[PATCH 27/34] KVM: selftests: Introduce VM "shape" to allow tests to specify the VM type

Fuad Tabba tabba at google.com
Mon Nov 6 08:17:09 PST 2023


On Mon, Nov 6, 2023 at 4:04 PM Sean Christopherson <seanjc at google.com> wrote:
>
> On Mon, Nov 06, 2023, Fuad Tabba wrote:
> > On Sun, Nov 5, 2023 at 4:34 PM Paolo Bonzini <pbonzini at redhat.com> wrote:
> > >
> > > From: Sean Christopherson <seanjc at google.com>
> > >
> > > Add a "vm_shape" structure to encapsulate the selftests-defined "mode",
> > > along with the KVM-defined "type" for use when creating a new VM.  "mode"
> > > tracks physical and virtual address properties, as well as the preferred
> > > backing memory type, while "type" corresponds to the VM type.
> > >
> > > Taking the VM type will allow adding tests for KVM_CREATE_GUEST_MEMFD,
> > > a.k.a. guest private memory, without needing an entirely separate set of
> > > helpers.  Guest private memory is effectively usable only by confidential
> > > VM types, and it's expected that x86 will double down and require unique
> > > VM types for TDX and SNP guests.
> > >
> > > Signed-off-by: Sean Christopherson <seanjc at google.com>
> > > Message-Id: <20231027182217.3615211-30-seanjc at google.com>
> > > Signed-off-by: Paolo Bonzini <pbonzini at redhat.com>
> > > ---
> >
> > nit: as in a prior selftest commit messages, references in the commit
> > message to guest _private_ memory. Should these be changed to just
> > guest memory?
>
> Hmm, no, "private" is mostly appropriate here.  At this point in time, only x86
> supports KVM_CREATE_GUEST_MEMFD, and x86 only supports it for private memory.
> And the purpose of letting x86 selftests specify KVM_X86_SW_PROTECTED_VM, i.e.
> the reason this patch exists, is purely to get private memory.
>
> Maybe tweak the second paragraph to this?
>
> Taking the VM type will allow adding tests for KVM_CREATE_GUEST_MEMFD
> without needing an entirely separate set of helpers.  At this time,
> guest_memfd is effectively usable only by confidential VM types in the
> form of guest private memory, and it's expected that x86 will double down
> and require unique VM types for TDX and SNP guests.

sgtm
/fuad



More information about the linux-riscv mailing list