[PATCH v3 08/27] KVM: arm64: nv: Use guest hypervisor's vSError state

[Oliver Upton <oliver.upton@linux.dev>]

When HCR_EL2.AMO is set, physical SErrors are routed to EL2 and virtual
SError injection is enabled for EL1. Conceptually treating
host-initiated SErrors as 'physical', this means we can delegate control
of the vSError injection context to the guest hypervisor when nesting &&
AMO is set.

Reviewed-by: Marc Zyngier <maz@kernel.org>
Signed-off-by: Oliver Upton <oliver.upton@linux.dev>
---
 arch/arm64/include/asm/kvm_emulate.h       |  5 +++
 arch/arm64/include/asm/kvm_host.h          |  3 ++
 arch/arm64/kvm/hyp/include/hyp/switch.h    | 45 +++++++++++++++++++---
 arch/arm64/kvm/hyp/include/hyp/sysreg-sr.h | 31 +++++++++++++--
 4 files changed, 76 insertions(+), 8 deletions(-)

diff --git a/arch/arm64/include/asm/kvm_emulate.h b/arch/arm64/include/asm/kvm_emulate.h
index daa0410aaebf..1ff52e66514c 100644
--- a/arch/arm64/include/asm/kvm_emulate.h
+++ b/arch/arm64/include/asm/kvm_emulate.h
@@ -257,6 +257,11 @@ static inline bool is_nested_ctxt(struct kvm_vcpu *vcpu)
 	return vcpu_has_nv(vcpu) && !is_hyp_ctxt(vcpu);
 }
 
+static inline bool vserror_state_is_nested(struct kvm_vcpu *vcpu)
+{
+	return is_nested_ctxt(vcpu) && vcpu_el2_amo_is_set(vcpu);
+}
+
 /*
  * The layout of SPSR for an AArch32 state is different when observed from an
  * AArch64 SPSR_ELx or an AArch32 SPSR_*. This function generates the AArch32
diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h
index 468df4538371..20d63ece3138 100644
--- a/arch/arm64/include/asm/kvm_host.h
+++ b/arch/arm64/include/asm/kvm_host.h
@@ -1682,6 +1682,9 @@ void kvm_set_vm_id_reg(struct kvm *kvm, u32 reg, u64 val);
 #define kvm_has_s1poe(k)				\
 	(kvm_has_feat((k), ID_AA64MMFR3_EL1, S1POE, IMP))
 
+#define kvm_has_ras(k)					\
+	(kvm_has_feat((k), ID_AA64PFR0_EL1, RAS, IMP))
+
 static inline bool kvm_arch_has_irq_bypass(void)
 {
 	return true;
diff --git a/arch/arm64/kvm/hyp/include/hyp/switch.h b/arch/arm64/kvm/hyp/include/hyp/switch.h
index 8a854ab5f705..84ec4e100fbb 100644
--- a/arch/arm64/kvm/hyp/include/hyp/switch.h
+++ b/arch/arm64/kvm/hyp/include/hyp/switch.h
@@ -476,21 +476,56 @@ static inline void ___activate_traps(struct kvm_vcpu *vcpu, u64 hcr)
 
 	write_sysreg_hcr(hcr);
 
-	if (cpus_have_final_cap(ARM64_HAS_RAS_EXTN) && (hcr & HCR_VSE))
-		write_sysreg_s(vcpu->arch.vsesr_el2, SYS_VSESR_EL2);
+	if (cpus_have_final_cap(ARM64_HAS_RAS_EXTN) && (hcr & HCR_VSE)) {
+		u64 vsesr;
+
+		/*
+		 * When HCR_EL2.AMO is set, physical SErrors are taken to EL2
+		 * and vSError injection is enabled for EL1. Conveniently, for
+		 * NV this means that it is never the case where a 'physical'
+		 * SError (injected by KVM or userspace) and vSError are
+		 * deliverable to the same context.
+		 *
+		 * As such, we can trivially select between the host or guest's
+		 * VSESR_EL2. Except for the case that FEAT_RAS hasn't been
+		 * exposed to the guest, where ESR propagation in hardware
+		 * occurs unconditionally.
+		 *
+		 * Paper over the architectural wart and use an IMPLEMENTATION
+		 * DEFINED ESR value in case FEAT_RAS is hidden from the guest.
+		 */
+		if (!vserror_state_is_nested(vcpu))
+			vsesr = vcpu->arch.vsesr_el2;
+		else if (kvm_has_ras(kern_hyp_va(vcpu->kvm)))
+			vsesr = __vcpu_sys_reg(vcpu, VSESR_EL2);
+		else
+			vsesr = ESR_ELx_ISV;
+
+		write_sysreg_s(vsesr, SYS_VSESR_EL2);
+	}
 }
 
 static inline void ___deactivate_traps(struct kvm_vcpu *vcpu)
 {
+	u64 *hcr;
+
+	if (vserror_state_is_nested(vcpu))
+		hcr = __ctxt_sys_reg(&vcpu->arch.ctxt, HCR_EL2);
+	else
+		hcr = &vcpu->arch.hcr_el2;
+
 	/*
 	 * If we pended a virtual abort, preserve it until it gets
 	 * cleared. See D1.14.3 (Virtual Interrupts) for details, but
 	 * the crucial bit is "On taking a vSError interrupt,
 	 * HCR_EL2.VSE is cleared to 0."
+	 *
+	 * Additionally, when in a nested context we need to propagate the
+	 * updated state to the guest hypervisor's HCR_EL2.
 	 */
-	if (vcpu->arch.hcr_el2 & HCR_VSE) {
-		vcpu->arch.hcr_el2 &= ~HCR_VSE;
-		vcpu->arch.hcr_el2 |= read_sysreg(hcr_el2) & HCR_VSE;
+	if (*hcr & HCR_VSE) {
+		*hcr &= ~HCR_VSE;
+		*hcr |= read_sysreg(hcr_el2) & HCR_VSE;
 	}
 }
 
diff --git a/arch/arm64/kvm/hyp/include/hyp/sysreg-sr.h b/arch/arm64/kvm/hyp/include/hyp/sysreg-sr.h
index 4d0dbea4c56f..fd9e3ea84d9f 100644
--- a/arch/arm64/kvm/hyp/include/hyp/sysreg-sr.h
+++ b/arch/arm64/kvm/hyp/include/hyp/sysreg-sr.h
@@ -109,6 +109,17 @@ static inline bool ctxt_has_s1poe(struct kvm_cpu_context *ctxt)
 	return kvm_has_s1poe(kern_hyp_va(vcpu->kvm));
 }
 
+static inline bool ctxt_has_ras(struct kvm_cpu_context *ctxt)
+{
+	struct kvm_vcpu *vcpu;
+
+	if (!cpus_have_final_cap(ARM64_HAS_RAS_EXTN))
+		return false;
+
+	vcpu = ctxt_to_vcpu(ctxt);
+	return kvm_has_ras(kern_hyp_va(vcpu->kvm));
+}
+
 static inline void __sysreg_save_el1_state(struct kvm_cpu_context *ctxt)
 {
 	ctxt_sys_reg(ctxt, SCTLR_EL1)	= read_sysreg_el1(SYS_SCTLR);
@@ -159,8 +170,13 @@ static inline void __sysreg_save_el2_return_state(struct kvm_cpu_context *ctxt)
 	if (!has_vhe() && ctxt->__hyp_running_vcpu)
 		ctxt->regs.pstate	= read_sysreg_el2(SYS_SPSR);
 
-	if (cpus_have_final_cap(ARM64_HAS_RAS_EXTN))
+	if (!cpus_have_final_cap(ARM64_HAS_RAS_EXTN))
+		return;
+
+	if (!vserror_state_is_nested(ctxt_to_vcpu(ctxt)))
 		ctxt_sys_reg(ctxt, DISR_EL1) = read_sysreg_s(SYS_VDISR_EL2);
+	else if (ctxt_has_ras(ctxt))
+		ctxt_sys_reg(ctxt, VDISR_EL2) = read_sysreg_s(SYS_VDISR_EL2);
 }
 
 static inline void __sysreg_restore_common_state(struct kvm_cpu_context *ctxt)
@@ -293,8 +309,17 @@ static inline void __sysreg_restore_el2_return_state(struct kvm_cpu_context *ctx
 	write_sysreg_el2(ctxt->regs.pc,			SYS_ELR);
 	write_sysreg_el2(pstate,			SYS_SPSR);
 
-	if (cpus_have_final_cap(ARM64_HAS_RAS_EXTN))
-		write_sysreg_s(ctxt_sys_reg(ctxt, DISR_EL1), SYS_VDISR_EL2);
+	if (!cpus_have_final_cap(ARM64_HAS_RAS_EXTN))
+		return;
+
+	if (!vserror_state_is_nested(ctxt_to_vcpu(ctxt)))
+		vdisr = ctxt_sys_reg(ctxt, DISR_EL1);
+	else if (ctxt_has_ras(ctxt))
+		vdisr = ctxt_sys_reg(ctxt, VDISR_EL2);
+	else
+		vdisr = 0;
+
+	write_sysreg_s(vdisr, SYS_VDISR_EL2);
 }
 
 static inline void __sysreg32_save_state(struct kvm_vcpu *vcpu)
-- 
2.39.5