Re: Patches for REALLY TINY 386 kernels

[Adrian Bunk <bunk@stusta.de>]

On Tue, Jul 24, 2007 at 01:50:35PM -0700, Yinghai Lu wrote:
> On 7/24/07, Helge Hafting <helge.hafting@aitel.hist.no> wrote:
>> Andi Kleen wrote:
>> >> Some people are putting Linux kernels in the "BIOS" (i.e. ROM chip) 
>> when
>> >> using LinuxBIOS (www.linuxbios.org). It _does_ make a lot of difference
>> >> there how big the kernel is. At the moment you can't do that with
>> >> anything smaller than a 1 MB chip. But if people could use 512 KB chips
>> >> because the kernel is small enough that would sure be a great thing.
>> >>
>> >
>> > I'm sure it would be possibel to save a lot of text size. But I don't
>> > think removing the relatively small CPUID code is the right way.
>> > That is just a big maintenance issue for little gain.
>> >
>> Well - anyone compiling linux for BIOS usage is targetting
>> a single machine.  So an ability to target a single machine is useful,
>> i.e. run the CPUID at compile-time, put the answer in a constant/macro,
>> let the optimizer prune the alternatives. :-)
>
> we are using AMD64 + LinuxBIOS + Kernel (without acpi) + kexec to load
> final kernel.
> So we can use drivers in kernel for any media (SCSI, SATA, IB,...),
> not like EFI need every driver re-porting. and We could use KVM in
> kernel to load other OS if needed.
>
> The problem is Kernel is getting bigger and bigger. and old Tiny
> kernel is stopping at 2.6.18...
>...

Please send:
- the .config for the last kernel small enough
- your size limit
- your gcc version
and I'll look at this.

> YH

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed