*BSD News Article 42737


Return to BSD News archive

Xref: sserve alt.comp.periphs.mainboard.asus:96 comp.os.386bsd.questions:16734 comp.periphs.scsi:29811 comp.sys.ibm.pc.hardware.systems:12605
Path: sserve!newshost.anu.edu.au!harbinger.cc.monash.edu.au!bunyip.cc.uq.oz.au!munnari.oz.au!spool.mu.edu!howland.reston.ans.net!Germany.EU.net!unibwh.unibw-hamburg.de!news.rrz.uni-hamburg.de!news.dkrz.de!news.tu-harburg.de!news.dfn.de!RRZ.Uni-Koeln.DE!usenet
From: se@fileserv1.MI.Uni-Koeln.DE (Stefan Esser)
Newsgroups: alt.comp.periphs.mainboard.asus,comp.os.386bsd.questions,comp.periphs.scsi,comp.sys.ibm.pc.hardware.systems
Subject: Re: Problems with PVI-486AP4, NCR53c810 and FBSD 2.0
Date: 21 Feb 1995 10:46:38 GMT
Organization: Institute for Mathematics, University of Cologne, Germany
Lines: 158
Distribution: world
Message-ID: <3icgae$bma@news.rrz.uni-koeln.de>
References: <3iab71$5qu@gate.sinica.edu.tw>
NNTP-Posting-Host: fileserv1.mi.uni-koeln.de

This has just been solved, with the help of 
Sebastian Strollo <seb@erix.ericsson.se>:

The Aries doesn't seem to support the address
we choose to map NCR chip to.

It works with I/O mapping, and I can make a 
kernel available to you via anon. FTP (or send 
you a patch).

We are working on a modification to the PCI 
attach code, that will make it chose a suitable 
address for all systems ...

In article <3iab71$5qu@gate.sinica.edu.tw>, lim@gate.sinica.edu.tw (Carmay Lim) writes:
|>     This is a semi-followup to my wildly popular "List of recommended
|> hardware components" thread.  ;-)  For some reason or another, the
|> dealer I talked to could not get an ASUS PCI/I-486SP3G for me, even
|> though ASUS is based right here in Taipei.  Anyway, he sent me an ASUS
|> PVI-486AP4 motherboard (Intel Aries chipset) with the SMC Super I/O
|> card and the PCI-SC200 SCSI-2 card (based on the NCR53c810) instead.
|> It has an AMD 486DX4/100 with 256K cache installed.  The NCR's boot
|> message says it is running "v3.0 SCSI BIOS, PCI Rev. 2.0,
|> NCRPCI-3.04.00".  A Quantum Empire 1080S is attached to the SCSI-2
|> card.
|> 
|>     When I try to boot with the FreeBSD 2.0-950210-SNAP disks, I get
|> through most of the device detection process.  When the PCI part comes
|> up though, I see this:
|> 
|> 
|> [...]
|> npx0: INT16 interface
|> pci0: scanning device 0..31, mechanism=1.
|> pci0:5: INTEL CORPORATION, device=0x486, class=old [not supported]
|> ncr0 <ncr53c810 scsi> int a (config) not bound on pci0:9
|>         reg20: virtual=0xf2d04000 physical=0xc0000000
|> CACHE TEST FAILED: host wrote 1, ncr read -1.
|> CACHE TEST FAILED: ncr wrote 2, host read 1.
|> CACHE TEST FAILED: ncr wrote 2, read back -1.
|> CACHE INCORRECTLY CONFIGURED.
|> vga0 <display device> on pci0:11
|> pci uses physical addresses from 0xc0000000 to 0xc0001000
|> changing root device to fd0c
|> [...]
|> 
|> 
|>     Question 1:  why does it say "class=old [not supported]" in the
|> third line?  Is this particular motherboard/chipset/CPU combination
|> not supported under FreeBSD?

Not supported just says that there is no special 
driver for that chip, it has been initialised by 
the BIOS (or you wouldn't have been able to boot
that far :).

|>     Question 2:  the NCR is plugged into the first PCI slot and the
|> Miro S3-964 2-meg graphics card is plugged into the third PCI slot.  I
|> assume this corresponds to the number after the "pci0:" label.  Are
|> there any particular BIOS settings that need to be changed?  I scanned
|> though some old messages and tried some of the ideas mentioned there,
|> but no luck so far. These are some of my BIOS settings (all set to
|> "Setup Defaults"):
|> 
|>   Slot 1 (master):  Latency timer:  80 PCI Clock
|>                     Using IRQ:      NA
|>                     Trigger method: Level (auto)
|> 
|>   Video BIOS Shadow: Enabled
|>   PCI Posted Write Buffer: Enabled
|>   CPU To PCI Bursting: Enabled
|>   CPU To PCI Byte Merging: Disabled

Should work with modern chip sets and I/O boards.

|>   Fast Page Code Read: Enabled
|>   Fast Page Data Read: Enabled
|>   Fast Page Write: Enabled
|>   Pipelined CAS: Disabled
|>   DRAM Timing/Mragin: 70ns/Max
|>   SRAM Timing: Normal
|>   Cache Update Scheme: Write-back
|>   Video BIOS Cacheable: Disabled

Seems all Ok.

|>     Question 3:  Which cache is it testing?  The Quantum's disk cache
|> or the SCSI controller's?  The manual mentions nothing about
|> configuring disk cache.

It's testing the secondary cache, connected to the
CPU chip set. The cache isn't bad on the ASUS AP4,
only the NCR can't be attached to the address we
choose, and the test fails for that reason ...
Will be corrected real soon now.

|>     Question 4:  When it tries to reboot, FreeBSD prints the usual
|> "Syncing disks" and "Rebooting..." messages, but it never actually
|> reboots the machine.  It just hangs.  I have to hit the hardware reset
|> myself.  Also, the memory check stops at 1024K and the subsequent BIOS
|> boot screen says I have 640K RAM and 0K extended.  FreeBSD then hangs
|> before it gets to its own memory scanner.  Sound like a motherboard
|> defect?

Can't say anything regarding that ...
Does it show that effect when booting from the 
Adaptec, too ?

|>     Note that when FreeBSD does continue to boot, sysinstall does not
|> recognize the drive and therefore I cannot fdisk it.  I was able to
|> swap in an Adaptec 1542C from another 486 and got as far as unpacking
|> the contents of the cpio floppy.  Now if I try to reboot with the
|> Adaptec, it goes into an endless loop at the FreeBSD boot manager
|> prompt.  It prints "Default: F?" (with the question mark) after listing
|> the two OS choices (DOS and BSD, in my case).  No matter what key I
|> hit, it just redisplays the prompt.

The OS selector and boot loader use BIOS accesses,
everything beyond that does direct NCR programming.
If the disk got a correct label written, then it 
will work under the NCR when installed using an 
Adaptec. But I found that in some cases I had to 
start a DOS install (just up to formatting drive C:)
to get a correct MBR ...

|>     If I put the NCR SCSI back, it boots up normally.  The same PCI
|> error messages are displayed, and the kernel panics when it can't mount
|> sd0a as the root filesystem.  I don't have another PCI system to play
|> with, so it could be a motherboard defect for all I know.  MS-DOS, of
|> course, doesn't give a hoot about what kind of drive it boots from,
|> but FreeBSD certainly takes exception to it.  Does someone have a
|> solution to this, or should I insist that the dealer find me a working
|> SP3G board?  Any advice is greatly appreciated.

The kernel can't access the NCR, since it doesn't 
get initialised after the cache test failed.
We have patches that will allow you to use the AP4
(it's in FreeBSD-current, use "options NCR_IOMAPPED"
in your kernel config file).

The boot messages should include a line reading:

 ncr0 scanning for targets 0..6 (V2 pl14 95/02/15)

(Will work with later versions, too, of course).


A fix that does work without any special config 
options will be in -current soon ...


Regards,
-- 
 Stefan Esser				Internet:	<se@ZPR.Uni-Koeln.DE>
 Zentrum fuer Paralleles Rechnen	Tel:		+49 221 4706019
 Universitaet zu Koeln			FAX:		+49 221 4705160
 Weyertal 80
 50931 Koeln