram per block group for keeping track
struct btrfs_path *path;
struct btrfs_block_group_cache *block_group;
struct btrfs_free_cluster *cluster;
+ struct btrfs_root *tree_root = root->fs_info->tree_root;
struct btrfs_key key;
+ struct inode *inode;
int ret;
int factor;
path = btrfs_alloc_path();
BUG_ON(!path);
+ inode = lookup_free_space_inode(root, block_group, path);
+ if (!IS_ERR(inode)) {
+ btrfs_orphan_add(trans, inode);
+ clear_nlink(inode);
+ /* One for the block groups ref */
+ spin_lock(&block_group->lock);
+ if (block_group->iref) {
+ block_group->iref = 0;
+ block_group->inode = NULL;
+ spin_unlock(&block_group->lock);
+ iput(inode);
+ } else {
+ spin_unlock(&block_group->lock);
+ }
+ /* One for our lookup ref */
+ iput(inode);
+ }
+
+ key.objectid = BTRFS_FREE_SPACE_OBJECTID;
+ key.offset = block_group->key.objectid;
+ key.type = 0;
+
+ ret = btrfs_search_slot(trans, tree_root, &key, path, -1, 1);
+ if (ret < 0)
+ goto out;
+ if (ret > 0)
+ btrfs_release_path(tree_root, path);
+ if (ret == 0) {
+ ret = btrfs_del_item(trans, tree_root, path);
+ if (ret)
+ goto out;
+ btrfs_release_path(tree_root, path);
+ }
+
spin_lock(&root->fs_info->block_group_cache_lock);
rb_erase(&block_group->cache_node,
&root->fs_info->block_group_cache_tree);
block_group->space_info->disk_total -= block_group->key.offset * factor;
spin_unlock(&block_group->space_info->lock);
+ memcpy(&key, &block_group->key, sizeof(key));
+
btrfs_clear_space_info_full(root->fs_info);
btrfs_put_block_group(block_group);