[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: RFC: live-initramfs 2.x features

Marco Amadori wrote:
>> it needs to be rewritten in a more modern, modular way.
> Not also changing language from shell to python/perl/C, right ? :-)

i guess you didn't mean it dead serious, but since otavio brought it up
yesterday too..

stuff in the initrd needs to be best written in a language that can be
executed within the capabilities of busybox itself. currently, that's
only shell and c. while bringing in another interpreter is possible, it
would blow up the initrd dramatically at not much gain.

in a first implementation, everything should be shell only, later when
things work, we could think about writing performance critical things in
c to make things faster.

however, compared to the current implementation, we can already make it
about twice as fast by just using proper shell consistently. that should
for a first rewrite be enough :)

> I'll prioritize that, for the sake of doing right the first time, or otherwise 
> nobody will enable this after the codebase grown a bit far.

ack. it should be done right after the booting of the 'essential' medias
(all block devices) is done.

> I see that the once dreamed feature of really fast boot times using "some 
> default common" resume image are now abandoned-ware ;-)

there was that thing called 'mach boot', where the was (tried) to make
it a big secret about and that afaik used some sort of lvm snapshot to
read in, compared with some ram dumps. i think, it's not worth all the
time to kill and hassle needed for that, and relay on such cases on
simple resuming capabilities of the kernel.


Address:        Daniel Baumann, Burgunderstrasse 3, CH-4562 Biberist
Email:          daniel.baumann@panthera-systems.net
Internet:       http://people.panthera-systems.net/~daniel-baumann/

Reply to: