Btrfs: delay commits during fsync to allow more writers
Signed-off-by: Chris Mason <chris.mason@oracle.com>
This commit is contained in:
parent
f2183bde1a
commit
15ee9bc7ed
7 changed files with 66 additions and 16 deletions
|
@ -25,6 +25,11 @@ struct btrfs_inode {
|
|||
struct btrfs_block_group_cache *block_group;
|
||||
struct btrfs_key location;
|
||||
struct inode vfs_inode;
|
||||
|
||||
/*
|
||||
* transid of the trans_handle that last modified this inode
|
||||
*/
|
||||
u64 last_trans;
|
||||
};
|
||||
static inline struct btrfs_inode *BTRFS_I(struct inode *inode)
|
||||
{
|
||||
|
|
|
@ -301,6 +301,7 @@ struct btrfs_fs_info {
|
|||
struct radix_tree_root extent_map_radix;
|
||||
struct radix_tree_root extent_ins_radix;
|
||||
u64 generation;
|
||||
u64 last_trans_committed;
|
||||
struct btrfs_transaction *running_transaction;
|
||||
struct btrfs_super_block *disk_super;
|
||||
struct btrfs_super_block super_copy;
|
||||
|
|
|
@ -435,6 +435,7 @@ struct btrfs_root *open_ctree(struct super_block *sb)
|
|||
INIT_LIST_HEAD(&fs_info->dead_roots);
|
||||
sb_set_blocksize(sb, 4096);
|
||||
fs_info->running_transaction = NULL;
|
||||
fs_info->last_trans_committed = 0;
|
||||
fs_info->tree_root = tree_root;
|
||||
fs_info->extent_root = extent_root;
|
||||
fs_info->sb = sb;
|
||||
|
|
|
@ -694,22 +694,36 @@ static int btrfs_sync_file(struct file *file,
|
|||
{
|
||||
struct inode *inode = dentry->d_inode;
|
||||
struct btrfs_root *root = BTRFS_I(inode)->root;
|
||||
int ret;
|
||||
int ret = 0;
|
||||
struct btrfs_trans_handle *trans;
|
||||
|
||||
/*
|
||||
* FIXME, use inode generation number to check if we can skip the
|
||||
* commit
|
||||
* 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) {
|
||||
BTRFS_I(inode)->last_trans = 0;
|
||||
mutex_unlock(&root->fs_info->trans_mutex);
|
||||
goto out;
|
||||
}
|
||||
mutex_unlock(&root->fs_info->trans_mutex);
|
||||
|
||||
/*
|
||||
* ok we haven't committed the transaction yet, lets do a commit
|
||||
*/
|
||||
trans = btrfs_start_transaction(root, 1);
|
||||
if (!trans) {
|
||||
ret = -ENOMEM;
|
||||
goto out;
|
||||
}
|
||||
ret = btrfs_commit_transaction(trans, root);
|
||||
mutex_unlock(&root->fs_info->fs_mutex);
|
||||
out:
|
||||
mutex_unlock(&root->fs_info->fs_mutex);
|
||||
return ret > 0 ? EIO : ret;
|
||||
}
|
||||
|
||||
|
|
|
@ -193,6 +193,7 @@ static int btrfs_update_inode(struct btrfs_trans_handle *trans,
|
|||
|
||||
fill_inode_item(inode_item, inode);
|
||||
btrfs_mark_buffer_dirty(path->nodes[0]);
|
||||
btrfs_set_inode_last_trans(trans, inode);
|
||||
ret = 0;
|
||||
failed:
|
||||
btrfs_release_path(root, path);
|
||||
|
@ -2234,6 +2235,7 @@ struct inode *btrfs_alloc_inode(struct super_block *sb)
|
|||
ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
|
||||
if (!ei)
|
||||
return NULL;
|
||||
ei->last_trans = 0;
|
||||
return &ei->vfs_inode;
|
||||
}
|
||||
|
||||
|
|
|
@ -55,7 +55,8 @@ static int join_transaction(struct btrfs_root *root)
|
|||
BUG_ON(!cur_trans);
|
||||
root->fs_info->generation++;
|
||||
root->fs_info->running_transaction = cur_trans;
|
||||
cur_trans->num_writers = 0;
|
||||
cur_trans->num_writers = 1;
|
||||
cur_trans->num_joined = 0;
|
||||
cur_trans->transid = root->fs_info->generation;
|
||||
init_waitqueue_head(&cur_trans->writer_wait);
|
||||
init_waitqueue_head(&cur_trans->commit_wait);
|
||||
|
@ -65,8 +66,11 @@ static int join_transaction(struct btrfs_root *root)
|
|||
cur_trans->start_time = get_seconds();
|
||||
list_add_tail(&cur_trans->list, &root->fs_info->trans_list);
|
||||
init_bit_radix(&cur_trans->dirty_pages);
|
||||
} else {
|
||||
cur_trans->num_writers++;
|
||||
cur_trans->num_joined++;
|
||||
}
|
||||
cur_trans->num_writers++;
|
||||
|
||||
return 0;
|
||||
}
|
||||
|
||||
|
@ -428,12 +432,14 @@ static int drop_dirty_roots(struct btrfs_root *tree_root,
|
|||
int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
|
||||
struct btrfs_root *root)
|
||||
{
|
||||
int ret = 0;
|
||||
unsigned long joined = 0;
|
||||
unsigned long timeout = 1;
|
||||
struct btrfs_transaction *cur_trans;
|
||||
struct btrfs_transaction *prev_trans = NULL;
|
||||
struct list_head dirty_fs_roots;
|
||||
struct radix_tree_root pinned_copy;
|
||||
DEFINE_WAIT(wait);
|
||||
int ret;
|
||||
|
||||
init_bit_radix(&pinned_copy);
|
||||
INIT_LIST_HEAD(&dirty_fs_roots);
|
||||
|
@ -448,7 +454,11 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
|
|||
mutex_unlock(&root->fs_info->fs_mutex);
|
||||
ret = wait_for_commit(root, cur_trans);
|
||||
BUG_ON(ret);
|
||||
|
||||
mutex_lock(&root->fs_info->trans_mutex);
|
||||
put_transaction(cur_trans);
|
||||
mutex_unlock(&root->fs_info->trans_mutex);
|
||||
|
||||
mutex_lock(&root->fs_info->fs_mutex);
|
||||
return 0;
|
||||
}
|
||||
|
@ -463,26 +473,35 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
|
|||
mutex_unlock(&root->fs_info->trans_mutex);
|
||||
|
||||
wait_for_commit(root, prev_trans);
|
||||
put_transaction(prev_trans);
|
||||
|
||||
mutex_lock(&root->fs_info->fs_mutex);
|
||||
mutex_lock(&root->fs_info->trans_mutex);
|
||||
put_transaction(prev_trans);
|
||||
}
|
||||
}
|
||||
while (trans->transaction->num_writers > 1) {
|
||||
|
||||
do {
|
||||
joined = cur_trans->num_joined;
|
||||
WARN_ON(cur_trans != trans->transaction);
|
||||
prepare_to_wait(&trans->transaction->writer_wait, &wait,
|
||||
prepare_to_wait(&cur_trans->writer_wait, &wait,
|
||||
TASK_UNINTERRUPTIBLE);
|
||||
if (trans->transaction->num_writers <= 1)
|
||||
break;
|
||||
|
||||
if (cur_trans->num_writers > 1)
|
||||
timeout = MAX_SCHEDULE_TIMEOUT;
|
||||
else
|
||||
timeout = 1;
|
||||
|
||||
mutex_unlock(&root->fs_info->fs_mutex);
|
||||
mutex_unlock(&root->fs_info->trans_mutex);
|
||||
schedule();
|
||||
|
||||
schedule_timeout(timeout);
|
||||
|
||||
mutex_lock(&root->fs_info->fs_mutex);
|
||||
mutex_lock(&root->fs_info->trans_mutex);
|
||||
finish_wait(&trans->transaction->writer_wait, &wait);
|
||||
}
|
||||
finish_wait(&trans->transaction->writer_wait, &wait);
|
||||
finish_wait(&cur_trans->writer_wait, &wait);
|
||||
} while (cur_trans->num_writers > 1 ||
|
||||
(cur_trans->num_joined != joined));
|
||||
|
||||
WARN_ON(cur_trans != trans->transaction);
|
||||
ret = add_dirty_roots(trans, &root->fs_info->fs_roots_radix,
|
||||
&dirty_fs_roots);
|
||||
|
@ -511,6 +530,7 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
|
|||
btrfs_finish_extent_commit(trans, root, &pinned_copy);
|
||||
mutex_lock(&root->fs_info->trans_mutex);
|
||||
cur_trans->commit_done = 1;
|
||||
root->fs_info->last_trans_committed = cur_trans->transid;
|
||||
wake_up(&cur_trans->commit_wait);
|
||||
put_transaction(cur_trans);
|
||||
put_transaction(cur_trans);
|
||||
|
|
|
@ -23,6 +23,7 @@
|
|||
struct btrfs_transaction {
|
||||
u64 transid;
|
||||
unsigned long num_writers;
|
||||
unsigned long num_joined;
|
||||
int in_commit;
|
||||
int use_count;
|
||||
int commit_done;
|
||||
|
@ -57,6 +58,12 @@ static inline void btrfs_update_inode_block_group(struct
|
|||
BTRFS_I(inode)->block_group = trans->block_group;
|
||||
}
|
||||
|
||||
static inline void btrfs_set_inode_last_trans(struct btrfs_trans_handle *trans,
|
||||
struct inode *inode)
|
||||
{
|
||||
BTRFS_I(inode)->last_trans = trans->transaction->transid;
|
||||
}
|
||||
|
||||
int btrfs_end_transaction(struct btrfs_trans_handle *trans,
|
||||
struct btrfs_root *root);
|
||||
struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root,
|
||||
|
|
Loading…
Reference in a new issue