Skip to contentSkip to conversation

block: blk-rq-qos: guard rq-qos helpers by static key

PATCH In review Jens Axboe 3 participants 4 messages Apr 15, 2025 – Apr 17, 2025 Block layer

lore
Share & tools
lore

Jens Axboe

author maintainer
#0 lore raw

PATCH 1/1 block: blk-rq-qos: guard rq-qos helpers by static key

Even if blk-rq-qos isn't used or configured, dipping into the queue to fetch ->rq_qos is a noticeable slowdown and visible in profiles. Add an unlikely static key around blk-rq-qos, to avoid fetching this cacheline if blk-iolatency or blk-wbt isn't configured or used.

---

block/blk-rq-qos.c +4 -0
index 95982bc46ba1..848591fb3c57 100644
@@ -2,6 +2,8 @@
 
 #include "blk-rq-qos.h"
 
+__read_mostly DEFINE_STATIC_KEY_FALSE(block_rq_qos);
+
 /*
  * Increment 'v', if 'v' is below 'below'. Returns true if we succeeded,
  * false if 'v' + 1 would be bigger than 'below'.
@@ -317,6 +319,7 @@ void rq_qos_exit(struct request_queue *q)
 		struct rq_qos *rqos = q->rq_qos;
 		q->rq_qos = rqos->next;
 		rqos->ops->exit(rqos);
+		static_branch_dec(&block_rq_qos);
 	}
 	mutex_unlock(&q->rq_qos_mutex);
 }
@@ -343,6 +346,7 @@ int rq_qos_add(struct rq_qos *rqos, struct gendisk *disk, enum rq_qos_id id,
 		goto ebusy;
 	rqos->next = q->rq_qos;
 	q->rq_qos = rqos;
+	static_branch_inc(&block_rq_qos);
 
 	blk_mq_unfreeze_queue(q, memflags);
 
block/blk-rq-qos.h +12 -9
index 37245c97ee61..39749f4066fb 100644
@@ -12,6 +12,7 @@
 #include "blk-mq-debugfs.h"
 
 struct blk_mq_debugfs_attr;
+extern struct static_key_false block_rq_qos;
 
 enum rq_qos_id {
 	RQ_QOS_WBT,
@@ -112,31 +113,33 @@ void __rq_qos_queue_depth_changed(struct rq_qos *rqos);
 
 static inline void rq_qos_cleanup(struct request_queue *q, struct bio *bio)
 {
-	if (q->rq_qos)
+	if (static_branch_unlikely(&block_rq_qos) && q->rq_qos)
 		__rq_qos_cleanup(q->rq_qos, bio);
 }
 
 static inline void rq_qos_done(struct request_queue *q, struct request *rq)
 {
-	if (q->rq_qos && !blk_rq_is_passthrough(rq))
+	if (static_branch_unlikely(&block_rq_qos) && q->rq_qos &&
+	    !blk_rq_is_passthrough(rq))
 		__rq_qos_done(q->rq_qos, rq);
 }
 
 static inline void rq_qos_issue(struct request_queue *q, struct request *rq)
 {
-	if (q->rq_qos)
+	if (static_branch_unlikely(&block_rq_qos) && q->rq_qos)
 		__rq_qos_issue(q->rq_qos, rq);
 }
 
 static inline void rq_qos_requeue(struct request_queue *q, struct request *rq)
 {
-	if (q->rq_qos)
+	if (static_branch_unlikely(&block_rq_qos) && q->rq_qos)
 		__rq_qos_requeue(q->rq_qos, rq);
 }
 
 static inline void rq_qos_done_bio(struct bio *bio)
 {
-	if (bio->bi_bdev && (bio_flagged(bio, BIO_QOS_THROTTLED) ||
+	if (static_branch_unlikely(&block_rq_qos) &&
+	    bio->bi_bdev && (bio_flagged(bio, BIO_QOS_THROTTLED) ||
 			     bio_flagged(bio, BIO_QOS_MERGED))) {
 		struct request_queue *q = bdev_get_queue(bio->bi_bdev);
 		if (q->rq_qos)
@@ -146,7 +149,7 @@ static inline void rq_qos_done_bio(struct bio *bio)
 
 static inline void rq_qos_throttle(struct request_queue *q, struct bio *bio)
 {
-	if (q->rq_qos) {
+	if (static_branch_unlikely(&block_rq_qos) && q->rq_qos) {
 		bio_set_flag(bio, BIO_QOS_THROTTLED);
 		__rq_qos_throttle(q->rq_qos, bio);
 	}
@@ -155,14 +158,14 @@ static inline void rq_qos_throttle(struct request_queue *q, struct bio *bio)
 static inline void rq_qos_track(struct request_queue *q, struct request *rq,
 				struct bio *bio)
 {
-	if (q->rq_qos)
+	if (static_branch_unlikely(&block_rq_qos) && q->rq_qos)
 		__rq_qos_track(q->rq_qos, rq, bio);
 }
 
 static inline void rq_qos_merge(struct request_queue *q, struct request *rq,
 				struct bio *bio)
 {
-	if (q->rq_qos) {
+	if (static_branch_unlikely(&block_rq_qos) && q->rq_qos) {
 		bio_set_flag(bio, BIO_QOS_MERGED);
 		__rq_qos_merge(q->rq_qos, rq, bio);
 	}
@@ -170,7 +173,7 @@ static inline void rq_qos_merge(struct request_queue *q, struct request *rq,
 
 static inline void rq_qos_queue_depth_changed(struct request_queue *q)
 {
-	if (q->rq_qos)
+	if (static_branch_unlikely(&block_rq_qos) && q->rq_qos)
 		__rq_qos_queue_depth_changed(q->rq_qos);
 }
 
Signature · 1 line
Jens Axboe
Reviewed-by on v1 1/1
#1 lore raw
Quoted 6 lines · Jens Axboe

Even if blk-rq-qos isn't used or configured, dipping into the queue to fetch ->rq_qos is a noticeable slowdown and visible in profiles. Add an unlikely static key around blk-rq-qos, to avoid fetching this cacheline if blk-iolatency or blk-wbt isn't configured or used.

Looks good.

-ck

Reviewed-by on v1 1/1
#2 lore raw
In reply to Jens Axboe
Quoted 9 lines · Jens Axboe

Even if blk-rq-qos isn't used or configured, dipping into the queue to fetch ->rq_qos is a noticeable slowdown and visible in profiles. Add an unlikely static key around blk-rq-qos, to avoid fetching this cacheline if blk-iolatency or blk-wbt isn't configured or used.

---

LGTM

BTW, do we want the same thing for q->td?

Thanks,
Kuai

Quoted 116 lines
block/blk-rq-qos.c +4 -0
index 95982bc46ba1..848591fb3c57 100644
@@ -2,6 +2,8 @@
 
  #include "blk-rq-qos.h"
 
+__read_mostly DEFINE_STATIC_KEY_FALSE(block_rq_qos);
+
  /*
   * Increment 'v', if 'v' is below 'below'. Returns true if we succeeded,
   * false if 'v' + 1 would be bigger than 'below'.
@@ -317,6 +319,7 @@ void rq_qos_exit(struct request_queue *q)
  		struct rq_qos *rqos = q->rq_qos;
  		q->rq_qos = rqos->next;
  		rqos->ops->exit(rqos);
+		static_branch_dec(&block_rq_qos);
  	}
  	mutex_unlock(&q->rq_qos_mutex);
  }
@@ -343,6 +346,7 @@ int rq_qos_add(struct rq_qos *rqos, struct gendisk *disk, enum rq_qos_id id,
  		goto ebusy;
  	rqos->next = q->rq_qos;
  	q->rq_qos = rqos;
+	static_branch_inc(&block_rq_qos);
 
  	blk_mq_unfreeze_queue(q, memflags);
 
block/blk-rq-qos.h +12 -9
index 37245c97ee61..39749f4066fb 100644
@@ -12,6 +12,7 @@
  #include "blk-mq-debugfs.h"
 
  struct blk_mq_debugfs_attr;
+extern struct static_key_false block_rq_qos;
 
  enum rq_qos_id {
  	RQ_QOS_WBT,
@@ -112,31 +113,33 @@ void __rq_qos_queue_depth_changed(struct rq_qos *rqos);
 
  static inline void rq_qos_cleanup(struct request_queue *q, struct bio *bio)
  {
-	if (q->rq_qos)
+	if (static_branch_unlikely(&block_rq_qos) && q->rq_qos)
  		__rq_qos_cleanup(q->rq_qos, bio);
  }
 
  static inline void rq_qos_done(struct request_queue *q, struct request *rq)
  {
-	if (q->rq_qos && !blk_rq_is_passthrough(rq))
+	if (static_branch_unlikely(&block_rq_qos) && q->rq_qos &&
+	    !blk_rq_is_passthrough(rq))
  		__rq_qos_done(q->rq_qos, rq);
  }
 
  static inline void rq_qos_issue(struct request_queue *q, struct request *rq)
  {
-	if (q->rq_qos)
+	if (static_branch_unlikely(&block_rq_qos) && q->rq_qos)
  		__rq_qos_issue(q->rq_qos, rq);
  }
 
  static inline void rq_qos_requeue(struct request_queue *q, struct request *rq)
  {
-	if (q->rq_qos)
+	if (static_branch_unlikely(&block_rq_qos) && q->rq_qos)
  		__rq_qos_requeue(q->rq_qos, rq);
  }
 
  static inline void rq_qos_done_bio(struct bio *bio)
  {
-	if (bio->bi_bdev && (bio_flagged(bio, BIO_QOS_THROTTLED) ||
+	if (static_branch_unlikely(&block_rq_qos) &&
+	    bio->bi_bdev && (bio_flagged(bio, BIO_QOS_THROTTLED) ||
  			     bio_flagged(bio, BIO_QOS_MERGED))) {
  		struct request_queue *q = bdev_get_queue(bio->bi_bdev);
  		if (q->rq_qos)
@@ -146,7 +149,7 @@ static inline void rq_qos_done_bio(struct bio *bio)
 
  static inline void rq_qos_throttle(struct request_queue *q, struct bio *bio)
  {
-	if (q->rq_qos) {
+	if (static_branch_unlikely(&block_rq_qos) && q->rq_qos) {
  		bio_set_flag(bio, BIO_QOS_THROTTLED);
  		__rq_qos_throttle(q->rq_qos, bio);
  	}
@@ -155,14 +158,14 @@ static inline void rq_qos_throttle(struct request_queue *q, struct bio *bio)
  static inline void rq_qos_track(struct request_queue *q, struct request *rq,
  				struct bio *bio)
  {
-	if (q->rq_qos)
+	if (static_branch_unlikely(&block_rq_qos) && q->rq_qos)
  		__rq_qos_track(q->rq_qos, rq, bio);
  }
 
  static inline void rq_qos_merge(struct request_queue *q, struct request *rq,
  				struct bio *bio)
  {
-	if (q->rq_qos) {
+	if (static_branch_unlikely(&block_rq_qos) && q->rq_qos) {
  		bio_set_flag(bio, BIO_QOS_MERGED);
  		__rq_qos_merge(q->rq_qos, rq, bio);
  	}
@@ -170,7 +173,7 @@ static inline void rq_qos_merge(struct request_queue *q, struct request *rq,
 
  static inline void rq_qos_queue_depth_changed(struct request_queue *q)
  {
-	if (q->rq_qos)
+	if (static_branch_unlikely(&block_rq_qos) && q->rq_qos)
  		__rq_qos_queue_depth_changed(q->rq_qos);
  }
 
author maintainer
#3 lore raw
Quoted 8 lines · Yu Kuai
Quoted 7 lines · Jens Axboe

Even if blk-rq-qos isn't used or configured, dipping into the queue to fetch ->rq_qos is a noticeable slowdown and visible in profiles. Add an unlikely static key around blk-rq-qos, to avoid fetching this cacheline if blk-iolatency or blk-wbt isn't configured or used.

Quoted 7 lines
Quoted 2 lines

---

LGTM

BTW, do we want the same thing for q->td?

We may want to do that - I haven't seen that in the hot path yet, but I also think I may have it configured out...

Signature · 1 line
Jens Axboe