xref: /linux/drivers/vhost/vsock.c (revision 4e0ae876f77bc01a7e77724dea57b4b82bd53244)
1 /*
2  * vhost transport for vsock
3  *
4  * Copyright (C) 2013-2015 Red Hat, Inc.
5  * Author: Asias He <asias@redhat.com>
6  *         Stefan Hajnoczi <stefanha@redhat.com>
7  *
8  * This work is licensed under the terms of the GNU GPL, version 2.
9  */
10 #include <linux/miscdevice.h>
11 #include <linux/atomic.h>
12 #include <linux/module.h>
13 #include <linux/mutex.h>
14 #include <linux/vmalloc.h>
15 #include <net/sock.h>
16 #include <linux/virtio_vsock.h>
17 #include <linux/vhost.h>
18 #include <linux/hashtable.h>
19 
20 #include <net/af_vsock.h>
21 #include "vhost.h"
22 
23 #define VHOST_VSOCK_DEFAULT_HOST_CID	2
24 
25 enum {
26 	VHOST_VSOCK_FEATURES = VHOST_FEATURES,
27 };
28 
29 /* Used to track all the vhost_vsock instances on the system. */
30 static DEFINE_MUTEX(vhost_vsock_mutex);
31 static DEFINE_READ_MOSTLY_HASHTABLE(vhost_vsock_hash, 8);
32 
33 struct vhost_vsock {
34 	struct vhost_dev dev;
35 	struct vhost_virtqueue vqs[2];
36 
37 	/* Link to global vhost_vsock_hash, writes use vhost_vsock_mutex */
38 	struct hlist_node hash;
39 
40 	struct vhost_work send_pkt_work;
41 	spinlock_t send_pkt_list_lock;
42 	struct list_head send_pkt_list;	/* host->guest pending packets */
43 
44 	atomic_t queued_replies;
45 
46 	u32 guest_cid;
47 };
48 
49 static u32 vhost_transport_get_local_cid(void)
50 {
51 	return VHOST_VSOCK_DEFAULT_HOST_CID;
52 }
53 
54 /* Callers that dereference the return value must hold vhost_vsock_mutex or the
55  * RCU read lock.
56  */
57 static struct vhost_vsock *vhost_vsock_get(u32 guest_cid)
58 {
59 	struct vhost_vsock *vsock;
60 
61 	hash_for_each_possible_rcu(vhost_vsock_hash, vsock, hash, guest_cid) {
62 		u32 other_cid = vsock->guest_cid;
63 
64 		/* Skip instances that have no CID yet */
65 		if (other_cid == 0)
66 			continue;
67 
68 		if (other_cid == guest_cid)
69 			return vsock;
70 
71 	}
72 
73 	return NULL;
74 }
75 
76 static void
77 vhost_transport_do_send_pkt(struct vhost_vsock *vsock,
78 			    struct vhost_virtqueue *vq)
79 {
80 	struct vhost_virtqueue *tx_vq = &vsock->vqs[VSOCK_VQ_TX];
81 	bool added = false;
82 	bool restart_tx = false;
83 
84 	mutex_lock(&vq->mutex);
85 
86 	if (!vq->private_data)
87 		goto out;
88 
89 	/* Avoid further vmexits, we're already processing the virtqueue */
90 	vhost_disable_notify(&vsock->dev, vq);
91 
92 	for (;;) {
93 		struct virtio_vsock_pkt *pkt;
94 		struct iov_iter iov_iter;
95 		unsigned out, in;
96 		size_t nbytes;
97 		size_t len;
98 		int head;
99 
100 		spin_lock_bh(&vsock->send_pkt_list_lock);
101 		if (list_empty(&vsock->send_pkt_list)) {
102 			spin_unlock_bh(&vsock->send_pkt_list_lock);
103 			vhost_enable_notify(&vsock->dev, vq);
104 			break;
105 		}
106 
107 		pkt = list_first_entry(&vsock->send_pkt_list,
108 				       struct virtio_vsock_pkt, list);
109 		list_del_init(&pkt->list);
110 		spin_unlock_bh(&vsock->send_pkt_list_lock);
111 
112 		head = vhost_get_vq_desc(vq, vq->iov, ARRAY_SIZE(vq->iov),
113 					 &out, &in, NULL, NULL);
114 		if (head < 0) {
115 			spin_lock_bh(&vsock->send_pkt_list_lock);
116 			list_add(&pkt->list, &vsock->send_pkt_list);
117 			spin_unlock_bh(&vsock->send_pkt_list_lock);
118 			break;
119 		}
120 
121 		if (head == vq->num) {
122 			spin_lock_bh(&vsock->send_pkt_list_lock);
123 			list_add(&pkt->list, &vsock->send_pkt_list);
124 			spin_unlock_bh(&vsock->send_pkt_list_lock);
125 
126 			/* We cannot finish yet if more buffers snuck in while
127 			 * re-enabling notify.
128 			 */
129 			if (unlikely(vhost_enable_notify(&vsock->dev, vq))) {
130 				vhost_disable_notify(&vsock->dev, vq);
131 				continue;
132 			}
133 			break;
134 		}
135 
136 		if (out) {
137 			virtio_transport_free_pkt(pkt);
138 			vq_err(vq, "Expected 0 output buffers, got %u\n", out);
139 			break;
140 		}
141 
142 		len = iov_length(&vq->iov[out], in);
143 		iov_iter_init(&iov_iter, READ, &vq->iov[out], in, len);
144 
145 		nbytes = copy_to_iter(&pkt->hdr, sizeof(pkt->hdr), &iov_iter);
146 		if (nbytes != sizeof(pkt->hdr)) {
147 			virtio_transport_free_pkt(pkt);
148 			vq_err(vq, "Faulted on copying pkt hdr\n");
149 			break;
150 		}
151 
152 		nbytes = copy_to_iter(pkt->buf, pkt->len, &iov_iter);
153 		if (nbytes != pkt->len) {
154 			virtio_transport_free_pkt(pkt);
155 			vq_err(vq, "Faulted on copying pkt buf\n");
156 			break;
157 		}
158 
159 		vhost_add_used(vq, head, sizeof(pkt->hdr) + pkt->len);
160 		added = true;
161 
162 		if (pkt->reply) {
163 			int val;
164 
165 			val = atomic_dec_return(&vsock->queued_replies);
166 
167 			/* Do we have resources to resume tx processing? */
168 			if (val + 1 == tx_vq->num)
169 				restart_tx = true;
170 		}
171 
172 		/* Deliver to monitoring devices all correctly transmitted
173 		 * packets.
174 		 */
175 		virtio_transport_deliver_tap_pkt(pkt);
176 
177 		virtio_transport_free_pkt(pkt);
178 	}
179 	if (added)
180 		vhost_signal(&vsock->dev, vq);
181 
182 out:
183 	mutex_unlock(&vq->mutex);
184 
185 	if (restart_tx)
186 		vhost_poll_queue(&tx_vq->poll);
187 }
188 
189 static void vhost_transport_send_pkt_work(struct vhost_work *work)
190 {
191 	struct vhost_virtqueue *vq;
192 	struct vhost_vsock *vsock;
193 
194 	vsock = container_of(work, struct vhost_vsock, send_pkt_work);
195 	vq = &vsock->vqs[VSOCK_VQ_RX];
196 
197 	vhost_transport_do_send_pkt(vsock, vq);
198 }
199 
200 static int
201 vhost_transport_send_pkt(struct virtio_vsock_pkt *pkt)
202 {
203 	struct vhost_vsock *vsock;
204 	int len = pkt->len;
205 
206 	rcu_read_lock();
207 
208 	/* Find the vhost_vsock according to guest context id  */
209 	vsock = vhost_vsock_get(le64_to_cpu(pkt->hdr.dst_cid));
210 	if (!vsock) {
211 		rcu_read_unlock();
212 		virtio_transport_free_pkt(pkt);
213 		return -ENODEV;
214 	}
215 
216 	if (pkt->reply)
217 		atomic_inc(&vsock->queued_replies);
218 
219 	spin_lock_bh(&vsock->send_pkt_list_lock);
220 	list_add_tail(&pkt->list, &vsock->send_pkt_list);
221 	spin_unlock_bh(&vsock->send_pkt_list_lock);
222 
223 	vhost_work_queue(&vsock->dev, &vsock->send_pkt_work);
224 
225 	rcu_read_unlock();
226 	return len;
227 }
228 
229 static int
230 vhost_transport_cancel_pkt(struct vsock_sock *vsk)
231 {
232 	struct vhost_vsock *vsock;
233 	struct virtio_vsock_pkt *pkt, *n;
234 	int cnt = 0;
235 	int ret = -ENODEV;
236 	LIST_HEAD(freeme);
237 
238 	rcu_read_lock();
239 
240 	/* Find the vhost_vsock according to guest context id  */
241 	vsock = vhost_vsock_get(vsk->remote_addr.svm_cid);
242 	if (!vsock)
243 		goto out;
244 
245 	spin_lock_bh(&vsock->send_pkt_list_lock);
246 	list_for_each_entry_safe(pkt, n, &vsock->send_pkt_list, list) {
247 		if (pkt->vsk != vsk)
248 			continue;
249 		list_move(&pkt->list, &freeme);
250 	}
251 	spin_unlock_bh(&vsock->send_pkt_list_lock);
252 
253 	list_for_each_entry_safe(pkt, n, &freeme, list) {
254 		if (pkt->reply)
255 			cnt++;
256 		list_del(&pkt->list);
257 		virtio_transport_free_pkt(pkt);
258 	}
259 
260 	if (cnt) {
261 		struct vhost_virtqueue *tx_vq = &vsock->vqs[VSOCK_VQ_TX];
262 		int new_cnt;
263 
264 		new_cnt = atomic_sub_return(cnt, &vsock->queued_replies);
265 		if (new_cnt + cnt >= tx_vq->num && new_cnt < tx_vq->num)
266 			vhost_poll_queue(&tx_vq->poll);
267 	}
268 
269 	ret = 0;
270 out:
271 	rcu_read_unlock();
272 	return ret;
273 }
274 
275 static struct virtio_vsock_pkt *
276 vhost_vsock_alloc_pkt(struct vhost_virtqueue *vq,
277 		      unsigned int out, unsigned int in)
278 {
279 	struct virtio_vsock_pkt *pkt;
280 	struct iov_iter iov_iter;
281 	size_t nbytes;
282 	size_t len;
283 
284 	if (in != 0) {
285 		vq_err(vq, "Expected 0 input buffers, got %u\n", in);
286 		return NULL;
287 	}
288 
289 	pkt = kzalloc(sizeof(*pkt), GFP_KERNEL);
290 	if (!pkt)
291 		return NULL;
292 
293 	len = iov_length(vq->iov, out);
294 	iov_iter_init(&iov_iter, WRITE, vq->iov, out, len);
295 
296 	nbytes = copy_from_iter(&pkt->hdr, sizeof(pkt->hdr), &iov_iter);
297 	if (nbytes != sizeof(pkt->hdr)) {
298 		vq_err(vq, "Expected %zu bytes for pkt->hdr, got %zu bytes\n",
299 		       sizeof(pkt->hdr), nbytes);
300 		kfree(pkt);
301 		return NULL;
302 	}
303 
304 	if (le16_to_cpu(pkt->hdr.type) == VIRTIO_VSOCK_TYPE_STREAM)
305 		pkt->len = le32_to_cpu(pkt->hdr.len);
306 
307 	/* No payload */
308 	if (!pkt->len)
309 		return pkt;
310 
311 	/* The pkt is too big */
312 	if (pkt->len > VIRTIO_VSOCK_MAX_PKT_BUF_SIZE) {
313 		kfree(pkt);
314 		return NULL;
315 	}
316 
317 	pkt->buf = kmalloc(pkt->len, GFP_KERNEL);
318 	if (!pkt->buf) {
319 		kfree(pkt);
320 		return NULL;
321 	}
322 
323 	nbytes = copy_from_iter(pkt->buf, pkt->len, &iov_iter);
324 	if (nbytes != pkt->len) {
325 		vq_err(vq, "Expected %u byte payload, got %zu bytes\n",
326 		       pkt->len, nbytes);
327 		virtio_transport_free_pkt(pkt);
328 		return NULL;
329 	}
330 
331 	return pkt;
332 }
333 
334 /* Is there space left for replies to rx packets? */
335 static bool vhost_vsock_more_replies(struct vhost_vsock *vsock)
336 {
337 	struct vhost_virtqueue *vq = &vsock->vqs[VSOCK_VQ_TX];
338 	int val;
339 
340 	smp_rmb(); /* paired with atomic_inc() and atomic_dec_return() */
341 	val = atomic_read(&vsock->queued_replies);
342 
343 	return val < vq->num;
344 }
345 
346 static void vhost_vsock_handle_tx_kick(struct vhost_work *work)
347 {
348 	struct vhost_virtqueue *vq = container_of(work, struct vhost_virtqueue,
349 						  poll.work);
350 	struct vhost_vsock *vsock = container_of(vq->dev, struct vhost_vsock,
351 						 dev);
352 	struct virtio_vsock_pkt *pkt;
353 	int head;
354 	unsigned int out, in;
355 	bool added = false;
356 
357 	mutex_lock(&vq->mutex);
358 
359 	if (!vq->private_data)
360 		goto out;
361 
362 	vhost_disable_notify(&vsock->dev, vq);
363 	for (;;) {
364 		u32 len;
365 
366 		if (!vhost_vsock_more_replies(vsock)) {
367 			/* Stop tx until the device processes already
368 			 * pending replies.  Leave tx virtqueue
369 			 * callbacks disabled.
370 			 */
371 			goto no_more_replies;
372 		}
373 
374 		head = vhost_get_vq_desc(vq, vq->iov, ARRAY_SIZE(vq->iov),
375 					 &out, &in, NULL, NULL);
376 		if (head < 0)
377 			break;
378 
379 		if (head == vq->num) {
380 			if (unlikely(vhost_enable_notify(&vsock->dev, vq))) {
381 				vhost_disable_notify(&vsock->dev, vq);
382 				continue;
383 			}
384 			break;
385 		}
386 
387 		pkt = vhost_vsock_alloc_pkt(vq, out, in);
388 		if (!pkt) {
389 			vq_err(vq, "Faulted on pkt\n");
390 			continue;
391 		}
392 
393 		len = pkt->len;
394 
395 		/* Deliver to monitoring devices all received packets */
396 		virtio_transport_deliver_tap_pkt(pkt);
397 
398 		/* Only accept correctly addressed packets */
399 		if (le64_to_cpu(pkt->hdr.src_cid) == vsock->guest_cid)
400 			virtio_transport_recv_pkt(pkt);
401 		else
402 			virtio_transport_free_pkt(pkt);
403 
404 		vhost_add_used(vq, head, sizeof(pkt->hdr) + len);
405 		added = true;
406 	}
407 
408 no_more_replies:
409 	if (added)
410 		vhost_signal(&vsock->dev, vq);
411 
412 out:
413 	mutex_unlock(&vq->mutex);
414 }
415 
416 static void vhost_vsock_handle_rx_kick(struct vhost_work *work)
417 {
418 	struct vhost_virtqueue *vq = container_of(work, struct vhost_virtqueue,
419 						poll.work);
420 	struct vhost_vsock *vsock = container_of(vq->dev, struct vhost_vsock,
421 						 dev);
422 
423 	vhost_transport_do_send_pkt(vsock, vq);
424 }
425 
426 static int vhost_vsock_start(struct vhost_vsock *vsock)
427 {
428 	struct vhost_virtqueue *vq;
429 	size_t i;
430 	int ret;
431 
432 	mutex_lock(&vsock->dev.mutex);
433 
434 	ret = vhost_dev_check_owner(&vsock->dev);
435 	if (ret)
436 		goto err;
437 
438 	for (i = 0; i < ARRAY_SIZE(vsock->vqs); i++) {
439 		vq = &vsock->vqs[i];
440 
441 		mutex_lock(&vq->mutex);
442 
443 		if (!vhost_vq_access_ok(vq)) {
444 			ret = -EFAULT;
445 			goto err_vq;
446 		}
447 
448 		if (!vq->private_data) {
449 			vq->private_data = vsock;
450 			ret = vhost_vq_init_access(vq);
451 			if (ret)
452 				goto err_vq;
453 		}
454 
455 		mutex_unlock(&vq->mutex);
456 	}
457 
458 	mutex_unlock(&vsock->dev.mutex);
459 	return 0;
460 
461 err_vq:
462 	vq->private_data = NULL;
463 	mutex_unlock(&vq->mutex);
464 
465 	for (i = 0; i < ARRAY_SIZE(vsock->vqs); i++) {
466 		vq = &vsock->vqs[i];
467 
468 		mutex_lock(&vq->mutex);
469 		vq->private_data = NULL;
470 		mutex_unlock(&vq->mutex);
471 	}
472 err:
473 	mutex_unlock(&vsock->dev.mutex);
474 	return ret;
475 }
476 
477 static int vhost_vsock_stop(struct vhost_vsock *vsock)
478 {
479 	size_t i;
480 	int ret;
481 
482 	mutex_lock(&vsock->dev.mutex);
483 
484 	ret = vhost_dev_check_owner(&vsock->dev);
485 	if (ret)
486 		goto err;
487 
488 	for (i = 0; i < ARRAY_SIZE(vsock->vqs); i++) {
489 		struct vhost_virtqueue *vq = &vsock->vqs[i];
490 
491 		mutex_lock(&vq->mutex);
492 		vq->private_data = NULL;
493 		mutex_unlock(&vq->mutex);
494 	}
495 
496 err:
497 	mutex_unlock(&vsock->dev.mutex);
498 	return ret;
499 }
500 
501 static void vhost_vsock_free(struct vhost_vsock *vsock)
502 {
503 	kvfree(vsock);
504 }
505 
506 static int vhost_vsock_dev_open(struct inode *inode, struct file *file)
507 {
508 	struct vhost_virtqueue **vqs;
509 	struct vhost_vsock *vsock;
510 	int ret;
511 
512 	/* This struct is large and allocation could fail, fall back to vmalloc
513 	 * if there is no other way.
514 	 */
515 	vsock = kvmalloc(sizeof(*vsock), GFP_KERNEL | __GFP_RETRY_MAYFAIL);
516 	if (!vsock)
517 		return -ENOMEM;
518 
519 	vqs = kmalloc_array(ARRAY_SIZE(vsock->vqs), sizeof(*vqs), GFP_KERNEL);
520 	if (!vqs) {
521 		ret = -ENOMEM;
522 		goto out;
523 	}
524 
525 	vsock->guest_cid = 0; /* no CID assigned yet */
526 
527 	atomic_set(&vsock->queued_replies, 0);
528 
529 	vqs[VSOCK_VQ_TX] = &vsock->vqs[VSOCK_VQ_TX];
530 	vqs[VSOCK_VQ_RX] = &vsock->vqs[VSOCK_VQ_RX];
531 	vsock->vqs[VSOCK_VQ_TX].handle_kick = vhost_vsock_handle_tx_kick;
532 	vsock->vqs[VSOCK_VQ_RX].handle_kick = vhost_vsock_handle_rx_kick;
533 
534 	vhost_dev_init(&vsock->dev, vqs, ARRAY_SIZE(vsock->vqs), UIO_MAXIOV);
535 
536 	file->private_data = vsock;
537 	spin_lock_init(&vsock->send_pkt_list_lock);
538 	INIT_LIST_HEAD(&vsock->send_pkt_list);
539 	vhost_work_init(&vsock->send_pkt_work, vhost_transport_send_pkt_work);
540 	return 0;
541 
542 out:
543 	vhost_vsock_free(vsock);
544 	return ret;
545 }
546 
547 static void vhost_vsock_flush(struct vhost_vsock *vsock)
548 {
549 	int i;
550 
551 	for (i = 0; i < ARRAY_SIZE(vsock->vqs); i++)
552 		if (vsock->vqs[i].handle_kick)
553 			vhost_poll_flush(&vsock->vqs[i].poll);
554 	vhost_work_flush(&vsock->dev, &vsock->send_pkt_work);
555 }
556 
557 static void vhost_vsock_reset_orphans(struct sock *sk)
558 {
559 	struct vsock_sock *vsk = vsock_sk(sk);
560 
561 	/* vmci_transport.c doesn't take sk_lock here either.  At least we're
562 	 * under vsock_table_lock so the sock cannot disappear while we're
563 	 * executing.
564 	 */
565 
566 	/* If the peer is still valid, no need to reset connection */
567 	if (vhost_vsock_get(vsk->remote_addr.svm_cid))
568 		return;
569 
570 	/* If the close timeout is pending, let it expire.  This avoids races
571 	 * with the timeout callback.
572 	 */
573 	if (vsk->close_work_scheduled)
574 		return;
575 
576 	sock_set_flag(sk, SOCK_DONE);
577 	vsk->peer_shutdown = SHUTDOWN_MASK;
578 	sk->sk_state = SS_UNCONNECTED;
579 	sk->sk_err = ECONNRESET;
580 	sk->sk_error_report(sk);
581 }
582 
583 static int vhost_vsock_dev_release(struct inode *inode, struct file *file)
584 {
585 	struct vhost_vsock *vsock = file->private_data;
586 
587 	mutex_lock(&vhost_vsock_mutex);
588 	if (vsock->guest_cid)
589 		hash_del_rcu(&vsock->hash);
590 	mutex_unlock(&vhost_vsock_mutex);
591 
592 	/* Wait for other CPUs to finish using vsock */
593 	synchronize_rcu();
594 
595 	/* Iterating over all connections for all CIDs to find orphans is
596 	 * inefficient.  Room for improvement here. */
597 	vsock_for_each_connected_socket(vhost_vsock_reset_orphans);
598 
599 	vhost_vsock_stop(vsock);
600 	vhost_vsock_flush(vsock);
601 	vhost_dev_stop(&vsock->dev);
602 
603 	spin_lock_bh(&vsock->send_pkt_list_lock);
604 	while (!list_empty(&vsock->send_pkt_list)) {
605 		struct virtio_vsock_pkt *pkt;
606 
607 		pkt = list_first_entry(&vsock->send_pkt_list,
608 				struct virtio_vsock_pkt, list);
609 		list_del_init(&pkt->list);
610 		virtio_transport_free_pkt(pkt);
611 	}
612 	spin_unlock_bh(&vsock->send_pkt_list_lock);
613 
614 	vhost_dev_cleanup(&vsock->dev);
615 	kfree(vsock->dev.vqs);
616 	vhost_vsock_free(vsock);
617 	return 0;
618 }
619 
620 static int vhost_vsock_set_cid(struct vhost_vsock *vsock, u64 guest_cid)
621 {
622 	struct vhost_vsock *other;
623 
624 	/* Refuse reserved CIDs */
625 	if (guest_cid <= VMADDR_CID_HOST ||
626 	    guest_cid == U32_MAX)
627 		return -EINVAL;
628 
629 	/* 64-bit CIDs are not yet supported */
630 	if (guest_cid > U32_MAX)
631 		return -EINVAL;
632 
633 	/* Refuse if CID is already in use */
634 	mutex_lock(&vhost_vsock_mutex);
635 	other = vhost_vsock_get(guest_cid);
636 	if (other && other != vsock) {
637 		mutex_unlock(&vhost_vsock_mutex);
638 		return -EADDRINUSE;
639 	}
640 
641 	if (vsock->guest_cid)
642 		hash_del_rcu(&vsock->hash);
643 
644 	vsock->guest_cid = guest_cid;
645 	hash_add_rcu(vhost_vsock_hash, &vsock->hash, vsock->guest_cid);
646 	mutex_unlock(&vhost_vsock_mutex);
647 
648 	return 0;
649 }
650 
651 static int vhost_vsock_set_features(struct vhost_vsock *vsock, u64 features)
652 {
653 	struct vhost_virtqueue *vq;
654 	int i;
655 
656 	if (features & ~VHOST_VSOCK_FEATURES)
657 		return -EOPNOTSUPP;
658 
659 	mutex_lock(&vsock->dev.mutex);
660 	if ((features & (1 << VHOST_F_LOG_ALL)) &&
661 	    !vhost_log_access_ok(&vsock->dev)) {
662 		mutex_unlock(&vsock->dev.mutex);
663 		return -EFAULT;
664 	}
665 
666 	for (i = 0; i < ARRAY_SIZE(vsock->vqs); i++) {
667 		vq = &vsock->vqs[i];
668 		mutex_lock(&vq->mutex);
669 		vq->acked_features = features;
670 		mutex_unlock(&vq->mutex);
671 	}
672 	mutex_unlock(&vsock->dev.mutex);
673 	return 0;
674 }
675 
676 static long vhost_vsock_dev_ioctl(struct file *f, unsigned int ioctl,
677 				  unsigned long arg)
678 {
679 	struct vhost_vsock *vsock = f->private_data;
680 	void __user *argp = (void __user *)arg;
681 	u64 guest_cid;
682 	u64 features;
683 	int start;
684 	int r;
685 
686 	switch (ioctl) {
687 	case VHOST_VSOCK_SET_GUEST_CID:
688 		if (copy_from_user(&guest_cid, argp, sizeof(guest_cid)))
689 			return -EFAULT;
690 		return vhost_vsock_set_cid(vsock, guest_cid);
691 	case VHOST_VSOCK_SET_RUNNING:
692 		if (copy_from_user(&start, argp, sizeof(start)))
693 			return -EFAULT;
694 		if (start)
695 			return vhost_vsock_start(vsock);
696 		else
697 			return vhost_vsock_stop(vsock);
698 	case VHOST_GET_FEATURES:
699 		features = VHOST_VSOCK_FEATURES;
700 		if (copy_to_user(argp, &features, sizeof(features)))
701 			return -EFAULT;
702 		return 0;
703 	case VHOST_SET_FEATURES:
704 		if (copy_from_user(&features, argp, sizeof(features)))
705 			return -EFAULT;
706 		return vhost_vsock_set_features(vsock, features);
707 	default:
708 		mutex_lock(&vsock->dev.mutex);
709 		r = vhost_dev_ioctl(&vsock->dev, ioctl, argp);
710 		if (r == -ENOIOCTLCMD)
711 			r = vhost_vring_ioctl(&vsock->dev, ioctl, argp);
712 		else
713 			vhost_vsock_flush(vsock);
714 		mutex_unlock(&vsock->dev.mutex);
715 		return r;
716 	}
717 }
718 
719 #ifdef CONFIG_COMPAT
720 static long vhost_vsock_dev_compat_ioctl(struct file *f, unsigned int ioctl,
721 					 unsigned long arg)
722 {
723 	return vhost_vsock_dev_ioctl(f, ioctl, (unsigned long)compat_ptr(arg));
724 }
725 #endif
726 
727 static const struct file_operations vhost_vsock_fops = {
728 	.owner          = THIS_MODULE,
729 	.open           = vhost_vsock_dev_open,
730 	.release        = vhost_vsock_dev_release,
731 	.llseek		= noop_llseek,
732 	.unlocked_ioctl = vhost_vsock_dev_ioctl,
733 #ifdef CONFIG_COMPAT
734 	.compat_ioctl   = vhost_vsock_dev_compat_ioctl,
735 #endif
736 };
737 
738 static struct miscdevice vhost_vsock_misc = {
739 	.minor = VHOST_VSOCK_MINOR,
740 	.name = "vhost-vsock",
741 	.fops = &vhost_vsock_fops,
742 };
743 
744 static struct virtio_transport vhost_transport = {
745 	.transport = {
746 		.get_local_cid            = vhost_transport_get_local_cid,
747 
748 		.init                     = virtio_transport_do_socket_init,
749 		.destruct                 = virtio_transport_destruct,
750 		.release                  = virtio_transport_release,
751 		.connect                  = virtio_transport_connect,
752 		.shutdown                 = virtio_transport_shutdown,
753 		.cancel_pkt               = vhost_transport_cancel_pkt,
754 
755 		.dgram_enqueue            = virtio_transport_dgram_enqueue,
756 		.dgram_dequeue            = virtio_transport_dgram_dequeue,
757 		.dgram_bind               = virtio_transport_dgram_bind,
758 		.dgram_allow              = virtio_transport_dgram_allow,
759 
760 		.stream_enqueue           = virtio_transport_stream_enqueue,
761 		.stream_dequeue           = virtio_transport_stream_dequeue,
762 		.stream_has_data          = virtio_transport_stream_has_data,
763 		.stream_has_space         = virtio_transport_stream_has_space,
764 		.stream_rcvhiwat          = virtio_transport_stream_rcvhiwat,
765 		.stream_is_active         = virtio_transport_stream_is_active,
766 		.stream_allow             = virtio_transport_stream_allow,
767 
768 		.notify_poll_in           = virtio_transport_notify_poll_in,
769 		.notify_poll_out          = virtio_transport_notify_poll_out,
770 		.notify_recv_init         = virtio_transport_notify_recv_init,
771 		.notify_recv_pre_block    = virtio_transport_notify_recv_pre_block,
772 		.notify_recv_pre_dequeue  = virtio_transport_notify_recv_pre_dequeue,
773 		.notify_recv_post_dequeue = virtio_transport_notify_recv_post_dequeue,
774 		.notify_send_init         = virtio_transport_notify_send_init,
775 		.notify_send_pre_block    = virtio_transport_notify_send_pre_block,
776 		.notify_send_pre_enqueue  = virtio_transport_notify_send_pre_enqueue,
777 		.notify_send_post_enqueue = virtio_transport_notify_send_post_enqueue,
778 
779 		.set_buffer_size          = virtio_transport_set_buffer_size,
780 		.set_min_buffer_size      = virtio_transport_set_min_buffer_size,
781 		.set_max_buffer_size      = virtio_transport_set_max_buffer_size,
782 		.get_buffer_size          = virtio_transport_get_buffer_size,
783 		.get_min_buffer_size      = virtio_transport_get_min_buffer_size,
784 		.get_max_buffer_size      = virtio_transport_get_max_buffer_size,
785 	},
786 
787 	.send_pkt = vhost_transport_send_pkt,
788 };
789 
790 static int __init vhost_vsock_init(void)
791 {
792 	int ret;
793 
794 	ret = vsock_core_init(&vhost_transport.transport);
795 	if (ret < 0)
796 		return ret;
797 	return misc_register(&vhost_vsock_misc);
798 };
799 
800 static void __exit vhost_vsock_exit(void)
801 {
802 	misc_deregister(&vhost_vsock_misc);
803 	vsock_core_exit();
804 };
805 
806 module_init(vhost_vsock_init);
807 module_exit(vhost_vsock_exit);
808 MODULE_LICENSE("GPL v2");
809 MODULE_AUTHOR("Asias He");
810 MODULE_DESCRIPTION("vhost transport for vsock ");
811 MODULE_ALIAS_MISCDEV(VHOST_VSOCK_MINOR);
812 MODULE_ALIAS("devname:vhost-vsock");
813