On Thu, Jan 01, 2004 at 03:02:41PM -0500, Bill Moran wrote: > Matthew Seaman wrote:
> >It's generally best either to kldload all of the netgraph modules you > >require for this purpose or to compile them all into the kernel. One > >way, or the other, but not a mixture of both. > I would go beyond "generally best" and straight into "it doesn't work". > My question is, "Is this fact documented somewhere, and I didn't see it?" > Because, if it's not, I'm going to put together a doc patch and file a PR. Hmmm... I've seen people mention similar effects where various other modules insisting on kldload'ing their dependencies even if they're already in the kernel. They pop up on various mailing lists over time. You may be right in saying "it doesn't work" -- but I have no idea whether that is correct in all cases. I get the feeling that most people will go to great pains to prevent their kernel code doing annoying things like that. Of course, if you compile, say, netgraph into your kernel, your buildkernel run shouldn't produce a separate netgraph.ko module. If there's one in /modules or wherever the equivalent is for 5.x then it's a hangover from a previous kernel. Chances are that if you get rid of the old netgraph.ko then things may well work as you expect. Cheers, Matthew -- Dr Matthew J Seaman MA, D.Phil. 26 The Paddocks Savill Way PGP: http://www.infracaninophile.co.uk/pgpkey Marlow Tel: +44 1628 476614 Bucks., SL7 1TH UK
pgp00000.pgp
Description: PGP signature