KVM: arm64: nv: Support multiple nested Stage-2 mmu structures

Add Stage-2 mmu data structures for virtual EL2 and for nested guests.
We don't yet populate shadow Stage-2 page tables, but we now have a
framework for getting to a shadow Stage-2 pgd.

We allocate twice the number of vcpus as Stage-2 mmu structures because
that's sufficient for each vcpu running two translation regimes without
having to flush the Stage-2 page tables.

Co-developed-by: Christoffer Dall <christoffer.dall@arm.com>
Signed-off-by: Christoffer Dall <christoffer.dall@arm.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20240614144552.2773592-2-maz@kernel.org
Signed-off-by: Oliver Upton <oliver.upton@linux.dev>

arch/arm64/include/asm/kvm_host.h

@@ -189,6 +189,33 @@ struct kvm_s2_mmu {
 	uint64_t split_page_chunk_size;
 
 	struct kvm_arch *arch;
+
+	/*
+	 * For a shadow stage-2 MMU, the virtual vttbr used by the
+	 * host to parse the guest S2.
+	 * This either contains:
+	 * - the virtual VTTBR programmed by the guest hypervisor with
+         *   CnP cleared
+	 * - The value 1 (VMID=0, BADDR=0, CnP=1) if invalid
+	 *
+	 * We also cache the full VTCR which gets used for TLB invalidation,
+	 * taking the ARM ARM's "Any of the bits in VTCR_EL2 are permitted
+	 * to be cached in a TLB" to the letter.
+	 */
+	u64	tlb_vttbr;
+	u64	tlb_vtcr;
+
+	/*
+	 * true when this represents a nested context where virtual
+	 * HCR_EL2.VM == 1
+	 */
+	bool	nested_stage2_enabled;
+
+	/*
+	 *  0: Nobody is currently using this, check vttbr for validity
+	 * >0: Somebody is actively using this.
+	 */
+	atomic_t refcnt;
 };
 
 struct kvm_arch_memory_slot {
@@ -256,6 +283,14 @@ struct kvm_arch {
 	 */
 	u64 fgu[__NR_FGT_GROUP_IDS__];
 
+	/*
+	 * Stage 2 paging state for VMs with nested S2 using a virtual
+	 * VMID.
+	 */
+	struct kvm_s2_mmu *nested_mmus;
+	size_t nested_mmus_size;
+	int nested_mmus_next;
+
 	/* Interrupt controller */
 	struct vgic_dist	vgic;
 
@@ -1306,6 +1341,7 @@ void kvm_vcpu_load_vhe(struct kvm_vcpu *vcpu);
 void kvm_vcpu_put_vhe(struct kvm_vcpu *vcpu);
 
 int __init kvm_set_ipa_limit(void);
+u32 kvm_get_pa_bits(struct kvm *kvm);
 
 #define __KVM_HAVE_ARCH_VM_ALLOC
 struct kvm *kvm_arch_alloc_vm(void);

arch/arm64/include/asm/kvm_mmu.h

@@ -98,6 +98,7 @@ alternative_cb_end
 #include <asm/mmu_context.h>
 #include <asm/kvm_emulate.h>
 #include <asm/kvm_host.h>
+#include <asm/kvm_nested.h>
 
 void kvm_update_va_mask(struct alt_instr *alt,
 			__le32 *origptr, __le32 *updptr, int nr_inst);
@@ -165,6 +166,8 @@ int create_hyp_exec_mappings(phys_addr_t phys_addr, size_t size,
 int create_hyp_stack(phys_addr_t phys_addr, unsigned long *haddr);
 void __init free_hyp_pgds(void);
 
+void kvm_stage2_unmap_range(struct kvm_s2_mmu *mmu, phys_addr_t start, u64 size);
+
 void stage2_unmap_vm(struct kvm *kvm);
 int kvm_init_stage2_mmu(struct kvm *kvm, struct kvm_s2_mmu *mmu, unsigned long type);
 void kvm_uninit_stage2_mmu(struct kvm *kvm);
@@ -326,5 +329,26 @@ static inline struct kvm *kvm_s2_mmu_to_kvm(struct kvm_s2_mmu *mmu)
 {
 	return container_of(mmu->arch, struct kvm, arch);
 }
+
+static inline u64 get_vmid(u64 vttbr)
+{
+	return (vttbr & VTTBR_VMID_MASK(kvm_get_vmid_bits())) >>
+		VTTBR_VMID_SHIFT;
+}
+
+static inline bool kvm_s2_mmu_valid(struct kvm_s2_mmu *mmu)
+{
+	return !(mmu->tlb_vttbr & VTTBR_CNP_BIT);
+}
+
+static inline bool kvm_is_nested_s2_mmu(struct kvm *kvm, struct kvm_s2_mmu *mmu)
+{
+	/*
+	 * Be careful, mmu may not be fully initialised so do look at
+	 * *any* of its fields.
+	 */
+	return &kvm->arch.mmu != mmu;
+}
+
 #endif /* __ASSEMBLY__ */
 #endif /* __ARM64_KVM_MMU_H__ */

arch/arm64/include/asm/kvm_nested.h

@@ -61,6 +61,12 @@ static inline u64 translate_ttbr0_el2_to_ttbr0_el1(u64 ttbr0)
 }
 
 extern bool forward_smc_trap(struct kvm_vcpu *vcpu);
+extern void kvm_init_nested(struct kvm *kvm);
+extern int kvm_vcpu_init_nested(struct kvm_vcpu *vcpu);
+extern void kvm_init_nested_s2_mmu(struct kvm_s2_mmu *mmu);
+extern struct kvm_s2_mmu *lookup_s2_mmu(struct kvm_vcpu *vcpu);
+extern void kvm_vcpu_load_hw_mmu(struct kvm_vcpu *vcpu);
+extern void kvm_vcpu_put_hw_mmu(struct kvm_vcpu *vcpu);
 
 int kvm_init_nv_sysregs(struct kvm *kvm);
 

arch/arm64/kvm/arm.c

@@ -170,6 +170,8 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
 	mutex_unlock(&kvm->lock);
 #endif
 
+	kvm_init_nested(kvm);
+
 	ret = kvm_share_hyp(kvm, kvm + 1);
 	if (ret)
 		return ret;
@@ -551,6 +553,9 @@ void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
 	struct kvm_s2_mmu *mmu;
 	int *last_ran;
 
+	if (vcpu_has_nv(vcpu))
+		kvm_vcpu_load_hw_mmu(vcpu);
+
 	mmu = vcpu->arch.hw_mmu;
 	last_ran = this_cpu_ptr(mmu->last_vcpu_ran);
 
@@ -601,6 +606,8 @@ void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
 	kvm_timer_vcpu_put(vcpu);
 	kvm_vgic_put(vcpu);
 	kvm_vcpu_pmu_restore_host(vcpu);
+	if (vcpu_has_nv(vcpu))
+		kvm_vcpu_put_hw_mmu(vcpu);
 	kvm_arm_vmid_clear_active();
 
 	vcpu_clear_on_unsupported_cpu(vcpu);
@@ -1459,6 +1466,10 @@ static int kvm_setup_vcpu(struct kvm_vcpu *vcpu)
 	if (kvm_vcpu_has_pmu(vcpu) && !kvm->arch.arm_pmu)
 		ret = kvm_arm_set_default_pmu(kvm);
 
+	/* Prepare for nested if required */
+	if (!ret && vcpu_has_nv(vcpu))
+		ret = kvm_vcpu_init_nested(vcpu);
+
 	return ret;
 }
 

arch/arm64/kvm/mmu.c

@@ -328,7 +328,7 @@ static void __unmap_stage2_range(struct kvm_s2_mmu *mmu, phys_addr_t start, u64
 				   may_block));
 }
 
-static void unmap_stage2_range(struct kvm_s2_mmu *mmu, phys_addr_t start, u64 size)
+void kvm_stage2_unmap_range(struct kvm_s2_mmu *mmu, phys_addr_t start, u64 size)
 {
 	__unmap_stage2_range(mmu, start, size, true);
 }
@@ -855,21 +855,9 @@ static struct kvm_pgtable_mm_ops kvm_s2_mm_ops = {
 	.icache_inval_pou	= invalidate_icache_guest_page,
 };
 
-/**
- * kvm_init_stage2_mmu - Initialise a S2 MMU structure
- * @kvm:	The pointer to the KVM structure
- * @mmu:	The pointer to the s2 MMU structure
- * @type:	The machine type of the virtual machine
- *
- * Allocates only the stage-2 HW PGD level table(s).
- * Note we don't need locking here as this is only called when the VM is
- * created, which can only be done once.
- */
-int kvm_init_stage2_mmu(struct kvm *kvm, struct kvm_s2_mmu *mmu, unsigned long type)
+static int kvm_init_ipa_range(struct kvm_s2_mmu *mmu, unsigned long type)
 {
 	u32 kvm_ipa_limit = get_kvm_ipa_limit();
-	int cpu, err;
-	struct kvm_pgtable *pgt;
 	u64 mmfr0, mmfr1;
 	u32 phys_shift;
 
@@ -896,11 +884,51 @@ int kvm_init_stage2_mmu(struct kvm *kvm, struct kvm_s2_mmu *mmu, unsigned long t
 	mmfr1 = read_sanitised_ftr_reg(SYS_ID_AA64MMFR1_EL1);
 	mmu->vtcr = kvm_get_vtcr(mmfr0, mmfr1, phys_shift);
 
+	return 0;
+}
+
+/**
+ * kvm_init_stage2_mmu - Initialise a S2 MMU structure
+ * @kvm:	The pointer to the KVM structure
+ * @mmu:	The pointer to the s2 MMU structure
+ * @type:	The machine type of the virtual machine
+ *
+ * Allocates only the stage-2 HW PGD level table(s).
+ * Note we don't need locking here as this is only called in two cases:
+ *
+ * - when the VM is created, which can't race against anything
+ *
+ * - when secondary kvm_s2_mmu structures are initialised for NV
+ *   guests, and the caller must hold kvm->lock as this is called on a
+ *   per-vcpu basis.
+ */
+int kvm_init_stage2_mmu(struct kvm *kvm, struct kvm_s2_mmu *mmu, unsigned long type)
+{
+	int cpu, err;
+	struct kvm_pgtable *pgt;
+
+	/*
+	 * If we already have our page tables in place, and that the
+	 * MMU context is the canonical one, we have a bug somewhere,
+	 * as this is only supposed to ever happen once per VM.
+	 *
+	 * Otherwise, we're building nested page tables, and that's
+	 * probably because userspace called KVM_ARM_VCPU_INIT more
+	 * than once on the same vcpu. Since that's actually legal,
+	 * don't kick a fuss and leave gracefully.
+	 */
 	if (mmu->pgt != NULL) {
+		if (kvm_is_nested_s2_mmu(kvm, mmu))
+			return 0;
+
 		kvm_err("kvm_arch already initialized?\n");
 		return -EINVAL;
 	}
 
+	err = kvm_init_ipa_range(mmu, type);
+	if (err)
+		return err;
+
 	pgt = kzalloc(sizeof(*pgt), GFP_KERNEL_ACCOUNT);
 	if (!pgt)
 		return -ENOMEM;
@@ -925,6 +953,10 @@ int kvm_init_stage2_mmu(struct kvm *kvm, struct kvm_s2_mmu *mmu, unsigned long t
 
 	mmu->pgt = pgt;
 	mmu->pgd_phys = __pa(pgt->pgd);
+
+	if (kvm_is_nested_s2_mmu(kvm, mmu))
+		kvm_init_nested_s2_mmu(mmu);
+
 	return 0;
 
 out_destroy_pgtable:
@@ -976,7 +1008,7 @@ static void stage2_unmap_memslot(struct kvm *kvm,
 
 		if (!(vma->vm_flags & VM_PFNMAP)) {
 			gpa_t gpa = addr + (vm_start - memslot->userspace_addr);
-			unmap_stage2_range(&kvm->arch.mmu, gpa, vm_end - vm_start);
+			kvm_stage2_unmap_range(&kvm->arch.mmu, gpa, vm_end - vm_start);
 		}
 		hva = vm_end;
 	} while (hva < reg_end);
@@ -2022,19 +2054,14 @@ void kvm_arch_memslots_updated(struct kvm *kvm, u64 gen)
 {
 }
 
-void kvm_arch_flush_shadow_all(struct kvm *kvm)
-{
-	kvm_uninit_stage2_mmu(kvm);
-}
-
 void kvm_arch_flush_shadow_memslot(struct kvm *kvm,
 				   struct kvm_memory_slot *slot)
 {
 	gpa_t gpa = slot->base_gfn << PAGE_SHIFT;
 	phys_addr_t size = slot->npages << PAGE_SHIFT;
 
 	write_lock(&kvm->mmu_lock);
-	unmap_stage2_range(&kvm->arch.mmu, gpa, size);
+	kvm_stage2_unmap_range(&kvm->arch.mmu, gpa, size);
 	write_unlock(&kvm->mmu_lock);
 }