Commit d7df2c79 authored by Josef Bacik's avatar Josef Bacik Committed by Chris Mason
Browse files

Btrfs: attach delayed ref updates to delayed ref heads 



Currently we have two rb-trees, one for delayed ref heads and one for all of the
delayed refs, including the delayed ref heads.  When we process the delayed refs
we have to hold onto the delayed ref lock for all of the selecting and merging
and such, which results in quite a bit of lock contention.  This was solved by
having a waitqueue and only one flusher at a time, however this hurts if we get
a lot of delayed refs queued up.

So instead just have an rb tree for the delayed ref heads, and then attach the
delayed ref updates to an rb tree that is per delayed ref head.  Then we only
need to take the delayed ref lock when adding new delayed refs and when
selecting a delayed ref head to process, all the rest of the time we deal with a
per delayed ref head lock which will be much less contentious.

The locking rules for this get a little more complicated since we have to lock
up to 3 things to properly process delayed refs, but I will address that problem
later.  For now this passes all of xfstests and my overnight stress tests.
Thanks,

Signed-off-by: default avatarJosef Bacik <jbacik@fb.com>
Signed-off-by: default avatarChris Mason <clm@fb.com>
parent 5039eddc

fs/btrfs/backref.c

+11 −12
Original line number Diff line number Diff line
@@ -538,14 +538,13 @@ static int __add_delayed_refs(struct btrfs_delayed_ref_head *head, u64 seq, 
	if (extent_op && extent_op->update_key)

		btrfs_disk_key_to_cpu(&op_key, &extent_op->key);



	while ((n = rb_prev(n))) {

	spin_lock(&head->lock);

	n = rb_first(&head->ref_root);

	while (n) {

		struct btrfs_delayed_ref_node *node;

		node = rb_entry(n, struct btrfs_delayed_ref_node,

				rb_node);

		if (node->bytenr != head->node.bytenr)

			break;

		WARN_ON(node->is_head);



		n = rb_next(n);

		if (node->seq > seq)

			continue;
@@ -612,10 +611,10 @@ static int __add_delayed_refs(struct btrfs_delayed_ref_head *head, u64 seq, 
			WARN_ON(1);

		}

		if (ret)

			return ret;

			break;

	}



	return 0;

	spin_unlock(&head->lock);

	return ret;

}



/*
@@ -882,16 +881,16 @@ again: 
				btrfs_put_delayed_ref(&head->node);

				goto again;

			}

			spin_unlock(&delayed_refs->lock);

			ret = __add_delayed_refs(head, time_seq,

						 &prefs_delayed);

			mutex_unlock(&head->mutex);

			if (ret) {

				spin_unlock(&delayed_refs->lock);

			if (ret)

				goto out;

			}

		}

		} else {

			spin_unlock(&delayed_refs->lock);

		}

	}



	if (path->slots[0]) {

		struct extent_buffer *leaf;

fs/btrfs/delayed-ref.c

+108 −115
Original line number Diff line number Diff line
@@ -269,39 +269,38 @@ int btrfs_delayed_ref_lock(struct btrfs_trans_handle *trans, 


static inline void drop_delayed_ref(struct btrfs_trans_handle *trans,

				    struct btrfs_delayed_ref_root *delayed_refs,

				    struct btrfs_delayed_ref_head *head,

				    struct btrfs_delayed_ref_node *ref)

{

	rb_erase(&ref->rb_node, &delayed_refs->root);

	if (btrfs_delayed_ref_is_head(ref)) {

		struct btrfs_delayed_ref_head *head;



		head = btrfs_delayed_node_to_head(ref);

		rb_erase(&head->href_node, &delayed_refs->href_root);

	} else {

		assert_spin_locked(&head->lock);

		rb_erase(&ref->rb_node, &head->ref_root);

	}

	ref->in_tree = 0;

	btrfs_put_delayed_ref(ref);

	delayed_refs->num_entries--;

	atomic_dec(&delayed_refs->num_entries);

	if (trans->delayed_ref_updates)

		trans->delayed_ref_updates--;

}



static int merge_ref(struct btrfs_trans_handle *trans,

		     struct btrfs_delayed_ref_root *delayed_refs,

		     struct btrfs_delayed_ref_head *head,

		     struct btrfs_delayed_ref_node *ref, u64 seq)

{

	struct rb_node *node;

	int merged = 0;

	int mod = 0;

	int done = 0;



	node = rb_prev(&ref->rb_node);

	while (node) {

	node = rb_next(&ref->rb_node);

	while (!done && node) {

		struct btrfs_delayed_ref_node *next;



		next = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);

		node = rb_prev(node);

		if (next->bytenr != ref->bytenr)

			break;

		node = rb_next(node);

		if (seq && next->seq >= seq)

			break;

		if (comp_entry(ref, next, 0))
@@ -321,12 +320,11 @@ static int merge_ref(struct btrfs_trans_handle *trans, 
			mod = -next->ref_mod;

		}



		merged++;

		drop_delayed_ref(trans, delayed_refs, next);

		drop_delayed_ref(trans, delayed_refs, head, next);

		ref->ref_mod += mod;

		if (ref->ref_mod == 0) {

			drop_delayed_ref(trans, delayed_refs, ref);

			break;

			drop_delayed_ref(trans, delayed_refs, head, ref);

			done = 1;

		} else {

			/*

			 * You can't have multiples of the same ref on a tree
@@ -335,13 +333,8 @@ static int merge_ref(struct btrfs_trans_handle *trans, 
			WARN_ON(ref->type == BTRFS_TREE_BLOCK_REF_KEY ||

				ref->type == BTRFS_SHARED_BLOCK_REF_KEY);

		}



		if (done)

			break;

		node = rb_prev(&ref->rb_node);

	}



	return merged;

	return done;

}



void btrfs_merge_delayed_refs(struct btrfs_trans_handle *trans,
@@ -352,6 +345,7 @@ void btrfs_merge_delayed_refs(struct btrfs_trans_handle *trans, 
	struct rb_node *node;

	u64 seq = 0;



	assert_spin_locked(&head->lock);

	/*

	 * We don't have too much refs to merge in the case of delayed data

	 * refs.
@@ -369,22 +363,19 @@ void btrfs_merge_delayed_refs(struct btrfs_trans_handle *trans, 
	}

	spin_unlock(&fs_info->tree_mod_seq_lock);



	node = rb_prev(&head->node.rb_node);

	node = rb_first(&head->ref_root);

	while (node) {

		struct btrfs_delayed_ref_node *ref;



		ref = rb_entry(node, struct btrfs_delayed_ref_node,

			       rb_node);

		if (ref->bytenr != head->node.bytenr)

			break;



		/* We can't merge refs that are outside of our seq count */

		if (seq && ref->seq >= seq)

			break;

		if (merge_ref(trans, delayed_refs, ref, seq))

			node = rb_prev(&head->node.rb_node);

		if (merge_ref(trans, delayed_refs, head, ref, seq))

			node = rb_first(&head->ref_root);

		else

			node = rb_prev(node);

			node = rb_next(&ref->rb_node);

	}

}
@@ -412,64 +403,52 @@ int btrfs_check_delayed_seq(struct btrfs_fs_info *fs_info, 
	return ret;

}



int btrfs_find_ref_cluster(struct btrfs_trans_handle *trans,

			   struct list_head *cluster, u64 start)

struct btrfs_delayed_ref_head *

btrfs_select_ref_head(struct btrfs_trans_handle *trans)

{

	int count = 0;

	struct btrfs_delayed_ref_root *delayed_refs;

	struct rb_node *node;

	struct btrfs_delayed_ref_head *head = NULL;

	struct btrfs_delayed_ref_head *head;

	u64 start;

	bool loop = false;



	delayed_refs = &trans->transaction->delayed_refs;

	node = rb_first(&delayed_refs->href_root);



	if (start) {

		find_ref_head(&delayed_refs->href_root, start + 1, &head, 1);

		if (head)

			node = &head->href_node;

	}

again:

	while (node && count < 32) {

		head = rb_entry(node, struct btrfs_delayed_ref_head, href_node);

		if (list_empty(&head->cluster)) {

			list_add_tail(&head->cluster, cluster);

			delayed_refs->run_delayed_start =

				head->node.bytenr;

			count++;



			WARN_ON(delayed_refs->num_heads_ready == 0);

			delayed_refs->num_heads_ready--;

		} else if (count) {

			/* the goal of the clustering is to find extents

			 * that are likely to end up in the same extent

			 * leaf on disk.  So, we don't want them spread

			 * all over the tree.  Stop now if we've hit

			 * a head that was already in use

			 */

			break;

		}

		node = rb_next(node);

	start = delayed_refs->run_delayed_start;

	head = find_ref_head(&delayed_refs->href_root, start, NULL, 1);

	if (!head && !loop) {

		delayed_refs->run_delayed_start = 0;

		start = 0;

		loop = true;

		head = find_ref_head(&delayed_refs->href_root, start, NULL, 1);

		if (!head)

			return NULL;

	} else if (!head && loop) {

		return NULL;

	}

	if (count) {

		return 0;

	} else if (start) {

		/*

		 * we've gone to the end of the rbtree without finding any

		 * clusters.  start from the beginning and try again

		 */



	while (head->processing) {

		struct rb_node *node;



		node = rb_next(&head->href_node);

		if (!node) {

			if (loop)

				return NULL;

			delayed_refs->run_delayed_start = 0;

			start = 0;

		node = rb_first(&delayed_refs->href_root);

			loop = true;

			goto again;

		}

	return 1;

		head = rb_entry(node, struct btrfs_delayed_ref_head,

				href_node);

	}



void btrfs_release_ref_cluster(struct list_head *cluster)

{

	struct list_head *pos, *q;



	list_for_each_safe(pos, q, cluster)

		list_del_init(pos);

	head->processing = 1;

	WARN_ON(delayed_refs->num_heads_ready == 0);

	delayed_refs->num_heads_ready--;

	delayed_refs->run_delayed_start = head->node.bytenr +

		head->node.num_bytes;

	return head;

}



/*
@@ -483,6 +462,7 @@ void btrfs_release_ref_cluster(struct list_head *cluster) 
static noinline void

update_existing_ref(struct btrfs_trans_handle *trans,

		    struct btrfs_delayed_ref_root *delayed_refs,

		    struct btrfs_delayed_ref_head *head,

		    struct btrfs_delayed_ref_node *existing,

		    struct btrfs_delayed_ref_node *update)

{
@@ -495,7 +475,7 @@ update_existing_ref(struct btrfs_trans_handle *trans, 
		 */

		existing->ref_mod--;

		if (existing->ref_mod == 0)

			drop_delayed_ref(trans, delayed_refs, existing);

			drop_delayed_ref(trans, delayed_refs, head, existing);

		else

			WARN_ON(existing->type == BTRFS_TREE_BLOCK_REF_KEY ||

				existing->type == BTRFS_SHARED_BLOCK_REF_KEY);
@@ -565,9 +545,13 @@ update_existing_head_ref(struct btrfs_delayed_ref_node *existing, 
		}

	}

	/*

	 * update the reference mod on the head to reflect this new operation

	 * update the reference mod on the head to reflect this new operation,

	 * only need the lock for this case cause we could be processing it

	 * currently, for refs we just added we know we're a-ok.

	 */

	spin_lock(&existing_ref->lock);

	existing->ref_mod += update->ref_mod;

	spin_unlock(&existing_ref->lock);

}



/*
@@ -575,13 +559,13 @@ update_existing_head_ref(struct btrfs_delayed_ref_node *existing, 
 * this does all the dirty work in terms of maintaining the correct

 * overall modification count.

 */

static noinline void add_delayed_ref_head(struct btrfs_fs_info *fs_info,

static noinline struct btrfs_delayed_ref_head *

add_delayed_ref_head(struct btrfs_fs_info *fs_info,

		     struct btrfs_trans_handle *trans,

					struct btrfs_delayed_ref_node *ref,

					u64 bytenr, u64 num_bytes,

					int action, int is_data)

		     struct btrfs_delayed_ref_node *ref, u64 bytenr,

		     u64 num_bytes, int action, int is_data)

{

	struct btrfs_delayed_ref_node *existing;

	struct btrfs_delayed_ref_head *existing;

	struct btrfs_delayed_ref_head *head_ref = NULL;

	struct btrfs_delayed_ref_root *delayed_refs;

	int count_mod = 1;
@@ -628,39 +612,43 @@ static noinline void add_delayed_ref_head(struct btrfs_fs_info *fs_info, 
	head_ref = btrfs_delayed_node_to_head(ref);

	head_ref->must_insert_reserved = must_insert_reserved;

	head_ref->is_data = is_data;

	head_ref->ref_root = RB_ROOT;

	head_ref->processing = 0;



	INIT_LIST_HEAD(&head_ref->cluster);

	spin_lock_init(&head_ref->lock);

	mutex_init(&head_ref->mutex);



	trace_add_delayed_ref_head(ref, head_ref, action);



	existing = tree_insert(&delayed_refs->root, &ref->rb_node);



	existing = htree_insert(&delayed_refs->href_root,

				&head_ref->href_node);

	if (existing) {

		update_existing_head_ref(existing, ref);

		update_existing_head_ref(&existing->node, ref);

		/*

		 * we've updated the existing ref, free the newly

		 * allocated ref

		 */

		kmem_cache_free(btrfs_delayed_ref_head_cachep, head_ref);

		head_ref = existing;

	} else {

		htree_insert(&delayed_refs->href_root, &head_ref->href_node);

		delayed_refs->num_heads++;

		delayed_refs->num_heads_ready++;

		delayed_refs->num_entries++;

		atomic_inc(&delayed_refs->num_entries);

		trans->delayed_ref_updates++;

	}

	return head_ref;

}



/*

 * helper to insert a delayed tree ref into the rbtree.

 */

static noinline void add_delayed_tree_ref(struct btrfs_fs_info *fs_info,

static noinline void

add_delayed_tree_ref(struct btrfs_fs_info *fs_info,

		     struct btrfs_trans_handle *trans,

					 struct btrfs_delayed_ref_node *ref,

					 u64 bytenr, u64 num_bytes, u64 parent,

					 u64 ref_root, int level, int action,

					 int for_cow)

		     struct btrfs_delayed_ref_head *head_ref,

		     struct btrfs_delayed_ref_node *ref, u64 bytenr,

		     u64 num_bytes, u64 parent, u64 ref_root, int level,

		     int action, int for_cow)

{

	struct btrfs_delayed_ref_node *existing;

	struct btrfs_delayed_tree_ref *full_ref;
@@ -696,30 +684,33 @@ static noinline void add_delayed_tree_ref(struct btrfs_fs_info *fs_info, 


	trace_add_delayed_tree_ref(ref, full_ref, action);



	existing = tree_insert(&delayed_refs->root, &ref->rb_node);



	spin_lock(&head_ref->lock);

	existing = tree_insert(&head_ref->ref_root, &ref->rb_node);

	if (existing) {

		update_existing_ref(trans, delayed_refs, existing, ref);

		update_existing_ref(trans, delayed_refs, head_ref, existing,

				    ref);

		/*

		 * we've updated the existing ref, free the newly

		 * allocated ref

		 */

		kmem_cache_free(btrfs_delayed_tree_ref_cachep, full_ref);

	} else {

		delayed_refs->num_entries++;

		atomic_inc(&delayed_refs->num_entries);

		trans->delayed_ref_updates++;

	}

	spin_unlock(&head_ref->lock);

}



/*

 * helper to insert a delayed data ref into the rbtree.

 */

static noinline void add_delayed_data_ref(struct btrfs_fs_info *fs_info,

static noinline void

add_delayed_data_ref(struct btrfs_fs_info *fs_info,

		     struct btrfs_trans_handle *trans,

					 struct btrfs_delayed_ref_node *ref,

					 u64 bytenr, u64 num_bytes, u64 parent,

					 u64 ref_root, u64 owner, u64 offset,

					 int action, int for_cow)

		     struct btrfs_delayed_ref_head *head_ref,

		     struct btrfs_delayed_ref_node *ref, u64 bytenr,

		     u64 num_bytes, u64 parent, u64 ref_root, u64 owner,

		     u64 offset, int action, int for_cow)

{

	struct btrfs_delayed_ref_node *existing;

	struct btrfs_delayed_data_ref *full_ref;
@@ -757,19 +748,21 @@ static noinline void add_delayed_data_ref(struct btrfs_fs_info *fs_info, 


	trace_add_delayed_data_ref(ref, full_ref, action);



	existing = tree_insert(&delayed_refs->root, &ref->rb_node);



	spin_lock(&head_ref->lock);

	existing = tree_insert(&head_ref->ref_root, &ref->rb_node);

	if (existing) {

		update_existing_ref(trans, delayed_refs, existing, ref);

		update_existing_ref(trans, delayed_refs, head_ref, existing,

				    ref);

		/*

		 * we've updated the existing ref, free the newly

		 * allocated ref

		 */

		kmem_cache_free(btrfs_delayed_data_ref_cachep, full_ref);

	} else {

		delayed_refs->num_entries++;

		atomic_inc(&delayed_refs->num_entries);

		trans->delayed_ref_updates++;

	}

	spin_unlock(&head_ref->lock);

}



/*
@@ -808,10 +801,10 @@ int btrfs_add_delayed_tree_ref(struct btrfs_fs_info *fs_info, 
	 * insert both the head node and the new ref without dropping

	 * the spin lock

	 */

	add_delayed_ref_head(fs_info, trans, &head_ref->node, bytenr,

				   num_bytes, action, 0);

	head_ref = add_delayed_ref_head(fs_info, trans, &head_ref->node,

					bytenr, num_bytes, action, 0);



	add_delayed_tree_ref(fs_info, trans, &ref->node, bytenr,

	add_delayed_tree_ref(fs_info, trans, head_ref, &ref->node, bytenr,

				   num_bytes, parent, ref_root, level, action,

				   for_cow);

	spin_unlock(&delayed_refs->lock);
@@ -856,10 +849,10 @@ int btrfs_add_delayed_data_ref(struct btrfs_fs_info *fs_info, 
	 * insert both the head node and the new ref without dropping

	 * the spin lock

	 */

	add_delayed_ref_head(fs_info, trans, &head_ref->node, bytenr,

				   num_bytes, action, 1);

	head_ref = add_delayed_ref_head(fs_info, trans, &head_ref->node,

					bytenr, num_bytes, action, 1);



	add_delayed_data_ref(fs_info, trans, &ref->node, bytenr,

	add_delayed_data_ref(fs_info, trans, head_ref, &ref->node, bytenr,

				   num_bytes, parent, ref_root, owner, offset,

				   action, for_cow);

	spin_unlock(&delayed_refs->lock);

fs/btrfs/delayed-ref.h

+7 −16
Original line number Diff line number Diff line
@@ -81,7 +81,8 @@ struct btrfs_delayed_ref_head { 
	 */

	struct mutex mutex;



	struct list_head cluster;

	spinlock_t lock;

	struct rb_root ref_root;



	struct rb_node href_node;
@@ -100,6 +101,7 @@ struct btrfs_delayed_ref_head { 
	 */

	unsigned int must_insert_reserved:1;

	unsigned int is_data:1;

	unsigned int processing:1;

};



struct btrfs_delayed_tree_ref {
@@ -118,8 +120,6 @@ struct btrfs_delayed_data_ref { 
};



struct btrfs_delayed_ref_root {

	struct rb_root root;



	/* head ref rbtree */

	struct rb_root href_root;
@@ -129,7 +129,7 @@ struct btrfs_delayed_ref_root { 
	/* how many delayed ref updates we've queued, used by the

	 * throttling code

	 */

	unsigned long num_entries;

	atomic_t num_entries;



	/* total number of head nodes in tree */

	unsigned long num_heads;
@@ -137,15 +137,6 @@ struct btrfs_delayed_ref_root { 
	/* total number of head nodes ready for processing */

	unsigned long num_heads_ready;



	/*

	 * bumped when someone is making progress on the delayed

	 * refs, so that other procs know they are just adding to

	 * contention intead of helping

	 */

	atomic_t procs_running_refs;

	atomic_t ref_seq;

	wait_queue_head_t wait;



	/*

	 * set when the tree is flushing before a transaction commit,

	 * used by the throttling code to decide if new updates need
@@ -231,9 +222,9 @@ static inline void btrfs_delayed_ref_unlock(struct btrfs_delayed_ref_head *head) 
	mutex_unlock(&head->mutex);

}



int btrfs_find_ref_cluster(struct btrfs_trans_handle *trans,

			   struct list_head *cluster, u64 search_start);

void btrfs_release_ref_cluster(struct list_head *cluster);



struct btrfs_delayed_ref_head *

btrfs_select_ref_head(struct btrfs_trans_handle *trans);



int btrfs_check_delayed_seq(struct btrfs_fs_info *fs_info,

			    struct btrfs_delayed_ref_root *delayed_refs,

fs/btrfs/disk-io.c

+38 −41
Original line number Diff line number Diff line
@@ -3801,58 +3801,55 @@ static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 
	delayed_refs = &trans->delayed_refs;



	spin_lock(&delayed_refs->lock);

	if (delayed_refs->num_entries == 0) {

	if (atomic_read(&delayed_refs->num_entries) == 0) {

		spin_unlock(&delayed_refs->lock);

		btrfs_info(root->fs_info, "delayed_refs has NO entry");

		return ret;

	}



	while ((node = rb_first(&delayed_refs->root)) != NULL) {

		struct btrfs_delayed_ref_head *head = NULL;

	while ((node = rb_first(&delayed_refs->href_root)) != NULL) {

		struct btrfs_delayed_ref_head *head;

		bool pin_bytes = false;



		ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);

		atomic_set(&ref->refs, 1);

		if (btrfs_delayed_ref_is_head(ref)) {



			head = btrfs_delayed_node_to_head(ref);

		head = rb_entry(node, struct btrfs_delayed_ref_head,

				href_node);

		if (!mutex_trylock(&head->mutex)) {

				atomic_inc(&ref->refs);

			atomic_inc(&head->node.refs);

			spin_unlock(&delayed_refs->lock);



				/* Need to wait for the delayed ref to run */

			mutex_lock(&head->mutex);

			mutex_unlock(&head->mutex);

				btrfs_put_delayed_ref(ref);



			btrfs_put_delayed_ref(&head->node);

			spin_lock(&delayed_refs->lock);

			continue;

		}



		spin_lock(&head->lock);

		while ((node = rb_first(&head->ref_root)) != NULL) {

			ref = rb_entry(node, struct btrfs_delayed_ref_node,

				       rb_node);

			ref->in_tree = 0;

			rb_erase(&ref->rb_node, &head->ref_root);

			atomic_dec(&delayed_refs->num_entries);

			btrfs_put_delayed_ref(ref);

			cond_resched_lock(&head->lock);

		}

		if (head->must_insert_reserved)

			pin_bytes = true;

		btrfs_free_delayed_extent_op(head->extent_op);

		delayed_refs->num_heads--;

			if (list_empty(&head->cluster))

		if (head->processing == 0)

			delayed_refs->num_heads_ready--;

			list_del_init(&head->cluster);

		}



		ref->in_tree = 0;

		rb_erase(&ref->rb_node, &delayed_refs->root);

		if (head)

		atomic_dec(&delayed_refs->num_entries);

		head->node.in_tree = 0;

		rb_erase(&head->href_node, &delayed_refs->href_root);



		delayed_refs->num_entries--;

		spin_unlock(&head->lock);

		spin_unlock(&delayed_refs->lock);

		if (head) {

			if (pin_bytes)

				btrfs_pin_extent(root, ref->bytenr,

						 ref->num_bytes, 1);

		mutex_unlock(&head->mutex);

		}

		btrfs_put_delayed_ref(ref);



		if (pin_bytes)

			btrfs_pin_extent(root, head->node.bytenr,

					 head->node.num_bytes, 1);

		btrfs_put_delayed_ref(&head->node);

		cond_resched();

		spin_lock(&delayed_refs->lock);

	}

fs/btrfs/extent-tree.c

+102 −215

File changed.

Preview size limit exceeded, changes collapsed.

CRA Git | Maintained and supported by SUSTech CRA and CCSE