blk-mq: improve support for shared tags maps

This adds support for active queue tracking, meaning that the
blk-mq tagging maintains a count of active users of a tag set.
This allows us to maintain a notion of fairness between users,
so that we can distribute the tag depth evenly without starving
some users while allowing others to try unfair deep queues.

If sharing of a tag set is detected, each hardware queue will
track the depth of its own queue. And if this exceeds the total
depth divided by the number of active queues, the user is actively
throttled down.

The active queue count is done lazily to avoid bouncing that data
between submitter and completer. Each hardware queue gets marked
active when it allocates its first tag, and gets marked inactive
when 1) the last tag is cleared, and 2) the queue timeout grace
period has passed.

Signed-off-by: Jens Axboe <axboe@fb.com>

3 files changed, 12 insertions, 0 deletions

diff --git a/include/linux/blk-mq.h b/include/linux/blk-mq.h
index f83d15f6e1c1..379f88d5c44d 100644
--- a/include/linux/blk-mq.h
+++ b/include/linux/blk-mq.h
@@ -48,6 +48,8 @@ struct blk_mq_hw_ctx {
 	unsigned int		numa_node;
 	unsigned int		cmd_size;	/* per-request extra data */
 
+	atomic_t		nr_active;
+
 	struct blk_mq_cpu_notifier	cpu_notifier;
 	struct kobject		kobj;
 };
@@ -64,6 +66,9 @@ struct blk_mq_tag_set {
 	void			*driver_data;
 
 	struct blk_mq_tags	**tags;
+
+	struct mutex		tag_list_lock;
+	struct list_head	tag_list;
 };
 
 typedef int (queue_rq_fn)(struct blk_mq_hw_ctx *, struct request *);
@@ -126,8 +131,10 @@ enum {
 
 	BLK_MQ_F_SHOULD_MERGE	= 1 << 0,
 	BLK_MQ_F_SHOULD_SORT	= 1 << 1,
+	BLK_MQ_F_TAG_SHARED	= 1 << 2,
 
 	BLK_MQ_S_STOPPED	= 0,
+	BLK_MQ_S_TAG_ACTIVE	= 1,
 
 	BLK_MQ_MAX_DEPTH	= 2048,
 
diff --git a/include/linux/blk_types.h b/include/linux/blk_types.h
index aa0eaa2d0bd8..d8e4cea23a25 100644
--- a/include/linux/blk_types.h
+++ b/include/linux/blk_types.h
@@ -190,6 +190,7 @@ enum rq_flag_bits {
 	__REQ_PM,		/* runtime pm request */
 	__REQ_END,		/* last of chain of requests */
 	__REQ_HASHED,		/* on IO scheduler merge hash */
+	__REQ_MQ_INFLIGHT,	/* track inflight for MQ */
 	__REQ_NR_BITS,		/* stops here */
 };
 
@@ -243,5 +244,6 @@ enum rq_flag_bits {
 #define REQ_PM			(1ULL << __REQ_PM)
 #define REQ_END			(1ULL << __REQ_END)
 #define REQ_HASHED		(1ULL << __REQ_HASHED)
+#define REQ_MQ_INFLIGHT		(1ULL << __REQ_MQ_INFLIGHT)
 
 #endif /* __LINUX_BLK_TYPES_H */
diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
index 94b27210641b..6bc011a09e82 100644
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -481,6 +481,9 @@ struct request_queue {
 	wait_queue_head_t	mq_freeze_wq;
 	struct percpu_counter	mq_usage_counter;
 	struct list_head	all_q_node;
+
+	struct blk_mq_tag_set	*tag_set;
+	struct list_head	tag_set_list;
 };
 
 #define QUEUE_FLAG_QUEUED	1	/* uses generic tag queueing */