Btrfs: Add a thread pool just for submit_bio

If a bio submission is after a lock holder waiting for the bio
on the work queue, it is possible to deadlock.  Move the bios
into their own pool.

Signed-off-by: Chris Mason <chris.mason@oracle.com>

3 files changed, 10 insertions, 1 deletions

diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 181c81d29897..dcea9d706d9b 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -529,9 +529,13 @@ struct btrfs_fs_info {
 	 * can run with FS locks held, and the writers may be waiting for
 	 * those locks.  We don't want ordering in the pending list to cause
 	 * deadlocks, and so the two are serviced separately.
+	 *
+	 * A third pool does submit_bio to avoid deadlocking with the other
+	 * two
 	 */
 	struct btrfs_workers workers;
 	struct btrfs_workers endio_workers;
+	struct btrfs_workers submit_workers;
 	int thread_pool_size;
 
 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index c6a710a668cb..e5c758e306d5 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -1233,8 +1233,10 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	 * cannot dynamically grow.
 	 */
 	btrfs_init_workers(&fs_info->workers, fs_info->thread_pool_size);
+	btrfs_init_workers(&fs_info->submit_workers, fs_info->thread_pool_size);
 	btrfs_init_workers(&fs_info->endio_workers, fs_info->thread_pool_size);
 	btrfs_start_workers(&fs_info->workers, 1);
+	btrfs_start_workers(&fs_info->submit_workers, 1);
 	btrfs_start_workers(&fs_info->endio_workers, fs_info->thread_pool_size);
 
 
@@ -1343,6 +1345,7 @@ fail_sb_buffer:
 	extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree);
 	btrfs_stop_workers(&fs_info->workers);
 	btrfs_stop_workers(&fs_info->endio_workers);
+	btrfs_stop_workers(&fs_info->submit_workers);
 fail_iput:
 	iput(fs_info->btree_inode);
 fail:
@@ -1597,6 +1600,7 @@ int close_ctree(struct btrfs_root *root)
 
 	btrfs_stop_workers(&fs_info->workers);
 	btrfs_stop_workers(&fs_info->endio_workers);
+	btrfs_stop_workers(&fs_info->submit_workers);
 
 	iput(fs_info->btree_inode);
 #if 0
diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index c57458ce6339..ba3968571024 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -2112,7 +2112,8 @@ int schedule_bio(struct btrfs_root *root, struct btrfs_device *device,
 	spin_unlock(&device->io_lock);
 
 	if (should_queue)
-		btrfs_queue_worker(&root->fs_info->workers, &device->work);
+		btrfs_queue_worker(&root->fs_info->submit_workers,
+				   &device->work);
 	return 0;
 }