[RFC PATCH 01/16] ARM: local timers: early device probing hooks

[Marc Zyngier]

On 17/06/11 15:56, Arnd Bergmann wrote:
> On Thursday 16 June 2011, Marc Zyngier wrote:
>> @@ -147,6 +148,19 @@ static int __init timer_init_syscore_ops(void)
>>  
>>  device_initcall(timer_init_syscore_ops);
>>  
>> +void (* __initdata arm_late_time_init)(void);
>> +
>> +static void __init __arm_late_time_init(void)
>> +{
>> +       if (arm_late_time_init)
>> +               arm_late_time_init();
>> +
>> +#ifdef CONFIG_LOCAL_TIMER_DEVICES
>> +       early_platform_driver_register_all("localtimer");
>> +       early_platform_driver_probe("localtimer", 1, 0);
>> +#endif
>> +}
>> +
>>  void __init time_init(void)
>>  {
>>         system_timer = machine_desc->timer;
> 
> Instead of having two separate global pointers, should we make the ARM specific
> callback a machine_desc callback?

Probably not. I only see this arm_late_time_init pointer as a temporary
measure. The only user (shmobile) uses this hook to install its own
early timers, and there is a patch floating around to move that code to
time.c. Unless we have other uses for such a callback, of course.

	M.
-- 
Jazz is not dead. It just smells funny...