[9fans] MBR and nvidia hangups

[9fans] MBR and nvidia hangups

[Rob Ristroph]

Hi, I've been doing some tests this Sunday morning and here are my
results:

A normal Plan 9 install, where Plan 9 is the only thing on a 60 GB
disk (except I also made a boot floppy):

Doesn't overwrite MBR, previously not working MBR gives an error
message (depends on what was left in there).  Also the boot floppy
wants to ream the filesystem and then panics when I use it.


Start over, re-install, this time at the [bootsetup] prompt in
addition to selecting plan9 as the method, I open another window and
type:

disk/mbr -m /386/pbslba /dev/sdC0

On reboot the machine hangs after "Verifying DMI Pool Data ......."
(which is also what happens if there is no MBR at all).

I boot from the install disk, make a new window, and run:

disk/mbr -m /386/pbs /dev/sdC0/data

This results in the message "I/0 err" followed by junk characters.

I did the same but this time used the command:

disk/mbr -m /386/mbr /dev/sdC0/data

And that worked well enough that I got to type "glenda" to log in for
the first time in weeks !  But it hung shortly after that on the "kfs
. . .version" stage.

Figuring I might finally have a working MBR, I re-imaged the install
floppy fresh and started in again.  It quickly did a "mountfs" and
then I chose to keep the plan 9 filesystem instead of wiping it; that
hung doing something for longer than it should take to wipe it, so I
rebooted, and this time wiped it and re-installed everything.

It now boots, allows me to select the disk or network, and log in as
glenda.  However, no rio comes up:

init: starting /bin/rc
aux/vga: nvidia: can't attach mmio segment
dossrv: serving #s/dos
rio: can't open display: initdisplay: /dev/draw/new: no frame buffer
init: rc exit status: rio 13: display open

init: starting /bin/rc
term%

(Note this is a card which was previously running Plan 9 in graphical
mode just fine, and will come up in graphical mode from the
installation floppy.)

I did

9fat:
cat /n/9fat/plan9.ini

but the plan9.ini file looked normal.

This is essentially where I was on the 6th of July, when I wiped my
drive with zeros in an attempt to make sure I had a fresh install.

However, I moved the drive to a machine that had an S3 video card
instead of the nvidia.  Then it booted fine, giving me a working Plan
9 with rio and everything.

In conclusion:

1) We have some work to do on the MBR/bootloader part of the
   installation procedure.  Once I finish configuring my machine I
   might try to take apart the installation floppy and find the
   installation program on it.  Any suggestions on that ?

2) We might have broken nvidia chip support ?  This seems improbable,
   but my card does still behave normally with knoppix.  I won't
   investigate this further right now, because I know where I can buy
   an S3 Virge for $5 and I'm on my way there now.

3) Now I can start figuring out how to authenticate and share services
   between my two Plan 9 machines, and shortly, other peoples !

Cheers,

--Rob

Re: [9fans] MBR and nvidia hangups

[David Presotto]

The new 'experimental' diskette will work if you edit the plan9.ini to say

nobootprompt=local!/boot/bzroot

instead of

nobootprompt=local!/bzroot

Things have changed...

I'll find a new way to do an experimental diskette tonight and
ask people to try it out.  I think I've fixed the mbr problem
and tonight we should also have a fix for the nvidia one.

The mbr problem (that I found) was that the install only stuck
in a new mbr if the plan9 partition information was bad.  That
meant that you had to zero 9fat to get a new mbr, not easily
guessed.

Re: [9fans] MBR and nvidia hangups

[Rob Ristroph]

>>>>> "David" == David Presotto <presotto@closedmind.org> writes:
David>
David> The new 'experimental' diskette will work if you edit the
David> plan9.ini to say nobootprompt=local!/boot/bzroot
David>
David> instead of
David>
David> nobootprompt=local!/bzroot
David>
David> Things have changed...
David>
David> I'll find a new way to do an experimental diskette tonight and
David> ask people to try it out.  I think I've fixed the mbr problem
David> and tonight we should also have a fix for the nvidia one.
David>
David> The mbr problem (that I found) was that the install only stuck
David> in a new mbr if the plan9 partition information was bad.  That
David> meant that you had to zero 9fat to get a new mbr, not easily
David> guessed.

Cool !  I'll find time for a round of testing Wednesday or Thursday.
I haven't configured the fresh install so I have nothing to loose by
repeating installs in a few different configurations, with large and
small disks.  I can also try having the Plan 9 disk be other than the
first IDE device, and any other weird cases folks think should tested.

--Rob

Re[2]: [9fans] MBR and nvidia hangups

[Martin Althoff]

David>> I'll find a new way to do an experimental diskette tonight and
David>> ask people to try it out.  I think I've fixed the mbr problem

> Cool !  I'll find time for a round of testing Wednesday or Thursday.

Same on my part! Just rigged up another system without case...

Ehmm... David, where do (will) I find that diskette? 'xcuse my
newbieishness...
cheers,
Martin

Re: Re[2]: [9fans] MBR and nvidia hangups

[David Presotto]

[-- Attachment #1: Type: text/plain, Size: 86 bytes --]

Sorry, about that.  My mom got sick and I got distracted.  I'll
get back to it today.

[-- Attachment #2: Type: message/rfc822, Size: 2601 bytes --]

From: Martin Althoff <martin.althoff@tiscali.co.uk>
To: Rob Ristroph <9fans@cse.psu.edu>
Subject: Re[2]: [9fans] MBR and nvidia hangups
Date: Tue, 22 Jul 2003 20:57:50 +0100
Message-ID: <842125109.20030722205750@tiscali.co.uk>

David>> I'll find a new way to do an experimental diskette tonight and
David>> ask people to try it out.  I think I've fixed the mbr problem

> Cool !  I'll find time for a round of testing Wednesday or Thursday.

Same on my part! Just rigged up another system without case...

Ehmm... David, where do (will) I find that diskette? 'xcuse my
newbieishness...
cheers,
Martin

Re: Re[2]: [9fans] MBR and nvidia hangups

[David Presotto]

[-- Attachment #1: Type: text/plain, Size: 81 bytes --]

Oops, I didn't mean to send that to the list.  Sorry and thanks
for the support.

[-- Attachment #2: Type: message/rfc822, Size: 4745 bytes --]

[-- Attachment #2.1.1: Type: text/plain, Size: 86 bytes --]

Sorry, about that.  My mom got sick and I got distracted.  I'll
get back to it today.

[-- Attachment #2.1.2: Type: message/rfc822, Size: 2601 bytes --]

From: Martin Althoff <martin.althoff@tiscali.co.uk>
To: Rob Ristroph <9fans@cse.psu.edu>
Subject: Re[2]: [9fans] MBR and nvidia hangups
Date: Tue, 22 Jul 2003 20:57:50 +0100
Message-ID: <842125109.20030722205750@tiscali.co.uk>

David>> I'll find a new way to do an experimental diskette tonight and
David>> ask people to try it out.  I think I've fixed the mbr problem

> Cool !  I'll find time for a round of testing Wednesday or Thursday.

Same on my part! Just rigged up another system without case...

Ehmm... David, where do (will) I find that diskette? 'xcuse my
newbieishness...
cheers,
Martin

Re: [9fans] MBR and nvidia hangups

[gene]

rgr@sdf.lonestar.org (Rob Ristroph) wrote in message news:<87d6g5gd8e.fsf@rgristroph-austin.ath.cx>...

> However, no rio comes up:
> 
> init: starting /bin/rc
> aux/vga: nvidia: can't attach mmio segment
> dossrv: serving #s/dos
> rio: can't open display: initdisplay: /dev/draw/new: no frame buffer
> init: rc exit status: rio 13: display open
> 
> init: starting /bin/rc
> term%

I also ran into this problem installing on a system with a Creative
Labs Riva TNT board and was able to get rio working by making a change
in 9/pc/vganvidia.c nvidiapci:

/* Nvidia is good about backwards compatibility -- any did > 0x20 is
fine */
static Pcidev*
nvidiapci(void)
{
	Pcidev *p;

	p = nil;
	while((p = pcimatch(p, 0x10DE, 0)) != nil){
		if(p->did > 0x20 && p->ccrb == 3)	/* video card */
			return p;
	}
	return nil;
}

I believe the test should be:
                
                 if(p->did >= 0x20 && p->ccrb == 3) /* video card */

An earlier lib/vgadb shows the following for Nvidia

ctlr
        vid=0x10DE did=0x0020           # Riva TNT
        vid=0x10DE did=0x0028           # Riva TNT2
        vid=0x10DE did=0x0029           # Riva TNT2 Ultra
        vid=0x10DE did=0x002C           # Riva TNT2 Vanta
        vid=0x10DE did=0x002D           # Riva TNT2 M64
        vid=0x10DE did=0x00A0           # Riva TNT2 Integrated
        vid=0x10DE did=0x0100           # GeForce 256
        vid=0x10DE did=0x0101           # GeForce DDR
        vid=0x10DE did=0x0103           # Quadro
        vid=0x10DE did=0x0110           # GeForce2 MMX
        vid=0x10DE did=0x0111           # GeForce2 MMX DDR
        vid=0x10DE did=0x0112           # GeForce2 Go
        vid=0x10DE did=0x0113           # Quadro 2 MXR
        vid=0x10DE did=0x0150           # GeForce2 GTS
        vid=0x10DE did=0x0151           # GeForce2 GTS (rev 1)
        vid=0x10DE did=0x0152           # GeForce2 Ultra
        vid=0x10DE did=0x0153           # Quadro 2 Pro  
        vid=0x10DE did=0x0200           # GeForce3
        link=vga
        ctlr=nvidia linear=1
        hwgc=nvidiahwgc


-gwg