Skip to contentSkip to conversation

KVM: arm64: Avoid NULL dereference in vgic-v2 device attr accessors

PATCH Applied Oliver Upton 2 participants 5 messages Apr 24, 2024

Subsystems (5)
Stable branch KVM Kselftest arm64 +1 more subsystems
lore
Share & tools
lore
#0 lore raw

cover KVM: arm64: Avoid NULL dereference in vgic-v2 device attr accessors

Alex reports that it is possible to trigger a NULL dereference via the vgic-v2 device attribute accessors, stemming from a lack of sanitization of user input...

Here's a fix + regression test for the bug. Obviously, I intend to take these as a fix ASAP.

Oliver Upton (2):
KVM: arm64: vgic-v2: Check for non-NULL vCPU in vgic_v2_parse_attr()
KVM: selftests: Add test for uaccesses to non-existent vgic-v2 CPUIF

 arch/arm64/kvm/vgic/vgic-kvm-device.c         |  8 +--
 .../testing/selftests/kvm/aarch64/vgic_init.c | 49 +++++++++++++++++++
 2 files changed, 53 insertions(+), 4 deletions(-)

base-commit: fec50db7033ea478773b159e0e2efb135270e3b7

Signature · 1 line
2.44.0.769.g3c40516874-goog
#1 lore raw

PATCH 1/2 KVM: arm64: vgic-v2: Check for non-NULL vCPU in vgic_v2_parse_attr()

vgic_v2_parse_attr() is responsible for finding the vCPU that matches the user-provided CPUID, which (of course) may not be valid. If the ID is invalid, kvm_get_vcpu_by_id() returns NULL, which isn't handled gracefully.

Similar to the GICv3 uaccess flow, check that kvm_get_vcpu_by_id() actually returns something and fail the ioctl if not.

Fixes 7d450e282171 ("KVM: arm/arm64: vgic-new: Add userland access to VGIC dist registers")
---
 arch/arm64/kvm/vgic/vgic-kvm-device.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
arch/arm64/kvm/vgic/vgic-kvm-device.c +4 -4
index f48b8dab8b3d..1d26bb5b02f4 100644
@@ -338,12 +338,12 @@ int kvm_register_vgic_device(unsigned long type)
 int vgic_v2_parse_attr(struct kvm_device *dev, struct kvm_device_attr *attr,
 		       struct vgic_reg_attr *reg_attr)
 {
-	int cpuid;
+	int cpuid = FIELD_GET(KVM_DEV_ARM_VGIC_CPUID_MASK, attr->attr);
 
-	cpuid = FIELD_GET(KVM_DEV_ARM_VGIC_CPUID_MASK, attr->attr);
-
-	reg_attr->vcpu = kvm_get_vcpu_by_id(dev->kvm, cpuid);
 	reg_attr->addr = attr->attr & KVM_DEV_ARM_VGIC_OFFSET_MASK;
+	reg_attr->vcpu = kvm_get_vcpu_by_id(dev->kvm, cpuid);
+	if (!reg_attr->vcpu)
+		return -EINVAL;
 
 	return 0;
 }
Signature · 1 line
2.44.0.769.g3c40516874-goog
#2 lore raw

PATCH 2/2 KVM: selftests: Add test for uaccesses to non-existent vgic-v2 CPUIF

In reply to Oliver Upton

Assert that accesses to a non-existent vgic-v2 CPU interface consistently fail across the various KVM device attr ioctls. This also serves as a regression test for a bug wherein KVM hits a NULL dereference when the CPUID specified in the ioctl is invalid.

Note that there is no need to print the observed errno, as TEST_ASSERT() will take care of it.

---
 .../testing/selftests/kvm/aarch64/vgic_init.c | 49 +++++++++++++++++++
 1 file changed, 49 insertions(+)
tools/testing/selftests/kvm/aarch64/vgic_init.c +49 -0
index eef816b80993..ca917c71ff60 100644
@@ -84,6 +84,18 @@ static struct vm_gic vm_gic_create_with_vcpus(uint32_t gic_dev_type,
 	return v;
 }
 
+static struct vm_gic vm_gic_create_barebones(uint32_t gic_dev_type)
+{
+	struct vm_gic v;
+
+	v.gic_dev_type = gic_dev_type;
+	v.vm = vm_create_barebones();
+	v.gic_fd = kvm_create_device(v.vm, gic_dev_type);
+
+	return v;
+}
+
+
 static void vm_gic_destroy(struct vm_gic *v)
 {
 	close(v->gic_fd);
@@ -357,6 +369,40 @@ static void test_vcpus_then_vgic(uint32_t gic_dev_type)
 	vm_gic_destroy(&v);
 }
 
+#define KVM_VGIC_V2_ATTR(offset, cpu) \
+	(FIELD_PREP(KVM_DEV_ARM_VGIC_OFFSET_MASK, offset) | \
+	 FIELD_PREP(KVM_DEV_ARM_VGIC_CPUID_MASK, cpu))
+
+#define GIC_CPU_CTRL	0x00
+
+static void test_v2_uaccess_cpuif_no_vcpus(void)
+{
+	struct vm_gic v;
+	u64 val = 0;
+	int ret;
+
+	v = vm_gic_create_barebones(KVM_DEV_TYPE_ARM_VGIC_V2);
+	subtest_dist_rdist(&v);
+
+	ret = __kvm_has_device_attr(v.gic_fd, KVM_DEV_ARM_VGIC_GRP_CPU_REGS,
+				    KVM_VGIC_V2_ATTR(GIC_CPU_CTRL, 0));
+	TEST_ASSERT(ret && errno == EINVAL,
+		    "accessed non-existent CPU interface, want errno: %i",
+		    EINVAL);
+	ret = __kvm_device_attr_get(v.gic_fd, KVM_DEV_ARM_VGIC_GRP_CPU_REGS,
+				    KVM_VGIC_V2_ATTR(GIC_CPU_CTRL, 0), &val);
+	TEST_ASSERT(ret && errno == EINVAL,
+		    "accessed non-existent CPU interface, want errno: %i",
+		    EINVAL);
+	ret = __kvm_device_attr_set(v.gic_fd, KVM_DEV_ARM_VGIC_GRP_CPU_REGS,
+				    KVM_VGIC_V2_ATTR(GIC_CPU_CTRL, 0), &val);
+	TEST_ASSERT(ret && errno == EINVAL,
+		    "accessed non-existent CPU interface, want errno: %i",
+		    EINVAL);
+
+	vm_gic_destroy(&v);
+}
+
 static void test_v3_new_redist_regions(void)
 {
 	struct kvm_vcpu *vcpus[NR_VCPUS];
@@ -675,6 +721,9 @@ void run_tests(uint32_t gic_dev_type)
 	test_vcpus_then_vgic(gic_dev_type);
 	test_vgic_then_vcpus(gic_dev_type);
 
+	if (VGIC_DEV_IS_V2(gic_dev_type))
+		test_v2_uaccess_cpuif_no_vcpus();
+
 	if (VGIC_DEV_IS_V3(gic_dev_type)) {
 		test_v3_new_redist_regions();
 		test_v3_typer_accesses();
Signature · 1 line
2.44.0.769.g3c40516874-goog
maintainer Reviewed-by on the v1 series
#3 lore raw
In reply to Oliver Upton
Quoted 12 lines · Oliver Upton

Alex reports that it is possible to trigger a NULL dereference via the vgic-v2 device attribute accessors, stemming from a lack of sanitization of user input...

Here's a fix + regression test for the bug. Obviously, I intend to take these as a fix ASAP.

Oliver Upton (2):
KVM: arm64: vgic-v2: Check for non-NULL vCPU in vgic_v2_parse_attr()
KVM: selftests: Add test for uaccesses to non-existent vgic-v2 CPUIF

Quoted 6 lines
 arch/arm64/kvm/vgic/vgic-kvm-device.c         |  8 +--
 .../testing/selftests/kvm/aarch64/vgic_init.c | 49 +++++++++++++++++++
 2 files changed, 53 insertions(+), 4 deletions(-)

base-commit: fec50db7033ea478773b159e0e2efb135270e3b7

Thanks Alex for the heads up!

Please queue this at your earliest convenience.

	M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
author Applied
#4 lore raw
In reply to Oliver Upton
Quoted 8 lines · Oliver Upton

Alex reports that it is possible to trigger a NULL dereference via the vgic-v2 device attribute accessors, stemming from a lack of sanitization of user input...

Here's a fix + regression test for the bug. Obviously, I intend to take these as a fix ASAP.

[...]

Applied to kvmarm/fixes, thanks!

[1/2] KVM: arm64: vgic-v2: Check for non-NULL vCPU in vgic_v2_parse_attr()

      https://git.kernel.org/kvmarm/kvmarm/c/6ddb4f372fc6
[2/2] KVM: selftests: Add test for uaccesses to non-existent vgic-v2 CPUIF
      https://git.kernel.org/kvmarm/kvmarm/c/160933e330f4

--
Best,
Oliver