diff --git a/drivers/md/bcache/bset.c b/drivers/md/bcache/bset.c
index fae5b7b..f7b5525 100644
--- a/drivers/md/bcache/bset.c
+++ b/drivers/md/bcache/bset.c
@@ -1196,7 +1196,7 @@
 	int ret;
 
 	memset(&t, 0, sizeof(struct bset_stats));
-	bch_btree_op_init_stack(&t.op);
+	bch_btree_op_init(&t.op, -1);
 
 	ret = bch_btree_map_nodes(&t.op, c, &ZERO_KEY, btree_bset_stats);
 	if (ret < 0)
diff --git a/drivers/md/bcache/btree.c b/drivers/md/bcache/btree.c
index 5cb59c3..cb1a490 100644
--- a/drivers/md/bcache/btree.c
+++ b/drivers/md/bcache/btree.c
@@ -115,13 +115,6 @@
 
 static struct workqueue_struct *btree_io_wq;
 
-void bch_btree_op_init_stack(struct btree_op *op)
-{
-	memset(op, 0, sizeof(struct btree_op));
-	closure_init_stack(&op->cl);
-	op->lock = -1;
-}
-
 static inline bool should_split(struct btree *b)
 {
 	struct bset *i = write_block(b);
@@ -965,8 +958,7 @@
  * bch_btree_node_get - find a btree node in the cache and lock it, reading it
  * in from disk if necessary.
  *
- * If IO is necessary, it uses the closure embedded in struct btree_op to wait;
- * if that closure is in non blocking mode, will return -EAGAIN.
+ * If IO is necessary and running under generic_make_request, returns -EAGAIN.
  *
  * The btree node will have either a read or a write lock held, depending on
  * level and op->lock.
@@ -1260,6 +1252,9 @@
 {
 	unsigned nodes = 0, keys = 0, blocks;
 	int i;
+	struct closure cl;
+
+	closure_init_stack(&cl);
 
 	while (nodes < GC_MERGE_NODES && r[nodes].b)
 		keys += r[nodes++].keys;
@@ -1353,9 +1348,7 @@
 {
 	void write(struct btree *r)
 	{
-		if (!r->written)
-			bch_btree_node_write(r, &op->cl);
-		else if (btree_node_dirty(r))
+		if (!r->written || btree_node_dirty(r))
 			bch_btree_node_write(r, writes);
 
 		up_write(&r->lock);
@@ -1431,6 +1424,9 @@
 	struct btree *n = NULL;
 	unsigned keys = 0;
 	int ret = 0, stale = btree_gc_mark_node(b, &keys, gc);
+	struct closure cl;
+
+	closure_init_stack(&cl);
 
 	if (b->level || stale > 10)
 		n = btree_node_alloc_replacement(b);
@@ -1442,11 +1438,11 @@
 		ret = btree_gc_recurse(b, op, writes, gc);
 
 	if (!b->written || btree_node_dirty(b)) {
-		bch_btree_node_write(b, n ? &op->cl : NULL);
+		bch_btree_node_write(b, n ? &cl : NULL);
 	}
 
 	if (!IS_ERR_OR_NULL(n)) {
-		closure_sync(&op->cl);
+		closure_sync(&cl);
 		bch_btree_set_root(b);
 		btree_node_free(n);
 		rw_unlock(true, b);
@@ -1545,15 +1541,13 @@
 
 	memset(&stats, 0, sizeof(struct gc_stat));
 	closure_init_stack(&writes);
-	bch_btree_op_init_stack(&op);
-	op.lock = SHRT_MAX;
+	bch_btree_op_init(&op, SHRT_MAX);
 
 	btree_gc_start(c);
 
 	atomic_inc(&c->prio_blocked);
 
 	ret = btree_root(gc_root, c, &op, &writes, &stats);
-	closure_sync(&op.cl);
 	closure_sync(&writes);
 
 	if (ret) {
@@ -1562,8 +1556,8 @@
 	}
 
 	/* Possibly wait for new UUIDs or whatever to hit disk */
-	bch_journal_meta(c, &op.cl);
-	closure_sync(&op.cl);
+	bch_journal_meta(c, &writes);
+	closure_sync(&writes);
 
 	available = bch_btree_gc_finish(c);
 
@@ -1671,8 +1665,7 @@
 	struct btree_op op;
 
 	memset(seen, 0, sizeof(seen));
-	bch_btree_op_init_stack(&op);
-	op.lock = SHRT_MAX;
+	bch_btree_op_init(&op, SHRT_MAX);
 
 	for (i = 0; c->cache[i]; i++) {
 		size_t n = DIV_ROUND_UP(c->cache[i]->sb.nbuckets, 8);
@@ -1980,6 +1973,9 @@
 	bool split;
 	struct btree *n1, *n2 = NULL, *n3 = NULL;
 	uint64_t start_time = local_clock();
+	struct closure cl;
+
+	closure_init_stack(&cl);
 
 	n1 = btree_node_alloc_replacement(b);
 	if (IS_ERR(n1))
@@ -2025,7 +2021,7 @@
 		bkey_copy_key(&n2->key, &b->key);
 
 		bch_keylist_add(parent_keys, &n2->key);
-		bch_btree_node_write(n2, &op->cl);
+		bch_btree_node_write(n2, &cl);
 		rw_unlock(true, n2);
 	} else {
 		trace_bcache_btree_node_compact(b, n1->sets[0].data->keys);
@@ -2034,23 +2030,23 @@
 	}
 
 	bch_keylist_add(parent_keys, &n1->key);
-	bch_btree_node_write(n1, &op->cl);
+	bch_btree_node_write(n1, &cl);
 
 	if (n3) {
 		/* Depth increases, make a new root */
 
 		bkey_copy_key(&n3->key, &MAX_KEY);
 		bch_btree_insert_keys(n3, op, parent_keys);
-		bch_btree_node_write(n3, &op->cl);
+		bch_btree_node_write(n3, &cl);
 
-		closure_sync(&op->cl);
+		closure_sync(&cl);
 		bch_btree_set_root(n3);
 		rw_unlock(true, n3);
 	} else if (!b->parent) {
 		/* Root filled up but didn't need to be split */
 
 		bch_keylist_reset(parent_keys);
-		closure_sync(&op->cl);
+		closure_sync(&cl);
 		bch_btree_set_root(n1);
 	} else {
 		unsigned i;
@@ -2065,7 +2061,7 @@
 		}
 
 		bch_keylist_push(parent_keys);
-		closure_sync(&op->cl);
+		closure_sync(&cl);
 		atomic_inc(&b->c->prio_blocked);
 	}
 
@@ -2126,10 +2122,15 @@
 			BUG_ON(write_block(b) != b->sets[b->nsets].data);
 
 			if (bch_btree_insert_keys(b, op, insert_keys)) {
-				if (!b->level)
+				if (!b->level) {
 					bch_btree_leaf_dirty(b, journal_ref);
-				else
-					bch_btree_node_write(b, &op->cl);
+				} else {
+					struct closure cl;
+
+					closure_init_stack(&cl);
+					bch_btree_node_write(b, &cl);
+					closure_sync(&cl);
+				}
 			}
 		}
 	} while (!bch_keylist_empty(&split_keys));
@@ -2204,12 +2205,6 @@
 {
 	int ret = 0;
 
-	/*
-	 * Don't want to block with the btree locked unless we have to,
-	 * otherwise we get deadlocks with try_harder and between split/gc
-	 */
-	clear_closure_blocking(&op->cl);
-
 	BUG_ON(bch_keylist_empty(keys));
 
 	while (!bch_keylist_empty(keys)) {
@@ -2217,8 +2212,8 @@
 		ret = btree_root(insert_recurse, c, op, keys, journal_ref);
 
 		if (ret == -EAGAIN) {
+			BUG();
 			ret = 0;
-			closure_sync(&op->cl);
 		} else if (ret) {
 			struct bkey *k;
 
@@ -2292,10 +2287,7 @@
 int __bch_btree_map_nodes(struct btree_op *op, struct cache_set *c,
 			  struct bkey *from, btree_map_nodes_fn *fn, int flags)
 {
-	int ret = btree_root(map_nodes_recurse, c, op, from, fn, flags);
-	if (closure_blocking(&op->cl))
-		closure_sync(&op->cl);
-	return ret;
+	return btree_root(map_nodes_recurse, c, op, from, fn, flags);
 }
 
 static int bch_btree_map_keys_recurse(struct btree *b, struct btree_op *op,
@@ -2328,10 +2320,7 @@
 int bch_btree_map_keys(struct btree_op *op, struct cache_set *c,
 		       struct bkey *from, btree_map_keys_fn *fn, int flags)
 {
-	int ret = btree_root(map_keys_recurse, c, op, from, fn, flags);
-	if (closure_blocking(&op->cl))
-		closure_sync(&op->cl);
-	return ret;
+	return btree_root(map_keys_recurse, c, op, from, fn, flags);
 }
 
 /* Keybuf code */
@@ -2409,7 +2398,7 @@
 
 	cond_resched();
 
-	bch_btree_op_init_stack(&refill.op);
+	bch_btree_op_init(&refill.op, -1);
 	refill.buf = buf;
 	refill.end = end;
 	refill.pred = pred;
diff --git a/drivers/md/bcache/btree.h b/drivers/md/bcache/btree.h
index 3f820b6..34ee535 100644
--- a/drivers/md/bcache/btree.h
+++ b/drivers/md/bcache/btree.h
@@ -237,8 +237,6 @@
 /* Recursing down the btree */
 
 struct btree_op {
-	struct closure		cl;
-
 	/* Btree level at which we start taking write locks */
 	short			lock;
 
@@ -253,7 +251,11 @@
 	BKEY_PADDED(replace);
 };
 
-void bch_btree_op_init_stack(struct btree_op *);
+static inline void bch_btree_op_init(struct btree_op *op, int write_lock_level)
+{
+	memset(op, 0, sizeof(struct btree_op));
+	op->lock = write_lock_level;
+}
 
 static inline void rw_lock(bool w, struct btree *b, int level)
 {
diff --git a/drivers/md/bcache/journal.c b/drivers/md/bcache/journal.c
index 725c8eb..20e900a 100644
--- a/drivers/md/bcache/journal.c
+++ b/drivers/md/bcache/journal.c
@@ -305,8 +305,7 @@
 	struct btree_op op;
 
 	bch_keylist_init(&keylist);
-	bch_btree_op_init_stack(&op);
-	op.lock = SHRT_MAX;
+	bch_btree_op_init(&op, SHRT_MAX);
 
 	list_for_each_entry(i, list, list) {
 		BUG_ON(i->pin && atomic_read(i->pin) != 1);
@@ -341,14 +340,13 @@
 
 	pr_info("journal replay done, %i keys in %i entries, seq %llu",
 		keys, entries, end);
-
+err:
 	while (!list_empty(list)) {
 		i = list_first_entry(list, struct journal_replay, list);
 		list_del(&i->list);
 		kfree(i);
 	}
-err:
-	closure_sync(&op.cl);
+
 	return ret;
 }
 
diff --git a/drivers/md/bcache/movinggc.c b/drivers/md/bcache/movinggc.c
index 80e30d7..219356f 100644
--- a/drivers/md/bcache/movinggc.c
+++ b/drivers/md/bcache/movinggc.c
@@ -108,8 +108,8 @@
 		s->op.type = BTREE_REPLACE;
 		bkey_copy(&s->op.replace, &io->w->key);
 
-		closure_init(&s->op.cl, cl);
-		bch_data_insert(&s->op.cl);
+		closure_init(&s->btree, cl);
+		bch_data_insert(&s->btree);
 	}
 
 	continue_at(cl, write_moving_finish, system_wq);
diff --git a/drivers/md/bcache/request.c b/drivers/md/bcache/request.c
index 818e2e3..5df44fb 100644
--- a/drivers/md/bcache/request.c
+++ b/drivers/md/bcache/request.c
@@ -215,8 +215,7 @@
 
 static void bch_data_insert_keys(struct closure *cl)
 {
-	struct btree_op *op = container_of(cl, struct btree_op, cl);
-	struct search *s = container_of(op, struct search, op);
+	struct search *s = container_of(cl, struct search, btree);
 	atomic_t *journal_ref = NULL;
 
 	/*
@@ -236,7 +235,7 @@
 					  s->flush_journal
 					  ? &s->cl : NULL);
 
-	if (bch_btree_insert(op, s->c, &s->insert_keys, journal_ref)) {
+	if (bch_btree_insert(&s->op, s->c, &s->insert_keys, journal_ref)) {
 		s->error		= -ENOMEM;
 		s->insert_data_done	= true;
 	}
@@ -433,8 +432,7 @@
 
 static void bch_data_invalidate(struct closure *cl)
 {
-	struct btree_op *op = container_of(cl, struct btree_op, cl);
-	struct search *s = container_of(op, struct search, op);
+	struct search *s = container_of(cl, struct search, btree);
 	struct bio *bio = s->cache_bio;
 
 	pr_debug("invalidating %i sectors from %llu",
@@ -461,8 +459,7 @@
 
 static void bch_data_insert_error(struct closure *cl)
 {
-	struct btree_op *op = container_of(cl, struct btree_op, cl);
-	struct search *s = container_of(op, struct search, op);
+	struct search *s = container_of(cl, struct search, btree);
 
 	/*
 	 * Our data write just errored, which means we've got a bunch of keys to
@@ -493,8 +490,7 @@
 static void bch_data_insert_endio(struct bio *bio, int error)
 {
 	struct closure *cl = bio->bi_private;
-	struct btree_op *op = container_of(cl, struct btree_op, cl);
-	struct search *s = container_of(op, struct search, op);
+	struct search *s = container_of(cl, struct search, btree);
 
 	if (error) {
 		/* TODO: We could try to recover from this. */
@@ -511,8 +507,7 @@
 
 static void bch_data_insert_start(struct closure *cl)
 {
-	struct btree_op *op = container_of(cl, struct btree_op, cl);
-	struct search *s = container_of(op, struct search, op);
+	struct search *s = container_of(cl, struct search, btree);
 	struct bio *bio = s->cache_bio, *n;
 
 	if (s->bypass)
@@ -630,8 +625,7 @@
  */
 void bch_data_insert(struct closure *cl)
 {
-	struct btree_op *op = container_of(cl, struct btree_op, cl);
-	struct search *s = container_of(op, struct search, op);
+	struct search *s = container_of(cl, struct search, btree);
 
 	bch_keylist_init(&s->insert_keys);
 	bio_get(s->cache_bio);
@@ -731,11 +725,10 @@
 
 static void cache_lookup(struct closure *cl)
 {
-	struct btree_op *op = container_of(cl, struct btree_op, cl);
-	struct search *s = container_of(op, struct search, op);
+	struct search *s = container_of(cl, struct search, btree);
 	struct bio *bio = &s->bio.bio;
 
-	int ret = bch_btree_map_keys(op, s->c,
+	int ret = bch_btree_map_keys(&s->op, s->c,
 				     &KEY(s->inode, bio->bi_sector, 0),
 				     cache_lookup_fn, MAP_END_KEY);
 	if (ret == -EAGAIN)
@@ -1064,7 +1057,7 @@
 	if (s->cache_bio &&
 	    !test_bit(CACHE_SET_STOPPING, &s->c->flags)) {
 		s->op.type = BTREE_REPLACE;
-		closure_call(&s->op.cl, bch_data_insert, NULL, cl);
+		closure_call(&s->btree, bch_data_insert, NULL, cl);
 	}
 
 	continue_at(cl, cached_dev_cache_miss_done, NULL);
@@ -1156,7 +1149,7 @@
 {
 	struct closure *cl = &s->cl;
 
-	closure_call(&s->op.cl, cache_lookup, NULL, cl);
+	closure_call(&s->btree, cache_lookup, NULL, cl);
 	continue_at(cl, cached_dev_read_done_bh, NULL);
 }
 
@@ -1239,7 +1232,7 @@
 		closure_bio_submit(bio, cl, s->d);
 	}
 
-	closure_call(&s->op.cl, bch_data_insert, NULL, cl);
+	closure_call(&s->btree, bch_data_insert, NULL, cl);
 	continue_at(cl, cached_dev_write_complete, NULL);
 }
 
@@ -1418,9 +1411,9 @@
 		s->writeback	= true;
 		s->cache_bio	= bio;
 
-		closure_call(&s->op.cl, bch_data_insert, NULL, cl);
+		closure_call(&s->btree, bch_data_insert, NULL, cl);
 	} else {
-		closure_call(&s->op.cl, cache_lookup, NULL, cl);
+		closure_call(&s->btree, cache_lookup, NULL, cl);
 	}
 
 	continue_at(cl, search_free, NULL);
diff --git a/drivers/md/bcache/request.h b/drivers/md/bcache/request.h
index 0f79177..ed578aa 100644
--- a/drivers/md/bcache/request.h
+++ b/drivers/md/bcache/request.h
@@ -6,6 +6,7 @@
 struct search {
 	/* Stack frame for bio_complete */
 	struct closure		cl;
+	struct closure		btree;
 
 	struct bcache_device	*d;
 	struct cache_set	*c;
diff --git a/drivers/md/bcache/writeback.c b/drivers/md/bcache/writeback.c
index b58c2bc..d0968e8 100644
--- a/drivers/md/bcache/writeback.c
+++ b/drivers/md/bcache/writeback.c
@@ -143,7 +143,7 @@
 		struct btree_op op;
 		struct keylist keys;
 
-		bch_btree_op_init_stack(&op);
+		bch_btree_op_init(&op, -1);
 		bch_keylist_init(&keys);
 
 		op.type = BTREE_REPLACE;
@@ -156,7 +156,6 @@
 			atomic_inc(&PTR_BUCKET(dc->disk.c, &w->key, i)->pin);
 
 		bch_btree_insert(&op, dc->disk.c, &keys, NULL);
-		closure_sync(&op.cl);
 
 		if (op.insert_collision)
 			trace_bcache_writeback_collision(&w->key);
@@ -457,7 +456,7 @@
 {
 	struct sectors_dirty_init op;
 
-	bch_btree_op_init_stack(&op.op);
+	bch_btree_op_init(&op.op, -1);
 	op.inode = dc->disk.id;
 
 	bch_btree_map_keys(&op.op, dc->disk.c, &KEY(op.inode, 0, 0),