Re: [OpenBIOS] Aladdin V chipset

[Ronald G. Minnich]

On Mon, 28 Feb 2000, Burl Nyswonger wrote:
> > when you say super socket 7 chipsets -- does this include AMD?
> > edwin
>
> I meant "Super Socket 7 boards" in general -- ie: the Asus P5 series, Acer
> boards (obviously), etc...  When I say "Aladdin V Chipset", I am referring

OK, this brings up an interesting issue for the OpenBIOS community. At
least one Socket 7 motherboard will support either alpha or pentium in the
socket. So it's the same motherboard with a different processor. All
startup tasks for the motherboard are pretty much the same.

I had an interesting set of discussions with vendors at extreme linux last
week. There's a lot of interest in linuxbios -- from Alpha motherboards
too (the alpha bios'es all have sucked very badly -- dec never got it
right).

As things stand openbios is too pentium-specific. So I'd like to propose
an idea.

1) get into protected mode early
2) get rid of as much assembly as possible
   there's a lot that can move to C

so we change the layout a bit. It's still early in the game, we can do
this.

A possible layout:

bios/arch/i386/{poweron.s,protected_mode.s,memory.s,motherboard.c,boot.c}
bios/arch/alpha/{poweron.s,protected_mode.s,memory.s,motherboard.c,boot.c}
bios/payload/{your c files here}

poweron.s is the very first set of things you do on poweron -- should be a
few instructions. protected_mode.s bumps you into protected mode. On the
pentium this is a few instructions. memory.s turns on memory -- this is
*much* easier on the pentium in protected mode. memory.s should set up a
stack and jump to C code. Motherboard.c does anything that has to be done
to further use the motherboard. Boot.c gunzips the payload (either a
bios-compatible code or Linux or FreeBSD or whatever) and jumps to it.

I'm talking to a commercial company that is going to do something like
this -- they don't want much assembly long-term, so they're recrafting
openbios. I don't think most people want to do assembly.

If we do this, we're not just for pentiums any more. That would be
wonderful! Any interest?

ron

-
To unsubscribe: send mail to [e-mail]
with 'unsubscribe openbios' in the body of the message

Aladdin V chipset

[Todd Whitesel]

> I had an interesting set of discussions with vendors at extreme linux last
> week. There's a lot of interest in linuxbios -- from Alpha motherboards
> too (the alpha bios'es all have sucked very badly -- dec never got it
> right).

I don't blame DEC for that. Try supporting hordes of cards with gnarly
BIOS code, all of it x86 assembly. They either wrote their own drivers
and supported a tiny list of cards, or wrote an x86 emulator and prayed.
> If we do this, we're not just for pentiums any more. That would be
> wonderful! Any interest?

Yes. For me at least, it is why I am on this list to begin with.

Todd Whitesel
toddpw @ best.com
-
To unsubscribe: send mail to [e-mail]
with 'unsubscribe openbios' in the body of the message

Aladdin V chipset

[Burl Nyswonger]

> On Mon, 28 Feb 2000, Burl Nyswonger wrote:
> > > when you say super socket 7 chipsets -- does this include AMD?

[...]
> get rid of as much assembly as possible
> there's a lot that can move to C
>
> bios/arch/i386/{poweron.s,protected_mode.s,memory.s,motherboard.c,boot.c}
> bios/arch/alpha/{poweron.s,protected_mode.s,memory.s,motherboard.c,boot.c}
> bios/payload/{your c files here}

I was thinking of the same scheme, ie: the Linux kernel layout...  How
about having a few directories beneath the arch-dependent dir (like
Linux), think of this: for __ARCH__ in i386 alpha ppc; do

bios/arch/$__ARCH__/include/	-- includes: *.h, *.s, *.inc, ...
  bios/arch/$__ARCH__/tools/	-- arch-dependent tools/scripts/etc...
  bios/arch/$__ARCH__/boot/	-- minimal boot-up code to get processor
    			   to have a stack and full linear memory.
> poweron.s is the very first set of things you do on poweron -- should be a
> few instructions. protected_mode.s bumps you into protected mode. On the
> pentium this is a few instructions. memory.s turns on memory -- this is
> *much* easier on the pentium in protected mode. memory.s should set up a
> stack and jump to C code. Motherboard.c does anything that has to be done
> to further use the motherboard. Boot.c gunzips the payload (either a

[...]

All of this stuff should perhaps live under the boot/ directory.

[...]
> ron

-- Burl.

-
To unsubscribe: send mail to [e-mail]
with 'unsubscribe openbios' in the body of the message

Aladdin V chipset

[Ronald G. Minnich]

On Tue, 29 Feb 2000, Burl Nyswonger wrote:
> bios/arch/$__ARCH__/include/	-- includes: *.h, *.s, *.inc, ...
>   bios/arch/$__ARCH__/tools/	-- arch-dependent tools/scripts/etc...
>   bios/arch/$__ARCH__/boot/	-- minimal boot-up code to get processor
>    			   to have a stack and full linear memory.

cool
> > poweron.s is the very first set of things you do on poweron -- should be a
> > few instructions. protected_mode.s bumps you into protected mode. On the
> > pentium this is a few instructions. memory.s turns on memory -- this is
> > *much* easier on the pentium in protected mode. memory.s should set up a
> > stack and jump to C code. Motherboard.c does anything that has to be done
> > to further use the motherboard. Boot.c gunzips the payload (either a
> All of this stuff should perhaps live under the boot/ directory.

I think .s all go in boot. Should .c? the .c for certain alpha and pentium
motherboards is almost the same code, and the gunzip.c is identical code
save for where you gunzip to.

Anyway we're on the same line of thought, I am not picky. OpenBIOS
configuration folks, is this set of changes acceptable to you?

ron

-
To unsubscribe: send mail to [e-mail]
with 'unsubscribe openbios' in the body of the message

Aladdin V chipset

Ronald G. Minnich  wrote:
> > > when you say super socket 7 chipsets -- does this include AMD?
> > > edwin
> >
> > I meant "Super Socket 7 boards" in general -- ie: the Asus P5 series, Acer
> > boards (obviously), etc...  When I say "Aladdin V Chipset", I am referring
>
> OK, this brings up an interesting issue for the OpenBIOS community. At
> least one Socket 7 motherboard will support either alpha or pentium in the
> socket. So it's the same motherboard with a different processor. All
> startup tasks for the motherboard are pretty much the same.

*shock*
Really ? I thought Alphas only ran on their own custom buses (EV etc...)
> I had an interesting set of discussions with vendors at extreme linux last
> week. There's a lot of interest in linuxbios -- from Alpha motherboards
> too (the alpha bios'es all have sucked very badly -- dec never got it
> right).

Hmm, fine hardware, not so good software.
Seems true of a quite a few vendors these days.
> As things stand openbios is too pentium-specific. So I'd like to propose
> an idea. 
> so we change the layout a bit. It's still early in the game, we can do
> this. 
> A possible layout:
> bios/arch/i386/{poweron.s,protected_mode.s,memory.s,motherboard.c,boot.c}
> bios/arch/alpha/{poweron.s,protected_mode.s,memory.s,motherboard.c,boot.c}
> bios/payload/{your c files here}

That seems quite logical.
> If we do this, we're not just for pentiums any more. That would be
> wonderful! Any interest?

Seems quite a good idea.
Finding someone with an Alpha to test things on would be a first step
though I suppose. That, and finding someone who knows the Alpha architecture.

regards,

Aladdin V chipset

[Ronald G. Minnich]

On 29 Feb 2000, Dave Jones wrote:
> *shock*
> Really ? I thought Alphas only ran on their own custom buses (EV etc...)

shock here too, but alpha mboards I used were socket-7 based ...
> Finding someone with an Alpha to test things on would be a first step
> though I suppose. That, and finding someone who knows the Alpha architecture.

not a problem.

So do we go for it?

ron

-
To unsubscribe: send mail to [e-mail]
with 'unsubscribe openbios' in the body of the message

Aladdin V chipset

[Jeff Garzik]

"Ronald G. Minnich" wrote:
> On 29 Feb 2000, Dave Jones wrote:
> > *shock*
> > Really ? I thought Alphas only ran on their own custom buses (EV etc...)
>
> shock here too, but alpha mboards I used were socket-7 based ...
> > Finding someone with an Alpha to test things on would be a first step
> > though I suppose. That, and finding someone who knows the Alpha architecture.
>
> not a problem.
>
> So do we go for it?

follow the linux model:  grow when someone steps up with a patch...  I
don't think it makes much sense to do more than rudimentary work on
Alpha when (a) we don't appear to have any bigtime Alpha hackers and (b)
we don't even have X86 working yet....

Aladdin V chipset

[Ronald G. Minnich]

On Wed, 1 Mar 2000, Jeff Garzik wrote:
> "Ronald G. Minnich" wrote:
> > So do we go for it?
>
> follow the linux model:  grow when someone steps up with a patch...  I

yeah the only thing I worry about is this is a pretty complete respin of
the directory structure. But, we'll do it here and see what you all think.

ron

-
To unsubscribe: send mail to [e-mail]
with 'unsubscribe openbios' in the body of the message

Aladdin V chipset

Ronald G. Minnich  wrote:
> > > So do we go for it?
> >
> > follow the linux model:  grow when someone steps up with a patch...  I
>
> yeah the only thing I worry about is this is a pretty complete respin of
> the directory structure. But, we'll do it here and see what you all think.

There's nothing wrong with declaring an arch directory, containing only x86
for now. As Jeff said, if someone needs alpha/sparc support, they'll submit
patches.

Getting the initial directory layout right makes everyones life a lot easier
though, and saves a lot of moving stuff around later. (Which is a real nightmare
with CVS access).

Aladdin V chipset

[Edwin Rhodes]

yeah man, go for it.

edwin
----- Original Message -----
From: Ronald G. Minnich 
To: 
Sent: Wednesday, March 01, 2000 8:01 AM
Subject: Re: [OpenBIOS] Aladdin V chipset
> On 29 Feb 2000, Dave Jones wrote:
> > *shock*
> > Really ? I thought Alphas only ran on their own custom buses (EV etc...)
>
> shock here too, but alpha mboards I used were socket-7 based ...
> > Finding someone with an Alpha to test things on would be a first step
> > though I suppose. That, and finding someone who knows the Alpha

architecture.
> not a problem.
>
> So do we go for it?
>
> ron
>
> To unsubscribe: send mail to [e-mail]
> with 'unsubscribe openbios' in the body of the message

-
To unsubscribe: send mail to [e-mail]
with 'unsubscribe openbios' in the body of the message

Aladdin V chipset

[Stefan Reinauer]

-----BEGIN PGP SIGNED MESSAGE-----

On 01-Mar-00 Ronald G. Minnich wrote:
> On 29 Feb 2000, Dave Jones wrote:
> > *shock*
> > Really ? I thought Alphas only ran on their own custom buses (EV etc...)
>
> shock here too, but alpha mboards I used were socket-7 based ...

What CPUs did you use? All 21164 Alpha CPUs I saw were socket-499,
which looks a lot like socket-7, but drives the ev56 protocol. 
So technically seen, it shouldn't even be possible to get an 21164
alpha CPU physically onto a socket 7 board. I don't know about older ones,
but they are just not interesting for any market anymore.

On the other hand, 21264s are available as Slot A, like the Athlon processor.
This is possible, since slot A Athlons use the EV6 protocol they "borrowed"
from the AXP CPUs. Here it's technically possible just to use another
CPU architecture type on the same board just by flashing a new firmware.
The problem here: Unlike older Alpha cpus, the 21164 are much more
expensive than the motherboard (21164: cpu ~200$, motherboard ~700$;
21264: CPU ~2000$, motherboard ~700$)
> > Finding someone with an Alpha to test things on would be a first step
> > though I suppose. That, and finding someone who knows the Alpha
> > architecture.
>
> not a problem.

I would be happy to help with any upcoming questions, though I might not bring
too much active work into the project for another while.

Best regards,
        Stefan
-----BEGIN PGP SIGNATURE-----
Version: 2.6.3i
Charset: noconv

iQEVAwUBOQGuJyu+zKndEmP5AQHakQf/SQJfP+hLSSaUqHdcU1pwhqysK9zDS3lL
IT9vKu6qbburS42czGhZw+75K67zi7WnaTz3eyyZbmULyPY8BOx+sw2NR8iF8RCN
bNG+PvOLgCwGhZmOPibZzcyaLxLZtLT211zMeX66+BFkMB2VawN0ttcASzdeZSer
OsxzPGF5R+qxM210qUcvgcIbbPg8/PKca9BOFn8Uzy5GWq6zbvKSmzKrRbgY2xJv
E8i60Eup5tHoullizWNVjSbGkrB3S2q0BUIt7NLykNT1ZiOC9RCctC77NfALkqnp
6nw+dqjayC2rxfqX2hZWnRjzNfazaJ51bVH8b9ugCKrnZRN3ljSixw==
=LmMp
-----END PGP SIGNATURE-----

-
To unsubscribe: send mail to [e-mail]
with 'unsubscribe openbios' in the body of the message

Aladdin V chipset

[Edwin Rhodes]

this souds great!
edwin
----- Original Message -----
From: Ronald G. Minnich 
To: 
Sent: Tuesday, February 29, 2000 6:52 AM
Subject: Re: [OpenBIOS] Aladdin V chipset
> On Mon, 28 Feb 2000, Burl Nyswonger wrote:
> > > when you say super socket 7 chipsets -- does this include AMD?
> > > edwin
> >
> > I meant "Super Socket 7 boards" in general -- ie: the Asus P5 series,

Acer
> > boards (obviously), etc...  When I say "Aladdin V Chipset", I am

referring
> OK, this brings up an interesting issue for the OpenBIOS community. At
> least one Socket 7 motherboard will support either alpha or pentium in the
> socket. So it's the same motherboard with a different processor. All
> startup tasks for the motherboard are pretty much the same.
>
> I had an interesting set of discussions with vendors at extreme linux last
> week. There's a lot of interest in linuxbios -- from Alpha motherboards
> too (the alpha bios'es all have sucked very badly -- dec never got it
> right).
>
> As things stand openbios is too pentium-specific. So I'd like to propose
> an idea.
> get into protected mode early
> get rid of as much assembly as possible
> there's a lot that can move to C
>
> so we change the layout a bit. It's still early in the game, we can do
> this.
>
> A possible layout:
>
> bios/arch/i386/{poweron.s,protected_mode.s,memory.s,motherboard.c,boot.c}
> bios/arch/alpha/{poweron.s,protected_mode.s,memory.s,motherboard.c,boot.c}
> bios/payload/{your c files here}
>
> poweron.s is the very first set of things you do on poweron -- should be a
> few instructions. protected_mode.s bumps you into protected mode. On the
> pentium this is a few instructions. memory.s turns on memory -- this is
> *much* easier on the pentium in protected mode. memory.s should set up a
> stack and jump to C code. Motherboard.c does anything that has to be done
> to further use the motherboard. Boot.c gunzips the payload (either a
> bios-compatible code or Linux or FreeBSD or whatever) and jumps to it.
>
> I'm talking to a commercial company that is going to do something like
> this -- they don't want much assembly long-term, so they're recrafting
> openbios. I don't think most people want to do assembly.
>
> If we do this, we're not just for pentiums any more. That would be
> wonderful! Any interest?
>
> ron
>
> To unsubscribe: send mail to [e-mail]
> with 'unsubscribe openbios' in the body of the message

-
To unsubscribe: send mail to [e-mail]
with 'unsubscribe openbios' in the body of the message