Btrfs: cleanup transaction on abort
authorJosef Bacik <jbacik@fusionio.com>
Fri, 27 Sep 2013 20:32:39 +0000 (16:32 -0400)
committerChris Mason <chris.mason@fusionio.com>
Tue, 12 Nov 2013 02:53:42 +0000 (21:53 -0500)
If we abort not during a transaction commit we won't clean up anything until we
unmount.  Unfortunately if we abort in the middle of writing out an ordered
extent we won't clean it up and if somebody is waiting on that ordered extent
they will wait forever.  To fix this just make the transaction kthread call the
cleanup transaction stuff if it notices theres an error, and make
btrfs_end_transaction wake up the transaction kthread if there is an error.
Thanks,

Signed-off-by: Josef Bacik <jbacik@fusionio.com>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
fs/btrfs/disk-io.c
fs/btrfs/transaction.c

index db2095486a4fb46081bdfc453f32d54181a62611..113cd43530bacc0b35c501ed1e0ef606fda18924 100644 (file)
@@ -1800,6 +1800,9 @@ sleep:
                wake_up_process(root->fs_info->cleaner_kthread);
                mutex_unlock(&root->fs_info->transaction_kthread_mutex);
 
+               if (unlikely(test_bit(BTRFS_FS_STATE_ERROR,
+                                     &root->fs_info->fs_state)))
+                       btrfs_cleanup_transaction(root);
                if (!try_to_freeze()) {
                        set_current_state(TASK_INTERRUPTIBLE);
                        if (!kthread_should_stop() &&
index 7138d6a3726e6d63df6def0107968a5800d92182..a1343e81c59df95656da812dabf4a228f696b6c8 100644 (file)
@@ -744,8 +744,10 @@ static int __btrfs_end_transaction(struct btrfs_trans_handle *trans,
                btrfs_run_delayed_iputs(root);
 
        if (trans->aborted ||
-           test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state))
+           test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) {
+               wake_up_process(info->transaction_kthread);
                err = -EIO;
+       }
        assert_qgroups_uptodate(trans);
 
        kmem_cache_free(btrfs_trans_handle_cachep, trans);