kvm: add device control API

Currently, devices that are emulated inside KVM are configured in a
hardcoded manner based on an assumption that any given architecture
only has one way to do it. If there's any need to access device state,
it is done through inflexible one-purpose-only IOCTLs (e.g.
KVM_GET/SET_LAPIC). Defining new IOCTLs for every little thing is
cumbersome and depletes a limited numberspace.

This API provides a mechanism to instantiate a device of a certain
type, returning an ID that can be used to set/get attributes of the
device. Attributes may include configuration parameters (e.g.
register base address), device state, operational commands, etc. It
is similar to the ONE_REG API, except that it acts on devices rather
than vcpus.

Both device types and individual attributes can be tested without having
to create the device or get/set the attribute, without the need for
separately managing enumerated capabilities.

Signed-off-by: Scott Wood <[e-mail]>
Signed-off-by: Alexander Graf <[e-mail]>

Documentation/virtual/kvm/api.txt

···
 written, then `n_invalid' invalid entries, invalidating any previously
 valid entries found.
 
+4.79 KVM_CREATE_DEVICE
+
+Capability: KVM_CAP_DEVICE_CTRL
+Type: vm ioctl
+Parameters: struct kvm_create_device (in/out)
+Returns: 0 on success, -1 on error
+Errors:
+  ENODEV: The device type is unknown or unsupported
+  EEXIST: Device already created, and this type of device may not
+          be instantiated multiple times
+
+  Other error conditions may be defined by individual device types or
+  have their standard meanings.
+
+Creates an emulated device in the kernel.  The file descriptor returned
+in fd can be used with KVM_SET/GET/HAS_DEVICE_ATTR.
+
+If the KVM_CREATE_DEVICE_TEST flag is set, only test whether the
+device type is supported (not necessarily whether it can be created
+in the current vm).
+
+Individual devices should not define flags.  Attributes should be used
+for specifying any behavior that is not implied by the device type
+number.
+
+struct kvm_create_device {
+	__u32	type;	/* in: KVM_DEV_TYPE_xxx */
+	__u32	fd;	/* out: device handle */
+	__u32	flags;	/* in: KVM_CREATE_DEVICE_xxx */
+};
+
+4.80 KVM_SET_DEVICE_ATTR/KVM_GET_DEVICE_ATTR
+
+Capability: KVM_CAP_DEVICE_CTRL
+Type: device ioctl
+Parameters: struct kvm_device_attr
+Returns: 0 on success, -1 on error
+Errors:
+  ENXIO:  The group or attribute is unknown/unsupported for this device
+  EPERM:  The attribute cannot (currently) be accessed this way
+          (e.g. read-only attribute, or attribute that only makes
+          sense when the device is in a different state)
+
+  Other error conditions may be defined by individual device types.
+
+Gets/sets a specified piece of device configuration and/or state.  The
+semantics are device-specific.  See individual device documentation in
+the "devices" directory.  As with ONE_REG, the size of the data
+transferred is defined by the particular attribute.
+
+struct kvm_device_attr {
+	__u32	flags;		/* no flags currently defined */
+	__u32	group;		/* device-defined */
+	__u64	attr;		/* group-defined */
+	__u64	addr;		/* userspace address of attr data */
+};
+
+4.81 KVM_HAS_DEVICE_ATTR
+
+Capability: KVM_CAP_DEVICE_CTRL
+Type: device ioctl
+Parameters: struct kvm_device_attr
+Returns: 0 on success, -1 on error
+Errors:
+  ENXIO:  The group or attribute is unknown/unsupported for this device
+
+Tests whether a device supports a particular attribute.  A successful
+return indicates the attribute is implemented.  It does not necessarily
+indicate that the attribute can be read or written in the device's
+current state.  "addr" is ignored.
 
 4.77 KVM_ARM_VCPU_INIT
 

Documentation/virtual/kvm/devices/README

···
+This directory contains specific device bindings for KVM_CAP_DEVICE_CTRL.

include/linux/kvm_host.h

···
 
 extern bool kvm_rebooting;
 
+struct kvm_device_ops;
+
+struct kvm_device {
+	struct kvm_device_ops *ops;
+	struct kvm *kvm;
+	atomic_t users;
+	void *private;
+};
+
+/* create, destroy, and name are mandatory */
+struct kvm_device_ops {
+	const char *name;
+	int (*create)(struct kvm_device *dev, u32 type);
+
+	/*
+	 * Destroy is responsible for freeing dev.
+	 *
+	 * Destroy may be called before or after destructors are called
+	 * on emulated I/O regions, depending on whether a reference is
+	 * held by a vcpu or other kvm component that gets destroyed
+	 * after the emulated I/O.
+	 */
+	void (*destroy)(struct kvm_device *dev);
+
+	int (*set_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
+	int (*get_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
+	int (*has_attr)(struct kvm_device *dev, struct kvm_device_attr *attr);
+	long (*ioctl)(struct kvm_device *dev, unsigned int ioctl,
+		      unsigned long arg);
+};
+
+void kvm_device_get(struct kvm_device *dev);
+void kvm_device_put(struct kvm_device *dev);
+struct kvm_device *kvm_device_from_filp(struct file *filp);
+
 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
 
 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)

include/uapi/linux/kvm.h

···
 #define KVM_CAP_PPC_EPR 86
 #define KVM_CAP_ARM_PSCI 87
 #define KVM_CAP_ARM_SET_DEVICE_ADDR 88
+#define KVM_CAP_DEVICE_CTRL 89
 
 #ifdef KVM_CAP_IRQ_ROUTING
 
···
 };
 
 /*
+ * Device control API, available with KVM_CAP_DEVICE_CTRL
+ */
+#define KVM_CREATE_DEVICE_TEST		1
+
+struct kvm_create_device {
+	__u32	type;	/* in: KVM_DEV_TYPE_xxx */
+	__u32	fd;	/* out: device handle */
+	__u32	flags;	/* in: KVM_CREATE_DEVICE_xxx */
+};
+
+struct kvm_device_attr {
+	__u32	flags;		/* no flags currently defined */
+	__u32	group;		/* device-defined */
+	__u64	attr;		/* group-defined */
+	__u64	addr;		/* userspace address of attr data */
+};
+
+/*
  * ioctls for VM fds
  */
 #define KVM_SET_MEMORY_REGION     _IOW(KVMIO,  0x40, struct kvm_memory_region)
···
 #define KVM_PPC_GET_HTAB_FD	  _IOW(KVMIO,  0xaa, struct kvm_get_htab_fd)
 /* Available with KVM_CAP_ARM_SET_DEVICE_ADDR */
 #define KVM_ARM_SET_DEVICE_ADDR	  _IOW(KVMIO,  0xab, struct kvm_arm_device_addr)
+
+/* ioctl for vm fd */
+#define KVM_CREATE_DEVICE	  _IOWR(KVMIO,  0xe0, struct kvm_create_device)
+
+/* ioctls for fds returned by KVM_CREATE_DEVICE */
+#define KVM_SET_DEVICE_ATTR	  _IOW(KVMIO,  0xe1, struct kvm_device_attr)
+#define KVM_GET_DEVICE_ATTR	  _IOW(KVMIO,  0xe2, struct kvm_device_attr)
+#define KVM_HAS_DEVICE_ATTR	  _IOW(KVMIO,  0xe3, struct kvm_device_attr)
 
 /*
  * ioctls for vcpu fds

virt/kvm/kvm_main.c

···
 }
 #endif
 
+static int kvm_device_ioctl_attr(struct kvm_device *dev,
+				 int (*accessor)(struct kvm_device *dev,
+						 struct kvm_device_attr *attr),
+				 unsigned long arg)
+{
+	struct kvm_device_attr attr;
+
+	if (!accessor)
+		return -EPERM;
+
+	if (copy_from_user(&attr, (void __user *)arg, sizeof(attr)))
+		return -EFAULT;
+
+	return accessor(dev, &attr);
+}
+
+static long kvm_device_ioctl(struct file *filp, unsigned int ioctl,
+			     unsigned long arg)
+{
+	struct kvm_device *dev = filp->private_data;
+
+	switch (ioctl) {
+	case KVM_SET_DEVICE_ATTR:
+		return kvm_device_ioctl_attr(dev, dev->ops->set_attr, arg);
+	case KVM_GET_DEVICE_ATTR:
+		return kvm_device_ioctl_attr(dev, dev->ops->get_attr, arg);
+	case KVM_HAS_DEVICE_ATTR:
+		return kvm_device_ioctl_attr(dev, dev->ops->has_attr, arg);
+	default:
+		if (dev->ops->ioctl)
+			return dev->ops->ioctl(dev, ioctl, arg);
+
+		return -ENOTTY;
+	}
+}
+
+void kvm_device_get(struct kvm_device *dev)
+{
+	atomic_inc(&dev->users);
+}
+
+void kvm_device_put(struct kvm_device *dev)
+{
+	if (atomic_dec_and_test(&dev->users))
+		dev->ops->destroy(dev);
+}
+
+static int kvm_device_release(struct inode *inode, struct file *filp)
+{
+	struct kvm_device *dev = filp->private_data;
+	struct kvm *kvm = dev->kvm;
+
+	kvm_device_put(dev);
+	kvm_put_kvm(kvm);
+	return 0;
+}
+
+static const struct file_operations kvm_device_fops = {
+	.unlocked_ioctl = kvm_device_ioctl,
+	.release = kvm_device_release,
+};
+
+struct kvm_device *kvm_device_from_filp(struct file *filp)
+{
+	if (filp->f_op != &kvm_device_fops)
+		return NULL;
+
+	return filp->private_data;
+}
+
+static int kvm_ioctl_create_device(struct kvm *kvm,
+				   struct kvm_create_device *cd)
+{
+	struct kvm_device_ops *ops = NULL;
+	struct kvm_device *dev;
+	bool test = cd->flags & KVM_CREATE_DEVICE_TEST;
+	int ret;
+
+	switch (cd->type) {
+	default:
+		return -ENODEV;
+	}
+
+	if (test)
+		return 0;
+
+	dev = kzalloc(sizeof(*dev), GFP_KERNEL);
+	if (!dev)
+		return -ENOMEM;
+
+	dev->ops = ops;
+	dev->kvm = kvm;
+	atomic_set(&dev->users, 1);
+
+	ret = ops->create(dev, cd->type);
+	if (ret < 0) {
+		kfree(dev);
+		return ret;
+	}
+
+	ret = anon_inode_getfd(ops->name, &kvm_device_fops, dev, O_RDWR);
+	if (ret < 0) {
+		ops->destroy(dev);
+		return ret;
+	}
+
+	kvm_get_kvm(kvm);
+	cd->fd = ret;
+	return 0;
+}
+
 static long kvm_vm_ioctl(struct file *filp,
 			   unsigned int ioctl, unsigned long arg)
 {
···
 		break;
 	}
 #endif /* CONFIG_HAVE_KVM_IRQ_ROUTING */
+	case KVM_CREATE_DEVICE: {
+		struct kvm_create_device cd;
+
+		r = -EFAULT;
+		if (copy_from_user(&cd, argp, sizeof(cd)))
+			goto out;
+
+		r = kvm_ioctl_create_device(kvm, &cd);
+		if (r)
+			goto out;
+
+		r = -EFAULT;
+		if (copy_to_user(argp, &cd, sizeof(cd)))
+			goto out;
+
+		r = 0;
+		break;
+	}
 	default:
 		r = kvm_arch_vm_ioctl(filp, ioctl, arg);
 		if (r == -ENOTTY)