cvs commit: src/sys/conf kern.post.mk kmod.mk

[Bruce Evans]

On Fri, 14 Nov 2003, Robert Watson wrote:

> On Fri, 14 Nov 2003, John Baldwin wrote:
>
> > On 14-Nov-2003 Brian Feldman wrote:
> > > green       2003/11/14 08:04:11 PST
> > >
> > >   FreeBSD src repository
> > >
> > >   Modified files:
> > >     sys/conf             kern.post.mk kmod.mk
> > >   Log:
> > >   Include opt_global.h in the modules build, when building from a normal
> > >   kernel build.  This makes it possible for me not to get pissed off that
> > >   random.ko crashes the system trying to rdtsc() when the i386/cpu.h
> > >   support code decides it's okay to call that op when neither I386_CPU or
> > >   I486_CPU is defined.  I guess it also makes WITNESS/INVARIANTS defines
> > >   get picked up by the modules.

Ugh.

> > WITNESS doesn't matter as all calls to it are done in the actual locking
> > code itself which does not lives in modules.  Modules always call the
> > locking functions and don't inline locking operations.

WITNESS actually has non-broken support for modules.  Modules cannot depend
on options, since they must work for all kernels.

Here is a fix to unbreak the rdtsc() call similarly.

%%%
Index: cpu.h
===================================================================
RCS file: /home/ncvs/src/sys/i386/include/cpu.h,v
retrieving revision 1.68
diff -u -2 -r1.68 cpu.h
--- cpu.h	16 Aug 2003 16:57:56 -0000	1.68
+++ cpu.h	14 Nov 2003 18:06:08 -0000
@@ -103,5 +104,5 @@
 get_cyclecount(void)
 {
-#if defined(I386_CPU) || defined(I486_CPU)
+#if defined(I386_CPU) || defined(I486_CPU) || defined(KLD_MODULE)
 	struct bintime bt;

%%%

All header files except the one with this bug seem to have had all the
necessary ifdefs on Ix86_CPU.

> I'd have picked up the syntax error in mac_test.c if this had been in
> place then :-).   And it makes mac_test and other regression test modules
> far more functional...

Is mac_test another broken module that depends on options? :-)

Bruce