[PATCH v2 0/3] KVM: arm64: support MTE in protected VMs

[PATCH v2 0/3] KVM: arm64: support MTE in protected VMs

[Peter Collingbourne]

Hi,

This patch series contains a proposed extension to pKVM that allows MTE
to be exposed to the protected guests. It is based on the base pKVM
series previously sent to the list [1] and later rebased to 5.19-rc3
and uploaded to [2].

This series takes precautions against host compromise of the guests
via direct access to their tag storage, by preventing the host from
accessing the tag storage via stage 2 page tables. The device tree
must describe the physical memory address of the tag storage, if any,
and the memory nodes must declare that the tag storage location is
described. Otherwise, the MTE feature is disabled in protected guests.

Now that we can easily do so, we also prevent the host from accessing
any unmapped reserved-memory regions without a driver, as the host
has no business accessing that memory.

A proposed extension to the devicetree specification is available at
[3], a patched version of QEMU that produces the required device tree
nodes is available at [4] and a patched version of the crosvm hypervisor
that enables MTE is available at [5].

v2:
- refcount the PTEs owned by NOBODY

[1] https://lore.kernel.org/all/20220519134204.5379-1-will@kernel.org/
[2] https://android-kvm.googlesource.com/linux/ for-upstream/pkvm-base-v2
[3] https://github.com/pcc/devicetree-specification mte-alloc
[4] https://github.com/pcc/qemu mte-shared-alloc
[5] https://chromium-review.googlesource.com/c/chromiumos/platform/crosvm/+/3719324

Peter Collingbourne (3):
  KVM: arm64: add a hypercall for disowning pages
  KVM: arm64: disown unused reserved-memory regions
  KVM: arm64: allow MTE in protected VMs if the tag storage is known

 arch/arm64/include/asm/kvm_asm.h              |  1 +
 arch/arm64/include/asm/kvm_host.h             |  6 ++
 arch/arm64/include/asm/kvm_pkvm.h             |  4 +-
 arch/arm64/kernel/image-vars.h                |  3 +
 arch/arm64/kvm/arm.c                          | 83 ++++++++++++++++++-
 arch/arm64/kvm/hyp/include/nvhe/mem_protect.h |  1 +
 arch/arm64/kvm/hyp/include/nvhe/pkvm.h        |  1 +
 arch/arm64/kvm/hyp/nvhe/hyp-main.c            |  9 ++
 arch/arm64/kvm/hyp/nvhe/mem_protect.c         | 11 +++
 arch/arm64/kvm/hyp/nvhe/pkvm.c                |  8 +-
 arch/arm64/kvm/mmu.c                          |  4 +-
 11 files changed, 123 insertions(+), 8 deletions(-)

-- 
2.37.0.144.g8ac04bfd2-goog

[PATCH v2 1/3] KVM: arm64: add a hypercall for disowning pages

[Peter Collingbourne]

Currently we only deny the host access to hyp and guest pages. However,
there may be other pages that could potentially be used to indirectly
compromise the hypervisor or the other guests. Therefore introduce a
__pkvm_disown_pages hypercall that the host kernel may use to deny its
future self access to those pages before deprivileging itself.

Signed-off-by: Peter Collingbourne <pcc@google.com>
---
v2:
- refcount the PTEs owned by NOBODY

 arch/arm64/include/asm/kvm_asm.h              |  1 +
 arch/arm64/kvm/hyp/include/nvhe/mem_protect.h |  1 +
 arch/arm64/kvm/hyp/include/nvhe/pkvm.h        |  1 +
 arch/arm64/kvm/hyp/nvhe/hyp-main.c            |  9 +++++++++
 arch/arm64/kvm/hyp/nvhe/mem_protect.c         | 11 +++++++++++
 5 files changed, 23 insertions(+)

diff --git a/arch/arm64/include/asm/kvm_asm.h b/arch/arm64/include/asm/kvm_asm.h
index 411cfbe3ebbd..1a177d9ed517 100644
--- a/arch/arm64/include/asm/kvm_asm.h
+++ b/arch/arm64/include/asm/kvm_asm.h
@@ -63,6 +63,7 @@ enum __kvm_host_smccc_func {
 	__KVM_HOST_SMCCC_FUNC___kvm_tlb_flush_vmid_ipa,
 	__KVM_HOST_SMCCC_FUNC___kvm_tlb_flush_vmid,
 	__KVM_HOST_SMCCC_FUNC___kvm_flush_cpu_context,
+	__KVM_HOST_SMCCC_FUNC___pkvm_disown_pages,
 	__KVM_HOST_SMCCC_FUNC___pkvm_prot_finalize,
 
 	/* Hypercalls available after pKVM finalisation */
diff --git a/arch/arm64/kvm/hyp/include/nvhe/mem_protect.h b/arch/arm64/kvm/hyp/include/nvhe/mem_protect.h
index e0bbb1726fa3..e88a9dab9cd5 100644
--- a/arch/arm64/kvm/hyp/include/nvhe/mem_protect.h
+++ b/arch/arm64/kvm/hyp/include/nvhe/mem_protect.h
@@ -58,6 +58,7 @@ enum pkvm_component_id {
 	PKVM_ID_HOST,
 	PKVM_ID_HYP,
 	PKVM_ID_GUEST,
+	PKVM_ID_NOBODY,
 };
 
 extern unsigned long hyp_nr_cpus;
diff --git a/arch/arm64/kvm/hyp/include/nvhe/pkvm.h b/arch/arm64/kvm/hyp/include/nvhe/pkvm.h
index c1987115b217..fbd991a46ab3 100644
--- a/arch/arm64/kvm/hyp/include/nvhe/pkvm.h
+++ b/arch/arm64/kvm/hyp/include/nvhe/pkvm.h
@@ -98,6 +98,7 @@ int __pkvm_init_shadow(struct kvm *kvm,
 		       unsigned long pgd_hva,
 		       unsigned long last_ran_hva, size_t last_ran_size);
 int __pkvm_teardown_shadow(unsigned int shadow_handle);
+int __pkvm_disown_pages(phys_addr_t phys, size_t size);
 
 struct kvm_shadow_vcpu_state *
 pkvm_load_shadow_vcpu_state(unsigned int shadow_handle, unsigned int vcpu_idx);
diff --git a/arch/arm64/kvm/hyp/nvhe/hyp-main.c b/arch/arm64/kvm/hyp/nvhe/hyp-main.c
index e575224244e6..0dab343734e8 100644
--- a/arch/arm64/kvm/hyp/nvhe/hyp-main.c
+++ b/arch/arm64/kvm/hyp/nvhe/hyp-main.c
@@ -1031,6 +1031,14 @@ static void handle___pkvm_teardown_shadow(struct kvm_cpu_context *host_ctxt)
 	cpu_reg(host_ctxt, 1) = __pkvm_teardown_shadow(shadow_handle);
 }
 
+static void handle___pkvm_disown_pages(struct kvm_cpu_context *host_ctxt)
+{
+	DECLARE_REG(phys_addr_t, phys, host_ctxt, 1);
+	DECLARE_REG(size_t, size, host_ctxt, 2);
+
+	cpu_reg(host_ctxt, 1) = __pkvm_disown_pages(phys, size);
+}
+
 typedef void (*hcall_t)(struct kvm_cpu_context *);
 
 #define HANDLE_FUNC(x)	[__KVM_HOST_SMCCC_FUNC_##x] = (hcall_t)handle_##x
@@ -1048,6 +1056,7 @@ static const hcall_t host_hcall[] = {
 	HANDLE_FUNC(__kvm_tlb_flush_vmid_ipa),
 	HANDLE_FUNC(__kvm_tlb_flush_vmid),
 	HANDLE_FUNC(__kvm_flush_cpu_context),
+	HANDLE_FUNC(__pkvm_disown_pages),
 	HANDLE_FUNC(__pkvm_prot_finalize),
 
 	HANDLE_FUNC(__pkvm_host_share_hyp),
diff --git a/arch/arm64/kvm/hyp/nvhe/mem_protect.c b/arch/arm64/kvm/hyp/nvhe/mem_protect.c
index d839bb573b49..b3a2ad8454cc 100644
--- a/arch/arm64/kvm/hyp/nvhe/mem_protect.c
+++ b/arch/arm64/kvm/hyp/nvhe/mem_protect.c
@@ -1756,3 +1756,14 @@ int __pkvm_host_reclaim_page(u64 pfn)
 
 	return ret;
 }
+
+int __pkvm_disown_pages(phys_addr_t phys, size_t size)
+{
+	int ret;
+
+	host_lock_component();
+	ret = host_stage2_set_owner_locked(phys, size, PKVM_ID_NOBODY);
+	host_unlock_component();
+
+	return ret;
+}
-- 
2.37.0.144.g8ac04bfd2-goog

[PATCH v2 2/3] KVM: arm64: disown unused reserved-memory regions

[Peter Collingbourne]

The meaning of no-map on a reserved-memory node is as follows:

      Indicates the operating system must not create a virtual mapping
      of the region as part of its standard mapping of system memory,
      nor permit speculative access to it under any circumstances other
      than under the control of the device driver using the region.

If there is no compatible property, there is no device driver, so the
host kernel has no business accessing the reserved-memory region. Since
these regions may represent a route through which the host kernel
can gain additional privileges, disown any such memory regions before
deprivileging ourselves.

Signed-off-by: Peter Collingbourne <pcc@google.com>
---
 arch/arm64/kvm/arm.c | 46 ++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 46 insertions(+)

diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
index c1fc4ef82f93..91ca128e7daa 100644
--- a/arch/arm64/kvm/arm.c
+++ b/arch/arm64/kvm/arm.c
@@ -4,6 +4,7 @@
  * Author: Christoffer Dall <c.dall@virtualopensystems.com>
  */
 
+#include <linux/acpi.h>
 #include <linux/bug.h>
 #include <linux/cpu_pm.h>
 #include <linux/entry-kvm.h>
@@ -12,6 +13,7 @@
 #include <linux/kvm_host.h>
 #include <linux/list.h>
 #include <linux/module.h>
+#include <linux/of.h>
 #include <linux/vmalloc.h>
 #include <linux/fs.h>
 #include <linux/mman.h>
@@ -1913,6 +1915,48 @@ static bool init_psci_relay(void)
 	return true;
 }
 
+static void disown_reserved_memory(struct device_node *node)
+{
+	int addr_cells = of_n_addr_cells(node);
+	int size_cells = of_n_size_cells(node);
+	const __be32 *reg, *end;
+	int len;
+
+	reg = of_get_property(node, "reg", &len);
+	if (len % (4 * (addr_cells + size_cells)))
+		return;
+
+	end = reg + (len / 4);
+	while (reg != end) {
+		u64 addr, size;
+
+		addr = of_read_number(reg, addr_cells);
+		reg += addr_cells;
+		size = of_read_number(reg, size_cells);
+		reg += size_cells;
+
+		kvm_call_hyp_nvhe(__pkvm_disown_pages, addr, size);
+	}
+}
+
+static void kvm_reserved_memory_init(void)
+{
+	struct device_node *parent, *node;
+
+	if (!acpi_disabled || !is_protected_kvm_enabled())
+		return;
+
+	parent = of_find_node_by_path("/reserved-memory");
+	if (!parent)
+		return;
+
+	for_each_child_of_node(parent, node) {
+		if (!of_get_property(node, "compatible", NULL) &&
+		    of_get_property(node, "no-map", NULL))
+			disown_reserved_memory(node);
+	}
+}
+
 static int init_subsystems(void)
 {
 	int err = 0;
@@ -1953,6 +1997,8 @@ static int init_subsystems(void)
 
 	kvm_register_perf_callbacks(NULL);
 
+	kvm_reserved_memory_init();
+
 out:
 	if (err || !is_protected_kvm_enabled())
 		on_each_cpu(_kvm_arch_hardware_disable, NULL, 1);
-- 
2.37.0.144.g8ac04bfd2-goog

[PATCH v2 3/3] KVM: arm64: allow MTE in protected VMs if the tag storage is known

[Peter Collingbourne]

Because the host may corrupt a protected guest's tag storage unless
protected by stage 2 page tables, we can't expose MTE to protected guests
if the location of the tag storage is not known.

Therefore, only allow protected VM guests to use MTE if the location of
the tag storage is described in the device tree, and only after disowning
any physical memory accessible tag storage regions.

To avoid exposing MTE tags from the host to protected VMs, sanitize
tags before donating pages.

Signed-off-by: Peter Collingbourne <pcc@google.com>
---
 arch/arm64/include/asm/kvm_host.h |  6 +++++
 arch/arm64/include/asm/kvm_pkvm.h |  4 +++-
 arch/arm64/kernel/image-vars.h    |  3 +++
 arch/arm64/kvm/arm.c              | 37 ++++++++++++++++++++++++++++---
 arch/arm64/kvm/hyp/nvhe/pkvm.c    |  8 ++++---
 arch/arm64/kvm/mmu.c              |  4 +++-
 6 files changed, 54 insertions(+), 8 deletions(-)

diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h
index e54e76afccc0..35cba0408eca 100644
--- a/arch/arm64/include/asm/kvm_host.h
+++ b/arch/arm64/include/asm/kvm_host.h
@@ -1037,6 +1037,12 @@ bool kvm_arm_vcpu_is_finalized(struct kvm_vcpu *vcpu);
 
 #define kvm_arm_vcpu_sve_finalized(vcpu) vcpu_get_flag(vcpu, VCPU_SVE_FINALIZED)
 
+DECLARE_STATIC_KEY_FALSE(pkvm_mte_supported);
+
+#define kvm_supports_mte(kvm)                                                  \
+	(system_supports_mte() &&                                              \
+	 (!kvm_vm_is_protected(kvm) ||                                         \
+	  static_branch_unlikely(&pkvm_mte_supported)))
 #define kvm_has_mte(kvm)					\
 	(system_supports_mte() &&				\
 	 test_bit(KVM_ARCH_FLAG_MTE_ENABLED, &(kvm)->arch.flags))
diff --git a/arch/arm64/include/asm/kvm_pkvm.h b/arch/arm64/include/asm/kvm_pkvm.h
index cd56438a34be..ef5d4870c043 100644
--- a/arch/arm64/include/asm/kvm_pkvm.h
+++ b/arch/arm64/include/asm/kvm_pkvm.h
@@ -73,10 +73,12 @@ void kvm_shadow_destroy(struct kvm *kvm);
  * Allow for protected VMs:
  * - Branch Target Identification
  * - Speculative Store Bypassing
+ * - Memory Tagging Extension
  */
 #define PVM_ID_AA64PFR1_ALLOW (\
 	ARM64_FEATURE_MASK(ID_AA64PFR1_BT) | \
-	ARM64_FEATURE_MASK(ID_AA64PFR1_SSBS) \
+	ARM64_FEATURE_MASK(ID_AA64PFR1_SSBS) | \
+	ARM64_FEATURE_MASK(ID_AA64PFR1_MTE) \
 	)
 
 /*
diff --git a/arch/arm64/kernel/image-vars.h b/arch/arm64/kernel/image-vars.h
index 2d4d6836ff47..26a9b31478aa 100644
--- a/arch/arm64/kernel/image-vars.h
+++ b/arch/arm64/kernel/image-vars.h
@@ -84,6 +84,9 @@ KVM_NVHE_ALIAS(__hyp_stub_vectors);
 KVM_NVHE_ALIAS(arm64_const_caps_ready);
 KVM_NVHE_ALIAS(cpu_hwcap_keys);
 
+/* Kernel symbol needed for kvm_supports_mte() check. */
+KVM_NVHE_ALIAS(pkvm_mte_supported);
+
 /* Static keys which are set if a vGIC trap should be handled in hyp. */
 KVM_NVHE_ALIAS(vgic_v2_cpuif_trap);
 KVM_NVHE_ALIAS(vgic_v3_cpuif_trap);
diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
index 91ca128e7daa..7c79a1be1e39 100644
--- a/arch/arm64/kvm/arm.c
+++ b/arch/arm64/kvm/arm.c
@@ -60,6 +60,7 @@ static bool vgic_present;
 
 static DEFINE_PER_CPU(unsigned char, kvm_arm_hardware_enabled);
 DEFINE_STATIC_KEY_FALSE(userspace_irqchip_in_use);
+DEFINE_STATIC_KEY_FALSE(pkvm_mte_supported);
 
 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu)
 {
@@ -96,9 +97,7 @@ int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
 		break;
 	case KVM_CAP_ARM_MTE:
 		mutex_lock(&kvm->lock);
-		if (!system_supports_mte() ||
-		    kvm_vm_is_protected(kvm) ||
-		    kvm->created_vcpus) {
+		if (!kvm_supports_mte(kvm) || kvm->created_vcpus) {
 			r = -EINVAL;
 		} else {
 			r = 0;
@@ -334,6 +333,9 @@ static int pkvm_check_extension(struct kvm *kvm, long ext, int kvm_cap)
 	case KVM_CAP_ARM_VM_IPA_SIZE:
 		r = kvm_cap;
 		break;
+	case KVM_CAP_ARM_MTE:
+		r = kvm_cap && static_branch_unlikely(&pkvm_mte_supported);
+		break;
 	case KVM_CAP_GUEST_DEBUG_HW_BPS:
 		r = min(kvm_cap, pkvm_get_max_brps());
 		break;
@@ -1954,9 +1956,36 @@ static void kvm_reserved_memory_init(void)
 		if (!of_get_property(node, "compatible", NULL) &&
 		    of_get_property(node, "no-map", NULL))
 			disown_reserved_memory(node);
+
+		if (of_device_is_compatible(node, "arm,mte-tag-storage"))
+			disown_reserved_memory(node);
 	}
 }
 
+static void kvm_mte_init(void)
+{
+	struct device_node *memory;
+
+	if (!system_supports_mte() || !acpi_disabled ||
+	    !is_protected_kvm_enabled())
+		return;
+
+	/*
+	 * It is only safe to turn on MTE for protected VMs if we can protect
+	 * the guests from host accesses to their tag storage. If every memory
+	 * region has an arm,mte-alloc property we know that all tag storage
+	 * regions exposed to physical memory, if any, are described by a
+	 * reserved-memory compatible with arm,mte-tag-storage. We can use these
+	 * descriptions to unmap these regions from the host's stage 2 page
+	 * tables (see kvm_reserved_memory_init).
+	 */
+	for_each_node_by_type(memory, "memory")
+		if (!of_get_property(memory, "arm,mte-alloc", NULL))
+			return;
+
+	static_branch_enable(&pkvm_mte_supported);
+}
+
 static int init_subsystems(void)
 {
 	int err = 0;
@@ -1999,6 +2028,8 @@ static int init_subsystems(void)
 
 	kvm_reserved_memory_init();
 
+	kvm_mte_init();
+
 out:
 	if (err || !is_protected_kvm_enabled())
 		on_each_cpu(_kvm_arch_hardware_disable, NULL, 1);
diff --git a/arch/arm64/kvm/hyp/nvhe/pkvm.c b/arch/arm64/kvm/hyp/nvhe/pkvm.c
index 80260e8e97f2..96538c984858 100644
--- a/arch/arm64/kvm/hyp/nvhe/pkvm.c
+++ b/arch/arm64/kvm/hyp/nvhe/pkvm.c
@@ -88,7 +88,7 @@ static void pvm_init_traps_aa64pfr1(struct kvm_vcpu *vcpu)
 	/* Memory Tagging: Trap and Treat as Untagged if not supported. */
 	if (!FIELD_GET(ARM64_FEATURE_MASK(ID_AA64PFR1_MTE), feature_ids)) {
 		hcr_set |= HCR_TID5;
-		hcr_clear |= HCR_DCT | HCR_ATA;
+		hcr_clear |= HCR_ATA;
 	}
 
 	vcpu->arch.hcr_el2 |= hcr_set;
@@ -179,8 +179,8 @@ static void pvm_init_trap_regs(struct kvm_vcpu *vcpu)
 	 * - Feature id registers: to control features exposed to guests
 	 * - Implementation-defined features
 	 */
-	vcpu->arch.hcr_el2 = HCR_GUEST_FLAGS |
-			     HCR_TID3 | HCR_TACR | HCR_TIDCP | HCR_TID1;
+	vcpu->arch.hcr_el2 = HCR_GUEST_FLAGS | HCR_TID3 | HCR_TACR | HCR_TIDCP |
+			     HCR_TID1 | HCR_ATA;
 
 	if (cpus_have_const_cap(ARM64_HAS_RAS_EXTN)) {
 		/* route synchronous external abort exceptions to EL2 */
@@ -473,6 +473,8 @@ static int init_shadow_structs(struct kvm *kvm, struct kvm_shadow_vm *vm,
 	vm->host_kvm = kvm;
 	vm->kvm.created_vcpus = nr_vcpus;
 	vm->kvm.arch.vtcr = host_kvm.arch.vtcr;
+	if (kvm_supports_mte(kvm) && test_bit(KVM_ARCH_FLAG_MTE_ENABLED, &kvm->arch.flags))
+		set_bit(KVM_ARCH_FLAG_MTE_ENABLED, &vm->kvm.arch.flags);
 	vm->kvm.arch.pkvm.enabled = READ_ONCE(kvm->arch.pkvm.enabled);
 	vm->kvm.arch.mmu.last_vcpu_ran = last_ran;
 	vm->last_ran_size = last_ran_size;
diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
index bca90b7354b9..5e079daf2d8e 100644
--- a/arch/arm64/kvm/mmu.c
+++ b/arch/arm64/kvm/mmu.c
@@ -1228,8 +1228,10 @@ static int pkvm_mem_abort(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
 		goto dec_account;
 	}
 
-	write_lock(&kvm->mmu_lock);
 	pfn = page_to_pfn(page);
+	sanitise_mte_tags(kvm, pfn, PAGE_SIZE);
+
+	write_lock(&kvm->mmu_lock);
 	ret = pkvm_host_map_guest(pfn, fault_ipa >> PAGE_SHIFT);
 	if (ret) {
 		if (ret == -EAGAIN)
-- 
2.37.0.144.g8ac04bfd2-goog

Re: [PATCH v2 0/3] KVM: arm64: support MTE in protected VMs

[Cornelia Huck]

On Fri, Jul 08 2022, Peter Collingbourne <pcc@google.com> wrote:

> Hi,
>
> This patch series contains a proposed extension to pKVM that allows MTE
> to be exposed to the protected guests. It is based on the base pKVM
> series previously sent to the list [1] and later rebased to 5.19-rc3
> and uploaded to [2].
>
> This series takes precautions against host compromise of the guests
> via direct access to their tag storage, by preventing the host from
> accessing the tag storage via stage 2 page tables. The device tree
> must describe the physical memory address of the tag storage, if any,
> and the memory nodes must declare that the tag storage location is
> described. Otherwise, the MTE feature is disabled in protected guests.
>
> Now that we can easily do so, we also prevent the host from accessing
> any unmapped reserved-memory regions without a driver, as the host
> has no business accessing that memory.
>
> A proposed extension to the devicetree specification is available at
> [3], a patched version of QEMU that produces the required device tree
> nodes is available at [4] and a patched version of the crosvm hypervisor
> that enables MTE is available at [5].

I'm unsure how this is supposed to work with QEMU + KVM, as your QEMU
patch adds mte-alloc properties to regions that are exposed as a
separate address space (which will not work with KVM). Is the magic in
that new shared section?

>
> v2:
> - refcount the PTEs owned by NOBODY
>
> [1] https://lore.kernel.org/all/20220519134204.5379-1-will@kernel.org/
> [2] https://android-kvm.googlesource.com/linux/ for-upstream/pkvm-base-v2
> [3] https://github.com/pcc/devicetree-specification mte-alloc
> [4] https://github.com/pcc/qemu mte-shared-alloc
> [5] https://chromium-review.googlesource.com/c/chromiumos/platform/crosvm/+/3719324
>
> Peter Collingbourne (3):
>   KVM: arm64: add a hypercall for disowning pages
>   KVM: arm64: disown unused reserved-memory regions
>   KVM: arm64: allow MTE in protected VMs if the tag storage is known
>
>  arch/arm64/include/asm/kvm_asm.h              |  1 +
>  arch/arm64/include/asm/kvm_host.h             |  6 ++
>  arch/arm64/include/asm/kvm_pkvm.h             |  4 +-
>  arch/arm64/kernel/image-vars.h                |  3 +
>  arch/arm64/kvm/arm.c                          | 83 ++++++++++++++++++-
>  arch/arm64/kvm/hyp/include/nvhe/mem_protect.h |  1 +
>  arch/arm64/kvm/hyp/include/nvhe/pkvm.h        |  1 +
>  arch/arm64/kvm/hyp/nvhe/hyp-main.c            |  9 ++
>  arch/arm64/kvm/hyp/nvhe/mem_protect.c         | 11 +++
>  arch/arm64/kvm/hyp/nvhe/pkvm.c                |  8 +-
>  arch/arm64/kvm/mmu.c                          |  4 +-
>  11 files changed, 123 insertions(+), 8 deletions(-)

Re: [PATCH v2 0/3] KVM: arm64: support MTE in protected VMs

[Peter Collingbourne]

On Tue, Jul 19, 2022 at 7:50 AM Cornelia Huck <cohuck@redhat.com> wrote:
>
> On Fri, Jul 08 2022, Peter Collingbourne <pcc@google.com> wrote:
>
> > Hi,
> >
> > This patch series contains a proposed extension to pKVM that allows MTE
> > to be exposed to the protected guests. It is based on the base pKVM
> > series previously sent to the list [1] and later rebased to 5.19-rc3
> > and uploaded to [2].
> >
> > This series takes precautions against host compromise of the guests
> > via direct access to their tag storage, by preventing the host from
> > accessing the tag storage via stage 2 page tables. The device tree
> > must describe the physical memory address of the tag storage, if any,
> > and the memory nodes must declare that the tag storage location is
> > described. Otherwise, the MTE feature is disabled in protected guests.
> >
> > Now that we can easily do so, we also prevent the host from accessing
> > any unmapped reserved-memory regions without a driver, as the host
> > has no business accessing that memory.
> >
> > A proposed extension to the devicetree specification is available at
> > [3], a patched version of QEMU that produces the required device tree
> > nodes is available at [4] and a patched version of the crosvm hypervisor
> > that enables MTE is available at [5].
>
> I'm unsure how this is supposed to work with QEMU + KVM, as your QEMU
> patch adds mte-alloc properties to regions that are exposed as a
> separate address space (which will not work with KVM). Is the magic in
> that new shared section?

Hi Cornelia,

The intent is that the mte-alloc property may be set on memory whose
allocation tag storage is not directly accessible via physical memory,
since in this case there is no need for the hypervisor to do anything
to protect allocation tag storage before exposing MTE to guests. In
the case of QEMU + KVM, I would expect the emulated system to not
expose the allocation tag storage directly, in which case it would be
able to set mte-alloc on all memory nodes without further action,
exactly as my patch implements for TCG. With the interface as
proposed, QEMU would need to reject the mte-shared-alloc option when
KVM is enabled, as there is currently no mechanism for KVM-accelerated
virtualized tag storage.

Note that these properties are only relevant for guest kernels running
under an emulated EL2 in which pKVM could conceivably run, which means
that the host would need to implement FEAT_NV2. As far as I know there
is currently no support for NV2 neither in QEMU TCG nor in the Linux
kernel, and I'm unaware of any available hardware that supports both
NV2 and MTE, so it'll be a while before any of this becomes relevant.

Peter

Re: [PATCH v2 0/3] KVM: arm64: support MTE in protected VMs

[Cornelia Huck]

On Tue, Jul 19 2022, Peter Collingbourne <pcc@google.com> wrote:

> On Tue, Jul 19, 2022 at 7:50 AM Cornelia Huck <cohuck@redhat.com> wrote:
>>
>> On Fri, Jul 08 2022, Peter Collingbourne <pcc@google.com> wrote:
>>
>> > Hi,
>> >
>> > This patch series contains a proposed extension to pKVM that allows MTE
>> > to be exposed to the protected guests. It is based on the base pKVM
>> > series previously sent to the list [1] and later rebased to 5.19-rc3
>> > and uploaded to [2].
>> >
>> > This series takes precautions against host compromise of the guests
>> > via direct access to their tag storage, by preventing the host from
>> > accessing the tag storage via stage 2 page tables. The device tree
>> > must describe the physical memory address of the tag storage, if any,
>> > and the memory nodes must declare that the tag storage location is
>> > described. Otherwise, the MTE feature is disabled in protected guests.
>> >
>> > Now that we can easily do so, we also prevent the host from accessing
>> > any unmapped reserved-memory regions without a driver, as the host
>> > has no business accessing that memory.
>> >
>> > A proposed extension to the devicetree specification is available at
>> > [3], a patched version of QEMU that produces the required device tree
>> > nodes is available at [4] and a patched version of the crosvm hypervisor
>> > that enables MTE is available at [5].
>>
>> I'm unsure how this is supposed to work with QEMU + KVM, as your QEMU
>> patch adds mte-alloc properties to regions that are exposed as a
>> separate address space (which will not work with KVM). Is the magic in
>> that new shared section?
>
> Hi Cornelia,
>
> The intent is that the mte-alloc property may be set on memory whose
> allocation tag storage is not directly accessible via physical memory,
> since in this case there is no need for the hypervisor to do anything
> to protect allocation tag storage before exposing MTE to guests. In
> the case of QEMU + KVM, I would expect the emulated system to not
> expose the allocation tag storage directly, in which case it would be
> able to set mte-alloc on all memory nodes without further action,
> exactly as my patch implements for TCG. With the interface as
> proposed, QEMU would need to reject the mte-shared-alloc option when
> KVM is enabled, as there is currently no mechanism for KVM-accelerated
> virtualized tag storage.

Ok, that makes sense.

>
> Note that these properties are only relevant for guest kernels running
> under an emulated EL2 in which pKVM could conceivably run, which means
> that the host would need to implement FEAT_NV2. As far as I know there
> is currently no support for NV2 neither in QEMU TCG nor in the Linux
> kernel, and I'm unaware of any available hardware that supports both
> NV2 and MTE, so it'll be a while before any of this becomes relevant.

Nod.

I'm mostly interested because I wanted to figure out how this feature
might interact with enabling MTE for QEMU+KVM. I'll keep it in mind.

Thanks!