[RFC PATCH v2 00/31] KVM: x86: Hyper-V VSM (Virtual Trust Levels) on VM planes
From: David Woodhouse
Date: Fri Nov 13 2026 - 14:14:02 EST
From: David Woodhouse <dwmw@xxxxxxxxxxxx>
This is v2 of the Hyper-V VSM support¹: Windows Virtualization Based
Security running natively on KVM planes, with the secure kernel in
VTL1 isolated from the normal kernel in VTL0, instead of using nested
virtualization.
This is still largely AI-written with minimal oversight so far; it's
presented to show the direction and for testing, not for detailed
review.
As before, it boots Windows Server 2016 through 2025 with Credential
Guard active, on both Intel and AMD hosts, with VTL1 protecting its
memory from VTL0, and survives hibernate/resume. It also runs VBS
enclaves: the full CreateEnclave / LoadEnclaveImage / InitializeEnclave
/ CallEnclave cycle behaves as it does on real Hyper-V. The userspace
side is still a separable, permissively-licensed library, pulled into
QEMU for testing.
Since v1:
• Rebased to v7.4-rc2. Paolo's memory protection attributes are now
merged in from kvm.git² rather than carried as patches.
• Three parts grew up and left home: the fixes to Jörg's series³,
the forwarding of unimplemented Hyper-V hypercalls to userspace⁴
and the generic plane userspace API⁷ are posted separately, and
this series is based on them.
• This is now the *minimum* VSM: what it takes to boot Windows with
Credential Guard and keep VTL1's memory protected. The parts it
can live without — KVM_TRANSLATE on plane fds, VTL-switch
statistics, the Hyper-V vCPU suspended state and the TLB flush
inhibit (KVM_HYPERV_SET_TLB_FLUSH_INHIBIT) — sit above it in git⁵
and will follow once this part makes sense to people.
• KVM_TRANSLATE2 is gone. The userspace library walks the guest
page tables itself for HvTranslateVirtualAddress, which it turned
out to need to do anyway.
• The hypercall exit gained a flags field (KVM_HYPERV_HCALL_UD) for
fault injection instead of stealing bit 63 of the result, so a
forwarded hypercall can return a full 64-bit result.
• A pile of fixes from review, each noted in its commit.
The base is Jörg's KVM Planes + SEV-SNP series⁶ with the fixes³ and
the plane userspace API⁷ applied; kvm-memprot-v4² and the hypercall
forwarding⁴ are merged on top. None of that is reposted here. The
whole assembly, and the userspace to drive it:
• https://git.infradead.org/?p=users/dwmw2/linux.git;a=shortlog;h=refs/heads/vsm2/vsm-min (tag vsm2-min1)
• https://git.infradead.org/?p=users/dwmw2/qemu.git;a=shortlog;h=refs/heads/vsm-v2
• https://git.infradead.org/?p=users/dwmw2/kvm-unit-tests.git;a=shortlog;h=refs/heads/vsm-v2
The 31 patches here are VSM itself: per-plane memory
attributes and overlays, the per-plane Hyper-V state (GUEST_OS_ID,
hypercall page, reference TSC page, SynIC, STIMER), VTL call/return
and the VTL-protection fault path, interrupt preemption and forced
VTL entry, and the register intercepts Windows actually uses.
Patch 28 (KVM: SVM: Keep TPR[3:0] across exits without a guest CR8
write) is a fix in its own right: the bug exists without planes, so
it should probably be applied standalone. I can send it separately
if wanted.
¹ https://lore.kernel.org/kvm/20261017122501.610248-1-dwmw2@xxxxxxxxxxxxx/
² git://git.kernel.org/pub/scm/virt/kvm/kvm.git kvm-memprot-v4
³ https://lore.kernel.org/kvm/20261111222911.477-1-dwmw2@xxxxxxxxxxxxx/
⁴ https://lore.kernel.org/kvm/20261111205836.4186584-1-dwmw2@xxxxxxxxxxxxx/
⁵ https://git.infradead.org/?p=users/dwmw2/linux.git;a=shortlog;h=refs/heads/vsm2/vsm-extras
⁶ https://lore.kernel.org/kvm/20260608144252.351443-1-joro@xxxxxxxxxx/
⁷ https://lore.kernel.org/kvm/20261113182607.82813-1-dwmw2@xxxxxxxxxxxxx/
David Woodhouse (30):
KVM: Implement per-plane memory attributes
KVM: x86: hyper-v: Bank GUEST_OS_ID and HYPERCALL MSRs per plane
KVM: x86: hyperv: Maintain a reference TSC page for each plane
KVM: x86: hyperv: Flush every plane twin on TLB flush hypercalls
KVM: x86: hyperv: Don't arm PV EOI for a higher plane's VP assist page
KVM: x86: hyper-v: Handle VTL call/return in the kernel
KVM: selftests: Cover single-step over a forwarded VTL call
KVM: x86: hyperv: Let higher-VTL interrupts preempt a lower VTL
KVM: x86: hyperv: Divert intercepted descriptor-table loads
KVM: x86: Keep APICv off on higher-plane vCPUs
KVM: x86: hyperv: Make KVM_REQ_HV_STIMER a per-plane request
KVM: x86: Kick lower twins when an event lands on a suspended plane
KVM: x86: hyperv: Suppress SynIC exits from higher planes
KVM: VMX: Report GVA and instruction bytes on plane-protection faults
KVM: x86: hyperv: Divert LLDT and LTR into register intercepts
KVM: Add per-plane memory overlays
KVM: x86: hyperv: Deliver higher-plane SynIC messages through overlays
KVM: x86: hyperv: Truncate higher-VTL vAPIC MSR writes
KVM: x86: Fold an NMI for a higher plane back to plane 0
KVM: x86: hyperv: Honour in_vtl in synthetic cluster IPIs
KVM: x86: Gate and re-arm forced VTL entry on interrupt deliverability
KVM: selftests: Test a plane suspending with an undelivered interrupt
KVM: selftests: Check that a masked higher-plane timer is no event
KVM: selftests: Check that a restored higher-plane interrupt forces an
exit
KVM: selftests: Check that a higher plane's NMI survives HLT and
suspension
KVM: selftests: Check that a VTL switch keeps bit 63 of RAX
KVM: SVM: Carry instruction state on reflected plane-protection faults
KVM: SVM: Keep TPR[3:0] across exits without a guest CR8 write
KVM: selftests: Add Hyper-V reference TSC page memory attribute test
KVM: Document the two scopes of KVM_REQUEST_PER_PLANE
Sriram Nambakam (1):
KVM: x86: hyperv: Do not divert descriptor-table writes in SMM
.../admin-guide/kernel-parameters.txt | 9 +
Documentation/virt/kvm/api.rst | 193 ++-
arch/x86/hyperv/hv_apic.c | 3 +-
arch/x86/include/asm/kvm-x86-ops.h | 1 +
arch/x86/include/asm/kvm_host.h | 95 +-
arch/x86/kvm/emulate.c | 21 +
arch/x86/kvm/hyperv.c | 1150 +++++++++++++++--
arch/x86/kvm/hyperv.h | 92 +-
arch/x86/kvm/kvm_emulate.h | 10 +
arch/x86/kvm/lapic.c | 80 +-
arch/x86/kvm/lapic.h | 3 +
arch/x86/kvm/mmu/mmu.c | 108 +-
arch/x86/kvm/mmu/paging_tmpl.h | 5 +-
arch/x86/kvm/mmu/tdp_mmu.c | 23 +-
arch/x86/kvm/msrs.c | 20 +-
arch/x86/kvm/svm/nested.c | 16 +
arch/x86/kvm/svm/sev.c | 34 +
arch/x86/kvm/svm/svm.c | 200 ++-
arch/x86/kvm/svm/svm.h | 13 +
arch/x86/kvm/vmx/main.c | 13 +
arch/x86/kvm/vmx/nested.c | 1 +
arch/x86/kvm/vmx/vmx.c | 131 +-
arch/x86/kvm/vmx/x86_ops.h | 1 +
arch/x86/kvm/x86.c | 546 +++++++-
arch/x86/kvm/x86.h | 3 +
include/hyperv/hvgdk_mini.h | 5 +-
include/hyperv/hvhdk_mini.h | 3 +-
include/linux/kvm_host.h | 144 +++
include/uapi/linux/kvm.h | 48 +-
tools/testing/selftests/kvm/Makefile.kvm | 6 +
.../testing/selftests/kvm/include/kvm_util.h | 1 +
.../selftests/kvm/include/x86/hyperv.h | 8 +
tools/testing/selftests/kvm/lib/kvm_util.c | 11 +
.../testing/selftests/kvm/memory_attributes.c | 6 +-
.../testing/selftests/kvm/x86/hyperv_clock.c | 7 -
.../selftests/kvm/x86/hyperv_forward_hcall.c | 92 +-
.../selftests/kvm/x86/hyperv_tsc_page_attrs.c | 131 ++
.../selftests/kvm/x86/planes_vtl_entry_test.c | 186 +++
.../kvm/x86/planes_vtl_nmi_halt_test.c | 305 +++++
.../kvm/x86/planes_vtl_restore_test.c | 121 ++
.../kvm/x86/planes_vtl_return_test.c | 204 +++
.../kvm/x86/planes_vtl_timer_mask_test.c | 177 +++
virt/kvm/kvm_main.c | 535 +++++++-
virt/kvm/pfncache.c | 3 +-
44 files changed, 4530 insertions(+), 234 deletions(-)
create mode 100644 tools/testing/selftests/kvm/x86/hyperv_tsc_page_attrs.c
create mode 100644 tools/testing/selftests/kvm/x86/planes_vtl_entry_test.c
create mode 100644 tools/testing/selftests/kvm/x86/planes_vtl_nmi_halt_test.c
create mode 100644 tools/testing/selftests/kvm/x86/planes_vtl_restore_test.c
create mode 100644 tools/testing/selftests/kvm/x86/planes_vtl_return_test.c
create mode 100644 tools/testing/selftests/kvm/x86/planes_vtl_timer_mask_test.c
--
2.50.1