[9fans] running in AMDs simnow

[9fans] running in AMDs simnow

[ron minnich]

I'm trying to get plan 9 to boot in AMDs simnow as part of tracking a
weird problem.

There's some weird issue between simnow and 9load.

Here is what happens, on floppy or iso:
initial probe, to find plan9.ini...

I then see 661,504 reads (in the case of the ISO) or few number of
reads (in the case of floppy)
and then it just hangs, then finally gets exception 6.

...FLAGS=10282 TRAP=6 ECODE=0 PC=3
  AX 00000000  BX 80815788  CX 00000000  DX 800fdf30
  SI 00001022  DI 00000024  BP 0000746b
  CS 0010 DS 0008  ES 0008  FS 0008  GS 0008
  CR0 80000011 CR2 00000000 CR3 0000c000
panic: exception/interrupt 6

Press almost any key to reset..

PC=3 is not encouraging :-)


The chipset is a standard 8111/8131 setup.

I'm confused, I'm going to try pointing it at the real ide drive next,
but suggestions are welcome.

Well, besides, 'Give up now", that is.

ron

Re: [9fans] running in AMDs simnow

[erik quanstrom]

> I'm trying to get plan 9 to boot in AMDs simnow as part of tracking a
> weird problem.
>
> There's some weird issue between simnow and 9load.
>
> Here is what happens, on floppy or iso:
> initial probe, to find plan9.ini...
>
> I then see 661,504 reads (in the case of the ISO) or few number of
> reads (in the case of floppy)
> and then it just hangs, then finally gets exception 6.
>
> ...FLAGS=10282 TRAP=6 ECODE=0 PC=3
>   AX 00000000  BX 80815788  CX 00000000  DX 800fdf30
>   SI 00001022  DI 00000024  BP 0000746b
>   CS 0010 DS 0008  ES 0008  FS 0008  GS 0008
>   CR0 80000011 CR2 00000000 CR3 0000c000
> panic: exception/interrupt 6
>
> Press almost any key to reset..
>
> PC=3 is not encouraging :-)
>
>
> The chipset is a standard 8111/8131 setup.
>
> I'm confused, I'm going to try pointing it at the real ide drive next,
> but suggestions are welcome.
>
> Well, besides, 'Give up now", that is.
>
> ron

hey.  Mathieu Lonjaret reported exactly this problem with
an ibm eserver with the same chipset.

i don't recall if he got it working.  does it work for you without simnow?

one silly idea.  any chance legacy usb is on.  system management mode is
direkt aus der hölle.  does 9loads top-of-memory match the e820 scan?

- erik

----

ELCR: 0C80
apm ax=f000 cx=f000 dx=40 di=ffff ebx=5d84 esi=-1
initial probe, to find plan9.ini... dev A0 port 1F0 config 0040 capabilities 2F00 mwdma 0007 udma 407F
dev A0 port 170 config 85C0 capabilities 2F00 mwdma 0007 udma 0407
FLAGS=10282 TRAP=6 ECODE=0 PC=3
 AX 00000000 BX 80815d78 CX 00000000 DX 800fdfc0
 SI 00001022 DI 00000024 BP 0000746b
 CS 0010 DS 0008 ES 0008 FS 0008 GS 0008
 CR0 80000011 CR2 0000000 CR3 0000c000
panic: exception/interrupt 6

Re: [9fans] running in AMDs simnow

[lejatorn]

[-- Attachment #1: Type: text/plain, Size: 1460 bytes --]

Yeah, nice coincidence as I was about to mail 9fans about my progress
on that matter.

So yes, with the help of Erik I managed to bake a bootable iso for
this ibm eserver 325 I have here, which seems to behave in a pretty
similar way to the one you describe here (look for the thread "plan 9
on an ibm eserver 325).

I'm attaching the 9load I used for that iso, which incidentally is the
one that actually works on my T61.  I don't have the sources for it as
Erik had built it for me when I had trouble with the T61.  So you
should hope that Erik still has the sources somewhere around or
directly try with that one.

For the record - although I may write a short wiki page about that
later if it's deemed usefull - here's how I made the iso:

disk/format -t 3½QD -db /386/pbs /usr/lejatorn/iso/cddisk /usr/lejatorn/iso/9load /sys/src/9/pc/9pccd.gz /sys/src/9/pc/9pcflop.gz /usr/lejatorn/iso/plan9.ini /lib/vgadb
mount /srv/9660 /n/9660 /dev/sdC0/data
bind /usr/lejatorn/iso/cddisk /n/9660/bootdisk.img
disk/mk9660 -9cj -v 'myiso' -s /n/9660 -p /sys/lib/sysconfig/proto/allproto -b bootdisk.img cd0

I'm attaching the plan9.ini I used as well.

Of course now the installed system fails in the same way it does when
booting with the default iso, since the "non-working" 9load was
installed, so I'll have to replace that one with the one that works
for me.  (Not sure I am crystal clear here, sorry).


Hope that helps,
Mathieu.

[-- Attachment #2: Type: message/rfc822, Size: 4731 bytes --]

From: erik quanstrom <quanstro@quanstro.net>
To: 9fans@9fans.net
Subject: Re: [9fans] running in AMDs simnow
Date: Tue, 13 May 2008 17:15:45 -0400
Message-ID: <dbaca129216b2e89209baf9f00aac581@quanstro.net>

> I'm trying to get plan 9 to boot in AMDs simnow as part of tracking a
> weird problem.
> 
> There's some weird issue between simnow and 9load.
> 
> Here is what happens, on floppy or iso:
> initial probe, to find plan9.ini...
> 
> I then see 661,504 reads (in the case of the ISO) or few number of
> reads (in the case of floppy)
> and then it just hangs, then finally gets exception 6.
> 
> ...FLAGS=10282 TRAP=6 ECODE=0 PC=3
>   AX 00000000  BX 80815788  CX 00000000  DX 800fdf30
>   SI 00001022  DI 00000024  BP 0000746b
>   CS 0010 DS 0008  ES 0008  FS 0008  GS 0008
>   CR0 80000011 CR2 00000000 CR3 0000c000
> panic: exception/interrupt 6
> 
> Press almost any key to reset..
> 
> PC=3 is not encouraging :-)
> 
> 
> The chipset is a standard 8111/8131 setup.
> 
> I'm confused, I'm going to try pointing it at the real ide drive next,
> but suggestions are welcome.
> 
> Well, besides, 'Give up now", that is.
> 
> ron

hey.  Mathieu Lonjaret reported exactly this problem with
an ibm eserver with the same chipset.

i don't recall if he got it working.  does it work for you without simnow?

one silly idea.  any chance legacy usb is on.  system management mode is
direkt aus der hölle.  does 9loads top-of-memory match the e820 scan?

- erik

----

ELCR: 0C80
apm ax=f000 cx=f000 dx=40 di=ffff ebx=5d84 esi=-1
initial probe, to find plan9.ini... dev A0 port 1F0 config 0040 capabilities 2F00 mwdma 0007 udma 407F
dev A0 port 170 config 85C0 capabilities 2F00 mwdma 0007 udma 0407
FLAGS=10282 TRAP=6 ECODE=0 PC=3
 AX 00000000 BX 80815d78 CX 00000000 DX 800fdfc0
 SI 00001022 DI 00000024 BP 0000746b
 CS 0010 DS 0008 ES 0008 FS 0008 GS 0008
 CR0 80000011 CR2 0000000 CR3 0000c000
panic: exception/interrupt 6


[-- Attachment #3: Type: text/plain, Size: 638 bytes --]

[menu]
menuitem=install, Install Plan 9 from this CD
menuitem=cdboot, Boot Plan 9 from this CD
menuitem=debug, Boot Plan 9 from this CD and debug 9load

[common]
*nobiosload=1
*nomp=1
*nodumpstack=1
partition=new
mouseport=ask
monitor=ask
vgasize=ask
dmamode=ask
adisk=/dev/sdD0/cdboot
cdboot=yes

[install]
bootargs=local!#S/sdD0/data
nobootprompt=local!/boot/bzroot
bootfile=sdD0!cdboot!9pcflop.gz
*nobiosload=1
#*noahciload=1

[cdboot]
bootargs=local!#S/sdD0/data
bootfile=sdD0!cdboot!9pccd.gz
*nobiosload=1
#*noahciload=1

[debug]
bootargs=local!#S/sdD0/data
bootfile=sdD0!cdboot!9pccd.gz
*debugload=1

[-- Attachment #4: Type: application/octet-stream, Size: 224956 bytes --]

Re: [9fans] running in AMDs simnow

[ron minnich]

On Tue, May 13, 2008 at 7:13 PM,  <lejatorn@gmail.com> wrote:

> mount /srv/9660 /n/9660 /dev/sdC0/data

This one lost me completely. /dev/sdC0/data? That's an iso on your system?
ron

Re: [9fans] running in AMDs simnow

[lejatorn]

On Tue, May 13, 2008 at 03:59:45PM -0700, ron minnich wrote:
> On Tue, May 13, 2008 at 7:13 PM,  <lejatorn@gmail.com> wrote:
>
> > mount /srv/9660 /n/9660 /dev/sdC0/data
>
> This one lost me completely. /dev/sdC0/data? That's an iso on your system?

Yeah sorry, I suspected it would not be obvious.
That's simply the "official" plan9 iso in my cd drive. The end result
is that I generate a new iso which has everything of the regular one,
except for the bootdisk.img which is replaced with the cddisk
previously created.

Mathieu.

Re: [9fans] running in AMDs simnow

[erik quanstrom]

> So yes, with the help of Erik I managed to bake a bootable iso for
> this ibm eserver 325 I have here, which seems to behave in a pretty
> similar way to the one you describe here (look for the thread "plan 9
> on an ibm eserver 325).
>
> I'm attaching the 9load I used for that iso, which incidentally is the
> one that actually works on my T61.  I don't have the sources for it as
> Erik had built it for me when I had trouble with the T61.  So you
> should hope that Erik still has the sources somewhere around or
> directly try with that one.

this wasn't ment to be secretive.  i believe the reason this version of
9load works is that it doesn't call bios.  (smm is aus der hölle.)

the source has been in /n/sources/contrib/quanstro/src/9loadaoe.
i just updated it to cover a fix for macbooks with ahci.

- erik

Re: [9fans] running in AMDs simnow

[lejatorn]

On Tue, May 13, 2008 at 07:10:20PM -0400, erik quanstrom wrote:
> > So yes, with the help of Erik I managed to bake a bootable iso for
> > this ibm eserver 325 I have here, which seems to behave in a pretty
> > similar way to the one you describe here (look for the thread "plan 9
> > on an ibm eserver 325).
> > 
> > I'm attaching the 9load I used for that iso, which incidentally is the
> > one that actually works on my T61.  I don't have the sources for it as
> > Erik had built it for me when I had trouble with the T61.  So you
> > should hope that Erik still has the sources somewhere around or
> > directly try with that one.
> 
> this wasn't ment to be secretive.  i believe the reason this version of
> 9load works is that it doesn't call bios.  (smm is aus der hölle.)

sorry if it sounded that way, I did not meant that.
It's just that as you once told me yourself, the fixes you make
sometimes never end up in the default distribution, either because you
forget or because as you point out below, it goes into contrib.

What is the rationale between stuff going into contrib and stuff going
into the distro anyway? If the 9load you made is the one allowing to
boot machines that don't boot with the one on the iso, shouldn't yours
go onto the iso so that newcomers benefit from it from the beginning?
What am I missing here?

> the source has been in /n/sources/contrib/quanstro/src/9loadaoe.
> i just updated it to cover a fix for macbooks with ahci.
> 
> - erik

Mathieu.

Re: [9fans] running in AMDs simnow

[erik quanstrom]

> sorry if it sounded that way, I did not meant that.
> It's just that as you once told me yourself, the fixes you make
> sometimes never end up in the default distribution, either because you
> forget or because as you point out below, it goes into contrib.
>
> What is the rationale between stuff going into contrib and stuff going
> into the distro anyway? If the 9load you made is the one allowing to
> boot machines that don't boot with the one on the iso, shouldn't yours
> go onto the iso so that newcomers benefit from it from the beginning?
> What am I missing here?

there's no rationale.

my use of contrib is mostly an unfortunate accident.  sometimes it takes
a while to get changes into the distribution.  i can, however put stuff in
contrib immediately.

sorry.

- erik

Re: [9fans] running in AMDs simnow

[ron minnich]

I love it. 9load is dying in real mode on:

1000:012C B80800           mov ax,0008
1000:012F 8ED8             mov ds,ax
                                   ^^^^^^^^^^^^^^^^^^^

takes a triple fault, which is interesting, as vm is not on.

But this is kind of intriguing because that's not how I recall setting
up the ds etc. when I've done it.

So I looked at how we did it long ago in coreboot (a.k.a. LinuxBIOS).
And we don't do this at all. We set up cs, jump to protected mode,
then set up ds etc.

But 9load works. On real hardware. Just not on the sim ...

well, I moved the ds set past the 32-bit mode point and it's back to
not reading all of disk again.

off to bed.

ron

Re: [9fans] running in AMDs simnow

[Bakul Shah]

On Tue, 13 May 2008 13:55:43 PDT "ron minnich" <rminnich@gmail.com>  wrote:
> I'm trying to get plan 9 to boot in AMDs simnow as part of tracking a
> weird problem.
>
> There's some weird issue between simnow and 9load.
>
> Here is what happens, on floppy or iso:
> initial probe, to find plan9.ini...
>
> I then see 661,504 reads (in the case of the ISO) or few number of
> reads (in the case of floppy)
> and then it just hangs, then finally gets exception 6.
	...
> I'm confused, I'm going to try pointing it at the real ide drive next,
> but suggestions are welcome.
>
> Well, besides, 'Give up now", that is.

Depending on what you are trying to do qemu-system-x86_64 may
be an option.  That is what I have been using for a while now.