> >Maybe it would be a good idea to wrap both the declaration and the definition
> >inside ifdef __SMP__.
>
> The following patch should go into 2.1.33 (I lost the e-mail address of the
> original poster, but Ingo Molnar and I agree on this one):
>
> [patch skipped]
Yes, that works nicely. Along the same lines, here's a patch for
arch/i386/kernel/irq.c which seems to do no harm.
Thanks
--------------------------------------
--- linux/arch/i386/kernel/irq.c.orig Sat Apr 5 18:59:09 1997
+++ linux/arch/i386/kernel/irq.c Sun Apr 6 00:36:43 1997
@@ -40,7 +40,9 @@
static unsigned char cache_21 = 0xff;
static unsigned char cache_A1 = 0xff;
+#ifdef __SMP__
unsigned int local_irq_count[NR_CPUS];
+#endif
#ifdef __SMP_PROF__
static unsigned int int_count[NR_CPUS][NR_IRQS] = {{0},};