insmod uses syscall

[Rob Landley]

On Friday 27 January 2006 22:26, Erik Andersen wrote:
> On Thu Jan 26, 2006 at 04:37:03PM -0700, Chris Kottaridis wrote:
> > However, I am still curious about why busybox is not calling
> > init_module directly, but rather calling it through syscall.
>
> Because there is no guarantee that libc happens to have an
> init_module() function to call and thus people kept complaining
> that insmod/modprobe didn't work with their C library which had
> been built vs 2.4 kernel headers but though they later decided
> they wanted to run a 2.6 kernel without rebuilding their C lib.
> Thus instead of getting a perfectly sane ENOSYS they instead got
> an undefined symbol, and complained, and complained, till I did
> something about it, not because it was a busybox problem, but
> because I didn't want to hear any more people moaning about
> busybox being broken when in fact it wasn't....

Between platform.h and the dummy.c checks, I suspect we'll grow infrastructure 
to be able to handle this sort of thing...  not exactly gracefully, but a 
little more flexibly at some point.

But not this week...

>  -Erik

Rob
-- 
Steve Ballmer: Innovation!  Inigo Montoya: You keep using that word.
I do not think it means what you think it means.