[PATCH v2 04/13] uprobes: allow arch-specific initialization

[Oleg Nesterov]

On 10/22, David Long wrote:
>
> On 10/19/13 12:42, Oleg Nesterov wrote:
>> On 10/15, David Long wrote:
>>>
>>> Add a weak function for any architecture-specific initialization.  ARM
>>> will use this to register the handlers for the undefined instructions it
>>> uses to implement uprobes.
>>
>> Could you explain why ARM can't simply do the necessary initialization in
>> arch/arm/kernel/uprobes-arm.c ?
>>
>>
>>> +int __weak __init arch_uprobes_init(void)
>>> +{
>>> +	return 0;
>>> +}
>>> +
>>>   static int __init init_uprobes(void)
>>>   {
>>> +	int ret;
>>>   	int i;
>>>
>>>   	for (i = 0; i < UPROBES_HASH_SZ; i++)
>>> @@ -1870,6 +1876,10 @@ static int __init init_uprobes(void)
>>>   	if (percpu_init_rwsem(&dup_mmap_sem))
>>>   		return -ENOMEM;
>>>
>>> +	ret = arch_uprobes_init();
>>> +	if (ret)
>>> +		return ret;
>>> +
>>>   	return register_die_notifier(&uprobe_exception_nb);
>>>   }
>>>   module_init(init_uprobes);
>>
>> IOW, why do we need to call arch_uprobes_init() from init_uprobes().
>>
>> Oleg
>>
>
> I don't know how you would do the initialization without invoking it
> through the module_init function, which I think you can only have one
> of.  Could you explain in more detail what you had in mind?

I simply do not understand why uprobes.c uses module_init/module_exit,
it can't be compiled as a module.

I think that module_exit/exit_uprobes should be killed, and module_init()
should be turned into __initcall(). uprobes-arm.c can have another one.

Oleg.