Merge branch kvm-arm64/nv-tcr2 into kvmarm/next

* kvm-arm64/nv-tcr2:
  : Fixes to the handling of TCR_EL1, courtesy of Marc Zyngier
  :
  : Series addresses a couple gaps that are present in KVM (from cover
  : letter):
  :
  :   - VM configuration: HCRX_EL2.TCR2En is forced to 1, and we blindly
  :     save/restore stuff.
  :
  :   - trap bit description and routing: none, obviously, since we make a
  :     point in not trapping.
  KVM: arm64: Honor trap routing for TCR2_EL1
  KVM: arm64: Make PIR{,E0}_EL1 save/restore conditional on FEAT_TCRX
  KVM: arm64: Make TCR2_EL1 save/restore dependent on the VM features
  KVM: arm64: Get rid of HCRX_GUEST_FLAGS
  KVM: arm64: Correctly honor the presence of FEAT_TCRX
Signed-off-by: Oliver Upton <oliver.upton@linux.dev>

arch/arm64/include/asm/kvm_arm.h

@@ -102,7 +102,6 @@
 #define HCR_HOST_NVHE_PROTECTED_FLAGS (HCR_HOST_NVHE_FLAGS | HCR_TSC)
 #define HCR_HOST_VHE_FLAGS (HCR_RW | HCR_TGE | HCR_E2H)
 
-#define HCRX_GUEST_FLAGS (HCRX_EL2_SMPME | HCRX_EL2_TCR2En)
 #define HCRX_HOST_FLAGS (HCRX_EL2_MSCEn | HCRX_EL2_TCR2En | HCRX_EL2_EnFPM)
 
 /* TCR_EL2 Registers bits */

arch/arm64/kvm/emulate-nested.c

@@ -83,6 +83,8 @@ enum cgt_group_id {
 	CGT_CPTR_TAM,
 	CGT_CPTR_TCPAC,
 
+	CGT_HCRX_TCR2En,
+
 	/*
 	 * Anything after this point is a combination of coarse trap
 	 * controls, which must all be evaluated to decide what to do.
@@ -93,6 +95,7 @@ enum cgt_group_id {
 	CGT_HCR_TTLB_TTLBIS,
 	CGT_HCR_TTLB_TTLBOS,
 	CGT_HCR_TVM_TRVM,
+	CGT_HCR_TVM_TRVM_HCRX_TCR2En,
 	CGT_HCR_TPU_TICAB,
 	CGT_HCR_TPU_TOCU,
 	CGT_HCR_NV1_nNV2_ENSCXT,
@@ -369,6 +372,12 @@ static const struct trap_bits coarse_trap_bits[] = {
 		.mask		= CPTR_EL2_TCPAC,
 		.behaviour	= BEHAVE_FORWARD_ANY,
 	},
+	[CGT_HCRX_TCR2En] = {
+		.index		= HCRX_EL2,
+		.value 		= 0,
+		.mask		= HCRX_EL2_TCR2En,
+		.behaviour	= BEHAVE_FORWARD_ANY,
+	},
 };
 
 #define MCB(id, ...)						\
@@ -383,6 +392,8 @@ static const enum cgt_group_id *coarse_control_combo[] = {
 	MCB(CGT_HCR_TTLB_TTLBIS,	CGT_HCR_TTLB, CGT_HCR_TTLBIS),
 	MCB(CGT_HCR_TTLB_TTLBOS,	CGT_HCR_TTLB, CGT_HCR_TTLBOS),
 	MCB(CGT_HCR_TVM_TRVM,		CGT_HCR_TVM, CGT_HCR_TRVM),
+	MCB(CGT_HCR_TVM_TRVM_HCRX_TCR2En,
+					CGT_HCR_TVM, CGT_HCR_TRVM, CGT_HCRX_TCR2En),
 	MCB(CGT_HCR_TPU_TICAB,		CGT_HCR_TPU, CGT_HCR_TICAB),
 	MCB(CGT_HCR_TPU_TOCU,		CGT_HCR_TPU, CGT_HCR_TOCU),
 	MCB(CGT_HCR_NV1_nNV2_ENSCXT,	CGT_HCR_NV1_nNV2, CGT_HCR_ENSCXT),
@@ -660,6 +671,7 @@ static const struct encoding_to_trap_config encoding_to_cgt[] __initconst = {
 	SR_TRAP(SYS_MAIR_EL1,		CGT_HCR_TVM_TRVM),
 	SR_TRAP(SYS_AMAIR_EL1,		CGT_HCR_TVM_TRVM),
 	SR_TRAP(SYS_CONTEXTIDR_EL1,	CGT_HCR_TVM_TRVM),
+	SR_TRAP(SYS_TCR2_EL1,		CGT_HCR_TVM_TRVM_HCRX_TCR2En),
 	SR_TRAP(SYS_DC_ZVA,		CGT_HCR_TDZ),
 	SR_TRAP(SYS_DC_GVA,		CGT_HCR_TDZ),
 	SR_TRAP(SYS_DC_GZVA,		CGT_HCR_TDZ),
@@ -1162,6 +1174,7 @@ static const struct encoding_to_trap_config encoding_to_fgt[] __initconst = {
 	SR_FGT(SYS_TPIDRRO_EL0,		HFGxTR, TPIDRRO_EL0, 1),
 	SR_FGT(SYS_TPIDR_EL1,		HFGxTR, TPIDR_EL1, 1),
 	SR_FGT(SYS_TCR_EL1,		HFGxTR, TCR_EL1, 1),
+	SR_FGT(SYS_TCR2_EL1,		HFGxTR, TCR_EL1, 1),
 	SR_FGT(SYS_SCXTNUM_EL0,		HFGxTR, SCXTNUM_EL0, 1),
 	SR_FGT(SYS_SCXTNUM_EL1, 	HFGxTR, SCXTNUM_EL1, 1),
 	SR_FGT(SYS_SCTLR_EL1, 		HFGxTR, SCTLR_EL1, 1),

arch/arm64/kvm/hyp/include/hyp/sysreg-sr.h

@@ -55,6 +55,17 @@ static inline bool ctxt_has_s1pie(struct kvm_cpu_context *ctxt)
 	return kvm_has_feat(kern_hyp_va(vcpu->kvm), ID_AA64MMFR3_EL1, S1PIE, IMP);
 }
 
+static inline bool ctxt_has_tcrx(struct kvm_cpu_context *ctxt)
+{
+	struct kvm_vcpu *vcpu;
+
+	if (!cpus_have_final_cap(ARM64_HAS_TCR2))
+		return false;
+
+	vcpu = ctxt_to_vcpu(ctxt);
+	return kvm_has_feat(kern_hyp_va(vcpu->kvm), ID_AA64MMFR3_EL1, TCRX, IMP);
+}
+
 static inline void __sysreg_save_el1_state(struct kvm_cpu_context *ctxt)
 {
 	ctxt_sys_reg(ctxt, SCTLR_EL1)	= read_sysreg_el1(SYS_SCTLR);
@@ -62,8 +73,14 @@ static inline void __sysreg_save_el1_state(struct kvm_cpu_context *ctxt)
 	ctxt_sys_reg(ctxt, TTBR0_EL1)	= read_sysreg_el1(SYS_TTBR0);
 	ctxt_sys_reg(ctxt, TTBR1_EL1)	= read_sysreg_el1(SYS_TTBR1);
 	ctxt_sys_reg(ctxt, TCR_EL1)	= read_sysreg_el1(SYS_TCR);
-	if (cpus_have_final_cap(ARM64_HAS_TCR2))
+	if (ctxt_has_tcrx(ctxt)) {
 		ctxt_sys_reg(ctxt, TCR2_EL1)	= read_sysreg_el1(SYS_TCR2);
+
+		if (ctxt_has_s1pie(ctxt)) {
+			ctxt_sys_reg(ctxt, PIR_EL1)	= read_sysreg_el1(SYS_PIR);
+			ctxt_sys_reg(ctxt, PIRE0_EL1)	= read_sysreg_el1(SYS_PIRE0);
+		}
+	}
 	ctxt_sys_reg(ctxt, ESR_EL1)	= read_sysreg_el1(SYS_ESR);
 	ctxt_sys_reg(ctxt, AFSR0_EL1)	= read_sysreg_el1(SYS_AFSR0);
 	ctxt_sys_reg(ctxt, AFSR1_EL1)	= read_sysreg_el1(SYS_AFSR1);
@@ -73,10 +90,6 @@ static inline void __sysreg_save_el1_state(struct kvm_cpu_context *ctxt)
 	ctxt_sys_reg(ctxt, CONTEXTIDR_EL1) = read_sysreg_el1(SYS_CONTEXTIDR);
 	ctxt_sys_reg(ctxt, AMAIR_EL1)	= read_sysreg_el1(SYS_AMAIR);
 	ctxt_sys_reg(ctxt, CNTKCTL_EL1)	= read_sysreg_el1(SYS_CNTKCTL);
-	if (ctxt_has_s1pie(ctxt)) {
-		ctxt_sys_reg(ctxt, PIR_EL1)	= read_sysreg_el1(SYS_PIR);
-		ctxt_sys_reg(ctxt, PIRE0_EL1)	= read_sysreg_el1(SYS_PIRE0);
-	}
 	ctxt_sys_reg(ctxt, PAR_EL1)	= read_sysreg_par();
 	ctxt_sys_reg(ctxt, TPIDR_EL1)	= read_sysreg(tpidr_el1);
 
@@ -138,8 +151,14 @@ static inline void __sysreg_restore_el1_state(struct kvm_cpu_context *ctxt)
 	write_sysreg_el1(ctxt_sys_reg(ctxt, CPACR_EL1),	SYS_CPACR);
 	write_sysreg_el1(ctxt_sys_reg(ctxt, TTBR0_EL1),	SYS_TTBR0);
 	write_sysreg_el1(ctxt_sys_reg(ctxt, TTBR1_EL1),	SYS_TTBR1);
-	if (cpus_have_final_cap(ARM64_HAS_TCR2))
+	if (ctxt_has_tcrx(ctxt)) {
 		write_sysreg_el1(ctxt_sys_reg(ctxt, TCR2_EL1),	SYS_TCR2);
+
+		if (ctxt_has_s1pie(ctxt)) {
+			write_sysreg_el1(ctxt_sys_reg(ctxt, PIR_EL1),	SYS_PIR);
+			write_sysreg_el1(ctxt_sys_reg(ctxt, PIRE0_EL1),	SYS_PIRE0);
+		}
+	}
 	write_sysreg_el1(ctxt_sys_reg(ctxt, ESR_EL1),	SYS_ESR);
 	write_sysreg_el1(ctxt_sys_reg(ctxt, AFSR0_EL1),	SYS_AFSR0);
 	write_sysreg_el1(ctxt_sys_reg(ctxt, AFSR1_EL1),	SYS_AFSR1);
@@ -149,10 +168,6 @@ static inline void __sysreg_restore_el1_state(struct kvm_cpu_context *ctxt)
 	write_sysreg_el1(ctxt_sys_reg(ctxt, CONTEXTIDR_EL1), SYS_CONTEXTIDR);
 	write_sysreg_el1(ctxt_sys_reg(ctxt, AMAIR_EL1),	SYS_AMAIR);
 	write_sysreg_el1(ctxt_sys_reg(ctxt, CNTKCTL_EL1), SYS_CNTKCTL);
-	if (ctxt_has_s1pie(ctxt)) {
-		write_sysreg_el1(ctxt_sys_reg(ctxt, PIR_EL1),	SYS_PIR);
-		write_sysreg_el1(ctxt_sys_reg(ctxt, PIRE0_EL1),	SYS_PIRE0);
-	}
 	write_sysreg(ctxt_sys_reg(ctxt, PAR_EL1),	par_el1);
 	write_sysreg(ctxt_sys_reg(ctxt, TPIDR_EL1),	tpidr_el1);
 

arch/arm64/kvm/sys_regs.c

@@ -384,6 +384,12 @@ static bool access_vm_reg(struct kvm_vcpu *vcpu,
 	bool was_enabled = vcpu_has_cache_enabled(vcpu);
 	u64 val, mask, shift;
 
+	if (reg_to_encoding(r) == SYS_TCR2_EL1 &&
+	    !kvm_has_feat(vcpu->kvm, ID_AA64MMFR3_EL1, TCRX, IMP)) {
+		kvm_inject_undefined(vcpu);
+		return false;
+	}
+
 	BUG_ON(!p->is_write);
 
 	get_access_mask(r, &mask, &shift);
@@ -4541,10 +4547,19 @@ void kvm_calculate_traps(struct kvm_vcpu *vcpu)
 	vcpu_set_hcr(vcpu);
 
 	if (cpus_have_final_cap(ARM64_HAS_HCX)) {
-		vcpu->arch.hcrx_el2 = HCRX_GUEST_FLAGS;
+		/*
+		 * In general, all HCRX_EL2 bits are gated by a feature.
+		 * The only reason we can set SMPME without checking any
+		 * feature is that its effects are not directly observable
+		 * from the guest.
+		 */
+		vcpu->arch.hcrx_el2 = HCRX_EL2_SMPME;
 
 		if (kvm_has_feat(kvm, ID_AA64ISAR2_EL1, MOPS, IMP))
 			vcpu->arch.hcrx_el2 |= (HCRX_EL2_MSCEn | HCRX_EL2_MCE2);
+
+		if (kvm_has_feat(kvm, ID_AA64MMFR3_EL1, TCRX, IMP))
+			vcpu->arch.hcrx_el2 |= HCRX_EL2_TCR2En;
 	}
 
 	if (test_bit(KVM_ARCH_FLAG_FGU_INITIALIZED, &kvm->arch.flags))