kern/72116: tagged queueing broken on quantum Atlas10K3
Holger Kipp
hk at alogis.com
Sun Sep 26 15:00:43 PDT 2004
>Number: 72116
>Category: kern
>Synopsis: tagged queueing broken on quantum Atlas10K3
>Confidential: no
>Severity: serious
>Priority: medium
>Responsible: freebsd-bugs
>State: open
>Quarter:
>Keywords:
>Date-Required:
>Class: update
>Submitter-Id: current-users
>Arrival-Date: Sun Sep 26 22:00:39 GMT 2004
>Closed-Date:
>Last-Modified:
>Originator: Holger Kipp
>Release: 4.8-RELEASE-p25
>Organization:
alogis AG
>Environment:
FreeBSD intserv.int1.b.intern 4.8-RELEASE-p25 FreeBSD 4.8-RELEASE-p25 #5: Sun Sep 26 23:12:45 CEST 2004 wartung at intserv.int1.b.intern:/usr/obj/usr/src/sys/B-INTSERV i386
>Description:
Quantum ATLAS10K3 tagged queueing seems to be broken.
On old ASUS P2B-S I encountered several BDRs and things like that
(that was with FreeBSD 4.2). Updating to 4.8 and switching hardware
except for harddisks gave the following error (U160 Dual Channel LSI):
mpt0: bullet missed in timeout
mpt1: bullet missed in timeout
This is with 3 of the offending disks in vinum raid5.
Initialization and rebuilding of raid had no problems, though -
maybe due to the fact that then a single drive receives only
read xor write commands, reading or writing sequential data.
>How-To-Repeat:
Buy Quantum Atlas 10K III hard disk, use eg within vinum.
>Fix:
add to sys/cam/cam_xpt.c something like
{
/*
* 74 Gig Atlas III, presumably same problem as the 9G/18G version.
* Reported by: Holger Kipp
* <holger.kipp at alogis.com>
*
* For future reference, the drive with the problem was:
* QUANTUM ATLAS10K3_73_WLS 020W
*/
{ T_DIRECT, SIP_MEDIA_FIXED, quantum, "ATLAS10K3*", "*" },
/*quirks*/0, /*mintags*/24, /*maxtags*/32
},
I have not checked if larger values would also work - merely took
those from my predecessor.
>Release-Note:
>Audit-Trail:
>Unformatted:
More information about the freebsd-bugs
mailing list