Relax canonical checks on some arch msrs
v2 cover Relax canonical checks on some arch msrs
Recently we came up upon a failure where likely the guest writes 0xff4547ceb1600000 to MSR_KERNEL_GS_BASE and later on, qemu sets this value via KVM_PUT_MSRS, and is rejected by the kernel, likely due to not being canonical in 4 level paging.
I did some reverse engineering and to my surprise I found out that both Intel and AMD have very loose checks in regard to non canonical addresses written to this and several other msrs, when the CPU supports 5 level paging.
Patch #1 addresses this, making KVM tolerate this.
Patch #2 is just a fix for a semi theoretical bug, found while trying to debug the issue.
V2: addressed a very good feedback from Chao Gao. Thanks!
Best regards,
Maxim Levitsky
Maxim Levitsky (2):
KVM: x86: relax canonical check for some x86 architectural msrs
KVM: SVM: fix emulation of msr reads/writes of MSR_FS_BASE and
MSR_GS_BASE
arch/x86/kvm/svm/svm.c | 12 ++++++++++++ arch/x86/kvm/x86.c | 11 ++++++++++- 2 files changed, 22 insertions(+), 1 deletion(-)
Signature · 1 line
2.40.1
PATCH v2 1/2 KVM: x86: relax canonical check for some x86 architectural msrs
Several architectural msrs (e.g MSR_KERNEL_GS_BASE) must contain a canonical address, and according to Intel PRM, this is enforced by a #GP canonical check during MSR write.
However as it turns out, the supported address width used for this canonical check is determined only by host cpu model:
if CPU *supports* 5 level paging, the width will be 57 regardless of the state of CR4.LA57.
Experemental tests on a Sapphire Rapids CPU and on a Zen4 CPU confirm this behavior.
In addition to that, the Intel ISA extension manual mentions that this might be the architectural behavior:
Architecture Instruction Set Extensions and Future Features Programming Reference [1]. Chapter 6.4:
"CANONICALITY CHECKING FOR DATA ADDRESSES WRITTEN TO CONTROL REGISTERS AND MSRS"
"In Processors that support LAM continue to require the addresses written to control registers or MSRs to be 57-bit canonical if the processor _supports_ 5-level paging or 48-bit canonical if it supports only 4-level paging"
[1]: https://cdrdv2.intel.com/v1/dl/getContent/671368
--- arch/x86/kvm/x86.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-)
arch/x86/kvm/x86.c +10 -1
@@ -1844,7 +1844,16 @@ static int __kvm_set_msr(struct kvm_vcpu *vcpu, u32 index, u64 data,case MSR_KERNEL_GS_BASE :case MSR_CSTAR :case MSR_LSTAR : -if (is_noncanonical_address (data, vcpu)) + +/* +* Both AMD and Intel cpus allow values which +* are canonical in the 5 level paging mode but are not +* canonical in the 4 level paging mode to be written +* to the above MSRs, as long as the host CPU supports +* 5 level paging, regardless of the state of the CR4.LA57. +*/ +if (!__is_canonical_address (data, +kvm_cpu_cap_has (X86_FEATURE_LA57 ) ?57 :48 ))return 1 ;break ;case MSR_IA32_SYSENTER_EIP :
Signature · 1 line
2.40.1
Sean Christopherson
maintainerRe: [PATCH v2 1/2] KVM: x86: relax canonical check for some x86 architectural msrs
Quoted 7 lines · Maxim Levitsky
Several architectural msrs (e.g
MSR_KERNEL_GS_BASE) must contain a canonical address, and according to Intel PRM, this is enforced by a #GP canonical check during MSR write.However as it turns out, the supported address width used for this canonical check is determined only by host cpu model:
Please try to wrap consistently and sanely, this is unnecessarily hard to read because every paragraph manages to wrap at a different column.
Quoted 36 lines
if CPU *supports* 5 level paging, the width will be 57 regardless of the state of CR4.LA57.
Experemental tests on a Sapphire Rapids CPU and on a Zen4 CPU confirm this behavior.
In addition to that, the Intel ISA extension manual mentions that this might be the architectural behavior:
Architecture Instruction Set Extensions and Future Features Programming Reference [1]. Chapter 6.4:
"CANONICALITY CHECKING FOR DATA ADDRESSES WRITTEN TO CONTROL REGISTERS AND MSRS"
"In Processors that support LAM continue to require the addresses written to control registers or MSRs to be 57-bit canonical if the processor _supports_ 5-level paging or 48-bit canonical if it supports only 4-level paging"
[1]: https://cdrdv2.intel.com/v1/dl/getContent/671368
Suggested-by Chao Gao <[email protected]>Signed-off-by Maxim Levitsky <[email protected]>--- arch/x86/kvm/x86.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-)arch/x86/kvm/x86.c +0
-1@@ -1844,7 +1844,16 @@ static int __kvm_set_msr(struct kvm_vcpu *vcpu, u32 index, u64 data,case MSR_KERNEL_GS_BASE :case MSR_CSTAR :case MSR_LSTAR : -if (is_noncanonical_address (data, vcpu))
arch/x86/kvm/x86.c static int __kvm_set_msr(struct kvm_vcpu *vcpu, u32 index, u64 data, · line 1856
Quoted 10 lines
arch/x86/kvm/x86.c +10
-0+ +/* +* Both AMD and Intel cpus allow values which +* are canonical in the 5 level paging mode but are not +* canonical in the 4 level paging mode to be written +* to the above MSRs, as long as the host CPU supports +* 5 level paging, regardless of the state of the CR4.LA57. +*/ +if (!__is_canonical_address (data, +kvm_cpu_cap_has (X86_FEATURE_LA57 ) ?57 :48 ))
Please align indentation.
Checking kvm_cpu_cap_has() is wrong. What the _host_ supports is irrelevant, what matters is what the guest CPU supports, i.e. this should check guest CPUID. Ah, but for safety, KVM also needs to check kvm_cpu_cap_has() to prevent faulting on a bad load into hardware. Which means adding a "governed" feature until my CPUID rework lands.
And I'm pretty sure this fix is incomplete, as nVMX's consistency checks on MSRs that are loaded via dedicated VMCS fields likely need the same treatment, e.g. presumably these checks should follow the MSR handling.
if (CC(is_noncanonical_address(vmcs12->host_ia32_sysenter_esp, vcpu)) || CC(is_noncanonical_address(vmcs12->host_ia32_sysenter_eip, vcpu))) return -EINVAL; (CC(is_noncanonical_address(vmcs12->guest_bndcfgs & PAGE_MASK, vcpu)) ||
So I think we probably need a dedicated helper for MSRs.
Hmm, and I suspect these are wrong too, but in a different way. Toggling host LA57 on VM-Exit is legal[*], so logically, KVM should use CR4.LA57 from vmcs12->host_cr4, not the vCPU's current CR4 value. Which makes me _really_ curious if Intel CPUs actually get that right.
if (CC(is_noncanonical_address(vmcs12->host_fs_base, vcpu)) || CC(is_noncanonical_address(vmcs12->host_gs_base, vcpu)) || CC(is_noncanonical_address(vmcs12->host_gdtr_base, vcpu)) || CC(is_noncanonical_address(vmcs12->host_idtr_base, vcpu)) || CC(is_noncanonical_address(vmcs12->host_tr_base, vcpu)) || CC(is_noncanonical_address(vmcs12->host_rip, vcpu))) return -EINVAL;
Re: [PATCH v2 1/2] KVM: x86: relax canonical check for some x86 architectural msrs
Quoted 10 lines
Checking
kvm_cpu_cap_has()is wrong. What the _host_ supports is irrelevant, what matters is what the guest CPU supports, i.e. this should check guest CPUID. Ah, but for safety, KVM also needs to checkkvm_cpu_cap_has()to prevent faulting on a bad load into hardware. Which means adding a "governed" feature until my CPUID rework lands.And I'm pretty sure this fix is incomplete, as nVMX's consistency checks on MSRs that are loaded via dedicated VMCS fields likely need the same treatment, e.g. presumably these checks should follow the MSR handling.
Quoted 8 lines
if (CC(is_noncanonical_address(vmcs12->host_ia32_sysenter_esp, vcpu)) || CC(is_noncanonical_address(vmcs12->host_ia32_sysenter_eip, vcpu))) return -EINVAL; (CC(is_noncanonical_address(vmcs12->guest_bndcfgs & PAGE_MASK, vcpu)) ||So I think we probably need a dedicated helper for MSRs.
+1
Quoted 2 lines
Hmm, and I suspect these are wrong too, but in a different way. Toggling host
Actually they are wrong in the _same_ way because ...
Quoted 3 lines
LA57 on VM-Exit is legal[*], so logically, KVM should use CR4.LA57 from
vmcs12->host_cr4, not the vCPU's current CR4 value. Which makes me _really_ curious if Intel CPUs actually get that right.
... according to Section 3.1 in 5-level paging whitepaper[1], on VM entry, the canonicality checking about the host/guest state (except guest RIP) is also based on whether the CPU supports 5-level paging, i.e., host or guest CR4.LA57 isn't checked.
[1]: https://cdrdv2.intel.com/v1/dl/getContent/671442?fileName=5-level-paging-white-paper.pdf
Quoted 7 lines
if (CC(is_noncanonical_address(vmcs12->host_fs_base, vcpu)) || CC(is_noncanonical_address(vmcs12->host_gs_base, vcpu)) || CC(is_noncanonical_address(vmcs12->host_gdtr_base, vcpu)) || CC(is_noncanonical_address(vmcs12->host_idtr_base, vcpu)) || CC(is_noncanonical_address(vmcs12->host_tr_base, vcpu)) || CC(is_noncanonical_address(vmcs12->host_rip, vcpu)))
not sure how host_rip should be handled because it isn't documented in that spec.
Quoted 3 lines
return -EINVAL;
Re: [PATCH v2 1/2] KVM: x86: relax canonical check for some x86 architectural msrs
In reply to Sean ChristophersonУ пт, 2024-08-02 у 08:53 -0700, Sean Christopherson пише:
Quoted 11 lines
Quoted 11 lines
Quoted 11 lines
Quoted 11 lines · Maxim Levitsky
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Several architectural msrs (e.g
MSR_KERNEL_GS_BASE) must contain a canonical address, and according to Intel PRM, this is enforced by a #GP canonical check during MSR write.However as it turns out, the supported address width used for this canonical check is determined only by host cpu model:
Please try to wrap consistently and sanely, this is unnecessarily hard to read because every paragraph manages to wrap at a different column.
I'll take a note.
Quoted 50 lines
Quoted 50 lines
Quoted 50 lines
Quoted 50 lines
Quoted 39 lines
Quoted 39 lines
Quoted 39 lines
Quoted 39 lines
if CPU *supports* 5 level paging, the width will be 57 regardless of the state of CR4.LA57.
Experemental tests on a Sapphire Rapids CPU and on a Zen4 CPU confirm this behavior.
In addition to that, the Intel ISA extension manual mentions that this might be the architectural behavior:
Architecture Instruction Set Extensions and Future Features Programming Reference [1]. Chapter 6.4:
"CANONICALITY CHECKING FOR DATA ADDRESSES WRITTEN TO CONTROL REGISTERS AND MSRS"
"In Processors that support LAM continue to require the addresses written to control registers or MSRs to be 57-bit canonical if the processor _supports_ 5-level paging or 48-bit canonical if it supports only 4-level paging"
[1]: https://cdrdv2.intel.com/v1/dl/getContent/671368
Suggested-by Chao Gao <[email protected]>Signed-off-by Maxim Levitsky <[email protected]>---
arch/x86/kvm/x86.c | 11 ++++++++++-1 file changed, 10 insertions(+), 1 deletion(-)arch/x86/kvm/x86.c +0
-0@@ -1844,7 +1844,16 @@ static int __kvm_set_msr(struct kvm_vcpu *vcpu, u32 index, u64 data,case
MSR_KERNEL_GS_BASE:
caseMSR_CSTAR:
caseMSR_LSTAR:
- if (is_noncanonical_address(data, vcpu)) +
+ /*Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
+ * Both AMD and Intel cpus allow values which
+ * are canonical in the 5 level paging mode but are not + * canonical in the 4 level paging mode to be written + * to the above MSRs, as long as the host CPU supports + * 5 level paging, regardless of the state of the CR4.LA57. + */
+ if (!__is_canonical_address(data,
+ kvm_cpu_cap_has(X86_FEATURE_LA57) ? 57 : 48))Please align indentation.
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Checking
kvm_cpu_cap_has()is wrong. What the _host_ supports is irrelevant, what matters is what the guest CPU supports, i.e. this should check guest CPUID. Ah, but for safety, KVM also needs to checkkvm_cpu_cap_has()to prevent faulting on a bad load into hardware. Which means adding a "governed" feature until my CPUID rework lands.
Well the problem is that we passthrough these MSRs, and that means that the guest can modify them at will, and only ucode can prevent it from doing so.
So even if the 5 level paging is disabled in the guest's CPUID, but host supports it, nothing will prevent the guest to write non canonical value to one of those MSRs, and later KVM during migration or just KVM_SET_SREGS will fail.
Thus I used kvm_cpu_cap_has on purpose to make KVM follow the actual ucode behavior.
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
And I'm pretty sure this fix is incomplete, as nVMX's consistency checks on MSRs that are loaded via dedicated VMCS fields likely need the same treatment, e.g. presumably these checks should follow the MSR handling.
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
if (
CC(is_noncanonical_address(vmcs12->host_ia32_sysenter_esp, vcpu))||CC(is_noncanonical_address(vmcs12->host_ia32_sysenter_eip, vcpu))) return-EINVAL;(
CC(is_noncanonical_address(vmcs12->guest_bndcfgs & PAGE_MASK, vcpu))||So I think we probably need a dedicated helper for MSRs.
This is a long story - I didn't want to make this patch explode in size, especially since it wasn't clear if this is architectural behavior.
Even now I can't be sure that it is architectural behavior - I haven't found anything in the latest official PRM of either AMD nor Intel.
I'll add a separate patch to fix the nested code path in the next version of the patches.
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
Hmm, and I suspect these are wrong too, but in a different way. Toggling host LA57 on VM-Exit is legal[*], so logically, KVM should use CR4.LA57 from
vmcs12->host_cr4, not the vCPU's current CR4 value. Which makes me _really_ curious if Intel CPUs actually get that right.
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
if (
CC(is_noncanonical_address(vmcs12->host_fs_base, vcpu))||CC(is_noncanonical_address(vmcs12->host_gs_base, vcpu))||CC(is_noncanonical_address(vmcs12->host_gdtr_base, vcpu)) || CC(is_noncanonical_address(vmcs12->host_idtr_base, vcpu)) ||
CC(is_noncanonical_address(vmcs12->host_tr_base, vcpu))||CC(is_noncanonical_address(vmcs12->host_rip, vcpu))) return-EINVAL;
This is a good question, I'll check this on bare metal.
Best regards,
Maxim Levitsky
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Sean Christopherson
maintainerRe: [PATCH v2 1/2] KVM: x86: relax canonical check for some x86 architectural msrs
Quoted 29 lines · On Mon, Aug 05, 2024, [email protected] wrote:
У пт, 2024-08-02 у 08:53 -0700, Sean Christopherson пише:
Quoted 22 lines
Quoted 22 lines
Quoted 22 lines
Quoted 22 lines
Quoted 11 lines
Quoted 11 lines
Quoted 11 lines
Quoted 11 lines
arch/x86/kvm/x86.c +0
-0@@ -1844,7 +1844,16 @@ static int __kvm_set_msr(struct kvm_vcpu *vcpu, u32 index, u64 data,case
MSR_KERNEL_GS_BASE:
caseMSR_CSTAR:
caseMSR_LSTAR:
- if (is_noncanonical_address(data, vcpu)) +
+ /*Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
+ * Both AMD and Intel cpus allow values which
+ * are canonical in the 5 level paging mode but are not + * canonical in the 4 level paging mode to be written + * to the above MSRs, as long as the host CPU supports + * 5 level paging, regardless of the state of the CR4.LA57. + */
+ if (!__is_canonical_address(data,
+ kvm_cpu_cap_has(X86_FEATURE_LA57) ? 57 : 48))Please align indentation.
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Checking
kvm_cpu_cap_has()is wrong. What the _host_ supports is irrelevant, what matters is what the guest CPU supports, i.e. this should check guest CPUID. Ah, but for safety, KVM also needs to checkkvm_cpu_cap_has()to prevent faulting on a bad load into hardware. Which means adding a "governed" feature until my CPUID rework lands.
Quoted 6 lines
Well the problem is that we passthrough these MSRs, and that means that the guest can modify them at will, and only ucode can prevent it from doing so.
So even if the 5 level paging is disabled in the guest's CPUID, but host supports it, nothing will prevent the guest to write non canonical value to one of those MSRs, and later KVM during migration or just
KVM_SET_SREGSwill fail.
Ahh, and now I recall the discussions around the virtualization holes with LA57.
Quoted 2 lines
Thus I used
kvm_cpu_cap_hason purpose to make KVM follow the actual ucode behavior.
I'm leaning towards having KVM do the right thing when emulation happens to be triggered. If KVM checks kvm_cpu_cap_has() instead of guest_cpu_cap_has() (looking at the future), then KVM will extend the virtualization hole to MSRs that are never passed through, and also to the nested VMX checks. Or I suppose we could add separate helpers for passthrough MSRs vs. non-passthrough, but that seems like it'd add very little value and a lot of maintenance burden.
Practically speaking, outside of tests, I can't imagine the guest will ever care if there is inconsistent behavior with respect to loading non-canonical values into MSRs.
Re: [PATCH v2 1/2] KVM: x86: relax canonical check for some x86 architectural msrs
У пн, 2024-08-05 у 09:39 -0700, Sean Christopherson пише:
Quoted 49 lines · On Mon, Aug 05, 2024, [email protected] wrote:
Quoted 29 lines
У пт, 2024-08-02 у 08:53 -0700, Sean Christopherson пише:
Quoted 22 lines
Quoted 22 lines
Quoted 22 lines
Quoted 22 lines
Quoted 11 lines
Quoted 11 lines
Quoted 11 lines
Quoted 11 lines
arch/x86/kvm/x86.c +0
-0@@ -1844,7 +1844,16 @@ static int __kvm_set_msr(struct kvm_vcpu *vcpu, u32 index, u64 data,case
MSR_KERNEL_GS_BASE:
caseMSR_CSTAR:
caseMSR_LSTAR:
- if (is_noncanonical_address(data, vcpu)) +
+ /*Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
+ * Both AMD and Intel cpus allow values which
+ * are canonical in the 5 level paging mode but are not + * canonical in the 4 level paging mode to be written + * to the above MSRs, as long as the host CPU supports + * 5 level paging, regardless of the state of the CR4.LA57. + */
+ if (!__is_canonical_address(data,
+ kvm_cpu_cap_has(X86_FEATURE_LA57) ? 57 : 48))Please align indentation.
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Checking
kvm_cpu_cap_has()is wrong. What the _host_ supports is irrelevant, what matters is what the guest CPU supports, i.e. this should check guest CPUID. Ah, but for safety, KVM also needs to checkkvm_cpu_cap_has()to prevent faulting on a bad load into hardware. Which means adding a "governed" feature until my CPUID rework lands.Quoted 6 lines
Well the problem is that we passthrough these MSRs, and that means that the guest can modify them at will, and only ucode can prevent it from doing so.
So even if the 5 level paging is disabled in the guest's CPUID, but host supports it, nothing will prevent the guest to write non canonical value to one of those MSRs, and later KVM during migration or just
KVM_SET_SREGSwill fail.
Ahh, and now I recall the discussions around the virtualization holes with LA57.Quoted 2 lines
Thus I used
kvm_cpu_cap_hason purpose to make KVM follow the actual ucode behavior.I'm leaning towards having KVM do the right thing when emulation happens to be triggered. If KVM checks
kvm_cpu_cap_has()instead ofguest_cpu_cap_has()(looking at the future), then KVM will extend the virtualization hole to MSRs that are never passed through, and also to the nested VMX checks. Or I suppose we could add separate helpers for passthrough MSRs vs. non-passthrough, but that seems like it'd add very little value and a lot of maintenance burden.
Quoted 4 lines
Practically speaking, outside of tests, I can't imagine the guest will ever care if there is inconsistent behavior with respect to loading non-canonical values into MSRs.
Hi,
If we weren't allowing the guest (and even nested guest assuming that L1 hypervisor allows it) to write these MSRs directly, I would have agreed with you, but we do allow this.
This means that for example a L2, even a malicious L2, can on purpose write non canonical value to one of these MSRs, and later on, KVM could kill the L0 due to canonical check.
Or L1 (not Linux, because it only lets canonical GS_BASE/FS_BASE), allow the untrusted userspace to write any value to say GS_BASE, thus allowing
malicious L1 userspace to crash L1 (also a security violation).
IMHO if we really want to do it right, we need to disable pass-though of these MSRs if ucode check is more lax than our check, that is if L1 is running without 5 level paging enabled but L0 does have it supported.
I don't know if this configuration is common, and thus how much this will affect performance.
Best regards,
Maxim Levitsky
Sean Christopherson
maintainerRe: [PATCH v2 1/2] KVM: x86: relax canonical check for some x86 architectural msrs
Quoted 33 lines · On Mon, Aug 05, 2024, [email protected] wrote:
У пн, 2024-08-05 у 09:39 -0700, Sean Christopherson пише:
Quoted 27 lines · On Mon, Aug 05, 2024, [email protected] wrote:
Quoted 7 lines
У пт, 2024-08-02 у 08:53 -0700, Sean Christopherson пише:
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
Checking
kvm_cpu_cap_has()is wrong. What the _host_ supports is irrelevant, what matters is what the guest CPU supports, i.e. this should check guest CPUID. Ah, but for safety, KVM also needs to checkkvm_cpu_cap_has()to prevent faulting on a bad load into hardware. Which means adding a "governed" feature until my CPUID rework lands.Quoted 6 lines
Well the problem is that we passthrough these MSRs, and that means that the guest can modify them at will, and only ucode can prevent it from doing so.
So even if the 5 level paging is disabled in the guest's CPUID, but host supports it, nothing will prevent the guest to write non canonical value to one of those MSRs, and later KVM during migration or just
KVM_SET_SREGSwill fail.
Ahh, and now I recall the discussions around the virtualization holes with LA57.Quoted 2 lines
Thus I used
kvm_cpu_cap_hason purpose to make KVM follow the actual ucode behavior.I'm leaning towards having KVM do the right thing when emulation happens to be triggered. If KVM checks
kvm_cpu_cap_has()instead ofguest_cpu_cap_has()(looking at the future), then KVM will extend the virtualization hole to MSRs that are never passed through, and also to the nested VMX checks. Or I suppose we could add separate helpers for passthrough MSRs vs. non-passthrough, but that seems like it'd add very little value and a lot of maintenance burden.Quoted 4 lines
Practically speaking, outside of tests, I can't imagine the guest will ever care if there is inconsistent behavior with respect to loading non-canonical values into MSRs.
Quoted 8 lines
Hi,
If we weren't allowing the guest (and even nested guest assuming that L1 hypervisor allows it) to write these MSRs directly, I would have agreed with you, but we do allow this.
This means that for example a L2, even a malicious L2, can on purpose write non canonical value to one of these MSRs, and later on, KVM could kill the L0
L1?
Quoted 1 line
due to canonical check.
Ugh, right, if L1 manually saves/restores MSRs and happens to trigger emulation on WRMSR at the 'wrong" time.
Host userspace save/restore would suffer the same problem. We could grant host userspace accesses an exception, but that's rather pointless.
Quoted 3 lines
Or L1 (not Linux, because it only lets canonical GS_BASE/FS_BASE), allow the untrusted userspace to write any value to say
GS_BASE, thus allowing malicious L1 userspace to crash L1 (also a security violation).
FWIW, I don't think this is possible. WR{FS,GS}BASE and other instructions that load FS/GS.base honor CR4.LA57, it's only WRMSR that does not.
Quoted 6 lines
IMHO if we really want to do it right, we need to disable pass-though of these MSRs if ucode check is more lax than our check, that is if L1 is running without 5 level paging enabled but L0 does have it supported.
I don't know if this configuration is common, and thus how much this will affect performance.
MSR_FS_BASE and SR_KERNEL_GS_BASE are hot spots when WR{FS,GS}BASE are unsupported, or if the guest kernels doesn't utilize those instructions.
All in all, I agree it's not worth trying to plug the virtualization hole for MSRs, especially since mimicking hardware yields much simpler code overall. E.g. add a dedicated MSR helper, and have that one check kvm_cpu_cap_has(), including in VM-Entry flows, but keep the existing is_noncanonical_address() for all non-WRMSR path.
Something like this?
static inline bool is_noncanonical_msr_value(u64 la)
{
u8 virt_addr_bits = kvm_cpu_cap_has(X86_FEATURE_LA57) ? 57 : 48;
return !__is_canonical_address(la, virt_addr_bits);
}PATCH v2 2/2 KVM: SVM: fix emulation of msr reads/writes of MSR_FS_BASE and MSR_GS_BASE
In reply to Maxim LevitskyIf these msrs are read by the emulator (e.g due to 'force emulation' prefix), SVM code currently fails to extract the corresponding segment bases, and return them to the emulator.
Fix that.
--- arch/x86/kvm/svm/svm.c | 12 ++++++++++++ 1 file changed, 12 insertions(+)
arch/x86/kvm/svm/svm.c +12 -0
@@ -2875,6 +2875,12 @@ static int svm_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)case MSR_CSTAR : msr_info->data = svm->vmcb01.ptr->save.cstar;break ; +case MSR_GS_BASE : + msr_info->data = svm->vmcb01.ptr->save.gs.base; +break ; +case MSR_FS_BASE : + msr_info->data = svm->vmcb01.ptr->save.fs.base; +break ;case MSR_KERNEL_GS_BASE : msr_info->data = svm->vmcb01.ptr->save.kernel_gs_base;break ;
@@ -3100,6 +3106,12 @@ static int svm_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr)case MSR_CSTAR : svm->vmcb01.ptr->save.cstar = data;break ; +case MSR_GS_BASE : + svm->vmcb01.ptr->save.gs.base = data; +break ; +case MSR_FS_BASE : + svm->vmcb01.ptr->save.fs.base = data; +break ;case MSR_KERNEL_GS_BASE : svm->vmcb01.ptr->save.kernel_gs_base = data;break ;
Signature · 1 line
2.40.1