Re: GENERIC and INSTALL kernels

[Alan Barrett <apb%cequrux.com@localhost>]

On Sat, 12 Feb 2011, Izumi Tsutsui wrote:
> The only problem on current module(7) is there is no way to specify
> alternative module path when new module files have problem.

It's not only when new module files have problems; it's also when you
want to test a new kernel, or test new modules.

My biggest problem with modules as currently implemented in NetBSD, is
that there is not way to specify a different module path for a different
kernel.  This makes safe updates effectively impossible.  (A requirement
for a safe update with a modular kernel is that it must be possible to
boot with a new kernel and new modules, test, and then go back to the
old kernel and old modules if desired.)

> Anyway, module(7) is required for users who won't compile their
> kernels,

Yes.  It's a pity that there's no safe way to use modules in the general
case.

By the way, I ue a kernel based on i386 MONOLITHIC, plus "options
modular".  The only module that I actually use is a ramdisk, and the
boot loader is capable of mixing different kernels with different
ramdisks, even though the ekrnel itself is not capable of using a
non-default module path.

--apb (Alan Barrett)