The 'flushoncommit' mount option forces any data dirtied by a write in a
prior transaction to commit as part of the current commit. This makes
the committed state a fully consistent view of the file system from the
application's perspective (i.e., it includes all completed file system
operations). This was previously the behavior only when a snapshot is
created.
While we're at it, make sync_fs also commit a consistent view (even
without 'flushoncommit') by moving the start_delalloc and
wait_ordered_extents into commit_transaction.
This is used by Ceph to ensure that completed writes make it to the
platter along with the metadata operations they are bound to (by
BTRFS_IOC_TRANS_{START,END}).
I'm not entirely sure why previously a snapshot creation didn't require
a start_delalloc_inodes but sync_fs did. I suspect that the call is
either also desirable if snap_pending in commit_transaction, or is not
needed by sync_fs either...?
Let me know if this looks reasonable, or if you would prefer a different
approach.
Thanks-
Signed-off-by: Sage Weil <sage@xxxxxxxxxxxx>
---
fs/btrfs/ctree.h | 1 +
fs/btrfs/disk-io.c | 6 +++---
fs/btrfs/extent-tree.c | 6 +++---
fs/btrfs/file.c | 4 ++--
fs/btrfs/inode.c | 2 +-
fs/btrfs/ioctl.c | 8 ++++----
fs/btrfs/super.c | 15 ++++++++-------
fs/btrfs/transaction.c | 12 +++++++++---
fs/btrfs/transaction.h | 3 ++-
fs/btrfs/tree-log.c | 2 +-
fs/btrfs/volumes.c | 4 ++--
11 files changed, 36 insertions(+), 27 deletions(-)
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 471fa67..019e7a7 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -951,6 +951,7 @@ struct btrfs_root {
#define BTRFS_MOUNT_DEGRADED (1 << 4)
#define BTRFS_MOUNT_COMPRESS (1 << 5)
#define BTRFS_MOUNT_NOTREELOG (1 << 6)
+#define BTRFS_MOUNT_FLUSHONCOMMIT (1 << 7)
#define btrfs_clear_opt(o, opt) ((o) &= ~BTRFS_MOUNT_##opt)
#define btrfs_set_opt(o, opt) ((o) |= BTRFS_MOUNT_##opt)
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 7feac5a..2d4e7c0 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -1443,7 +1443,7 @@ static int transaction_kthread(void *arg)
}
mutex_unlock(&root->fs_info->trans_mutex);
trans = btrfs_start_transaction(root, 1);
- ret = btrfs_commit_transaction(trans, root);
+ ret = btrfs_commit_transaction(trans, root, 0);
sleep:
wake_up_process(root->fs_info->cleaner_kthread);
mutex_unlock(&root->fs_info->transaction_kthread_mutex);
@@ -2192,11 +2192,11 @@ int btrfs_commit_super(struct btrfs_root *root)
btrfs_clean_old_snapshots(root);
mutex_unlock(&root->fs_info->cleaner_mutex);
trans = btrfs_start_transaction(root, 1);
- ret = btrfs_commit_transaction(trans, root);
+ ret = btrfs_commit_transaction(trans, root, 0);
BUG_ON(ret);
/* run commit again to drop the original snapshot */
trans = btrfs_start_transaction(root, 1);
- btrfs_commit_transaction(trans, root);
+ btrfs_commit_transaction(trans, root, 0);
ret = btrfs_write_and_wait_transaction(NULL, root);
BUG_ON(ret);
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 3b26f09..b06d857 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -5021,7 +5021,7 @@ int btrfs_cleanup_reloc_trees(struct btrfs_root *root)
if (found) {
trans = btrfs_start_transaction(root, 1);
BUG_ON(!trans);
- ret = btrfs_commit_transaction(trans, root);
+ ret = btrfs_commit_transaction(trans, root, 0);
BUG_ON(ret);
}
@@ -5642,7 +5642,7 @@ again:
cur_byte = key.objectid;
trans = btrfs_start_transaction(info->tree_root, 1);
- btrfs_commit_transaction(trans, info->tree_root);
+ btrfs_commit_transaction(trans, info->tree_root, 0);
mutex_lock(&root->fs_info->cleaner_mutex);
btrfs_clean_old_snapshots(info->tree_root);
@@ -5728,7 +5728,7 @@ next:
/* unpin extents in this range */
trans = btrfs_start_transaction(info->tree_root, 1);
- btrfs_commit_transaction(trans, info->tree_root);
+ btrfs_commit_transaction(trans, info->tree_root, 0);
spin_lock(&block_group->lock);
WARN_ON(block_group->pinned > 0);
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index 3e8023e..158963a 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -1160,7 +1160,7 @@ out_nolock:
btrfs_sync_log(trans, root);
btrfs_end_transaction(trans, root);
} else {
- btrfs_commit_transaction(trans, root);
+ btrfs_commit_transaction(trans, root, 0);
}
}
if (file->f_flags & O_DIRECT) {
@@ -1248,7 +1248,7 @@ int btrfs_sync_file(struct file *file, struct dentry *dentry, int datasync)
mutex_unlock(&file->f_dentry->d_inode->i_mutex);
if (ret > 0) {
- ret = btrfs_commit_transaction(trans, root);
+ ret = btrfs_commit_transaction(trans, root, 0);
} else {
btrfs_sync_log(trans, root);
ret = btrfs_end_transaction(trans, root);
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 288c2cd..553278c 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -3285,7 +3285,7 @@ int btrfs_write_inode(struct inode *inode, int wait)
if (wait) {
trans = btrfs_join_transaction(root, 1);
btrfs_set_trans_block_group(trans, inode);
- ret = btrfs_commit_transaction(trans, root);
+ ret = btrfs_commit_transaction(trans, root, 0);
}
return ret;
}
diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
index 988fdc8..f793814 100644
--- a/fs/btrfs/ioctl.c
+++ b/fs/btrfs/ioctl.c
@@ -166,7 +166,7 @@ static noinline int create_subvol(struct btrfs_root *root,
BUG_ON(ret);
- ret = btrfs_commit_transaction(trans, root);
+ ret = btrfs_commit_transaction(trans, root, 0);
if (ret)
goto fail_commit;
@@ -183,7 +183,7 @@ static noinline int create_subvol(struct btrfs_root *root,
fail:
nr = trans->blocks_used;
- err = btrfs_commit_transaction(trans, new_root);
+ err = btrfs_commit_transaction(trans, new_root, 0);
if (err && !ret)
ret = err;
fail_commit:
@@ -226,7 +226,7 @@ static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
pending_snapshot->root = root;
list_add(&pending_snapshot->list,
&trans->transaction->pending_snapshots);
- err = btrfs_commit_transaction(trans, root);
+ err = btrfs_commit_transaction(trans, root, 0);
fail_unlock:
btrfs_btree_balance_dirty(root, nr);
@@ -538,7 +538,7 @@ static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg)
if (new_size > old_size) {
trans = btrfs_start_transaction(root, 1);
ret = btrfs_grow_device(trans, device, new_size);
- btrfs_commit_transaction(trans, root);
+ btrfs_commit_transaction(trans, root, 0);
} else {
ret = btrfs_shrink_device(device, new_size);
}
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index d8c664c..4c9f661 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -67,7 +67,7 @@ enum {
Opt_degraded, Opt_subvol, Opt_device, Opt_nodatasum, Opt_nodatacow,
Opt_max_extent, Opt_max_inline, Opt_alloc_start, Opt_nobarrier,
Opt_ssd, Opt_thread_pool, Opt_noacl, Opt_compress, Opt_notreelog,
- Opt_err,
+ Opt_flushoncommit, Opt_err,
};
static match_table_t tokens = {
@@ -85,6 +85,7 @@ static match_table_t tokens = {
{Opt_ssd, "ssd"},
{Opt_noacl, "noacl"},
{Opt_notreelog, "notreelog"},
+ {Opt_flushoncommit, "flushoncommit"},
{Opt_err, NULL},
};
@@ -228,6 +229,10 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
printk(KERN_INFO "btrfs: disabling tree log\n");
btrfs_set_opt(info->mount_opt, NOTREELOG);
break;
+ case Opt_flushoncommit:
+ printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
+ btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
+ break;
default:
break;
}
@@ -369,9 +374,8 @@ fail_close:
int btrfs_sync_fs(struct super_block *sb, int wait)
{
struct btrfs_trans_handle *trans;
- struct btrfs_root *root;
+ struct btrfs_root *root = btrfs_sb(sb);
int ret;
- root = btrfs_sb(sb);
if (sb->s_flags & MS_RDONLY)
return 0;
@@ -382,12 +386,9 @@ int btrfs_sync_fs(struct super_block *sb, int wait)
return 0;
}
- btrfs_start_delalloc_inodes(root);
- btrfs_wait_ordered_extents(root, 0);
-
btrfs_clean_old_snapshots(root);
trans = btrfs_start_transaction(root, 1);
- ret = btrfs_commit_transaction(trans, root);
+ ret = btrfs_commit_transaction(trans, root, 1);
sb->s_dirt = 0;
return ret;
}
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 919172d..f687e66 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -881,7 +881,8 @@ static noinline int finish_pending_snapshots(struct btrfs_trans_handle *trans,
}
int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
- struct btrfs_root *root)
+ struct btrfs_root *root,
+ int ordered)
{
unsigned long joined = 0;
unsigned long timeout = 1;
@@ -893,6 +894,9 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
DEFINE_WAIT(wait);
int ret;
+ if (btrfs_test_opt(root, FLUSHONCOMMIT))
+ ordered = 1;
+
INIT_LIST_HEAD(&dirty_fs_roots);
mutex_lock(&root->fs_info->trans_mutex);
if (trans->transaction->in_commit) {
@@ -951,8 +955,10 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
timeout = 1;
mutex_unlock(&root->fs_info->trans_mutex);
-
- if (snap_pending) {
+
+ if (ordered || snap_pending) {
+ if (ordered)
+ ret = btrfs_start_delalloc_inodes(root);
ret = btrfs_wait_ordered_extents(root, 1);
BUG_ON(ret);
}
diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h
index ea29211..e167b70 100644
--- a/fs/btrfs/transaction.h
+++ b/fs/btrfs/transaction.h
@@ -96,7 +96,8 @@ int btrfs_add_dead_root(struct btrfs_root *root, struct btrfs_root *latest);
int btrfs_defrag_root(struct btrfs_root *root, int cacheonly);
int btrfs_clean_old_snapshots(struct btrfs_root *root);
int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
- struct btrfs_root *root);
+ struct btrfs_root *root,
+ int ordered);
int btrfs_end_transaction_throttle(struct btrfs_trans_handle *trans,
struct btrfs_root *root);
void btrfs_throttle(struct btrfs_root *root);
diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c
index ac58991..b01d6c2 100644
--- a/fs/btrfs/tree-log.c
+++ b/fs/btrfs/tree-log.c
@@ -2877,7 +2877,7 @@ again:
fs_info->log_root_recovering = 0;
/* step 4: commit the transaction, which also unpins the blocks */
- btrfs_commit_transaction(trans, fs_info->tree_root);
+ btrfs_commit_transaction(trans, fs_info->tree_root, 0);
kfree(log_root_tree);
return 0;
diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index fd0bedb..6cfec73 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -964,7 +964,7 @@ static int btrfs_rm_dev_item(struct btrfs_root *root,
out:
btrfs_free_path(path);
unlock_chunks(root);
- btrfs_commit_transaction(trans, root);
+ btrfs_commit_transaction(trans, root, 0);
return ret;
}
@@ -1368,7 +1368,7 @@ int btrfs_init_new_device(struct btrfs_root *root, char *device_path)
}
unlock_chunks(root);
- btrfs_commit_transaction(trans, root);
+ btrfs_commit_transaction(trans, root, 0);
if (seeding_dev) {
mutex_unlock(&uuid_mutex);
--
1.5.6.5
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
