*BSD News Article 8393


Return to BSD News archive

Path: sserve!manuel.anu.edu.au!munnari.oz.au!news.hawaii.edu!sh.wide!wnoc-tyo-news!rena!specgw!amuraim
From: amurai@tama.spec.co.jp (Atsushi Murai)
Newsgroups: comp.unix.bsd,fj.os.386bsd
Subject: [386BSD] Problem of New SCSI system for BT742A (w/investgation)
Keywords: BT742A
Message-ID: <23@tama.spec.co.jp>
Date: 1 Dec 92 16:32:23 GMT
Sender: "Atsushi Murai" <amurai@tama.spec.co.jp>
Followup-To: comp.unix.bsd,fj.os.386bsd
Distribution: world,fj,local
Organization: System Planning and Engineering Co,.ltd
Lines: 44

Julian Elischer (julian@tfs.com) wrote:

>if you can find someone with internet access, you can try the kernel
>running on ref.tfs.com
>(there are japanese users, some may be close to you)
>>Yes, I will. I'm going to ask someone for sending it.
<stuff deleted>
>I suspect a faulty board.. I have 3 running here
<stuff deleted>
>I cannot get at the boards now to look at the version, but I have been running
>this drive now for nearly 2 years. NEVER with this sort of problem.
>some 20 or 30 busteks running under both 386bsd and MACH2.6.
>If there have been version changes over those 2 years, they have all worked.

 After following by Mr. Julian Elischer, I have no luck to get neither the
kernel running on ref.tfs.com nor get(break?) into his living room ;-)
But during investgation myself, I sucess the system is brought up without
this problem.

 When I just set the debug flag(bt_debug) as BT_SHOWCCBS, the system
is brought up after having a lot of "<start ccb(xxxxxxxx)>","<int
ccb(xxxxxxxx)>" messages. And then I attempted do heavy disk activity
such a compiling the 386BSD kernel, fsck -n, dd if=<swap> of=/dev/null
and so on. It's still alive! 
  
 From this result, I assume that the cpu is too fast for I/O or
relating Disk drive performance. so I just increse the value of
delaycount to 50% and insert "spinwait(1);" line before "if
(BT_SHOWCCB & bt_debug)" in bt_scsi_cmd routine. Finaly it's up and
running a couple of days without timeouted!  (Note: original value of
delaycount is 0x1591 on NICE EISA 486-33MHz + 256B-20ns Ex-Cash +
8MB-70ns DRAM)

But I still don't understand mechanism of this problem, this fix could
be wrong.  So I would appreciate any suggestion, information ( firm
ware version ant so on) about this issue. 

Atsushi Murai.

# Sorry, I'm STILL not permited doing a mail from/to out side of JAPAN.
# Thank you.
--
Atsushi Murai                                        TEL:+81-3-3833-5341
System Planning & Engineering co,.                   FAX:+81-3-3832-2779