KVM: s390: add floating irq controller

This patch adds a floating irq controller as a kvm_device.
It will be necessary for migration of floating interrupts as well
as for hardening the reset code by allowing user space to explicitly
remove all pending floating interrupts.
Signed-off-by: Jens Freimann <jfrei@linux.vnet.ibm.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>

Documentation/virtual/kvm/devices/s390_flic.txt

+FLIC (floating interrupt controller)
+====================================
+
+FLIC handles floating (non per-cpu) interrupts, i.e. I/O, service and some
+machine check interruptions. All interrupts are stored in a per-vm list of
+pending interrupts. FLIC performs operations on this list.
+
+Only one FLIC instance may be instantiated.
+
+FLIC provides support to
+- add interrupts (KVM_DEV_FLIC_ENQUEUE)
+- inspect currently pending interrupts (KVM_FLIC_GET_ALL_IRQS)
+- purge all pending floating interrupts (KVM_DEV_FLIC_CLEAR_IRQS)
+
+Groups:
+  KVM_DEV_FLIC_ENQUEUE
+    Passes a buffer and length into the kernel which are then injected into
+    the list of pending interrupts.
+    attr->addr contains the pointer to the buffer and attr->attr contains
+    the length of the buffer.
+    The format of the data structure kvm_s390_irq as it is copied from userspace
+    is defined in usr/include/linux/kvm.h.
+
+  KVM_DEV_FLIC_GET_ALL_IRQS
+    Copies all floating interrupts into a buffer provided by userspace.
+    When the buffer is too small it returns -ENOMEM, which is the indication
+    for userspace to try again with a bigger buffer.
+    All interrupts remain pending, i.e. are not deleted from the list of
+    currently pending interrupts.
+    attr->addr contains the userspace address of the buffer into which all
+    interrupt data will be copied.
+    attr->attr contains the size of the buffer in bytes.
+
+  KVM_DEV_FLIC_CLEAR_IRQS
+    Simply deletes all elements from the list of currently pending floating
+    interrupts.  No interrupts are injected into the guest.

arch/s390/include/asm/kvm_host.h

@@ -243,6 +243,7 @@ struct kvm_arch{
 	struct sca_block *sca;
 	debug_info_t *dbf;
 	struct kvm_s390_float_interrupt float_int;
+	struct kvm_device *flic;
 	struct gmap *gmap;
 	int css_support;
 };

arch/s390/include/uapi/asm/kvm.h

@@ -16,6 +16,20 @@
 
 #define __KVM_S390
 
+/* Device control API: s390-specific devices */
+#define KVM_DEV_FLIC_GET_ALL_IRQS	1
+#define KVM_DEV_FLIC_ENQUEUE		2
+#define KVM_DEV_FLIC_CLEAR_IRQS		3
+/*
+ * We can have up to 4*64k pending subchannels + 8 adapter interrupts,
+ * as well as up  to ASYNC_PF_PER_VCPU*KVM_MAX_VCPUS pfault done interrupts.
+ * There are also sclp and machine checks. This gives us
+ * sizeof(kvm_s390_irq)*(4*65536+8+64*64+1+1) = 72 * 266250 = 19170000
+ * Lets round up to 8192 pages.
+ */
+
+#define KVM_S390_FLIC_MAX_BUFFER	0x2000000
+
 /* for KVM_GET_REGS and KVM_SET_REGS */
 struct kvm_regs {
 	/* general purpose regs for s390 */

arch/s390/kvm/interrupt.c

@@ -659,53 +659,86 @@ struct kvm_s390_interrupt_info *kvm_s390_get_io_int(struct kvm *kvm,
 	return inti;
 }
 
-int kvm_s390_inject_vm(struct kvm *kvm,
-		       struct kvm_s390_interrupt *s390int)
+static void __inject_vm(struct kvm *kvm, struct kvm_s390_interrupt_info *inti)
 {
 	struct kvm_s390_local_interrupt *li;
 	struct kvm_s390_float_interrupt *fi;
-	struct kvm_s390_interrupt_info *inti, *iter;
+	struct kvm_s390_interrupt_info *iter;
 	int sigcpu;
 
+	mutex_lock(&kvm->lock);
+	fi = &kvm->arch.float_int;
+	spin_lock(&fi->lock);
+	if (!is_ioint(inti->type)) {
+		list_add_tail(&inti->list, &fi->list);
+	} else {
+		u64 isc_bits = int_word_to_isc_bits(inti->io.io_int_word);
+
+		/* Keep I/O interrupts sorted in isc order. */
+		list_for_each_entry(iter, &fi->list, list) {
+			if (!is_ioint(iter->type))
+				continue;
+			if (int_word_to_isc_bits(iter->io.io_int_word)
+			    <= isc_bits)
+				continue;
+			break;
+		}
+		list_add_tail(&inti->list, &iter->list);
+	}
+	atomic_set(&fi->active, 1);
+	sigcpu = find_first_bit(fi->idle_mask, KVM_MAX_VCPUS);
+	if (sigcpu == KVM_MAX_VCPUS) {
+		do {
+			sigcpu = fi->next_rr_cpu++;
+			if (sigcpu == KVM_MAX_VCPUS)
+				sigcpu = fi->next_rr_cpu = 0;
+		} while (fi->local_int[sigcpu] == NULL);
+	}
+	li = fi->local_int[sigcpu];
+	spin_lock_bh(&li->lock);
+	atomic_set_mask(CPUSTAT_EXT_INT, li->cpuflags);
+	if (waitqueue_active(li->wq))
+		wake_up_interruptible(li->wq);
+	spin_unlock_bh(&li->lock);
+	spin_unlock(&fi->lock);
+	mutex_unlock(&kvm->lock);
+}
+
+int kvm_s390_inject_vm(struct kvm *kvm,
+		       struct kvm_s390_interrupt *s390int)
+{
+	struct kvm_s390_interrupt_info *inti;
+
 	inti = kzalloc(sizeof(*inti), GFP_KERNEL);
 	if (!inti)
 		return -ENOMEM;
 
-	switch (s390int->type) {
+	inti->type = s390int->type;
+	switch (inti->type) {
 	case KVM_S390_INT_VIRTIO:
 		VM_EVENT(kvm, 5, "inject: virtio parm:%x,parm64:%llx",
 			 s390int->parm, s390int->parm64);
-		inti->type = s390int->type;
 		inti->ext.ext_params = s390int->parm;
 		inti->ext.ext_params2 = s390int->parm64;
 		break;
 	case KVM_S390_INT_SERVICE:
 		VM_EVENT(kvm, 5, "inject: sclp parm:%x", s390int->parm);
-		inti->type = s390int->type;
 		inti->ext.ext_params = s390int->parm;
 		break;
-	case KVM_S390_PROGRAM_INT:
-	case KVM_S390_SIGP_STOP:
-	case KVM_S390_INT_EXTERNAL_CALL:
-	case KVM_S390_INT_EMERGENCY:
-		kfree(inti);
-		return -EINVAL;
 	case KVM_S390_MCHK:
 		VM_EVENT(kvm, 5, "inject: machine check parm64:%llx",
 			 s390int->parm64);
-		inti->type = s390int->type;
 		inti->mchk.cr14 = s390int->parm; /* upper bits are not used */
 		inti->mchk.mcic = s390int->parm64;
 		break;
 	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
-		if (s390int->type & IOINT_AI_MASK)
+		if (inti->type & IOINT_AI_MASK)
 			VM_EVENT(kvm, 5, "%s", "inject: I/O (AI)");
 		else
 			VM_EVENT(kvm, 5, "inject: I/O css %x ss %x schid %04x",
 				 s390int->type & IOINT_CSSID_MASK,
 				 s390int->type & IOINT_SSID_MASK,
 				 s390int->type & IOINT_SCHID_MASK);
-		inti->type = s390int->type;
 		inti->io.subchannel_id = s390int->parm >> 16;
 		inti->io.subchannel_nr = s390int->parm & 0x0000ffffu;
 		inti->io.io_int_parm = s390int->parm64 >> 32;
@@ -718,42 +751,7 @@ int kvm_s390_inject_vm(struct kvm *kvm,
 	trace_kvm_s390_inject_vm(s390int->type, s390int->parm, s390int->parm64,
 				 2);
 
-	mutex_lock(&kvm->lock);
-	fi = &kvm->arch.float_int;
-	spin_lock(&fi->lock);
-	if (!is_ioint(inti->type))
-		list_add_tail(&inti->list, &fi->list);
-	else {
-		u64 isc_bits = int_word_to_isc_bits(inti->io.io_int_word);
-
-		/* Keep I/O interrupts sorted in isc order. */
-		list_for_each_entry(iter, &fi->list, list) {
-			if (!is_ioint(iter->type))
-				continue;
-			if (int_word_to_isc_bits(iter->io.io_int_word)
-			    <= isc_bits)
-				continue;
-			break;
-		}
-		list_add_tail(&inti->list, &iter->list);
-	}
-	atomic_set(&fi->active, 1);
-	sigcpu = find_first_bit(fi->idle_mask, KVM_MAX_VCPUS);
-	if (sigcpu == KVM_MAX_VCPUS) {
-		do {
-			sigcpu = fi->next_rr_cpu++;
-			if (sigcpu == KVM_MAX_VCPUS)
-				sigcpu = fi->next_rr_cpu = 0;
-		} while (fi->local_int[sigcpu] == NULL);
-	}
-	li = fi->local_int[sigcpu];
-	spin_lock_bh(&li->lock);
-	atomic_set_mask(CPUSTAT_EXT_INT, li->cpuflags);
-	if (waitqueue_active(li->wq))
-		wake_up_interruptible(li->wq);
-	spin_unlock_bh(&li->lock);
-	spin_unlock(&fi->lock);
-	mutex_unlock(&kvm->lock);
+	__inject_vm(kvm, inti);
 	return 0;
 }
 
@@ -841,3 +839,207 @@ int kvm_s390_inject_vcpu(struct kvm_vcpu *vcpu,
 	mutex_unlock(&vcpu->kvm->lock);
 	return 0;
 }
+
+static void clear_floating_interrupts(struct kvm *kvm)
+{
+	struct kvm_s390_float_interrupt *fi;
+	struct kvm_s390_interrupt_info	*n, *inti = NULL;
+
+	mutex_lock(&kvm->lock);
+	fi = &kvm->arch.float_int;
+	spin_lock(&fi->lock);
+	list_for_each_entry_safe(inti, n, &fi->list, list) {
+		list_del(&inti->list);
+		kfree(inti);
+	}
+	atomic_set(&fi->active, 0);
+	spin_unlock(&fi->lock);
+	mutex_unlock(&kvm->lock);
+}
+
+static inline int copy_irq_to_user(struct kvm_s390_interrupt_info *inti,
+				   u8 *addr)
+{
+	struct kvm_s390_irq __user *uptr = (struct kvm_s390_irq __user *) addr;
+	struct kvm_s390_irq irq = {0};
+
+	irq.type = inti->type;
+	switch (inti->type) {
+	case KVM_S390_INT_VIRTIO:
+	case KVM_S390_INT_SERVICE:
+		irq.u.ext = inti->ext;
+		break;
+	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
+		irq.u.io = inti->io;
+		break;
+	case KVM_S390_MCHK:
+		irq.u.mchk = inti->mchk;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	if (copy_to_user(uptr, &irq, sizeof(irq)))
+		return -EFAULT;
+
+	return 0;
+}
+
+static int get_all_floating_irqs(struct kvm *kvm, __u8 *buf, __u64 len)
+{
+	struct kvm_s390_interrupt_info *inti;
+	struct kvm_s390_float_interrupt *fi;
+	int ret = 0;
+	int n = 0;
+
+	mutex_lock(&kvm->lock);
+	fi = &kvm->arch.float_int;
+	spin_lock(&fi->lock);
+
+	list_for_each_entry(inti, &fi->list, list) {
+		if (len < sizeof(struct kvm_s390_irq)) {
+			/* signal userspace to try again */
+			ret = -ENOMEM;
+			break;
+		}
+		ret = copy_irq_to_user(inti, buf);
+		if (ret)
+			break;
+		buf += sizeof(struct kvm_s390_irq);
+		len -= sizeof(struct kvm_s390_irq);
+		n++;
+	}
+
+	spin_unlock(&fi->lock);
+	mutex_unlock(&kvm->lock);
+
+	return ret < 0 ? ret : n;
+}
+
+static int flic_get_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
+{
+	int r;
+
+	switch (attr->group) {
+	case KVM_DEV_FLIC_GET_ALL_IRQS:
+		r = get_all_floating_irqs(dev->kvm, (u8 *) attr->addr,
+					  attr->attr);
+		break;
+	default:
+		r = -EINVAL;
+	}
+
+	return r;
+}
+
+static inline int copy_irq_from_user(struct kvm_s390_interrupt_info *inti,
+				     u64 addr)
+{
+	struct kvm_s390_irq __user *uptr = (struct kvm_s390_irq __user *) addr;
+	void *target = NULL;
+	void __user *source;
+	u64 size;
+
+	if (get_user(inti->type, (u64 __user *)addr))
+		return -EFAULT;
+
+	switch (inti->type) {
+	case KVM_S390_INT_VIRTIO:
+	case KVM_S390_INT_SERVICE:
+		target = (void *) &inti->ext;
+		source = &uptr->u.ext;
+		size = sizeof(inti->ext);
+		break;
+	case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
+		target = (void *) &inti->io;
+		source = &uptr->u.io;
+		size = sizeof(inti->io);
+		break;
+	case KVM_S390_MCHK:
+		target = (void *) &inti->mchk;
+		source = &uptr->u.mchk;
+		size = sizeof(inti->mchk);
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	if (copy_from_user(target, source, size))
+		return -EFAULT;
+
+	return 0;
+}
+
+static int enqueue_floating_irq(struct kvm_device *dev,
+				struct kvm_device_attr *attr)
+{
+	struct kvm_s390_interrupt_info *inti = NULL;
+	int r = 0;
+	int len = attr->attr;
+
+	if (len % sizeof(struct kvm_s390_irq) != 0)
+		return -EINVAL;
+	else if (len > KVM_S390_FLIC_MAX_BUFFER)
+		return -EINVAL;
+
+	while (len >= sizeof(struct kvm_s390_irq)) {
+		inti = kzalloc(sizeof(*inti), GFP_KERNEL);
+		if (!inti)
+			return -ENOMEM;
+
+		r = copy_irq_from_user(inti, attr->addr);
+		if (r) {
+			kfree(inti);
+			return r;
+		}
+		__inject_vm(dev->kvm, inti);
+		len -= sizeof(struct kvm_s390_irq);
+		attr->addr += sizeof(struct kvm_s390_irq);
+	}
+
+	return r;
+}
+
+static int flic_set_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
+{
+	int r = 0;
+
+	switch (attr->group) {
+	case KVM_DEV_FLIC_ENQUEUE:
+		r = enqueue_floating_irq(dev, attr);
+		break;
+	case KVM_DEV_FLIC_CLEAR_IRQS:
+		r = 0;
+		clear_floating_interrupts(dev->kvm);
+		break;
+	default:
+		r = -EINVAL;
+	}
+
+	return r;
+}
+
+static int flic_create(struct kvm_device *dev, u32 type)
+{
+	if (!dev)
+		return -EINVAL;
+	if (dev->kvm->arch.flic)
+		return -EINVAL;
+	dev->kvm->arch.flic = dev;
+	return 0;
+}
+
+static void flic_destroy(struct kvm_device *dev)
+{
+	dev->kvm->arch.flic = NULL;
+	kfree(dev);
+}
+
+/* s390 floating irq controller (flic) */
+struct kvm_device_ops kvm_flic_ops = {
+	.name = "kvm-flic",
+	.get_attr = flic_get_attr,
+	.set_attr = flic_set_attr,
+	.create = flic_create,
+	.destroy = flic_destroy,
+};

arch/s390/kvm/kvm-s390.c

@@ -157,6 +157,7 @@ int kvm_dev_ioctl_check_extension(long ext)
 	case KVM_CAP_ENABLE_CAP:
 	case KVM_CAP_S390_CSS_SUPPORT:
 	case KVM_CAP_IOEVENTFD:
+	case KVM_CAP_DEVICE_CTRL:
 		r = 1;
 		break;
 	case KVM_CAP_NR_VCPUS:

include/linux/kvm_host.h

@@ -1064,6 +1064,7 @@ extern struct kvm_device_ops kvm_mpic_ops;
 extern struct kvm_device_ops kvm_xics_ops;
 extern struct kvm_device_ops kvm_vfio_ops;
 extern struct kvm_device_ops kvm_arm_vgic_v2_ops;
+extern struct kvm_device_ops kvm_flic_ops;
 
 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
 

include/uapi/linux/kvm.h

@@ -918,6 +918,7 @@ struct kvm_device_attr {
 #define   KVM_DEV_VFIO_GROUP_ADD			1
 #define   KVM_DEV_VFIO_GROUP_DEL			2
 #define KVM_DEV_TYPE_ARM_VGIC_V2	5
+#define KVM_DEV_TYPE_FLIC		6
 
 /*
  * ioctls for VM fds

virt/kvm/kvm_main.c

@@ -2283,6 +2283,11 @@ static int kvm_ioctl_create_device(struct kvm *kvm,
 	case KVM_DEV_TYPE_ARM_VGIC_V2:
 		ops = &kvm_arm_vgic_v2_ops;
 		break;
+#endif
+#ifdef CONFIG_S390
+	case KVM_DEV_TYPE_FLIC:
+		ops = &kvm_flic_ops;
+		break;
 #endif
 	default:
 		return -ENODEV;