* [PATCH v2 1/2] KVM: x86: Fix array_index_nospec() protection in kvm_vcpu_ioctl_x86_set_mce()
2026-06-09 10:55 [PATCH v2 0/2] KVM: x86: MCE fixes Carlos López
@ 2026-06-09 10:55 ` Carlos López
2026-06-09 10:55 ` [PATCH v2 2/2] KVM: x86: Do not gate MCE logging based on IA32_MCi_CTL Carlos López
1 sibling, 0 replies; 3+ messages in thread
From: Carlos López @ 2026-06-09 10:55 UTC (permalink / raw)
To: kvm, seanjc, pbonzini
Cc: linux-kernel, x86, tglx, mingo, dave.hansen, hpa,
Carlos López, Borislav Petkov, Jue Wang
Commit aebc3ca19063 ("KVM: x86: Enable CMCI capability by default and
handle injected UCNA errors") introduced kvm_vcpu_x86_set_ucna(), which
accesses @vcpu->arch.mci_ctl2_banks[] using @mce->bank as the index. The
@mce struct is user-controlled, provided via the KVM_X86_SET_MCE ioctl.
The caller of this function, kvm_vcpu_ioctl_x86_set_mce(), bounds-checks
@mce->bank and applies array_index_nospec() to advance the @banks
pointer, but @mce->bank itself is passed through unclamped. On a
speculative path that bypasses the bounds check, the raw @mce->bank
value can index mci_ctl2_banks[] out-of-bounds.
In practice this is a very weak gadget, and would at most allow leaking
a single bit in a 64-bit integer, but prevent potential future issues by
clamping @mce->bank in place with array_index_nospec(), before passing
the struct to kvm_vcpu_x86_set_ucna().
Fixes: aebc3ca19063 ("KVM: x86: Enable CMCI capability by default and handle injected UCNA errors")
Signed-off-by: Carlos López <clopez@suse.de>
---
arch/x86/kvm/x86.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index cf122b8c3210..77a780177c4e 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -5472,7 +5472,8 @@ static int kvm_vcpu_ioctl_x86_set_mce(struct kvm_vcpu *vcpu,
if (mce->bank >= bank_num || !(mce->status & MCI_STATUS_VAL))
return -EINVAL;
- banks += array_index_nospec(4 * mce->bank, 4 * bank_num);
+ mce->bank = array_index_nospec(mce->bank, bank_num);
+ banks += 4 * mce->bank;
if (is_ucna(mce))
return kvm_vcpu_x86_set_ucna(vcpu, mce, banks);
--
2.51.0
^ permalink raw reply [flat|nested] 3+ messages in thread* [PATCH v2 2/2] KVM: x86: Do not gate MCE logging based on IA32_MCi_CTL
2026-06-09 10:55 [PATCH v2 0/2] KVM: x86: MCE fixes Carlos López
2026-06-09 10:55 ` [PATCH v2 1/2] KVM: x86: Fix array_index_nospec() protection in kvm_vcpu_ioctl_x86_set_mce() Carlos López
@ 2026-06-09 10:55 ` Carlos López
1 sibling, 0 replies; 3+ messages in thread
From: Carlos López @ 2026-06-09 10:55 UTC (permalink / raw)
To: kvm, seanjc, pbonzini
Cc: linux-kernel, x86, tglx, mingo, dave.hansen, hpa,
Carlos López, Borislav Petkov, Ying Huang, Avi Kivity
When userspace issues KVM_X86_SET_MCE, kvm_vcpu_ioctl_x86_set_mce()
decides whether to log an uncorrectable MCE by looking at the
corresponding IA32_MCi_CTL MSR. This is not the behavior specified in
the Intel SDM (17.3.2.1 IA32_MCi_CTL MSRs):
Setting an EEj flag enables signaling #MC of the associated error and
clearing it disables signaling of the error. Error logging happens
regardless of the setting of these bits. The processor drops writes to
bits that are not implemented.
Hoist the logging code, which writes to the corresponding error banks,
above the MCi_CTL check. Keep the #MC delivery after the check, as this
is correct behavior.
Fixes: 890ca9aefa78 ("KVM: Add MCE support")
Signed-off-by: Carlos López <clopez@suse.de>
---
arch/x86/kvm/x86.c | 29 ++++++++++++++++++-----------
1 file changed, 18 insertions(+), 11 deletions(-)
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 77a780177c4e..722bead405b5 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -5485,24 +5485,31 @@ static int kvm_vcpu_ioctl_x86_set_mce(struct kvm_vcpu *vcpu,
if ((mce->status & MCI_STATUS_UC) && (mcg_cap & MCG_CTL_P) &&
vcpu->arch.mcg_ctl != ~(u64)0)
return 0;
- /*
- * if IA32_MCi_CTL is not all 1s, the uncorrected error
- * reporting is disabled for the bank
- */
- if ((mce->status & MCI_STATUS_UC) && banks[0] != ~(u64)0)
- return 0;
+
if (mce->status & MCI_STATUS_UC) {
+ /*
+ * Per the SDM, error logging always happens independently of
+ * IA32_MCi_CTL
+ */
+ if (banks[1] & MCI_STATUS_VAL)
+ mce->status |= MCI_STATUS_OVER;
+ banks[2] = mce->addr;
+ banks[3] = mce->misc;
+ banks[1] = mce->status;
+
+ /*
+ * if IA32_MCi_CTL is not all 1s, the uncorrected error
+ * reporting is disabled for the bank
+ */
+ if (banks[0] != ~(u64)0)
+ return 0;
+
if ((vcpu->arch.mcg_status & MCG_STATUS_MCIP) ||
!kvm_is_cr4_bit_set(vcpu, X86_CR4_MCE)) {
kvm_make_request(KVM_REQ_TRIPLE_FAULT, vcpu);
return 0;
}
- if (banks[1] & MCI_STATUS_VAL)
- mce->status |= MCI_STATUS_OVER;
- banks[2] = mce->addr;
- banks[3] = mce->misc;
vcpu->arch.mcg_status = mce->mcg_status;
- banks[1] = mce->status;
kvm_queue_exception(vcpu, MC_VECTOR);
} else if (!(banks[1] & MCI_STATUS_VAL)
|| !(banks[1] & MCI_STATUS_UC)) {
--
2.51.0
^ permalink raw reply [flat|nested] 3+ messages in thread