KVM: x86: XSAVE/XRSTOR live migration support

This patch enable save/restore of xsave state.

Signed-off-by: Sheng Yang <sheng@linux.intel.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>

Documentation/kvm/api.txt

@@ -922,6 +922,80 @@ Define which vcpu is the Bootstrap Processor (BSP).  Values are the same
 as the vcpu id in KVM_CREATE_VCPU.  If this ioctl is not called, the default
 is vcpu 0.
 
+4.41 KVM_GET_XSAVE
+
+Capability: KVM_CAP_XSAVE
+Architectures: x86
+Type: vcpu ioctl
+Parameters: struct kvm_xsave (out)
+Returns: 0 on success, -1 on error
+
+struct kvm_xsave {
+	__u32 region[1024];
+};
+
+This ioctl would copy current vcpu's xsave struct to the userspace.
+
+4.42 KVM_SET_XSAVE
+
+Capability: KVM_CAP_XSAVE
+Architectures: x86
+Type: vcpu ioctl
+Parameters: struct kvm_xsave (in)
+Returns: 0 on success, -1 on error
+
+struct kvm_xsave {
+	__u32 region[1024];
+};
+
+This ioctl would copy userspace's xsave struct to the kernel.
+
+4.43 KVM_GET_XCRS
+
+Capability: KVM_CAP_XCRS
+Architectures: x86
+Type: vcpu ioctl
+Parameters: struct kvm_xcrs (out)
+Returns: 0 on success, -1 on error
+
+struct kvm_xcr {
+	__u32 xcr;
+	__u32 reserved;
+	__u64 value;
+};
+
+struct kvm_xcrs {
+	__u32 nr_xcrs;
+	__u32 flags;
+	struct kvm_xcr xcrs[KVM_MAX_XCRS];
+	__u64 padding[16];
+};
+
+This ioctl would copy current vcpu's xcrs to the userspace.
+
+4.44 KVM_SET_XCRS
+
+Capability: KVM_CAP_XCRS
+Architectures: x86
+Type: vcpu ioctl
+Parameters: struct kvm_xcrs (in)
+Returns: 0 on success, -1 on error
+
+struct kvm_xcr {
+	__u32 xcr;
+	__u32 reserved;
+	__u64 value;
+};
+
+struct kvm_xcrs {
+	__u32 nr_xcrs;
+	__u32 flags;
+	struct kvm_xcr xcrs[KVM_MAX_XCRS];
+	__u64 padding[16];
+};
+
+This ioctl would set vcpu's xcr to the value userspace specified.
+
 5. The kvm_run structure
 
 Application code obtains a pointer to the kvm_run structure by

arch/x86/include/asm/kvm.h

@@ -22,6 +22,8 @@
 #define __KVM_HAVE_XEN_HVM
 #define __KVM_HAVE_VCPU_EVENTS
 #define __KVM_HAVE_DEBUGREGS
+#define __KVM_HAVE_XSAVE
+#define __KVM_HAVE_XCRS
 
 /* Architectural interrupt line count. */
 #define KVM_NR_INTERRUPTS 256
@@ -299,4 +301,24 @@ struct kvm_debugregs {
 	__u64 reserved[9];
 };
 
+/* for KVM_CAP_XSAVE */
+struct kvm_xsave {
+	__u32 region[1024];
+};
+
+#define KVM_MAX_XCRS	16
+
+struct kvm_xcr {
+	__u32 xcr;
+	__u32 reserved;
+	__u64 value;
+};
+
+struct kvm_xcrs {
+	__u32 nr_xcrs;
+	__u32 flags;
+	struct kvm_xcr xcrs[KVM_MAX_XCRS];
+	__u64 padding[16];
+};
+
 #endif /* _ASM_X86_KVM_H */

arch/x86/include/asm/xsave.h

@@ -13,8 +13,11 @@
 
 #define FXSAVE_SIZE	512
 
-#define XSTATE_YMM_SIZE 256
+#define XSAVE_HDR_SIZE	    64
-#define XSTATE_YMM_OFFSET (512 + 64)
+#define XSAVE_HDR_OFFSET    FXSAVE_SIZE
+
+#define XSAVE_YMM_SIZE	    256
+#define XSAVE_YMM_OFFSET    (XSAVE_HDR_SIZE + XSAVE_HDR_OFFSET)
 
 /*
  * These are the features that the OS can handle currently.

arch/x86/kvm/x86.c

@@ -1680,6 +1680,7 @@ int kvm_dev_ioctl_check_extension(long ext)
 	case KVM_CAP_PCI_SEGMENT:
 	case KVM_CAP_DEBUGREGS:
 	case KVM_CAP_X86_ROBUST_SINGLESTEP:
+	case KVM_CAP_XSAVE:
 		r = 1;
 		break;
 	case KVM_CAP_COALESCED_MMIO:
@@ -1703,6 +1704,9 @@ int kvm_dev_ioctl_check_extension(long ext)
 	case KVM_CAP_MCE:
 		r = KVM_MAX_MCE_BANKS;
 		break;
+	case KVM_CAP_XCRS:
+		r = cpu_has_xsave;
+		break;
 	default:
 		r = 0;
 		break;
@@ -2355,6 +2359,77 @@ static int kvm_vcpu_ioctl_x86_set_debugregs(struct kvm_vcpu *vcpu,
 	return 0;
 }
 
+static void kvm_vcpu_ioctl_x86_get_xsave(struct kvm_vcpu *vcpu,
+					 struct kvm_xsave *guest_xsave)
+{
+	if (cpu_has_xsave)
+		memcpy(guest_xsave->region,
+			&vcpu->arch.guest_fpu.state->xsave,
+			sizeof(struct xsave_struct));
+	else {
+		memcpy(guest_xsave->region,
+			&vcpu->arch.guest_fpu.state->fxsave,
+			sizeof(struct i387_fxsave_struct));
+		*(u64 *)&guest_xsave->region[XSAVE_HDR_OFFSET / sizeof(u32)] =
+			XSTATE_FPSSE;
+	}
+}
+
+static int kvm_vcpu_ioctl_x86_set_xsave(struct kvm_vcpu *vcpu,
+					struct kvm_xsave *guest_xsave)
+{
+	u64 xstate_bv =
+		*(u64 *)&guest_xsave->region[XSAVE_HDR_OFFSET / sizeof(u32)];
+
+	if (cpu_has_xsave)
+		memcpy(&vcpu->arch.guest_fpu.state->xsave,
+			guest_xsave->region, sizeof(struct xsave_struct));
+	else {
+		if (xstate_bv & ~XSTATE_FPSSE)
+			return -EINVAL;
+		memcpy(&vcpu->arch.guest_fpu.state->fxsave,
+			guest_xsave->region, sizeof(struct i387_fxsave_struct));
+	}
+	return 0;
+}
+
+static void kvm_vcpu_ioctl_x86_get_xcrs(struct kvm_vcpu *vcpu,
+					struct kvm_xcrs *guest_xcrs)
+{
+	if (!cpu_has_xsave) {
+		guest_xcrs->nr_xcrs = 0;
+		return;
+	}
+
+	guest_xcrs->nr_xcrs = 1;
+	guest_xcrs->flags = 0;
+	guest_xcrs->xcrs[0].xcr = XCR_XFEATURE_ENABLED_MASK;
+	guest_xcrs->xcrs[0].value = vcpu->arch.xcr0;
+}
+
+static int kvm_vcpu_ioctl_x86_set_xcrs(struct kvm_vcpu *vcpu,
+				       struct kvm_xcrs *guest_xcrs)
+{
+	int i, r = 0;
+
+	if (!cpu_has_xsave)
+		return -EINVAL;
+
+	if (guest_xcrs->nr_xcrs > KVM_MAX_XCRS || guest_xcrs->flags)
+		return -EINVAL;
+
+	for (i = 0; i < guest_xcrs->nr_xcrs; i++)
+		/* Only support XCR0 currently */
+		if (guest_xcrs->xcrs[0].xcr == XCR_XFEATURE_ENABLED_MASK) {
+			r = __kvm_set_xcr(vcpu, XCR_XFEATURE_ENABLED_MASK,
+				guest_xcrs->xcrs[0].value);
+			break;
+		}
+	if (r)
+		r = -EINVAL;
+	return r;
+}
+
 long kvm_arch_vcpu_ioctl(struct file *filp,
 			 unsigned int ioctl, unsigned long arg)
 {
@@ -2556,6 +2631,70 @@ long kvm_arch_vcpu_ioctl(struct file *filp,
 		r = kvm_vcpu_ioctl_x86_set_debugregs(vcpu, &dbgregs);
 		break;
 	}
+	case KVM_GET_XSAVE: {
+		struct kvm_xsave *xsave;
+
+		xsave = kzalloc(sizeof(struct kvm_xsave), GFP_KERNEL);
+		r = -ENOMEM;
+		if (!xsave)
+			break;
+
+		kvm_vcpu_ioctl_x86_get_xsave(vcpu, xsave);
+
+		r = -EFAULT;
+		if (copy_to_user(argp, xsave, sizeof(struct kvm_xsave)))
+			break;
+		r = 0;
+		break;
+	}
+	case KVM_SET_XSAVE: {
+		struct kvm_xsave *xsave;
+
+		xsave = kzalloc(sizeof(struct kvm_xsave), GFP_KERNEL);
+		r = -ENOMEM;
+		if (!xsave)
+			break;
+
+		r = -EFAULT;
+		if (copy_from_user(xsave, argp, sizeof(struct kvm_xsave)))
+			break;
+
+		r = kvm_vcpu_ioctl_x86_set_xsave(vcpu, xsave);
+		break;
+	}
+	case KVM_GET_XCRS: {
+		struct kvm_xcrs *xcrs;
+
+		xcrs = kzalloc(sizeof(struct kvm_xcrs), GFP_KERNEL);
+		r = -ENOMEM;
+		if (!xcrs)
+			break;
+
+		kvm_vcpu_ioctl_x86_get_xcrs(vcpu, xcrs);
+
+		r = -EFAULT;
+		if (copy_to_user(argp, xcrs,
+				 sizeof(struct kvm_xcrs)))
+			break;
+		r = 0;
+		break;
+	}
+	case KVM_SET_XCRS: {
+		struct kvm_xcrs *xcrs;
+
+		xcrs = kzalloc(sizeof(struct kvm_xcrs), GFP_KERNEL);
+		r = -ENOMEM;
+		if (!xcrs)
+			break;
+
+		r = -EFAULT;
+		if (copy_from_user(xcrs, argp,
+				   sizeof(struct kvm_xcrs)))
+			break;
+
+		r = kvm_vcpu_ioctl_x86_set_xcrs(vcpu, xcrs);
+		break;
+	}
 	default:
 		r = -EINVAL;
 	}

include/linux/kvm.h

@@ -524,6 +524,12 @@ struct kvm_enable_cap {
 #define KVM_CAP_PPC_OSI 52
 #define KVM_CAP_PPC_UNSET_IRQ 53
 #define KVM_CAP_ENABLE_CAP 54
+#ifdef __KVM_HAVE_XSAVE
+#define KVM_CAP_XSAVE 55
+#endif
+#ifdef __KVM_HAVE_XCRS
+#define KVM_CAP_XCRS 56
+#endif
 
 #ifdef KVM_CAP_IRQ_ROUTING
 
@@ -714,6 +720,12 @@ struct kvm_clock_data {
 #define KVM_GET_DEBUGREGS         _IOR(KVMIO,  0xa1, struct kvm_debugregs)
 #define KVM_SET_DEBUGREGS         _IOW(KVMIO,  0xa2, struct kvm_debugregs)
 #define KVM_ENABLE_CAP            _IOW(KVMIO,  0xa3, struct kvm_enable_cap)
+/* Available with KVM_CAP_XSAVE */
+#define KVM_GET_XSAVE		  _IOR(KVMIO,  0xa4, struct kvm_xsave)
+#define KVM_SET_XSAVE		  _IOW(KVMIO,  0xa5, struct kvm_xsave)
+/* Available with KVM_CAP_XCRS */
+#define KVM_GET_XCRS		  _IOR(KVMIO,  0xa6, struct kvm_xcrs)
+#define KVM_SET_XCRS		  _IOW(KVMIO,  0xa7, struct kvm_xcrs)
 
 #define KVM_DEV_ASSIGN_ENABLE_IOMMU	(1 << 0)