| Message ID | 20170928013859.1214-1-martin.petersen@oracle.com (mailing list archive) |
|---|---|
| State | Accepted |
| Headers | show |
On Wed, 2017-09-27 at 21:38 -0400, Martin K. Petersen wrote: > A user may lower the max_sectors_kb setting in sysfs to accommodate > certain workloads. Previously we would always set the max I/O size to > either the block layer default or the optional preferred I/O size > reported by the device. > > Keep the current heuristics for the initial setting of > max_sectors_kb. > For subsequent invocations, only update the current queue limit if it > exceeds the capabilities of the hardware. > > Reported-by: Don Brace <don.brace@microsemi.com> > Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> This looks good to me. I agree that it's superior to the original suggestion, because it sets the soft limit to the hard limit when the device is scanned for the first time. Regards Martin
> -----Original Message----- > From: linux-scsi-owner@vger.kernel.org [mailto:linux-scsi- > owner@vger.kernel.org] On Behalf Of Martin K. Petersen > Sent: Wednesday, September 27, 2017 8:39 PM > To: linux-scsi@vger.kernel.org > Cc: Martin K. Petersen <martin.petersen@oracle.com> > Subject: [PATCH] scsi: sd: Do not override max_sectors_kb sysfs setting > > EXTERNAL EMAIL > > > A user may lower the max_sectors_kb setting in sysfs to accommodate > certain workloads. Previously we would always set the max I/O size to > either the block layer default or the optional preferred I/O size > reported by the device. > > Keep the current heuristics for the initial setting of max_sectors_kb. > For subsequent invocations, only update the current queue limit if it > exceeds the capabilities of the hardware. > > Reported-by: Don Brace <don.brace@microsemi.com> > Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Tested-by: Don Brace <don.brace@microsemi.com> Really appreciate your attention to this matter. Thanks, Don Brace ESC - Smart Storage Microsemi Corporation
diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c index 6549e5ce09ca..b18ba3235900 100644 --- a/drivers/scsi/sd.c +++ b/drivers/scsi/sd.c @@ -3098,8 +3098,6 @@ static int sd_revalidate_disk(struct gendisk *disk) sd_read_security(sdkp, buffer); } - sdkp->first_scan = 0; - /* * We now have all cache related info, determine how we deal * with flush requests. @@ -3114,7 +3112,7 @@ static int sd_revalidate_disk(struct gendisk *disk) q->limits.max_dev_sectors = logical_to_sectors(sdp, dev_max); /* - * Use the device's preferred I/O size for reads and writes + * Determine the device's preferred I/O size for reads and writes * unless the reported value is unreasonably small, large, or * garbage. */ @@ -3128,8 +3126,19 @@ static int sd_revalidate_disk(struct gendisk *disk) rw_max = min_not_zero(logical_to_sectors(sdp, dev_max), (sector_t)BLK_DEF_MAX_SECTORS); - /* Combine with controller limits */ - q->limits.max_sectors = min(rw_max, queue_max_hw_sectors(q)); + /* Do not exceed controller limit */ + rw_max = min(rw_max, queue_max_hw_sectors(q)); + + /* + * Only update max_sectors if previously unset or if the current value + * exceeds the capabilities of the hardware. + */ + if (sdkp->first_scan || + q->limits.max_sectors > q->limits.max_dev_sectors || + q->limits.max_sectors > q->limits.max_hw_sectors) + q->limits.max_sectors = rw_max; + + sdkp->first_scan = 0; set_capacity(disk, logical_to_sectors(sdp, sdkp->capacity)); sd_config_write_same(sdkp);
A user may lower the max_sectors_kb setting in sysfs to accommodate certain workloads. Previously we would always set the max I/O size to either the block layer default or the optional preferred I/O size reported by the device. Keep the current heuristics for the initial setting of max_sectors_kb. For subsequent invocations, only update the current queue limit if it exceeds the capabilities of the hardware. Reported-by: Don Brace <don.brace@microsemi.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> --- drivers/scsi/sd.c | 19 ++++++++++++++----- 1 file changed, 14 insertions(+), 5 deletions(-)