[benefits of swappable kernel code...]
> Some candidates I can think of for being swappable:
>
> - PnP code (especially if we get full blown ISA PnP)
> - legacy PCI cruft
> - fib (suggested by Alan Cox)
> - math emulation
> - bits of some drivers (have to be very careful here)
> - networking protocols with no active sockets
> - filesystems which aren't presently in use
>
> The latter three being swapable only at certain times (don't swap anything
> that might be required to swap back in), so I'm not sure if that would be
> feasible.
Hmmm. On my system, network protocols with no active sockets and unused
filesystems are already "swapped out" - or, at least, they do not eat my
memory as I compiled them as modules...
I was actually a little suprised that the math emulation code was unable
to me compiler as a module, for precisely memory use reasons. I would
have figured that making it into a module would let someone build a
nice, modular kernel, portable across whatever CPU setup you had,
irrespective of the presence of a math co-processor.
So, unless I have missed something *really* obvious here, the only real
advantage of being able to swap kernel code would come if Linux either
(a) swapped little bits of active code, such as a few pages of the IPX
code while other parts were in use, or (b) follows the NT model and
implements everything in kernel space, giving a 12MB kernel with 1.5MB
swapped in (as this box has right now...)
Of course, I make no claim to being more than a beginner when it comes
to kernel hacking, at least under Linux, so I have probably missed
something in my analysis. Oh - and I have no access to the kernel
source right now, so 'read the code' comments will help me not at all.
I hate hardware failure :(
Daniel Pittman
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html