* Ronald G Minnich <[EMAIL PROTECTED]> [020715 06:47]:
>
> OK, I am going to put in one day next week to see if I should either
> - get linuxbios CURRENT back to working on ds10
> - back-port etherboot to my ANCIENT working linuxbios on ds10
I'd definitely prefer the first one. How universal is the code that
was written back then btw? I saw a lot of stuff is hard coded for
the DS10/Tsunami stuff but it looks like with only a little bit more
effort we could get this working on more Alphas (especially the older
machines that _have_ to use milo now come to my mind. A boot solution
including LinuxBIOS looks a lot more sane to me than forward porting
milo to every little Kernel change.
There's no market potential in this, but at least keeping those machines
that are already out there in a sane state would be very nice and milo
is anything but sane.
> I could use some comments from those who have tried this recently: what
> exactly went wrong? Does it look huge to fix or not that big a deal?
I sent a patch to this list some time ago that made the stuff compile
again at least. One of the things to do is to move the LinuxBIOS table
stuff out of the i386 dependant code so it fits nicer into the scheme
when using it on other architectures.
The rest is more like glueing things together in the right order again
and getting rid of the Universal Boot(?) stuff
> I have 104 of these crummy boxes so I don't mind frying 10 or so to get
> this working, but I don't have a whole month to do it either. It's going
> to have to converge pretty quickly since I don't want to put too much
Are all machines capable of switching between 2 firmware images in rom?
how can I determine is this is possible on a certain box? Is the second
image visible when booting the first one? I.e. do I have to flash the
LinuxBIOS image to a certain position and take care not to overwrite
other areas in flash? I have 2 machines left, one of them has a flakey
motherboard or power supply, but i think it is good enough for
kill^H^H^H^Htesting linuxbios on it. Though i want to keep risk as low
as possible after I killed one machine a while ago but flashing the
kernel image with a position offset of 512bytes from what LinuxBIOS
expected (stupid me)
> effort into these machines -- I like 'em but compa-- er, HP doesn't --
> they killed the Alpha, after all.
Unlike a lot of modern CPUs Alphas are said to be designed without
heavy use of autorouting stuff, so it is rather expensive for them
to move to smaller production methods. This might be one of the reasons
why the Alpha does not really seem to be a loved child since a very long
time. It helps nothing though. The architecture is nice and keeping it
alive a bit longer is probably worth some effort. From what I heard
Alpha is going to be developed until somewhen after 2010, whereas ia32
is said to be dead in 2006 - not really a dead arch if you look at both
players.
Stefan
--
The x86 isn't all that complex - it just doesn't make a lot of
sense. -- Mike Johnson, Leader of 80x86 Design at AMD
Microprocessor Report (1994)
