[PATCH] Re: BUG: 2.6.6-rc3 on SMP/SPARC64 (Sun E3000)?

From: Patrick Finnegan
Date: Tue May 11 2004 - 20:07:27 EST


On Friday 07 May 2004 10:49, you wrote:
> I've been trying to get 2.6.6-rc3 to work in SMP mode on my E3000
> without much success yet. It boots fine with a uniprocessor kernel,
> but trying to enable SMP gives me this as the last few lines of the
> kernel messages (booted with -p early printk option):

It appears that the problem is that on Sparc64, smp_processor_id() gives
the hardware ID, not a logical number (ie 0..n for the first n+1
processors). I had NR_CPUS set to 8, and the first (boot) CPU was
numbered by the hardware to be CPU 10.

The patch catches if the boot CPU is greater than NR_CPUS, since that is
possible with Sparc64. I see that the condition is explicitly checked
for in smp_tick_init(), but that must not get called soon enough to
catch the problem (which manifested in sched_init()s call to
wake_up_forked_process() ), so I moved it to smp_prepare_boot_cpu().

Comments?

Pat
--
Purdue University ITAP/RCS --- http://www.itap.purdue.edu/rcs/
The Computer Refuge --- http://computer-refuge.org

--- linux-2.6.6.orig/arch/sparc64/kernel/smp.c 2004-05-09 21:31:55.000000000 -0500
+++ linux-2.6.6/arch/sparc64/kernel/smp.c 2004-05-11 19:46:15.692007000 -0500
@@ -1108,11 +1108,6 @@
boot_cpu_id = hard_smp_processor_id();
current_tick_offset = timer_tick_offset;

- if (boot_cpu_id >= NR_CPUS) {
- prom_printf("Serious problem, boot cpu id >= NR_CPUS\n");
- prom_halt();
- }
-
cpu_set(boot_cpu_id, cpu_online_map);
prof_counter(boot_cpu_id) = prof_multiplier(boot_cpu_id) = 1;
}
@@ -1254,6 +1249,11 @@

void __devinit smp_prepare_boot_cpu(void)
{
+ if (hard_smp_processor_id() >= NR_CPUS) {
+ prom_printf("Serious problem, boot cpu id >= NR_CPUS\n");
+ prom_halt();
+ }
+
current_thread_info()->cpu = hard_smp_processor_id();
cpu_set(smp_processor_id(), cpu_online_map);
cpu_set(smp_processor_id(), phys_cpu_present_map);
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/