public inbox for linux-scsi@vger.kernel.org
 help / color / mirror / Atom feed
* [PATCH] libosd: osd_req_read_sg, optimize the single entry case
@ 2011-01-23 15:53 Boaz Harrosh
  2011-01-23 15:54 ` Boaz Harrosh
  0 siblings, 1 reply; 2+ messages in thread
From: Boaz Harrosh @ 2011-01-23 15:53 UTC (permalink / raw)
  To: James Bottomley, linux-scsi, open-osd


Since sg-read is a bidi operation, it is a gain to convert
a single sg entry into a regular read. Better do this in the
generic layer then force each caller to do so.

Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
---
 drivers/scsi/osd/osd_initiator.c |   20 ++++++++++++++++----
 1 files changed, 16 insertions(+), 4 deletions(-)

diff --git a/drivers/scsi/osd/osd_initiator.c b/drivers/scsi/osd/osd_initiator.c
index 0433ea6..19ddf65 100644
--- a/drivers/scsi/osd/osd_initiator.c
+++ b/drivers/scsi/osd/osd_initiator.c
@@ -1005,11 +1005,23 @@ int osd_req_read_sg(struct osd_request *or,
 	const struct osd_sg_entry *sglist, unsigned numentries)
 {
 	u64 len;
-	int ret = _add_sg_continuation_descriptor(or, sglist, numentries, &len);
+	u64 off;
+	int ret;
 
-	if (ret)
-		return ret;
-	osd_req_read(or, obj, 0, bio, len);
+	if (numentries > 1) {
+		off = 0;
+		ret = _add_sg_continuation_descriptor(or, sglist, numentries,
+						      &len);
+		if (ret)
+			return ret;
+	} else {
+		/* Optimize the case of single segment, read_sg is a
+		 * bidi operation.
+		 */
+		len = sglist->len;
+		off = sglist->offset;
+	}
+	osd_req_read(or, obj, off, bio, len);
 
 	return 0;
 }
-- 
1.7.2.3


^ permalink raw reply related	[flat|nested] 2+ messages in thread

* Re: [PATCH] libosd: osd_req_read_sg, optimize the single entry case
  2011-01-23 15:53 [PATCH] libosd: osd_req_read_sg, optimize the single entry case Boaz Harrosh
@ 2011-01-23 15:54 ` Boaz Harrosh
  0 siblings, 0 replies; 2+ messages in thread
From: Boaz Harrosh @ 2011-01-23 15:54 UTC (permalink / raw)
  To: James Bottomley, linux-scsi, open-osd

On 01/23/2011 05:53 PM, Boaz Harrosh wrote:
> 
> Since sg-read is a bidi operation, it is a gain to convert
> a single sg entry into a regular read. Better do this in the
> generic layer then force each caller to do so.
> 
> Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
> ---
>  drivers/scsi/osd/osd_initiator.c |   20 ++++++++++++++++----
>  1 files changed, 16 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/scsi/osd/osd_initiator.c b/drivers/scsi/osd/osd_initiator.c
> index 0433ea6..19ddf65 100644
> --- a/drivers/scsi/osd/osd_initiator.c
> +++ b/drivers/scsi/osd/osd_initiator.c
> @@ -1005,11 +1005,23 @@ int osd_req_read_sg(struct osd_request *or,
>  	const struct osd_sg_entry *sglist, unsigned numentries)
>  {
>  	u64 len;
> -	int ret = _add_sg_continuation_descriptor(or, sglist, numentries, &len);
> +	u64 off;
> +	int ret;
>  
> -	if (ret)
> -		return ret;
> -	osd_req_read(or, obj, 0, bio, len);
> +	if (numentries > 1) {
> +		off = 0;
> +		ret = _add_sg_continuation_descriptor(or, sglist, numentries,
> +						      &len);
> +		if (ret)
> +			return ret;
> +	} else {
> +		/* Optimize the case of single segment, read_sg is a
> +		 * bidi operation.
> +		 */
> +		len = sglist->len;
> +		off = sglist->offset;
> +	}
> +	osd_req_read(or, obj, off, bio, len);
>  
>  	return 0;
>  }

James please include this patch into scsi-misc for the next merge window.

Thanks
Boaz

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2011-01-23 15:54 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2011-01-23 15:53 [PATCH] libosd: osd_req_read_sg, optimize the single entry case Boaz Harrosh
2011-01-23 15:54 ` Boaz Harrosh

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox