[coreboot] Error executing VGA ROM with coreboot

[Peter Stuge]

Mansoor wrote:
> I have attached the YABEL output.

Thanks. I think it is helpful. Please follow Kevin's suggestions to
find out more about what the ROM is trying to do.


> Even though coreboot does only the hardware initialization, it has
> to do a VGA rom init for payloads other than seabios?

VGA is often not very important for payloads, and option ROMs in
general fundamentally require a BIOS, which coreboot does not, and
shall not, offer. If a BIOS environment is needed, then SeaBIOS is
the way to go.


> I my case I was planning to use FILO as payload to boot Linux. FILO
> doesn't do any VGA init.

In theory it can, if libpayload has native support for your graphics
hardware. Since this isn't the case, there is no init done. There
exists only very few graphics drivers in libpayload, which would
support the idea that graphics is not very important for payloads.


I think there are a few options that you can explore further. First,
maybe the graphics driver in Linux has the ability to initialize
hardware completely? Because Linux will be started pretty quickly
with coreboot+FILO, the delay from power on until graphics enabled
will not be very long even if it is postponed to Linux.

Second, you can use both SeaBIOS and FILO together. SeaBIOS supports
booting using another payload file than itself in CBFS. Normally this
is used as a last resort when there is no local storage or option ROM
to boot from, but that behavior should be very easy to change in
SeaBIOS, so that FILO is always used for booting.

Using multiple payloads will increase boot time, but if legacy VGA
ROM initialization is a hard requirement then I'm afraid there is no
substitute for a BIOS environment, such as provided by SeaBIOS.

But, don't rule out the Linux graphics driver alternative. There
exists at least one driver (for AMD Geode) which does this, one or
several of the VIA hardware drivers may also be able to do it. Linux
2.6.31 got kernel modesetting support in the Intel framebuffer driver
so there is a chance. Also, the Intel graphics drivers developers are
very active in the Linux community, and maybe they can quickly add
any features that are discovered to be missing at the moment.


//Peter