Subject: kern/2156: time outs of Exabytes
To: None <gnats-bugs@NetBSD.ORG>
From: Klaus Helbing <helbing@phop5.physik.uni-bonn.de>
List: netbsd-bugs
Date: 03/01/1996 21:22:16
>Number: 2156
>Category: kern
>Synopsis: time outs of Exabytes
>Confidential: no
>Severity: critical
>Priority: medium
>Responsible: kern-bug-people (Kernel Bug People)
>State: open
>Class: change-request
>Submitter-Id: net
>Arrival-Date: Fri Mar 1 15:50:02 1996
>Last-Modified:
>Originator: Klaus Helbing
>Organization:
Physikalisches Institut Uni Bonn
>Release: NetBSD-current Feb 7
>Environment:
System: NetBSD phop5.physik.uni-bonn.de 1.1A NetBSD 1.1A (PHOP) #0: Wed Feb 7 15:50:38 MET 1996 helbing@phop5.physik.uni-bonn.de:/usr/src/sys/arch/i386/compile/PHOP i386
>Description:
NetBSD crashes every once in a while during write to Exabyte-Tape.
I tried to find out what happens and compiled /sys/scsi/st.c with
-DSCSIDEBUG into the kernel. The last lines before the crash in
/var/log/messages look the following:
Feb 29 15:39:40 phop5 /netbsd: st0(ahc0:4:0): ststart
Feb 29 15:39:40 phop5 /netbsd: st0(ahc0:4:0): ststart st0(ahc0:4:0): ststrategy 65536 bytes @ blk 1613440
Feb 29 15:39:40 phop5 /netbsd: st0(ahc0:4:0): ststart st0(ahc0:4:0): ststart
Feb 29 15:39:40 phop5 /netbsd: st0(ahc0:4:0): ststrategy 65536 bytes @ blk 1613568
Feb 29 15:39:40 phop5 /netbsd: st0(ahc0:4:0): ststart
Feb 29 15:39:40 phop5 /netbsd: st0(ahc0:4:0): ststart st0(ahc0:4:0): ststrategy 65536 bytes @ blk 1613696
Feb 29 15:39:40 phop5 /netbsd: st0(ahc0:4:0): ststart st0(ahc0:4:0): ststart
Feb 29 15:39:41 phop5 /netbsd: st0(ahc0:4:0): ststrategy 65536 bytes @ blk 1613824
Feb 29 15:39:41 phop5 /netbsd: st0(ahc0:4:0): ststart
Feb 29 15:39:41 phop5 /netbsd: st0(ahc0:4:0): ststart st0(ahc0:4:0): ststrategy 65536 bytes @ blk 1613952
Feb 29 15:39:41 phop5 /netbsd: st0(ahc0:4:0): ststart st0(ahc0:4:0): ststart
Feb 29 15:39:41 phop5 /netbsd: st0(ahc0:4:0): ststrategy 65536 bytes @ blk 1614080
Feb 29 15:39:41 phop5 /netbsd: st0(ahc0:4:0): ststart
Feb 29 15:39:41 phop5 /netbsd: st0(ahc0:4:0): timed out
Feb 29 15:39:41 phop5 /netbsd: st0(ahc0:4:0): closing
This behavior is not specific to the Adaptec 2940 SCSI controller
used in this example. The same happens with an NCR53815 controller.
>How-To-Repeat:
Write to Exabyte using tar (no dependency on block-size found).
The amount of data should be of the order of 20GB.
>Fix:
I increased the time outs in all scsi_scsi_cmd calls in st.c to
900000 which lead to "more" seldom crashes.
>Audit-Trail:
>Unformatted: