Re: [PATCH 1/1] x86/irq: Add hardcoded hypervisor interrupts to /proc/stat

[Sean Christopherson]

On Wed, Mar 22, 2023, Michael Kelley (LINUX) wrote:
> From: Dave Hansen <dave.hansen@xxxxxxxxx> Sent: Wednesday, March 22, 2023 11:07 AM
> > 
> > On 2/27/23 10:46, Michael Kelley wrote:
> > > diff --git a/arch/x86/kernel/irq.c b/arch/x86/kernel/irq.c
> > > index 766ffe3..9f668d2 100644
> > > --- a/arch/x86/kernel/irq.c
> > > +++ b/arch/x86/kernel/irq.c
> > > @@ -211,6 +211,13 @@ u64 arch_irq_stat_cpu(unsigned int cpu)
> > >  #ifdef CONFIG_X86_MCE_THRESHOLD
> > >  	sum += irq_stats(cpu)->irq_threshold_count;
> > >  #endif
> > > +#ifdef CONFIG_X86_HV_CALLBACK_VECTOR
> > > +	sum += irq_stats(cpu)->irq_hv_callback_count;
> > > +#endif
> > > +#if IS_ENABLED(CONFIG_HYPERV)
> > > +	sum += irq_stats(cpu)->irq_hv_reenlightenment_count;
> > > +	sum += irq_stats(cpu)->hyperv_stimer0_count;
> > > +#endif
> > >  #ifdef CONFIG_X86_MCE
> > >  	sum += per_cpu(mce_exception_count, cpu);
> > >  	sum += per_cpu(mce_poll_count, cpu);
> > 
> > This seems fine, especially since arch_show_interrupts() has them.  But,
> > what's with the "#if IS_ENABLED" versus the plain #ifdef?  Is there some
> > difference I'm missing?  Why not just be consistent with the other code
> > and use a plain #ifdef for both?
> 
> I'm following the coding pattern in arch_show_interrupts(), in irq_cpustat_t,
> and most other places that test CONFIG_HYPERV.   Maybe all those existing
> cases are a mis-application of Documentation/process/coding-style.rst
> Section 21, which prefers "if (IS_ENABLED(CONFIG_HYPERV))" over
> "#ifdef CONFIG_HYPERV".  "#if IS_ENABLED()" is not the same as
> "if (IS_ENABLED())".  :-)
> 
>  Net, I don't have a strong preference either way.

Using IS_ENABLED() is mandatory because CONFIG_HYPERV is a tri-state, i.e. can
be a module and thus #define CONFIG_HYPER_MODULE instead of CONFIG_HYPERV.