[lkml]   [1998]   [Jun]   [28]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRFD: Do we want modules loading modules?

I was wondering how people felt on this issue, i.e. is it something
we want to encourage or not? It might be beneficial to make a general
decision on this issue before the decision is made implicitly by having
several drivers making use of this trick.

Case in point (feel free to skip reading this):

Personally I don't care how the above decision goes. However,
I recently adapted all the 8390 based network drivers to try and
autoload 8390.o if kmod was enabled, in an effort to make things
more user-friendly for the novice doing an install. As an example
consider a recent & typical post (regarding wd module) that I just saw:

Message-ID: <>
I'm a total newbie, so when a post replies "insmod [yada yada yada]", it
doesn't help me coz I don't know insmod and when I tried it, I get errors
like "unresolved symbols". And lilo commands are the same thing. And the

With my autoload-8390 support (actually credit where credit is due, it was
Donald's idea -- the basic plan is in 2.0.34 ne2k-pci.c) the user wouldn't
have a problem. The kmod-aware wd.o module checks if 8390 support is
in-kernel or already loaded, and if not, it checks if the kernel has kmod
support and if so, tries to load the 8390.o module. If kmod support isn't
available, they still don't get an "unresolved symbol" error, as the wd.o
module is able to tell them that kmod isn't available, and hence it can't
load the 8390 module.

I felt this was a slight step forward in making linux more luser-friendly,
and furthermore, if you didn't enable kmod (or use a modular driver for
that matter) it didn't cost you any extra code at all.

However, I recently got a rather hostile e-mail from Yggdrasil calling
this change "inconsistent", "wasted cycles" (it doesn't touch the
driver core?!?) and that it "breaks any code that would try to read
modules.dep" and that "We will not ship your driver with that kmod
code enabled".

As I said, I don't really care either way, but given the above negative
feedback, I was interested to know what others thought of the idea of
modules loading modules.

[ I was intending to #if 0 out the #define LOAD_8390_BY_KMOD from 8390.h
anyways until the module locking issues were resolved and until the
get_module_symbol() function handled versioned module symbols correctly ]


To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to

 \ /
  Last update: 2005-03-22 13:43    [W:0.033 / U:4.716 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site