Btrfs: Change btree locking to use explicit blocking points

Most of the btrfs metadata operations can be protected by a spinlock,
but some operations still need to schedule.

So far, btrfs has been using a mutex along with a trylock loop,
most of the time it is able to avoid going for the full mutex, so
the trylock loop is a big performance gain.

This commit is step one for getting rid of the blocking locks entirely.
btrfs_tree_lock takes a spinlock, and the code explicitly switches
to a blocking lock when it starts an operation that can schedule.

We'll be able get rid of the blocking locks in smaller pieces over time.
Tracing allows us to find the most common cause of blocking, so we
can start with the hot spots first.

The basic idea is:

btrfs_tree_lock() returns with the spin lock held

btrfs_set_lock_blocking() sets the EXTENT_BUFFER_BLOCKING bit in
the extent buffer flags, and then drops the spin lock.  The buffer is
still considered locked by all of the btrfs code.

If btrfs_tree_lock gets the spinlock but finds the blocking bit set, it drops
the spin lock and waits on a wait queue for the blocking bit to go away.

Much of the code that needs to set the blocking bit finishes without actually
blocking a good percentage of the time.  So, an adaptive spin is still
used against the blocking bit to avoid very high context switch rates.

btrfs_clear_lock_blocking() clears the blocking bit and returns
with the spinlock held again.

btrfs_tree_unlock() can be called on either blocking or spinning locks,
it does the right thing based on the blocking bit.

ctree.c has a helper function to set/clear all the locked buffers in a
path as blocking.

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

1 files changed, 226 insertions, 8 deletions

diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index 3b6e35aafc9..3af777357ac 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -54,6 +54,31 @@ struct btrfs_path *btrfs_alloc_path(void)
 	return path;
 }
 
+/*
+ * set all locked nodes in the path to blocking locks.  This should
+ * be done before scheduling
+ */
+noinline void btrfs_set_path_blocking(struct btrfs_path *p)
+{
+	int i;
+	for (i = 0; i < BTRFS_MAX_LEVEL; i++) {
+		if (p->nodes[i] && p->locks[i])
+			btrfs_set_lock_blocking(p->nodes[i]);
+	}
+}
+
+/*
+ * reset all the locked nodes in the patch to spinning locks.
+ */
+noinline void btrfs_clear_path_blocking(struct btrfs_path *p)
+{
+	int i;
+	for (i = 0; i < BTRFS_MAX_LEVEL; i++) {
+		if (p->nodes[i] && p->locks[i])
+			btrfs_clear_lock_blocking(p->nodes[i]);
+	}
+}
+
 /* this also releases the path */
 void btrfs_free_path(struct btrfs_path *p)
 {
@@ -272,6 +297,8 @@ static noinline int __btrfs_cow_block(struct btrfs_trans_handle *trans,
 	if (IS_ERR(cow))
 		return PTR_ERR(cow);
 
+	/* cow is set to blocking by btrfs_init_new_buffer */
+
 	copy_extent_buffer(cow, buf, 0, 0, cow->len);
 	btrfs_set_header_bytenr(cow, cow->start);
 	btrfs_set_header_generation(cow, trans->transid);
@@ -397,6 +424,11 @@ noinline int btrfs_cow_block(struct btrfs_trans_handle *trans,
 	}
 
 	search_start = buf->start & ~((u64)(1024 * 1024 * 1024) - 1);
+
+	if (parent)
+		btrfs_set_lock_blocking(parent);
+	btrfs_set_lock_blocking(buf);
+
 	ret = __btrfs_cow_block(trans, root, buf, parent,
 				 parent_slot, cow_ret, search_start, 0,
 				 prealloc_dest);
@@ -502,6 +534,8 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 	if (parent_nritems == 1)
 		return 0;
 
+	btrfs_set_lock_blocking(parent);
+
 	for (i = start_slot; i < end_slot; i++) {
 		int close = 1;
 
@@ -562,6 +596,7 @@ int btrfs_realloc_node(struct btrfs_trans_handle *trans,
 			search_start = last_block;
 
 		btrfs_tree_lock(cur);
+		btrfs_set_lock_blocking(cur);
 		err = __btrfs_cow_block(trans, root, cur, parent, i,
 					&cur, search_start,
 					min(16 * blocksize,
@@ -860,6 +895,7 @@ static noinline int balance_level(struct btrfs_trans_handle *trans,
 		return 0;
 
 	mid = path->nodes[level];
+
 	WARN_ON(!path->locks[level]);
 	WARN_ON(btrfs_header_generation(mid) != trans->transid);
 
@@ -882,6 +918,7 @@ static noinline int balance_level(struct btrfs_trans_handle *trans,
 		/* promote the child to a root */
 		child = read_node_slot(root, mid, 0);
 		btrfs_tree_lock(child);
+		btrfs_set_lock_blocking(child);
 		BUG_ON(!child);
 		ret = btrfs_cow_block(trans, root, child, mid, 0, &child, 0);
 		BUG_ON(ret);
@@ -898,6 +935,7 @@ static noinline int balance_level(struct btrfs_trans_handle *trans,
 
 		add_root_to_dirty_list(root);
 		btrfs_tree_unlock(child);
+
 		path->locks[level] = 0;
 		path->nodes[level] = NULL;
 		clean_tree_block(trans, root, mid);
@@ -922,6 +960,7 @@ static noinline int balance_level(struct btrfs_trans_handle *trans,
 	left = read_node_slot(root, parent, pslot - 1);
 	if (left) {
 		btrfs_tree_lock(left);
+		btrfs_set_lock_blocking(left);
 		wret = btrfs_cow_block(trans, root, left,
 				       parent, pslot - 1, &left, 0);
 		if (wret) {
@@ -932,6 +971,7 @@ static noinline int balance_level(struct btrfs_trans_handle *trans,
 	right = read_node_slot(root, parent, pslot + 1);
 	if (right) {
 		btrfs_tree_lock(right);
+		btrfs_set_lock_blocking(right);
 		wret = btrfs_cow_block(trans, root, right,
 				       parent, pslot + 1, &right, 0);
 		if (wret) {
@@ -1107,6 +1147,8 @@ static noinline int push_nodes_for_insert(struct btrfs_trans_handle *trans,
 		u32 left_nr;
 
 		btrfs_tree_lock(left);
+		btrfs_set_lock_blocking(left);
+
 		left_nr = btrfs_header_nritems(left);
 		if (left_nr >= BTRFS_NODEPTRS_PER_BLOCK(root) - 1) {
 			wret = 1;
@@ -1153,7 +1195,10 @@ static noinline int push_nodes_for_insert(struct btrfs_trans_handle *trans,
 	 */
 	if (right) {
 		u32 right_nr;
+
 		btrfs_tree_lock(right);
+		btrfs_set_lock_blocking(right);
+
 		right_nr = btrfs_header_nritems(right);
 		if (right_nr >= BTRFS_NODEPTRS_PER_BLOCK(root) - 1) {
 			wret = 1;
@@ -1265,6 +1310,68 @@ static noinline void reada_for_search(struct btrfs_root *root,
 }
 
 /*
+ * returns -EAGAIN if it had to drop the path, or zero if everything was in
+ * cache
+ */
+static noinline int reada_for_balance(struct btrfs_root *root,
+				      struct btrfs_path *path, int level)
+{
+	int slot;
+	int nritems;
+	struct extent_buffer *parent;
+	struct extent_buffer *eb;
+	u64 gen;
+	u64 block1 = 0;
+	u64 block2 = 0;
+	int ret = 0;
+	int blocksize;
+
+	parent = path->nodes[level - 1];
+	if (!parent)
+		return 0;
+
+	nritems = btrfs_header_nritems(parent);
+	slot = path->slots[level];
+	blocksize = btrfs_level_size(root, level);
+
+	if (slot > 0) {
+		block1 = btrfs_node_blockptr(parent, slot - 1);
+		gen = btrfs_node_ptr_generation(parent, slot - 1);
+		eb = btrfs_find_tree_block(root, block1, blocksize);
+		if (eb && btrfs_buffer_uptodate(eb, gen))
+			block1 = 0;
+		free_extent_buffer(eb);
+	}
+	if (slot < nritems) {
+		block2 = btrfs_node_blockptr(parent, slot + 1);
+		gen = btrfs_node_ptr_generation(parent, slot + 1);
+		eb = btrfs_find_tree_block(root, block2, blocksize);
+		if (eb && btrfs_buffer_uptodate(eb, gen))
+			block2 = 0;
+		free_extent_buffer(eb);
+	}
+	if (block1 || block2) {
+		ret = -EAGAIN;
+		btrfs_release_path(root, path);
+		if (block1)
+			readahead_tree_block(root, block1, blocksize, 0);
+		if (block2)
+			readahead_tree_block(root, block2, blocksize, 0);
+
+		if (block1) {
+			eb = read_tree_block(root, block1, blocksize, 0);
+			free_extent_buffer(eb);
+		}
+		if (block1) {
+			eb = read_tree_block(root, block2, blocksize, 0);
+			free_extent_buffer(eb);
+		}
+	}
+	return ret;
+}
+
+
+/*
  * when we walk down the tree, it is usually safe to unlock the higher layers
  * in the tree.  The exceptions are when our path goes through slot 0, because
  * operations on the tree might require changing key pointers higher up in the
@@ -1315,6 +1422,32 @@ static noinline void unlock_up(struct btrfs_path *path, int level,
 }
 
 /*
+ * This releases any locks held in the path starting at level and
+ * going all the way up to the root.
+ *
+ * btrfs_search_slot will keep the lock held on higher nodes in a few
+ * corner cases, such as COW of the block at slot zero in the node.  This
+ * ignores those rules, and it should only be called when there are no
+ * more updates to be done higher up in the tree.
+ */
+noinline void btrfs_unlock_up_safe(struct btrfs_path *path, int level)
+{
+	int i;
+
+	if (path->keep_locks || path->lowest_level)
+		return;
+
+	for (i = level; i < BTRFS_MAX_LEVEL; i++) {
+		if (!path->nodes[i])
+			break;
+		if (!path->locks[i])
+			break;
+		btrfs_tree_unlock(path->nodes[i]);
+		path->locks[i] = 0;
+	}
+}
+
+/*
  * look for key in the tree.  path is filled in with nodes along the way
  * if key is found, we return zero and you can find the item in the leaf
  * level of the path (level 0)
@@ -1385,6 +1518,7 @@ again:
 			 */
 			if (prealloc_block.objectid &&
 			    prealloc_block.offset != b->len) {
+				btrfs_set_path_blocking(p);
 				btrfs_free_reserved_extent(root,
 					   prealloc_block.objectid,
 					   prealloc_block.offset);
@@ -1409,6 +1543,8 @@ again:
 				goto again;
 			}
 
+			btrfs_set_path_blocking(p);
+
 			wret = btrfs_cow_block(trans, root, b,
 					       p->nodes[level + 1],
 					       p->slots[level + 1],
@@ -1430,6 +1566,22 @@ cow_done:
 		if (!p->skip_locking)
 			p->locks[level] = 1;
 
+		btrfs_clear_path_blocking(p);
+
+		/*
+		 * we have a lock on b and as long as we aren't changing
+		 * the tree, there is no way to for the items in b to change.
+		 * It is safe to drop the lock on our parent before we
+		 * go through the expensive btree search on b.
+		 *
+		 * If cow is true, then we might be changing slot zero,
+		 * which may require changing the parent.  So, we can't
+		 * drop the lock until after we know which slot we're
+		 * operating on.
+		 */
+		if (!cow)
+			btrfs_unlock_up_safe(p, level + 1);
+
 		ret = check_block(root, p, level);
 		if (ret) {
 			ret = -1;
@@ -1437,6 +1589,7 @@ cow_done:
 		}
 
 		ret = bin_search(b, key, level, &slot);
+
 		if (level != 0) {
 			if (ret && slot > 0)
 				slot -= 1;
@@ -1444,7 +1597,16 @@ cow_done:
 			if ((p->search_for_split || ins_len > 0) &&
 			    btrfs_header_nritems(b) >=
 			    BTRFS_NODEPTRS_PER_BLOCK(root) - 3) {
-				int sret = split_node(trans, root, p, level);
+				int sret;
+
+				sret = reada_for_balance(root, p, level);
+				if (sret)
+					goto again;
+
+				btrfs_set_path_blocking(p);
+				sret = split_node(trans, root, p, level);
+				btrfs_clear_path_blocking(p);
+
 				BUG_ON(sret > 0);
 				if (sret) {
 					ret = sret;
@@ -1453,8 +1615,16 @@ cow_done:
 				b = p->nodes[level];
 				slot = p->slots[level];
 			} else if (ins_len < 0) {
-				int sret = balance_level(trans, root, p,
-							 level);
+				int sret;
+
+				sret = reada_for_balance(root, p, level);
+				if (sret)
+					goto again;
+
+				btrfs_set_path_blocking(p);
+				sret = balance_level(trans, root, p, level);
+				btrfs_clear_path_blocking(p);
+
 				if (sret) {
 					ret = sret;
 					goto done;
@@ -1488,7 +1658,7 @@ cow_done:
 				 * of the btree by dropping locks before
 				 * we read.
 				 */
-				if (level > 1) {
+				if (level > 0) {
 					btrfs_release_path(NULL, p);
 					if (tmp)
 						free_extent_buffer(tmp);
@@ -1503,6 +1673,7 @@ cow_done:
 						free_extent_buffer(tmp);
 					goto again;
 				} else {
+					btrfs_set_path_blocking(p);
 					if (tmp)
 						free_extent_buffer(tmp);
 					if (should_reada)
@@ -1512,14 +1683,29 @@ cow_done:
 					b = read_node_slot(root, b, slot);
 				}
 			}
-			if (!p->skip_locking)
-				btrfs_tree_lock(b);
+			if (!p->skip_locking) {
+				int lret;
+
+				btrfs_clear_path_blocking(p);
+				lret = btrfs_try_spin_lock(b);
+
+				if (!lret) {
+					btrfs_set_path_blocking(p);
+					btrfs_tree_lock(b);
+					btrfs_clear_path_blocking(p);
+				}
+			}
 		} else {
 			p->slots[level] = slot;
 			if (ins_len > 0 &&
 			    btrfs_leaf_free_space(root, b) < ins_len) {
-				int sret = split_leaf(trans, root, key,
+				int sret;
+
+				btrfs_set_path_blocking(p);
+				sret = split_leaf(trans, root, key,
 						      p, ins_len, ret == 0);
+				btrfs_clear_path_blocking(p);
+
 				BUG_ON(sret > 0);
 				if (sret) {
 					ret = sret;
@@ -1533,12 +1719,16 @@ cow_done:
 	}
 	ret = 1;
 done:
+	/*
+	 * we don't really know what they plan on doing with the path
+	 * from here on, so for now just mark it as blocking
+	 */
+	btrfs_set_path_blocking(p);
 	if (prealloc_block.objectid) {
 		btrfs_free_reserved_extent(root,
 			   prealloc_block.objectid,
 			   prealloc_block.offset);
 	}
-
 	return ret;
 }
 
@@ -1562,6 +1752,8 @@ int btrfs_merge_path(struct btrfs_trans_handle *trans,
 	ret = btrfs_cow_block(trans, root, eb, NULL, 0, &eb, 0);
 	BUG_ON(ret);
 
+	btrfs_set_lock_blocking(eb);
+
 	parent = eb;
 	while (1) {
 		level = btrfs_header_level(parent);
@@ -1586,6 +1778,7 @@ int btrfs_merge_path(struct btrfs_trans_handle *trans,
 			eb = read_tree_block(root, bytenr, blocksize,
 					     generation);
 			btrfs_tree_lock(eb);
+			btrfs_set_lock_blocking(eb);
 		}
 
 		/*
@@ -1610,6 +1803,7 @@ int btrfs_merge_path(struct btrfs_trans_handle *trans,
 				eb = read_tree_block(root, bytenr, blocksize,
 						generation);
 				btrfs_tree_lock(eb);
+				btrfs_set_lock_blocking(eb);
 			}
 
 			ret = btrfs_cow_block(trans, root, eb, parent, slot,
@@ -2156,6 +2350,8 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root
 
 	right = read_node_slot(root, upper, slot + 1);
 	btrfs_tree_lock(right);
+	btrfs_set_lock_blocking(right);
+
 	free_space = btrfs_leaf_free_space(root, right);
 	if (free_space < data_size)
 		goto out_unlock;
@@ -2351,6 +2547,8 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
 
 	left = read_node_slot(root, path->nodes[1], slot - 1);
 	btrfs_tree_lock(left);
+	btrfs_set_lock_blocking(left);
+
 	free_space = btrfs_leaf_free_space(root, left);
 	if (free_space < data_size) {
 		ret = 1;
@@ -2809,6 +3007,12 @@ int btrfs_split_item(struct btrfs_trans_handle *trans,
 	path->keep_locks = 0;
 	BUG_ON(ret);
 
+	/*
+	 * make sure any changes to the path from split_leaf leave it
+	 * in a blocking state
+	 */
+	btrfs_set_path_blocking(path);
+
 	leaf = path->nodes[0];
 	BUG_ON(btrfs_leaf_free_space(root, leaf) < sizeof(struct btrfs_item));
 
@@ -3338,6 +3542,7 @@ int btrfs_insert_empty_items(struct btrfs_trans_handle *trans,
 		BUG();
 	}
 out:
+	btrfs_unlock_up_safe(path, 1);
 	return ret;
 }
 
@@ -3705,12 +3910,14 @@ find_next_key:
 		 */
 		if (slot >= nritems) {
 			path->slots[level] = slot;
+			btrfs_set_path_blocking(path);
 			sret = btrfs_find_next_key(root, path, min_key, level,
 						  cache_only, min_trans);
 			if (sret == 0) {
 				btrfs_release_path(root, path);
 				goto again;
 			} else {
+				btrfs_clear_path_blocking(path);
 				goto out;
 			}
 		}
@@ -3722,16 +3929,20 @@ find_next_key:
 			unlock_up(path, level, 1);
 			goto out;
 		}
+		btrfs_set_path_blocking(path);
 		cur = read_node_slot(root, cur, slot);
 
 		btrfs_tree_lock(cur);
+
 		path->locks[level - 1] = 1;
 		path->nodes[level - 1] = cur;
 		unlock_up(path, level, 1);
+		btrfs_clear_path_blocking(path);
 	}
 out:
 	if (ret == 0)
 		memcpy(min_key, &found_key, sizeof(found_key));
+	btrfs_set_path_blocking(path);
 	return ret;
 }
 
@@ -3827,6 +4038,7 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
 	if (ret < 0)
 		return ret;
 
+	btrfs_set_path_blocking(path);
 	nritems = btrfs_header_nritems(path->nodes[0]);
 	/*
 	 * by releasing the path above we dropped all our locks.  A balance
@@ -3857,6 +4069,7 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
 			free_extent_buffer(next);
 		}
 
+		/* the path was set to blocking above */
 		if (level == 1 && (path->locks[1] || path->skip_locking) &&
 		    path->reada)
 			reada_for_search(root, path, level, slot, 0);
@@ -3865,6 +4078,7 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
 		if (!path->skip_locking) {
 			WARN_ON(!btrfs_tree_locked(c));
 			btrfs_tree_lock(next);
+			btrfs_set_lock_blocking(next);
 		}
 		break;
 	}
@@ -3881,12 +4095,15 @@ int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
 			path->locks[level] = 1;
 		if (!level)
 			break;
+
+		btrfs_set_path_blocking(path);
 		if (level == 1 && path->locks[1] && path->reada)
 			reada_for_search(root, path, level, slot, 0);
 		next = read_node_slot(root, next, 0);
 		if (!path->skip_locking) {
 			WARN_ON(!btrfs_tree_locked(path->nodes[level]));
 			btrfs_tree_lock(next);
+			btrfs_set_lock_blocking(next);
 		}
 	}
 done:
@@ -3911,6 +4128,7 @@ int btrfs_previous_item(struct btrfs_root *root,
 
 	while (1) {
 		if (path->slots[0] == 0) {
+			btrfs_set_path_blocking(path);
 			ret = btrfs_prev_leaf(root, path);
 			if (ret != 0)
 				return ret;