* [PATCH V2 05/12] Btrfs: Introduce contexts for metadata reservation
@ 2010-04-26 10:44 Yan, Zheng
0 siblings, 0 replies; only message in thread
From: Yan, Zheng @ 2010-04-26 10:44 UTC (permalink / raw)
To: linux-btrfs; +Cc: Chris Mason
Introducing metadata reseravtion contexts has two major advantages.
First, it makes metadata reseravtion more traceable. Second, it can
reclaim freed space and re-add them to the itself after transaction
committed.
Besides add btrfs_block_rsv structure and related helper functions,
This patch contains following changes:
Move code that decides if freed tree block should be pinned into
btrfs_free_tree_block().
Make space accounting more accuracy, mainly for handling read only
block groups.
diff -urp 2/fs/btrfs/ctree.c 3/fs/btrfs/ctree.c
--- 2/fs/btrfs/ctree.c 2010-04-26 17:24:55.227081691 +0800
+++ 3/fs/btrfs/ctree.c 2010-04-26 17:24:55.245079481 +0800
@@ -279,7 +279,8 @@ int btrfs_block_can_be_shared(struct btr
static noinline int update_ref_for_cow(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct extent_buffer *buf,
- struct extent_buffer *cow)
+ struct extent_buffer *cow,
+ int *last_ref)
{
u64 refs;
u64 owner;
@@ -365,6 +366,7 @@ static noinline int update_ref_for_cow(s
BUG_ON(ret);
}
clean_tree_block(trans, root, buf);
+ *last_ref = 1;
}
return 0;
}
@@ -391,6 +393,7 @@ static noinline int __btrfs_cow_block(st
struct btrfs_disk_key disk_key;
struct extent_buffer *cow;
int level;
+ int last_ref = 0;
int unlock_orig = 0;
u64 parent_start;
@@ -441,7 +444,7 @@ static noinline int __btrfs_cow_block(st
(unsigned long)btrfs_header_fsid(cow),
BTRFS_FSID_SIZE);
- update_ref_for_cow(trans, root, buf, cow);
+ update_ref_for_cow(trans, root, buf, cow, &last_ref);
if (buf == root->node) {
WARN_ON(parent && parent != buf);
@@ -456,8 +459,8 @@ static noinline int __btrfs_cow_block(st
extent_buffer_get(cow);
spin_unlock(&root->node_lock);
- btrfs_free_tree_block(trans, root, buf->start, buf->len,
- parent_start, root->root_key.objectid, level);
+ btrfs_free_tree_block(trans, root, buf, parent_start,
+ last_ref);
free_extent_buffer(buf);
add_root_to_dirty_list(root);
} else {
@@ -472,8 +475,8 @@ static noinline int __btrfs_cow_block(st
btrfs_set_node_ptr_generation(parent, parent_slot,
trans->transid);
btrfs_mark_buffer_dirty(parent);
- btrfs_free_tree_block(trans, root, buf->start, buf->len,
- parent_start, root->root_key.objectid, level);
+ btrfs_free_tree_block(trans, root, buf, parent_start,
+ last_ref);
}
if (unlock_orig)
btrfs_tree_unlock(buf);
@@ -948,6 +951,22 @@ int btrfs_bin_search(struct extent_buffe
return bin_search(eb, key, level, slot);
}
+static void root_add_used(struct btrfs_root *root, u32 size)
+{
+ spin_lock(&root->accounting_lock);
+ btrfs_set_root_used(&root->root_item,
+ btrfs_root_used(&root->root_item) + size);
+ spin_unlock(&root->accounting_lock);
+}
+
+static void root_sub_used(struct btrfs_root *root, u32 size)
+{
+ spin_lock(&root->accounting_lock);
+ btrfs_set_root_used(&root->root_item,
+ btrfs_root_used(&root->root_item) - size);
+ spin_unlock(&root->accounting_lock);
+}
+
/* given a node and slot number, this reads the blocks it points to. The
* extent buffer is returned with a reference taken (but unlocked).
* NULL is returned on error.
@@ -1018,7 +1037,11 @@ static noinline int balance_level(struct
btrfs_tree_lock(child);
btrfs_set_lock_blocking(child);
ret = btrfs_cow_block(trans, root, child, mid, 0, &child);
- BUG_ON(ret);
+ if (ret) {
+ btrfs_tree_unlock(child);
+ free_extent_buffer(child);
+ goto enospc;
+ }
spin_lock(&root->node_lock);
root->node = child;
@@ -1033,11 +1056,12 @@ static noinline int balance_level(struct
btrfs_tree_unlock(mid);
/* once for the path */
free_extent_buffer(mid);
- ret = btrfs_free_tree_block(trans, root, mid->start, mid->len,
- 0, root->root_key.objectid, level);
+
+ root_sub_used(root, mid->len);
+ btrfs_free_tree_block(trans, root, mid, 0, 1);
/* once for the root ptr */
free_extent_buffer(mid);
- return ret;
+ return 0;
}
if (btrfs_header_nritems(mid) >
BTRFS_NODEPTRS_PER_BLOCK(root) / 4)
@@ -1087,23 +1111,16 @@ static noinline int balance_level(struct
if (wret < 0 && wret != -ENOSPC)
ret = wret;
if (btrfs_header_nritems(right) == 0) {
- u64 bytenr = right->start;
- u32 blocksize = right->len;
-
clean_tree_block(trans, root, right);
btrfs_tree_unlock(right);
- free_extent_buffer(right);
- right = NULL;
wret = del_ptr(trans, root, path, level + 1, pslot +
1);
if (wret)
ret = wret;
- wret = btrfs_free_tree_block(trans, root,
- bytenr, blocksize, 0,
- root->root_key.objectid,
- level);
- if (wret)
- ret = wret;
+ root_sub_used(root, right->len);
+ btrfs_free_tree_block(trans, root, right, 0, 1);
+ free_extent_buffer(right);
+ right = NULL;
} else {
struct btrfs_disk_key right_key;
btrfs_node_key(right, &right_key, 0);
@@ -1135,21 +1152,15 @@ static noinline int balance_level(struct
BUG_ON(wret == 1);
}
if (btrfs_header_nritems(mid) == 0) {
- /* we've managed to empty the middle node, drop it */
- u64 bytenr = mid->start;
- u32 blocksize = mid->len;
-
clean_tree_block(trans, root, mid);
btrfs_tree_unlock(mid);
- free_extent_buffer(mid);
- mid = NULL;
wret = del_ptr(trans, root, path, level + 1, pslot);
if (wret)
ret = wret;
- wret = btrfs_free_tree_block(trans, root, bytenr, blocksize,
- 0, root->root_key.objectid, level);
- if (wret)
- ret = wret;
+ root_sub_used(root, mid->len);
+ btrfs_free_tree_block(trans, root, mid, 0, 1);
+ free_extent_buffer(mid);
+ mid = NULL;
} else {
/* update the parent key to reflect our changes */
struct btrfs_disk_key mid_key;
@@ -1739,7 +1750,6 @@ again:
p->nodes[level + 1],
p->slots[level + 1], &b);
if (err) {
- free_extent_buffer(b);
ret = err;
goto done;
}
@@ -2075,6 +2085,8 @@ static noinline int insert_new_root(stru
if (IS_ERR(c))
return PTR_ERR(c);
+ root_add_used(root, root->nodesize);
+
memset_extent_buffer(c, 0, 0, sizeof(struct btrfs_header));
btrfs_set_header_nritems(c, 1);
btrfs_set_header_level(c, level);
@@ -2133,6 +2145,7 @@ static int insert_ptr(struct btrfs_trans
int nritems;
BUG_ON(!path->nodes[level]);
+ btrfs_assert_tree_locked(path->nodes[level]);
lower = path->nodes[level];
nritems = btrfs_header_nritems(lower);
BUG_ON(slot > nritems);
@@ -2201,6 +2214,8 @@ static noinline int split_node(struct bt
if (IS_ERR(split))
return PTR_ERR(split);
+ root_add_used(root, root->nodesize);
+
memset_extent_buffer(split, 0, 0, sizeof(struct btrfs_header));
btrfs_set_header_level(split, btrfs_header_level(c));
btrfs_set_header_bytenr(split, split->start);
@@ -2414,6 +2429,9 @@ static noinline int __push_leaf_right(st
if (left_nritems)
btrfs_mark_buffer_dirty(left);
+ else
+ clean_tree_block(trans, root, left);
+
btrfs_mark_buffer_dirty(right);
btrfs_item_key(right, &disk_key, 0);
@@ -2659,6 +2677,8 @@ static noinline int __push_leaf_left(str
btrfs_mark_buffer_dirty(left);
if (right_nritems)
btrfs_mark_buffer_dirty(right);
+ else
+ clean_tree_block(trans, root, right);
btrfs_item_key(right, &disk_key, 0);
wret = fixup_low_keys(trans, root, path, &disk_key, 1);
@@ -2668,8 +2688,6 @@ static noinline int __push_leaf_left(str
/* then fixup the leaf pointer in the path */
if (path->slots[0] < push_items) {
path->slots[0] += old_left_nritems;
- if (btrfs_header_nritems(path->nodes[0]) == 0)
- clean_tree_block(trans, root, path->nodes[0]);
btrfs_tree_unlock(path->nodes[0]);
free_extent_buffer(path->nodes[0]);
path->nodes[0] = left;
@@ -2931,10 +2949,10 @@ again:
right = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
root->root_key.objectid,
&disk_key, 0, l->start, 0);
- if (IS_ERR(right)) {
- BUG_ON(1);
+ if (IS_ERR(right))
return PTR_ERR(right);
- }
+
+ root_add_used(root, root->leafsize);
memset_extent_buffer(right, 0, 0, sizeof(struct btrfs_header));
btrfs_set_header_bytenr(right, right->start);
@@ -3053,7 +3071,8 @@ static noinline int setup_leaf_for_split
btrfs_set_path_blocking(path);
ret = split_leaf(trans, root, &key, path, ins_len, 1);
- BUG_ON(ret);
+ if (ret)
+ goto err;
path->keep_locks = 0;
btrfs_unlock_up_safe(path, 1);
@@ -3795,9 +3814,10 @@ static noinline int btrfs_del_leaf(struc
*/
btrfs_unlock_up_safe(path, 0);
- ret = btrfs_free_tree_block(trans, root, leaf->start, leaf->len,
- 0, root->root_key.objectid, 0);
- return ret;
+ root_sub_used(root, leaf->len);
+
+ btrfs_free_tree_block(trans, root, leaf, 0, 1);
+ return 0;
}
/*
* delete the item at the leaf level in path. If that empties
@@ -3864,6 +3884,8 @@ int btrfs_del_items(struct btrfs_trans_h
if (leaf == root->node) {
btrfs_set_header_level(leaf, 0);
} else {
+ btrfs_set_path_blocking(path);
+ clean_tree_block(trans, root, leaf);
ret = btrfs_del_leaf(trans, root, path, leaf);
BUG_ON(ret);
}
diff -urp 2/fs/btrfs/ctree.h 3/fs/btrfs/ctree.h
--- 2/fs/btrfs/ctree.h 2010-04-26 17:24:55.237085267 +0800
+++ 3/fs/btrfs/ctree.h 2010-04-26 17:24:55.246080391 +0800
@@ -706,6 +706,20 @@ struct btrfs_space_info {
atomic_t caching_threads;
};
+struct btrfs_block_rsv {
+ u64 size;
+ u64 reserved;
+ u64 freed[2];
+ struct btrfs_space_info *space_info;
+ struct list_head list;
+ spinlock_t lock;
+ atomic_t usage;
+ unsigned int priority:8;
+ unsigned int durable:1;
+ unsigned int refill_used:1;
+ unsigned int full:1;
+};
+
/*
* free clusters are used to claim free space in relatively large chunks,
* allowing us to do less seeky writes. They are used for all metadata
@@ -756,6 +770,7 @@ struct btrfs_block_group_cache {
spinlock_t lock;
u64 pinned;
u64 reserved;
+ u64 reserved_pinned;
u64 bytes_super;
u64 flags;
u64 sectorsize;
@@ -821,6 +836,22 @@ struct btrfs_fs_info {
/* logical->physical extent mapping */
struct btrfs_mapping_tree mapping_tree;
+ /* block reservation for extent, checksum and root tree */
+ struct btrfs_block_rsv global_block_rsv;
+ /* block reservation for delay allocation */
+ struct btrfs_block_rsv delalloc_block_rsv;
+ /* block reservation for metadata operations */
+ struct btrfs_block_rsv trans_block_rsv;
+ /* block reservation for chunk tree */
+ struct btrfs_block_rsv chunk_block_rsv;
+
+ struct btrfs_block_rsv empty_block_rsv;
+
+ /* list of block reservations that cross multiple transactions */
+ struct list_head durable_block_rsv_list;
+
+ struct mutex durable_block_rsv_mutex;
+
u64 generation;
u64 last_trans_committed;
@@ -1007,6 +1038,9 @@ struct btrfs_root {
struct completion kobj_unregister;
struct mutex objectid_mutex;
+ spinlock_t accounting_lock;
+ struct btrfs_block_rsv *block_rsv;
+
struct mutex log_mutex;
wait_queue_head_t log_writer_wait;
wait_queue_head_t log_commit_wait[2];
@@ -1979,10 +2013,10 @@ struct extent_buffer *btrfs_alloc_free_b
u64 parent, u64 root_objectid,
struct btrfs_disk_key *key, int level,
u64 hint, u64 empty_size);
-int btrfs_free_tree_block(struct btrfs_trans_handle *trans,
- struct btrfs_root *root,
- u64 bytenr, u32 blocksize,
- u64 parent, u64 root_objectid, int level);
+void btrfs_free_tree_block(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ struct extent_buffer *buf,
+ u64 parent, int last_ref);
struct extent_buffer *btrfs_init_new_buffer(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
u64 bytenr, u32 blocksize,
@@ -2036,9 +2070,6 @@ int btrfs_make_block_group(struct btrfs_
u64 size);
int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
struct btrfs_root *root, u64 group_start);
-int btrfs_prepare_block_group_relocation(struct btrfs_root *root,
- struct btrfs_block_group_cache *group);
-
u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags);
void btrfs_set_inode_space_info(struct btrfs_root *root, struct inode *ionde);
void btrfs_clear_space_info_full(struct btrfs_fs_info *info);
@@ -2057,6 +2088,30 @@ void btrfs_delalloc_reserve_space(struct
u64 bytes);
void btrfs_delalloc_free_space(struct btrfs_root *root, struct inode *inode,
u64 bytes);
+void btrfs_init_block_rsv(struct btrfs_block_rsv *rsv);
+struct btrfs_block_rsv *btrfs_alloc_block_rsv(struct btrfs_root *root);
+void btrfs_free_block_rsv(struct btrfs_root *root,
+ struct btrfs_block_rsv *rsv);
+void btrfs_add_durable_block_rsv(struct btrfs_fs_info *fs_info,
+ struct btrfs_block_rsv *rsv);
+int btrfs_block_rsv_add(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ struct btrfs_block_rsv *block_rsv,
+ u64 num_bytes, int *retries);
+int btrfs_block_rsv_check(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ struct btrfs_block_rsv *block_rsv,
+ u64 min_reserved, int min_factor);
+int btrfs_block_rsv_migrate(struct btrfs_block_rsv *src_rsv,
+ struct btrfs_block_rsv *dst_rsv,
+ u64 num_bytes);
+void btrfs_block_rsv_release(struct btrfs_root *root,
+ struct btrfs_block_rsv *block_rsv,
+ u64 num_bytes);
+int btrfs_set_block_group_ro(struct btrfs_root *root,
+ struct btrfs_block_group_cache *cache);
+int btrfs_set_block_group_rw(struct btrfs_root *root,
+ struct btrfs_block_group_cache *cache);
/* ctree.c */
int btrfs_bin_search(struct extent_buffer *eb, struct btrfs_key *key,
int level, int *slot);
diff -urp 2/fs/btrfs/disk-io.c 3/fs/btrfs/disk-io.c
--- 2/fs/btrfs/disk-io.c 2010-04-26 17:24:55.225081619 +0800
+++ 3/fs/btrfs/disk-io.c 2010-04-26 17:24:55.247080601 +0800
@@ -904,6 +904,7 @@ static int __setup_root(u32 nodesize, u3
root->name = NULL;
root->in_sysfs = 0;
root->inode_tree = RB_ROOT;
+ root->block_rsv = NULL;
INIT_LIST_HEAD(&root->dirty_list);
INIT_LIST_HEAD(&root->orphan_list);
@@ -911,6 +912,7 @@ static int __setup_root(u32 nodesize, u3
spin_lock_init(&root->node_lock);
spin_lock_init(&root->list_lock);
spin_lock_init(&root->inode_lock);
+ spin_lock_init(&root->accounting_lock);
mutex_init(&root->objectid_mutex);
mutex_init(&root->log_mutex);
init_waitqueue_head(&root->log_writer_wait);
@@ -1629,6 +1631,13 @@ struct btrfs_root *open_ctree(struct sup
INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
INIT_LIST_HEAD(&fs_info->space_info);
btrfs_mapping_init(&fs_info->mapping_tree);
+ btrfs_init_block_rsv(&fs_info->global_block_rsv);
+ btrfs_init_block_rsv(&fs_info->delalloc_block_rsv);
+ btrfs_init_block_rsv(&fs_info->trans_block_rsv);
+ btrfs_init_block_rsv(&fs_info->chunk_block_rsv);
+ btrfs_init_block_rsv(&fs_info->empty_block_rsv);
+ INIT_LIST_HEAD(&fs_info->durable_block_rsv_list);
+ mutex_init(&fs_info->durable_block_rsv_mutex);
atomic_set(&fs_info->nr_async_submits, 0);
atomic_set(&fs_info->async_delalloc_pages, 0);
atomic_set(&fs_info->async_submit_draining, 0);
diff -urp 2/fs/btrfs/extent-tree.c 3/fs/btrfs/extent-tree.c
--- 2/fs/btrfs/extent-tree.c 2010-04-26 17:24:55.239079753 +0800
+++ 3/fs/btrfs/extent-tree.c 2010-04-26 17:25:29.283857877 +0800
@@ -34,10 +34,9 @@
static int update_block_group(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
- u64 bytenr, u64 num_bytes, int alloc,
- int mark_free);
-static int update_reserved_extents(struct btrfs_block_group_cache *cache,
- u64 num_bytes, int reserve);
+ u64 bytenr, u64 num_bytes, int alloc);
+static int update_reserved_bytes(struct btrfs_block_group_cache *cache,
+ u64 num_bytes, int reserve, int sinfo);
static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
u64 bytenr, u64 num_bytes, u64 parent,
@@ -60,12 +59,6 @@ static int alloc_reserved_tree_block(str
static int do_chunk_alloc(struct btrfs_trans_handle *trans,
struct btrfs_root *extent_root, u64 alloc_bytes,
u64 flags, int force);
-static int pin_down_bytes(struct btrfs_trans_handle *trans,
- struct btrfs_root *root,
- struct btrfs_path *path,
- u64 bytenr, u64 num_bytes,
- int is_data, int reserved,
- struct extent_buffer **must_clean);
static int find_next_key(struct btrfs_path *path, int level,
struct btrfs_key *key);
static void dump_space_info(struct btrfs_space_info *info, u64 bytes,
@@ -96,8 +89,12 @@ void btrfs_get_block_group(struct btrfs_
void btrfs_put_block_group(struct btrfs_block_group_cache *cache)
{
- if (atomic_dec_and_test(&cache->count))
+ if (atomic_dec_and_test(&cache->count)) {
+ WARN_ON(cache->pinned > 0);
+ WARN_ON(cache->reserved > 0);
+ WARN_ON(cache->reserved_pinned > 0);
kfree(cache);
+ }
}
/*
@@ -324,7 +321,7 @@ static int caching_kthread(void *data)
exclude_super_stripes(extent_root, block_group);
spin_lock(&block_group->space_info->lock);
- block_group->space_info->bytes_super += block_group->bytes_super;
+ block_group->space_info->bytes_readonly += block_group->bytes_super;
spin_unlock(&block_group->space_info->lock);
last = max_t(u64, block_group->key.objectid, BTRFS_SUPER_INFO_OFFSET);
@@ -1879,7 +1876,6 @@ static int run_delayed_tree_ref(struct b
return ret;
}
-
/* helper function to actually process a single delayed ref entry */
static int run_one_delayed_ref(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
@@ -1899,32 +1895,14 @@ static int run_one_delayed_ref(struct bt
BUG_ON(extent_op);
head = btrfs_delayed_node_to_head(node);
if (insert_reserved) {
- int mark_free = 0;
- struct extent_buffer *must_clean = NULL;
-
- ret = pin_down_bytes(trans, root, NULL,
- node->bytenr, node->num_bytes,
- head->is_data, 1, &must_clean);
- if (ret > 0)
- mark_free = 1;
-
- if (must_clean) {
- clean_tree_block(NULL, root, must_clean);
- btrfs_tree_unlock(must_clean);
- free_extent_buffer(must_clean);
- }
+ btrfs_pin_extent(root, node->bytenr,
+ node->num_bytes, 1);
if (head->is_data) {
ret = btrfs_del_csums(trans, root,
node->bytenr,
node->num_bytes);
BUG_ON(ret);
}
- if (mark_free) {
- ret = btrfs_free_reserved_extent(root,
- node->bytenr,
- node->num_bytes);
- BUG_ON(ret);
- }
}
mutex_unlock(&head->mutex);
return 0;
@@ -2355,6 +2333,8 @@ int btrfs_cross_ref_exist(struct btrfs_t
ret = 0;
out:
btrfs_free_path(path);
+ if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID)
+ WARN_ON(ret > 0);
return ret;
}
@@ -2705,7 +2685,7 @@ static int update_space_info(struct btrf
found->bytes_pinned = 0;
found->bytes_reserved = 0;
found->bytes_readonly = 0;
- found->bytes_delalloc = 0;
+ found->bytes_may_use = 0;
found->full = 0;
found->force_alloc = 0;
*space_info = found;
@@ -2730,19 +2710,6 @@ static void set_avail_alloc_bits(struct
}
}
-static void set_block_group_readonly(struct btrfs_block_group_cache *cache)
-{
- spin_lock(&cache->space_info->lock);
- spin_lock(&cache->lock);
- if (!cache->ro) {
- cache->space_info->bytes_readonly += cache->key.offset -
- btrfs_block_group_used(&cache->item);
- cache->ro = 1;
- }
- spin_unlock(&cache->lock);
- spin_unlock(&cache->space_info->lock);
-}
-
u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags)
{
u64 num_devices = root->fs_info->fs_devices->rw_devices;
@@ -2801,11 +2768,8 @@ static u64 btrfs_get_alloc_profile(struc
void btrfs_set_inode_space_info(struct btrfs_root *root, struct inode *inode)
{
- u64 alloc_target;
-
- alloc_target = btrfs_get_alloc_profile(root, 1);
BTRFS_I(inode)->space_info = __find_space_info(root->fs_info,
- alloc_target);
+ BTRFS_BLOCK_GROUP_DATA);
}
static u64 calculate_bytes_needed(struct btrfs_root *root, int num_items)
@@ -3411,10 +3375,334 @@ static int shrink_delalloc(struct btrfs_
return reclaimed >= to_reclaim;
}
+static int should_retry_reserve(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ struct btrfs_block_rsv *block_rsv,
+ u64 num_bytes, int *retries)
+{
+ struct btrfs_space_info *space_info = block_rsv->space_info;
+ int ret;
+
+ if ((*retries) > 2)
+ return -ENOSPC;
+
+ ret = maybe_allocate_chunk(trans, root, space_info, num_bytes);
+ if (ret)
+ return 1;
+
+ if (trans && trans->transaction->in_commit)
+ return -ENOSPC;
+
+ ret = shrink_delalloc(trans, root, space_info, num_bytes);
+ if (ret)
+ return ret;
+
+ spin_lock(&space_info->lock);
+ if (space_info->bytes_pinned < num_bytes)
+ ret = 1;
+ spin_unlock(&space_info->lock);
+ if (ret)
+ return -ENOSPC;
+
+ (*retries)++;
+
+ if (trans)
+ return -EAGAIN;
+
+ trans = btrfs_join_transaction(root, 1);
+ BUG_ON(IS_ERR(trans));
+ ret = btrfs_commit_transaction(trans, root);
+ BUG_ON(ret);
+
+ return 1;
+}
+
+static int reserve_metadata_bytes(struct btrfs_block_rsv *block_rsv,
+ u64 num_bytes)
+{
+ struct btrfs_space_info *space_info = block_rsv->space_info;
+ u64 unused;
+ int ret = -ENOSPC;
+
+ spin_lock(&space_info->lock);
+ unused = space_info->bytes_used + space_info->bytes_reserved +
+ space_info->bytes_pinned + space_info->bytes_readonly;
+
+ if (unused < space_info->total_bytes)
+ unused = space_info->total_bytes - unused;
+ else
+ unused = 0;
+
+ if (unused >= num_bytes) {
+ if (block_rsv->priority >= 10) {
+ space_info->bytes_reserved += num_bytes;
+ ret = 0;
+ } else {
+ if ((unused + block_rsv->reserved) *
+ block_rsv->priority >=
+ (num_bytes + block_rsv->reserved) * 10) {
+ space_info->bytes_reserved += num_bytes;
+ ret = 0;
+ }
+ }
+ }
+ spin_unlock(&space_info->lock);
+
+ return ret;
+}
+
+static struct btrfs_block_rsv *get_block_rsv(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root)
+{
+ struct btrfs_block_rsv *block_rsv;
+ if (root->ref_cows)
+ block_rsv = trans->block_rsv;
+ else
+ block_rsv = root->block_rsv;
+
+ if (!block_rsv)
+ block_rsv = &root->fs_info->empty_block_rsv;
+
+ return block_rsv;
+}
+
+static int block_rsv_use_bytes(struct btrfs_block_rsv *block_rsv,
+ u64 num_bytes)
+{
+ int ret = -ENOSPC;
+ spin_lock(&block_rsv->lock);
+ if (block_rsv->reserved >= num_bytes) {
+ block_rsv->reserved -= num_bytes;
+ if (block_rsv->reserved < block_rsv->size)
+ block_rsv->full = 0;
+ ret = 0;
+ }
+ spin_unlock(&block_rsv->lock);
+ return ret;
+}
+
+static void block_rsv_add_bytes(struct btrfs_block_rsv *block_rsv,
+ u64 num_bytes, int update_size)
+{
+ spin_lock(&block_rsv->lock);
+ block_rsv->reserved += num_bytes;
+ if (update_size)
+ block_rsv->size += num_bytes;
+ else if (block_rsv->reserved >= block_rsv->size)
+ block_rsv->full = 1;
+ spin_unlock(&block_rsv->lock);
+}
+
+void block_rsv_release_bytes(struct btrfs_block_rsv *block_rsv,
+ struct btrfs_block_rsv *dest, u64 num_bytes)
+{
+ struct btrfs_space_info *space_info = block_rsv->space_info;
+
+ spin_lock(&block_rsv->lock);
+ if (num_bytes == (u64)-1)
+ num_bytes = block_rsv->size;
+ block_rsv->size -= num_bytes;
+ if (block_rsv->reserved >= block_rsv->size) {
+ num_bytes = block_rsv->reserved - block_rsv->size;
+ block_rsv->reserved = block_rsv->size;
+ block_rsv->full = 1;
+ } else {
+ num_bytes = 0;
+ }
+ spin_unlock(&block_rsv->lock);
+
+ if (num_bytes > 0) {
+ if (dest) {
+ block_rsv_add_bytes(dest, num_bytes, 0);
+ } else {
+ spin_lock(&space_info->lock);
+ space_info->bytes_reserved -= num_bytes;
+ spin_unlock(&space_info->lock);
+ }
+ }
+}
+
+static int block_rsv_migrate_bytes(struct btrfs_block_rsv *src,
+ struct btrfs_block_rsv *dst, u64 num_bytes)
+{
+ int ret;
+
+ ret = block_rsv_use_bytes(src, num_bytes);
+ if (ret)
+ return ret;
+
+ block_rsv_add_bytes(dst, num_bytes, 1);
+ return 0;
+}
+
+void btrfs_init_block_rsv(struct btrfs_block_rsv *rsv)
+{
+ memset(rsv, 0, sizeof(*rsv));
+ spin_lock_init(&rsv->lock);
+ atomic_set(&rsv->usage, 1);
+ rsv->priority = 6;
+ INIT_LIST_HEAD(&rsv->list);
+}
+
+struct btrfs_block_rsv *btrfs_alloc_block_rsv(struct btrfs_root *root)
+{
+ struct btrfs_block_rsv *block_rsv;
+ struct btrfs_fs_info *fs_info = root->fs_info;
+ u64 alloc_target;
+
+ block_rsv = kmalloc(sizeof(*block_rsv), GFP_NOFS);
+ if (!block_rsv)
+ return NULL;
+
+ btrfs_init_block_rsv(block_rsv);
+
+ alloc_target = btrfs_get_alloc_profile(root, 0);
+ block_rsv->space_info = __find_space_info(fs_info,
+ BTRFS_BLOCK_GROUP_METADATA);
+
+ return block_rsv;
+}
+
+void btrfs_free_block_rsv(struct btrfs_root *root,
+ struct btrfs_block_rsv *rsv)
+{
+ if (rsv && atomic_dec_and_test(&rsv->usage)) {
+ btrfs_block_rsv_release(root, rsv, (u64)-1);
+ if (!rsv->durable)
+ kfree(rsv);
+ }
+}
+
+/*
+ * make the block_rsv struct be able to capture freed space.
+ * the captured space will re-add to the the block_rsv struct
+ * after transaction commit
+ */
+void btrfs_add_durable_block_rsv(struct btrfs_fs_info *fs_info,
+ struct btrfs_block_rsv *block_rsv)
+{
+ block_rsv->durable = 1;
+ mutex_lock(&fs_info->durable_block_rsv_mutex);
+ list_add_tail(&block_rsv->list, &fs_info->durable_block_rsv_list);
+ mutex_unlock(&fs_info->durable_block_rsv_mutex);
+}
+
+int btrfs_block_rsv_add(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ struct btrfs_block_rsv *block_rsv,
+ u64 num_bytes, int *retries)
+{
+ int ret;
+
+ if (num_bytes == 0)
+ return 0;
+again:
+ ret = reserve_metadata_bytes(block_rsv, num_bytes);
+ if (!ret) {
+ block_rsv_add_bytes(block_rsv, num_bytes, 1);
+ return 0;
+ }
+
+ ret = should_retry_reserve(trans, root, block_rsv, num_bytes, retries);
+ if (ret > 0)
+ goto again;
+
+ return ret;
+}
+
+int btrfs_block_rsv_check(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ struct btrfs_block_rsv *block_rsv,
+ u64 min_reserved, int min_factor)
+{
+ u64 num_bytes = 0;
+ int commit_trans = 0;
+ int ret = -ENOSPC;
+
+ if (!block_rsv)
+ return 0;
+
+ spin_lock(&block_rsv->lock);
+ if (min_factor > 0)
+ num_bytes = div_factor(block_rsv->size, min_factor);
+ if (min_reserved > num_bytes)
+ num_bytes = min_reserved;
+
+ if (block_rsv->reserved >= num_bytes) {
+ ret = 0;
+ } else {
+ num_bytes -= block_rsv->reserved;
+ if (block_rsv->durable &&
+ block_rsv->freed[0] + block_rsv->freed[1] >= num_bytes)
+ commit_trans = 1;
+ }
+ spin_unlock(&block_rsv->lock);
+ if (!ret)
+ return 0;
+
+ if (block_rsv->refill_used) {
+ ret = reserve_metadata_bytes(block_rsv, num_bytes);
+ if (!ret) {
+ block_rsv_add_bytes(block_rsv, num_bytes, 0);
+ return 0;
+ }
+ }
+
+ if (commit_trans) {
+ if (trans)
+ return -EAGAIN;
+
+ trans = btrfs_join_transaction(root, 1);
+ BUG_ON(IS_ERR(trans));
+ ret = btrfs_commit_transaction(trans, root);
+ return 0;
+ }
+
+ WARN_ON(1);
+ printk(KERN_INFO"block_rsv size %llu reserved %llu freed %llu %llu\n",
+ block_rsv->size, block_rsv->reserved,
+ block_rsv->freed[0], block_rsv->freed[1]);
+
+ return -ENOSPC;
+}
+
+int btrfs_block_rsv_migrate(struct btrfs_block_rsv *src_rsv,
+ struct btrfs_block_rsv *dst_rsv,
+ u64 num_bytes)
+{
+ return block_rsv_migrate_bytes(src_rsv, dst_rsv, num_bytes);
+}
+
+void btrfs_block_rsv_release(struct btrfs_root *root,
+ struct btrfs_block_rsv *block_rsv,
+ u64 num_bytes)
+{
+ struct btrfs_block_rsv *global_rsv = &root->fs_info->global_block_rsv;
+ if (global_rsv->full || global_rsv == block_rsv ||
+ block_rsv->space_info != global_rsv->space_info)
+ global_rsv = NULL;
+ block_rsv_release_bytes(block_rsv, global_rsv, num_bytes);
+}
+
+static void init_global_block_rsv(struct btrfs_fs_info *fs_info)
+{
+ struct btrfs_space_info *space_info;
+
+ space_info = __find_space_info(fs_info, BTRFS_BLOCK_GROUP_SYSTEM);
+ fs_info->chunk_block_rsv.space_info = space_info;
+ fs_info->chunk_block_rsv.priority = 10;
+
+ space_info = __find_space_info(fs_info, BTRFS_BLOCK_GROUP_METADATA);
+ fs_info->trans_block_rsv.space_info = space_info;
+ fs_info->empty_block_rsv.space_info = space_info;
+ fs_info->empty_block_rsv.priority = 10;
+
+ fs_info->chunk_root->block_rsv = &fs_info->chunk_block_rsv;
+}
+
static int update_block_group(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
- u64 bytenr, u64 num_bytes, int alloc,
- int mark_free)
+ u64 bytenr, u64 num_bytes, int alloc)
{
struct btrfs_block_group_cache *cache;
struct btrfs_fs_info *info = root->fs_info;
@@ -3458,30 +3746,21 @@ static int update_block_group(struct btr
cache->space_info->bytes_reserved -= num_bytes;
cache->space_info->bytes_used += num_bytes;
cache->space_info->disk_used += num_bytes * factor;
- if (cache->ro)
- cache->space_info->bytes_readonly -= num_bytes;
spin_unlock(&cache->lock);
spin_unlock(&cache->space_info->lock);
} else {
old_val -= num_bytes;
btrfs_set_block_group_used(&cache->item, old_val);
+ cache->pinned += num_bytes;
+ cache->space_info->bytes_pinned += num_bytes;
cache->space_info->bytes_used -= num_bytes;
cache->space_info->disk_used -= num_bytes * factor;
- if (cache->ro)
- cache->space_info->bytes_readonly += num_bytes;
spin_unlock(&cache->lock);
spin_unlock(&cache->space_info->lock);
- if (mark_free) {
- int ret;
- ret = btrfs_discard_extent(root, bytenr,
- num_bytes);
- WARN_ON(ret);
-
- ret = btrfs_add_free_space(cache, bytenr,
- num_bytes);
- WARN_ON(ret);
- }
+ set_extent_dirty(info->pinned_extents,
+ bytenr, bytenr + num_bytes - 1,
+ GFP_NOFS | __GFP_NOFAIL);
}
btrfs_put_block_group(cache);
total -= num_bytes;
@@ -3505,18 +3784,10 @@ static u64 first_logical_byte(struct btr
return bytenr;
}
-/*
- * this function must be called within transaction
- */
-int btrfs_pin_extent(struct btrfs_root *root,
- u64 bytenr, u64 num_bytes, int reserved)
+static int pin_down_extent(struct btrfs_root *root,
+ struct btrfs_block_group_cache *cache,
+ u64 bytenr, u64 num_bytes, int reserved)
{
- struct btrfs_fs_info *fs_info = root->fs_info;
- struct btrfs_block_group_cache *cache;
-
- cache = btrfs_lookup_block_group(fs_info, bytenr);
- BUG_ON(!cache);
-
spin_lock(&cache->space_info->lock);
spin_lock(&cache->lock);
cache->pinned += num_bytes;
@@ -3528,28 +3799,68 @@ int btrfs_pin_extent(struct btrfs_root *
spin_unlock(&cache->lock);
spin_unlock(&cache->space_info->lock);
- btrfs_put_block_group(cache);
+ set_extent_dirty(root->fs_info->pinned_extents, bytenr,
+ bytenr + num_bytes - 1, GFP_NOFS | __GFP_NOFAIL);
+ return 0;
+}
- set_extent_dirty(fs_info->pinned_extents,
- bytenr, bytenr + num_bytes - 1, GFP_NOFS);
+/*
+ * this function must be called within transaction
+ */
+int btrfs_pin_extent(struct btrfs_root *root,
+ u64 bytenr, u64 num_bytes, int reserved)
+{
+ struct btrfs_block_group_cache *cache;
+
+ cache = btrfs_lookup_block_group(root->fs_info, bytenr);
+ BUG_ON(!cache);
+
+ pin_down_extent(root, cache, bytenr, num_bytes, reserved);
+
+ btrfs_put_block_group(cache);
return 0;
}
-static int update_reserved_extents(struct btrfs_block_group_cache *cache,
- u64 num_bytes, int reserve)
+/*
+ * update size of reserved extents. this function may return -EAGAIN
+ * if 'reserve' is true or 'sinfo' is false.
+ */
+static int update_reserved_bytes(struct btrfs_block_group_cache *cache,
+ u64 num_bytes, int reserve, int sinfo)
{
- spin_lock(&cache->space_info->lock);
- spin_lock(&cache->lock);
- if (reserve) {
- cache->reserved += num_bytes;
- cache->space_info->bytes_reserved += num_bytes;
+ int ret = 0;
+ if (sinfo) {
+ struct btrfs_space_info *space_info = cache->space_info;
+ spin_lock(&space_info->lock);
+ spin_lock(&cache->lock);
+ if (reserve) {
+ if (cache->ro) {
+ ret = -EAGAIN;
+ } else {
+ cache->reserved += num_bytes;
+ space_info->bytes_reserved += num_bytes;
+ }
+ } else {
+ if (cache->ro)
+ space_info->bytes_readonly += num_bytes;
+ cache->reserved -= num_bytes;
+ space_info->bytes_reserved -= num_bytes;
+ }
+ spin_unlock(&cache->lock);
+ spin_unlock(&space_info->lock);
} else {
- cache->reserved -= num_bytes;
- cache->space_info->bytes_reserved -= num_bytes;
+ spin_lock(&cache->lock);
+ if (cache->ro) {
+ ret = -EAGAIN;
+ } else {
+ if (reserve)
+ cache->reserved += num_bytes;
+ else
+ cache->reserved -= num_bytes;
+ }
+ spin_unlock(&cache->lock);
}
- spin_unlock(&cache->lock);
- spin_unlock(&cache->space_info->lock);
- return 0;
+ return ret;
}
int btrfs_prepare_extent_commit(struct btrfs_trans_handle *trans,
@@ -3606,14 +3917,21 @@ static int unpin_extent_range(struct btr
btrfs_add_free_space(cache, start, len);
}
+ start += len;
+
spin_lock(&cache->space_info->lock);
spin_lock(&cache->lock);
cache->pinned -= len;
cache->space_info->bytes_pinned -= len;
+ if (cache->ro) {
+ cache->space_info->bytes_readonly += len;
+ } else if (cache->reserved_pinned > 0) {
+ len = min(len, cache->reserved_pinned);
+ cache->reserved_pinned -= len;
+ cache->space_info->bytes_reserved += len;
+ }
spin_unlock(&cache->lock);
spin_unlock(&cache->space_info->lock);
-
- start += len;
}
if (cache)
@@ -3626,8 +3944,11 @@ int btrfs_finish_extent_commit(struct bt
{
struct btrfs_fs_info *fs_info = root->fs_info;
struct extent_io_tree *unpin;
+ struct btrfs_block_rsv *block_rsv;
+ struct btrfs_block_rsv *next_rsv;
u64 start;
u64 end;
+ int idx;
int ret;
if (fs_info->pinned_extents == &fs_info->freed_extents[0])
@@ -3648,59 +3969,30 @@ int btrfs_finish_extent_commit(struct bt
cond_resched();
}
- return ret;
-}
-
-static int pin_down_bytes(struct btrfs_trans_handle *trans,
- struct btrfs_root *root,
- struct btrfs_path *path,
- u64 bytenr, u64 num_bytes,
- int is_data, int reserved,
- struct extent_buffer **must_clean)
-{
- int err = 0;
- struct extent_buffer *buf;
-
- if (is_data)
- goto pinit;
-
- /*
- * discard is sloooow, and so triggering discards on
- * individual btree blocks isn't a good plan. Just
- * pin everything in discard mode.
- */
- if (btrfs_test_opt(root, DISCARD))
- goto pinit;
-
- buf = btrfs_find_tree_block(root, bytenr, num_bytes);
- if (!buf)
- goto pinit;
-
- /* we can reuse a block if it hasn't been written
- * and it is from this transaction. We can't
- * reuse anything from the tree log root because
- * it has tiny sub-transactions.
- */
- if (btrfs_buffer_uptodate(buf, 0) &&
- btrfs_try_tree_lock(buf)) {
- u64 header_owner = btrfs_header_owner(buf);
- u64 header_transid = btrfs_header_generation(buf);
- if (header_owner != BTRFS_TREE_LOG_OBJECTID &&
- header_transid == trans->transid &&
- !btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) {
- *must_clean = buf;
- return 1;
+ mutex_lock(&fs_info->durable_block_rsv_mutex);
+ list_for_each_entry_safe(block_rsv, next_rsv,
+ &fs_info->durable_block_rsv_list, list) {
+
+ idx = trans->transid & 0x1;
+ if (block_rsv->freed[idx] > 0) {
+ block_rsv_add_bytes(block_rsv,
+ block_rsv->freed[idx], 0);
+ block_rsv->freed[idx] = 0;
+ }
+ if (atomic_read(&block_rsv->usage) == 0) {
+ btrfs_block_rsv_release(root, block_rsv, (u64)-1);
+
+ if (block_rsv->freed[0] == 0 &&
+ block_rsv->freed[1] == 0) {
+ list_del_init(&block_rsv->list);
+ kfree(block_rsv);
+ }
+ } else {
+ btrfs_block_rsv_release(root, block_rsv, 0);
}
- btrfs_tree_unlock(buf);
}
- free_extent_buffer(buf);
-pinit:
- if (path)
- btrfs_set_path_blocking(path);
- /* unlocks the pinned mutex */
- btrfs_pin_extent(root, bytenr, num_bytes, reserved);
+ mutex_unlock(&fs_info->durable_block_rsv_mutex);
- BUG_ON(err < 0);
return 0;
}
@@ -3861,9 +4153,6 @@ static int __btrfs_free_extent(struct bt
BUG_ON(ret);
}
} else {
- int mark_free = 0;
- struct extent_buffer *must_clean = NULL;
-
if (found_extent) {
BUG_ON(is_data && refs_to_drop !=
extent_data_ref_count(root, path, iref));
@@ -3876,31 +4165,11 @@ static int __btrfs_free_extent(struct bt
}
}
- ret = pin_down_bytes(trans, root, path, bytenr,
- num_bytes, is_data, 0, &must_clean);
- if (ret > 0)
- mark_free = 1;
- BUG_ON(ret < 0);
- /*
- * it is going to be very rare for someone to be waiting
- * on the block we're freeing. del_items might need to
- * schedule, so rather than get fancy, just force it
- * to blocking here
- */
- if (must_clean)
- btrfs_set_lock_blocking(must_clean);
-
ret = btrfs_del_items(trans, extent_root, path, path->slots[0],
num_to_del);
BUG_ON(ret);
btrfs_release_path(extent_root, path);
- if (must_clean) {
- clean_tree_block(NULL, root, must_clean);
- btrfs_tree_unlock(must_clean);
- free_extent_buffer(must_clean);
- }
-
if (is_data) {
ret = btrfs_del_csums(trans, root, bytenr, num_bytes);
BUG_ON(ret);
@@ -3910,8 +4179,7 @@ static int __btrfs_free_extent(struct bt
(bytenr + num_bytes - 1) >> PAGE_CACHE_SHIFT);
}
- ret = update_block_group(trans, root, bytenr, num_bytes, 0,
- mark_free);
+ ret = update_block_group(trans, root, bytenr, num_bytes, 0);
BUG_ON(ret);
}
btrfs_free_path(path);
@@ -3919,7 +4187,7 @@ static int __btrfs_free_extent(struct bt
}
/*
- * when we free an extent, it is possible (and likely) that we free the last
+ * when we free an block, it is possible (and likely) that we free the last
* delayed ref for that extent as well. This searches the delayed ref tree for
* a given extent, and if there are no other delayed refs to be processed, it
* removes it from the tree.
@@ -3931,7 +4199,7 @@ static noinline int check_ref_cleanup(st
struct btrfs_delayed_ref_root *delayed_refs;
struct btrfs_delayed_ref_node *ref;
struct rb_node *node;
- int ret;
+ int ret = 0;
delayed_refs = &trans->transaction->delayed_refs;
spin_lock(&delayed_refs->lock);
@@ -3983,17 +4251,99 @@ static noinline int check_ref_cleanup(st
list_del_init(&head->cluster);
spin_unlock(&delayed_refs->lock);
- ret = run_one_delayed_ref(trans, root->fs_info->tree_root,
- &head->node, head->extent_op,
- head->must_insert_reserved);
- BUG_ON(ret);
+ BUG_ON(head->extent_op);
+ if (head->must_insert_reserved)
+ ret = 1;
+
+ mutex_unlock(&head->mutex);
btrfs_put_delayed_ref(&head->node);
- return 0;
+ return ret;
out:
spin_unlock(&delayed_refs->lock);
return 0;
}
+void btrfs_free_tree_block(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ struct extent_buffer *buf,
+ u64 parent, int last_ref)
+{
+ struct btrfs_block_rsv *block_rsv;
+ struct btrfs_block_group_cache *cache = NULL;
+ int ret;
+
+ if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
+ ret = btrfs_add_delayed_tree_ref(trans, buf->start, buf->len,
+ parent, root->root_key.objectid,
+ btrfs_header_level(buf),
+ BTRFS_DROP_DELAYED_REF, NULL);
+ BUG_ON(ret);
+ }
+
+ if (!last_ref)
+ return;
+
+ block_rsv = get_block_rsv(trans, root);
+ cache = btrfs_lookup_block_group(root->fs_info, buf->start);
+ BUG_ON(block_rsv->space_info != cache->space_info);
+
+ if (btrfs_header_generation(buf) == trans->transid) {
+ if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
+ ret = check_ref_cleanup(trans, root, buf->start);
+ if (!ret)
+ goto pin;
+ }
+
+ if (btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) {
+ pin_down_extent(root, cache, buf->start, buf->len, 1);
+ goto pin;
+ }
+
+ WARN_ON(test_bit(EXTENT_BUFFER_DIRTY, &buf->bflags));
+
+ btrfs_add_free_space(cache, buf->start, buf->len);
+ ret = update_reserved_bytes(cache, buf->len, 0, 0);
+ if (ret == -EAGAIN) {
+ /* block group became read-only */
+ update_reserved_bytes(cache, buf->len, 0, 1);
+ goto out;
+ }
+
+ ret = 1;
+ spin_lock(&block_rsv->lock);
+ if (block_rsv->reserved < block_rsv->size) {
+ block_rsv->reserved += buf->len;
+ ret = 0;
+ }
+ spin_unlock(&block_rsv->lock);
+
+ if (ret) {
+ spin_lock(&cache->space_info->lock);
+ cache->space_info->bytes_reserved -= buf->len;
+ spin_unlock(&cache->space_info->lock);
+ }
+ goto out;
+ }
+pin:
+ if (block_rsv->durable && !cache->ro) {
+ ret = 0;
+ spin_lock(&cache->lock);
+ if (!cache->ro) {
+ cache->reserved_pinned += buf->len;
+ ret = 1;
+ }
+ spin_unlock(&cache->lock);
+
+ if (ret) {
+ spin_lock(&block_rsv->lock);
+ block_rsv->freed[trans->transid & 0x1] += buf->len;
+ spin_unlock(&block_rsv->lock);
+ }
+ }
+out:
+ btrfs_put_block_group(cache);
+}
+
int btrfs_free_extent(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
u64 bytenr, u64 num_bytes, u64 parent,
@@ -4015,8 +4365,6 @@ int btrfs_free_extent(struct btrfs_trans
parent, root_objectid, (int)owner,
BTRFS_DROP_DELAYED_REF, NULL);
BUG_ON(ret);
- ret = check_ref_cleanup(trans, root, bytenr);
- BUG_ON(ret);
} else {
ret = btrfs_add_delayed_data_ref(trans, bytenr, num_bytes,
parent, root_objectid, owner,
@@ -4026,21 +4374,6 @@ int btrfs_free_extent(struct btrfs_trans
return ret;
}
-int btrfs_free_tree_block(struct btrfs_trans_handle *trans,
- struct btrfs_root *root,
- u64 bytenr, u32 blocksize,
- u64 parent, u64 root_objectid, int level)
-{
- u64 used;
- spin_lock(&root->node_lock);
- used = btrfs_root_used(&root->root_item) - blocksize;
- btrfs_set_root_used(&root->root_item, used);
- spin_unlock(&root->node_lock);
-
- return btrfs_free_extent(trans, root, bytenr, blocksize,
- parent, root_objectid, level, 0);
-}
-
static u64 stripe_align(struct btrfs_root *root, u64 val)
{
u64 mask = ((u64)root->stripesize - 1);
@@ -4130,7 +4463,6 @@ static noinline int find_free_extent(str
u64 num_bytes, u64 empty_size,
u64 search_start, u64 search_end,
u64 hint_byte, struct btrfs_key *ins,
- u64 exclude_start, u64 exclude_nr,
int data)
{
int ret = 0;
@@ -4142,8 +4474,8 @@ static noinline int find_free_extent(str
int done_chunk_alloc = 0;
struct btrfs_space_info *space_info;
int last_ptr_loop = 0;
- int index = 0;
int loop = 0;
+ int index = 0;
bool found_uncached_bg = false;
bool failed_cluster_refill = false;
bool failed_alloc = false;
@@ -4414,23 +4746,22 @@ checks:
goto loop;
}
- if (exclude_nr > 0 &&
- (search_start + num_bytes > exclude_start &&
- search_start < exclude_start + exclude_nr)) {
- search_start = exclude_start + exclude_nr;
+ ins->objectid = search_start;
+ ins->offset = num_bytes;
+ if (offset < search_start)
+ btrfs_add_free_space(block_group, offset,
+ search_start - offset);
+ BUG_ON(offset > search_start);
+
+ ret = update_reserved_bytes(block_group, num_bytes, 1,
+ (data & BTRFS_BLOCK_GROUP_DATA));
+ if (ret == -EAGAIN) {
btrfs_add_free_space(block_group, offset, num_bytes);
- /*
- * if search_start is still in this block group
- * then we just re-search this block group
- */
- if (search_start >= block_group->key.objectid &&
- search_start < (block_group->key.objectid +
- block_group->key.offset))
- goto have_block_group;
goto loop;
}
+ /* we are all good, lets return */
ins->objectid = search_start;
ins->offset = num_bytes;
@@ -4438,10 +4769,6 @@ checks:
btrfs_add_free_space(block_group, offset,
search_start - offset);
BUG_ON(offset > search_start);
-
- update_reserved_extents(block_group, num_bytes, 1);
-
- /* we are all good, lets return */
break;
loop:
failed_cluster_refill = false;
@@ -4615,9 +4942,8 @@ again:
WARN_ON(num_bytes < root->sectorsize);
ret = find_free_extent(trans, root, num_bytes, empty_size,
- search_start, search_end, hint_byte, ins,
- trans->alloc_exclude_start,
- trans->alloc_exclude_nr, data);
+ search_start, search_end, hint_byte,
+ ins, data);
if (ret == -ENOSPC && num_bytes > min_alloc_size) {
num_bytes = num_bytes >> 1;
@@ -4655,7 +4981,7 @@ int btrfs_free_reserved_extent(struct bt
ret = btrfs_discard_extent(root, start, len);
btrfs_add_free_space(cache, start, len);
- update_reserved_extents(cache, len, 0);
+ update_reserved_bytes(cache, len, 0, 1);
btrfs_put_block_group(cache);
return ret;
@@ -4718,8 +5044,7 @@ static int alloc_reserved_file_extent(st
btrfs_mark_buffer_dirty(path->nodes[0]);
btrfs_free_path(path);
- ret = update_block_group(trans, root, ins->objectid, ins->offset,
- 1, 0);
+ ret = update_block_group(trans, root, ins->objectid, ins->offset, 1);
if (ret) {
printk(KERN_ERR "btrfs update block group failed for %llu "
"%llu\n", (unsigned long long)ins->objectid,
@@ -4779,8 +5104,7 @@ static int alloc_reserved_tree_block(str
btrfs_mark_buffer_dirty(leaf);
btrfs_free_path(path);
- ret = update_block_group(trans, root, ins->objectid, ins->offset,
- 1, 0);
+ ret = update_block_group(trans, root, ins->objectid, ins->offset, 1);
if (ret) {
printk(KERN_ERR "btrfs update block group failed for %llu "
"%llu\n", (unsigned long long)ins->objectid,
@@ -4856,73 +5180,14 @@ int btrfs_alloc_logged_file_extent(struc
put_caching_control(caching_ctl);
}
- update_reserved_extents(block_group, ins->offset, 1);
+ ret = update_reserved_bytes(block_group, ins->offset, 1, 1);
+ BUG_ON(ret);
btrfs_put_block_group(block_group);
ret = alloc_reserved_file_extent(trans, root, 0, root_objectid,
0, owner, offset, ins, 1);
return ret;
}
-/*
- * finds a free extent and does all the dirty work required for allocation
- * returns the key for the extent through ins, and a tree buffer for
- * the first block of the extent through buf.
- *
- * returns 0 if everything worked, non-zero otherwise.
- */
-static int alloc_tree_block(struct btrfs_trans_handle *trans,
- struct btrfs_root *root,
- u64 num_bytes, u64 parent, u64 root_objectid,
- struct btrfs_disk_key *key, int level,
- u64 empty_size, u64 hint_byte, u64 search_end,
- struct btrfs_key *ins)
-{
- int ret;
- u64 flags = 0;
-
- ret = btrfs_reserve_extent(trans, root, num_bytes, num_bytes,
- empty_size, hint_byte, search_end,
- ins, 0);
- if (ret)
- return ret;
-
- if (root_objectid == BTRFS_TREE_RELOC_OBJECTID) {
- if (parent == 0)
- parent = ins->objectid;
- flags |= BTRFS_BLOCK_FLAG_FULL_BACKREF;
- } else
- BUG_ON(parent > 0);
-
- if (root_objectid != BTRFS_TREE_LOG_OBJECTID) {
- struct btrfs_delayed_extent_op *extent_op;
- extent_op = kmalloc(sizeof(*extent_op), GFP_NOFS);
- BUG_ON(!extent_op);
- if (key)
- memcpy(&extent_op->key, key, sizeof(extent_op->key));
- else
- memset(&extent_op->key, 0, sizeof(extent_op->key));
- extent_op->flags_to_set = flags;
- extent_op->update_key = 1;
- extent_op->update_flags = 1;
- extent_op->is_data = 0;
-
- ret = btrfs_add_delayed_tree_ref(trans, ins->objectid,
- ins->offset, parent, root_objectid,
- level, BTRFS_ADD_DELAYED_EXTENT,
- extent_op);
- BUG_ON(ret);
- }
-
- if (root_objectid == root->root_key.objectid) {
- u64 used;
- spin_lock(&root->node_lock);
- used = btrfs_root_used(&root->root_item) + num_bytes;
- btrfs_set_root_used(&root->root_item, used);
- spin_unlock(&root->node_lock);
- }
- return ret;
-}
-
struct extent_buffer *btrfs_init_new_buffer(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
u64 bytenr, u32 blocksize,
@@ -4961,8 +5226,45 @@ struct extent_buffer *btrfs_init_new_buf
return buf;
}
+static struct btrfs_block_rsv *
+use_block_rsv(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root, u32 blocksize)
+{
+ struct btrfs_block_rsv *block_rsv;
+ int ret;
+
+ block_rsv = get_block_rsv(trans, root);
+
+ if (block_rsv->size == 0) {
+ ret = reserve_metadata_bytes(block_rsv, blocksize);
+ if (ret)
+ return ERR_PTR(ret);
+ return block_rsv;
+ }
+
+ ret = block_rsv_use_bytes(block_rsv, blocksize);
+ if (!ret)
+ return block_rsv;
+
+ WARN_ON(1);
+ printk(KERN_INFO"block_rsv size %llu reserved %llu freed %llu %llu\n",
+ block_rsv->size, block_rsv->reserved,
+ block_rsv->freed[0], block_rsv->freed[1]);
+
+ return ERR_PTR(-ENOSPC);
+}
+
+static void unuse_block_rsv(struct btrfs_block_rsv *block_rsv, u32 blocksize)
+{
+ block_rsv_add_bytes(block_rsv, blocksize, 0);
+ block_rsv_release_bytes(block_rsv, NULL, 0);
+}
+
/*
- * helper function to allocate a block for a given tree
+ * finds a free extent and does all the dirty work required for allocation
+ * returns the key for the extent through ins, and a tree buffer for
+ * the first block of the extent through buf.
+ *
* returns the tree buffer or NULL.
*/
struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
@@ -4972,18 +5274,53 @@ struct extent_buffer *btrfs_alloc_free_b
u64 hint, u64 empty_size)
{
struct btrfs_key ins;
- int ret;
+ struct btrfs_block_rsv *block_rsv;
struct extent_buffer *buf;
+ u64 flags = 0;
+ int ret;
+
+
+ block_rsv = use_block_rsv(trans, root, blocksize);
+ if (IS_ERR(block_rsv))
+ return ERR_CAST(block_rsv);
- ret = alloc_tree_block(trans, root, blocksize, parent, root_objectid,
- key, level, empty_size, hint, (u64)-1, &ins);
+ ret = btrfs_reserve_extent(trans, root, blocksize, blocksize,
+ empty_size, hint, (u64)-1, &ins, 0);
if (ret) {
- BUG_ON(ret > 0);
+ unuse_block_rsv(block_rsv, blocksize);
return ERR_PTR(ret);
}
buf = btrfs_init_new_buffer(trans, root, ins.objectid,
blocksize, level);
+ BUG_ON(IS_ERR(buf));
+
+ if (root_objectid == BTRFS_TREE_RELOC_OBJECTID) {
+ if (parent == 0)
+ parent = ins.objectid;
+ flags |= BTRFS_BLOCK_FLAG_FULL_BACKREF;
+ } else
+ BUG_ON(parent > 0);
+
+ if (root_objectid != BTRFS_TREE_LOG_OBJECTID) {
+ struct btrfs_delayed_extent_op *extent_op;
+ extent_op = kmalloc(sizeof(*extent_op), GFP_NOFS);
+ BUG_ON(!extent_op);
+ if (key)
+ memcpy(&extent_op->key, key, sizeof(extent_op->key));
+ else
+ memset(&extent_op->key, 0, sizeof(extent_op->key));
+ extent_op->flags_to_set = flags;
+ extent_op->update_key = 1;
+ extent_op->update_flags = 1;
+ extent_op->is_data = 0;
+
+ ret = btrfs_add_delayed_tree_ref(trans, ins.objectid,
+ ins.offset, parent, root_objectid,
+ level, BTRFS_ADD_DELAYED_EXTENT,
+ extent_op);
+ BUG_ON(ret);
+ }
return buf;
}
@@ -5308,7 +5645,7 @@ static noinline int walk_up_proc(struct
struct btrfs_path *path,
struct walk_control *wc)
{
- int ret = 0;
+ int ret;
int level = wc->level;
struct extent_buffer *eb = path->nodes[level];
u64 parent = 0;
@@ -5386,13 +5723,11 @@ static noinline int walk_up_proc(struct
btrfs_header_owner(path->nodes[level + 1]));
}
- ret = btrfs_free_extent(trans, root, eb->start, eb->len, parent,
- root->root_key.objectid, level, 0);
- BUG_ON(ret);
+ btrfs_free_tree_block(trans, root, eb, parent, wc->refs[level] == 1);
out:
wc->refs[level] = 0;
wc->flags[level] = 0;
- return ret;
+ return 0;
}
static noinline int walk_down_tree(struct btrfs_trans_handle *trans,
@@ -7215,48 +7550,80 @@ static u64 update_block_group_flags(stru
return flags;
}
-static int __alloc_chunk_for_shrink(struct btrfs_root *root,
- struct btrfs_block_group_cache *shrink_block_group,
- int force)
+static int set_block_group_ro(struct btrfs_block_group_cache *cache)
{
- struct btrfs_trans_handle *trans;
- u64 new_alloc_flags;
- u64 calc;
+ struct btrfs_space_info *sinfo = cache->space_info;
+ u64 num_bytes;
+ int ret = -ENOSPC;
- spin_lock(&shrink_block_group->lock);
- if (btrfs_block_group_used(&shrink_block_group->item) +
- shrink_block_group->reserved > 0) {
- spin_unlock(&shrink_block_group->lock);
+ if (cache->ro)
+ return 0;
- trans = btrfs_start_transaction(root, 1);
- spin_lock(&shrink_block_group->lock);
+ spin_lock(&sinfo->lock);
+ spin_lock(&cache->lock);
+ num_bytes = cache->key.offset - cache->reserved - cache->pinned -
+ cache->bytes_super - btrfs_block_group_used(&cache->item);
- new_alloc_flags = update_block_group_flags(root,
- shrink_block_group->flags);
- if (new_alloc_flags != shrink_block_group->flags) {
- calc =
- btrfs_block_group_used(&shrink_block_group->item);
- } else {
- calc = shrink_block_group->key.offset;
- }
- spin_unlock(&shrink_block_group->lock);
+ if (sinfo->bytes_used + sinfo->bytes_reserved + sinfo->bytes_pinned +
+ sinfo->bytes_may_use + sinfo->bytes_readonly +
+ cache->reserved_pinned + num_bytes < sinfo->total_bytes) {
+ sinfo->bytes_readonly += num_bytes;
+ sinfo->bytes_reserved += cache->reserved_pinned;
+ cache->reserved_pinned = 0;
+ cache->ro = 1;
+ ret = 0;
+ }
+ spin_unlock(&cache->lock);
+ spin_unlock(&sinfo->lock);
+ return ret;
+}
- do_chunk_alloc(trans, root->fs_info->extent_root,
- calc + 2 * 1024 * 1024, new_alloc_flags, force);
+int btrfs_set_block_group_ro(struct btrfs_root *root,
+ struct btrfs_block_group_cache *cache)
- btrfs_end_transaction(trans, root);
- } else
- spin_unlock(&shrink_block_group->lock);
- return 0;
-}
+{
+ struct btrfs_trans_handle *trans;
+ u64 alloc_flags;
+ int ret;
+ BUG_ON(cache->ro);
-int btrfs_prepare_block_group_relocation(struct btrfs_root *root,
- struct btrfs_block_group_cache *group)
+ trans = btrfs_join_transaction(root, 1);
+ BUG_ON(IS_ERR(trans));
+ alloc_flags = update_block_group_flags(root, cache->flags);
+ if (alloc_flags != cache->flags)
+ do_chunk_alloc(trans, root, 2 * 1024 * 1024, alloc_flags, 1);
+
+ ret = set_block_group_ro(cache);
+ if (!ret)
+ goto out;
+ alloc_flags = get_alloc_profile(root, cache->space_info->flags);
+ ret = do_chunk_alloc(trans, root, 2 * 1024 * 1024, alloc_flags, 1);
+ if (ret < 0)
+ goto out;
+ ret = set_block_group_ro(cache);
+out:
+ btrfs_end_transaction(trans, root);
+ return ret;
+}
+
+int btrfs_set_block_group_rw(struct btrfs_root *root,
+ struct btrfs_block_group_cache *cache)
{
- __alloc_chunk_for_shrink(root, group, 1);
- set_block_group_readonly(group);
+ struct btrfs_space_info *sinfo = cache->space_info;
+ u64 num_bytes;
+
+ BUG_ON(!cache->ro);
+
+ spin_lock(&sinfo->lock);
+ spin_lock(&cache->lock);
+ num_bytes = cache->key.offset - cache->reserved - cache->pinned -
+ cache->bytes_super - btrfs_block_group_used(&cache->item);
+ sinfo->bytes_readonly -= num_bytes;
+ cache->ro = 0;
+ spin_unlock(&cache->lock);
+ spin_unlock(&sinfo->lock);
return 0;
}
@@ -7427,7 +7794,11 @@ int btrfs_free_block_groups(struct btrfs
space_info = list_entry(info->space_info.next,
struct btrfs_space_info,
list);
-
+ if (space_info->bytes_pinned > 0 ||
+ space_info->bytes_reserved > 0) {
+ WARN_ON(1);
+ dump_space_info(space_info, 0, 0);
+ }
list_del(&space_info->list);
kfree(space_info);
}
@@ -7475,7 +7846,7 @@ int btrfs_read_block_groups(struct btrfs
cache = kzalloc(sizeof(*cache), GFP_NOFS);
if (!cache) {
ret = -ENOMEM;
- break;
+ goto error;
}
atomic_set(&cache->count, 1);
@@ -7532,7 +7903,7 @@ int btrfs_read_block_groups(struct btrfs
BUG_ON(ret);
cache->space_info = space_info;
spin_lock(&cache->space_info->lock);
- cache->space_info->bytes_super += cache->bytes_super;
+ cache->space_info->bytes_readonly += cache->bytes_super;
spin_unlock(&cache->space_info->lock);
__link_block_group(space_info, cache);
@@ -7542,7 +7913,7 @@ int btrfs_read_block_groups(struct btrfs
set_avail_alloc_bits(root->fs_info, cache->flags);
if (btrfs_chunk_readonly(root, cache->key.objectid))
- set_block_group_readonly(cache);
+ set_block_group_ro(cache);
}
list_for_each_entry_rcu(space_info, &root->fs_info->space_info, list) {
@@ -7556,10 +7927,12 @@ int btrfs_read_block_groups(struct btrfs
* mirrored block groups.
*/
list_for_each_entry(cache, &space_info->block_groups[3], list)
- set_block_group_readonly(cache);
+ set_block_group_ro(cache);
list_for_each_entry(cache, &space_info->block_groups[4], list)
- set_block_group_readonly(cache);
+ set_block_group_ro(cache);
}
+
+ init_global_block_rsv(info);
ret = 0;
error:
btrfs_free_path(path);
@@ -7620,7 +7993,7 @@ int btrfs_make_block_group(struct btrfs_
BUG_ON(ret);
spin_lock(&cache->space_info->lock);
- cache->space_info->bytes_super += cache->bytes_super;
+ cache->space_info->bytes_readonly += cache->bytes_super;
spin_unlock(&cache->space_info->lock);
__link_block_group(cache->space_info, cache);
diff -urp 2/fs/btrfs/relocation.c 3/fs/btrfs/relocation.c
--- 2/fs/btrfs/relocation.c 2010-04-26 17:24:55.238079821 +0800
+++ 3/fs/btrfs/relocation.c 2010-04-26 17:24:55.253080191 +0800
@@ -3513,6 +3513,7 @@ int btrfs_relocate_block_group(struct bt
struct btrfs_fs_info *fs_info = extent_root->fs_info;
struct reloc_control *rc;
int ret;
+ int rw = 0;
int err = 0;
rc = kzalloc(sizeof(*rc), GFP_NOFS);
@@ -3523,15 +3524,22 @@ int btrfs_relocate_block_group(struct bt
extent_io_tree_init(&rc->processed_blocks, NULL, GFP_NOFS);
INIT_LIST_HEAD(&rc->reloc_roots);
+ rc->extent_root = extent_root;
rc->block_group = btrfs_lookup_block_group(fs_info, group_start);
BUG_ON(!rc->block_group);
+ if (!rc->block_group->ro) {
+ ret = btrfs_set_block_group_ro(extent_root, rc->block_group);
+ if (ret) {
+ err = ret;
+ goto out;
+ }
+ rw = 1;
+ }
+
btrfs_init_workers(&rc->workers, "relocate",
fs_info->thread_pool_size, NULL);
- rc->extent_root = extent_root;
- btrfs_prepare_block_group_relocation(extent_root, rc->block_group);
-
rc->data_inode = create_reloc_inode(fs_info, rc->block_group);
if (IS_ERR(rc->data_inode)) {
err = PTR_ERR(rc->data_inode);
@@ -3596,6 +3604,8 @@ int btrfs_relocate_block_group(struct bt
WARN_ON(rc->block_group->reserved > 0);
WARN_ON(btrfs_block_group_used(&rc->block_group->item) > 0);
out:
+ if (err && rw)
+ btrfs_set_block_group_rw(extent_root, rc->block_group);
iput(rc->data_inode);
btrfs_stop_workers(&rc->workers);
btrfs_put_block_group(rc->block_group);
diff -urp 2/fs/btrfs/transaction.c 3/fs/btrfs/transaction.c
--- 2/fs/btrfs/transaction.c 2010-04-26 17:24:55.222858330 +0800
+++ 3/fs/btrfs/transaction.c 2010-04-26 17:24:55.253080191 +0800
@@ -184,9 +184,8 @@ static struct btrfs_trans_handle *start_
h->blocks_reserved = num_blocks;
h->blocks_used = 0;
h->block_group = 0;
- h->alloc_exclude_nr = 0;
- h->alloc_exclude_start = 0;
h->delayed_ref_updates = 0;
+ h->block_rsv = NULL;
if (!current->journal_info && type != TRANS_USERSPACE)
current->journal_info = h;
diff -urp 2/fs/btrfs/transaction.h 3/fs/btrfs/transaction.h
--- 2/fs/btrfs/transaction.h 2010-04-26 17:24:55.226081480 +0800
+++ 3/fs/btrfs/transaction.h 2010-04-26 17:24:55.254080681 +0800
@@ -45,13 +45,13 @@ struct btrfs_transaction {
struct btrfs_trans_handle {
u64 transid;
+ u64 block_group;
+ u64 bytes_reserved;
unsigned long blocks_reserved;
unsigned long blocks_used;
- struct btrfs_transaction *transaction;
- u64 block_group;
- u64 alloc_exclude_start;
- u64 alloc_exclude_nr;
unsigned long delayed_ref_updates;
+ struct btrfs_transaction *transaction;
+ struct btrfs_block_rsv *block_rsv;
};
struct btrfs_pending_snapshot {
Signed-off-by: Yan Zheng <zheng.yan@oracle.com>
---
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2010-04-26 10:44 UTC | newest]
Thread overview: (only message) (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2010-04-26 10:44 [PATCH V2 05/12] Btrfs: Introduce contexts for metadata reservation Yan, Zheng
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.