Re: [PATCH v10 00/28] KVM: Enable FRED support with KVM VMX

[Shivansh Dhiman <shivansh.dhiman@amd.com>]

On 24-09-26 19:43, David Woodhouse wrote:
> On Thu, 2026-09-24 at 16:44 +0530, Shivansh Dhiman wrote:
>> On 12-09-26 03:06, Sohil Mehta wrote:
>>> Opens
>>> -----
>>
>>> AMD compatibility: The SVM enabling [5] builds on the same common
>>> plumbing but I haven't had a chance to look at the series yet. Shivansh,
>>> please let me know if there are any incompatibilities. I assume most of
>>> the opens apply to SVM as well so resolutions would help both vendors.
>>
>> Thanks for the series Sohil. I've been able to run SVM side of vFRED with
>> changes rebased on top of this series. So, no incompatibilities in common
>> plumbing as of now. Will let you know if anything comes up.
> 
> Can I remove the "Some AMD CPUs (Venice) incorrectly..." part of the
> commit message of
> https://git.infradead.org/?p=users/dwmw2/linux.git;a=commitdiff;h=fred


Hi David,

Yes, you may go ahead and drop that sentence. AMD Zen 6 (Venice) supports
event type 5 for INT1, and the FRED Virtualization spec [1] has been
updated to reflect this.

I also ran your int1_fred_test selftest on top of my SVM FRED series [2],
and all three tests pass. Note that the series stops intercepting ICEBP
for FRED-enabled guests, which is what lets the guest see event type 5.

$ sudo ./int1_fred_test
Random seed: 0x55d6fcb8
Testing FRED event types with EPT fault on stack
  ICEBP (EVENT_TYPE_PRIV_SWEXC): PASSED
  INT3 (EVENT_TYPE_SWEXC): PASSED
  INT $0x20 (EVENT_TYPE_SWINT): PASSED


One small nit: could the message above say "EPT/NPT" instead of "EPT",
since the test also runs on AMD?

Thanks,
Shivansh

[1]: https://docs.amd.com/v/u/en-US/69191-PUB
[2]: https://lore.kernel.org/kvm/20260925114017.2208531-1-shivansh.dhiman@amd.com/