On Thu, 14 Dec 2000, Levi Khatskevitch wrote:
> 1) Linuxbios code is aimed to be the simplest possible boot loader with the
> ablity to load and start a linux kernel from the same NVRAM/flash chip.
> Anything that is not required *before* loading a kernel should go into the
> kernel itself.
yes. There are lots of reasons for this, and it conforms to current trends
in both FreeBSD, Linux, and other systems that are pulling ACPI out of the
BIOS and into kernel services.
> 2) After the kernel has been started linuxbios should not provide any services
> to it nor should it remain in the memory.
yes. One reason it should not provide any services is that OSes have
changing requirements for an execution environment for modules they
call. I don't want to try to have linuxbios correctly provide services to
the OS.
> 3) All additinal chipset support (like fans, clock speed etc.) that is
> either not requitred on startup or can be used by running kernel should be
> built into the kernel the usual way (drivers, modules, patches). Even when this
> means dublicating some code from linuxbios boot loader. This is acceptible
> since while linuxbios will the code in it's simplest possile way linux
> kernel should implement it in the most convenient way.
yes, and the reason is that this stuff is coming anyway. FreeBSD and Linux
will have the capability to monitor all this in-kernel. More importantly,
as problems are found and resolved the FreeBSD and Linux support for these
operations will continue to improve. LinuxBIOS should not try to do work
that the kernels will do better.
Also, keep in mind, as the kernels evolve, they will take on more and more
BIOS capability due to things such as "hot plug".
I hope this is OK.
ron
