Re: [RFC PATCH 1/2] bulk: Replace !CONFIG_SOFTMMU -> CONFIG_USER_ONLY

[Philippe Mathieu-Daudé]

On 3/6/23 05:49, Richard Henderson wrote:
> On 6/2/23 15:58, Philippe Mathieu-Daudé wrote:
> > CONFIG_USER_ONLY is the opposite of CONFIG_SOFTMMU.
> > Replace !CONFIG_SOFTMMU negation by the positive form
> > which is clearer when reviewing code.
>
> CONFIG_SOFTMMU should be reserved for the actual softmmu tlb, which we 
> *should* be able to enable for user-only.  It is the only way to handle 
> some of our host/guest page size problems.  Further, CONFIG_SOFTMMU 
> should go away as a #define and become a runtime test (forced to true 
> for system mode).  Pie in the sky stuff.

This would be:

  bool has_softmmu(void)
  {
  #ifdef CONFIG_USER_ONLY
      /* TODO: implement */
      return false;
  #else
      return true;
  #endif
  }

?

> It is quite likely that all uses of CONFIG_SOFTMMU outside of tcg/, 
> accel/tcg/, and random bits of include/ should only be using 
> CONFIG_USER_ONLY.

I see.

Thanks,

Phil.