Re: [PATCH v13 06/14] x86/mm: use broadcast TLB flushing for page reclaim TLB flushing

From: Tom Lendacky
Date: Fri Feb 28 2025 - 15:43:12 EST


On 2/28/25 09:02, Tom Lendacky wrote:
> On 2/27/25 19:13, Rik van Riel wrote:
>> On Wed, 2025-02-26 at 12:12 -0600, Tom Lendacky wrote:
>>>
>>> As long as you keep the ASID value in EDX[15:0] as 0, then you won't
>>> #GP. ASID 0 is the host/hypervisor. An ASID > 0 belongs to a guest.
>>>
>> I've been spending some time reading the KVM code,
>> and I don't think invlpgb would be currently useful
>> with KVM.
>>
>> From reading pre_svm_run(), new_asid(), and svm_vcpu_run(),
>> it looks like the ASID number used might be different for
>> each VCPU, assigned on a per (physical host) CPU basis.
>>
>> It would take some surgery to change that around.
>>
>> Some googling around also suggests that the ASID address
>> space is even more limited than the PCID address space :(
>
> Right, to support using INVLPGB in guests you need a global ASID, which is
> an ASID that doesn't change over the VMs lifetime and is used on all
> vCPUs. Global ASIDs are only available and used today with SEV guests. At
> that point you would not intercept the instruction and, based on APM vol
> 3, the ASID value is replaced with the guest ASID value.
>
> "A guest that executes a legal INVLPGB that is not intercepted will have
> the requested ASID field replaced by the current ASID and the valid ASID
> bit set before doing the broadcast invalidation."
>
> So I'm in the process of verifying that issuing INVLPLG in a guest with
> the ASID valid bit set and an ASID value of 0 (EDX[15:0]) won't #GP, but
> will just replace the specified ASID value with the guest ASID value in
> hardware.

I verified that when (a non-intercepted) INVLPGB is issued in a guest,
hardware will set the ASID valid bit and use the guest ASID value
(regardless of the value specified in EDX[15:0]) before doing the
broadcast invalidation.

So the implementation of setting the ASID-valid bit and specifying ASID 0
is not incompatible in the guest.

Thanks,
Tom

>
> For non-SEV guests, INVLPGB would need to be intercepted and somehow
> emulated or just not advertised to the guest so that the IPI path is used.
>
> Thanks,
> Tom
>
>>