Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

[9fans] MBR and nvidia hangups

0 views
Skip to first unread message

Rob Ristroph

unread,
Jul 20, 2003, 4:21:38 PM7/20/03
to

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

David Presotto

unread,
Jul 22, 2003, 7:52:39 AM7/22/03
to
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.

Rob Ristroph

unread,
Jul 22, 2003, 12:31:35 PM7/22/03
to

>>>>> "David" == David Presotto <pres...@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

Martin Althoff

unread,
Jul 22, 2003, 3:58:28 PM7/22/03
to
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


David Presotto

unread,
Jul 29, 2003, 9:25:36 AM7/29/03
to
Sorry, about that. My mom got sick and I got distracted. I'll
get back to it today.

David Presotto

unread,
Jul 29, 2003, 9:30:36 AM7/29/03
to
Oops, I didn't mean to send that to the list. Sorry and thanks
for the support.

gene

unread,
Aug 7, 2003, 4:39:02 AM8/7/03
to
r...@sdf.lonestar.org (Rob Ristroph) wrote in message news:<87d6g5g...@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

0 new messages