In message <20000803091540.A20044@noc.untraceable.net>, Andrew Brown writes:
>>Any objections to this approach?
>
>only that i don't think it will work.

OK, I don't think that's an objection ;-)

The first goal of this is to make sure the problem doesn't get
any worse. The secondary goal is to fix it.

>or that it will be very difficult at the very least.  at one i needed
>to include sys/lockf.h in a userland program.  that got me this
>warning:

Sigh. Perhaps we need to apply use of

#if defined(LKM) && !defined(_KERNEL)

and discourage LKM's from _KERNEL-ing? Ugh, this does get painful.

There are a lot of cases where this can be fixed without such pain,
though, and the arla case is one such.

--jhawk