KVM: x86: KVM_CAP_SYNC_REGS

This commit implements an enhanced x86 version of S390
KVM_CAP_SYNC_REGS functionality. KVM_CAP_SYNC_REGS "allow[s]
userspace to access certain guest registers without having
to call SET/GET_*REGS”. This reduces ioctl overhead which
is particularly important when userspace is making synchronous
guest state modifications (e.g. when emulating and/or intercepting
instructions).

Originally implemented upstream for the S390, the x86 differences
follow:
- userspace can select the register sets to be synchronized with kvm_run
using bit-flags in the kvm_valid_registers and kvm_dirty_registers
fields.
- vcpu_events is available in addition to the regs and sregs register
sets.

Signed-off-by: Ken Hofsass <hofsass@google.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
[Removed wrapper around check for reserved kvm_valid_regs. - Radim]
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>

2 files changed, 105 insertions, 16 deletions

diff --git a/arch/x86/include/uapi/asm/kvm.h b/arch/x86/include/uapi/asm/kvm.h
index f3a960488eae..c535c2fdea13 100644
--- a/arch/x86/include/uapi/asm/kvm.h
+++ b/arch/x86/include/uapi/asm/kvm.h
@@ -354,8 +354,25 @@ struct kvm_xcrs {
 	__u64 padding[16];
 };
 
-/* definition of registers in kvm_run */
+#define KVM_SYNC_X86_REGS      (1UL << 0)
+#define KVM_SYNC_X86_SREGS     (1UL << 1)
+#define KVM_SYNC_X86_EVENTS    (1UL << 2)
+
+#define KVM_SYNC_X86_VALID_FIELDS \
+	(KVM_SYNC_X86_REGS| \
+	 KVM_SYNC_X86_SREGS| \
+	 KVM_SYNC_X86_EVENTS)
+
+/* kvm_sync_regs struct included by kvm_run struct */
 struct kvm_sync_regs {
+	/* Members of this structure are potentially malicious.
+	 * Care must be taken by code reading, esp. interpreting,
+	 * data fields from them inside KVM to prevent TOCTOU and
+	 * double-fetch types of vulnerabilities.
+	 */
+	struct kvm_regs regs;
+	struct kvm_sregs sregs;
+	struct kvm_vcpu_events events;
 };
 
 #define KVM_X86_QUIRK_LINT0_REENABLED	(1 << 0)
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 3b6b7ee9fa8f..967cc1388770 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -102,6 +102,8 @@ static void update_cr8_intercept(struct kvm_vcpu *vcpu);
 static void process_nmi(struct kvm_vcpu *vcpu);
 static void enter_smm(struct kvm_vcpu *vcpu);
 static void __kvm_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags);
+static void store_regs(struct kvm_vcpu *vcpu);
+static int sync_regs(struct kvm_vcpu *vcpu);
 
 struct kvm_x86_ops *kvm_x86_ops __read_mostly;
 EXPORT_SYMBOL_GPL(kvm_x86_ops);
@@ -2829,6 +2831,9 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
 	case KVM_CAP_GET_MSR_FEATURES:
 		r = 1;
 		break;
+	case KVM_CAP_SYNC_REGS:
+		r = KVM_SYNC_X86_VALID_FIELDS;
+		break;
 	case KVM_CAP_ADJUST_CLOCK:
 		r = KVM_CLOCK_TSC_STABLE;
 		break;
@@ -7510,7 +7515,6 @@ static int complete_emulated_mmio(struct kvm_vcpu *vcpu)
 	return 0;
 }
 
-
 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
 {
 	int r;
@@ -7536,6 +7540,17 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
 		goto out;
 	}
 
+	if (vcpu->run->kvm_valid_regs & ~KVM_SYNC_X86_VALID_FIELDS) {
+		r = -EINVAL;
+		goto out;
+	}
+
+	if (vcpu->run->kvm_dirty_regs) {
+		r = sync_regs(vcpu);
+		if (r != 0)
+			goto out;
+	}
+
 	/* re-sync apic's tpr */
 	if (!lapic_in_kernel(vcpu)) {
 		if (kvm_set_cr8(vcpu, kvm_run->cr8) != 0) {
@@ -7560,6 +7575,8 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
 
 out:
 	kvm_put_guest_fpu(vcpu);
+	if (vcpu->run->kvm_valid_regs)
+		store_regs(vcpu);
 	post_kvm_run_save(vcpu);
 	kvm_sigset_deactivate(vcpu);
 
@@ -7567,10 +7584,8 @@ out:
 	return r;
 }
 
-int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
+static void __get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
 {
-	vcpu_load(vcpu);
-
 	if (vcpu->arch.emulate_regs_need_sync_to_vcpu) {
 		/*
 		 * We are here if userspace calls get_regs() in the middle of
@@ -7603,15 +7618,18 @@ int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
 
 	regs->rip = kvm_rip_read(vcpu);
 	regs->rflags = kvm_get_rflags(vcpu);
+}
 
+int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
+{
+	vcpu_load(vcpu);
+	__get_regs(vcpu, regs);
 	vcpu_put(vcpu);
 	return 0;
 }
 
-int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
+static void __set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
 {
-	vcpu_load(vcpu);
-
 	vcpu->arch.emulate_regs_need_sync_from_vcpu = true;
 	vcpu->arch.emulate_regs_need_sync_to_vcpu = false;
 
@@ -7640,7 +7658,12 @@ int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
 	vcpu->arch.exception.pending = false;
 
 	kvm_make_request(KVM_REQ_EVENT, vcpu);
+}
 
+int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs)
+{
+	vcpu_load(vcpu);
+	__set_regs(vcpu, regs);
 	vcpu_put(vcpu);
 	return 0;
 }
@@ -7655,13 +7678,10 @@ void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l)
 }
 EXPORT_SYMBOL_GPL(kvm_get_cs_db_l_bits);
 
-int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
-				  struct kvm_sregs *sregs)
+static void __get_sregs(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
 {
 	struct desc_ptr dt;
 
-	vcpu_load(vcpu);
-
 	kvm_get_segment(vcpu, &sregs->cs, VCPU_SREG_CS);
 	kvm_get_segment(vcpu, &sregs->ds, VCPU_SREG_DS);
 	kvm_get_segment(vcpu, &sregs->es, VCPU_SREG_ES);
@@ -7692,7 +7712,13 @@ int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
 	if (vcpu->arch.interrupt.pending && !vcpu->arch.interrupt.soft)
 		set_bit(vcpu->arch.interrupt.nr,
 			(unsigned long *)sregs->interrupt_bitmap);
+}
 
+int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
+				  struct kvm_sregs *sregs)
+{
+	vcpu_load(vcpu);
+	__get_sregs(vcpu, sregs);
 	vcpu_put(vcpu);
 	return 0;
 }
@@ -7787,8 +7813,7 @@ int kvm_valid_sregs(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
 	return 0;
 }
 
-int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
-				  struct kvm_sregs *sregs)
+static int __set_sregs(struct kvm_vcpu *vcpu, struct kvm_sregs *sregs)
 {
 	struct msr_data apic_base_msr;
 	int mmu_reset_needed = 0;
@@ -7796,8 +7821,6 @@ int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
 	struct desc_ptr dt;
 	int ret = -EINVAL;
 
-	vcpu_load(vcpu);
-
 	if (!guest_cpuid_has(vcpu, X86_FEATURE_XSAVE) &&
 			(sregs->cr4 & X86_CR4_OSXSAVE))
 		goto out;
@@ -7876,6 +7899,16 @@ int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
 
 	ret = 0;
 out:
+	return ret;
+}
+
+int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
+				  struct kvm_sregs *sregs)
+{
+	int ret;
+
+	vcpu_load(vcpu);
+	ret = __set_sregs(vcpu, sregs);
 	vcpu_put(vcpu);
 	return ret;
 }
@@ -8002,6 +8035,45 @@ int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
 	return 0;
 }
 
+static void store_regs(struct kvm_vcpu *vcpu)
+{
+	BUILD_BUG_ON(sizeof(struct kvm_sync_regs) > SYNC_REGS_SIZE_BYTES);
+
+	if (vcpu->run->kvm_valid_regs & KVM_SYNC_X86_REGS)
+		__get_regs(vcpu, &vcpu->run->s.regs.regs);
+
+	if (vcpu->run->kvm_valid_regs & KVM_SYNC_X86_SREGS)
+		__get_sregs(vcpu, &vcpu->run->s.regs.sregs);
+
+	if (vcpu->run->kvm_valid_regs & KVM_SYNC_X86_EVENTS)
+		kvm_vcpu_ioctl_x86_get_vcpu_events(
+				vcpu, &vcpu->run->s.regs.events);
+}
+
+static int sync_regs(struct kvm_vcpu *vcpu)
+{
+	if (vcpu->run->kvm_dirty_regs & ~KVM_SYNC_X86_VALID_FIELDS)
+		return -EINVAL;
+
+	if (vcpu->run->kvm_dirty_regs & KVM_SYNC_X86_REGS) {
+		__set_regs(vcpu, &vcpu->run->s.regs.regs);
+		vcpu->run->kvm_dirty_regs &= ~KVM_SYNC_X86_REGS;
+	}
+	if (vcpu->run->kvm_dirty_regs & KVM_SYNC_X86_SREGS) {
+		if (__set_sregs(vcpu, &vcpu->run->s.regs.sregs))
+			return -EINVAL;
+		vcpu->run->kvm_dirty_regs &= ~KVM_SYNC_X86_SREGS;
+	}
+	if (vcpu->run->kvm_dirty_regs & KVM_SYNC_X86_EVENTS) {
+		if (kvm_vcpu_ioctl_x86_set_vcpu_events(
+				vcpu, &vcpu->run->s.regs.events))
+			return -EINVAL;
+		vcpu->run->kvm_dirty_regs &= ~KVM_SYNC_X86_EVENTS;
+	}
+
+	return 0;
+}
+
 static void fx_init(struct kvm_vcpu *vcpu)
 {
 	fpstate_init(&vcpu->arch.guest_fpu.state);