Re: 2.6.16-rc3-mm1

[Brice Goglin <Brice.Goglin@ens-lyon.org>]

Con Kolivas wrote:

>On Wednesday 15 February 2006 07:28, Andrew Morton wrote:
>  
>
>>Brice Goglin <Brice.Goglin@ens-lyon.org> wrote:
>>    
>>
>>>WARNING: speedstep-centrino.ko needs unknown symbol cpu_online_map
>>>
>>>This symbol is in include/linux/cpumask.h but actually only defined and
>>>exported in smpboot.c which is not compiled on UP.
>>>      
>>>
>>diff -puN
>>arch/i386/kernel/cpu/cpufreq/speedstep-centrino.c~git-acpi-up-fix-2
>>arch/i386/kernel/cpu/cpufreq/speedstep-centrino.c ---
>>devel/arch/i386/kernel/cpu/cpufreq/speedstep-centrino.c~git-acpi-up-fix-2	2
>>006-02-14 12:27:41.000000000 -0800 +++
>>devel-akpm/arch/i386/kernel/cpu/cpufreq/speedstep-centrino.c	2006-02-14
>>12:27:41.000000000 -0800 @@ -654,8 +654,10 @@ static int centrino_target
>>(struct cpufr
>> 		return -EINVAL;
>> 	}
>>
>>+#ifdef CONFIG_SMP
>> 	/* cpufreq holds the hotplug lock, so we are safe from here on */
>> 	cpus_and(online_policy_cpus, cpu_online_map, policy->cpus);
>>+#endif
>>    
>>
>
>Shouldn't the cpu_online_map be hardcoded to the first/only cpu on UP instead?
>
>Cheers,
>Con
>  
>
It's actually defined on UP in kernel/sched.c:

#ifndef CONFIG_SMP
cpumask_t cpu_online_map __read_mostly = CPU_MASK_ALL;
cpumask_t cpu_possible_map __read_mostly = CPU_MASK_ALL;
#endif

What about adding an EXPORT_SYMBOL_GPL there ?

Brice