All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH V2] scsi: put hot fields of scsi_host_template into one cacheline
@ 2020-04-22  9:54 Ming Lei
  2020-04-22 10:06 ` John Garry
                   ` (4 more replies)
  0 siblings, 5 replies; 9+ messages in thread
From: Ming Lei @ 2020-04-22  9:54 UTC (permalink / raw)
  To: James Bottomley, linux-scsi, Martin K . Petersen
  Cc: Ming Lei, Bart Van Assche, Christoph Hellwig, John Garry

The following three fields of scsi_host_template are referenced in
scsi IO submission path, so put them together into one cacheline:

- cmd_size
- queuecommand
- commit_rqs

Cc: Bart Van Assche <bvanassche@acm.org>
Cc: Christoph Hellwig <hch@lst.de>
Cc: John Garry <john.garry@huawei.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
---
V2:
	- move the 3 fields at the beginning of scsi_host_template
	- comment why we do this way

 include/scsi/scsi_host.h | 72 ++++++++++++++++++++++------------------
 1 file changed, 39 insertions(+), 33 deletions(-)

diff --git a/include/scsi/scsi_host.h b/include/scsi/scsi_host.h
index 822e8cda8d9b..eae5b9f52dfe 100644
--- a/include/scsi/scsi_host.h
+++ b/include/scsi/scsi_host.h
@@ -30,40 +30,15 @@ struct scsi_transport_template;
 #define MODE_TARGET 0x02
 
 struct scsi_host_template {
-	struct module *module;
-	const char *name;
-
 	/*
-	 * The info function will return whatever useful information the
-	 * developer sees fit.  If not provided, then the name field will
-	 * be used instead.
-	 *
-	 * Status: OPTIONAL
+	 * Put fields referenced in IO submission path together in
+	 * same cacheline
 	 */
-	const char *(* info)(struct Scsi_Host *);
 
 	/*
-	 * Ioctl interface
-	 *
-	 * Status: OPTIONAL
-	 */
-	int (*ioctl)(struct scsi_device *dev, unsigned int cmd,
-		     void __user *arg);
-
-
-#ifdef CONFIG_COMPAT
-	/* 
-	 * Compat handler. Handle 32bit ABI.
-	 * When unknown ioctl is passed return -ENOIOCTLCMD.
-	 *
-	 * Status: OPTIONAL
+	 * Additional per-command data allocated for the driver.
 	 */
-	int (*compat_ioctl)(struct scsi_device *dev, unsigned int cmd,
-			    void __user *arg);
-#endif
-
-	int (*init_cmd_priv)(struct Scsi_Host *shost, struct scsi_cmnd *cmd);
-	int (*exit_cmd_priv)(struct Scsi_Host *shost, struct scsi_cmnd *cmd);
+	unsigned int cmd_size;
 
 	/*
 	 * The queuecommand function is used to queue up a scsi
@@ -111,6 +86,41 @@ struct scsi_host_template {
 	 */
 	void (*commit_rqs)(struct Scsi_Host *, u16);
 
+	struct module *module;
+	const char *name;
+
+	/*
+	 * The info function will return whatever useful information the
+	 * developer sees fit.  If not provided, then the name field will
+	 * be used instead.
+	 *
+	 * Status: OPTIONAL
+	 */
+	const char *(* info)(struct Scsi_Host *);
+
+	/*
+	 * Ioctl interface
+	 *
+	 * Status: OPTIONAL
+	 */
+	int (*ioctl)(struct scsi_device *dev, unsigned int cmd,
+		     void __user *arg);
+
+
+#ifdef CONFIG_COMPAT
+	/* 
+	 * Compat handler. Handle 32bit ABI.
+	 * When unknown ioctl is passed return -ENOIOCTLCMD.
+	 *
+	 * Status: OPTIONAL
+	 */
+	int (*compat_ioctl)(struct scsi_device *dev, unsigned int cmd,
+			    void __user *arg);
+#endif
+
+	int (*init_cmd_priv)(struct Scsi_Host *shost, struct scsi_cmnd *cmd);
+	int (*exit_cmd_priv)(struct Scsi_Host *shost, struct scsi_cmnd *cmd);
+
 	/*
 	 * This is an error handling strategy routine.  You don't need to
 	 * define one of these if you don't want to - there is a default
@@ -468,10 +478,6 @@ struct scsi_host_template {
 	 */
 	u64 vendor_id;
 
-	/*
-	 * Additional per-command data allocated for the driver.
-	 */
-	unsigned int cmd_size;
 	struct scsi_host_cmd_pool *cmd_pool;
 
 	/* Delay for runtime autosuspend */
-- 
2.25.2


^ permalink raw reply related	[flat|nested] 9+ messages in thread

* Re: [PATCH V2] scsi: put hot fields of scsi_host_template into one cacheline
  2020-04-22  9:54 [PATCH V2] scsi: put hot fields of scsi_host_template into one cacheline Ming Lei
@ 2020-04-22 10:06 ` John Garry
  2020-04-22 14:36 ` Bart Van Assche
                   ` (3 subsequent siblings)
  4 siblings, 0 replies; 9+ messages in thread
From: John Garry @ 2020-04-22 10:06 UTC (permalink / raw)
  To: Ming Lei, James Bottomley, linux-scsi, Martin K . Petersen
  Cc: Bart Van Assche, Christoph Hellwig

On 22/04/2020 10:54, Ming Lei wrote:
> The following three fields of scsi_host_template are referenced in
> scsi IO submission path, so put them together into one cacheline:
> 
> - cmd_size
> - queuecommand
> - commit_rqs
> 
> Cc: Bart Van Assche <bvanassche@acm.org>
> Cc: Christoph Hellwig <hch@lst.de>
> Cc: John Garry <john.garry@huawei.com>
> Signed-off-by: Ming Lei <ming.lei@redhat.com>

FWIW,
Reviewed-by: John Garry <john.garry@huawei.com>

> ---
> V2:
> 	- move the 3 fields at the beginning of scsi_host_template
> 	- comment why we do this way
> 
>   include/scsi/scsi_host.h | 72 ++++++++++++++++++++++------------------
>   1 file changed, 39 insertions(+), 33 deletions(-)
> 
> diff --git a/include/scsi/scsi_host.h b/include/scsi/scsi_host.h
> index 822e8cda8d9b..eae5b9f52dfe 100644
> --- a/include/scsi/scsi_host.h
> +++ b/include/scsi/scsi_host.h
> @@ -30,40 +30,15 @@ struct scsi_transport_template;
>   #define MODE_TARGET 0x02
>   
>   struct scsi_host_template {
> -	struct module *module;
> -	const char *name;
> -
>   	/*
> -	 * The info function will return whatever useful information the
> -	 * developer sees fit.  If not provided, then the name field will
> -	 * be used instead.
> -	 *
> -	 * Status: OPTIONAL
> +	 * Put fields referenced in IO submission path together in
> +	 * same cacheline
>   	 */
> -	const char *(* info)(struct Scsi_Host *);
>   
>   	/*
> -	 * Ioctl interface
> -	 *
> -	 * Status: OPTIONAL
> -	 */
> -	int (*ioctl)(struct scsi_device *dev, unsigned int cmd,
> -		     void __user *arg);
> -
> -
> -#ifdef CONFIG_COMPAT
> -	/*
> -	 * Compat handler. Handle 32bit ABI.
> -	 * When unknown ioctl is passed return -ENOIOCTLCMD.
> -	 *
> -	 * Status: OPTIONAL
> +	 * Additional per-command data allocated for the driver.
>   	 */
> -	int (*compat_ioctl)(struct scsi_device *dev, unsigned int cmd,
> -			    void __user *arg);
> -#endif
> -
> -	int (*init_cmd_priv)(struct Scsi_Host *shost, struct scsi_cmnd *cmd);
> -	int (*exit_cmd_priv)(struct Scsi_Host *shost, struct scsi_cmnd *cmd);
> +	unsigned int cmd_size;
>   
>   	/*
>   	 * The queuecommand function is used to queue up a scsi
> @@ -111,6 +86,41 @@ struct scsi_host_template {
>   	 */
>   	void (*commit_rqs)(struct Scsi_Host *, u16);
>   
> +	struct module *module;
> +	const char *name;
> +
> +	/*
> +	 * The info function will return whatever useful information the
> +	 * developer sees fit.  If not provided, then the name field will
> +	 * be used instead.
> +	 *
> +	 * Status: OPTIONAL
> +	 */
> +	const char *(* info)(struct Scsi_Host *);
> +
> +	/*
> +	 * Ioctl interface
> +	 *
> +	 * Status: OPTIONAL
> +	 */
> +	int (*ioctl)(struct scsi_device *dev, unsigned int cmd,
> +		     void __user *arg);
> +
> +
> +#ifdef CONFIG_COMPAT
> +	/*
> +	 * Compat handler. Handle 32bit ABI.
> +	 * When unknown ioctl is passed return -ENOIOCTLCMD.
> +	 *
> +	 * Status: OPTIONAL
> +	 */
> +	int (*compat_ioctl)(struct scsi_device *dev, unsigned int cmd,
> +			    void __user *arg);
> +#endif
> +
> +	int (*init_cmd_priv)(struct Scsi_Host *shost, struct scsi_cmnd *cmd);
> +	int (*exit_cmd_priv)(struct Scsi_Host *shost, struct scsi_cmnd *cmd);
> +
>   	/*
>   	 * This is an error handling strategy routine.  You don't need to
>   	 * define one of these if you don't want to - there is a default
> @@ -468,10 +478,6 @@ struct scsi_host_template {
>   	 */
>   	u64 vendor_id;
>   
> -	/*
> -	 * Additional per-command data allocated for the driver.
> -	 */
> -	unsigned int cmd_size;
>   	struct scsi_host_cmd_pool *cmd_pool;
>   
>   	/* Delay for runtime autosuspend */
> 


^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [PATCH V2] scsi: put hot fields of scsi_host_template into one cacheline
  2020-04-22  9:54 [PATCH V2] scsi: put hot fields of scsi_host_template into one cacheline Ming Lei
  2020-04-22 10:06 ` John Garry
@ 2020-04-22 14:36 ` Bart Van Assche
  2020-04-22 14:41 ` James Bottomley
                   ` (2 subsequent siblings)
  4 siblings, 0 replies; 9+ messages in thread
From: Bart Van Assche @ 2020-04-22 14:36 UTC (permalink / raw)
  To: Ming Lei, James Bottomley, linux-scsi, Martin K . Petersen
  Cc: Christoph Hellwig, John Garry

On 4/22/20 2:54 AM, Ming Lei wrote:
> The following three fields of scsi_host_template are referenced in
> scsi IO submission path, so put them together into one cacheline:
> 
> - cmd_size
> - queuecommand
> - commit_rqs

Reviewed-by: Bart Van Assche <bvanassche@acm.org>

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [PATCH V2] scsi: put hot fields of scsi_host_template into one cacheline
  2020-04-22  9:54 [PATCH V2] scsi: put hot fields of scsi_host_template into one cacheline Ming Lei
  2020-04-22 10:06 ` John Garry
  2020-04-22 14:36 ` Bart Van Assche
@ 2020-04-22 14:41 ` James Bottomley
  2020-04-23  1:48   ` Ming Lei
  2020-04-23  6:02 ` Christoph Hellwig
  2020-05-01 22:27 ` Douglas Gilbert
  4 siblings, 1 reply; 9+ messages in thread
From: James Bottomley @ 2020-04-22 14:41 UTC (permalink / raw)
  To: Ming Lei, linux-scsi, Martin K . Petersen
  Cc: Bart Van Assche, Christoph Hellwig, John Garry

On Wed, 2020-04-22 at 17:54 +0800, Ming Lei wrote:
> The following three fields of scsi_host_template are referenced in
> scsi IO submission path, so put them together into one cacheline:
> 
> - cmd_size
> - queuecommand
> - commit_rqs

Are there benchmarks to show this actually makes a difference, if so,
how much?

James


^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [PATCH V2] scsi: put hot fields of scsi_host_template into one cacheline
  2020-04-22 14:41 ` James Bottomley
@ 2020-04-23  1:48   ` Ming Lei
  0 siblings, 0 replies; 9+ messages in thread
From: Ming Lei @ 2020-04-23  1:48 UTC (permalink / raw)
  To: James Bottomley
  Cc: linux-scsi, Martin K . Petersen, Bart Van Assche,
	Christoph Hellwig, John Garry

On Wed, Apr 22, 2020 at 07:41:30AM -0700, James Bottomley wrote:
> On Wed, 2020-04-22 at 17:54 +0800, Ming Lei wrote:
> > The following three fields of scsi_host_template are referenced in
> > scsi IO submission path, so put them together into one cacheline:
> > 
> > - cmd_size
> > - queuecommand
> > - commit_rqs
> 
> Are there benchmarks to show this actually makes a difference, if so,
> how much?

Originally I observed 40% IOPS boost on scsi_debug by moving the three
fields into scsi_host, but I lost that test environment now.

When I run fio on scsi_debug in another machine, IOPS gets ~10% boost
with this patch.

Thanks,
Ming


^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [PATCH V2] scsi: put hot fields of scsi_host_template into one cacheline
  2020-04-22  9:54 [PATCH V2] scsi: put hot fields of scsi_host_template into one cacheline Ming Lei
                   ` (2 preceding siblings ...)
  2020-04-22 14:41 ` James Bottomley
@ 2020-04-23  6:02 ` Christoph Hellwig
  2020-04-24 18:14   ` Martin K. Petersen
  2020-05-01 22:27 ` Douglas Gilbert
  4 siblings, 1 reply; 9+ messages in thread
From: Christoph Hellwig @ 2020-04-23  6:02 UTC (permalink / raw)
  To: Ming Lei
  Cc: James Bottomley, linux-scsi, Martin K . Petersen,
	Bart Van Assche, Christoph Hellwig, John Garry

Looks good, although you might want to throw in a blurb on the
actually measured benefits:

Reviewed-by: Christoph Hellwig <hch@lst.de>

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [PATCH V2] scsi: put hot fields of scsi_host_template into one cacheline
  2020-04-23  6:02 ` Christoph Hellwig
@ 2020-04-24 18:14   ` Martin K. Petersen
  0 siblings, 0 replies; 9+ messages in thread
From: Martin K. Petersen @ 2020-04-24 18:14 UTC (permalink / raw)
  To: Christoph Hellwig
  Cc: Ming Lei, James Bottomley, linux-scsi, Martin K . Petersen,
	Bart Van Assche, John Garry


> Looks good, although you might want to throw in a blurb on the
> actually measured benefits:

Yes, please!

-- 
Martin K. Petersen	Oracle Linux Engineering

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [PATCH V2] scsi: put hot fields of scsi_host_template into one cacheline
  2020-04-22  9:54 [PATCH V2] scsi: put hot fields of scsi_host_template into one cacheline Ming Lei
                   ` (3 preceding siblings ...)
  2020-04-23  6:02 ` Christoph Hellwig
@ 2020-05-01 22:27 ` Douglas Gilbert
  2020-05-04  1:09   ` Ming Lei
  4 siblings, 1 reply; 9+ messages in thread
From: Douglas Gilbert @ 2020-05-01 22:27 UTC (permalink / raw)
  To: Ming Lei, James Bottomley, linux-scsi, Martin K . Petersen
  Cc: Bart Van Assche, Christoph Hellwig, John Garry

On 2020-04-22 5:54 a.m., Ming Lei wrote:
> The following three fields of scsi_host_template are referenced in
> scsi IO submission path, so put them together into one cacheline:
> 
> - cmd_size
> - queuecommand
> - commit_rqs
> 
> Cc: Bart Van Assche <bvanassche@acm.org>
> Cc: Christoph Hellwig <hch@lst.de>
> Cc: John Garry <john.garry@huawei.com>
> Signed-off-by: Ming Lei <ming.lei@redhat.com>
> ---
> V2:
> 	- move the 3 fields at the beginning of scsi_host_template
> 	- comment why we do this way
> 
>   include/scsi/scsi_host.h | 72 ++++++++++++++++++++++------------------
>   1 file changed, 39 insertions(+), 33 deletions(-)

Hi,
I would like to test this patch. However it doesn't apply on Martin's
tree (all 3 chunks fail). Could you post a clean version?

Doug Gilbert


^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [PATCH V2] scsi: put hot fields of scsi_host_template into one cacheline
  2020-05-01 22:27 ` Douglas Gilbert
@ 2020-05-04  1:09   ` Ming Lei
  0 siblings, 0 replies; 9+ messages in thread
From: Ming Lei @ 2020-05-04  1:09 UTC (permalink / raw)
  To: Douglas Gilbert
  Cc: James Bottomley, linux-scsi, Martin K . Petersen,
	Bart Van Assche, Christoph Hellwig, John Garry

On Fri, May 01, 2020 at 06:27:07PM -0400, Douglas Gilbert wrote:
> On 2020-04-22 5:54 a.m., Ming Lei wrote:
> > The following three fields of scsi_host_template are referenced in
> > scsi IO submission path, so put them together into one cacheline:
> > 
> > - cmd_size
> > - queuecommand
> > - commit_rqs
> > 
> > Cc: Bart Van Assche <bvanassche@acm.org>
> > Cc: Christoph Hellwig <hch@lst.de>
> > Cc: John Garry <john.garry@huawei.com>
> > Signed-off-by: Ming Lei <ming.lei@redhat.com>
> > ---
> > V2:
> > 	- move the 3 fields at the beginning of scsi_host_template
> > 	- comment why we do this way
> > 
> >   include/scsi/scsi_host.h | 72 ++++++++++++++++++++++------------------
> >   1 file changed, 39 insertions(+), 33 deletions(-)
> 
> Hi,
> I would like to test this patch. However it doesn't apply on Martin's
> tree (all 3 chunks fail). Could you post a clean version?
> 
> Doug Gilbert
> 

Hi Doug,

Please try the following one:

From 356220ede9f06a20d90fa387cc84eb63eea2335d Mon Sep 17 00:00:00 2001
From: Ming Lei <ming.lei@redhat.com>
Date: Tue, 21 Apr 2020 17:27:30 +0800
Subject: [PATCH] scsi: put hot fields of scsi_host_template into one cacheline

The following three fields of scsi_host_template are referenced in
scsi IO submission path, so put them together into one cacheline:

- cmd_size
- queuecommand
- commit_rqs

Cc: Bart Van Assche <bvanassche@acm.org>
Cc: Christoph Hellwig <hch@lst.de>
Cc: John Garry <john.garry@huawei.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
---
 include/scsi/scsi_host.h | 72 ++++++++++++++++++++++------------------
 1 file changed, 39 insertions(+), 33 deletions(-)

diff --git a/include/scsi/scsi_host.h b/include/scsi/scsi_host.h
index 46ef8cccc982..6fa2697638b4 100644
--- a/include/scsi/scsi_host.h
+++ b/include/scsi/scsi_host.h
@@ -30,40 +30,15 @@ struct scsi_transport_template;
 #define MODE_TARGET 0x02
 
 struct scsi_host_template {
-	struct module *module;
-	const char *name;
-
 	/*
-	 * The info function will return whatever useful information the
-	 * developer sees fit.  If not provided, then the name field will
-	 * be used instead.
-	 *
-	 * Status: OPTIONAL
+	 * Put fields referenced in IO submission path together in
+	 * same cacheline
 	 */
-	const char *(* info)(struct Scsi_Host *);
 
 	/*
-	 * Ioctl interface
-	 *
-	 * Status: OPTIONAL
-	 */
-	int (*ioctl)(struct scsi_device *dev, unsigned int cmd,
-		     void __user *arg);
-
-
-#ifdef CONFIG_COMPAT
-	/* 
-	 * Compat handler. Handle 32bit ABI.
-	 * When unknown ioctl is passed return -ENOIOCTLCMD.
-	 *
-	 * Status: OPTIONAL
+	 * Additional per-command data allocated for the driver.
 	 */
-	int (*compat_ioctl)(struct scsi_device *dev, unsigned int cmd,
-			    void __user *arg);
-#endif
-
-	int (*init_cmd_priv)(struct Scsi_Host *shost, struct scsi_cmnd *cmd);
-	int (*exit_cmd_priv)(struct Scsi_Host *shost, struct scsi_cmnd *cmd);
+	unsigned int cmd_size;
 
 	/*
 	 * The queuecommand function is used to queue up a scsi
@@ -111,6 +86,41 @@ struct scsi_host_template {
 	 */
 	void (*commit_rqs)(struct Scsi_Host *, u16);
 
+	struct module *module;
+	const char *name;
+
+	/*
+	 * The info function will return whatever useful information the
+	 * developer sees fit.  If not provided, then the name field will
+	 * be used instead.
+	 *
+	 * Status: OPTIONAL
+	 */
+	const char *(*info)(struct Scsi_Host *);
+
+	/*
+	 * Ioctl interface
+	 *
+	 * Status: OPTIONAL
+	 */
+	int (*ioctl)(struct scsi_device *dev, unsigned int cmd,
+		     void __user *arg);
+
+
+#ifdef CONFIG_COMPAT
+	/*
+	 * Compat handler. Handle 32bit ABI.
+	 * When unknown ioctl is passed return -ENOIOCTLCMD.
+	 *
+	 * Status: OPTIONAL
+	 */
+	int (*compat_ioctl)(struct scsi_device *dev, unsigned int cmd,
+			    void __user *arg);
+#endif
+
+	int (*init_cmd_priv)(struct Scsi_Host *shost, struct scsi_cmnd *cmd);
+	int (*exit_cmd_priv)(struct Scsi_Host *shost, struct scsi_cmnd *cmd);
+
 	/*
 	 * This is an error handling strategy routine.  You don't need to
 	 * define one of these if you don't want to - there is a default
@@ -475,10 +485,6 @@ struct scsi_host_template {
 	 */
 	u64 vendor_id;
 
-	/*
-	 * Additional per-command data allocated for the driver.
-	 */
-	unsigned int cmd_size;
 	struct scsi_host_cmd_pool *cmd_pool;
 
 	/* Delay for runtime autosuspend */
-- 
2.25.2



-- 
Ming


^ permalink raw reply related	[flat|nested] 9+ messages in thread

end of thread, other threads:[~2020-05-04  1:09 UTC | newest]

Thread overview: 9+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-04-22  9:54 [PATCH V2] scsi: put hot fields of scsi_host_template into one cacheline Ming Lei
2020-04-22 10:06 ` John Garry
2020-04-22 14:36 ` Bart Van Assche
2020-04-22 14:41 ` James Bottomley
2020-04-23  1:48   ` Ming Lei
2020-04-23  6:02 ` Christoph Hellwig
2020-04-24 18:14   ` Martin K. Petersen
2020-05-01 22:27 ` Douglas Gilbert
2020-05-04  1:09   ` Ming Lei

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.