Robert Boehm wrote:

> Wow...that is VERY interesting...as I wouldn't think that the BIOS
> would have
> anything to do with it....so I'll put that one in the memory banks and
> back
> pocket.  Interestingly, I just heard when I was at the computer shop
> that a
> person just updated his BIOS and it messed up all of his Linux
> stuff...to the
> point that it wouldn't even boot anymore...so the BIOS can be all
> important,
> I guess...and some of the VERY LATEST bios's don't play nice with the
> Linux
> Kernel...especially some of the newest ACPI schemes, I'm finding out
> (unrelated
> to this thread, I know...but you brought up an interesting problem).

I would have thought that the BIOS would be out-of-the-picture as soon
as the kernel started.  But, here's what went through my head when I
decided to believe it:

   1. The BIOS initializes some of the hardware before Linux gets to
      play with it; if it's initialized in a subtly wrong way[0], that
      could cause the behavior I'm seeing.  Also, since this is a mostly
      jumperless board and since it has both "modern" and "legacy" modes
      for IDE, SATA, and other devices, this could be a real issue.
   2. When I was reading about how the PCI PnP stuff works, the
      documentation I read talked about asking the PCI BIOS for
      information about the cards.  The way it was written implied that
      on a modern system, the BIOS does a lot more after the system is
      booted than the old BIOSes did on my trusty old AMD 386DX/40...

In any case, VMWare works my machine pretty hard.  I think I'm going to
go buy a 1GB DIMM on my way home from work.  :-)

Also, for the record, the BIOS that I installed was version 1011.  The
bios that shipped with the motherboard (both in the flash and on the
driver-CD) was version 1007.  I'll also re-iterate that my motherboard
is an Asus P4P800S-X -- so that the search-engines can find this message
if someone else has the same problem!

-Luke

[0] I'm thinking about a flag on the that gets referenced only in rare
error-cases.  For instance, what does the controller do if both hda and
hdb try to transmit a sector at the same time?  Or data-errors on the
IDE channel?  If overlapping I/O is used (I haven't checked to see if
this is supported), there are probably dozens of weird errors that would
happen once per hour or even a couple of times a day...

-- 
Luke Scharf, Systems Administrator
Virginia Tech Aerospace and Ocean Engineering