kernel/3048: scsi disk errors, sd"x": not queued, error 5

From: <tylczak(at)alrc.doe.gov>
Date: Mon Dec 23 2002 - 18:05:36 EST

>Number: 3048
>Category: kernel
>Synopsis: scsi disk errors, sd"x": not queued, error 5
>Confidential: no
net
>Environment:

	Gateway 6400 server with four ibm scsi hard drives using 53c1010 scsi chip controller
	System      : OpenBSD 3.2
	Architecture: OpenBSD.i386
	Machine     : i386

>Description:

I have a couple Gateway 6400 servers I have been happily running OpenBSD on. After my upgrade from 3.1 to 3.2 I have been starting to see errors under load. The errors I see are:

Dec 20 09:00:20 drc /bsd: sd0: not queued, error 5
Dec 20 09:00:20 drc last message repeated 6 times
Dec 20 09:00:20 drc /bsd: sd3: not queued, error 5
Dec 20 09:00:20 drc /bsd: sd0: not queued, error 5

The machine does not crash immediately, but it stops talking to the disk involved with the queue error. The issue occurs on both machines using the GENERIC kernel, and "tuned" ones. The system does continue to allow logins, but will not complete a "sync", nor will it reboot.         

>How-To-Repeat:

I seem to be able to get this to occure by loading up the disks with i/o. I have seen it by doing a compile, while gunziping srcsys.tar, while doing an additional compile.         

>Fix:

Don't do any work on the machine.. I wonder if this has to to with the reintergration of the NetBSD siop device driver, between release 3.1 and 3.2.

>Release-Note:
Received on Sun Dec 29 22:08:03 2002