FWI. I need lots of testing on this. Only real SCSI controllers,
please, not RAID controllers (except for MPT-SCSI with integrated
mirroring). So Adaptec, LSI, Symbios, Buslogic, Tekram, SME, etc,
users, please try this and get back to me. The patch should apply
to FreeBSD 7 as well. FreeBSD 6 is only affected by this problem
when CAM_NEW_TRAN_CODE is enabled.
Scott
-------- Original Message --------
Subject: svn commit: r188671 - head/sys/cam
Date: Mon, 16 Feb 2009 14:57:15 +0000 (UTC)
From: Scott Long <scottl@FreeBSD.org>
To: src-committers@FreeBSD.org, svn-src-all@FreeBSD.org,
svn-src-head@FreeBSD.org
Author: scottl
Date: Mon Feb 16 14:57:15 2009
New Revision: 188671
URL: http://svn.freebsd.org/changeset/base/188671
Log:
Fix parallel SCSI negotiation in the CAM_NEW_TRAN_CODE world order.
Overzealous sanity checks were locking the sync_rate and offset values to
zero, thanks to a twisty maze of recursive code.
Modified:
head/sys/cam/cam_xpt.c
Modified: head/sys/cam/cam_xpt.c
=============================================================================---
head/sys/cam/cam_xpt.c Mon Feb 16 14:38:52 2009 (r188670)
+++ head/sys/cam/cam_xpt.c Mon Feb 16 14:57:15 2009 (r188671)
@@ -6679,9 +6679,7 @@ xpt_set_transfer_settings(struct ccb_tra
if (((device->flags & CAM_DEV_INQUIRY_DATA_VALID) != 0
&& (inq_data->flags & SID_Sync) == 0
&& cts->type == CTS_TYPE_CURRENT_SETTINGS)
- || ((cpi.hba_inquiry & PI_SDTR_ABLE) == 0)
- || (spi->sync_offset == 0)
- || (spi->sync_period == 0)) {
+ || ((cpi.hba_inquiry & PI_SDTR_ABLE) == 0)) {
/* Force async */
spi->sync_period = 0;
spi->sync_offset = 0;
@@ -6729,7 +6727,8 @@ xpt_set_transfer_settings(struct ccb_tra
if (spi->bus_width == 0)
spi->ppr_options = 0;
- if ((spi->flags & CTS_SPI_FLAGS_DISC_ENB) == 0) {
+ if ((spi->valid & CTS_SPI_VALID_DISC)
+ && ((spi->flags & CTS_SPI_FLAGS_DISC_ENB) == 0)) {
/*
* Can't tag queue without disconnection.
*/
7.1-STABLE FreeBSD 7.1-STABLE #3: Tue Feb 17 14:58:42
asr0@pci0:3:3:0: class=0x010400 card=0xc0341044 chip=0xa5111044 rev=0x01
hdr=0x00
vendor = 'Adaptec (Formerly: Distributed Processing Technology
(DPT))'
device = 'Raptor SmartRAID Controller'
class = mass storage
subclass = RAID
root:# camcontrol tags da0
(pass0:asr0:0:0:0): device openings: 1
---------
6.2-STABLE FreeBSD 6.2-STABLE #1: Mon Oct 15 16:53:04
asr0@pci3:3:0: class=0x010400 card=0xc0341044 chip=0xa5111044 rev=0x01
hdr=0x00
vendor = 'Adaptec (Formerly: Distributed Processing Technology
(DPT))'
device = 'Raptor SmartRAID Controller'
class = mass storage
subclass = RAID
root:# camcontrol tags da0
(pass0:asr0:0:0:0): device openings: 255
On Monday 16 February 2009, Scott Long wrote:> FWI. I need lots of testing on this. Only real SCSI controllers,
> please, not RAID controllers (except for MPT-SCSI with integrated
> mirroring). So Adaptec, LSI, Symbios, Buslogic, Tekram, SME, etc,
> users, please try this and get back to me. The patch should apply
> to FreeBSD 7 as well. FreeBSD 6 is only affected by this problem
> when CAM_NEW_TRAN_CODE is enabled.
>
> Scott
>
>
> -------- Original Message --------
> Subject: svn commit: r188671 - head/sys/cam
> Date: Mon, 16 Feb 2009 14:57:15 +0000 (UTC)
> From: Scott Long <scottl@FreeBSD.org>
> To: src-committers@FreeBSD.org, svn-src-all@FreeBSD.org,
> svn-src-head@FreeBSD.org
>
> Author: scottl
> Date: Mon Feb 16 14:57:15 2009
> New Revision: 188671
> URL: http://svn.freebsd.org/changeset/base/188671
>
> Log:
> Fix parallel SCSI negotiation in the CAM_NEW_TRAN_CODE world order.
> Overzealous sanity checks were locking the sync_rate and offset values
to> zero, thanks to a twisty maze of recursive code.
>
> Modified:
> head/sys/cam/cam_xpt.c
>
> Modified: head/sys/cam/cam_xpt.c
>
=============================================================================>
--- head/sys/cam/cam_xpt.c Mon Feb 16 14:38:52 2009
(r188670)> +++ head/sys/cam/cam_xpt.c Mon Feb 16 14:57:15 2009 (r188671)
> @@ -6679,9 +6679,7 @@ xpt_set_transfer_settings(struct ccb_tra
> if (((device->flags & CAM_DEV_INQUIRY_DATA_VALID) != 0
> && (inq_data->flags & SID_Sync) == 0
> && cts->type == CTS_TYPE_CURRENT_SETTINGS)
> - || ((cpi.hba_inquiry & PI_SDTR_ABLE) == 0)
> - || (spi->sync_offset == 0)
> - || (spi->sync_period == 0)) {
> + || ((cpi.hba_inquiry & PI_SDTR_ABLE) == 0)) {
> /* Force async */
> spi->sync_period = 0;
> spi->sync_offset = 0;
> @@ -6729,7 +6727,8 @@ xpt_set_transfer_settings(struct ccb_tra
> if (spi->bus_width == 0)
> spi->ppr_options = 0;
>
> - if ((spi->flags & CTS_SPI_FLAGS_DISC_ENB) == 0) {
> + if ((spi->valid & CTS_SPI_VALID_DISC)
> + && ((spi->flags & CTS_SPI_FLAGS_DISC_ENB) == 0)) {
> /*
> * Can't tag queue without disconnection.
> */
> _______________________________________________
> freebsd-stable@freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-stable
> To unsubscribe, send any mail to
"freebsd-stable-unsubscribe@freebsd.org"
>
>
--
EMIT-RIPN, EVM7-RIPE
On Mon, 16 Feb 2009 08:09:35 -0700 Scott Long <scottl@samsco.org> wrote:> FWI. I need lots of testing on this. Only real SCSI controllers, > please, not RAID controllers (except for MPT-SCSI with integrated > mirroring). So Adaptec, LSI, Symbios, Buslogic, Tekram, SME, etc, > users, please try this and get back to me. The patch should apply > to FreeBSD 7 as well. FreeBSD 6 is only affected by this problem > when CAM_NEW_TRAN_CODE is enabled. >I tested this with an Adaptec 29160. I saw no real improvement in performance, but also no regressions. I suspect that the old disk I had attached just didn't have enough performance reserves to show an improvement. My test scenario was buildworld. Since /usr/src and /usr/obj were both on the one disk it got a pretty good workout. AMD64 X2 (2.5 GHz) with 4GB of RAM. BTW under a very fresh 8.0-current. --- Gary Jennejohn