vhost/vsock: add IOTLB API support

This patch enables the IOTLB API support for vhost-vsock devices,
allowing the userspace to emulate an IOMMU for the guest.

These changes were made following vhost-net, in details this patch:
- exposes VIRTIO_F_ACCESS_PLATFORM feature and inits the iotlb
  device if the feature is acked
- implements VHOST_GET_BACKEND_FEATURES and
  VHOST_SET_BACKEND_FEATURES ioctls
- calls vq_meta_prefetch() before vq processing to prefetch vq
  metadata address in IOTLB
- provides .read_iter, .write_iter, and .poll callbacks for the
  chardev; they are used by the userspace to exchange IOTLB messages

This patch was tested specifying "intel_iommu=strict" in the guest
kernel command line. I used QEMU with a patch applied [1] to fix a
simple issue (that patch was merged in QEMU v5.2.0):
    $ qemu -M q35,accel=kvm,kernel-irqchip=split \
           -drive file=fedora.qcow2,format=qcow2,if=virtio \
           -device intel-iommu,intremap=on,device-iotlb=on \
           -device vhost-vsock-pci,guest-cid=3,iommu_platform=on,ats=on

[1] https://lists.gnu.org/archive/html/qemu-devel/2020-10/msg09077.html

Reviewed-by: Stefan Hajnoczi <[email protected]>
Signed-off-by: Stefano Garzarella <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Michael S. Tsirkin <[email protected]>
Acked-by: Jason Wang <[email protected]>

Author: stefano-garzarella

drivers/vhost/vsock.c

@@ -30,7 +30,12 @@
 #define VHOST_VSOCK_PKT_WEIGHT 256
 
 enum {
-	VHOST_VSOCK_FEATURES = VHOST_FEATURES,
+	VHOST_VSOCK_FEATURES = VHOST_FEATURES |
+			       (1ULL << VIRTIO_F_ACCESS_PLATFORM)
+};
+
+enum {
+	VHOST_VSOCK_BACKEND_FEATURES = (1ULL << VHOST_BACKEND_F_IOTLB_MSG_V2)
 };
 
 /* Used to track all the vhost_vsock instances on the system. */
@@ -94,6 +99,9 @@ vhost_transport_do_send_pkt(struct vhost_vsock *vsock,
 	if (!vhost_vq_get_backend(vq))
 		goto out;
 
+	if (!vq_meta_prefetch(vq))
+		goto out;
+
 	/* Avoid further vmexits, we're already processing the virtqueue */
 	vhost_disable_notify(&vsock->dev, vq);
 
@@ -449,6 +457,9 @@ static void vhost_vsock_handle_tx_kick(struct vhost_work *work)
 	if (!vhost_vq_get_backend(vq))
 		goto out;
 
+	if (!vq_meta_prefetch(vq))
+		goto out;
+
 	vhost_disable_notify(&vsock->dev, vq);
 	do {
 		u32 len;
@@ -766,8 +777,12 @@ static int vhost_vsock_set_features(struct vhost_vsock *vsock, u64 features)
 	mutex_lock(&vsock->dev.mutex);
 	if ((features & (1 << VHOST_F_LOG_ALL)) &&
 	    !vhost_log_access_ok(&vsock->dev)) {
-		mutex_unlock(&vsock->dev.mutex);
-		return -EFAULT;
+		goto err;
+	}
+
+	if ((features & (1ULL << VIRTIO_F_ACCESS_PLATFORM))) {
+		if (vhost_init_device_iotlb(&vsock->dev, true))
+			goto err;
 	}
 
 	for (i = 0; i < ARRAY_SIZE(vsock->vqs); i++) {
@@ -778,6 +793,10 @@ static int vhost_vsock_set_features(struct vhost_vsock *vsock, u64 features)
 	}
 	mutex_unlock(&vsock->dev.mutex);
 	return 0;
+
+err:
+	mutex_unlock(&vsock->dev.mutex);
+	return -EFAULT;
 }
 
 static long vhost_vsock_dev_ioctl(struct file *f, unsigned int ioctl,
@@ -811,6 +830,18 @@ static long vhost_vsock_dev_ioctl(struct file *f, unsigned int ioctl,
 		if (copy_from_user(&features, argp, sizeof(features)))
 			return -EFAULT;
 		return vhost_vsock_set_features(vsock, features);
+	case VHOST_GET_BACKEND_FEATURES:
+		features = VHOST_VSOCK_BACKEND_FEATURES;
+		if (copy_to_user(argp, &features, sizeof(features)))
+			return -EFAULT;
+		return 0;
+	case VHOST_SET_BACKEND_FEATURES:
+		if (copy_from_user(&features, argp, sizeof(features)))
+			return -EFAULT;
+		if (features & ~VHOST_VSOCK_BACKEND_FEATURES)
+			return -EOPNOTSUPP;
+		vhost_set_backend_features(&vsock->dev, features);
+		return 0;
 	default:
 		mutex_lock(&vsock->dev.mutex);
 		r = vhost_dev_ioctl(&vsock->dev, ioctl, argp);
@@ -823,13 +854,44 @@ static long vhost_vsock_dev_ioctl(struct file *f, unsigned int ioctl,
 	}
 }
 
+static ssize_t vhost_vsock_chr_read_iter(struct kiocb *iocb, struct iov_iter *to)
+{
+	struct file *file = iocb->ki_filp;
+	struct vhost_vsock *vsock = file->private_data;
+	struct vhost_dev *dev = &vsock->dev;
+	int noblock = file->f_flags & O_NONBLOCK;
+
+	return vhost_chr_read_iter(dev, to, noblock);
+}
+
+static ssize_t vhost_vsock_chr_write_iter(struct kiocb *iocb,
+					struct iov_iter *from)
+{
+	struct file *file = iocb->ki_filp;
+	struct vhost_vsock *vsock = file->private_data;
+	struct vhost_dev *dev = &vsock->dev;
+
+	return vhost_chr_write_iter(dev, from);
+}
+
+static __poll_t vhost_vsock_chr_poll(struct file *file, poll_table *wait)
+{
+	struct vhost_vsock *vsock = file->private_data;
+	struct vhost_dev *dev = &vsock->dev;
+
+	return vhost_chr_poll(file, dev, wait);
+}
+
 static const struct file_operations vhost_vsock_fops = {
 	.owner          = THIS_MODULE,
 	.open           = vhost_vsock_dev_open,
 	.release        = vhost_vsock_dev_release,
 	.llseek		= noop_llseek,
 	.unlocked_ioctl = vhost_vsock_dev_ioctl,
 	.compat_ioctl   = compat_ptr_ioctl,
+	.read_iter      = vhost_vsock_chr_read_iter,
+	.write_iter     = vhost_vsock_chr_write_iter,
+	.poll           = vhost_vsock_chr_poll,
 };
 
 static struct miscdevice vhost_vsock_misc = {