[PATCH 0/5 v3] KVM: nSVM: Check addresses of MSR bitmap and IO bitmap tables on vmrun of nested guests

[PATCH 0/5 v3] KVM: nSVM: Check addresses of MSR bitmap and IO bitmap tables on vmrun of nested guests

[Krish Sadhukhan]

v2 -> v3:
	1. Moved the #defines for IOPM_ALLOC_ORDER and MSRPM_ALLOC_ORDER so
	   that they can be used by nested_vmcb_check_controls()
	2. Fixed the wrong check in nested_vmcb_check_controls() in patch# 2
	   (which was patch# 1 in v2).
	3. Added a clean-up patch for nested_svm_vmrun().


[PATCH 1/5 v3] KVM: SVM: Move IOPM_ALLOC_ORDER and MSRPM_ALLOC_ORDER
[PATCH 2/5 v3] nSVM: Check addresses of MSR and IO bitmap
[PATCH 3/5 v3] KVM: nSVM: Cleanup in nested_svm_vmrun()
[PATCH 4/5 v3] Test: nSVM: Test MSR and IO bitmap address
[PATCH 5/5 v3] Test: SVM: Use ALIGN macro when aligning 'io_bitmap_area'

 arch/x86/kvm/svm/nested.c | 63 +++++++++++++++++++++++++++--------------------
 arch/x86/kvm/svm/svm.c    |  3 ---
 arch/x86/kvm/svm/svm.h    |  3 +++
 3 files changed, 39 insertions(+), 30 deletions(-)

Krish Sadhukhan (3):
      KVM: SVM: Move IOPM_ALLOC_ORDER and MSRPM_ALLOC_ORDER #defines to svm.h
      nSVM: Check addresses of MSR and IO bitmap
      KVM: nSVM: Cleanup in nested_svm_vmrun()

 x86/svm.c       |  2 +-
 x86/svm_tests.c | 38 ++++++++++++++++++++++++++++++++++++++
 2 files changed, 39 insertions(+), 1 deletion(-)

Krish Sadhukhan (2):
      nSVM: Test MSR and IO bitmap address
      SVM: Use ALIGN macro when aligning 'io_bitmap_area'

[PATCH 1/5 v3] KVM: SVM: Move IOPM_ALLOC_ORDER and MSRPM_ALLOC_ORDER #defines to svm.h

[Krish Sadhukhan]

These #defines will be used by nested.c in the next patch. So move these to
svm.h.

Signed-off-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
---
 arch/x86/kvm/svm/svm.c | 3 ---
 arch/x86/kvm/svm/svm.h | 3 +++
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/arch/x86/kvm/svm/svm.c b/arch/x86/kvm/svm/svm.c
index f923e14e87df..1641cb8ac5dd 100644
--- a/arch/x86/kvm/svm/svm.c
+++ b/arch/x86/kvm/svm/svm.c
@@ -55,9 +55,6 @@ static const struct x86_cpu_id svm_cpu_id[] = {
 MODULE_DEVICE_TABLE(x86cpu, svm_cpu_id);
 #endif
 
-#define IOPM_ALLOC_ORDER 2
-#define MSRPM_ALLOC_ORDER 1
-
 #define SEG_TYPE_LDT 2
 #define SEG_TYPE_BUSY_TSS16 3
 
diff --git a/arch/x86/kvm/svm/svm.h b/arch/x86/kvm/svm/svm.h
index 0fe874ae5498..f529a259a03e 100644
--- a/arch/x86/kvm/svm/svm.h
+++ b/arch/x86/kvm/svm/svm.h
@@ -42,6 +42,9 @@ static const struct svm_host_save_msrs {
 };
 #define NR_HOST_SAVE_USER_MSRS ARRAY_SIZE(host_save_user_msrs)
 
+#define IOPM_ALLOC_ORDER 2
+#define MSRPM_ALLOC_ORDER 1
+
 #define MAX_DIRECT_ACCESS_MSRS	18
 #define MSRPM_OFFSETS	16
 extern u32 msrpm_offsets[MSRPM_OFFSETS] __read_mostly;
-- 
2.27.0

[PATCH 2/5 v3] nSVM: Check addresses of MSR and IO bitmap

[Krish Sadhukhan]

According to section "Canonicalization and Consistency Checks" in APM vol 2,
the following guest state is illegal:

        "The MSR or IOIO intercept tables extend to a physical address that
         is greater than or equal to the maximum supported physical address."

Also check that these addresses are aligned on page boundary.

Signed-off-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
---
 arch/x86/kvm/svm/nested.c | 20 ++++++++++++++------
 1 file changed, 14 insertions(+), 6 deletions(-)

diff --git a/arch/x86/kvm/svm/nested.c b/arch/x86/kvm/svm/nested.c
index 7a605ad8254d..caf285e643db 100644
--- a/arch/x86/kvm/svm/nested.c
+++ b/arch/x86/kvm/svm/nested.c
@@ -214,7 +214,8 @@ static bool svm_get_nested_state_pages(struct kvm_vcpu *vcpu)
 	return true;
 }
 
-static bool nested_vmcb_check_controls(struct vmcb_control_area *control)
+static bool nested_vmcb_check_controls(struct kvm_vcpu *vcpu,
+				       struct vmcb_control_area *control)
 {
 	if ((vmcb_is_intercept(control, INTERCEPT_VMRUN)) == 0)
 		return false;
@@ -226,10 +227,17 @@ static bool nested_vmcb_check_controls(struct vmcb_control_area *control)
 	    !npt_enabled)
 		return false;
 
+	if (!page_address_valid(vcpu, control->msrpm_base_pa +
+	    MSRPM_ALLOC_ORDER * PAGE_SIZE))
+		return false;
+	if (!page_address_valid(vcpu, control->iopm_base_pa +
+	    IOPM_ALLOC_ORDER * PAGE_SIZE))
+		return false;
+
 	return true;
 }
 
-static bool nested_vmcb_checks(struct vcpu_svm *svm, struct vmcb *vmcb12)
+static bool nested_vmcb_checks(struct kvm_vcpu *vcpu, struct vmcb *vmcb12)
 {
 	bool vmcb12_lma;
 
@@ -258,10 +266,10 @@ static bool nested_vmcb_checks(struct vcpu_svm *svm, struct vmcb *vmcb12)
 		    (vmcb12->save.cr3 & MSR_CR3_LONG_MBZ_MASK))
 			return false;
 	}
-	if (!kvm_is_valid_cr4(&svm->vcpu, vmcb12->save.cr4))
+	if (!kvm_is_valid_cr4(vcpu, vmcb12->save.cr4))
 		return false;
 
-	return nested_vmcb_check_controls(&vmcb12->control);
+	return nested_vmcb_check_controls(vcpu, &vmcb12->control);
 }
 
 static void load_nested_vmcb_control(struct vcpu_svm *svm,
@@ -488,7 +496,7 @@ int nested_svm_vmrun(struct vcpu_svm *svm)
 	if (WARN_ON_ONCE(!svm->nested.initialized))
 		return -EINVAL;
 
-	if (!nested_vmcb_checks(svm, vmcb12)) {
+	if (!nested_vmcb_checks(&svm->vcpu, vmcb12)) {
 		vmcb12->control.exit_code    = SVM_EXIT_ERR;
 		vmcb12->control.exit_code_hi = 0;
 		vmcb12->control.exit_info_1  = 0;
@@ -1176,7 +1184,7 @@ static int svm_set_nested_state(struct kvm_vcpu *vcpu,
 		goto out_free;
 
 	ret = -EINVAL;
-	if (!nested_vmcb_check_controls(ctl))
+	if (!nested_vmcb_check_controls(vcpu, ctl))
 		goto out_free;
 
 	/*
-- 
2.27.0

[PATCH 3/5 v3] KVM: nSVM: Cleanup in nested_svm_vmrun()

[Krish Sadhukhan]

Use local variables to derefence svm->vcpu and svm->vmcb as they make the
code tidier.

Suggested-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
---
 arch/x86/kvm/svm/nested.c | 45 ++++++++++++++++++++-------------------
 1 file changed, 23 insertions(+), 22 deletions(-)

diff --git a/arch/x86/kvm/svm/nested.c b/arch/x86/kvm/svm/nested.c
index caf285e643db..e9228fdac9b7 100644
--- a/arch/x86/kvm/svm/nested.c
+++ b/arch/x86/kvm/svm/nested.c
@@ -470,33 +470,34 @@ int nested_svm_vmrun(struct vcpu_svm *svm)
 {
 	int ret;
 	struct vmcb *vmcb12;
+	struct kvm_vcpu *vcpu = &svm->vcpu;
 	struct vmcb *hsave = svm->nested.hsave;
 	struct vmcb *vmcb = svm->vmcb;
 	struct kvm_host_map map;
 	u64 vmcb12_gpa;
 
-	if (is_smm(&svm->vcpu)) {
-		kvm_queue_exception(&svm->vcpu, UD_VECTOR);
+	if (is_smm(vcpu)) {
+		kvm_queue_exception(vcpu, UD_VECTOR);
 		return 1;
 	}
 
-	vmcb12_gpa = svm->vmcb->save.rax;
-	ret = kvm_vcpu_map(&svm->vcpu, gpa_to_gfn(vmcb12_gpa), &map);
+	vmcb12_gpa = vmcb->save.rax;
+	ret = kvm_vcpu_map(vcpu, gpa_to_gfn(vmcb12_gpa), &map);
 	if (ret == -EINVAL) {
-		kvm_inject_gp(&svm->vcpu, 0);
+		kvm_inject_gp(vcpu, 0);
 		return 1;
 	} else if (ret) {
-		return kvm_skip_emulated_instruction(&svm->vcpu);
+		return kvm_skip_emulated_instruction(vcpu);
 	}
 
-	ret = kvm_skip_emulated_instruction(&svm->vcpu);
+	ret = kvm_skip_emulated_instruction(vcpu);
 
 	vmcb12 = map.hva;
 
 	if (WARN_ON_ONCE(!svm->nested.initialized))
 		return -EINVAL;
 
-	if (!nested_vmcb_checks(&svm->vcpu, vmcb12)) {
+	if (!nested_vmcb_checks(vcpu, vmcb12)) {
 		vmcb12->control.exit_code    = SVM_EXIT_ERR;
 		vmcb12->control.exit_code_hi = 0;
 		vmcb12->control.exit_info_1  = 0;
@@ -504,7 +505,7 @@ int nested_svm_vmrun(struct vcpu_svm *svm)
 		goto out;
 	}
 
-	trace_kvm_nested_vmrun(svm->vmcb->save.rip, vmcb12_gpa,
+	trace_kvm_nested_vmrun(vmcb->save.rip, vmcb12_gpa,
 			       vmcb12->save.rip,
 			       vmcb12->control.int_ctl,
 			       vmcb12->control.event_inj,
@@ -518,8 +519,8 @@ int nested_svm_vmrun(struct vcpu_svm *svm)
 				    vmcb12->control.intercepts[INTERCEPT_WORD5]);
 
 	/* Clear internal status */
-	kvm_clear_exception_queue(&svm->vcpu);
-	kvm_clear_interrupt_queue(&svm->vcpu);
+	kvm_clear_exception_queue(vcpu);
+	kvm_clear_interrupt_queue(vcpu);
 
 	/*
 	 * Save the old vmcb, so we don't need to pick what we save, but can
@@ -531,17 +532,17 @@ int nested_svm_vmrun(struct vcpu_svm *svm)
 	hsave->save.ds     = vmcb->save.ds;
 	hsave->save.gdtr   = vmcb->save.gdtr;
 	hsave->save.idtr   = vmcb->save.idtr;
-	hsave->save.efer   = svm->vcpu.arch.efer;
-	hsave->save.cr0    = kvm_read_cr0(&svm->vcpu);
-	hsave->save.cr4    = svm->vcpu.arch.cr4;
-	hsave->save.rflags = kvm_get_rflags(&svm->vcpu);
-	hsave->save.rip    = kvm_rip_read(&svm->vcpu);
+	hsave->save.efer   = vcpu->arch.efer;
+	hsave->save.cr0    = kvm_read_cr0(vcpu);
+	hsave->save.cr4    = vcpu->arch.cr4;
+	hsave->save.rflags = kvm_get_rflags(vcpu);
+	hsave->save.rip    = kvm_rip_read(vcpu);
 	hsave->save.rsp    = vmcb->save.rsp;
 	hsave->save.rax    = vmcb->save.rax;
 	if (npt_enabled)
 		hsave->save.cr3    = vmcb->save.cr3;
 	else
-		hsave->save.cr3    = kvm_read_cr3(&svm->vcpu);
+		hsave->save.cr3    = kvm_read_cr3(vcpu);
 
 	copy_vmcb_control_area(&hsave->control, &vmcb->control);
 
@@ -556,15 +557,15 @@ int nested_svm_vmrun(struct vcpu_svm *svm)
 out_exit_err:
 	svm->nested.nested_run_pending = 0;
 
-	svm->vmcb->control.exit_code    = SVM_EXIT_ERR;
-	svm->vmcb->control.exit_code_hi = 0;
-	svm->vmcb->control.exit_info_1  = 0;
-	svm->vmcb->control.exit_info_2  = 0;
+	vmcb->control.exit_code    = SVM_EXIT_ERR;
+	vmcb->control.exit_code_hi = 0;
+	vmcb->control.exit_info_1  = 0;
+	vmcb->control.exit_info_2  = 0;
 
 	nested_svm_vmexit(svm);
 
 out:
-	kvm_vcpu_unmap(&svm->vcpu, &map, true);
+	kvm_vcpu_unmap(vcpu, &map, true);
 
 	return ret;
 }
-- 
2.27.0

[PATCH 4/5 v3] Test: nSVM: Test MSR and IO bitmap address

[Krish Sadhukhan]

According to section "Canonicalization and Consistency Checks" in APM vol 2,
the following guest state is illegal:

    "The MSR or IOIO intercept tables extend to a physical address that
     is greater than or equal to the maximum supported physical address."

Also test that these addresses are aligned on page boundary.

Signed-off-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
---
 x86/svm_tests.c | 38 ++++++++++++++++++++++++++++++++++++++
 1 file changed, 38 insertions(+)

diff --git a/x86/svm_tests.c b/x86/svm_tests.c
index dc86efd..929a3e1 100644
--- a/x86/svm_tests.c
+++ b/x86/svm_tests.c
@@ -2304,6 +2304,43 @@ static void test_dr(void)
 	vmcb->save.dr7 = dr_saved;
 }
 
+extern u8 msr_bitmap_area[];
+extern u8 io_bitmap_area[];
+
+#define TEST_BITMAP_ADDR(prot_type, bitmap_addr, msg)  {                \
+	vmcb->control.intercept = 1ULL << prot_type;                    \
+	addr_unalign = virt_to_phys(bitmap_addr);                       \
+	if (prot_type == INTERCEPT_MSR_PROT)                            \
+		vmcb->control.msrpm_base_pa = addr_unalign;             \
+	else                                                            \
+		vmcb->control.iopm_base_pa = addr_unalign;              \
+	report(svm_vmrun() == SVM_EXIT_ERR, "Test %s address: %lx", msg,\
+	    addr_unalign);                                              \
+	vmcb->control.msrpm_base_pa = addr_spill_beyond_ram;            \
+	report(svm_vmrun() == SVM_EXIT_ERR, "Test %s address: %lx", msg,\
+	    addr_spill_beyond_ram);                                     \
+}                                                                       \
+
+/*
+ * If the MSR or IOIO intercept table extends to a physical address that
+ * is greater than or equal to the maximum supported physical address, the
+ * guest state is illegal.
+ *
+ * [ APM vol 2]
+ */
+static void test_msrpm_iopm_bitmap_addrs(void)
+{
+	u64 addr_unalign;
+	u64 addr_spill_beyond_ram =
+	    (u64)(((u64)1 << cpuid_maxphyaddr()) - 4096);
+
+	/* MSR bitmap address */
+	TEST_BITMAP_ADDR(INTERCEPT_MSR_PROT, msr_bitmap_area, "MSRPM");
+
+	/* MSR bitmap address */
+	TEST_BITMAP_ADDR(INTERCEPT_IOIO_PROT, io_bitmap_area, "IOPM");
+}
+
 static void svm_guest_state_test(void)
 {
 	test_set_guest(basic_guest_main);
@@ -2313,6 +2350,7 @@ static void svm_guest_state_test(void)
 	test_cr3();
 	test_cr4();
 	test_dr();
+	test_msrpm_iopm_bitmap_addrs();
 }
 
 struct svm_test svm_tests[] = {
-- 
2.27.0

[PATCH 5/5 v3] Test: SVM: Use ALIGN macro when aligning 'io_bitmap_area'

[Krish Sadhukhan]

Since the macro is available and we already use it for MSR bitmap table, use
it for aligning IO bitmap table also.

Signed-off-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
---
 x86/svm.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/x86/svm.c b/x86/svm.c
index a1808c7..846cf2a 100644
--- a/x86/svm.c
+++ b/x86/svm.c
@@ -298,7 +298,7 @@ static void setup_svm(void)
 	wrmsr(MSR_EFER, rdmsr(MSR_EFER) | EFER_SVME);
 	wrmsr(MSR_EFER, rdmsr(MSR_EFER) | EFER_NX);
 
-	io_bitmap = (void *) (((ulong)io_bitmap_area + 4095) & ~4095);
+	io_bitmap = (void *) ALIGN((ulong)io_bitmap_area, PAGE_SIZE);
 
 	msr_bitmap = (void *) ALIGN((ulong)msr_bitmap_area, PAGE_SIZE);
 
-- 
2.27.0

Re: [PATCH 2/5 v3] nSVM: Check addresses of MSR and IO bitmap

[Paolo Bonzini]

On 03/02/21 01:40, Krish Sadhukhan wrote:
> According to section "Canonicalization and Consistency Checks" in APM vol 2,
> the following guest state is illegal:
> 
>          "The MSR or IOIO intercept tables extend to a physical address that
>           is greater than or equal to the maximum supported physical address."
> 
> Also check that these addresses are aligned on page boundary.
> 
> Signed-off-by: Krish Sadhukhan <krish.sadhukhan@oracle.com>
> ---
>   arch/x86/kvm/svm/nested.c | 20 ++++++++++++++------
>   1 file changed, 14 insertions(+), 6 deletions(-)
> 
> diff --git a/arch/x86/kvm/svm/nested.c b/arch/x86/kvm/svm/nested.c
> index 7a605ad8254d..caf285e643db 100644
> --- a/arch/x86/kvm/svm/nested.c
> +++ b/arch/x86/kvm/svm/nested.c
> @@ -214,7 +214,8 @@ static bool svm_get_nested_state_pages(struct kvm_vcpu *vcpu)
>   	return true;
>   }
>   
> -static bool nested_vmcb_check_controls(struct vmcb_control_area *control)
> +static bool nested_vmcb_check_controls(struct kvm_vcpu *vcpu,
> +				       struct vmcb_control_area *control)
>   {
>   	if ((vmcb_is_intercept(control, INTERCEPT_VMRUN)) == 0)
>   		return false;
> @@ -226,10 +227,17 @@ static bool nested_vmcb_check_controls(struct vmcb_control_area *control)
>   	    !npt_enabled)
>   		return false;
>   
> +	if (!page_address_valid(vcpu, control->msrpm_base_pa +
> +	    MSRPM_ALLOC_ORDER * PAGE_SIZE))
> +		return false;
> +	if (!page_address_valid(vcpu, control->iopm_base_pa +
> +	    IOPM_ALLOC_ORDER * PAGE_SIZE))

There are four problems:

1) The value does not have to be page-aligned

2) you also have an off-by-one here, the value to be checked is the last 
byte of the previous page

3) ORDER is a shift count not a number of pages

4) there could be an overflow

1-3 can be fixed by something like this:

if (!page_address_valid(vcpu,
                         PAGE_ALIGN(control->xyz_pa) +
                         ((PAGE_SIZE << XYZ_ALLOC_ORDER) - 1)));

but it's even better to extract everything to a new function and not use 
page_address_valid at all.

static inline nested_check_pa(struct kvm_vcpu *vcpu, uint64_t pa,
                               unsigned int order)
{
	uint64_t last_pa = PAGE_ALIGN(pa) + (PAGE_SIZE << order) - 1;
	return last_pa > pa && !(last_pa >> cpuid_maxphyaddr(vcpu));
}

Paolo

> +		return false;
> +
>   	return true;
>   }
>   
> -static bool nested_vmcb_checks(struct vcpu_svm *svm, struct vmcb *vmcb12)
> +static bool nested_vmcb_checks(struct kvm_vcpu *vcpu, struct vmcb *vmcb12)
>   {
>   	bool vmcb12_lma;
>   
> @@ -258,10 +266,10 @@ static bool nested_vmcb_checks(struct vcpu_svm *svm, struct vmcb *vmcb12)
>   		    (vmcb12->save.cr3 & MSR_CR3_LONG_MBZ_MASK))
>   			return false;
>   	}
> -	if (!kvm_is_valid_cr4(&svm->vcpu, vmcb12->save.cr4))
> +	if (!kvm_is_valid_cr4(vcpu, vmcb12->save.cr4))
>   		return false;
>   
> -	return nested_vmcb_check_controls(&vmcb12->control);
> +	return nested_vmcb_check_controls(vcpu, &vmcb12->control);
>   }
>   
>   static void load_nested_vmcb_control(struct vcpu_svm *svm,
> @@ -488,7 +496,7 @@ int nested_svm_vmrun(struct vcpu_svm *svm)
>   	if (WARN_ON_ONCE(!svm->nested.initialized))
>   		return -EINVAL;
>   
> -	if (!nested_vmcb_checks(svm, vmcb12)) {
> +	if (!nested_vmcb_checks(&svm->vcpu, vmcb12)) {
>   		vmcb12->control.exit_code    = SVM_EXIT_ERR;
>   		vmcb12->control.exit_code_hi = 0;
>   		vmcb12->control.exit_info_1  = 0;
> @@ -1176,7 +1184,7 @@ static int svm_set_nested_state(struct kvm_vcpu *vcpu,
>   		goto out_free;
>   
>   	ret = -EINVAL;
> -	if (!nested_vmcb_check_controls(ctl))
> +	if (!nested_vmcb_check_controls(vcpu, ctl))
>   		goto out_free;
>   
>   	/*
>