[PATCH 2/2] KVM: riscv: selftests: Detect supported vm modes

[Andrew Jones]

On Mon, Oct 20, 2025 at 09:32:57PM +0200, Radim Krčmář wrote:
> 2025-10-20T21:26:18+08:00, Wu Fei <atwufei at 163.com>:
> > On 10/18/25 00:04, Andrew Jones wrote:
> >>> -               guest_mode_append(VM_MODE_P56V57_4K, sz >= 59);
> >>> -               guest_mode_append(VM_MODE_P50V48_4K, sz >= 50);
> >>> -               guest_mode_append(VM_MODE_P41V39_4K, sz >= 41);
> >>> +               guest_mode_append(VM_MODE_P56V57_4K, satp_mode >= SATP_MODE_57 && sz >= 59);
> >>> +               guest_mode_append(VM_MODE_P50V48_4K, satp_mode >= SATP_MODE_48 && sz >= 50);
> >>> +               guest_mode_append(VM_MODE_P41V39_4K, satp_mode >= SATP_MODE_39 && sz >= 41);
> >> 
> >> Something along these lines was what I was suggesting in the last patch,
> >> but where we select from many more combinations of P and V for modes.
> >>
> >> VM_MODE_P41V39_4K
> >> VM_MODE_P41V48_4K
> >> VM_MODE_P41V57_4K
> >> VM_MODE_P50V48_4K
> >> VM_MODE_P50V57_4K
> >> VM_MODE_P59V57_4K
> >
> > Do we need more? e.g. for sv39, there could be P50V39 and P56V39.
> 
> Bare is probably the most important first stage mode, as it's the only
> one that might fully test the two extra bits of Sv57x4.

Indeed, but for that we need a mode named something like VM_MODE_BARE59.

> 
> I think a single combination per "x4" mode might be sufficient, since
> all non-bare first stages generate 56-bit physical addresses anyway, but
> it makes sense to have all of them if we're doing more than one.

Yeah, might as well.

> 
> >> 
> >>> +
> >>> +               // set the first (maximum) supported mode as default
> >> 
> >> Please, no C++ comments, and don't we want a V39 mode to be the default?
> >> In fact, VM_MODE_P41V39_4K should always be supported, so we can just set
> >> VM_MODE_DEFAULT to that.
> >
> > It's likely the maximum supported mode is the most common configuration 
> > on production?
> 
> Yes, KVM currently picks the widest mode, although I think the optimal
> choice of hgatp is the smallest one that fully covers the whole guest
> physical address space.
> 
> (e.g. if all guest memory is mapped below 2 TiB, KVM should pick Sv39x4
>  for its hgatp, regardless of what is the widest supported mode.)

Yes, and for the tests, I think the guest config can just default to the
minimal. Tests that need particular modes can select them when supported.
We ensure tests run the same code on all hosts this way when the mode
selected isn't what's under test.

Thanks,
drew