Re: [tlug] pcg-c1mrx / pcg-c1mv

[bhoward@example.com]

Christopher> PCI BIOS is not assigning interrupts to its devices -- it
Christopher> expects the OS to do so.

that explains the "no irq assigned" message but a search of google
shows many other people with that same error starting up fine.  the
fact that i can get it to boot with pci=conf2 suggests to me that
conf1 mode is probing some device it a way that locks up the system or
maybe there is some kind of fight between devices sharing an irq?

another suspicious clue is that the ide driver is setting the bus
speed for PIO modes to 50mhz for the successfully booting pci=conf2
mode and to 30mhz for the not-successful conf1 mode.

Christopher> Laptops are notorious for this behavior, with Sony being
Christopher> the worst of the lot.

Christopher> Linux's PCI init code seems to have varying levels of
Christopher> success with PCI fixups.  Compile a kernel with every PCI
Christopher> option available and see what happens.

guess it is time to spend some more time in the kernel sources.

>> As an aside, I tried OpenBSD 3.0 to see if it had the same problem
>> -- it did seem to have better luck recognizing devices but I
>> haven't

Christopher> Ah!  You're not averse to *BSD!

heh.  it's all unix of one sort or another to me.  i started my life
in the unix world with system iii, research unix and 3bsd and still
have trouble accepting that bsd is now the "conservative" end of the
unix spectrum ;-) i find the great battles over the relatively subtle
differences between modern bsd variants pretty funny when i think back
to the situation in my university days.

Christopher> This works for NetBSD and possibly OpenBSD -- compile a
Christopher> kernel with:

i'll try what you suggest with netbsd/openbsd.  if it works, maybe can
backport whatever it is doing to linux.

cheers, chris.

						bruce