Re: [PATCH 10/15] static_call: Add basic static call infrastructure

[Nadav Amit <namit@vmware.com>]

> On Jun 7, 2019, at 1:49 AM, Ard Biesheuvel <ard.biesheuvel@linaro.org> wrote:
> 
> On Fri, 7 Jun 2019 at 10:29, Peter Zijlstra <peterz@infradead.org> wrote:
>> On Thu, Jun 06, 2019 at 10:44:23PM +0000, Nadav Amit wrote:
>>>> + * Usage example:
>>>> + *
>>>> + *   # Start with the following functions (with identical prototypes):
>>>> + *   int func_a(int arg1, int arg2);
>>>> + *   int func_b(int arg1, int arg2);
>>>> + *
>>>> + *   # Define a 'my_key' reference, associated with func_a() by default
>>>> + *   DEFINE_STATIC_CALL(my_key, func_a);
>>>> + *
>>>> + *   # Call func_a()
>>>> + *   static_call(my_key, arg1, arg2);
>>>> + *
>>>> + *   # Update 'my_key' to point to func_b()
>>>> + *   static_call_update(my_key, func_b);
>>>> + *
>>>> + *   # Call func_b()
>>>> + *   static_call(my_key, arg1, arg2);
>>> 
>>> I think that this calling interface is not very intuitive.
>> 
>> Yeah, it is somewhat unfortunate..
> 
> Another thing I brought up at the time is that it would be useful to
> have the ability to 'reset' a static call to its default target. E.g.,
> for crypto modules that implement an accelerated version of a library
> interface, removing the module should revert those call sites back to
> the original target, without putting a disproportionate burden on the
> module itself to implement the logic to support this.
> 
> 
>>> I understand that
>>> the macros/objtool cannot allow the calling interface to be completely
>>> transparent (as compiler plugin could). But, can the macros be used to
>>> paste the key with the “static_call”? I think that having something like:
>>> 
>>>  static_call__func(arg1, arg2)
>>> 
>>> Is more readable than
>>> 
>>>  static_call(func, arg1, arg2)
>> 
>> Doesn't really make it much better for me; I think I'd prefer to switch
>> to the GCC plugin scheme over this.  ISTR there being some propotypes
>> there, but I couldn't quickly locate them.
> 
> I implemented the GCC plugin here
> 
> https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgit.kernel.org%2Fpub%2Fscm%2Flinux%2Fkernel%2Fgit%2Fardb%2Flinux.git%2Flog%2F%3Fh%3Dstatic-calls&amp;data=02%7C01%7Cnamit%40vmware.com%7Cd31c4713640c44a651bf08d6eb250faa%7Cb39138ca3cee4b4aa4d6cd83d9dd62f0%7C0%7C0%7C636954941771964758&amp;sdata=h7RtT33E9FMapLZbAu9aTfjREP5kXrM0o2QQ1WpbDCM%3D&amp;reserved=0
> 
> but IIRC, all it does is annotate call sites exactly how objtool does it.

I did not see your version before I made mine for a similar (but slightly
different) purpose:

https://lore.kernel.org/lkml/20181231072112.21051-4-namit@vmware.com/

My version, I think, is more generic (I don’t think yours consider calls
that have a return value). Anyhow, I am sure you know more about GCC plugins
than I do.

I do have a version that can take annotations to say which call should be
static and to get the symbol it uses.

I also think that this implementation would disallow keys that reside within
structs. This would mean that paravirt, for instance, would need to go
through many changes to use this infrastructure.