Re: fixing 2.6.10 UP builds

[Jesse Barnes <jbarnes@engr.sgi.com>]

On Wednesday, January 12, 2005 4:02 pm, Luck, Tony wrote:
> >I've regenerated this patch, original developed by Jesse
> >Barnes, against 2.6.10, which fixes UP builds.  I've also removed the
> > bogus header removals from the cyclone driver.
> >
> >I won't pretend to be able to explain the details, hopefully
> >Jesse recalls :)
>
> I think that I pulled together all the pieces, but I still see
> a couple of problems when trying to build a generic-UP kernel:
>
>   CC      drivers/char/mmtimer.o
> drivers/char/mmtimer.c: In function `mmtimer_tasklet':
> drivers/char/mmtimer.c:502: structure has no member named `write_lock'
> make[1]: *** [drivers/char/mmtimer.o] Error 1
> make: *** [drivers/char/mmtimer.o] Error 2

Looks like we need rwlock_is_write_locked for this one.  Christoph?

>   CC      arch/ia64/sn/kernel/sn2/sn_hwperf.o
> In file included from arch/ia64/sn/kernel/sn2/sn_hwperf.c:33:
> include/linux/nodemask.h: In function `__first_unset_node':
> include/linux/nodemask.h:244: warning: passing arg 1 of
> `__find_next_zero_bit' discards qualifiers from pointer target type

This warning was introduced upstream with the nodemask stuff.  It should be 
fixed but is unrelated to the UP stuff.

> arch/ia64/sn/kernel/sn2/sn_hwperf.c: In function `sn_hwperf_op_cpu':
> arch/ia64/sn/kernel/sn2/sn_hwperf.c:360: warning: implicit declaration of
> function `smp_call_function_single'

Looks like we need a !CONFIG_SMP version of smp_call_function_single.  
include/linux/smp.h has a non-smp version of smp_call_function that just 
returns 0, should smp_call_function_single do the same thing?

Jesse