Hi Wei Lin,
On Wed, Mar 25, 2026 at 12:36:20AM +0000, Wei-Lin Chang wrote:
> Introduce library functions for setting up guest stage-2 page tables,
> then use that to give L2 an identity mapped stage-2 and enable it.
>
> The translation and stage-2 page table built is simple, start level 0,
> 4 levels, 4KB granules, normal cachable, 48-bit IA, 40-bit OA.
>
> The nested page table code is adapted from lib/x86/vmx.c.
>
> Signed-off-by: Wei-Lin Chang <[email protected]>
> ---
> .../selftests/kvm/include/arm64/nested.h | 7 ++
> .../selftests/kvm/include/arm64/processor.h | 9 ++
> .../testing/selftests/kvm/lib/arm64/nested.c | 97 ++++++++++++++++++-
> 3 files changed, 111 insertions(+), 2 deletions(-)
>
> diff --git a/tools/testing/selftests/kvm/include/arm64/nested.h
> b/tools/testing/selftests/kvm/include/arm64/nested.h
> index 739ff2ee0161..0be10a775e48 100644
> --- a/tools/testing/selftests/kvm/include/arm64/nested.h
> +++ b/tools/testing/selftests/kvm/include/arm64/nested.h
> @@ -6,6 +6,13 @@
> #ifndef SELFTEST_KVM_NESTED_H
> #define SELFTEST_KVM_NESTED_H
>
> +uint64_t get_l1_vtcr(void);
Using a type u64 is simpler? And I think you configure guest
hypervisor's stage 2 translation table, I felt this gives us
an impression somewhere the configuration IA and OA sizes etc
are stored.
> +
> +void nested_map(struct kvm_vm *vm, vm_paddr_t guest_pgd,
> + uint64_t nested_paddr, uint64_t paddr, uint64_t size);
> +void nested_map_memslot(struct kvm_vm *vm, vm_paddr_t guest_pgd,
> + uint32_t memslot);
> +
> void prepare_l2_stack(struct kvm_vm *vm, struct kvm_vcpu *vcpu);
> void prepare_hyp_state(struct kvm_vm *vm, struct kvm_vcpu *vcpu);
> void prepare_eret_destination(struct kvm_vm *vm, struct kvm_vcpu *vcpu, void
> *l2_pc);
> diff --git a/tools/testing/selftests/kvm/include/arm64/processor.h
> b/tools/testing/selftests/kvm/include/arm64/processor.h
> index ac97a1c436fc..5de2e932d95a 100644
> --- a/tools/testing/selftests/kvm/include/arm64/processor.h
> +++ b/tools/testing/selftests/kvm/include/arm64/processor.h
> @@ -104,6 +104,15 @@
> #define TCR_HA (UL(1) << 39)
> #define TCR_DS (UL(1) << 59)
>
> +/* VTCR_EL2 specific flags */
> +#define VTCR_EL2_T0SZ_BITS(x) ((UL(64) - (x)) << VTCR_EL2_T0SZ_SHIFT)
> +
> +#define VTCR_EL2_SL0_LV0_4K (UL(2) << VTCR_EL2_SL0_SHIFT)
> +#define VTCR_EL2_SL0_LV1_4K (UL(1) << VTCR_EL2_SL0_SHIFT)
> +#define VTCR_EL2_SL0_LV2_4K (UL(0) << VTCR_EL2_SL0_SHIFT)
> +
> +#define VTCR_EL2_PS_40_BITS (UL(2) << VTCR_EL2_PS_SHIFT)
> +
> /*
> * AttrIndx[2:0] encoding (mapping attributes defined in the MAIR*
> registers).
> */
> diff --git a/tools/testing/selftests/kvm/lib/arm64/nested.c
> b/tools/testing/selftests/kvm/lib/arm64/nested.c
> index 111d02f44cfe..910f8cd30f96 100644
> --- a/tools/testing/selftests/kvm/lib/arm64/nested.c
> +++ b/tools/testing/selftests/kvm/lib/arm64/nested.c
> @@ -1,8 +1,11 @@
> // SPDX-License-Identifier: GPL-2.0
> /*
> - * ARM64 Nested virtualization helpers
> + * ARM64 Nested virtualization helpers, nested page table code adapted from
> + * ../x86/vmx.c.
> */
>
> +#include <linux/sizes.h>
> +
> #include "kvm_util.h"
> #include "nested.h"
> #include "processor.h"
> @@ -18,6 +21,87 @@ static void hvc_handler(struct ex_regs *regs)
> regs->pc = (u64)after_hvc;
> }
>
> +uint64_t get_l1_vtcr(void)
> +{
> + return VTCR_EL2_PS_40_BITS | VTCR_EL2_TG0_4K | VTCR_EL2_ORGN0_WBWA |
> + VTCR_EL2_IRGN0_WBWA | VTCR_EL2_SL0_LV0_4K |
> VTCR_EL2_T0SZ_BITS(48);
> +}
> +
> +static void __nested_pg_map(struct kvm_vm *vm, uint64_t guest_pgd,
> + uint64_t nested_paddr, uint64_t paddr, uint64_t flags)
> +{
> + uint8_t attr_idx = flags & (PTE_ATTRINDX_MASK >> PTE_ATTRINDX_SHIFT);
> + uint64_t pg_attr;
> + uint64_t *ptep;
> +
> + TEST_ASSERT((nested_paddr % vm->page_size) == 0,
> + "L2 IPA not on page boundary,\n"
> + " nested_paddr: 0x%lx vm->page_size: 0x%x", nested_paddr,
> vm->page_size);
> + TEST_ASSERT((paddr % vm->page_size) == 0,
> + "Guest physical address not on page boundary,\n"
> + " paddr: 0x%lx vm->page_size: 0x%x", paddr, vm->page_size);
> + TEST_ASSERT((paddr >> vm->page_shift) <= vm->max_gfn,
> + "Physical address beyond maximum supported,\n"
> + " paddr: 0x%lx vm->max_gfn: 0x%lx vm->page_size: 0x%x",
> + paddr, vm->max_gfn, vm->page_size);
> +
> + ptep = addr_gpa2hva(vm, guest_pgd) + ((nested_paddr >> 39) & 0x1ffu) *
> 8;
> + if (!*ptep)
> + *ptep = (vm_alloc_page_table(vm) & GENMASK(47, 12)) |
> PGD_TYPE_TABLE | PTE_VALID;
Same but given this is stage 2 translation tables, KVM_PTE_VALID?
Thanks,
Itaru.
> + ptep = addr_gpa2hva(vm, *ptep & GENMASK(47, 12)) + ((nested_paddr >>
> 30) & 0x1ffu) * 8;
> + if (!*ptep)
> + *ptep = (vm_alloc_page_table(vm) & GENMASK(47, 12)) |
> PUD_TYPE_TABLE | PTE_VALID;
> + ptep = addr_gpa2hva(vm, *ptep & GENMASK(47, 12)) + ((nested_paddr >>
> 21) & 0x1ffu) * 8;
> + if (!*ptep)
> + *ptep = (vm_alloc_page_table(vm) & GENMASK(47, 12)) |
> PMD_TYPE_TABLE | PTE_VALID;
> + ptep = addr_gpa2hva(vm, *ptep & GENMASK(47, 12)) + ((nested_paddr >>
> 12) & 0x1ffu) * 8;
> +
> + pg_attr = PTE_AF | PTE_ATTRINDX(attr_idx) | PTE_TYPE_PAGE | PTE_VALID;
> + pg_attr |= PTE_SHARED;
> +
> + *ptep = (paddr & GENMASK(47, 12)) | pg_attr;
> +}
> +
> +void nested_map(struct kvm_vm *vm, vm_paddr_t guest_pgd,
> + uint64_t nested_paddr, uint64_t paddr, uint64_t size)
> +{
> + size_t npages = size / SZ_4K;
> +
> + TEST_ASSERT(nested_paddr + size > nested_paddr, "Vaddr overflow");
> + TEST_ASSERT(paddr + size > paddr, "Paddr overflow");
> +
> + while (npages--) {
> + __nested_pg_map(vm, guest_pgd, nested_paddr, paddr, MT_NORMAL);
> + nested_paddr += SZ_4K;
> + paddr += SZ_4K;
> + }
> +}
> +
> +/*
> + * Prepare an identity shadow page table that maps all the
> + * physical pages in VM.
> + */
> +void nested_map_memslot(struct kvm_vm *vm, vm_paddr_t guest_pgd,
> + uint32_t memslot)
> +{
> + sparsebit_idx_t i, last;
> + struct userspace_mem_region *region =
> + memslot2region(vm, memslot);
> +
> + i = (region->region.guest_phys_addr >> vm->page_shift) - 1;
> + last = i + (region->region.memory_size >> vm->page_shift);
> + for (;;) {
> + i = sparsebit_next_clear(region->unused_phy_pages, i);
> + if (i > last)
> + break;
> +
> + nested_map(vm, guest_pgd,
> + (uint64_t)i << vm->page_shift,
> + (uint64_t)i << vm->page_shift,
> + 1 << vm->page_shift);
> + }
> +}
> +
> void prepare_l2_stack(struct kvm_vm *vm, struct kvm_vcpu *vcpu)
> {
> size_t l2_stack_size;
> @@ -32,7 +116,16 @@ void prepare_l2_stack(struct kvm_vm *vm, struct kvm_vcpu
> *vcpu)
>
> void prepare_hyp_state(struct kvm_vm *vm, struct kvm_vcpu *vcpu)
> {
> - vcpu_set_reg(vcpu, KVM_ARM64_SYS_REG(SYS_HCR_EL2), HCR_EL2_RW);
> + vm_paddr_t guest_pgd;
> +
> + guest_pgd = vm_phy_pages_alloc(vm, 1,
> + KVM_GUEST_PAGE_TABLE_MIN_PADDR,
> + vm->memslots[MEM_REGION_PT]);
> + nested_map_memslot(vm, guest_pgd, 0);
> +
> + vcpu_set_reg(vcpu, KVM_ARM64_SYS_REG(SYS_HCR_EL2), HCR_EL2_RW |
> HCR_EL2_VM);
> + vcpu_set_reg(vcpu, KVM_ARM64_SYS_REG(SYS_VTTBR_EL2), guest_pgd);
> + vcpu_set_reg(vcpu, KVM_ARM64_SYS_REG(SYS_VTCR_EL2), get_l1_vtcr());
> }
>
> void prepare_eret_destination(struct kvm_vm *vm, struct kvm_vcpu *vcpu, void
> *l2_pc)
> --
> 2.43.0
>
