Re: [PATCH] KVM: TDX: Fix APIC MSR ranges in tdx_has_emulated_msr()

[Binbin Wu]

On 3/19/2026 3:42 AM, Dave Hansen wrote:
> On 3/18/26 12:01, Dmytro Maluka wrote:
>> +		case X2APIC_MSR(APIC_ISR) ... X2APIC_MSR(APIC_ISR) + APIC_ISR_NR - 1:
>> +		case X2APIC_MSR(APIC_TMR) ... X2APIC_MSR(APIC_TMR) + APIC_ISR_NR - 1:
>> +		case X2APIC_MSR(APIC_IRR) ... X2APIC_MSR(APIC_IRR) + APIC_ISR_NR - 1:
> 
> Thanks for the patch, Dmytro.
> 
> <sigh>
> 
> So this code never worked (at least for a big chunk of the ranges.
> Isaku, could you please go try to figure out if there are tests for this
> somewhere, and why this never bit us?

The bug doesn't cause problems for TDs because:
- These x2apic MSRs (TASKPRI, PROCPRI, EOI, ISRx, TMRx, IRRx) are virtualized by CPU,
  when a TD accesses these MSRs, it doesn't cause #VE, thus no TDVMCALL from the TD to
  request the emulation of these MSRs.
- The bug make the "false" range of APIC MSRs smaller, so it doesn't impact the result
  for the rest of the APIC MSRs.

The bug could be triggered if a TD issues a TDVMCALL directly to request the
read/write operations for these x2apic MSRs, but a sane TD will not do it. 

Currently, we don't have dedicated KVM selftests code to call TDVMCALL directly to request
the emulation for these x2apic MSRs.

> 
> It might also be handy to have a:
> 
> #define X2APIC_LAST_MSR(r)	(X2APIC_MSR(x)+APIC_ISR_NR-1)
> 
> so that the resulting code is a bit more readable:
> 
> 	case X2APIC_MSR(APIC_IRR) ... X2APIC_LAST_MSR(APIC_IRR):
> 
> Dmytro, if you feel a burning need to respin this, don't let me stop
> you. I can probably just fix this up when it gets applied, or Isaku can
> make those changes and resend it too.
>