[patch v2 0/3] block: optimize flush for non-queueable flush drive

[patch v2 0/3] block: optimize flush for non-queueable flush drive

[shaohua.li]

Flush request isn't queueable. When it's running, other request
can't. We can optimize flush performance according to this knowledge
In my test, I got about 20% performance boost.

[patch v2 1/3] block: add a non-queueable flush flag

[shaohua.li]

[-- Attachment #1: blk-nonqueueable-flush-flag.patch --]
[-- Type: text/plain, Size: 1233 bytes --]

flush request isn't queueable in some drives. Add a flag to let driver
notify block layer about this. We can optimize flush performance with the
knowledge.

Signed-off-by: Shaohua Li <shaohua.li@intel.com>
---
 include/linux/blkdev.h |   11 +++++++++++
 1 file changed, 11 insertions(+)

Index: linux/include/linux/blkdev.h
===================================================================
--- linux.orig/include/linux/blkdev.h	2011-05-04 14:23:42.000000000 +0800
+++ linux/include/linux/blkdev.h	2011-05-04 14:24:40.000000000 +0800
@@ -364,6 +364,7 @@ struct request_queue
 	 * for flush operations
 	 */
 	unsigned int		flush_flags;
+	unsigned int		flush_not_queueable:1;
 	unsigned int		flush_pending_idx:1;
 	unsigned int		flush_running_idx:1;
 	unsigned long		flush_pending_since;
@@ -549,6 +550,16 @@ static inline void blk_clear_queue_full(
 		queue_flag_clear(QUEUE_FLAG_ASYNCFULL, q);
 }
 
+static inline void blk_set_queue_flush_queueable(struct request_queue *q,
+		bool queueable)
+{
+	q->flush_not_queueable = !queueable;
+}
+
+static inline bool blk_queue_flush_queueable(struct request_queue *q)
+{
+	return !q->flush_not_queueable;
+}
 
 /*
  * mergeable request must not have _NOMERGE or _BARRIER bit set, nor may

[patch v2 2/3] block: hold queue if flush is running for non-queueable flush drive

[shaohua.li]

[-- Attachment #1: blk-hold-queue-for-nonqueueable-flush.patch --]
[-- Type: text/plain, Size: 2850 bytes --]

In some drives, flush requests are non-queueable. When flush request is running,
normal read/write requests can't run. If block layer dispatches such request,
driver can't handle it and requeue it.
Tejun suggested we can hold the queue when flush is running. This can avoid
unnecessary requeue.
Also this can improve performance. Say we have requests f1, w1, f2 (f is flush
request, and w is write request). When f1 is running, queue will be hold, so w1
will not be added to queue list. Just after f1 is finished, f2 will be
dispatched. Since f1 already flushs cache out, f2 can be finished very quickly.
In my test, the queue holding completely solves a regression introduced by
commit 53d63e6b0dfb9588, which is about 20% regression running a sysbench fileio
workload.

Signed-off-by: Shaohua Li <shaohua.li@intel.com>
---
 block/blk-flush.c      |    3 +++
 block/blk.h            |   12 +++++++++++-
 include/linux/blkdev.h |    1 +
 3 files changed, 15 insertions(+), 1 deletion(-)

Index: linux/block/blk-flush.c
===================================================================
--- linux.orig/block/blk-flush.c	2011-05-04 14:20:33.000000000 +0800
+++ linux/block/blk-flush.c	2011-05-04 15:23:50.000000000 +0800
@@ -199,6 +199,9 @@ static void flush_end_io(struct request
 
 	BUG_ON(q->flush_pending_idx == q->flush_running_idx);
 
+	queued |= q->flush_queue_delayed;
+	q->flush_queue_delayed = 0;
+
 	/* account completion of the flush request */
 	q->flush_running_idx ^= 1;
 	elv_completed_request(q, flush_rq);
Index: linux/include/linux/blkdev.h
===================================================================
--- linux.orig/include/linux/blkdev.h	2011-05-04 14:24:40.000000000 +0800
+++ linux/include/linux/blkdev.h	2011-05-04 14:29:29.000000000 +0800
@@ -365,6 +365,7 @@ struct request_queue
 	 */
 	unsigned int		flush_flags;
 	unsigned int		flush_not_queueable:1;
+	unsigned int		flush_queue_delayed:1;
 	unsigned int		flush_pending_idx:1;
 	unsigned int		flush_running_idx:1;
 	unsigned long		flush_pending_since;
Index: linux/block/blk.h
===================================================================
--- linux.orig/block/blk.h	2011-05-04 14:20:33.000000000 +0800
+++ linux/block/blk.h	2011-05-04 16:09:42.000000000 +0800
@@ -61,7 +61,17 @@ static inline struct request *__elv_next
 			rq = list_entry_rq(q->queue_head.next);
 			return rq;
 		}
-
+		/*
+		 *  Flush request is running and flush request isn't queeueable
+		 *  in the drive, we can hold the queue till flush request is
+		 *  finished. Even we don't do this, driver can't dispatch next
+		 *  requests and will requeue them.
+		 */
+		if (q->flush_pending_idx != q->flush_running_idx &&
+				!blk_queue_flush_queueable(q)) {
+			q->flush_queue_delayed = 1;
+			return NULL;
+		}
 		if (!q->elevator->ops->elevator_dispatch_fn(q, 0))
 			return NULL;
 	}

[patch v2 3/3] SATA: enable non-queueable flush flag

[shaohua.li]

[-- Attachment #1: sata-add-nonqueueable.patch --]
[-- Type: text/plain, Size: 770 bytes --]

Enable non-queueable flush flag for SATA.

Signed-off-by: Shaohua Li <shaohua.li@intel.com>
---
 drivers/ata/libata-scsi.c |    2 ++
 1 file changed, 2 insertions(+)

Index: linux/drivers/ata/libata-scsi.c
===================================================================
--- linux.orig/drivers/ata/libata-scsi.c	2011-05-04 14:20:41.000000000 +0800
+++ linux/drivers/ata/libata-scsi.c	2011-05-04 14:28:28.000000000 +0800
@@ -3424,7 +3424,9 @@ void ata_scsi_scan_host(struct ata_port
 			sdev = __scsi_add_device(ap->scsi_host, channel, id, 0,
 						 NULL);
 			if (!IS_ERR(sdev)) {
+				struct request_queue *q = sdev->request_queue;
 				dev->sdev = sdev;
+				blk_set_queue_flush_queueable(q, false);
 				scsi_device_put(sdev);
 			} else {
 				dev->sdev = NULL;

Re: [patch v2 3/3] SATA: enable non-queueable flush flag

[Tejun Heo]

On Wed, May 04, 2011 at 04:17:28PM +0800, shaohua.li@intel.com wrote:
> Enable non-queueable flush flag for SATA.
> 
> Signed-off-by: Shaohua Li <shaohua.li@intel.com>
> ---
>  drivers/ata/libata-scsi.c |    2 ++
>  1 file changed, 2 insertions(+)
> 
> Index: linux/drivers/ata/libata-scsi.c
> ===================================================================
> --- linux.orig/drivers/ata/libata-scsi.c	2011-05-04 14:20:41.000000000 +0800
> +++ linux/drivers/ata/libata-scsi.c	2011-05-04 14:28:28.000000000 +0800
> @@ -3424,7 +3424,9 @@ void ata_scsi_scan_host(struct ata_port
>  			sdev = __scsi_add_device(ap->scsi_host, channel, id, 0,
>  						 NULL);
>  			if (!IS_ERR(sdev)) {
> +				struct request_queue *q = sdev->request_queue;
>  				dev->sdev = sdev;
> +				blk_set_queue_flush_queueable(q, false);
>  				scsi_device_put(sdev);
>  			} else {
>  				dev->sdev = NULL;

Isn't ata_scsi_dev_config() better place for this?

Thakns.

-- 
tejun

Re: [patch v2 1/3] block: add a non-queueable flush flag

[Tejun Heo]

Hello,

On Wed, May 04, 2011 at 04:17:26PM +0800, shaohua.li@intel.com wrote:
> flush request isn't queueable in some drives. Add a flag to let driver
> notify block layer about this. We can optimize flush performance with the
> knowledge.
> 
> Signed-off-by: Shaohua Li <shaohua.li@intel.com>
> ---
>  include/linux/blkdev.h |   11 +++++++++++
>  1 file changed, 11 insertions(+)
> 
> Index: linux/include/linux/blkdev.h
> ===================================================================
> --- linux.orig/include/linux/blkdev.h	2011-05-04 14:23:42.000000000 +0800
> +++ linux/include/linux/blkdev.h	2011-05-04 14:24:40.000000000 +0800
> @@ -364,6 +364,7 @@ struct request_queue
>  	 * for flush operations
>  	 */
>  	unsigned int		flush_flags;
> +	unsigned int		flush_not_queueable:1;
>  	unsigned int		flush_pending_idx:1;
>  	unsigned int		flush_running_idx:1;
>  	unsigned long		flush_pending_since;
> @@ -549,6 +550,16 @@ static inline void blk_clear_queue_full(
>  		queue_flag_clear(QUEUE_FLAG_ASYNCFULL, q);
>  }
>  
> +static inline void blk_set_queue_flush_queueable(struct request_queue *q,
> +		bool queueable)
> +{
> +	q->flush_not_queueable = !queueable;
> +}
> +
> +static inline bool blk_queue_flush_queueable(struct request_queue *q)
> +{
> +	return !q->flush_not_queueable;
> +}

In the block layer, these setters live in block/blk-settings.c and
usually don't have _set in their names.

Also, negation in the interface is slightly bothering but I can't
think of better name meaning !queueable either, so unless someone can
come up with better name, it should do, I think.

Thank you.

-- 
tejun

Re: [patch v2 2/3] block: hold queue if flush is running for non-queueable flush drive

[Tejun Heo]

Hello,

On Wed, May 04, 2011 at 04:17:27PM +0800, shaohua.li@intel.com wrote:
> In some drives, flush requests are non-queueable. When flush request is running,
> normal read/write requests can't run. If block layer dispatches such request,
> driver can't handle it and requeue it.
> Tejun suggested we can hold the queue when flush is running. This can avoid
> unnecessary requeue.
> Also this can improve performance. Say we have requests f1, w1, f2 (f is flush
> request, and w is write request). When f1 is running, queue will be hold, so w1
> will not be added to queue list. Just after f1 is finished, f2 will be
> dispatched. Since f1 already flushs cache out, f2 can be finished very quickly.
> In my test, the queue holding completely solves a regression introduced by
> commit 53d63e6b0dfb9588, which is about 20% regression running a sysbench fileio
> workload.
> 
> Signed-off-by: Shaohua Li <shaohua.li@intel.com>

It looks good to me now, but some nitpicks.

> Index: linux/block/blk-flush.c
> ===================================================================
> --- linux.orig/block/blk-flush.c	2011-05-04 14:20:33.000000000 +0800
> +++ linux/block/blk-flush.c	2011-05-04 15:23:50.000000000 +0800
> @@ -199,6 +199,9 @@ static void flush_end_io(struct request
>  
>  	BUG_ON(q->flush_pending_idx == q->flush_running_idx);
>  
> +	queued |= q->flush_queue_delayed;
> +	q->flush_queue_delayed = 0;
> +
>  	/* account completion of the flush request */
>  	q->flush_running_idx ^= 1;
>  	elv_completed_request(q, flush_rq);

Can you please do if (queued || q->flush_queue_delayed) instead of
setting queued?  And please also update the comment above the if
statement so that it explains the flush_queue_delayed case too.

> Index: linux/block/blk.h
> ===================================================================
> --- linux.orig/block/blk.h	2011-05-04 14:20:33.000000000 +0800
> +++ linux/block/blk.h	2011-05-04 16:09:42.000000000 +0800
> @@ -61,7 +61,17 @@ static inline struct request *__elv_next
>  			rq = list_entry_rq(q->queue_head.next);
>  			return rq;
>  		}
> -
> +		/*
> +		 *  Flush request is running and flush request isn't queeueable
> +		 *  in the drive, we can hold the queue till flush request is
> +		 *  finished. Even we don't do this, driver can't dispatch next
> +		 *  requests and will requeue them.
> +		 */

Please explain the f1, w1, f2 case here as that's the biggest reason
this optimization is implemented and also explain the use of
flush_queue_delayed (just explain briefly and refer to
flush_end_io()).

Thank you.

-- 
tejun

Re: [patch v2 3/3] SATA: enable non-queueable flush flag

[Jeff Garzik]

On 05/04/2011 04:53 AM, Tejun Heo wrote:
> On Wed, May 04, 2011 at 04:17:28PM +0800, shaohua.li@intel.com wrote:
>> @@ -3424,7 +3424,9 @@ void ata_scsi_scan_host(struct ata_port
>>   			sdev = __scsi_add_device(ap->scsi_host, channel, id, 0,
>>   						 NULL);
>>   			if (!IS_ERR(sdev)) {
>> +				struct request_queue *q = sdev->request_queue;
>>   				dev->sdev = sdev;
>> +				blk_set_queue_flush_queueable(q, false);
>>   				scsi_device_put(sdev);
>>   			} else {
>>   				dev->sdev = NULL;

> Isn't ata_scsi_dev_config() better place for this?

Yes :)

Re: [patch v2 2/3] block: hold queue if flush is running for non-queueable flush drive

[Sergei Shtylyov]

Hello.

On 04-05-2011 12:17, shaohua.li@intel.com wrote:

> In some drives, flush requests are non-queueable. When flush request is running,
> normal read/write requests can't run. If block layer dispatches such request,
> driver can't handle it and requeue it.
> Tejun suggested we can hold the queue when flush is running. This can avoid
> unnecessary requeue.
> Also this can improve performance. Say we have requests f1, w1, f2 (f is flush
> request, and w is write request). When f1 is running, queue will be hold, so w1
> will not be added to queue list. Just after f1 is finished, f2 will be
> dispatched. Since f1 already flushs cache out, f2 can be finished very quickly.
> In my test, the queue holding completely solves a regression introduced by
> commit 53d63e6b0dfb9588, which is about 20% regression running a sysbench fileio

    Please specify that commit's summary -- for human readers. The ID is only 
immediately usable to gitweb.

> workload.

> Signed-off-by: Shaohua Li <shaohua.li@intel.com>

[...]

> Index: linux/block/blk.h
> ===================================================================
> --- linux.orig/block/blk.h	2011-05-04 14:20:33.000000000 +0800
> +++ linux/block/blk.h	2011-05-04 16:09:42.000000000 +0800
> @@ -61,7 +61,17 @@ static inline struct request *__elv_next
>   			rq = list_entry_rq(q->queue_head.next);
>   			return rq;
>   		}
> -
> +		/*
> +		 *  Flush request is running and flush request isn't queeueable

    Queueable.

WBR, Sergei