NetBSD Problem Report #4815

Received: (qmail 3149 invoked from network); 13 Jan 1998 13:51:01 -0000
Message-Id: <199801131349.OAA00440@cosinus.cs.uni-bonn.de>
Date: Tue, 13 Jan 1998 14:49:56 +0100 (CET)
From: Ignatios Souvatzis <ignatios@cosinus.cs.uni-bonn.de>
Reply-To: is@netbsd.org
To: gnats-bugs@gnats.netbsd.org
Subject: AHC driver can't make ARCHIVE VIPER 150 stream.
X-Send-Pr-Version: 3.95

>Number:         4815
>Category:       kern
>Synopsis:       AHC driver can't make ARCHIVE VIPER 150 stream.
>Confidential:   no
>Severity:       non-critical
>Priority:       low
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Tue Jan 13 06:05:02 +0000 1998
>Closed-Date:    
>Last-Modified:  Thu Mar 22 23:23:37 +0000 2001
>Originator:     Ignatios Souvatzis
>Release:        1.3
>Organization:
computer science department, university of Bonn, Germany
>Environment:

System: NetBSD cosinus.cs.uni-bonn.de 1.3_ALPHA NetBSD 1.3_ALPHA (COSINUS) #1: Sat Jan 10 01:49:13 CET 1998 ignatios@cosinus.cs.uni-bonn.de:/usr/local/obj/kernel/COSINUS i386

	Release branch as of early January 1998

>Description:
When writing to an Archive Viper 150, using a DC6150 tape (density code 16),
from a Pentium 166 MMX + Adaptec 2940, the tape doesn't stream. Instead it
moves about 3 seconds forth, and 1 or 2 backwards.

The data rate needed for streaming is about 100 kbytes/s sustained.

Note that:
- the tape does stream when reading.
- the same tape drive (yes, i carried my home one to work to make sure)
and the same tape in it do stream, if driven by NetBSD/Amiga on DraCo
(68060/50MHz, ncr53c710).

>How-To-Repeat:
cosinus# dmesg | grep st1
st1 at scsibus0 targ 5 lun 0: <ARCHIVE, VIPER 150  21247, -005> SCSI1 1/sequential removable
st1: rogue, density code 0x0, 512-byte blocks, write-enabled
cosinus# mknod /dev/nrst1h c 14 21
cosinus# dd if=/dev/zero of=/dev/nrst1h bs=32k count=100
100+0 records in
100+0 records out
3276800 bytes transferred in 140 secs (23405 bytes/sec)
cosinus# mt -f /dev/nrst1h rewind
cosinus# dd if=/dev/nrst1h of=/dev/null bs=32k
100+0 records in
100+0 records out
3276800 bytes transferred in 34 secs (96376 bytes/sec)


>Fix:
	not known.
>Release-Note:
>Audit-Trail:

From: Ignatios Souvatzis <ignatios@cosinus.cs.uni-bonn.de>
To: gnats-bugs@netbsd.org
Cc:  Subject: kern/4815
Date: Tue, 16 Jun 1998 11:09:47 +0200

 Further tests with different jumper settings of the Archive Viper 150 show
 this:

 a) originally, the buffer size of the drive was 16 kBytes.

 b) when increasing it to 32 kBytes, the performance was marginally better:
    dd if=/dev/zero of=/dev/nrst0h bs=64k	and
    dd if=/dev/zero of=/dev/nrst0h bs=128k

 are streaming, tar cbvf 128 /dev/nrst0h is streaming, tar with other 
 block sizes and dump at any block size are not streaming.

 The comparison drive normally on the NetBSD/Amiga machine mentioned
 in the original PR is set to 16 kBytes.
>Unformatted:

Home	PR Database Search