PCI BIOS Extension ROM in v3

[Myles Watson]

Coreboot v3 refuses to initialize the VGA in QEMU because the PCI device
ID's are mismatched.

It turns out that the header has been overwritten by a copyright string, so
that v3 is following a bad pointer into the ROM looking for the PCI device
IDs and not finding it.  Coreboot v2 still successfully initializes the VGA
because it skips the vendor check.

The question is if we should make the vendor check a warning instead of
failing.  I realize that the right thing to do is fix the ROM, but there may
be other broken ROMs as well.

Another thing we could do is only fail on a mismatched ID if we find a
correct PCIR signature on the data section.

Myles

[Carl-Daniel Hailfinger]

On 22.01.2008 18:59, Myles Watson wrote:
> Coreboot v3 refuses to initialize the VGA in QEMU because the PCI device
> ID's are mismatched.
>
> It turns out that the header has been overwritten by a copyright string, so
> that v3 is following a bad pointer into the ROM looking for the PCI device
> IDs and not finding it.  Coreboot v2 still successfully initializes the VGA
> because it skips the vendor check.
>
> The question is if we should make the vendor check a warning instead of
> failing.  I realize that the right thing to do is fix the ROM, but there may
> be other broken ROMs as well.

I'd say we refuse to execute the ROM in v3. If the ROM is broken, we
have no way to determine how broken it is. Unconditional execution may
prevent booting the machine, thereby forcing an out-of-system flash of
coreboot with ROM execution disabled.

How many ROMs of real hardware are broken that way? If their number is
close to zero, printing a scary warning message will be best. If there
is a large number of such broken ROMs, we can make execution dependent
on some NVRAM setting, which will refuse execution after NVRAM clear.

Do you know which versions of Qemu ship the broken ROM? Is there already
a patch for it?

Regards,
Carl-Daniel

[ron minnich]

On Jan 22, 2008 10:20 AM, Carl-Daniel Hailfinger
 wrote:
> I'd say we refuse to execute the ROM in v3. If the ROM is broken, we
> have no way to determine how broken it is. Unconditional execution may
> prevent booting the machine, thereby forcing an out-of-system flash of
> coreboot with ROM execution disabled.

I agree. I have yet to see a real ROM broken in this way.

Print the warning message and refuse to run it. And get QEMU to fix
both their broken VGA rom and their broken BIOS :-)

ron

[Myles Watson]

> -----Original Message-----
> From: ron minnich [mailto:[e-mail]]
>
> On Jan 22, 2008 10:20 AM, Carl-Daniel Hailfinger
>  wrote:
> > I'd say we refuse to execute the ROM in v3. If the ROM is broken, we
> > have no way to determine how broken it is. Unconditional execution may
> > prevent booting the machine, thereby forcing an out-of-system flash of
> > coreboot with ROM execution disabled.
>
> I agree. I have yet to see a real ROM broken in this way.
>
> Print the warning message and refuse to run it. And get QEMU to fix
> both their broken VGA rom and their broken BIOS :-)

The VGA BIOS is now fixed in CVS if you download the latest.  I don't know
when QEMU will pick it up.

Myles