Merge tag 'v3.9-rc5' into wq/for-3.10

Writeback conversion to workqueue will be based on top of wq/for-3.10
branch to take advantage of custom attrs and NUMA support for unbound
workqueues.  Mainline currently contains two commits which result in
non-trivial merge conflicts with wq/for-3.10 and because
block/for-3.10/core is based on v3.9-rc3 which contains one of the
conflicting commits, we need a pre-merge-window merge anyway.  Let's
pull v3.9-rc5 into wq/for-3.10 so that the block tree doesn't suffer
from workqueue merge conflicts.

The two conflicts and their resolutions:

* e68035fb65 ("workqueue: convert to idr_alloc()") in mainline changes
  worker_pool_assign_id() to use idr_alloc() instead of the old idr
  interface.  worker_pool_assign_id() goes through multiple locking
  changes in wq/for-3.10 causing the following conflict.

  static int worker_pool_assign_id(struct worker_pool *pool)
  {
	  int ret;

  <<<<<<< HEAD
	  lockdep_assert_held(&wq_pool_mutex);

	  do {
		  if (!idr_pre_get(&worker_pool_idr, GFP_KERNEL))
			  return -ENOMEM;
		  ret = idr_get_new(&worker_pool_idr, pool, &pool->id);
	  } while (ret == -EAGAIN);
  =======
	  mutex_lock(&worker_pool_idr_mutex);
	  ret = idr_alloc(&worker_pool_idr, pool, 0, 0, GFP_KERNEL);
	  if (ret >= 0)
		  pool->id = ret;
	  mutex_unlock(&worker_pool_idr_mutex);
  >>>>>>> c67bf5361e7e66a0ff1f4caf95f89347d55dfb89

	  return ret < 0 ? ret : 0;
  }

  We want locking from the former and idr_alloc() usage from the
  latter, which can be combined to the following.

  static int worker_pool_assign_id(struct worker_pool *pool)
  {
	  int ret;

	  lockdep_assert_held(&wq_pool_mutex);

	  ret = idr_alloc(&worker_pool_idr, pool, 0, 0, GFP_KERNEL);
	  if (ret >= 0) {
		  pool->id = ret;
		  return 0;
	  }
	  return ret;
   }

* eb2834285c ("workqueue: fix possible pool stall bug in
  wq_unbind_fn()") updated wq_unbind_fn() such that it has single
  larger for_each_std_worker_pool() loop instead of two separate loops
  with a schedule() call inbetween.  wq/for-3.10 renamed
  pool->assoc_mutex to pool->manager_mutex causing the following
  conflict (earlier function body and comments omitted for brevity).

  static void wq_unbind_fn(struct work_struct *work)
  {
  ...
		  spin_unlock_irq(&pool->lock);
  <<<<<<< HEAD
		  mutex_unlock(&pool->manager_mutex);
	  }
  =======
		  mutex_unlock(&pool->assoc_mutex);
  >>>>>>> c67bf5361e7e66a0ff1f4caf95f89347d55dfb89

		  schedule();

  <<<<<<< HEAD
	  for_each_cpu_worker_pool(pool, cpu)
  =======
  >>>>>>> c67bf5361e7e66a0ff1f4caf95f89347d55dfb89
		  atomic_set(&pool->nr_running, 0);

		  spin_lock_irq(&pool->lock);
		  wake_up_worker(pool);
		  spin_unlock_irq(&pool->lock);
	  }
  }

  The resolution is mostly trivial.  We want the control flow of the
  latter with the rename of the former.

  static void wq_unbind_fn(struct work_struct *work)
  {
  ...
		  spin_unlock_irq(&pool->lock);
		  mutex_unlock(&pool->manager_mutex);

		  schedule();

		  atomic_set(&pool->nr_running, 0);

		  spin_lock_irq(&pool->lock);
		  wake_up_worker(pool);
		  spin_unlock_irq(&pool->lock);
	  }
  }

Signed-off-by: Tejun Heo <tj@kernel.org>

1 files changed, 90 insertions, 61 deletions

diff --git a/fs/btrfs/delayed-inode.c b/fs/btrfs/delayed-inode.c
index 0b278b117cbe..14fce27b4780 100644
--- a/fs/btrfs/delayed-inode.c
+++ b/fs/btrfs/delayed-inode.c
@@ -22,8 +22,9 @@
 #include "disk-io.h"
 #include "transaction.h"
 
-#define BTRFS_DELAYED_WRITEBACK		400
-#define BTRFS_DELAYED_BACKGROUND	100
+#define BTRFS_DELAYED_WRITEBACK		512
+#define BTRFS_DELAYED_BACKGROUND	128
+#define BTRFS_DELAYED_BATCH		16
 
 static struct kmem_cache *delayed_node_cache;
 
@@ -494,6 +495,15 @@ static int __btrfs_add_delayed_deletion_item(struct btrfs_delayed_node *node,
 					BTRFS_DELAYED_DELETION_ITEM);
 }
 
+static void finish_one_item(struct btrfs_delayed_root *delayed_root)
+{
+	int seq = atomic_inc_return(&delayed_root->items_seq);
+	if ((atomic_dec_return(&delayed_root->items) <
+	    BTRFS_DELAYED_BACKGROUND || seq % BTRFS_DELAYED_BATCH == 0) &&
+	    waitqueue_active(&delayed_root->wait))
+		wake_up(&delayed_root->wait);
+}
+
 static void __btrfs_remove_delayed_item(struct btrfs_delayed_item *delayed_item)
 {
 	struct rb_root *root;
@@ -512,10 +522,8 @@ static void __btrfs_remove_delayed_item(struct btrfs_delayed_item *delayed_item)
 
 	rb_erase(&delayed_item->rb_node, root);
 	delayed_item->delayed_node->count--;
-	if (atomic_dec_return(&delayed_root->items) <
-	    BTRFS_DELAYED_BACKGROUND &&
-	    waitqueue_active(&delayed_root->wait))
-		wake_up(&delayed_root->wait);
+
+	finish_one_item(delayed_root);
 }
 
 static void btrfs_release_delayed_item(struct btrfs_delayed_item *item)
@@ -1056,10 +1064,7 @@ static void btrfs_release_delayed_inode(struct btrfs_delayed_node *delayed_node)
 		delayed_node->count--;
 
 		delayed_root = delayed_node->root->fs_info->delayed_root;
-		if (atomic_dec_return(&delayed_root->items) <
-		    BTRFS_DELAYED_BACKGROUND &&
-		    waitqueue_active(&delayed_root->wait))
-			wake_up(&delayed_root->wait);
+		finish_one_item(delayed_root);
 	}
 }
 
@@ -1304,35 +1309,44 @@ void btrfs_remove_delayed_node(struct inode *inode)
 	btrfs_release_delayed_node(delayed_node);
 }
 
-struct btrfs_async_delayed_node {
-	struct btrfs_root *root;
-	struct btrfs_delayed_node *delayed_node;
+struct btrfs_async_delayed_work {
+	struct btrfs_delayed_root *delayed_root;
+	int nr;
 	struct btrfs_work work;
 };
 
-static void btrfs_async_run_delayed_node_done(struct btrfs_work *work)
+static void btrfs_async_run_delayed_root(struct btrfs_work *work)
 {
-	struct btrfs_async_delayed_node *async_node;
+	struct btrfs_async_delayed_work *async_work;
+	struct btrfs_delayed_root *delayed_root;
 	struct btrfs_trans_handle *trans;
 	struct btrfs_path *path;
 	struct btrfs_delayed_node *delayed_node = NULL;
 	struct btrfs_root *root;
 	struct btrfs_block_rsv *block_rsv;
-	int need_requeue = 0;
+	int total_done = 0;
 
-	async_node = container_of(work, struct btrfs_async_delayed_node, work);
+	async_work = container_of(work, struct btrfs_async_delayed_work, work);
+	delayed_root = async_work->delayed_root;
 
 	path = btrfs_alloc_path();
 	if (!path)
 		goto out;
-	path->leave_spinning = 1;
 
-	delayed_node = async_node->delayed_node;
+again:
+	if (atomic_read(&delayed_root->items) < BTRFS_DELAYED_BACKGROUND / 2)
+		goto free_path;
+
+	delayed_node = btrfs_first_prepared_delayed_node(delayed_root);
+	if (!delayed_node)
+		goto free_path;
+
+	path->leave_spinning = 1;
 	root = delayed_node->root;
 
 	trans = btrfs_join_transaction(root);
 	if (IS_ERR(trans))
-		goto free_path;
+		goto release_path;
 
 	block_rsv = trans->block_rsv;
 	trans->block_rsv = &root->fs_info->delayed_block_rsv;
@@ -1363,57 +1377,47 @@ static void btrfs_async_run_delayed_node_done(struct btrfs_work *work)
 	 * Task1 will sleep until the transaction is commited.
 	 */
 	mutex_lock(&delayed_node->mutex);
-	if (delayed_node->count)
-		need_requeue = 1;
-	else
-		btrfs_dequeue_delayed_node(root->fs_info->delayed_root,
-					   delayed_node);
+	btrfs_dequeue_delayed_node(root->fs_info->delayed_root, delayed_node);
 	mutex_unlock(&delayed_node->mutex);
 
 	trans->block_rsv = block_rsv;
 	btrfs_end_transaction_dmeta(trans, root);
 	btrfs_btree_balance_dirty_nodelay(root);
+
+release_path:
+	btrfs_release_path(path);
+	total_done++;
+
+	btrfs_release_prepared_delayed_node(delayed_node);
+	if (async_work->nr == 0 || total_done < async_work->nr)
+		goto again;
+
 free_path:
 	btrfs_free_path(path);
 out:
-	if (need_requeue)
-		btrfs_requeue_work(&async_node->work);
-	else {
-		btrfs_release_prepared_delayed_node(delayed_node);
-		kfree(async_node);
-	}
+	wake_up(&delayed_root->wait);
+	kfree(async_work);
 }
 
+
 static int btrfs_wq_run_delayed_node(struct btrfs_delayed_root *delayed_root,
-				     struct btrfs_root *root, int all)
+				     struct btrfs_root *root, int nr)
 {
-	struct btrfs_async_delayed_node *async_node;
-	struct btrfs_delayed_node *curr;
-	int count = 0;
+	struct btrfs_async_delayed_work *async_work;
 
-again:
-	curr = btrfs_first_prepared_delayed_node(delayed_root);
-	if (!curr)
+	if (atomic_read(&delayed_root->items) < BTRFS_DELAYED_BACKGROUND)
 		return 0;
 
-	async_node = kmalloc(sizeof(*async_node), GFP_NOFS);
-	if (!async_node) {
-		btrfs_release_prepared_delayed_node(curr);
+	async_work = kmalloc(sizeof(*async_work), GFP_NOFS);
+	if (!async_work)
 		return -ENOMEM;
-	}
-
-	async_node->root = root;
-	async_node->delayed_node = curr;
-
-	async_node->work.func = btrfs_async_run_delayed_node_done;
-	async_node->work.flags = 0;
 
-	btrfs_queue_worker(&root->fs_info->delayed_workers, &async_node->work);
-	count++;
-
-	if (all || count < 4)
-		goto again;
+	async_work->delayed_root = delayed_root;
+	async_work->work.func = btrfs_async_run_delayed_root;
+	async_work->work.flags = 0;
+	async_work->nr = nr;
 
+	btrfs_queue_worker(&root->fs_info->delayed_workers, &async_work->work);
 	return 0;
 }
 
@@ -1424,30 +1428,55 @@ void btrfs_assert_delayed_root_empty(struct btrfs_root *root)
 	WARN_ON(btrfs_first_delayed_node(delayed_root));
 }
 
+static int refs_newer(struct btrfs_delayed_root *delayed_root,
+		      int seq, int count)
+{
+	int val = atomic_read(&delayed_root->items_seq);
+
+	if (val < seq || val >= seq + count)
+		return 1;
+	return 0;
+}
+
 void btrfs_balance_delayed_items(struct btrfs_root *root)
 {
 	struct btrfs_delayed_root *delayed_root;
+	int seq;
 
 	delayed_root = btrfs_get_delayed_root(root);
 
 	if (atomic_read(&delayed_root->items) < BTRFS_DELAYED_BACKGROUND)
 		return;
 
+	seq = atomic_read(&delayed_root->items_seq);
+
 	if (atomic_read(&delayed_root->items) >= BTRFS_DELAYED_WRITEBACK) {
 		int ret;
-		ret = btrfs_wq_run_delayed_node(delayed_root, root, 1);
+		DEFINE_WAIT(__wait);
+
+		ret = btrfs_wq_run_delayed_node(delayed_root, root, 0);
 		if (ret)
 			return;
 
-		wait_event_interruptible_timeout(
-				delayed_root->wait,
-				(atomic_read(&delayed_root->items) <
-				 BTRFS_DELAYED_BACKGROUND),
-				HZ);
-		return;
+		while (1) {
+			prepare_to_wait(&delayed_root->wait, &__wait,
+					TASK_INTERRUPTIBLE);
+
+			if (refs_newer(delayed_root, seq,
+				       BTRFS_DELAYED_BATCH) ||
+			    atomic_read(&delayed_root->items) <
+			    BTRFS_DELAYED_BACKGROUND) {
+				break;
+			}
+			if (!signal_pending(current))
+				schedule();
+			else
+				break;
+		}
+		finish_wait(&delayed_root->wait, &__wait);
 	}
 
-	btrfs_wq_run_delayed_node(delayed_root, root, 0);
+	btrfs_wq_run_delayed_node(delayed_root, root, BTRFS_DELAYED_BATCH);
 }
 
 /* Will return 0 or -ENOMEM */