Skip to contentSkip to conversation

blk-mq: optimize the size of struct request

PATCH Superseded [email protected] 3 participants 14 messages Jun 29, 2023 – Jul 6, 2023 Block layer

v2 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

v2 cover blk-mq: optimize the size of struct request

From: Chengming Zhou <[email protected]>

v2:
- Change to use call_single_data_t, which use __aligned() to avoid
to use 2 cache lines for 1 csd. Thanks Ming Lei.
- [v1] https://lore.kernel.org/all/[email protected]/

Hello,

After the commit be4c427809b0 ("blk-mq: use the I/O scheduler for writes from the flush state machine"), rq->flush can't reuse rq->elv anymore, since flush_data requests can go into io scheduler now.

That increased the size of struct request by 24 bytes, but this patchset can decrease the size by 40 bytes, which is good I think.

patch 1 use percpu csd to do remote complete instead of per-rq csd, decrease the size by 24 bytes.

patch 2-3 reuse rq->queuelist in flush state machine pending list, and maintain a u64 counter of inflight flush_data requests, decrease the size by 16 bytes.

patch 4 is just cleanup by the way.

Thanks for comments!

Chengming Zhou (4):
blk-mq: use percpu csd to remote complete instead of per-rq csd
blk-flush: count inflight flush_data requests
blk-flush: reuse rq queuelist in flush state machine
blk-mq: delete unused completion_data in struct request

 block/blk-flush.c      | 19 +++++++++----------
 block/blk-mq.c         | 12 ++++++++----
 block/blk.h            |  5 ++---
 include/linux/blk-mq.h | 10 ++--------
 4 files changed, 21 insertions(+), 25 deletions(-)
Signature · 1 line
2.39.2
#1 lore raw

PATCH v2 1/4 blk-mq: use percpu csd to remote complete instead of per-rq csd

From: Chengming Zhou <[email protected]>

If request need to be completed remotely, we insert it into percpu llist, and smp_call_function_single_async() if llist is empty previously.

We don't need to use per-rq csd, percpu csd is enough. And the size of struct request is decreased by 24 bytes.

This way is cleaner, and looks correct, given block softirq is guaranteed to be scheduled to consume the list if one new request is added to this percpu list, either smp_call_function_single_async() returns -EBUSY or 0.

---
v2:
- Change to use call_single_data_t, which avoid to use 2 cache lines for 1 csd, as suggested by Ming Lei.
- Improve the commit log, the explanation is copied from Ming Lei.

---
 block/blk-mq.c         | 12 ++++++++----
 include/linux/blk-mq.h |  5 +----
 2 files changed, 9 insertions(+), 8 deletions(-)
block/blk-mq.c +8 -4
index decb6ab2d508..e52200edd2b1 100644
@@ -43,6 +43,7 @@
 #include "blk-ioprio.h"
 
 static DEFINE_PER_CPU(struct llist_head, blk_cpu_done);
+static DEFINE_PER_CPU(call_single_data_t, blk_cpu_csd);
 
 static void blk_mq_insert_request(struct request *rq, blk_insert_t flags);
 static void blk_mq_request_bypass_insert(struct request *rq,
@@ -1156,13 +1157,13 @@ static void blk_mq_complete_send_ipi(struct request *rq)
 {
 	struct llist_head *list;
 	unsigned int cpu;
+	call_single_data_t *csd;
 
 	cpu = rq->mq_ctx->cpu;
 	list = &per_cpu(blk_cpu_done, cpu);
-	if (llist_add(&rq->ipi_list, list)) {
-		INIT_CSD(&rq->csd, __blk_mq_complete_request_remote, rq);
-		smp_call_function_single_async(cpu, &rq->csd);
-	}
+	csd = &per_cpu(blk_cpu_csd, cpu);
+	if (llist_add(&rq->ipi_list, list))
+		smp_call_function_single_async(cpu, csd);
 }
 
 static void blk_mq_raise_softirq(struct request *rq)
@@ -4796,6 +4797,9 @@ static int __init blk_mq_init(void)
 
 	for_each_possible_cpu(i)
 		init_llist_head(&per_cpu(blk_cpu_done, i));
+	for_each_possible_cpu(i)
+		INIT_CSD(&per_cpu(blk_cpu_csd, i),
+			 __blk_mq_complete_request_remote, NULL);
 	open_softirq(BLOCK_SOFTIRQ, blk_done_softirq);
 
 	cpuhp_setup_state_nocalls(CPUHP_BLOCK_SOFTIRQ_DEAD,
include/linux/blk-mq.h +1 -4
index f401067ac03a..070551197c0e 100644
@@ -182,10 +182,7 @@ struct request {
 		rq_end_io_fn		*saved_end_io;
 	} flush;
 
-	union {
-		struct __call_single_data csd;
-		u64 fifo_time;
-	};
+	u64 fifo_time;
 
 	/*
 	 * completion callback.
Signature · 1 line
2.39.2
Reviewed-by on v2 1/4
#2 lore raw

Re: [PATCH v2 1/4] blk-mq: use percpu csd to remote complete instead of per-rq csd

Quoted 12 lines · On Thu, Jun 29, 2023 at 07:03:56PM +0800, [email protected] wrote:

From: Chengming Zhou <[email protected]>

If request need to be completed remotely, we insert it into percpu llist, and smp_call_function_single_async() if llist is empty previously.

We don't need to use per-rq csd, percpu csd is enough. And the size of struct request is decreased by 24 bytes.

This way is cleaner, and looks correct, given block softirq is guaranteed to be scheduled to consume the list if one new request is added to this percpu list, either smp_call_function_single_async() returns -EBUSY or 0.

Quoted 6 lines

---
v2:
- Change to use call_single_data_t, which avoid to use 2 cache lines for 1 csd, as suggested by Ming Lei.
- Improve the commit log, the explanation is copied from Ming Lei.

Thanks,
Ming

#3 lore raw

Re: [PATCH v2 1/4] blk-mq: use percpu csd to remote complete instead of per-rq csd

In reply to [email protected]
Quoted 11 lines · On Thu, Jun 29, 2023 at 07:03:56PM +0800, [email protected] wrote:

From: Chengming Zhou <[email protected]>

If request need to be completed remotely, we insert it into percpu llist, and smp_call_function_single_async() if llist is empty previously.

We don't need to use per-rq csd, percpu csd is enough. And the size of struct request is decreased by 24 bytes.

This way is cleaner, and looks correct, given block softirq is guaranteed to be scheduled to consume the list if one new request is added to this percpu list, either smp_call_function_single_async() returns -EBUSY or 0.

Please trim your commit logs to 73 characters per line so that they are readable in git log output.

Quoted 7 lines

static void blk_mq_request_bypass_insert(struct request *rq,

block/blk-mq.c +1 -0
@@ -1156,13 +1157,13 @@ static void blk_mq_complete_send_ipi(struct request *rq)
 {
 	struct llist_head *list;
 	unsigned int cpu;
+	call_single_data_t *csd;
 

block/blk-mq.c static void blk_mq_complete_send_ipi(struct request *rq) · line 1167

Quoted 10 lines
block/blk-mq.c +3 -4
 	cpu = rq->mq_ctx->cpu;
 	list = &per_cpu(blk_cpu_done, cpu);
-	if (llist_add(&rq->ipi_list, list)) {
-		INIT_CSD(&rq->csd, __blk_mq_complete_request_remote, rq);
-		smp_call_function_single_async(cpu, &rq->csd);
-	}
+	csd = &per_cpu(blk_cpu_csd, cpu);
+	if (llist_add(&rq->ipi_list, list))
+		smp_call_function_single_async(cpu, csd);
 }

No need for the list and csd variables here as they are only used once.

But I think this code has a rpboem when it is preemptd between the llist_add and smp_call_function_single_async. We either need a get_cpu/put_cpu around them, or instroduce a structure with the list and csd, and then you can use one pointer from per_cpu and still ensure the list and csd are for the same CPU.

#4 lore raw

Re: [PATCH v2 1/4] blk-mq: use percpu csd to remote complete instead of per-rq csd

Quoted 9 lines · Christoph Hellwig
Quoted 8 lines · On Thu, Jun 29, 2023 at 07:03:56PM +0800, [email protected] wrote:

From: Chengming Zhou <[email protected]>

If request need to be completed remotely, we insert it into percpu llist, and smp_call_function_single_async() if llist is empty previously.

We don't need to use per-rq csd, percpu csd is enough. And the size of struct request is decreased by 24 bytes.

Quoted 6 lines
Quoted 3 lines

This way is cleaner, and looks correct, given block softirq is guaranteed to be scheduled to consume the list if one new request is added to this percpu list, either smp_call_function_single_async() returns -EBUSY or 0.

Please trim your commit logs to 73 characters per line so that they are readable in git log output.

Ok, will fix in the next version.

Quoted 8 lines
Quoted 8 lines

static void blk_mq_request_bypass_insert(struct request *rq,

block/blk-mq.c +1 -0
@@ -1156,13 +1157,13 @@ static void blk_mq_complete_send_ipi(struct request *rq)
 {
 	struct llist_head *list;
 	unsigned int cpu;
+	call_single_data_t *csd;
 
Quoted 13 lines

block/blk-mq.c static void blk_mq_complete_send_ipi(struct request *rq) · line 1167

Quoted 10 lines
block/blk-mq.c +3 -4
 	cpu = rq->mq_ctx->cpu;
 	list = &per_cpu(blk_cpu_done, cpu);
-	if (llist_add(&rq->ipi_list, list)) {
-		INIT_CSD(&rq->csd, __blk_mq_complete_request_remote, rq);
-		smp_call_function_single_async(cpu, &rq->csd);
-	}
+	csd = &per_cpu(blk_cpu_csd, cpu);
+	if (llist_add(&rq->ipi_list, list))
+		smp_call_function_single_async(cpu, csd);
 }

No need for the list and csd variables here as they are only used once.

Yes, should I change like below? Looks like much long code. :-)

if (llist_add(&rq->ipi_list, &per_cpu(blk_cpu_done, cpu)))

	smp_call_function_single_async(cpu, &per_cpu(blk_cpu_csd, cpu));
Quoted 7 lines

But I think this code has a rpboem when it is preemptd between the llist_add and smp_call_function_single_async. We either need a get_cpu/put_cpu around them, or instroduce a structure with the list and csd, and then you can use one pointer from per_cpu and still ensure the list and csd are for the same CPU.

cpu = rq->mq_ctx->cpu; So it's certainly the same CPU, right?

Thanks!

#5 lore raw

Re: [PATCH v2 1/4] blk-mq: use percpu csd to remote complete instead of per-rq csd

Quoted 4 lines · Chengming Zhou

Yes, should I change like below? Looks like much long code. :-)

if (llist_add(&rq->ipi_list, &per_cpu(blk_cpu_done, cpu)))

	smp_call_function_single_async(cpu, &per_cpu(blk_cpu_csd, cpu));

Doesn't look bad too me.

Quoted 11 lines
Quoted 9 lines

But I think this code has a rpboem when it is preemptd between the llist_add and smp_call_function_single_async. We either need a get_cpu/put_cpu around them, or instroduce a structure with the list and csd, and then you can use one pointer from per_cpu and still ensure the list and csd are for the same CPU.

cpu = rq->mq_ctx->cpu; So it's certainly the same CPU, right?

You're right of couse - cpu is the submitting cpu and not the current one and thus not affected by preemption. Sorry for the noise.

#6 lore raw

PATCH v2 2/4 blk-flush: count inflight flush_data requests

In reply to [email protected]

From: Chengming Zhou <[email protected]>

The flush state machine use a double list to link all inflight flush_data requests, to avoid issuing separate post-flushes for these flush_data requests which shared PREFLUSH.

So we can't reuse rq->queuelist, this is why we need rq->flush.list

In preparation of the next patch that reuse rq->queuelist for flush state machine, we change the double linked list to unsigned long counter, which count all inflight flush_data requests.

This is ok since we only need to know if there is any inflight flush_data request, so unsigned long counter is good.

---
 block/blk-flush.c | 9 +++++----
 block/blk.h       | 5 ++---
 2 files changed, 7 insertions(+), 7 deletions(-)
block/blk-flush.c +5 -4
index dba392cf22be..bb7adfc2a5da 100644
@@ -187,7 +187,8 @@ static void blk_flush_complete_seq(struct request *rq,
 		break;
 
 	case REQ_FSEQ_DATA:
-		list_move_tail(&rq->flush.list, &fq->flush_data_in_flight);
+		list_del_init(&rq->flush.list);
+		fq->flush_data_in_flight++;
 		spin_lock(&q->requeue_lock);
 		list_add_tail(&rq->queuelist, &q->flush_list);
 		spin_unlock(&q->requeue_lock);
@@ -299,7 +300,7 @@ static void blk_kick_flush(struct request_queue *q, struct blk_flush_queue *fq,
 		return;
 
 	/* C2 and C3 */
-	if (!list_empty(&fq->flush_data_in_flight) &&
+	if (fq->flush_data_in_flight &&
 	    time_before(jiffies,
 			fq->flush_pending_since + FLUSH_PENDING_TIMEOUT))
 		return;
@@ -374,6 +375,7 @@ static enum rq_end_io_ret mq_flush_data_end_io(struct request *rq,
 	 * the comment in flush_end_io().
 	 */
 	spin_lock_irqsave(&fq->mq_flush_lock, flags);
+	fq->flush_data_in_flight--;
 	blk_flush_complete_seq(rq, fq, REQ_FSEQ_DATA, error);
 	spin_unlock_irqrestore(&fq->mq_flush_lock, flags);
 
@@ -445,7 +447,7 @@ bool blk_insert_flush(struct request *rq)
 		blk_rq_init_flush(rq);
 		rq->flush.seq |= REQ_FSEQ_POSTFLUSH;
 		spin_lock_irq(&fq->mq_flush_lock);
-		list_move_tail(&rq->flush.list, &fq->flush_data_in_flight);
+		fq->flush_data_in_flight++;
 		spin_unlock_irq(&fq->mq_flush_lock);
 		return false;
 	default:
@@ -496,7 +498,6 @@ struct blk_flush_queue *blk_alloc_flush_queue(int node, int cmd_size,
 
 	INIT_LIST_HEAD(&fq->flush_queue[0]);
 	INIT_LIST_HEAD(&fq->flush_queue[1]);
-	INIT_LIST_HEAD(&fq->flush_data_in_flight);
 
 	return fq;
 
block/blk.h +2 -3
index 608c5dcc516b..686712e13835 100644
@@ -15,15 +15,14 @@ struct elevator_type;
 extern struct dentry *blk_debugfs_root;
 
 struct blk_flush_queue {
+	spinlock_t		mq_flush_lock;
 	unsigned int		flush_pending_idx:1;
 	unsigned int		flush_running_idx:1;
 	blk_status_t 		rq_status;
 	unsigned long		flush_pending_since;
 	struct list_head	flush_queue[2];
-	struct list_head	flush_data_in_flight;
+	unsigned long		flush_data_in_flight;
 	struct request		*flush_rq;
-
-	spinlock_t		mq_flush_lock;
 };
 
 bool is_flush_rq(struct request *req);
Signature · 1 line
2.39.2
Reviewed-by on v2 2/4
#7 lore raw

Re: [PATCH v2 2/4] blk-flush: count inflight flush_data requests

Quoted 78 lines · On Thu, Jun 29, 2023 at 07:03:57PM +0800, [email protected] wrote:

From: Chengming Zhou <[email protected]>

The flush state machine use a double list to link all inflight flush_data requests, to avoid issuing separate post-flushes for these flush_data requests which shared PREFLUSH.

So we can't reuse rq->queuelist, this is why we need rq->flush.list

In preparation of the next patch that reuse rq->queuelist for flush state machine, we change the double linked list to unsigned long counter, which count all inflight flush_data requests.

This is ok since we only need to know if there is any inflight flush_data request, so unsigned long counter is good.

---
 block/blk-flush.c | 9 +++++----
 block/blk.h       | 5 ++---
 2 files changed, 7 insertions(+), 7 deletions(-)
block/blk-flush.c +5 -4
index dba392cf22be..bb7adfc2a5da 100644
@@ -187,7 +187,8 @@ static void blk_flush_complete_seq(struct request *rq,
 		break;
 
 	case REQ_FSEQ_DATA:
-		list_move_tail(&rq->flush.list, &fq->flush_data_in_flight);
+		list_del_init(&rq->flush.list);
+		fq->flush_data_in_flight++;
 		spin_lock(&q->requeue_lock);
 		list_add_tail(&rq->queuelist, &q->flush_list);
 		spin_unlock(&q->requeue_lock);
@@ -299,7 +300,7 @@ static void blk_kick_flush(struct request_queue *q, struct blk_flush_queue *fq,
 		return;
 
 	/* C2 and C3 */
-	if (!list_empty(&fq->flush_data_in_flight) &&
+	if (fq->flush_data_in_flight &&
 	    time_before(jiffies,
 			fq->flush_pending_since + FLUSH_PENDING_TIMEOUT))
 		return;
@@ -374,6 +375,7 @@ static enum rq_end_io_ret mq_flush_data_end_io(struct request *rq,
 	 * the comment in flush_end_io().
 	 */
 	spin_lock_irqsave(&fq->mq_flush_lock, flags);
+	fq->flush_data_in_flight--;
 	blk_flush_complete_seq(rq, fq, REQ_FSEQ_DATA, error);
 	spin_unlock_irqrestore(&fq->mq_flush_lock, flags);
 
@@ -445,7 +447,7 @@ bool blk_insert_flush(struct request *rq)
 		blk_rq_init_flush(rq);
 		rq->flush.seq |= REQ_FSEQ_POSTFLUSH;
 		spin_lock_irq(&fq->mq_flush_lock);
-		list_move_tail(&rq->flush.list, &fq->flush_data_in_flight);
+		fq->flush_data_in_flight++;
 		spin_unlock_irq(&fq->mq_flush_lock);
 		return false;
 	default:
@@ -496,7 +498,6 @@ struct blk_flush_queue *blk_alloc_flush_queue(int node, int cmd_size,
 
 	INIT_LIST_HEAD(&fq->flush_queue[0]);
 	INIT_LIST_HEAD(&fq->flush_queue[1]);
-	INIT_LIST_HEAD(&fq->flush_data_in_flight);
 
 	return fq;
 
block/blk.h +1 -0
index 608c5dcc516b..686712e13835 100644
@@ -15,15 +15,14 @@ struct elevator_type;
 extern struct dentry *blk_debugfs_root;
 
 struct blk_flush_queue {
+	spinlock_t		mq_flush_lock;

block/blk.h struct elevator_type; · line 26

Quoted 10 lines
block/blk.h +1 -3
 	unsigned int		flush_pending_idx:1;
 	unsigned int		flush_running_idx:1;
 	blk_status_t 		rq_status;
 	unsigned long		flush_pending_since;
 	struct list_head	flush_queue[2];
-	struct list_head	flush_data_in_flight;
+	unsigned long		flush_data_in_flight;
 	struct request		*flush_rq;
-
-	spinlock_t		mq_flush_lock;

Looks fine,

Thanks,
Ming

#9 lore raw

PATCH v2 3/4 blk-flush: reuse rq queuelist in flush state machine

In reply to [email protected]

From: Chengming Zhou <[email protected]>

Since we don't need to maintain inflight flush_data requests list anymore, we can reuse rq->queuelist for flush pending list.

This patch decrease the size of struct request by 16 bytes.

---
 block/blk-flush.c      | 12 +++++-------
 include/linux/blk-mq.h |  1 -
 2 files changed, 5 insertions(+), 8 deletions(-)
block/blk-flush.c +5 -7
index bb7adfc2a5da..81588edbe8b0 100644
@@ -183,14 +183,13 @@ static void blk_flush_complete_seq(struct request *rq,
 		/* queue for flush */
 		if (list_empty(pending))
 			fq->flush_pending_since = jiffies;
-		list_move_tail(&rq->flush.list, pending);
+		list_move_tail(&rq->queuelist, pending);
 		break;
 
 	case REQ_FSEQ_DATA:
-		list_del_init(&rq->flush.list);
 		fq->flush_data_in_flight++;
 		spin_lock(&q->requeue_lock);
-		list_add_tail(&rq->queuelist, &q->flush_list);
+		list_move_tail(&rq->queuelist, &q->flush_list);
 		spin_unlock(&q->requeue_lock);
 		blk_mq_kick_requeue_list(q);
 		break;
@@ -202,7 +201,7 @@ static void blk_flush_complete_seq(struct request *rq,
 		 * flush data request completion path.  Restore @rq for
 		 * normal completion and end it.
 		 */
-		list_del_init(&rq->flush.list);
+		list_del_init(&rq->queuelist);
 		blk_flush_restore_request(rq);
 		blk_mq_end_request(rq, error);
 		break;
@@ -258,7 +257,7 @@ static enum rq_end_io_ret flush_end_io(struct request *flush_rq,
 	fq->flush_running_idx ^= 1;
 
 	/* and push the waiting requests to the next stage */
-	list_for_each_entry_safe(rq, n, running, flush.list) {
+	list_for_each_entry_safe(rq, n, running, queuelist) {
 		unsigned int seq = blk_flush_cur_seq(rq);
 
 		BUG_ON(seq != REQ_FSEQ_PREFLUSH && seq != REQ_FSEQ_POSTFLUSH);
@@ -292,7 +291,7 @@ static void blk_kick_flush(struct request_queue *q, struct blk_flush_queue *fq,
 {
 	struct list_head *pending = &fq->flush_queue[fq->flush_pending_idx];
 	struct request *first_rq =
-		list_first_entry(pending, struct request, flush.list);
+		list_first_entry(pending, struct request, queuelist);
 	struct request *flush_rq = fq->flush_rq;
 
 	/* C1 described at the top of this file */
@@ -386,7 +385,6 @@ static enum rq_end_io_ret mq_flush_data_end_io(struct request *rq,
 static void blk_rq_init_flush(struct request *rq)
 {
 	rq->flush.seq = 0;
-	INIT_LIST_HEAD(&rq->flush.list);
 	rq->rq_flags |= RQF_FLUSH_SEQ;
 	rq->flush.saved_end_io = rq->end_io; /* Usually NULL */
 	rq->end_io = mq_flush_data_end_io;
include/linux/blk-mq.h +0 -1
index 070551197c0e..96644d6f8d18 100644
@@ -178,7 +178,6 @@ struct request {
 
 	struct {
 		unsigned int		seq;
-		struct list_head	list;
 		rq_end_io_fn		*saved_end_io;
 	} flush;
 
Signature · 1 line
2.39.2
#11 lore raw

PATCH v2 4/4 blk-mq: delete unused completion_data in struct request

In reply to [email protected]

From: Chengming Zhou <[email protected]>

After global search, I found "completion_data" in struct request is not used anywhere, so just clean it up by the way.

---
 include/linux/blk-mq.h | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)
include/linux/blk-mq.h +1 -3
index 96644d6f8d18..ab790eba5fcf 100644
@@ -158,13 +158,11 @@ struct request {
 
 	/*
 	 * The rb_node is only used inside the io scheduler, requests
-	 * are pruned when moved to the dispatch queue. So let the
-	 * completion_data share space with the rb_node.
+	 * are pruned when moved to the dispatch queue.
 	 */
 	union {
 		struct rb_node rb_node;	/* sort/lookup */
 		struct bio_vec special_vec;
-		void *completion_data;
 	};
 
 	/*
Signature · 1 line
2.39.2
#13 lore raw

Re: [PATCH v2 4/4] blk-mq: delete unused completion_data in struct request

Quoted 6 lines · Christoph Hellwig

Looks good:

Although I'd order it first or even send it separately as it doesn't need any of the other changes.

Ok, will order it first in the next version.

I thought it's too minor to send a separate patch, so put it in this series.

Thanks.