Re: [PATCH 5/8] x86/msr: Switch rdmsr_safe_on_cpu() to return a 64-bit quantity

[Ingo Molnar]

* Ingo Molnar <mingo@xxxxxxxxxx> wrote:

> 
> * Ingo Molnar <mingo@xxxxxxxxxx> wrote:
> 
> > 
> > * Juergen Gross <jgross@xxxxxxxx> wrote:
> > 
> > > In order to prepare retiring rdmsrq_safe_on_cpu() switch
> > > rdmsr_safe_on_cpu() to have the same interface as rdmsrq_safe_on_cpu().
> > 
> > Patch #1.
> 
> Note that to do this, a new name should be introduced,
> such as rdmsrq_safe_on_cpu() or so. This principle applies
> to the other patches as well.
> 
> Ie. since we are changing the function signature anyway,
> do not keep 32-bit API names, and introduce the new ones
> cleanly in 3 patches:
> 
>   - introduce rdmsrq_safe_on_cpu()
>   - change users from rdmsr_safe_on_cpu() => rdmsrq_safe_on_cpu()
>   - remove obsolete rdmsr_safe_on_cpu()
> 
> This will make it easier to bisect any breakages, and will
> make it easier to transition the APIs.

Addendum #2:

So instead of doing:

  x86/msr: Switch rdmsr_safe_on_cpu() to return a 64-bit quantity
  x86/msr: Switch all callers of rdmsrq_safe_on_cpu() to use rdmsr_safe_on_cpu()

We want something like:

  x86/msr: Switch all callers of rdmsr_safe_on_cpu() to use rdmsrq_safe_on_cpu()
  x86/msr: Remove unused rdmsr_safe_on_cpu() API

Ie. instead of having this confusing mixture of 32-bit and 64-bit
API names and transition, just transition to the 64-bit API cleanly,
and remove the 32-bit API in a separate patch once it's unused.

Thanks,

	Ingo