Skip to contentSkip to conversation

A few bugfix and cleancode patch for bfq

PATCH New Kemeng Shi 1 participant 11 messages Nov 8, 2022

Subsystems (2)
lore
Share & tools
lore

Kemeng Shi

author
#0 lore raw

cover A few bugfix and cleancode patch for bfq

Hi, this is another series contain some bugfix and cleancode patches to fix temporarily inject limit raise in bfq_choose_bfqq_for_injection, correctly initialize decrease_time_jif to trigger inject limit update in time and some random cleancode.
Thanks!

Kemeng Shi (10):
block, bfq: correctly raise inject limit in
bfq_choose_bfqq_for_injection
block, bfq: remove unsed parameter reason in bfq_bfqq_is_slow block, bfq: initialize bfqq->decrease_time_jif correctly block, bfq: use helper macro RQ_BFQQ to get bfqq of request block, bfq: remove unnecessary pointer reference to get async_bfqq block, bfq: remove redundant bfqd->rq_in_driver > 0 check in
bfq_add_request
block, bfq: remove redundant check in bfq_put_cooperator block, bfq: remove unnecessary goto tag in bfq_dispatch_rq_from_bfqq block, bfq: remove unused bfq_wr_max_time in struct bfq_data block, bfq: remove check of bfq_wr_max_softrt_rate which is always
greater than 0

 block/bfq-iosched.c | 49 +++++++++++++++++----------------------------
 block/bfq-iosched.h |  2 --
 2 files changed, 18 insertions(+), 33 deletions(-)
Signature · 1 line
2.30.0
author
#1 lore raw

PATCH 1/10 block, bfq: correctly raise inject limit in bfq_choose_bfqq_for_injection

Inject limit could be temporarily raised if current inject_limit is 0. raised limit is saved in local variable "limit". The traverse below will reset raised "limit" to bfqd->in_service_queue which is 0 for limit raised condition and will invalidate the raised limit.

After passing bfqd->rq_in_driver >= limit check above, we can be sure about two things in traverse:
1. Local variable "limit" is greater than 0 or bfqd->rq_in_driver >= limit check is always met.
2. For normal case (else case for large request to non-rotational drives), no need to check bfqd->rq_in_driver < limit again if local variable "limit" is not changed.

Fix this by not overwriting local variable "limit" in traverse. As metioned in first thing above that limit is greater than 0, so result of min_t(unsigned int, 1, limit) is always 1. we can simply check whether rq_in_driver is less than 1 for case of large request to non-rotational drives and remove assignment to local variable "limit" in traverse. As metioned in second thing above, in normal case no futher check is needed if local variable "limit" is not chaged, so return directly in normal case.

---
 block/bfq-iosched.c | 10 ++++------
 1 file changed, 4 insertions(+), 6 deletions(-)
block/bfq-iosched.c +4 -6
index 7ea427817f7f..b0bee8ab65e6 100644
@@ -4707,12 +4707,10 @@ bfq_choose_bfqq_for_injection(struct bfq_data *bfqd)
 			 */
 			if (blk_queue_nonrot(bfqd->queue) &&
 			    blk_rq_sectors(bfqq->next_rq) >=
-			    BFQQ_SECT_THR_NONROT)
-				limit = min_t(unsigned int, 1, limit);
-			else
-				limit = in_serv_bfqq->inject_limit;
-
-			if (bfqd->rq_in_driver < limit) {
+			    BFQQ_SECT_THR_NONROT &&
+			    bfqd->rq_in_driver >= 1)
+				continue;
+			else {
 				bfqd->rqs_injected = true;
 				return bfqq;
 			}
Signature · 1 line
2.30.0
author
#2 lore raw

PATCH 2/10 block, bfq: remove unsed parameter reason in bfq_bfqq_is_slow

In reply to Kemeng Shi

Parameter reason is never used, just remove it.

---
 block/bfq-iosched.c | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)
block/bfq-iosched.c +2 -3
index b0bee8ab65e6..fc71181a7e5d 100644
@@ -4132,8 +4132,7 @@ static void __bfq_bfqq_recalc_budget(struct bfq_data *bfqd,
  * function to evaluate the I/O speed of a process.
  */
 static bool bfq_bfqq_is_slow(struct bfq_data *bfqd, struct bfq_queue *bfqq,
-				 bool compensate, enum bfqq_expiration reason,
-				 unsigned long *delta_ms)
+				 bool compensate, unsigned long *delta_ms)
 {
 	ktime_t delta_ktime;
 	u32 delta_usecs;
@@ -4329,7 +4328,7 @@ void bfq_bfqq_expire(struct bfq_data *bfqd,
 	/*
 	 * Check whether the process is slow (see bfq_bfqq_is_slow).
 	 */
-	slow = bfq_bfqq_is_slow(bfqd, bfqq, compensate, reason, &delta);
+	slow = bfq_bfqq_is_slow(bfqd, bfqq, compensate, &delta);
 
 	/*
 	 * As above explained, charge slow (typically seeky) and
Signature · 1 line
2.30.0
author
#3 lore raw

PATCH 3/10 block, bfq: initialize bfqq->decrease_time_jif correctly

In reply to Kemeng Shi

Currently, decrease_time_jif is initialized to 0 and will be updated in function bfq_reset_inject_limit and bfq_update_inject_limit while these two functions are called when jiffies pasts decrease_time_jif with some extra time or think-time state changes. If jiffies is slightly larger than MAX_JIFFY_OFFSET, it will take a long time to meet the first condition. The second condition is heavily relies on process behavior. To be more specific:

Function bfq_update_inject_limit maybe triggered when jiffies pasts decrease_time_jif + msecs_to_jiffies(10) in bfq_add_request by setting bfqd->wait_dispatch to true.

Function bfq_reset_inject_limit are called in two conditions:
1. jiffies pasts bfqq->decrease_time_jif + msecs_to_jiffies(1000) in function bfq_add_request.
2. jiffies pasts bfqq->decrease_time_jif + msecs_to_jiffies(100) or bfq think-time state change from short to long.

In worst case that jiffies is slightly larger than MAX_JIFFY_OFFSET and think-time state never changes, the service injection may be not triggered for a long time.

Fix this by initializing bfqq->decrease_time_jif to current jiffies to trigger service injection soon when service injection conditions are met.

---
 block/bfq-iosched.c | 2 ++
 1 file changed, 2 insertions(+)
block/bfq-iosched.c +2 -0
index fc71181a7e5d..01fa16047eb5 100644
@@ -5557,6 +5557,8 @@ static void bfq_init_bfqq(struct bfq_data *bfqd, struct bfq_queue *bfqq,
 
 	/* first request is almost certainly seeky */
 	bfqq->seek_history = 1;
+
+	bfqq->decrease_time_jif = jiffies;
 }
 
 static struct bfq_queue **bfq_async_queue_prio(struct bfq_data *bfqd,
Signature · 1 line
2.30.0
author
#4 lore raw

PATCH 4/10 block, bfq: use helper macro RQ_BFQQ to get bfqq of request

In reply to Kemeng Shi

Use helper macro RQ_BFQQ to get bfqq of request.

---
 block/bfq-iosched.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
block/bfq-iosched.c +3 -3
index 01fa16047eb5..2c4fe286ead7 100644
@@ -6748,14 +6748,14 @@ static struct bfq_queue *bfq_init_rq(struct request *rq)
 		return NULL;
 
 	/*
-	 * Assuming that elv.priv[1] is set only if everything is set
+	 * Assuming that RQ_BFQQ(rq) is set only if everything is set
 	 * for this rq. This holds true, because this function is
 	 * invoked only for insertion or merging, and, after such
 	 * events, a request cannot be manipulated any longer before
 	 * being removed from bfq.
 	 */
-	if (rq->elv.priv[1])
-		return rq->elv.priv[1];
+	if (RQ_BFQQ(rq))
+		return RQ_BFQQ(rq);
 
 	bic = icq_to_bic(rq->elv.icq);
 
Signature · 1 line
2.30.0
author
#5 lore raw

PATCH 5/10 block, bfq: remove unnecessary pointer reference to get async_bfqq

In reply to Kemeng Shi

The async_bfqq is assigned with bfqq->bic->bfqq[0], use it directly.

---
 block/bfq-iosched.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
block/bfq-iosched.c +1 -1
index 2c4fe286ead7..aad88ca5e5a5 100644
@@ -4901,7 +4901,7 @@ static struct bfq_queue *bfq_select_queue(struct bfq_data *bfqd)
 		    icq_to_bic(async_bfqq->next_rq->elv.icq) == bfqq->bic &&
 		    bfq_serv_to_charge(async_bfqq->next_rq, async_bfqq) <=
 		    bfq_bfqq_budget_left(async_bfqq))
-			bfqq = bfqq->bic->bfqq[0];
+			bfqq = async_bfqq;
 		else if (bfqq->waker_bfqq &&
 			   bfq_bfqq_busy(bfqq->waker_bfqq) &&
 			   bfqq->waker_bfqq->next_rq &&
Signature · 1 line
2.30.0
author
#6 lore raw

PATCH 6/10 block, bfq: remove redundant bfqd->rq_in_driver > 0 check in bfq_add_request

In reply to Kemeng Shi

The bfqd->rq_in_driver > 0 check is along with previous "bfqd->rq_in_driver == 0 ||" check, so no need to re-check bfqd->rq_in_driver > 0.

---
 block/bfq-iosched.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
block/bfq-iosched.c +1 -2
index aad88ca5e5a5..be923be48647 100644
@@ -2254,8 +2254,7 @@ static void bfq_add_request(struct request *rq)
 		 */
 		if (bfqq == bfqd->in_service_queue &&
 		    (bfqd->rq_in_driver == 0 ||
-		     (bfqq->last_serv_time_ns > 0 &&
-		      bfqd->rqs_injected && bfqd->rq_in_driver > 0)) &&
+		     (bfqq->last_serv_time_ns > 0 && bfqd->rqs_injected)) &&
 		    time_is_before_eq_jiffies(bfqq->decrease_time_jif +
 					      msecs_to_jiffies(10))) {
 			bfqd->last_empty_occupied_ns = ktime_get_ns();
Signature · 1 line
2.30.0
author
#7 lore raw

PATCH 7/10 block, bfq: remove redundant check in bfq_put_cooperator

In reply to Kemeng Shi

We have already avoided a circular list in bfq_setup_merge (see comments in bfq_setup_merge() for details), so bfq_queue will not appear in it's new_bfqq list. Just remove this check.

---
 block/bfq-iosched.c | 2 --
 1 file changed, 2 deletions(-)
block/bfq-iosched.c +0 -2
index be923be48647..cffc5ff66ee6 100644
@@ -5339,8 +5339,6 @@ void bfq_put_cooperator(struct bfq_queue *bfqq)
 	 */
 	__bfqq = bfqq->new_bfqq;
 	while (__bfqq) {
-		if (__bfqq == bfqq)
-			break;
 		next = __bfqq->new_bfqq;
 		bfq_put_queue(__bfqq);
 		__bfqq = next;
Signature · 1 line
2.30.0
author
#8 lore raw

PATCH 8/10 block, bfq: remove unnecessary goto tag in bfq_dispatch_rq_from_bfqq

In reply to Kemeng Shi

We jump to tag only for returning current rq. Return directly to remove this tag.

---
 block/bfq-iosched.c | 9 +++------
 1 file changed, 3 insertions(+), 6 deletions(-)
block/bfq-iosched.c +3 -6
index cffc5ff66ee6..64c1249d8eda 100644
@@ -5031,7 +5031,7 @@ static struct request *bfq_dispatch_rq_from_bfqq(struct bfq_data *bfqd,
 	bfq_dispatch_remove(bfqd->queue, rq);
 
 	if (bfqq != bfqd->in_service_queue)
-		goto return_rq;
+		return rq;
 
 	/*
 	 * If weight raising has to terminate for bfqq, then next
@@ -5051,12 +5051,9 @@ static struct request *bfq_dispatch_rq_from_bfqq(struct bfq_data *bfqd,
 	 * belongs to CLASS_IDLE and other queues are waiting for
 	 * service.
 	 */
-	if (!(bfq_tot_busy_queues(bfqd) > 1 && bfq_class_idle(bfqq)))
-		goto return_rq;
+	if ((bfq_tot_busy_queues(bfqd) > 1 && bfq_class_idle(bfqq)))
+		bfq_bfqq_expire(bfqd, bfqq, false, BFQQE_BUDGET_EXHAUSTED);
 
-	bfq_bfqq_expire(bfqd, bfqq, false, BFQQE_BUDGET_EXHAUSTED);
-
-return_rq:
 	return rq;
 }
 
Signature · 1 line
2.30.0
author
#9 lore raw

PATCH 9/10 block, bfq: remove unused bfq_wr_max_time in struct bfq_data

In reply to Kemeng Shi

bfqd->bfq_wr_max_time is set to 0 in bfq_init_queue and is never changed. It is only used in bfq_wr_duration when bfq_wr_max_time > 0 which never meets, so bfqd->bfq_wr_max_time is not used actually. Just remove it.

---
 block/bfq-iosched.c | 4 ----
 block/bfq-iosched.h | 2 --
 2 files changed, 6 deletions(-)
block/bfq-iosched.c +0 -4
index 64c1249d8eda..40ac219b2d8f 100644
@@ -1120,9 +1120,6 @@ static unsigned int bfq_wr_duration(struct bfq_data *bfqd)
 {
 	u64 dur;
 
-	if (bfqd->bfq_wr_max_time > 0)
-		return bfqd->bfq_wr_max_time;
-
 	dur = bfqd->rate_dur_prod;
 	do_div(dur, bfqd->peak_rate);
 
@@ -7138,7 +7135,6 @@ static int bfq_init_queue(struct request_queue *q, struct elevator_type *e)
 	 */
 	bfqd->bfq_wr_coeff = 30;
 	bfqd->bfq_wr_rt_max_time = msecs_to_jiffies(300);
-	bfqd->bfq_wr_max_time = 0;
 	bfqd->bfq_wr_min_idle_time = msecs_to_jiffies(2000);
 	bfqd->bfq_wr_min_inter_arr_async = msecs_to_jiffies(500);
 	bfqd->bfq_wr_max_softrt_rate = 7000; /*
block/bfq-iosched.h +0 -2
index 71f721670ab6..48c147e5f9e0 100644
@@ -719,8 +719,6 @@ struct bfq_data {
 	 * is multiplied.
 	 */
 	unsigned int bfq_wr_coeff;
-	/* maximum duration of a weight-raising period (jiffies) */
-	unsigned int bfq_wr_max_time;
 
 	/* Maximum weight-raising duration for soft real-time processes */
 	unsigned int bfq_wr_rt_max_time;
Signature · 1 line
2.30.0
author
#10 lore raw

PATCH 10/10 block, bfq: remove check of bfq_wr_max_softrt_rate which is always greater than 0

In reply to Kemeng Shi

bfqd->bfq_wr_max_softrt_rate is assigned with 7000 in bfq_init_queue and never changed. So we can remove bfqd->bfq_wr_max_softrt_rate > 0 check which is always true.

---
 block/bfq-iosched.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)
block/bfq-iosched.c +2 -4
index 40ac219b2d8f..d0a867b2569b 100644
@@ -1840,8 +1840,7 @@ static void bfq_bfqq_handle_idle_busy_switch(struct bfq_data *bfqd,
 	 *   to control its weight explicitly)
 	 */
 	in_burst = bfq_bfqq_in_large_burst(bfqq);
-	soft_rt = bfqd->bfq_wr_max_softrt_rate > 0 &&
-		!BFQQ_TOTALLY_SEEKY(bfqq) &&
+	soft_rt = !BFQQ_TOTALLY_SEEKY(bfqq) &&
 		!in_burst &&
 		time_is_before_jiffies(bfqq->soft_rt_next_start) &&
 		bfqq->dispatched == 0 &&
@@ -4350,8 +4349,7 @@ void bfq_bfqq_expire(struct bfq_data *bfqd,
 	if (bfqd->low_latency && bfqq->wr_coeff == 1)
 		bfqq->last_wr_start_finish = jiffies;
 
-	if (bfqd->low_latency && bfqd->bfq_wr_max_softrt_rate > 0 &&
-	    RB_EMPTY_ROOT(&bfqq->sort_list)) {
+	if (bfqd->low_latency && RB_EMPTY_ROOT(&bfqq->sort_list)) {
 		/*
 		 * If we get here, and there are no outstanding
 		 * requests, then the request pattern is isochronous
Signature · 1 line
2.30.0