KVM: x86: MCE fixes
v3 cover KVM: x86: MCE fixes
These two patches are somewhat unrelated, but patch 2 came out of Sean's suggestions from reviewing the first patch (see [1]). Patch 1 is unchanged.
v3:
* Address Sashiko's review. While incorrect in pointing out MCIP's role, it was partially correct in the sense that the MCE logging rules were still not fully compliant to the SDM.
v2:
* Add patch 2.
[1] https://lore.kernel.org/all/[email protected]/
Carlos López (2):
KVM: x86: Fix array_index_nospec() protection in
kvm_vcpu_ioctl_x86_set_mce()
KVM: x86: Fix MCE logging rules for KVM_X86_SET_MCE
arch/x86/kvm/x86.c | 65 +++++++++++++++++++++++++++++----------------- 1 file changed, 41 insertions(+), 24 deletions(-)
base-commit: de3a35be92d2391ece4bf3143ef2887192625fd0
Signature · 1 line
2.51.0
PATCH v3 1/2 KVM: x86: Fix array_index_nospec() protection in kvm_vcpu_ioctl_x86_set_mce()
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().
--- arch/x86/kvm/x86.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
arch/x86/kvm/x86.c +2 -1
@@ -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);
Signature · 1 line
2.51.0
PATCH v3 2/2 KVM: x86: Fix MCE logging rules for KVM_X86_SET_MCE
In reply to Carlos LópezWhen 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.
Perform the logging before checking MCi_CTL, unless there is already a valid UC error logged for the bank, in which case the SDM (17.3.2.2 "IA32_MCi_STATUS MSRS") specifies that error information should not be overwritten.
To avoid even more complex control flow, hoist the logging logic into a separate function, which then enables the removal of the non-UC branch in kvm_vcpu_ioctl_x86_set_mce(), which only existed to perform logging.
Note that the SDM is ambiguous regarding the effects of IA32_MCG_CTL on logging, so preserve the existing logic (i.e. do not log the error if MCG_CTL is disabled).
--- arch/x86/kvm/x86.c | 62 +++++++++++++++++++++++++++++----------------- 1 file changed, 39 insertions(+), 23 deletions(-)
arch/x86/kvm/x86.c +39 -23
@@ -5462,6 +5462,28 @@ static int kvm_vcpu_x86_set_ucna(struct kvm_vcpu *vcpu, struct kvm_x86_mce *mce,return 0 ; } +/* +* Record @mce into @banks per the SDM logging rules: +* - Per 17.3.2.1, logging happens regardless of IA32_MCi_CTL. +* - Per 17.3.2.2.1 and 17.3.2.2, a UC error does not overwrite a +* previous valid UC log in the same bank; only OVER is set. +* Any other prior state (invalid, or valid non-UC) is replaced, +* with OVER set if a prior valid log was overwritten. +*/ +static void kvm_log_mce (u64 *banks,struct kvm_x86_mce *mce) +{ +bool overflow = banks[1 ] &MCI_STATUS_VAL ; + +if (!overflow || !(banks[1 ] &MCI_STATUS_UC )) { + banks[2 ] = mce->addr; + banks[3 ] = mce->misc; + banks[1 ] = mce->status; + } + +if (overflow) + banks[1 ] |=MCI_STATUS_OVER ; +} +static int kvm_vcpu_ioctl_x86_set_mce (struct kvm_vcpu *vcpu,struct kvm_x86_mce *mce) {
@@ -5485,34 +5507,28 @@ 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 ; + +kvm_log_mce (banks, mce); + +if (!(mce->status &MCI_STATUS_UC )) +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 ) +if (banks[0 ] != ~(u64 )0 )return 0 ; -if (mce->status &MCI_STATUS_UC ) { -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 )) { -if (banks[1 ] &MCI_STATUS_VAL ) - mce->status |=MCI_STATUS_OVER ; - banks[2 ] = mce->addr; - banks[3 ] = mce->misc; - banks[1 ] = mce->status; - }else - banks[1 ] |=MCI_STATUS_OVER ; + +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 ; + } + + vcpu->arch.mcg_status = mce->mcg_status; +kvm_queue_exception (vcpu,MC_VECTOR ); +return 0 ; }
Signature · 1 line
2.51.0
Quoted 10 lines · Carlos López
These two patches are somewhat unrelated, but patch 2 came out of Sean's suggestions from reviewing the first patch (see [1]). Patch 1 is unchanged.
v3:
* Address Sashiko's review. While incorrect in pointing out MCIP's role, it was partially correct in the sense that the MCE logging rules were still not fully compliant to the SDM.[...]
Applied patch 1 to kvm-x86 misc so it doesn't get waylaid any longer. I'll respond to patch 2 at some point (hopefully today, but it might not be until next week).
[1/2] KVM: x86: Fix array_index_nospec() protection in kvm_vcpu_ioctl_x86_set_mce()
https://github.com/kvm-x86/linux/commit/249f9be58c7f
Hi Sean,
Quoted 10 lines · Sean Christopherson
Quoted 9 lines · Carlos López
These two patches are somewhat unrelated, but patch 2 came out of Sean's suggestions from reviewing the first patch (see [1]). Patch 1 is unchanged.
v3:
* Address Sashiko's review. While incorrect in pointing out MCIP's role, it was partially correct in the sense that the MCE logging rules were still not fully compliant to the SDM.
Quoted 5 lines
Quoted 1 line
[...]
Applied patch 1 to kvm-x86 misc so it doesn't get waylaid any longer. I'll respond to patch 2 at some point (hopefully today, but it might not be until next week).
I believe this fell through the cracks. Did you have time to take another look at patch 2? Not urgent.
Quoted 5 lines
[1/2] KVM: x86: Fix
array_index_nospec()protection inkvm_vcpu_ioctl_x86_set_mce()https://github.com/kvm-x86/linux/commit/249f9be58c7f
Thanks,
Carlos