io-uring.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [FIO PATCH] engines/io_uring: add sqthread_poll_percpu option
@ 2020-09-10  7:05 Xiaoguang Wang
  2020-09-10  7:07 ` Xiaoguang Wang
  0 siblings, 1 reply; 4+ messages in thread
From: Xiaoguang Wang @ 2020-09-10  7:05 UTC (permalink / raw)
  To: io-uring; +Cc: axboe, asml.silence, joseph.qi, Xiaoguang Wang

This option is only meaningful when sqthread_poll and sqthread_poll_cpu
are both set. If this option is effective, for multiple io_uring instances
which are all bound to one same cpu, only a kernel thread is created for
this cpu to perform these io_uring instances' submission queue polling.

Signed-off-by: Xiaoguang Wang <xiaoguang.wang@linux.alibaba.com>
---
 engines/io_uring.c  | 13 +++++++++++++
 fio.1               |  6 ++++++
 os/linux/io_uring.h |  1 +
 3 files changed, 20 insertions(+)

diff --git a/engines/io_uring.c b/engines/io_uring.c
index ec8cb18a..8a447b62 100644
--- a/engines/io_uring.c
+++ b/engines/io_uring.c
@@ -77,6 +77,7 @@ struct ioring_options {
 	unsigned int fixedbufs;
 	unsigned int registerfiles;
 	unsigned int sqpoll_thread;
+	unsigned int sqpoll_thread_percpu;
 	unsigned int sqpoll_set;
 	unsigned int sqpoll_cpu;
 	unsigned int nonvectored;
@@ -160,6 +161,16 @@ static struct fio_option options[] = {
 		.category = FIO_OPT_C_ENGINE,
 		.group	= FIO_OPT_G_IOURING,
 	},
+	{
+		.name	= "sqthread_poll_percpu",
+		.lname	= "Kernel percpu SQ thread polling",
+		.type	= FIO_OPT_INT,
+		.off1	= offsetof(struct ioring_options, sqpoll_thread_percpu),
+		.help	= "Offload submission/completion to kernel thread, use percpu thread",
+		.category = FIO_OPT_C_ENGINE,
+		.group	= FIO_OPT_G_IOURING,
+	},
+
 	{
 		.name	= "sqthread_poll_cpu",
 		.lname	= "SQ Thread Poll CPU",
@@ -596,6 +607,8 @@ static int fio_ioring_queue_init(struct thread_data *td)
 		p.flags |= IORING_SETUP_IOPOLL;
 	if (o->sqpoll_thread) {
 		p.flags |= IORING_SETUP_SQPOLL;
+		if (o->sqpoll_thread_percpu)
+			p.flags |= IORING_SETUP_SQPOLL_PERCPU;
 		if (o->sqpoll_set) {
 			p.flags |= IORING_SETUP_SQ_AFF;
 			p.sq_thread_cpu = o->sqpoll_cpu;
diff --git a/fio.1 b/fio.1
index 1c90e4a5..d367134f 100644
--- a/fio.1
+++ b/fio.1
@@ -1851,6 +1851,12 @@ the cost of using more CPU in the system.
 When `sqthread_poll` is set, this option provides a way to define which CPU
 should be used for the polling thread.
 .TP
+.BI (io_uring)sqthread_poll_percpu
+This option is only meaningful when `sqthread_poll` and `sqthread_poll_cpu` are
+both set. If this option is effective, for multiple io_uring instances which are all
+bound to one same cpu, only a kernel thread is created for this cpu to perform these
+io_uring instances' submission queue polling.
+.TP
 .BI (libaio)userspace_reap
 Normally, with the libaio engine in use, fio will use the
 \fBio_getevents\fR\|(3) system call to reap newly returned events. With
diff --git a/os/linux/io_uring.h b/os/linux/io_uring.h
index d39b45fd..0a4c8a35 100644
--- a/os/linux/io_uring.h
+++ b/os/linux/io_uring.h
@@ -99,6 +99,7 @@ enum {
 #define IORING_SETUP_CQSIZE	(1U << 3)	/* app defines CQ size */
 #define IORING_SETUP_CLAMP	(1U << 4)	/* clamp SQ/CQ ring sizes */
 #define IORING_SETUP_ATTACH_WQ	(1U << 5)	/* attach to existing wq */
+#define IORING_SETUP_SQPOLL_PERCPU (1U << 7)       /* sq_thread_cpu is valid */
 
 enum {
 	IORING_OP_NOP,
-- 
2.17.2


^ permalink raw reply related	[flat|nested] 4+ messages in thread

* Re: [FIO PATCH] engines/io_uring: add sqthread_poll_percpu option
  2020-09-10  7:05 [FIO PATCH] engines/io_uring: add sqthread_poll_percpu option Xiaoguang Wang
@ 2020-09-10  7:07 ` Xiaoguang Wang
  0 siblings, 0 replies; 4+ messages in thread
From: Xiaoguang Wang @ 2020-09-10  7:07 UTC (permalink / raw)
  To: io-uring; +Cc: axboe, asml.silence, joseph.qi

hi,

You can use this patch to test IORING_SETUP_SQPOLL_PERCPU flag introduced in
patch " io_uring: support multiple rings to share same poll thread by specifying same cpu".
Below is a example fio job:
[global]
ioengine=io_uring
sqthread_poll=1
sqthread_poll_percpu=1
registerfiles=1
fixedbufs=1
hipri=0
thread=1
bs=4k
direct=1
rw=randread
time_based=1
runtime=60
randrepeat=0
filename=/dev/vdb

[job1]
cpus_allowed=2
sqthread_poll_cpu=17
iodepth=128

[job2]
cpus_allowed=3
sqthread_poll_cpu=17
iodepth=128

[job3]
cpus_allowed=4
sqthread_poll_cpu=17
iodepth=128

[job4]
cpus_allowed=5
sqthread_poll_cpu=17
iodepth=128

Regards,
Xiaoguang Wang

> This option is only meaningful when sqthread_poll and sqthread_poll_cpu
> are both set. If this option is effective, for multiple io_uring instances
> which are all bound to one same cpu, only a kernel thread is created for
> this cpu to perform these io_uring instances' submission queue polling.
> 
> Signed-off-by: Xiaoguang Wang <xiaoguang.wang@linux.alibaba.com>
> ---
>   engines/io_uring.c  | 13 +++++++++++++
>   fio.1               |  6 ++++++
>   os/linux/io_uring.h |  1 +
>   3 files changed, 20 insertions(+)
> 
> diff --git a/engines/io_uring.c b/engines/io_uring.c
> index ec8cb18a..8a447b62 100644
> --- a/engines/io_uring.c
> +++ b/engines/io_uring.c
> @@ -77,6 +77,7 @@ struct ioring_options {
>   	unsigned int fixedbufs;
>   	unsigned int registerfiles;
>   	unsigned int sqpoll_thread;
> +	unsigned int sqpoll_thread_percpu;
>   	unsigned int sqpoll_set;
>   	unsigned int sqpoll_cpu;
>   	unsigned int nonvectored;
> @@ -160,6 +161,16 @@ static struct fio_option options[] = {
>   		.category = FIO_OPT_C_ENGINE,
>   		.group	= FIO_OPT_G_IOURING,
>   	},
> +	{
> +		.name	= "sqthread_poll_percpu",
> +		.lname	= "Kernel percpu SQ thread polling",
> +		.type	= FIO_OPT_INT,
> +		.off1	= offsetof(struct ioring_options, sqpoll_thread_percpu),
> +		.help	= "Offload submission/completion to kernel thread, use percpu thread",
> +		.category = FIO_OPT_C_ENGINE,
> +		.group	= FIO_OPT_G_IOURING,
> +	},
> +
>   	{
>   		.name	= "sqthread_poll_cpu",
>   		.lname	= "SQ Thread Poll CPU",
> @@ -596,6 +607,8 @@ static int fio_ioring_queue_init(struct thread_data *td)
>   		p.flags |= IORING_SETUP_IOPOLL;
>   	if (o->sqpoll_thread) {
>   		p.flags |= IORING_SETUP_SQPOLL;
> +		if (o->sqpoll_thread_percpu)
> +			p.flags |= IORING_SETUP_SQPOLL_PERCPU;
>   		if (o->sqpoll_set) {
>   			p.flags |= IORING_SETUP_SQ_AFF;
>   			p.sq_thread_cpu = o->sqpoll_cpu;
> diff --git a/fio.1 b/fio.1
> index 1c90e4a5..d367134f 100644
> --- a/fio.1
> +++ b/fio.1
> @@ -1851,6 +1851,12 @@ the cost of using more CPU in the system.
>   When `sqthread_poll` is set, this option provides a way to define which CPU
>   should be used for the polling thread.
>   .TP
> +.BI (io_uring)sqthread_poll_percpu
> +This option is only meaningful when `sqthread_poll` and `sqthread_poll_cpu` are
> +both set. If this option is effective, for multiple io_uring instances which are all
> +bound to one same cpu, only a kernel thread is created for this cpu to perform these
> +io_uring instances' submission queue polling.
> +.TP
>   .BI (libaio)userspace_reap
>   Normally, with the libaio engine in use, fio will use the
>   \fBio_getevents\fR\|(3) system call to reap newly returned events. With
> diff --git a/os/linux/io_uring.h b/os/linux/io_uring.h
> index d39b45fd..0a4c8a35 100644
> --- a/os/linux/io_uring.h
> +++ b/os/linux/io_uring.h
> @@ -99,6 +99,7 @@ enum {
>   #define IORING_SETUP_CQSIZE	(1U << 3)	/* app defines CQ size */
>   #define IORING_SETUP_CLAMP	(1U << 4)	/* clamp SQ/CQ ring sizes */
>   #define IORING_SETUP_ATTACH_WQ	(1U << 5)	/* attach to existing wq */
> +#define IORING_SETUP_SQPOLL_PERCPU (1U << 7)       /* sq_thread_cpu is valid */
>   
>   enum {
>   	IORING_OP_NOP,
> 

^ permalink raw reply	[flat|nested] 4+ messages in thread

* [FIO PATCH] engines/io_uring: add sqthread_poll_percpu option
@ 2021-03-31 16:01 Xiaoguang Wang
  0 siblings, 0 replies; 4+ messages in thread
From: Xiaoguang Wang @ 2021-03-31 16:01 UTC (permalink / raw)
  To: io-uring; +Cc: axboe, asml.silence, joseph.qi

This option is only meaningful when sqthread_poll and sqthread_poll_cpu
are both set. If this option is effective, for multiple io_uring instances
which are all bound to one same cpu, only one kernel thread is created for
this cpu to perform these io_uring instances' submission queue polling.

Signed-off-by: Xiaoguang Wang <xiaoguang.wang@linux.alibaba.com>
---
 engines/io_uring.c  | 12 ++++++++++++
 fio.1               |  6 ++++++
 os/linux/io_uring.h |  1 +
 3 files changed, 19 insertions(+)

diff --git a/engines/io_uring.c b/engines/io_uring.c
index b962e804..638e5b71 100644
--- a/engines/io_uring.c
+++ b/engines/io_uring.c
@@ -78,6 +78,7 @@ struct ioring_options {
 	unsigned int fixedbufs;
 	unsigned int registerfiles;
 	unsigned int sqpoll_thread;
+	unsigned int sqpoll_thread_percpu;
 	unsigned int sqpoll_set;
 	unsigned int sqpoll_cpu;
 	unsigned int nonvectored;
@@ -162,6 +163,15 @@ static struct fio_option options[] = {
 		.category = FIO_OPT_C_ENGINE,
 		.group	= FIO_OPT_G_IOURING,
 	},
+	{
+		.name   = "sqthread_poll_percpu",
+		.lname  = "Kernel percpu SQ thread polling",
+		.type   = FIO_OPT_INT,
+		.off1   = offsetof(struct ioring_options, sqpoll_thread_percpu),
+		.help   = "Offload submission/completion to kernel thread, use percpu thread",
+		.category = FIO_OPT_C_ENGINE,
+		.group  = FIO_OPT_G_IOURING,
+	},
 	{
 		.name	= "sqthread_poll_cpu",
 		.lname	= "SQ Thread Poll CPU",
@@ -615,6 +625,8 @@ static int fio_ioring_queue_init(struct thread_data *td)
 		p.flags |= IORING_SETUP_IOPOLL;
 	if (o->sqpoll_thread) {
 		p.flags |= IORING_SETUP_SQPOLL;
+		if (o->sqpoll_thread_percpu)
+			p.flags |= IORING_SETUP_SQ_PERCPU;
 		if (o->sqpoll_set) {
 			p.flags |= IORING_SETUP_SQ_AFF;
 			p.sq_thread_cpu = o->sqpoll_cpu;
diff --git a/fio.1 b/fio.1
index ad4a662b..7925a511 100644
--- a/fio.1
+++ b/fio.1
@@ -1925,6 +1925,12 @@ the cost of using more CPU in the system.
 When `sqthread_poll` is set, this option provides a way to define which CPU
 should be used for the polling thread.
 .TP
+.BI (io_uring)sqthread_poll_percpu
+This option is only meaningful when `sqthread_poll` and `sqthread_poll_cpu` are
+both set. If this option is effective, for multiple io_uring instances which are all
+bound to one same cpu, only one polling thread in the kernel is created to perform
+these io_uring instances' submission queue polling.
+.TP
 .BI (libaio)userspace_reap
 Normally, with the libaio engine in use, fio will use the
 \fBio_getevents\fR\|(3) system call to reap newly returned events. With
diff --git a/os/linux/io_uring.h b/os/linux/io_uring.h
index d39b45fd..7839e487 100644
--- a/os/linux/io_uring.h
+++ b/os/linux/io_uring.h
@@ -99,6 +99,7 @@ enum {
 #define IORING_SETUP_CQSIZE	(1U << 3)	/* app defines CQ size */
 #define IORING_SETUP_CLAMP	(1U << 4)	/* clamp SQ/CQ ring sizes */
 #define IORING_SETUP_ATTACH_WQ	(1U << 5)	/* attach to existing wq */
+#define IORING_SETUP_SQ_PERCPU	(1U << 7)	/* use percpu SQ poll thread */
 
 enum {
 	IORING_OP_NOP,
-- 
2.17.2


^ permalink raw reply related	[flat|nested] 4+ messages in thread

* [FIO PATCH] engines/io_uring: add sqthread_poll_percpu option
@ 2020-11-03  6:19 Xiaoguang Wang
  0 siblings, 0 replies; 4+ messages in thread
From: Xiaoguang Wang @ 2020-11-03  6:19 UTC (permalink / raw)
  To: io-uring; +Cc: axboe, joseph.qi

This option is only meaningful when sqthread_poll and sqthread_poll_cpu
are both set. If this option is effective, for multiple io_uring instances
which are all bound to one same cpu, only a kernel thread is created for
this cpu to perform these io_uring instances' submission queue polling.

Signed-off-by: Xiaoguang Wang <xiaoguang.wang@linux.alibaba.com>
---
 engines/io_uring.c  | 13 +++++++++++++
 fio.1               |  6 ++++++
 os/linux/io_uring.h |  1 +
 3 files changed, 20 insertions(+)

diff --git a/engines/io_uring.c b/engines/io_uring.c
index ec8cb18a..8a447b62 100644
--- a/engines/io_uring.c
+++ b/engines/io_uring.c
@@ -77,6 +77,7 @@ struct ioring_options {
 	unsigned int fixedbufs;
 	unsigned int registerfiles;
 	unsigned int sqpoll_thread;
+	unsigned int sqpoll_thread_percpu;
 	unsigned int sqpoll_set;
 	unsigned int sqpoll_cpu;
 	unsigned int nonvectored;
@@ -160,6 +161,16 @@ static struct fio_option options[] = {
 		.category = FIO_OPT_C_ENGINE,
 		.group	= FIO_OPT_G_IOURING,
 	},
+	{
+		.name	= "sqthread_poll_percpu",
+		.lname	= "Kernel percpu SQ thread polling",
+		.type	= FIO_OPT_INT,
+		.off1	= offsetof(struct ioring_options, sqpoll_thread_percpu),
+		.help	= "Offload submission/completion to kernel thread, use percpu thread",
+		.category = FIO_OPT_C_ENGINE,
+		.group	= FIO_OPT_G_IOURING,
+	},
+
 	{
 		.name	= "sqthread_poll_cpu",
 		.lname	= "SQ Thread Poll CPU",
@@ -596,6 +607,8 @@ static int fio_ioring_queue_init(struct thread_data *td)
 		p.flags |= IORING_SETUP_IOPOLL;
 	if (o->sqpoll_thread) {
 		p.flags |= IORING_SETUP_SQPOLL;
+		if (o->sqpoll_thread_percpu)
+			p.flags |= IORING_SETUP_SQPOLL_PERCPU;
 		if (o->sqpoll_set) {
 			p.flags |= IORING_SETUP_SQ_AFF;
 			p.sq_thread_cpu = o->sqpoll_cpu;
diff --git a/fio.1 b/fio.1
index 1c90e4a5..d367134f 100644
--- a/fio.1
+++ b/fio.1
@@ -1851,6 +1851,12 @@ the cost of using more CPU in the system.
 When `sqthread_poll` is set, this option provides a way to define which CPU
 should be used for the polling thread.
 .TP
+.BI (io_uring)sqthread_poll_percpu
+This option is only meaningful when `sqthread_poll` and `sqthread_poll_cpu` are
+both set. If this option is effective, for multiple io_uring instances which are all
+bound to one same cpu, only a kernel thread is created for this cpu to perform these
+io_uring instances' submission queue polling.
+.TP
 .BI (libaio)userspace_reap
 Normally, with the libaio engine in use, fio will use the
 \fBio_getevents\fR\|(3) system call to reap newly returned events. With
diff --git a/os/linux/io_uring.h b/os/linux/io_uring.h
index d39b45fd..0a4c8a35 100644
--- a/os/linux/io_uring.h
+++ b/os/linux/io_uring.h
@@ -99,6 +99,7 @@ enum {
 #define IORING_SETUP_CQSIZE	(1U << 3)	/* app defines CQ size */
 #define IORING_SETUP_CLAMP	(1U << 4)	/* clamp SQ/CQ ring sizes */
 #define IORING_SETUP_ATTACH_WQ	(1U << 5)	/* attach to existing wq */
+#define IORING_SETUP_SQPOLL_PERCPU (1U << 7)       /* sq_thread_cpu is valid */
 
 enum {
 	IORING_OP_NOP,
-- 
2.17.2


^ permalink raw reply related	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2021-03-31 16:02 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-09-10  7:05 [FIO PATCH] engines/io_uring: add sqthread_poll_percpu option Xiaoguang Wang
2020-09-10  7:07 ` Xiaoguang Wang
2020-11-03  6:19 Xiaoguang Wang
2021-03-31 16:01 Xiaoguang Wang

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).