Merge branch 'for-3.17/core' of git://git.kernel.dk/linux-block

Pull block core bits from Jens Axboe:
 "Small round this time, after the massive blk-mq dump for 3.16.  This
  pull request contains:

   - Fixes for max_sectors overflow in ioctls from Akinoby Mita.

   - Partition off-by-one bug fix in aix partitions from Dan Carpenter.

   - Various small partition cleanups from Fabian Frederick.

   - Fix for the block integrity code sometimes returning the wrong
     vector count from Gu Zheng.

   - Cleanup an re-org of the blk-mq queue enter/exit percpu counters
     from Tejun.  Dependent on the percpu pull for 3.17 (which was in
     the block tree too), that you have already pulled in.

   - A blkcg oops fix, also from Tejun"

* 'for-3.17/core' of git://git.kernel.dk/linux-block:
  partitions: aix.c: off by one bug
  blkcg: don't call into policy draining if root_blkg is already gone
  Revert "bio: modify __bio_add_page() to accept pages that don't start a new segment"
  bio: modify __bio_add_page() to accept pages that don't start a new segment
  block: fix SG_[GS]ET_RESERVED_SIZE ioctl when max_sectors is huge
  block: fix BLKSECTGET ioctl when max_sectors is greater than USHRT_MAX
  block/partitions/efi.c: kerneldoc fixing
  block/partitions/msdos.c: code clean-up
  block/partitions/amiga.c: replace nolevel printk by pr_err
  block/partitions/aix.c: replace count*size kzalloc by kcalloc
  bio-integrity: add "bip_max_vcnt" into struct bio_integrity_payload
  blk-mq: use percpu_ref for mq usage count
  blk-mq: collapse __blk_mq_drain_queue() into blk_mq_freeze_queue()
  blk-mq: decouble blk-mq freezing from generic bypassing
  block, blk-mq: draining can't be skipped even if bypass_depth was non-zero
  blk-mq: fix a memory ordering bug in blk_mq_queue_enter()

1 files changed, 29 insertions, 52 deletions

diff --git a/block/blk-mq.c b/block/blk-mq.c
index ad69ef657e85..5189cb1e478a 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -78,68 +78,47 @@ static void blk_mq_hctx_clear_pending(struct blk_mq_hw_ctx *hctx,
 
 static int blk_mq_queue_enter(struct request_queue *q)
 {
-	int ret;
-
-	__percpu_counter_add(&q->mq_usage_counter, 1, 1000000);
-	smp_wmb();
-
-	/* we have problems freezing the queue if it's initializing */
-	if (!blk_queue_dying(q) &&
-	    (!blk_queue_bypass(q) || !blk_queue_init_done(q)))
-		return 0;
-
-	__percpu_counter_add(&q->mq_usage_counter, -1, 1000000);
+	while (true) {
+		int ret;
 
-	spin_lock_irq(q->queue_lock);
-	ret = wait_event_interruptible_lock_irq(q->mq_freeze_wq,
-		!blk_queue_bypass(q) || blk_queue_dying(q),
-		*q->queue_lock);
-	/* inc usage with lock hold to avoid freeze_queue runs here */
-	if (!ret && !blk_queue_dying(q))
-		__percpu_counter_add(&q->mq_usage_counter, 1, 1000000);
-	else if (blk_queue_dying(q))
-		ret = -ENODEV;
-	spin_unlock_irq(q->queue_lock);
+		if (percpu_ref_tryget_live(&q->mq_usage_counter))
+			return 0;
 
-	return ret;
+		ret = wait_event_interruptible(q->mq_freeze_wq,
+				!q->mq_freeze_depth || blk_queue_dying(q));
+		if (blk_queue_dying(q))
+			return -ENODEV;
+		if (ret)
+			return ret;
+	}
 }
 
 static void blk_mq_queue_exit(struct request_queue *q)
 {
-	__percpu_counter_add(&q->mq_usage_counter, -1, 1000000);
+	percpu_ref_put(&q->mq_usage_counter);
 }
 
-void blk_mq_drain_queue(struct request_queue *q)
+static void blk_mq_usage_counter_release(struct percpu_ref *ref)
 {
-	while (true) {
-		s64 count;
-
-		spin_lock_irq(q->queue_lock);
-		count = percpu_counter_sum(&q->mq_usage_counter);
-		spin_unlock_irq(q->queue_lock);
+	struct request_queue *q =
+		container_of(ref, struct request_queue, mq_usage_counter);
 
-		if (count == 0)
-			break;
-		blk_mq_start_hw_queues(q);
-		msleep(10);
-	}
+	wake_up_all(&q->mq_freeze_wq);
 }
 
 /*
  * Guarantee no request is in use, so we can change any data structure of
  * the queue afterward.
  */
-static void blk_mq_freeze_queue(struct request_queue *q)
+void blk_mq_freeze_queue(struct request_queue *q)
 {
-	bool drain;
-
 	spin_lock_irq(q->queue_lock);
-	drain = !q->bypass_depth++;
-	queue_flag_set(QUEUE_FLAG_BYPASS, q);
+	q->mq_freeze_depth++;
 	spin_unlock_irq(q->queue_lock);
 
-	if (drain)
-		blk_mq_drain_queue(q);
+	percpu_ref_kill(&q->mq_usage_counter);
+	blk_mq_run_queues(q, false);
+	wait_event(q->mq_freeze_wq, percpu_ref_is_zero(&q->mq_usage_counter));
 }
 
 static void blk_mq_unfreeze_queue(struct request_queue *q)
@@ -147,14 +126,13 @@ static void blk_mq_unfreeze_queue(struct request_queue *q)
 	bool wake = false;
 
 	spin_lock_irq(q->queue_lock);
-	if (!--q->bypass_depth) {
-		queue_flag_clear(QUEUE_FLAG_BYPASS, q);
-		wake = true;
-	}
-	WARN_ON_ONCE(q->bypass_depth < 0);
+	wake = !--q->mq_freeze_depth;
+	WARN_ON_ONCE(q->mq_freeze_depth < 0);
 	spin_unlock_irq(q->queue_lock);
-	if (wake)
+	if (wake) {
+		percpu_ref_reinit(&q->mq_usage_counter);
 		wake_up_all(&q->mq_freeze_wq);
+	}
 }
 
 bool blk_mq_can_queue(struct blk_mq_hw_ctx *hctx)
@@ -1798,7 +1776,7 @@ struct request_queue *blk_mq_init_queue(struct blk_mq_tag_set *set)
 	if (!q)
 		goto err_hctxs;
 
-	if (percpu_counter_init(&q->mq_usage_counter, 0))
+	if (percpu_ref_init(&q->mq_usage_counter, blk_mq_usage_counter_release))
 		goto err_map;
 
 	setup_timer(&q->timeout, blk_mq_rq_timer, (unsigned long) q);
@@ -1891,7 +1869,7 @@ void blk_mq_free_queue(struct request_queue *q)
 	blk_mq_exit_hw_queues(q, set, set->nr_hw_queues);
 	blk_mq_free_hw_queues(q, set);
 
-	percpu_counter_destroy(&q->mq_usage_counter);
+	percpu_ref_exit(&q->mq_usage_counter);
 
 	free_percpu(q->queue_ctx);
 	kfree(q->queue_hw_ctx);
@@ -2050,8 +2028,7 @@ static int __init blk_mq_init(void)
 {
 	blk_mq_cpu_init();
 
-	/* Must be called after percpu_counter_hotcpu_callback() */
-	hotcpu_notifier(blk_mq_queue_reinit_notify, -10);
+	hotcpu_notifier(blk_mq_queue_reinit_notify, 0);
 
 	return 0;
 }