Btrfs: introduce per-subvolume ordered extent list
The reason we introduce per-subvolume ordered extent list is the same
as the per-subvolume delalloc inode list.
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c
index 1ddd728..665c640 100644
--- a/fs/btrfs/ordered-data.c
+++ b/fs/btrfs/ordered-data.c
@@ -24,6 +24,7 @@
#include "transaction.h"
#include "btrfs_inode.h"
#include "extent_io.h"
+#include "disk-io.h"
static struct kmem_cache *btrfs_ordered_extent_cache;
@@ -184,6 +185,7 @@
u64 start, u64 len, u64 disk_len,
int type, int dio, int compress_type)
{
+ struct btrfs_root *root = BTRFS_I(inode)->root;
struct btrfs_ordered_inode_tree *tree;
struct rb_node *node;
struct btrfs_ordered_extent *entry;
@@ -227,10 +229,18 @@
ordered_data_tree_panic(inode, -EEXIST, file_offset);
spin_unlock_irq(&tree->lock);
- spin_lock(&BTRFS_I(inode)->root->fs_info->ordered_extent_lock);
+ spin_lock(&root->ordered_extent_lock);
list_add_tail(&entry->root_extent_list,
- &BTRFS_I(inode)->root->fs_info->ordered_extents);
- spin_unlock(&BTRFS_I(inode)->root->fs_info->ordered_extent_lock);
+ &root->ordered_extents);
+ root->nr_ordered_extents++;
+ if (root->nr_ordered_extents == 1) {
+ spin_lock(&root->fs_info->ordered_root_lock);
+ BUG_ON(!list_empty(&root->ordered_root));
+ list_add_tail(&root->ordered_root,
+ &root->fs_info->ordered_roots);
+ spin_unlock(&root->fs_info->ordered_root_lock);
+ }
+ spin_unlock(&root->ordered_extent_lock);
return 0;
}
@@ -516,8 +526,9 @@
set_bit(BTRFS_ORDERED_COMPLETE, &entry->flags);
spin_unlock_irq(&tree->lock);
- spin_lock(&root->fs_info->ordered_extent_lock);
+ spin_lock(&root->ordered_extent_lock);
list_del_init(&entry->root_extent_list);
+ root->nr_ordered_extents--;
trace_btrfs_ordered_extent_remove(inode, entry);
@@ -530,7 +541,14 @@
!mapping_tagged(inode->i_mapping, PAGECACHE_TAG_DIRTY)) {
list_del_init(&BTRFS_I(inode)->ordered_operations);
}
- spin_unlock(&root->fs_info->ordered_extent_lock);
+
+ if (!root->nr_ordered_extents) {
+ spin_lock(&root->fs_info->ordered_root_lock);
+ BUG_ON(list_empty(&root->ordered_root));
+ list_del_init(&root->ordered_root);
+ spin_unlock(&root->fs_info->ordered_root_lock);
+ }
+ spin_unlock(&root->ordered_extent_lock);
wake_up(&entry->wait);
}
@@ -550,7 +568,6 @@
void btrfs_wait_ordered_extents(struct btrfs_root *root, int delay_iput)
{
struct list_head splice, works;
- struct list_head *cur;
struct btrfs_ordered_extent *ordered, *next;
struct inode *inode;
@@ -558,35 +575,34 @@
INIT_LIST_HEAD(&works);
mutex_lock(&root->fs_info->ordered_operations_mutex);
- spin_lock(&root->fs_info->ordered_extent_lock);
- list_splice_init(&root->fs_info->ordered_extents, &splice);
+ spin_lock(&root->ordered_extent_lock);
+ list_splice_init(&root->ordered_extents, &splice);
while (!list_empty(&splice)) {
- cur = splice.next;
- ordered = list_entry(cur, struct btrfs_ordered_extent,
- root_extent_list);
- list_del_init(&ordered->root_extent_list);
- atomic_inc(&ordered->refs);
-
+ ordered = list_first_entry(&splice, struct btrfs_ordered_extent,
+ root_extent_list);
+ list_move_tail(&ordered->root_extent_list,
+ &root->ordered_extents);
/*
* the inode may be getting freed (in sys_unlink path).
*/
inode = igrab(ordered->inode);
-
- spin_unlock(&root->fs_info->ordered_extent_lock);
-
- if (inode) {
- ordered->flush_work.func = btrfs_run_ordered_extent_work;
- list_add_tail(&ordered->work_list, &works);
- btrfs_queue_worker(&root->fs_info->flush_workers,
- &ordered->flush_work);
- } else {
- btrfs_put_ordered_extent(ordered);
+ if (!inode) {
+ cond_resched_lock(&root->ordered_extent_lock);
+ continue;
}
+ atomic_inc(&ordered->refs);
+ spin_unlock(&root->ordered_extent_lock);
+
+ ordered->flush_work.func = btrfs_run_ordered_extent_work;
+ list_add_tail(&ordered->work_list, &works);
+ btrfs_queue_worker(&root->fs_info->flush_workers,
+ &ordered->flush_work);
+
cond_resched();
- spin_lock(&root->fs_info->ordered_extent_lock);
+ spin_lock(&root->ordered_extent_lock);
}
- spin_unlock(&root->fs_info->ordered_extent_lock);
+ spin_unlock(&root->ordered_extent_lock);
list_for_each_entry_safe(ordered, next, &works, work_list) {
list_del_init(&ordered->work_list);
@@ -604,6 +620,33 @@
mutex_unlock(&root->fs_info->ordered_operations_mutex);
}
+void btrfs_wait_all_ordered_extents(struct btrfs_fs_info *fs_info,
+ int delay_iput)
+{
+ struct btrfs_root *root;
+ struct list_head splice;
+
+ INIT_LIST_HEAD(&splice);
+
+ spin_lock(&fs_info->ordered_root_lock);
+ list_splice_init(&fs_info->ordered_roots, &splice);
+ while (!list_empty(&splice)) {
+ root = list_first_entry(&splice, struct btrfs_root,
+ ordered_root);
+ root = btrfs_grab_fs_root(root);
+ BUG_ON(!root);
+ list_move_tail(&root->ordered_root,
+ &fs_info->ordered_roots);
+ spin_unlock(&fs_info->ordered_root_lock);
+
+ btrfs_wait_ordered_extents(root, delay_iput);
+ btrfs_put_fs_root(root);
+
+ spin_lock(&fs_info->ordered_root_lock);
+ }
+ spin_unlock(&fs_info->ordered_root_lock);
+}
+
/*
* this is used during transaction commit to write all the inodes
* added to the ordered operation list. These files must be fully on
@@ -629,7 +672,7 @@
INIT_LIST_HEAD(&works);
mutex_lock(&root->fs_info->ordered_operations_mutex);
- spin_lock(&root->fs_info->ordered_extent_lock);
+ spin_lock(&root->fs_info->ordered_root_lock);
list_splice_init(&cur_trans->ordered_operations, &splice);
while (!list_empty(&splice)) {
btrfs_inode = list_entry(splice.next, struct btrfs_inode,
@@ -648,17 +691,17 @@
if (!wait)
list_add_tail(&BTRFS_I(inode)->ordered_operations,
&cur_trans->ordered_operations);
- spin_unlock(&root->fs_info->ordered_extent_lock);
+ spin_unlock(&root->fs_info->ordered_root_lock);
work = btrfs_alloc_delalloc_work(inode, wait, 1);
if (!work) {
- spin_lock(&root->fs_info->ordered_extent_lock);
+ spin_lock(&root->fs_info->ordered_root_lock);
if (list_empty(&BTRFS_I(inode)->ordered_operations))
list_add_tail(&btrfs_inode->ordered_operations,
&splice);
list_splice_tail(&splice,
&cur_trans->ordered_operations);
- spin_unlock(&root->fs_info->ordered_extent_lock);
+ spin_unlock(&root->fs_info->ordered_root_lock);
ret = -ENOMEM;
goto out;
}
@@ -667,9 +710,9 @@
&work->work);
cond_resched();
- spin_lock(&root->fs_info->ordered_extent_lock);
+ spin_lock(&root->fs_info->ordered_root_lock);
}
- spin_unlock(&root->fs_info->ordered_extent_lock);
+ spin_unlock(&root->fs_info->ordered_root_lock);
out:
list_for_each_entry_safe(work, next, &works, list) {
list_del_init(&work->list);
@@ -1055,12 +1098,12 @@
if (last_mod < root->fs_info->last_trans_committed)
return;
- spin_lock(&root->fs_info->ordered_extent_lock);
+ spin_lock(&root->fs_info->ordered_root_lock);
if (list_empty(&BTRFS_I(inode)->ordered_operations)) {
list_add_tail(&BTRFS_I(inode)->ordered_operations,
&cur_trans->ordered_operations);
}
- spin_unlock(&root->fs_info->ordered_extent_lock);
+ spin_unlock(&root->fs_info->ordered_root_lock);
}
int __init ordered_data_init(void)