[RFC] Kernel and initramfs plans

Alexander E. Patrakov patrakov at ums.usu.ru
Sun Mar 4 21:20:27 PST 2007


Jeremy Huntwork wrote:
> Alexander E. Patrakov wrote:
>> 2) Follow the other distributions and load the needed drivers (and only 
>> them) from initramfs with the help of udev. This requires a full redesign of 
>> the initramfs and dropping the C-based /init binary.
>>
> 
> I prefer this option, mostly because it reduces the size of the kernel 
> (I assume anything that can be modular will be...) by a great deal. And 
> it does seem to be the standard approach - there's something to be said 
> for that. Still, why does this require dropping the C init?
> 
> I'm not really objecting to doing that (at least for the x86 cd - if I 
> ever do get around to building a sparc cd I may have to continue using 
> the C-based init), I just want to understand fully why it is required to 
> use a shell-based init.

After going to a fully modular kernel, it has to do complex parsing of the 
command line (such as "load=qlogicfas blacklist=ne2k-pci"), which is better 
done in a language specially suited to that. Besides, some users want to 
know how it works, and this option is completely withdrawn from them with 
binary init. Third, everyone uses shell-based initramfs.

-- 
Alexander E. Patrakov



More information about the livecd mailing list