Re: [PATCH 12/16] KVM: TDX: Inhibit APICv for TDX guest

[Binbin Wu <binbin.wu@linux.intel.com>]

On 1/7/2025 11:24 AM, Chao Gao wrote:
>> Note, kvm_use_posted_timer_interrupt() uses a heuristic of HLT/MWAIT interception
>> being disabled to detect that it's worth trying to post a timer interrupt, but off
>> the top of my head I'm pretty sure that's unnecessary and pointless.  The
> Commit 1714a4eb6fb0 gives an explanation:
>
>    if only some guests isolated and others not, they would not
>    have any benefit from posted timer interrupts, and at the same time lose
>    VMX preemption timer fast paths because kvm_can_post_timer_interrupt()
>    returns true and therefore forces kvm_can_use_hv_timer() to false.
Userspace uses KVM_CAP_X86_DISABLE_EXITS to enable mwait_in_guest or/and
hlt_in_guest for non TDX guest. The code doesn't work for TDX guests.
- Whether mwait in guest is allowed for TDX depends on the cpuid
   configuration in TD_PARAMS, not the capability of physical cpu checked by
   kvm_can_mwait_in_guest().
- hlt for TDX is via TDVMCALL, i.e. hlt_in_guest should always be false
   for TDX guests.

For TDX guests, not sure if it worths to fix kvm_{mwait,hlt}_in_guest()
or add special handling (i.e., skip checking mwait/hlt in guest) because
VMX preemption timer can't be used anyway, in order to allow housekeeping
CPU to post timer interrupt for TDX vCPUs when nohz_full is configured
after changing APICv active to true.


>
>> "vcpu->mode == IN_GUEST_MODE" is super cheap, and I can't think of any harm in
>> posting the time interrupt if the target vCPU happens to be in guest mode even
>> if the host wasn't configured just so.
>>
>>> Another scenario I was thinking of was hrtimer expiry during vcpu exit
>>> being handled in KVM/userspace, which will cause timer interrupt
>>> injection after the next exit [1] delaying timer interrupt to guest.
>> No, the IRQ won't be delayed.  Expiration from the hrtimer callback will set
>> KVM_REQ_UNBLOCK, which will prevent actually entering the guest (see the call
>> to kvm_request_pending() in kvm_vcpu_exit_request()).
>>
>>> This scenario is not specific to TDX VMs though.
>>>
>>> [1] https://git.kernel.org/pub/scm/virt/kvm/kvm.git/tree/arch/x86/kvm/x86.c?h=kvm-coco-queue#n11263