Linux 2.5.23 cpu_online_map undeclared

I am running on a UP system, so I don't believe that cpu_online_map should be declared. Any suggestions?

gcc -Wp,-MD,./.sched.o.d -D__KERNEL__ -I/usr/src/linux-2.5.23/include
-Wall -Wstrict-prototypes -Wno-trigraphs -O2 -fomit-frame-pointer -fno-strict-aliasing -fno-common -pipe -mpreferred-stack-boundary=2 -march=i686 -nostdinc -iwithprefix include -fno-omit-frame-pointer -DKBUILD_BASENAME=sched -c -o sched.o sched.c
sched.c: In function `sys_sched_setaffinity':
sched.c:1332: `cpu_online_map' undeclared (first use in this function) sched.c:1332: (Each undeclared identifier is reported only once sched.c:1332: for each function it appears in.)
sched.c: In function `sys_sched_getaffinity':
sched.c:1391: `cpu_online_map' undeclared (first use in this function) make[1]: *** [sched.o] Error 1
make[1]: Leaving directory `/usr/src/linux-2.5.23/kernel'
make: *** [kernel] Error 2

-o) J o h n W e b e r
/\\ [email protected]
_\/v http://www.linuxhq.com/people/weber/

In article <[email protected]>,

Actually, it _should_ be declared, it's just that on UP it should be defined to the constant 1.

Somehow that #define got dropped by the hotplug-CPU stuff.

To fix, just add a

	#define cpu_online_map		1

to the non-SMP parts of include/linus/smp.h. The patch looks something like the appended (totally untested) thing.

Does that make UP happier?

Linus

@@ -86,6 +86,7 @@
 #define smp_call_function(func,info,retry,wait)	({ 0; })
 static inline void smp_send_reschedule(int cpu) { }
 static inline void smp_send_reschedule_all(void) { }
+#define cpu_online_map				1
 #define cpu_online(cpu)				1
 #define num_online_cpus()			1
 #define __per_cpu_data

yesterday i've tested a number of UP kernels that had an equivalent fix and this works as expected, nothing else broke.

Ingo

Actually, was deliberate. Not exposing the mask when almost everyone just wants to know "are any of these cpus online" was a preemptive strike against NR_CPUS > BITS_PER_LONG (I wouldn't have bothered but I was so close anyway).

My mistake was not noticing that x86 SMP was still exposing it and that scheduler changes since my original patch were using it. Fixed with cpu_online_mask(res, mask) in prior patch.

Apologies, and hope that clarifies,
Rusty.

   there are those who do and those who hang on and you don't see too
   many doers quoting their contemporaries.  -- Larry McVoy