Skip to contentSkip to conversation

KVM: arm64: Uphold 64bit-only behavior on asymmetric systems

PATCH Applied Oliver Upton 3 participants 8 messages Aug 16, 2022 – Aug 17, 2022

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

v2 cover KVM: arm64: Uphold 64bit-only behavior on asymmetric systems

Small series to fix a couple issues around when 64bit-only behavior is applied. As KVM is more restrictive than the kernel in terms of 32bit support (no asymmetry), we really needed our own predicate when the meaning of system_supports_32bit_el0() changed in commit 2122a833316f ("arm64: Allow mismatched 32-bit EL0 support").

Lightly tested as I do not have any asymmetric systems on hand at the moment. Attention on patch 2 would be appreciated as it affects ABI.

Applies to 6.0-rc1.

v1 -> v2:
- Fix a silly logic inversion in vcpu_mode_is_bad_32bit()
- Rebase to 6.0-rc1

Oliver Upton (2):
KVM: arm64: Treat PMCR_EL1.LC as RES1 on asymmetric systems
KVM: arm64: Reject 32bit user PSTATE on asymmetric systems

 arch/arm64/include/asm/kvm_host.h | 4 ++++
 arch/arm64/kvm/arm.c              | 3 +--
 arch/arm64/kvm/guest.c            | 2 +-
 arch/arm64/kvm/sys_regs.c         | 4 ++--
 4 files changed, 8 insertions(+), 5 deletions(-)

base-commit: 568035b01cfb107af8d2e4bd2fb9aea22cf5b868

Signature · 1 line
2.37.1.595.g718a3a8f04-goog
#1 lore raw

PATCH v2 1/2 KVM: arm64: Treat PMCR_EL1.LC as RES1 on asymmetric systems

KVM does not support AArch32 on asymmetric systems. To that end, enforce AArch64-only behavior on PMCR_EL1.LC when on an asymmetric system.

Fixes 2122a833316f ("arm64: Allow mismatched 32-bit EL0 support")
---
 arch/arm64/include/asm/kvm_host.h | 4 ++++
 arch/arm64/kvm/arm.c              | 3 +--
 arch/arm64/kvm/sys_regs.c         | 4 ++--
 3 files changed, 7 insertions(+), 4 deletions(-)
arch/arm64/include/asm/kvm_host.h +4 -0
index f38ef299f13b..e9c9388ccc02 100644
@@ -929,6 +929,10 @@ bool kvm_arm_vcpu_is_finalized(struct kvm_vcpu *vcpu);
 	(system_supports_mte() &&				\
 	 test_bit(KVM_ARCH_FLAG_MTE_ENABLED, &(kvm)->arch.flags))
 
+#define kvm_supports_32bit_el0()				\
+	(system_supports_32bit_el0() &&				\
+	 !static_branch_unlikely(&arm64_mismatched_32bit_el0))
+
 int kvm_trng_call(struct kvm_vcpu *vcpu);
 #ifdef CONFIG_KVM
 extern phys_addr_t hyp_mem_base;
arch/arm64/kvm/arm.c +1 -2
index 986cee6fbc7f..2ff0ef62abad 100644
@@ -757,8 +757,7 @@ static bool vcpu_mode_is_bad_32bit(struct kvm_vcpu *vcpu)
 	if (likely(!vcpu_mode_is_32bit(vcpu)))
 		return false;
 
-	return !system_supports_32bit_el0() ||
-		static_branch_unlikely(&arm64_mismatched_32bit_el0);
+	return !kvm_supports_32bit_el0();
 }
 
 /**
arch/arm64/kvm/sys_regs.c +2 -2
index c059b259aea6..3234f50b8c4b 100644
@@ -652,7 +652,7 @@ static void reset_pmcr(struct kvm_vcpu *vcpu, const struct sys_reg_desc *r)
 	 */
 	val = ((pmcr & ~ARMV8_PMU_PMCR_MASK)
 	       | (ARMV8_PMU_PMCR_MASK & 0xdecafbad)) & (~ARMV8_PMU_PMCR_E);
-	if (!system_supports_32bit_el0())
+	if (!kvm_supports_32bit_el0())
 		val |= ARMV8_PMU_PMCR_LC;
 	__vcpu_sys_reg(vcpu, r->reg) = val;
 }
@@ -701,7 +701,7 @@ static bool access_pmcr(struct kvm_vcpu *vcpu, struct sys_reg_params *p,
 		val = __vcpu_sys_reg(vcpu, PMCR_EL0);
 		val &= ~ARMV8_PMU_PMCR_MASK;
 		val |= p->regval & ARMV8_PMU_PMCR_MASK;
-		if (!system_supports_32bit_el0())
+		if (!kvm_supports_32bit_el0())
 			val |= ARMV8_PMU_PMCR_LC;
 		__vcpu_sys_reg(vcpu, PMCR_EL0) = val;
 		kvm_pmu_handle_pmcr(vcpu, val);
Signature · 1 line
2.37.1.595.g718a3a8f04-goog
#2 lore raw

PATCH v2 2/2 KVM: arm64: Reject 32bit user PSTATE on asymmetric systems

In reply to Oliver Upton

KVM does not support AArch32 EL0 on asymmetric systems. To that end, prevent userspace from configuring a vCPU in such a state through setting PSTATE.

It is already ABI that KVM rejects such a write on a system where AArch32 EL0 is unsupported. Though the kernel's definition of a 32bit system changed in commit 2122a833316f ("arm64: Allow mismatched 32-bit EL0 support"), KVM's did not.

Fixes 2122a833316f ("arm64: Allow mismatched 32-bit EL0 support")
---
 arch/arm64/kvm/guest.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/arm64/kvm/guest.c +1 -1
index 8c607199cad1..f802a3b3f8db 100644
@@ -242,7 +242,7 @@ static int set_core_reg(struct kvm_vcpu *vcpu, const struct kvm_one_reg *reg)
 		u64 mode = (*(u64 *)valp) & PSR_AA32_MODE_MASK;
 		switch (mode) {
 		case PSR_AA32_MODE_USR:
-			if (!system_supports_32bit_el0())
+			if (!kvm_supports_32bit_el0())
 				return -EINVAL;
 			break;
 		case PSR_AA32_MODE_FIQ:
Signature · 1 line
2.37.1.595.g718a3a8f04-goog
maintainer Applied
#3 lore raw
In reply to Oliver Upton
Quoted 10 lines · Oliver Upton

Small series to fix a couple issues around when 64bit-only behavior is applied. As KVM is more restrictive than the kernel in terms of 32bit support (no asymmetry), we really needed our own predicate when the meaning of system_supports_32bit_el0() changed in commit 2122a833316f ("arm64: Allow mismatched 32-bit EL0 support").

Lightly tested as I do not have any asymmetric systems on hand at the moment. Attention on patch 2 would be appreciated as it affects ABI.

[...]

Applied to fixes, thanks!

[1/2] KVM: arm64: Treat PMCR_EL1.LC as RES1 on asymmetric systems

      commit: f3c6efc72f3b20ec23566e768979802f0a398f04
[2/2] KVM: arm64: Reject 32bit user PSTATE on asymmetric systems
      commit: b10d86fb8e46cc812171728bcd326df2f34e9ed5

Cheers,

	M.
Signature · 1 line
Marc Zyngier <[email protected]>
maintainer
#4 lore raw
In reply to Oliver Upton
Quoted 9 lines · Oliver Upton

Small series to fix a couple issues around when 64bit-only behavior is applied. As KVM is more restrictive than the kernel in terms of 32bit support (no asymmetry), we really needed our own predicate when the meaning of system_supports_32bit_el0() changed in commit 2122a833316f ("arm64: Allow mismatched 32-bit EL0 support").

Lightly tested as I do not have any asymmetric systems on hand at the moment. Attention on patch 2 would be appreciated as it affects ABI.

I don't think this significantly affect the ABI, as it is pretty unlikely that you'd have been able to execute the result, at least on VM creation (set PSTATE.M=USR, start executing, get the page fault on the first instruction... bang).

You could have tricked it in other ways, but at the end of the day you're running a broken hypervisor on an even more broken system...

Anyway, I've applied this to fixes.

Thanks,

	M.
Signature · 5 lines
Without deviation from the norm, progress is not possible.
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
#5 lore raw

Hi,

Quoted 12 lines · Marc Zyngier
Quoted 9 lines · Oliver Upton

Small series to fix a couple issues around when 64bit-only behavior is applied. As KVM is more restrictive than the kernel in terms of 32bit support (no asymmetry), we really needed our own predicate when the meaning of system_supports_32bit_el0() changed in commit 2122a833316f ("arm64: Allow mismatched 32-bit EL0 support").

Lightly tested as I do not have any asymmetric systems on hand at the moment. Attention on patch 2 would be appreciated as it affects ABI.

Quoted 7 lines

I don't think this significantly affect the ABI, as it is pretty unlikely that you'd have been able to execute the result, at least on VM creation (set PSTATE.M=USR, start executing, get the page fault on the first instruction... bang).

You could have tricked it in other ways, but at the end of the day you're running a broken hypervisor on an even more broken system...

Just FYI, you can create such a system on models, by running two clusters and setting clusterX.max_32bit_el=-1. Or you can have even crazier configurations, where AArch32 support is present on only one cluster, and only for EL0.

Thanks,
Alex

Quoted 9 lines

Anyway, I've applied this to fixes.

Thanks,

	M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
maintainer
#6 lore raw
Quoted 21 lines · Alexandru Elisei

Hi,

Quoted 17 lines · Marc Zyngier
Quoted 9 lines · Oliver Upton

Small series to fix a couple issues around when 64bit-only behavior is applied. As KVM is more restrictive than the kernel in terms of 32bit support (no asymmetry), we really needed our own predicate when the meaning of system_supports_32bit_el0() changed in commit 2122a833316f ("arm64: Allow mismatched 32-bit EL0 support").

Lightly tested as I do not have any asymmetric systems on hand at the moment. Attention on patch 2 would be appreciated as it affects ABI.

I don't think this significantly affect the ABI, as it is pretty unlikely that you'd have been able to execute the result, at least on VM creation (set PSTATE.M=USR, start executing, get the page fault on the first instruction... bang).

Quoted 7 lines
Quoted 2 lines

You could have tricked it in other ways, but at the end of the day you're running a broken hypervisor on an even more broken system...

Just FYI, you can create such a system on models, by running two clusters and setting clusterX.max_32bit_el=-1. Or you can have even crazier configurations, where AArch32 support is present on only one cluster, and only for EL0.

You mean, just as crazy as some of the systems out there? :D

Thanks,

	M.
Signature · 5 lines
Without deviation from the norm, progress is not possible.
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
#7 lore raw
Quoted 18 lines · Marc Zyngier
Quoted 16 lines · Alexandru Elisei

Hi,

Quoted 12 lines · Marc Zyngier
Quoted 9 lines · Oliver Upton

Small series to fix a couple issues around when 64bit-only behavior is applied. As KVM is more restrictive than the kernel in terms of 32bit support (no asymmetry), we really needed our own predicate when the meaning of system_supports_32bit_el0() changed in commit 2122a833316f ("arm64: Allow mismatched 32-bit EL0 support").

Lightly tested as I do not have any asymmetric systems on hand at the moment. Attention on patch 2 would be appreciated as it affects ABI.

Quoted 7 lines
Quoted 7 lines
Quoted 7 lines

I don't think this significantly affect the ABI, as it is pretty unlikely that you'd have been able to execute the result, at least on VM creation (set PSTATE.M=USR, start executing, get the page fault on the first instruction... bang).

You could have tricked it in other ways, but at the end of the day you're running a broken hypervisor on an even more broken system...

Lol, fair enough. Just wanted to make sure we're all happy with how we turn the guest into rubble on the other end :)

Quoted 4 lines
Quoted 4 lines

Just FYI, you can create such a system on models, by running two clusters and setting clusterX.max_32bit_el=-1. Or you can have even crazier configurations, where AArch32 support is present on only one cluster, and only for EL0.

Doh! Forgot about the fast model.

--
Thanks,
Oliver
_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm