Subject: kern/172: SCSI tape soft error bugreport
To: None <gnats-admin>
From: Hellmuth Michaelis <hm@hcshh.hcs.de>
List: netbsd-bugs
Date: 03/19/1994 09:50:02
>Number: 172
>Category: kern
>Synopsis: soft error while rd/wr to scsi-tape stops backup operation
>Confidential: no
>Severity: serious
>Priority: medium
>Responsible: gnats-admin (Kernel Bug People)
>State: open
>Class: sw-bug
>Submitter-Id: net
>Arrival-Date: Sat Mar 19 09:50:01 1994
>Originator: Hellmuth Michaelis
>Organization:
HCS Hanseatischer Computerservice GmbH
>Release: NetBSD-current as of March 16th 1994
>Environment:
I486 ISA, NetBSD-current, Cipher ST150-S1
System: NetBSD bert.hcs.de 0.9a BERT#2 i386
>Description:
This system has a SCSI-1 QIC Streamer Cipher ST150-S1 connected
to an Adaptec 1542B.
While reading/writing and a soft error occures, this tape drive
generates a check condition status and request sense reads a sense
key of 01h which is a "recoverable error".
No data loss occured by the above described procedure.
This behaviour unhides a bug in the currently used SCSI - driver
code, in the event of a soft error (while writing with tar), the
driver/tar writes the following lines to the console and terminates
operation:
st0: soft error (corrected)
st0: error ignored
tar: only wrote 0 of 10240 bytes to /dev/rst0
The driver announces the correct behaviour (= to ignore this type
of error) but the application program terminates.
>How-To-Repeat:
try to write using tar to a scsi-tape drive which alarms the driver
that a soft error occured.
>Fix:
Modify the scsi tape driver in such a way, that a soft error
condition is really ignored.
This behaviour has been fixed more than a year ago in the original
version of Julian's driver.
Hope this was filled-out right ...
hellmuth
--
Hellmuth Michaelis HCS Hanseatischer Computerservice GmbH Hamburg, Europe
There is a difference between an open mind and a hole in the head
>Audit-Trail:
>Unformatted:
------------------------------------------------------------------------------