* [PATCH v3 0/2] KVM: arm64: Force HCR_EL2.NV1 to RES0 unconditionally
@ 2026-11-18 20:36 Colton Lewis
2026-11-18 20:36 ` [PATCH v3 1/2] " Colton Lewis
2026-11-18 20:36 ` [PATCH v3 2/2] KVM: arm64: selftests: Verify HCR_EL2.NV1 is RES0 when E2H0 is NI_NV1 Colton Lewis
0 siblings, 2 replies; 4+ messages in thread
From: Colton Lewis @ 2026-11-18 20:36 UTC (permalink / raw)
To: Marc Zyngier, Oliver Upton
Cc: Leonardo Bras, Joey Gouly, Suzuki K Poulose, Zenghui Yu,
Fuad Tabba, kvmarm, linux-arm-kernel, kvm, linux-kernel,
Colton Lewis
When commit b2a324ff01fe ("KVM: arm64: Use HCR_EL2 feature map to drive
fixed-value bits") migrated HCR_EL2 fixed-bit masking to hcr_feat_map[],
feat_nv2_e2h0_ni() was written as !kvm_has_feat(kvm, FEAT_E2H0). Because
ID_AA64MMFR4_EL1.E2H0 is a signed 4-bit field where IMP is 0, NI is -1,
and NI_NV1 is -2, !kvm_has_feat(kvm, FEAT_E2H0) evaluates to true when
E2H0 < 0 (including NI_NV1), leaving HCR_EL2.NV1 writable instead of
RES0 for VHE-only nested guests where KVM advertises E2H0 = NI_NV1.
KVM never supports a configuration where HCR_EL2.NV1 is non-RES0
(limit_nv_id_reg() exposes either VHE-only EL2 with E2H0 = NI_NV1 where
E2H0 is not userspace-writable, or FEAT_E2H0 without FEAT_VHE or
FEAT_NV).
Patch 1 drops feat_nv2_e2h0_ni() and marks HCR_EL2.NV1 unconditionally
RES0 using NEEDS_FEAT_FLAG(HCR_EL2_NV1, FORCE_RESx).
Patch 2 extends the arm64 hello_el2 selftest to verify via
KVM_SET_ONE_REG and KVM_GET_ONE_REG that HCR_EL2.E2H remains RES1 and
HCR_EL2.NV1 remains RES0 when E2H0 is NI_NV1.
Changes in v3:
- Drop feat_nv2_nv1_not_res0() and force HCR_EL2.NV1 to RES0
unconditionally via NEEDS_FEAT_FLAG(HCR_EL2_NV1, FORCE_RESx) (Marc).
- Add Suggested-by: Marc Zyngier <maz@kernel.org> on patch 1.
- Add a comment in hello_el2.c explaining why RESx enforcement is
checked from userspace rather than inside the guest (Oliver).
- Link to v2: https://lore.kernel.org/r/20261117204535.4036191-1-coltonlewis@google.com
Tested under QEMU 11.1.2 TCG (arm64, kvm-arm.mode=nested): hello_el2
fails with "HCR_EL2.NV1 should be RES0" on unpatched v7.4-rc3 and passes
with this series applied, with no check_feature_map() warnings at boot.
Changes in v2:
- Rename feat_nv2_e2h0_ni() to feat_nv2_nv1_not_res0() (Leonardo).
- Drop the guest EL2 MSR/MRS HCR_EL2 check in hello_el2 (Marc).
- Link to v1: https://lore.kernel.org/r/20261116200418.3535619-1-coltonlewis@google.com
Colton Lewis (2):
KVM: arm64: Force HCR_EL2.NV1 to RES0 unconditionally
KVM: arm64: selftests: Verify HCR_EL2.NV1 is RES0 when E2H0 is NI_NV1
arch/arm64/kvm/config.c | 7 +------
tools/testing/selftests/kvm/arm64/hello_el2.c | 15 +++++++++++++++
2 files changed, 16 insertions(+), 6 deletions(-)
base-commit: aebb7274476403040d582c61478df8f4a7bf9cd3
--
2.56.1.750.gf70dc682cf-goog
^ permalink raw reply [flat|nested] 4+ messages in thread* [PATCH v3 1/2] KVM: arm64: Force HCR_EL2.NV1 to RES0 unconditionally 2026-11-18 20:36 [PATCH v3 0/2] KVM: arm64: Force HCR_EL2.NV1 to RES0 unconditionally Colton Lewis @ 2026-11-18 20:36 ` Colton Lewis 2026-11-19 12:09 ` Leonardo Bras 2026-11-18 20:36 ` [PATCH v3 2/2] KVM: arm64: selftests: Verify HCR_EL2.NV1 is RES0 when E2H0 is NI_NV1 Colton Lewis 1 sibling, 1 reply; 4+ messages in thread From: Colton Lewis @ 2026-11-18 20:36 UTC (permalink / raw) To: Marc Zyngier, Oliver Upton Cc: Leonardo Bras, Joey Gouly, Suzuki K Poulose, Zenghui Yu, Fuad Tabba, kvmarm, linux-arm-kernel, kvm, linux-kernel, Colton Lewis ID_AA64MMFR4_EL1.E2H0 is a signed 4-bit field where: - 0b0000 (IMP, 0): FEAT_E2H0 is implemented. - 0b1111 (NI, -1): FEAT_E2H0 is not implemented, except HCR_EL2.NV1 is implemented. - 0b1110 (NI_NV1, -2): Neither HCR_EL2.E2H == 0 nor HCR_EL2.NV1 is implemented, and HCR_EL2.NV1 is RES0. When commit b2a324ff01fe ("KVM: arm64: Use HCR_EL2 feature map to drive fixed-value bits") converted HCR_EL2 to hcr_feat_map[], feat_nv2_e2h0_ni() was written as !kvm_has_feat(kvm, FEAT_E2H0), where FEAT_E2H0 checks E2H0 >= IMP (0). This inverted the condition: feat_nv2_e2h0_ni() evaluates to true when E2H0 < 0 (including NI_NV1), leaving HCR_EL2.NV1 writable instead of RES0 for VHE-only nested guests where KVM advertises ID_AA64MMFR4_EL1.E2H0 = NI_NV1. In practice, KVM cannot handle HCR_EL2.NV1 being non-RES0 in any configuration: limit_nv_id_reg() exposes either VHE-only EL2 with ID_AA64MMFR4_EL1.E2H0 = NI_NV1 (and E2H0 is not userspace-writable), or FEAT_E2H0 without FEAT_VHE or FEAT_NV. Checking E2H0 in a feature helper is therefore unnecessary. Drop feat_nv2_e2h0_ni() and mark HCR_EL2_NV1 as an unconditional RES0 bit using NEEDS_FEAT_FLAG(HCR_EL2_NV1, FORCE_RESx) in hcr_feat_map[] (shared by HCR_EL2 and NVHCR_EL2). Because HCR_EL2_NV1 is not part of the architectural HCR_EL2_RES0 mask, check_feat_map() accounts for it as an implementation-defined FORCE_RESx bit. Fixes: b2a324ff01fe ("KVM: arm64: Use HCR_EL2 feature map to drive fixed-value bits") Suggested-by: Marc Zyngier <maz@kernel.org> Signed-off-by: Colton Lewis <coltonlewis@google.com> --- arch/arm64/kvm/config.c | 7 +------ 1 file changed, 1 insertion(+), 6 deletions(-) diff --git a/arch/arm64/kvm/config.c b/arch/arm64/kvm/config.c index adaf26ab93b85..2e81f45dc9921 100644 --- a/arch/arm64/kvm/config.c +++ b/arch/arm64/kvm/config.c @@ -256,11 +256,6 @@ static bool feat_nv2(struct kvm *kvm) kvm_has_feat(kvm, ID_AA64MMFR2_EL1, NV, NV2)); } -static bool feat_nv2_e2h0_ni(struct kvm *kvm) -{ - return feat_nv2(kvm) && !kvm_has_feat(kvm, FEAT_E2H0); -} - static bool feat_rasv1p1(struct kvm *kvm) { return (kvm_has_feat(kvm, ID_AA64PFR0_EL1, RAS, V1P1) || @@ -1034,7 +1029,7 @@ static const struct reg_bits_to_feat_map hcr_feat_map[] = { HCR_EL2_NV2 | HCR_EL2_NV, feat_nv2), - NEEDS_FEAT(HCR_EL2_NV1, feat_nv2_e2h0_ni), /* Missing from JSON */ + NEEDS_FEAT_FLAG(HCR_EL2_NV1, FORCE_RESx), /* Missing from JSON, RES0 as an IMPDEF choice */ NEEDS_FEAT(HCR_EL2_API | HCR_EL2_APK, feat_pauth), -- 2.56.1.750.gf70dc682cf-goog ^ permalink raw reply related [flat|nested] 4+ messages in thread
* Re: [PATCH v3 1/2] KVM: arm64: Force HCR_EL2.NV1 to RES0 unconditionally 2026-11-18 20:36 ` [PATCH v3 1/2] " Colton Lewis @ 2026-11-19 12:09 ` Leonardo Bras 0 siblings, 0 replies; 4+ messages in thread From: Leonardo Bras @ 2026-11-19 12:09 UTC (permalink / raw) To: Colton Lewis Cc: Leonardo Bras, Marc Zyngier, Oliver Upton, Joey Gouly, Suzuki K Poulose, Zenghui Yu, Fuad Tabba, kvmarm, linux-arm-kernel, kvm, linux-kernel On Wed, Nov 18, 2026 at 08:36:24PM +0000, Colton Lewis wrote: > ID_AA64MMFR4_EL1.E2H0 is a signed 4-bit field where: > - 0b0000 (IMP, 0): FEAT_E2H0 is implemented. > - 0b1111 (NI, -1): FEAT_E2H0 is not implemented, except HCR_EL2.NV1 > is implemented. > - 0b1110 (NI_NV1, -2): Neither HCR_EL2.E2H == 0 nor HCR_EL2.NV1 is > implemented, and HCR_EL2.NV1 is RES0. > > When commit b2a324ff01fe ("KVM: arm64: Use HCR_EL2 feature map to drive > fixed-value bits") converted HCR_EL2 to hcr_feat_map[], feat_nv2_e2h0_ni() > was written as !kvm_has_feat(kvm, FEAT_E2H0), where FEAT_E2H0 checks > E2H0 >= IMP (0). This inverted the condition: feat_nv2_e2h0_ni() evaluates > to true when E2H0 < 0 (including NI_NV1), leaving HCR_EL2.NV1 writable > instead of RES0 for VHE-only nested guests where KVM advertises > ID_AA64MMFR4_EL1.E2H0 = NI_NV1. > > In practice, KVM cannot handle HCR_EL2.NV1 being non-RES0 in any > configuration: limit_nv_id_reg() exposes either VHE-only EL2 with > ID_AA64MMFR4_EL1.E2H0 = NI_NV1 (and E2H0 is not userspace-writable), or > FEAT_E2H0 without FEAT_VHE or FEAT_NV. Checking E2H0 in a feature helper > is therefore unnecessary. > > Drop feat_nv2_e2h0_ni() and mark HCR_EL2_NV1 as an unconditional RES0 bit > using NEEDS_FEAT_FLAG(HCR_EL2_NV1, FORCE_RESx) in hcr_feat_map[] (shared > by HCR_EL2 and NVHCR_EL2). Because HCR_EL2_NV1 is not part of the > architectural HCR_EL2_RES0 mask, check_feat_map() accounts for it as an > implementation-defined FORCE_RESx bit. > > Fixes: b2a324ff01fe ("KVM: arm64: Use HCR_EL2 feature map to drive fixed-value bits") > Suggested-by: Marc Zyngier <maz@kernel.org> > Signed-off-by: Colton Lewis <coltonlewis@google.com> > --- > arch/arm64/kvm/config.c | 7 +------ > 1 file changed, 1 insertion(+), 6 deletions(-) > > diff --git a/arch/arm64/kvm/config.c b/arch/arm64/kvm/config.c > index adaf26ab93b85..2e81f45dc9921 100644 > --- a/arch/arm64/kvm/config.c > +++ b/arch/arm64/kvm/config.c > @@ -256,11 +256,6 @@ static bool feat_nv2(struct kvm *kvm) > kvm_has_feat(kvm, ID_AA64MMFR2_EL1, NV, NV2)); > } > > -static bool feat_nv2_e2h0_ni(struct kvm *kvm) > -{ > - return feat_nv2(kvm) && !kvm_has_feat(kvm, FEAT_E2H0); > -} > - > static bool feat_rasv1p1(struct kvm *kvm) > { > return (kvm_has_feat(kvm, ID_AA64PFR0_EL1, RAS, V1P1) || > @@ -1034,7 +1029,7 @@ static const struct reg_bits_to_feat_map hcr_feat_map[] = { > HCR_EL2_NV2 | > HCR_EL2_NV, > feat_nv2), > - NEEDS_FEAT(HCR_EL2_NV1, feat_nv2_e2h0_ni), /* Missing from JSON */ > + NEEDS_FEAT_FLAG(HCR_EL2_NV1, FORCE_RESx), /* Missing from JSON, RES0 as an IMPDEF choice */ > NEEDS_FEAT(HCR_EL2_API | > HCR_EL2_APK, > feat_pauth), > -- > 2.56.1.750.gf70dc682cf-goog > Right, KVM nested requires VHE, which uses E2H, which conflicts with NV1. FWIW: Reviewed-by: Leonardo Bras <leo.bras@arm.com> Thanks! Leo ^ permalink raw reply [flat|nested] 4+ messages in thread
* [PATCH v3 2/2] KVM: arm64: selftests: Verify HCR_EL2.NV1 is RES0 when E2H0 is NI_NV1 2026-11-18 20:36 [PATCH v3 0/2] KVM: arm64: Force HCR_EL2.NV1 to RES0 unconditionally Colton Lewis 2026-11-18 20:36 ` [PATCH v3 1/2] " Colton Lewis @ 2026-11-18 20:36 ` Colton Lewis 1 sibling, 0 replies; 4+ messages in thread From: Colton Lewis @ 2026-11-18 20:36 UTC (permalink / raw) To: Marc Zyngier, Oliver Upton Cc: Leonardo Bras, Joey Gouly, Suzuki K Poulose, Zenghui Yu, Fuad Tabba, kvmarm, linux-arm-kernel, kvm, linux-kernel, Colton Lewis In a VHE-only EL2 guest (KVM_ARM_VCPU_HAS_EL2 without KVM_ARM_VCPU_HAS_EL2_E2H0), KVM advertises ID_AA64MMFR4_EL1.E2H0 as NI_NV1, making HCR_EL2.E2H RES1 and HCR_EL2.NV1 RES0. Extend hello_el2 to attempt clearing HCR_EL2.E2H and setting HCR_EL2.NV1 from userspace via KVM_SET_ONE_REG, verifying that HCR_EL2.E2H remains RES1 and HCR_EL2.NV1 is masked to RES0. Add a comment explaining why the check is performed from userspace rather than inside the guest. Signed-off-by: Colton Lewis <coltonlewis@google.com> --- tools/testing/selftests/kvm/arm64/hello_el2.c | 15 +++++++++++++++ 1 file changed, 15 insertions(+) diff --git a/tools/testing/selftests/kvm/arm64/hello_el2.c b/tools/testing/selftests/kvm/arm64/hello_el2.c index bbe6862c6ab16..3bd3596f5fb1e 100644 --- a/tools/testing/selftests/kvm/arm64/hello_el2.c +++ b/tools/testing/selftests/kvm/arm64/hello_el2.c @@ -45,6 +45,7 @@ int main(void) struct kvm_vcpu *vcpu; struct kvm_vm *vm; struct ucall uc; + u64 hcr; TEST_REQUIRE(kvm_check_cap(KVM_CAP_ARM_EL2)); @@ -66,6 +67,20 @@ int main(void) TEST_FAIL("Unhandled ucall: %ld\n", uc.cmd); } + /* + * Verify RES0/RES1 enforcement from userspace rather than inside + * the guest. From the guest's perspective RESx bits may be + * stateful (reading back as written) even though their value is + * ignored, whereas KVM_SET_ONE_REG applies the effective masks + * directly. + */ + hcr = vcpu_get_reg(vcpu, KVM_ARM64_SYS_REG(SYS_HCR_EL2)); + vcpu_set_reg(vcpu, KVM_ARM64_SYS_REG(SYS_HCR_EL2), + (hcr & ~HCR_EL2_E2H) | HCR_EL2_NV1); + hcr = vcpu_get_reg(vcpu, KVM_ARM64_SYS_REG(SYS_HCR_EL2)); + TEST_ASSERT(hcr & HCR_EL2_E2H, "HCR_EL2.E2H should be RES1"); + TEST_ASSERT(!(hcr & HCR_EL2_NV1), "HCR_EL2.NV1 should be RES0"); + kvm_vm_free(vm); return 0; } -- 2.56.1.750.gf70dc682cf-goog ^ permalink raw reply related [flat|nested] 4+ messages in thread
end of thread, other threads:[~2026-11-19 12:09 UTC | newest] Thread overview: 4+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2026-11-18 20:36 [PATCH v3 0/2] KVM: arm64: Force HCR_EL2.NV1 to RES0 unconditionally Colton Lewis 2026-11-18 20:36 ` [PATCH v3 1/2] " Colton Lewis 2026-11-19 12:09 ` Leonardo Bras 2026-11-18 20:36 ` [PATCH v3 2/2] KVM: arm64: selftests: Verify HCR_EL2.NV1 is RES0 when E2H0 is NI_NV1 Colton Lewis
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox