btrfs: convert snapshot/nocow exlcusion to drew lock

This patch removes all haphazard code implementing nocow writers
exclusion from pending snapshot creation and switches to using the drew
lock to ensure this invariant still holds.

'Readers' are snapshot creators from create_snapshot and 'writers' are
nocow writers from buffered write path or btrfs_setsize. This locking
scheme allows for multiple snapshots to happen while any nocow writers
are blocked, since writes to page cache in the nocow path will make
snapshots inconsistent.

So for performance reasons we'd like to have the ability to run multiple
concurrent snapshots and also favors readers in this case. And in case
there aren't pending snapshots (which will be the majority of the cases)
we rely on the percpu's writers counter to avoid cacheline contention.

The main gain from using the drew lock is it's now a lot easier to
reason about the guarantees of the locking scheme and whether there is
some silent breakage lurking.

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

1 files changed, 10 insertions, 35 deletions

diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 770d469e1d9c..06819c41e4f4 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -1104,32 +1104,6 @@ void btrfs_clean_tree_block(struct extent_buffer *buf)
 	}
 }
 
-static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void)
-{
-	struct btrfs_subvolume_writers *writers;
-	int ret;
-
-	writers = kmalloc(sizeof(*writers), GFP_NOFS);
-	if (!writers)
-		return ERR_PTR(-ENOMEM);
-
-	ret = percpu_counter_init(&writers->counter, 0, GFP_NOFS);
-	if (ret < 0) {
-		kfree(writers);
-		return ERR_PTR(ret);
-	}
-
-	init_waitqueue_head(&writers->wait);
-	return writers;
-}
-
-static void
-btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers)
-{
-	percpu_counter_destroy(&writers->counter);
-	kfree(writers);
-}
-
 static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info,
 			 u64 objectid)
 {
@@ -1178,7 +1152,6 @@ static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info,
 	atomic_set(&root->log_writers, 0);
 	atomic_set(&root->log_batch, 0);
 	refcount_set(&root->refs, 1);
-	atomic_set(&root->will_be_snapshotted, 0);
 	atomic_set(&root->snapshot_force_cow, 0);
 	atomic_set(&root->nr_swapfiles, 0);
 	root->log_transid = 0;
@@ -1450,7 +1423,7 @@ alloc_fail:
 static int btrfs_init_fs_root(struct btrfs_root *root)
 {
 	int ret;
-	struct btrfs_subvolume_writers *writers;
+	unsigned int nofs_flag;
 
 	root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
 	root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
@@ -1460,12 +1433,15 @@ static int btrfs_init_fs_root(struct btrfs_root *root)
 		goto fail;
 	}
 
-	writers = btrfs_alloc_subvolume_writers();
-	if (IS_ERR(writers)) {
-		ret = PTR_ERR(writers);
+	/*
+	 * We might be called under a transaction (e.g. indirect backref
+	 * resolution) which could deadlock if it triggers memory reclaim
+	 */
+	nofs_flag = memalloc_nofs_save();
+	ret = btrfs_drew_lock_init(&root->snapshot_lock);
+	memalloc_nofs_restore(nofs_flag);
+	if (ret)
 		goto fail;
-	}
-	root->subv_writers = writers;
 
 	if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
 		set_bit(BTRFS_ROOT_REF_COWS, &root->state);
@@ -3961,8 +3937,7 @@ void btrfs_free_fs_root(struct btrfs_root *root)
 	WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
 	if (root->anon_dev)
 		free_anon_bdev(root->anon_dev);
-	if (root->subv_writers)
-		btrfs_free_subvolume_writers(root->subv_writers);
+	btrfs_drew_lock_destroy(&root->snapshot_lock);
 	free_extent_buffer(root->node);
 	free_extent_buffer(root->commit_root);
 	kfree(root->free_ino_ctl);