--- Begin Message ---
- To: Debian Bug Tracking System <submit@bugs.debian.org>
- Subject: linux-source-2.6.16: bogus scsi timeouts with qla1280 driver
- From: "R. Scott Bailey" <scott.bailey@eds.com>
- Date: Wed, 10 May 2006 13:25:54 -0400
- Message-id: <20060510172554.6775.28540.reportbug@hamster.dsddi.eds.com>
Package: linux-source-2.6.16
Version: 2.6.16-12
Severity: normal
Tags: patch
I have been aggravated by tape problems since upgrading to 2.6 on my
Alphaserver using the qla1280 driver with ISP1020/1040 support enabled.
It appears on inspection that the problem lies with the qla1280 driver
itself, which enforces a 30-second timeout on all (?) SCSI operations it
issues. The result is that operations on my TZ89 DLT4 tape drive which
take longer than that to complete (for example, "mt eod" or "mt fsf" in
particular) abort and force a device reset.
The attached kludgy patch demonstrates the fix I am using, namely to
increase the timeout to 900 seconds. This is fine because my hardware is
all solid and I'm not worried at the moment about a disk going bad and
hanging my system for 15 minutes while the I/O times out... :-)
Probably the timeout could be reduced to perhaps 5 minutes, but I am
erring on the side of longer since these errant "errors" really screw up
my backups and restores.
I have posted to linux-scsi on this but haven't gotten much of a
response. I'll report this there too but figured it would be better to
have more eyes on the problem.
Thanks,
Scott Bailey
scott.bailey@eds.com
--- drivers/scsi/qla1280.c 2006-03-20 00:53:29.000000000 -0500
+++ drivers/scsi/qla1280.c 2006-05-10 10:02:39.000000000 -0400
@@ -17,9 +17,11 @@
* General Public License for more details.
*
******************************************************************************/
-#define QLA1280_VERSION "3.26"
+#define QLA1280_VERSION "3.26.rsb"
/*****************************************************************************
Revision History:
+ Rev 3.26.rsb, May 10, 2006 R. Scott Bailey
+ - Increase timeouts (?) from 30 to 900 sec for tape support
Rev 3.26, January 16, 2006 Jes Sorensen
- Ditch all < 2.6 support
Rev 3.25.1, February 10, 2005 Christoph Hellwig
@@ -2886,7 +2888,7 @@
memset(((char *)pkt + 8), 0, (REQUEST_ENTRY_SIZE - 8));
/* Set ISP command timeout. */
- pkt->timeout = cpu_to_le16(30);
+ pkt->timeout = cpu_to_le16(900); /* was 30 - RSB */
/* Set device target ID and LUN */
pkt->lun = SCSI_LUN_32(cmd);
@@ -3185,7 +3187,7 @@
memset(((char *)pkt + 8), 0, (REQUEST_ENTRY_SIZE - 8));
/* Set ISP command timeout. */
- pkt->timeout = cpu_to_le16(30);
+ pkt->timeout = cpu_to_le16(900); /* was 30 - RSB */
/* Set device target ID and LUN */
pkt->lun = SCSI_LUN_32(cmd);
-- System Information:
Debian Release: testing/unstable
Architecture: alpha
Shell: /bin/sh linked to /bin/bash
Kernel: Linux 2.6.16-scsi.12
Locale: LANG=C, LC_CTYPE=C (charmap=ANSI_X3.4-1968)
Versions of packages linux-source-2.6.16 depends on:
ii binutils 2.16.1cvs20060413-1 The GNU assembler, linker and bina
ii bzip2 1.0.3-2 high-quality block-sorting file co
Versions of packages linux-source-2.6.16 recommends:
ii gcc 4:4.0.2-2 The GNU C compiler
ii libc6.1-dev [libc6-dev] 2.3.6-7 GNU C Library: Development Librari
ii make 3.81-1 The GNU version of the "make" util
-- no debconf information
--- End Message ---
--- Begin Message ---
- To: 366730-done@bugs.debian.org
- Subject: Re: qla1280 times out on long operations such as tape rewind
- From: maximilian attems <max@stro.at>
- Date: Sun, 24 Jun 2007 23:04:54 +0200
- Message-id: <20070624210454.GG4869@stro.at>
- In-reply-to: <F3598542BAC24B4783B7A440D45522FF04B5D899@usahm236.amer.corp.eds.com>
- References: <F3598542BAC24B4783B7A440D45522FF04B5D899@usahm236.amer.corp.eds.com>
Version: 2.6.20-1
On Thu, 24 Aug 2006, Bailey, Scott wrote:
> Just tuning in briefly to report:
>
> 1. I have been running my 3-processor Alphaserver 4100 with Ian's patch
> http://bugzilla.kernel.org/show_bug.cgi?id=6275 for a little over a
> month now, first with 2.6.16 and now with 2.6.17, and it continues to
> work like a champ. I very much would like to see this fix incorporated
> into the Debian kernel or preferably upstream, since my DLT drive is
> pretty much useless without it.
>
> 2. Unfortunately, I see no sign that anybody has looked at the kernel
> bugzilla report above.
aboves mentioned patch got merged for 2.6.19
and might be a backport candidate for etch.
closing with the version 2.6.20 as unstable never saw an 2.6.19.
regards
--
maks
--- End Message ---