Re: the creation of boot_cpu_init() is wrong and accessinguninitialised data

[James Bottomley]

On Tue, 2006-06-27 at 17:04 -0700, Andrew Morton wrote:
> +static void __init __attribute__((weak)) smp_setup_processor_id(void)

If you're really sure this works .. then it looks OK.  However, I
thought weak was a linker attribute, not a compiler one.  How does the
compiler know if the static inline needs to be incorporated or if a
strong symbol is going to override it when the whole thing is linked
together at the time it just compiles main.c?

> +{
> +}
> +
>  asmlinkage void __init start_kernel(void)
>  {
>  	char * command_line;
>  	extern struct kernel_param __start___param[], __stop___param[];
> +
> +	smp_setup_processor_id();
> +
>  /*
>   * Interrupts are still disabled. Do necessary setups, then
>   * enable them
> diff -puN arch/i386/mach-voyager/voyager_smp.c~add-smp_setup_processor_id arch/i386/mach-voyager/voyager_smp.c
> --- a/arch/i386/mach-voyager/voyager_smp.c~add-smp_setup_processor_id
> +++ a/arch/i386/mach-voyager/voyager_smp.c
> @@ -1938,3 +1938,9 @@ smp_cpus_done(unsigned int max_cpus)
>  {
>  	zap_low_mappings();
>  }
> +
> +void __init
> +smp_setup_processor_id(void)
> +{
> +	current_thread_info()->cpu = hard_smp_processor_id();
> +}
> diff -puN include/linux/smp.h~add-smp_setup_processor_id include/linux/smp.h
> --- a/include/linux/smp.h~add-smp_setup_processor_id
> +++ a/include/linux/smp.h
> @@ -125,4 +125,6 @@ static inline void smp_send_reschedule(i
>  #define put_cpu()		preempt_enable()
>  #define put_cpu_no_resched()	preempt_enable_no_resched()
>  
> +void smp_setup_processor_id(void);
> +
>  #endif /* __LINUX_SMP_H */
> _
> 
> 
> Note that I moved the call to smp_setup_processor_id() to be super-early. 
> Before the lock_kernel(), before everything.  It seems better that way. 
> And there's lockdep init stuff in -mm which is called immediately on entry
> to start_kernel() which might want to use smp_processor_id().
> 
> Is that all OK?

I'll give it a whirl tomorrow when I get access to one of the voyager
systems in Columbia.

James


-
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/