Btrfs: Replace the big fs_mutex with a collection of other locks
Extent alloctions are still protected by a large alloc_mutex.
Objectid allocations are covered by a objectid mutex
Other btree operations are protected by a lock on individual btree nodes
Signed-off-by: Chris Mason <chris.mason@oracle.com>
diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 75625c6..dff4da0 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -63,6 +63,9 @@
void btrfs_release_path(struct btrfs_root *root, struct btrfs_path *p)
{
int i;
+ int skip = p->skip_locking;
+ int keep = p->keep_locks;
+
for (i = 0; i < BTRFS_MAX_LEVEL; i++) {
if (!p->nodes[i])
continue;
@@ -73,6 +76,8 @@
free_extent_buffer(p->nodes[i]);
}
memset(p, 0, sizeof(*p));
+ p->skip_locking = skip;
+ p->keep_locks = keep;
}
struct extent_buffer *btrfs_root_node(struct btrfs_root *root)
@@ -1202,13 +1207,19 @@
u32 nritems;
t = path->nodes[i];
nritems = btrfs_header_nritems(t);
- if (path->slots[i] >= nritems - 1) {
+ if (nritems < 2 || path->slots[i] >= nritems - 2) {
+if (path->keep_locks) {
+//printk("path %p skip level now %d\n", path, skip_level);
+}
skip_level = i + 1;
continue;
}
}
t = path->nodes[i];
if (i >= lowest_unlock && i > skip_level && path->locks[i]) {
+if (path->keep_locks) {
+//printk("path %p unlocking level %d slot %d nritems %d skip_level %d\n", path, i, path->slots[i], btrfs_header_nritems(t), skip_level);
+}
btrfs_tree_unlock(t);
path->locks[i] = 0;
}
@@ -1243,7 +1254,6 @@
lowest_level = p->lowest_level;
WARN_ON(lowest_level && ins_len);
WARN_ON(p->nodes[0] != NULL);
- // WARN_ON(!mutex_is_locked(&root->fs_info->fs_mutex));
WARN_ON(root == root->fs_info->extent_root &&
!mutex_is_locked(&root->fs_info->alloc_mutex));
WARN_ON(root == root->fs_info->chunk_root &&
@@ -1321,7 +1331,7 @@
b = read_node_slot(root, b, slot);
if (!p->skip_locking)
btrfs_tree_lock(b);
- unlock_up(p, level, lowest_unlock);
+ unlock_up(p, level + 1, lowest_unlock);
} else {
p->slots[level] = slot;
if (ins_len > 0 && btrfs_leaf_free_space(root, b) <
@@ -1804,6 +1814,8 @@
if (slot >= btrfs_header_nritems(upper) - 1)
return 1;
+ WARN_ON(!btrfs_tree_locked(path->nodes[1]));
+
right = read_node_slot(root, upper, slot + 1);
btrfs_tree_lock(right);
free_space = btrfs_leaf_free_space(root, right);
@@ -1981,6 +1993,8 @@
return 1;
}
+ WARN_ON(!btrfs_tree_locked(path->nodes[1]));
+
left = read_node_slot(root, path->nodes[1], slot - 1);
btrfs_tree_lock(left);
free_space = btrfs_leaf_free_space(root, left);
@@ -2957,15 +2971,16 @@
btrfs_item_key_to_cpu(path->nodes[0], &key, nritems - 1);
- path->keep_locks = 1;
btrfs_release_path(root, path);
+ path->keep_locks = 1;
ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
path->keep_locks = 0;
if (ret < 0)
return ret;
- if (path->slots[0] < nritems - 1) {
+ nritems = btrfs_header_nritems(path->nodes[0]);
+ if (nritems > 0 && path->slots[0] < nritems - 1) {
goto done;
}
@@ -2992,8 +3007,17 @@
reada_for_search(root, path, level, slot, 0);
next = read_node_slot(root, c, slot);
- if (!path->skip_locking)
+ if (!path->skip_locking) {
+ if (!btrfs_tree_locked(c)) {
+ int i;
+ WARN_ON(1);
+printk("path %p no lock on level %d\n", path, level);
+for (i = 0; i < BTRFS_MAX_LEVEL; i++) {
+printk("path %p level %d slot %d nritems %d\n", path, i, path->slots[i], btrfs_header_nritems(path->nodes[i]));
+}
+ }
btrfs_tree_lock(next);
+ }
break;
}
path->slots[level] = slot;
@@ -3011,8 +3035,10 @@
if (level == 1 && path->locks[1] && path->reada)
reada_for_search(root, path, level, slot, 0);
next = read_node_slot(root, next, 0);
- if (!path->skip_locking)
+ if (!path->skip_locking) {
+ WARN_ON(!btrfs_tree_locked(path->nodes[level]));
btrfs_tree_lock(next);
+ }
}
done:
unlock_up(path, 0, 1);
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 50891b3..692b8ea 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -519,9 +519,9 @@
struct backing_dev_info bdi;
spinlock_t hash_lock;
struct mutex trans_mutex;
- struct mutex fs_mutex;
struct mutex alloc_mutex;
struct mutex chunk_mutex;
+ struct mutex drop_mutex;
struct list_head trans_list;
struct list_head hashers;
struct list_head dead_roots;
@@ -554,7 +554,7 @@
struct completion kobj_unregister;
int do_barriers;
int closing;
- unsigned long throttles;
+ atomic_t throttles;
u64 total_pinned;
struct list_head dirty_cowonly_roots;
@@ -594,6 +594,7 @@
struct inode *inode;
struct kobject root_kobj;
struct completion kobj_unregister;
+ struct mutex objectid_mutex;
u64 objectid;
u64 last_trans;
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index fe40bdd..f638803 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -724,6 +724,7 @@
INIT_LIST_HEAD(&root->dirty_list);
spin_lock_init(&root->node_lock);
+ mutex_init(&root->objectid_mutex);
memset(&root->root_key, 0, sizeof(root->root_key));
memset(&root->root_item, 0, sizeof(root->root_item));
memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
@@ -1146,6 +1147,7 @@
INIT_LIST_HEAD(&fs_info->space_info);
btrfs_mapping_init(&fs_info->mapping_tree);
atomic_set(&fs_info->nr_async_submits, 0);
+ atomic_set(&fs_info->throttles, 0);
fs_info->sb = sb;
fs_info->max_extent = (u64)-1;
fs_info->max_inline = 8192 * 1024;
@@ -1199,7 +1201,7 @@
mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
mutex_init(&fs_info->trans_mutex);
- mutex_init(&fs_info->fs_mutex);
+ mutex_init(&fs_info->drop_mutex);
mutex_init(&fs_info->alloc_mutex);
mutex_init(&fs_info->chunk_mutex);
@@ -1278,8 +1280,6 @@
goto fail_sb_buffer;
}
- mutex_lock(&fs_info->fs_mutex);
-
mutex_lock(&fs_info->chunk_mutex);
ret = btrfs_read_sys_array(tree_root);
mutex_unlock(&fs_info->chunk_mutex);
@@ -1342,7 +1342,6 @@
fs_info->metadata_alloc_profile = (u64)-1;
fs_info->system_alloc_profile = fs_info->metadata_alloc_profile;
- mutex_unlock(&fs_info->fs_mutex);
return tree_root;
fail_extent_root:
@@ -1350,7 +1349,6 @@
fail_tree_root:
free_extent_buffer(tree_root->node);
fail_sys_array:
- mutex_unlock(&fs_info->fs_mutex);
fail_sb_buffer:
extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree);
btrfs_stop_workers(&fs_info->workers);
@@ -1562,8 +1560,9 @@
struct btrfs_fs_info *fs_info = root->fs_info;
fs_info->closing = 1;
+ smp_mb();
+
btrfs_transaction_flush_work(root);
- mutex_lock(&fs_info->fs_mutex);
btrfs_defrag_dirty_roots(root->fs_info);
trans = btrfs_start_transaction(root, 1);
ret = btrfs_commit_transaction(trans, root);
@@ -1574,7 +1573,6 @@
BUG_ON(ret);
write_ctree_super(NULL, root);
- mutex_unlock(&fs_info->fs_mutex);
btrfs_transaction_flush_work(root);
@@ -1679,7 +1677,8 @@
struct backing_dev_info *bdi;
bdi = &root->fs_info->bdi;
- if (root->fs_info->throttles && bdi_write_congested(bdi)) {
+ if (atomic_read(&root->fs_info->throttles) &&
+ bdi_write_congested(bdi)) {
#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,18)
congestion_wait(WRITE, HZ/20);
#else
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 7e40c51..890b9e9 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -1577,9 +1577,11 @@
}
/* block accounting for super block */
+ spin_lock_irq(&info->delalloc_lock);
super_used = btrfs_super_bytes_used(&info->super_copy);
btrfs_set_super_bytes_used(&info->super_copy,
super_used - num_bytes);
+ spin_unlock_irq(&info->delalloc_lock);
/* block accounting for root item */
root_used = btrfs_root_used(&root->root_item);
@@ -1968,8 +1970,10 @@
}
/* block accounting for super block */
+ spin_lock_irq(&info->delalloc_lock);
super_used = btrfs_super_bytes_used(&info->super_copy);
btrfs_set_super_bytes_used(&info->super_copy, super_used + num_bytes);
+ spin_unlock_irq(&info->delalloc_lock);
/* block accounting for root item */
root_used = btrfs_root_used(&root->root_item);
@@ -2172,12 +2176,12 @@
continue;
}
}
- mutex_unlock(&root->fs_info->fs_mutex);
+ mutex_unlock(&root->fs_info->alloc_mutex);
ret = readahead_tree_block(root, bytenr, blocksize,
btrfs_node_ptr_generation(node, i));
last = bytenr + blocksize;
cond_resched();
- mutex_lock(&root->fs_info->fs_mutex);
+ mutex_lock(&root->fs_info->alloc_mutex);
if (ret)
break;
}
@@ -2254,11 +2258,9 @@
free_extent_buffer(next);
reada_walk_down(root, cur, path->slots[*level]);
- mutex_unlock(&root->fs_info->fs_mutex);
mutex_unlock(&root->fs_info->alloc_mutex);
next = read_tree_block(root, bytenr, blocksize,
ptr_gen);
- mutex_lock(&root->fs_info->fs_mutex);
mutex_lock(&root->fs_info->alloc_mutex);
/* we've dropped the lock, double check */
@@ -2381,6 +2383,7 @@
int orig_level;
struct btrfs_root_item *root_item = &root->root_item;
+ WARN_ON(!mutex_is_locked(&root->fs_info->drop_mutex));
path = btrfs_alloc_path();
BUG_ON(!path);
@@ -2710,7 +2713,6 @@
*last_file_root == ref_root)
goto out;
- mutex_unlock(&extent_root->fs_info->fs_mutex);
inode = btrfs_iget_locked(extent_root->fs_info->sb,
ref_objectid, found_root);
if (inode->i_state & I_NEW) {
@@ -2727,7 +2729,6 @@
* the latest version of the tree root
*/
if (is_bad_inode(inode)) {
- mutex_lock(&extent_root->fs_info->fs_mutex);
goto out;
}
*last_file_objectid = inode->i_ino;
@@ -2736,7 +2737,6 @@
relocate_inode_pages(inode, ref_offset, extent_key->offset);
iput(inode);
- mutex_lock(&extent_root->fs_info->fs_mutex);
} else {
struct btrfs_trans_handle *trans;
struct extent_buffer *eb;
@@ -3033,9 +3033,7 @@
if (progress && need_resched()) {
memcpy(&key, &found_key, sizeof(key));
- mutex_unlock(&root->fs_info->fs_mutex);
cond_resched();
- mutex_lock(&root->fs_info->fs_mutex);
btrfs_release_path(root, path);
btrfs_search_slot(NULL, root, &key, path, 0, 0);
progress = 0;
@@ -3068,9 +3066,7 @@
trans = btrfs_start_transaction(tree_root, 1);
btrfs_commit_transaction(trans, tree_root);
- mutex_unlock(&root->fs_info->fs_mutex);
btrfs_clean_old_snapshots(tree_root);
- mutex_lock(&root->fs_info->fs_mutex);
trans = btrfs_start_transaction(tree_root, 1);
btrfs_commit_transaction(trans, tree_root);
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index 73c6d08..18bbe10 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -252,7 +252,6 @@
end_of_last_block = start_pos + num_bytes - 1;
lock_extent(io_tree, start_pos, end_of_last_block, GFP_NOFS);
- mutex_lock(&root->fs_info->fs_mutex);
trans = btrfs_start_transaction(root, 1);
if (!trans) {
err = -ENOMEM;
@@ -341,7 +340,6 @@
failed:
err = btrfs_end_transaction(trans, root);
out_unlock:
- mutex_unlock(&root->fs_info->fs_mutex);
unlock_extent(io_tree, start_pos, end_of_last_block, GFP_NOFS);
return err;
}
@@ -905,9 +903,7 @@
WARN_ON(num_pages > nrptrs);
memset(pages, 0, sizeof(pages));
- mutex_lock(&root->fs_info->fs_mutex);
ret = btrfs_check_free_space(root, write_bytes, 0);
- mutex_unlock(&root->fs_info->fs_mutex);
if (ret)
goto out;
@@ -998,9 +994,9 @@
* check the transaction that last modified this inode
* and see if its already been committed
*/
- mutex_lock(&root->fs_info->fs_mutex);
if (!BTRFS_I(inode)->last_trans)
goto out;
+
mutex_lock(&root->fs_info->trans_mutex);
if (BTRFS_I(inode)->last_trans <=
root->fs_info->last_trans_committed) {
@@ -1023,7 +1019,6 @@
}
ret = btrfs_commit_transaction(trans, root);
out:
- mutex_unlock(&root->fs_info->fs_mutex);
return ret > 0 ? EIO : ret;
}
diff --git a/fs/btrfs/inode-map.c b/fs/btrfs/inode-map.c
index a0925ea..298346a 100644
--- a/fs/btrfs/inode-map.c
+++ b/fs/btrfs/inode-map.c
@@ -69,6 +69,12 @@
struct btrfs_key search_key;
u64 search_start = dirid;
+ mutex_lock(&root->objectid_mutex);
+ if (root->last_inode_alloc) {
+ *objectid = ++root->last_inode_alloc;
+ mutex_unlock(&root->objectid_mutex);
+ return 0;
+ }
path = btrfs_alloc_path();
BUG_ON(!path);
search_start = root->last_inode_alloc;
@@ -124,9 +130,11 @@
btrfs_release_path(root, path);
btrfs_free_path(path);
BUG_ON(*objectid < search_start);
+ mutex_unlock(&root->objectid_mutex);
return 0;
error:
btrfs_release_path(root, path);
btrfs_free_path(path);
+ mutex_unlock(&root->objectid_mutex);
return ret;
}
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 61bd895..b2251e2 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -79,12 +79,15 @@
int btrfs_check_free_space(struct btrfs_root *root, u64 num_required,
int for_del)
{
- u64 total = btrfs_super_total_bytes(&root->fs_info->super_copy);
- u64 used = btrfs_super_bytes_used(&root->fs_info->super_copy);
+ u64 total;
+ u64 used;
u64 thresh;
unsigned long flags;
int ret = 0;
+ spin_lock_irqsave(&root->fs_info->delalloc_lock, flags);
+ total = btrfs_super_total_bytes(&root->fs_info->super_copy);
+ used = btrfs_super_bytes_used(&root->fs_info->super_copy);
if (for_del)
thresh = total * 90;
else
@@ -92,7 +95,6 @@
do_div(thresh, 100);
- spin_lock_irqsave(&root->fs_info->delalloc_lock, flags);
if (used + root->fs_info->delalloc_bytes + num_required > thresh)
ret = -ENOSPC;
spin_unlock_irqrestore(&root->fs_info->delalloc_lock, flags);
@@ -115,7 +117,6 @@
trans = btrfs_start_transaction(root, 1);
BUG_ON(!trans);
btrfs_set_trans_block_group(trans, inode);
- mutex_unlock(&root->fs_info->fs_mutex);
num_bytes = (end - start + blocksize) & ~(blocksize - 1);
num_bytes = max(blocksize, num_bytes);
@@ -160,7 +161,6 @@
btrfs_add_ordered_inode(inode);
btrfs_update_inode(trans, root, inode);
out:
- mutex_lock(&root->fs_info->fs_mutex);
btrfs_end_transaction(trans, root);
return ret;
}
@@ -269,14 +269,13 @@
{
struct btrfs_root *root = BTRFS_I(inode)->root;
int ret;
- mutex_lock(&root->fs_info->fs_mutex);
+
if (btrfs_test_opt(root, NODATACOW) ||
btrfs_test_flag(inode, NODATACOW))
ret = run_delalloc_nocow(inode, start, end);
else
ret = cow_file_range(inode, start, end);
- mutex_unlock(&root->fs_info->fs_mutex);
return ret;
}
@@ -349,17 +348,13 @@
ret = btrfs_csum_one_bio(root, bio, &sums);
BUG_ON(ret);
- mutex_lock(&root->fs_info->fs_mutex);
trans = btrfs_start_transaction(root, 1);
- mutex_unlock(&root->fs_info->fs_mutex);
btrfs_set_trans_block_group(trans, inode);
btrfs_csum_file_blocks(trans, root, inode, bio, sums);
- mutex_lock(&root->fs_info->fs_mutex);
ret = btrfs_end_transaction(trans, root);
BUG_ON(ret);
- mutex_unlock(&root->fs_info->fs_mutex);
kfree(sums);
@@ -404,7 +399,6 @@
btrfs_test_flag(inode, NODATASUM))
return 0;
- mutex_lock(&root->fs_info->fs_mutex);
path = btrfs_alloc_path();
item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0);
if (IS_ERR(item)) {
@@ -422,7 +416,6 @@
out:
if (path)
btrfs_free_path(path);
- mutex_unlock(&root->fs_info->fs_mutex);
return ret;
}
@@ -616,7 +609,6 @@
path = btrfs_alloc_path();
BUG_ON(!path);
- mutex_lock(&root->fs_info->fs_mutex);
memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
@@ -662,8 +654,6 @@
btrfs_free_path(path);
inode_item = NULL;
- mutex_unlock(&root->fs_info->fs_mutex);
-
switch (inode->i_mode & S_IFMT) {
case S_IFREG:
inode->i_mapping->a_ops = &btrfs_aops;
@@ -691,9 +681,7 @@
return;
make_bad:
- btrfs_release_path(root, path);
btrfs_free_path(path);
- mutex_unlock(&root->fs_info->fs_mutex);
make_bad_inode(inode);
}
@@ -758,7 +746,6 @@
btrfs_set_inode_last_trans(trans, inode);
ret = 0;
failed:
- btrfs_release_path(root, path);
btrfs_free_path(path);
return ret;
}
@@ -849,7 +836,6 @@
unsigned long nr = 0;
root = BTRFS_I(dir)->root;
- mutex_lock(&root->fs_info->fs_mutex);
ret = btrfs_check_free_space(root, 1, 1);
if (ret)
@@ -871,7 +857,6 @@
btrfs_end_transaction(trans, root);
fail:
- mutex_unlock(&root->fs_info->fs_mutex);
btrfs_btree_balance_dirty(root, nr);
btrfs_throttle(root);
return ret;
@@ -890,7 +875,6 @@
return -ENOTEMPTY;
}
- mutex_lock(&root->fs_info->fs_mutex);
ret = btrfs_check_free_space(root, 1, 1);
if (ret)
goto fail;
@@ -907,7 +891,6 @@
nr = trans->blocks_used;
ret = btrfs_end_transaction(trans, root);
fail:
- mutex_unlock(&root->fs_info->fs_mutex);
btrfs_btree_balance_dirty(root, nr);
btrfs_throttle(root);
@@ -1129,7 +1112,6 @@
ret = btrfs_del_items(trans, root, path, pending_del_slot,
pending_del_nr);
}
- btrfs_release_path(root, path);
btrfs_free_path(path);
inode->i_sb->s_dirt = 1;
return ret;
@@ -1234,9 +1216,7 @@
if (attr->ia_size <= hole_start)
goto out;
- mutex_lock(&root->fs_info->fs_mutex);
err = btrfs_check_free_space(root, 1, 0);
- mutex_unlock(&root->fs_info->fs_mutex);
if (err)
goto fail;
@@ -1245,7 +1225,6 @@
lock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
hole_size = block_end - hole_start;
- mutex_lock(&root->fs_info->fs_mutex);
trans = btrfs_start_transaction(root, 1);
btrfs_set_trans_block_group(trans, inode);
err = btrfs_drop_extents(trans, root, inode,
@@ -1262,7 +1241,6 @@
btrfs_check_file(root, inode);
}
btrfs_end_transaction(trans, root);
- mutex_unlock(&root->fs_info->fs_mutex);
unlock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
if (err)
return err;
@@ -1286,7 +1264,6 @@
}
inode->i_size = 0;
- mutex_lock(&root->fs_info->fs_mutex);
trans = btrfs_start_transaction(root, 1);
btrfs_set_trans_block_group(trans, inode);
@@ -1298,7 +1275,6 @@
clear_inode(inode);
btrfs_end_transaction(trans, root);
- mutex_unlock(&root->fs_info->fs_mutex);
btrfs_btree_balance_dirty(root, nr);
btrfs_throttle(root);
return;
@@ -1306,7 +1282,6 @@
no_delete_lock:
nr = trans->blocks_used;
btrfs_end_transaction(trans, root);
- mutex_unlock(&root->fs_info->fs_mutex);
btrfs_btree_balance_dirty(root, nr);
btrfs_throttle(root);
no_delete:
@@ -1402,7 +1377,6 @@
path = btrfs_alloc_path();
BUG_ON(!path);
- mutex_lock(&root->fs_info->fs_mutex);
*sub_root = btrfs_read_fs_root(root->fs_info, location,
dentry->d_name.name,
@@ -1416,7 +1390,6 @@
location->offset = 0;
btrfs_free_path(path);
- mutex_unlock(&root->fs_info->fs_mutex);
return 0;
}
@@ -1482,9 +1455,7 @@
if (dentry->d_name.len > BTRFS_NAME_LEN)
return ERR_PTR(-ENAMETOOLONG);
- mutex_lock(&root->fs_info->fs_mutex);
ret = btrfs_inode_by_name(dir, dentry, &location);
- mutex_unlock(&root->fs_info->fs_mutex);
if (ret < 0)
return ERR_PTR(ret);
@@ -1559,7 +1530,6 @@
filp->f_pos = 1;
}
- mutex_lock(&root->fs_info->fs_mutex);
key.objectid = inode->i_ino;
path = btrfs_alloc_path();
path->reada = 2;
@@ -1668,9 +1638,7 @@
nopos:
ret = 0;
err:
- btrfs_release_path(root, path);
btrfs_free_path(path);
- mutex_unlock(&root->fs_info->fs_mutex);
return ret;
}
@@ -1681,11 +1649,9 @@
int ret = 0;
if (wait) {
- mutex_lock(&root->fs_info->fs_mutex);
trans = btrfs_start_transaction(root, 1);
btrfs_set_trans_block_group(trans, inode);
ret = btrfs_commit_transaction(trans, root);
- mutex_unlock(&root->fs_info->fs_mutex);
}
return ret;
}
@@ -1701,12 +1667,10 @@
struct btrfs_root *root = BTRFS_I(inode)->root;
struct btrfs_trans_handle *trans;
- mutex_lock(&root->fs_info->fs_mutex);
trans = btrfs_start_transaction(root, 1);
btrfs_set_trans_block_group(trans, inode);
btrfs_update_inode(trans, root, inode);
btrfs_end_transaction(trans, root);
- mutex_unlock(&root->fs_info->fs_mutex);
}
static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
@@ -1874,7 +1838,6 @@
if (!new_valid_dev(rdev))
return -EINVAL;
- mutex_lock(&root->fs_info->fs_mutex);
err = btrfs_check_free_space(root, 1, 0);
if (err)
goto fail;
@@ -1912,8 +1875,6 @@
nr = trans->blocks_used;
btrfs_end_transaction(trans, root);
fail:
- mutex_unlock(&root->fs_info->fs_mutex);
-
if (drop_inode) {
inode_dec_link_count(inode);
iput(inode);
@@ -1934,7 +1895,6 @@
unsigned long nr = 0;
u64 objectid;
- mutex_lock(&root->fs_info->fs_mutex);
err = btrfs_check_free_space(root, 1, 0);
if (err)
goto fail;
@@ -1980,8 +1940,6 @@
nr = trans->blocks_used;
btrfs_end_transaction(trans, root);
fail:
- mutex_unlock(&root->fs_info->fs_mutex);
-
if (drop_inode) {
inode_dec_link_count(inode);
iput(inode);
@@ -2009,7 +1967,6 @@
#else
inc_nlink(inode);
#endif
- mutex_lock(&root->fs_info->fs_mutex);
err = btrfs_check_free_space(root, 1, 0);
if (err)
goto fail;
@@ -2032,8 +1989,6 @@
nr = trans->blocks_used;
btrfs_end_transaction(trans, root);
fail:
- mutex_unlock(&root->fs_info->fs_mutex);
-
if (drop_inode) {
inode_dec_link_count(inode);
iput(inode);
@@ -2053,7 +2008,6 @@
u64 objectid = 0;
unsigned long nr = 1;
- mutex_lock(&root->fs_info->fs_mutex);
err = btrfs_check_free_space(root, 1, 0);
if (err)
goto out_unlock;
@@ -2106,7 +2060,6 @@
btrfs_end_transaction(trans, root);
out_unlock:
- mutex_unlock(&root->fs_info->fs_mutex);
if (drop_on_err)
iput(inode);
btrfs_btree_balance_dirty(root, nr);
@@ -2199,7 +2152,6 @@
path = btrfs_alloc_path();
BUG_ON(!path);
- mutex_lock(&root->fs_info->fs_mutex);
again:
spin_lock(&em_tree->lock);
@@ -2402,7 +2354,6 @@
if (!err)
err = ret;
}
- mutex_unlock(&root->fs_info->fs_mutex);
if (err) {
free_extent_map(em);
WARN_ON(1);
@@ -2584,9 +2535,7 @@
int ret;
u64 page_start;
- mutex_lock(&root->fs_info->fs_mutex);
ret = btrfs_check_free_space(root, PAGE_CACHE_SIZE, 0);
- mutex_unlock(&root->fs_info->fs_mutex);
if (ret)
goto out;
@@ -2631,7 +2580,6 @@
btrfs_truncate_page(inode->i_mapping, inode->i_size);
- mutex_lock(&root->fs_info->fs_mutex);
trans = btrfs_start_transaction(root, 1);
btrfs_set_trans_block_group(trans, inode);
@@ -2643,7 +2591,6 @@
ret = btrfs_end_transaction(trans, root);
BUG_ON(ret);
- mutex_unlock(&root->fs_info->fs_mutex);
btrfs_btree_balance_dirty(root, nr);
btrfs_throttle(root);
}
@@ -2827,7 +2774,6 @@
struct inode *new_inode = new_dentry->d_inode;
struct inode *old_inode = old_dentry->d_inode;
struct timespec ctime = CURRENT_TIME;
- struct btrfs_path *path;
int ret;
if (S_ISDIR(old_inode->i_mode) && new_inode &&
@@ -2835,7 +2781,6 @@
return -ENOTEMPTY;
}
- mutex_lock(&root->fs_info->fs_mutex);
ret = btrfs_check_free_space(root, 1, 0);
if (ret)
goto out_unlock;
@@ -2843,11 +2788,6 @@
trans = btrfs_start_transaction(root, 1);
btrfs_set_trans_block_group(trans, new_dir);
- path = btrfs_alloc_path();
- if (!path) {
- ret = -ENOMEM;
- goto out_fail;
- }
old_dentry->d_inode->i_nlink++;
old_dir->i_ctime = old_dir->i_mtime = ctime;
@@ -2869,10 +2809,8 @@
goto out_fail;
out_fail:
- btrfs_free_path(path);
btrfs_end_transaction(trans, root);
out_unlock:
- mutex_unlock(&root->fs_info->fs_mutex);
return ret;
}
@@ -2898,7 +2836,6 @@
if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
return -ENAMETOOLONG;
- mutex_lock(&root->fs_info->fs_mutex);
err = btrfs_check_free_space(root, 1, 0);
if (err)
goto out_fail;
@@ -2979,7 +2916,6 @@
nr = trans->blocks_used;
btrfs_end_transaction(trans, root);
out_fail:
- mutex_unlock(&root->fs_info->fs_mutex);
if (drop_inode) {
inode_dec_link_count(inode);
iput(inode);
diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
index 3fbf74e..6002eb6 100644
--- a/fs/btrfs/ioctl.c
+++ b/fs/btrfs/ioctl.c
@@ -63,7 +63,6 @@
u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
unsigned long nr = 1;
- mutex_lock(&root->fs_info->fs_mutex);
ret = btrfs_check_free_space(root, 1, 0);
if (ret)
goto fail_commit;
@@ -164,7 +163,6 @@
if (err && !ret)
ret = err;
fail_commit:
- mutex_unlock(&root->fs_info->fs_mutex);
btrfs_btree_balance_dirty(root, nr);
btrfs_throttle(root);
return ret;
@@ -181,7 +179,6 @@
if (!root->ref_cows)
return -EINVAL;
- mutex_lock(&root->fs_info->fs_mutex);
ret = btrfs_check_free_space(root, 1, 0);
if (ret)
goto fail_unlock;
@@ -208,7 +205,6 @@
err = btrfs_commit_transaction(trans, root);
fail_unlock:
- mutex_unlock(&root->fs_info->fs_mutex);
btrfs_btree_balance_dirty(root, nr);
btrfs_throttle(root);
return ret;
@@ -228,9 +224,7 @@
unsigned long i;
int ret;
- mutex_lock(&root->fs_info->fs_mutex);
ret = btrfs_check_free_space(root, inode->i_size, 0);
- mutex_unlock(&root->fs_info->fs_mutex);
if (ret)
return -ENOSPC;
@@ -315,7 +309,8 @@
goto out;
}
- mutex_lock(&root->fs_info->fs_mutex);
+ mutex_lock(&root->fs_info->alloc_mutex);
+ mutex_lock(&root->fs_info->chunk_mutex);
sizestr = vol_args->name;
devstr = strchr(sizestr, ':');
if (devstr) {
@@ -385,7 +380,8 @@
}
out_unlock:
- mutex_unlock(&root->fs_info->fs_mutex);
+ mutex_lock(&root->fs_info->alloc_mutex);
+ mutex_lock(&root->fs_info->chunk_mutex);
out:
kfree(vol_args);
return ret;
@@ -428,11 +424,9 @@
}
root_dirid = root->fs_info->sb->s_root->d_inode->i_ino,
- mutex_lock(&root->fs_info->fs_mutex);
di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root,
path, root_dirid,
vol_args->name, namelen, 0);
- mutex_unlock(&root->fs_info->fs_mutex);
btrfs_free_path(path);
if (di && !IS_ERR(di)) {
@@ -445,10 +439,12 @@
goto out;
}
+ mutex_lock(&root->fs_info->drop_mutex);
if (root == root->fs_info->tree_root)
ret = create_subvol(root, vol_args->name, namelen);
else
ret = create_snapshot(root, vol_args->name, namelen);
+ mutex_unlock(&root->fs_info->drop_mutex);
out:
kfree(vol_args);
return ret;
@@ -461,10 +457,8 @@
switch (inode->i_mode & S_IFMT) {
case S_IFDIR:
- mutex_lock(&root->fs_info->fs_mutex);
btrfs_defrag_root(root, 0);
btrfs_defrag_root(root->fs_info->extent_root, 0);
- mutex_unlock(&root->fs_info->fs_mutex);
break;
case S_IFREG:
btrfs_defrag_file(file);
@@ -588,7 +582,6 @@
unlock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS);
}
- mutex_lock(&root->fs_info->fs_mutex);
trans = btrfs_start_transaction(root, 0);
path = btrfs_alloc_path();
if (!path) {
@@ -685,7 +678,6 @@
unlock_extent(&BTRFS_I(src)->io_tree, 0, (u64)-1, GFP_NOFS);
btrfs_end_transaction(trans, root);
- mutex_unlock(&root->fs_info->fs_mutex);
out_unlock:
mutex_unlock(&src->i_mutex);
@@ -711,7 +703,6 @@
if (!capable(CAP_SYS_ADMIN))
return -EPERM;
- mutex_lock(&root->fs_info->fs_mutex);
if (file->private_data) {
ret = -EINPROGRESS;
goto out;
@@ -723,7 +714,6 @@
ret = -ENOMEM;
/*printk(KERN_INFO "btrfs_ioctl_trans_start on %p\n", file);*/
out:
- mutex_unlock(&root->fs_info->fs_mutex);
return ret;
}
@@ -740,7 +730,6 @@
struct btrfs_trans_handle *trans;
int ret = 0;
- mutex_lock(&root->fs_info->fs_mutex);
trans = file->private_data;
if (!trans) {
ret = -EINVAL;
@@ -749,7 +738,6 @@
btrfs_end_transaction(trans, root);
file->private_data = 0;
out:
- mutex_unlock(&root->fs_info->fs_mutex);
return ret;
}
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 196d0e2..b61ded7 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -366,12 +366,10 @@
return 0;
}
btrfs_clean_old_snapshots(root);
- mutex_lock(&root->fs_info->fs_mutex);
btrfs_defrag_dirty_roots(root->fs_info);
trans = btrfs_start_transaction(root, 1);
ret = btrfs_commit_transaction(trans, root);
sb->s_dirt = 0;
- mutex_unlock(&root->fs_info->fs_mutex);
return ret;
}
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 1ed433a..5a1ee06 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -370,6 +370,7 @@
struct btrfs_trans_handle *trans;
unsigned long nr;
+ smp_mb();
if (root->defrag_running)
return 0;
trans = btrfs_start_transaction(root, 1);
@@ -378,16 +379,15 @@
ret = btrfs_defrag_leaves(trans, root, cacheonly);
nr = trans->blocks_used;
btrfs_end_transaction(trans, root);
- mutex_unlock(&info->fs_mutex);
btrfs_btree_balance_dirty(info->tree_root, nr);
cond_resched();
- mutex_lock(&info->fs_mutex);
trans = btrfs_start_transaction(root, 1);
if (ret != -EAGAIN)
break;
}
root->defrag_running = 0;
+ smp_mb();
radix_tree_tag_clear(&info->fs_roots_radix,
(unsigned long)root->root_key.objectid,
BTRFS_ROOT_DEFRAG_TAG);
@@ -435,14 +435,14 @@
while(!list_empty(list)) {
struct btrfs_root *root;
- mutex_lock(&tree_root->fs_info->fs_mutex);
dirty = list_entry(list->next, struct dirty_root, list);
list_del_init(&dirty->list);
num_bytes = btrfs_root_used(&dirty->root->root_item);
root = dirty->latest_root;
- root->fs_info->throttles++;
+ atomic_inc(&root->fs_info->throttles);
+ mutex_lock(&root->fs_info->drop_mutex);
while(1) {
trans = btrfs_start_transaction(tree_root, 1);
ret = btrfs_drop_snapshot(trans, dirty->root);
@@ -459,14 +459,16 @@
nr = trans->blocks_used;
ret = btrfs_end_transaction(trans, tree_root);
BUG_ON(ret);
- mutex_unlock(&tree_root->fs_info->fs_mutex);
+
+ mutex_unlock(&root->fs_info->drop_mutex);
btrfs_btree_balance_dirty(tree_root, nr);
cond_resched();
- mutex_lock(&tree_root->fs_info->fs_mutex);
+ mutex_lock(&root->fs_info->drop_mutex);
}
BUG_ON(ret);
- root->fs_info->throttles--;
+ atomic_dec(&root->fs_info->throttles);
+ mutex_lock(&root->fs_info->alloc_mutex);
num_bytes -= btrfs_root_used(&dirty->root->root_item);
bytes_used = btrfs_root_used(&root->root_item);
if (num_bytes) {
@@ -474,11 +476,15 @@
btrfs_set_root_used(&root->root_item,
bytes_used - num_bytes);
}
+ mutex_unlock(&root->fs_info->alloc_mutex);
+
ret = btrfs_del_root(trans, tree_root, &dirty->root->root_key);
if (ret) {
BUG();
break;
}
+ mutex_unlock(&root->fs_info->drop_mutex);
+
nr = trans->blocks_used;
ret = btrfs_end_transaction(trans, tree_root);
BUG_ON(ret);
@@ -486,7 +492,6 @@
free_extent_buffer(dirty->root->node);
kfree(dirty->root);
kfree(dirty);
- mutex_unlock(&tree_root->fs_info->fs_mutex);
btrfs_btree_balance_dirty(tree_root, nr);
cond_resched();
@@ -503,7 +508,7 @@
u64 objectid = 0;
int ret;
- root->fs_info->throttles++;
+ atomic_inc(&root->fs_info->throttles);
while(1) {
ret = btrfs_find_first_ordered_inode(
&cur_trans->ordered_inode_tree,
@@ -512,7 +517,6 @@
break;
mutex_unlock(&root->fs_info->trans_mutex);
- mutex_unlock(&root->fs_info->fs_mutex);
if (S_ISREG(inode->i_mode)) {
atomic_inc(&BTRFS_I(inode)->ordered_writeback);
@@ -521,7 +525,6 @@
}
iput(inode);
- mutex_lock(&root->fs_info->fs_mutex);
mutex_lock(&root->fs_info->trans_mutex);
}
while(1) {
@@ -533,7 +536,6 @@
if (!ret)
break;
mutex_unlock(&root->fs_info->trans_mutex);
- mutex_unlock(&root->fs_info->fs_mutex);
if (S_ISREG(inode->i_mode)) {
atomic_inc(&BTRFS_I(inode)->ordered_writeback);
@@ -543,10 +545,9 @@
atomic_dec(&inode->i_count);
iput(inode);
- mutex_lock(&root->fs_info->fs_mutex);
mutex_lock(&root->fs_info->trans_mutex);
}
- root->fs_info->throttles--;
+ atomic_dec(&root->fs_info->throttles);
return 0;
}
@@ -661,7 +662,6 @@
mutex_unlock(&root->fs_info->trans_mutex);
btrfs_end_transaction(trans, root);
- mutex_unlock(&root->fs_info->fs_mutex);
ret = wait_for_commit(root, cur_trans);
BUG_ON(ret);
@@ -669,7 +669,6 @@
put_transaction(cur_trans);
mutex_unlock(&root->fs_info->trans_mutex);
- mutex_lock(&root->fs_info->fs_mutex);
return 0;
}
@@ -687,12 +686,10 @@
struct btrfs_transaction, list);
if (!prev_trans->commit_done) {
prev_trans->use_count++;
- mutex_unlock(&root->fs_info->fs_mutex);
mutex_unlock(&root->fs_info->trans_mutex);
wait_for_commit(root, prev_trans);
- mutex_lock(&root->fs_info->fs_mutex);
mutex_lock(&root->fs_info->trans_mutex);
put_transaction(prev_trans);
}
@@ -709,12 +706,10 @@
else
timeout = 1;
- mutex_unlock(&root->fs_info->fs_mutex);
mutex_unlock(&root->fs_info->trans_mutex);
schedule_timeout(timeout);
- mutex_lock(&root->fs_info->fs_mutex);
mutex_lock(&root->fs_info->trans_mutex);
finish_wait(&cur_trans->writer_wait, &wait);
ret = btrfs_write_ordered_inodes(trans, root);
@@ -755,12 +750,10 @@
btrfs_copy_pinned(root, pinned_copy);
mutex_unlock(&root->fs_info->trans_mutex);
- mutex_unlock(&root->fs_info->fs_mutex);
ret = btrfs_write_and_wait_transaction(trans, root);
BUG_ON(ret);
write_ctree_super(trans, root);
- mutex_lock(&root->fs_info->fs_mutex);
btrfs_finish_extent_commit(trans, root, pinned_copy);
mutex_lock(&root->fs_info->trans_mutex);
@@ -781,9 +774,7 @@
kmem_cache_free(btrfs_trans_handle_cachep, trans);
if (root->fs_info->closing) {
- mutex_unlock(&root->fs_info->fs_mutex);
drop_dirty_roots(root->fs_info->tree_root, &dirty_fs_roots);
- mutex_lock(&root->fs_info->fs_mutex);
}
return ret;
}
@@ -823,7 +814,7 @@
unsigned long delay = HZ * 30;
int ret;
- mutex_lock(&root->fs_info->fs_mutex);
+ smp_mb();
if (root->fs_info->closing)
goto out;
@@ -844,7 +835,6 @@
trans = btrfs_start_transaction(root, 1);
ret = btrfs_commit_transaction(trans, root);
out:
- mutex_unlock(&root->fs_info->fs_mutex);
btrfs_clean_old_snapshots(root);
btrfs_transaction_queue_work(root, delay);
}
diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index ba396857..869864d 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -866,7 +866,8 @@
u64 devid;
int ret = 0;
- mutex_lock(&root->fs_info->fs_mutex);
+ mutex_lock(&root->fs_info->alloc_mutex);
+ mutex_lock(&root->fs_info->chunk_mutex);
mutex_lock(&uuid_mutex);
all_avail = root->fs_info->avail_data_alloc_bits |
@@ -984,7 +985,8 @@
close_bdev_excl(bdev);
out:
mutex_unlock(&uuid_mutex);
- mutex_unlock(&root->fs_info->fs_mutex);
+ mutex_unlock(&root->fs_info->chunk_mutex);
+ mutex_unlock(&root->fs_info->alloc_mutex);
return ret;
}
@@ -1003,7 +1005,10 @@
if (!bdev) {
return -EIO;
}
- mutex_lock(&root->fs_info->fs_mutex);
+
+ mutex_lock(&root->fs_info->alloc_mutex);
+ mutex_lock(&root->fs_info->chunk_mutex);
+
trans = btrfs_start_transaction(root, 1);
devices = &root->fs_info->fs_devices->devices;
list_for_each(cur, devices) {
@@ -1057,7 +1062,9 @@
root->fs_info->fs_devices->open_devices++;
out:
btrfs_end_transaction(trans, root);
- mutex_unlock(&root->fs_info->fs_mutex);
+ mutex_unlock(&root->fs_info->chunk_mutex);
+ mutex_unlock(&root->fs_info->alloc_mutex);
+
return ret;
out_close_bdev:
@@ -1297,9 +1304,10 @@
struct btrfs_key found_key;
+ BUG(); /* FIXME, needs locking */
+
dev_root = dev_root->fs_info->dev_root;
- mutex_lock(&dev_root->fs_info->fs_mutex);
/* step one make some room on all the devices */
list_for_each(cur, devices) {
device = list_entry(cur, struct btrfs_device, dev_list);
@@ -1368,7 +1376,6 @@
ret = 0;
error:
btrfs_free_path(path);
- mutex_unlock(&dev_root->fs_info->fs_mutex);
return ret;
}
diff --git a/fs/btrfs/xattr.c b/fs/btrfs/xattr.c
index f63488d..3e9eb91 100644
--- a/fs/btrfs/xattr.c
+++ b/fs/btrfs/xattr.c
@@ -153,7 +153,6 @@
return -ENOMEM;
}
- mutex_lock(&root->fs_info->fs_mutex);
/* lookup the xattr by name */
di = btrfs_lookup_xattr(NULL, root, path, inode->i_ino, name,
strlen(name), 0);
@@ -181,7 +180,6 @@
ret = btrfs_dir_data_len(leaf, di);
out:
- mutex_unlock(&root->fs_info->fs_mutex);
kfree(name);
btrfs_free_path(path);
return ret;
@@ -210,7 +208,6 @@
return -ENOMEM;
}
- mutex_lock(&root->fs_info->fs_mutex);
trans = btrfs_start_transaction(root, 1);
btrfs_set_trans_block_group(trans, inode);
@@ -260,7 +257,6 @@
}
btrfs_end_transaction(trans, root);
- mutex_unlock(&root->fs_info->fs_mutex);
kfree(name);
btrfs_free_path(path);
@@ -297,8 +293,6 @@
return -ENOMEM;
path->reada = 2;
- mutex_lock(&root->fs_info->fs_mutex);
-
/* search for our xattrs */
ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
if (ret < 0)
@@ -379,15 +373,13 @@
ret = total_size;
err:
- mutex_unlock(&root->fs_info->fs_mutex);
btrfs_free_path(path);
return ret;
}
/*
- * delete all the xattrs associated with the inode. fs_mutex should be
- * held when we come into here
+ * delete all the xattrs associated with the inode.
*/
int btrfs_delete_xattrs(struct btrfs_trans_handle *trans,
struct btrfs_root *root, struct inode *inode)
