KVM: arm64: Propagate and handle Fine-Grained UNDEF bits

In order to correctly honor our FGU bits, they must be converted
into a set of FGT bits. They get merged as part of the existing
FGT setting.

Similarly, the UNDEF injection phase takes place when handling
the trap.

This results in a bit of rework in the FGT macros in order to
help with the code generation, as burying per-CPU accesses in
macros results in a lot of expansion, not to mention the vcpu->kvm
access on nvhe (kern_hyp_va() is not optimisation-friendly).
Reviewed-by: Joey Gouly <joey.gouly@arm.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20240214131827.2856277-19-maz@kernel.orgSigned-off-by: Oliver Upton <oliver.upton@linux.dev>

arch/arm64/kvm/emulate-nested.c

@@ -2017,6 +2017,17 @@ bool triage_sysreg_trap(struct kvm_vcpu *vcpu, int *sr_index)
 	if (!tc.val)
 		goto local;
 
+	/*
+	 * If a sysreg can be trapped using a FGT, first check whether we
+	 * trap for the purpose of forbidding the feature. In that case,
+	 * inject an UNDEF.
+	 */
+	if (tc.fgt != __NO_FGT_GROUP__ &&
+	    (vcpu->kvm->arch.fgu[tc.fgt] & BIT(tc.bit))) {
+		kvm_inject_undefined(vcpu);
+		return true;
+	}
+
 	/*
 	 * If we're not nesting, immediately return to the caller, with the
 	 * sysreg index, should we have it.

arch/arm64/kvm/hyp/include/hyp/switch.h

@@ -79,14 +79,48 @@ static inline void __activate_traps_fpsimd32(struct kvm_vcpu *vcpu)
 		clr |= ~hfg & __ ## reg ## _nMASK; 			\
 	} while(0)
 
-#define update_fgt_traps_cs(vcpu, reg, clr, set)			\
+#define reg_to_fgt_group_id(reg)					\
+	({								\
+		enum fgt_group_id id;					\
+		switch(reg) {						\
+		case HFGRTR_EL2:					\
+		case HFGWTR_EL2:					\
+			id = HFGxTR_GROUP;				\
+			break;						\
+		case HFGITR_EL2:					\
+			id = HFGITR_GROUP;				\
+			break;						\
+		case HDFGRTR_EL2:					\
+		case HDFGWTR_EL2:					\
+			id = HDFGRTR_GROUP;				\
+			break;						\
+		case HAFGRTR_EL2:					\
+			id = HAFGRTR_GROUP;				\
+			break;						\
+		default:						\
+			BUILD_BUG_ON(1);				\
+		}							\
+									\
+		id;							\
+	})
+
+#define compute_undef_clr_set(vcpu, kvm, reg, clr, set)			\
+	do {								\
+		u64 hfg = kvm->arch.fgu[reg_to_fgt_group_id(reg)];	\
+		set |= hfg & __ ## reg ## _MASK;			\
+		clr |= hfg & __ ## reg ## _nMASK; 			\
+	} while(0)
+
+#define update_fgt_traps_cs(hctxt, vcpu, kvm, reg, clr, set)		\
 	do {								\
-		struct kvm_cpu_context *hctxt =				\
-			&this_cpu_ptr(&kvm_host_data)->host_ctxt;	\
 		u64 c = 0, s = 0;					\
 									\
 		ctxt_sys_reg(hctxt, reg) = read_sysreg_s(SYS_ ## reg);	\
-		compute_clr_set(vcpu, reg, c, s);			\
+		if (vcpu_has_nv(vcpu) && !is_hyp_ctxt(vcpu))		\
+			compute_clr_set(vcpu, reg, c, s);		\
+									\
+		compute_undef_clr_set(vcpu, kvm, reg, c, s);		\
+									\
 		s |= set;						\
 		c |= clr;						\
 		if (c || s) {						\
@@ -97,8 +131,8 @@ static inline void __activate_traps_fpsimd32(struct kvm_vcpu *vcpu)
 		}							\
 	} while(0)
 
-#define update_fgt_traps(vcpu, reg)		\
-	update_fgt_traps_cs(vcpu, reg, 0, 0)
+#define update_fgt_traps(hctxt, vcpu, kvm, reg)		\
+	update_fgt_traps_cs(hctxt, vcpu, kvm, reg, 0, 0)
 
 /*
  * Validate the fine grain trap masks.
@@ -122,6 +156,7 @@ static inline bool cpu_has_amu(void)
 static inline void __activate_traps_hfgxtr(struct kvm_vcpu *vcpu)
 {
 	struct kvm_cpu_context *hctxt = &this_cpu_ptr(&kvm_host_data)->host_ctxt;
+	struct kvm *kvm = kern_hyp_va(vcpu->kvm);
 	u64 r_clr = 0, w_clr = 0, r_set = 0, w_set = 0, tmp;
 	u64 r_val, w_val;
 
@@ -157,6 +192,9 @@ static inline void __activate_traps_hfgxtr(struct kvm_vcpu *vcpu)
 		compute_clr_set(vcpu, HFGWTR_EL2, w_clr, w_set);
 	}
 
+	compute_undef_clr_set(vcpu, kvm, HFGRTR_EL2, r_clr, r_set);
+	compute_undef_clr_set(vcpu, kvm, HFGWTR_EL2, w_clr, w_set);
+
 	/* The default to trap everything not handled or supported in KVM. */
 	tmp = HFGxTR_EL2_nAMAIR2_EL1 | HFGxTR_EL2_nMAIR2_EL1 | HFGxTR_EL2_nS2POR_EL1 |
 	      HFGxTR_EL2_nPOR_EL1 | HFGxTR_EL2_nPOR_EL0 | HFGxTR_EL2_nACCDATA_EL1;
@@ -172,20 +210,26 @@ static inline void __activate_traps_hfgxtr(struct kvm_vcpu *vcpu)
 	write_sysreg_s(r_val, SYS_HFGRTR_EL2);
 	write_sysreg_s(w_val, SYS_HFGWTR_EL2);
 
-	if (!vcpu_has_nv(vcpu) || is_hyp_ctxt(vcpu))
-		return;
-
-	update_fgt_traps(vcpu, HFGITR_EL2);
-	update_fgt_traps(vcpu, HDFGRTR_EL2);
-	update_fgt_traps(vcpu, HDFGWTR_EL2);
+	update_fgt_traps(hctxt, vcpu, kvm, HFGITR_EL2);
+	update_fgt_traps(hctxt, vcpu, kvm, HDFGRTR_EL2);
+	update_fgt_traps(hctxt, vcpu, kvm, HDFGWTR_EL2);
 
 	if (cpu_has_amu())
-		update_fgt_traps(vcpu, HAFGRTR_EL2);
+		update_fgt_traps(hctxt, vcpu, kvm, HAFGRTR_EL2);
 }
 
+#define __deactivate_fgt(htcxt, vcpu, kvm, reg)				\
+	do {								\
+		if ((vcpu_has_nv(vcpu) && !is_hyp_ctxt(vcpu)) ||	\
+		    kvm->arch.fgu[reg_to_fgt_group_id(reg)])		\
+			write_sysreg_s(ctxt_sys_reg(hctxt, reg),	\
+				       SYS_ ## reg);			\
+	} while(0)
+
 static inline void __deactivate_traps_hfgxtr(struct kvm_vcpu *vcpu)
 {
 	struct kvm_cpu_context *hctxt = &this_cpu_ptr(&kvm_host_data)->host_ctxt;
+	struct kvm *kvm = kern_hyp_va(vcpu->kvm);
 
 	if (!cpus_have_final_cap(ARM64_HAS_FGT))
 		return;
@@ -193,15 +237,12 @@ static inline void __deactivate_traps_hfgxtr(struct kvm_vcpu *vcpu)
 	write_sysreg_s(ctxt_sys_reg(hctxt, HFGRTR_EL2), SYS_HFGRTR_EL2);
 	write_sysreg_s(ctxt_sys_reg(hctxt, HFGWTR_EL2), SYS_HFGWTR_EL2);
 
-	if (!vcpu_has_nv(vcpu) || is_hyp_ctxt(vcpu))
-		return;
-
-	write_sysreg_s(ctxt_sys_reg(hctxt, HFGITR_EL2), SYS_HFGITR_EL2);
-	write_sysreg_s(ctxt_sys_reg(hctxt, HDFGRTR_EL2), SYS_HDFGRTR_EL2);
-	write_sysreg_s(ctxt_sys_reg(hctxt, HDFGWTR_EL2), SYS_HDFGWTR_EL2);
+	__deactivate_fgt(hctxt, vcpu, kvm, HFGITR_EL2);
+	__deactivate_fgt(hctxt, vcpu, kvm, HDFGRTR_EL2);
+	__deactivate_fgt(hctxt, vcpu, kvm, HDFGWTR_EL2);
 
 	if (cpu_has_amu())
-		write_sysreg_s(ctxt_sys_reg(hctxt, HAFGRTR_EL2), SYS_HAFGRTR_EL2);
+		__deactivate_fgt(hctxt, vcpu, kvm, HAFGRTR_EL2);
 }
 
 static inline void __activate_traps_common(struct kvm_vcpu *vcpu)