1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * common code for virtio vsock
4  *
5  * Copyright (C) 2013-2015 Red Hat, Inc.
6  * Author: Asias He <asias@redhat.com>
7  *         Stefan Hajnoczi <stefanha@redhat.com>
8  */
9 #include <linux/spinlock.h>
10 #include <linux/module.h>
11 #include <linux/sched/signal.h>
12 #include <linux/ctype.h>
13 #include <linux/list.h>
14 #include <linux/virtio_vsock.h>
15 #include <uapi/linux/vsockmon.h>
16 
17 #include <net/sock.h>
18 #include <net/af_vsock.h>
19 
20 #define CREATE_TRACE_POINTS
21 #include <trace/events/vsock_virtio_transport_common.h>
22 
23 /* How long to wait for graceful shutdown of a connection */
24 #define VSOCK_CLOSE_TIMEOUT (8 * HZ)
25 
26 /* Threshold for detecting small packets to copy */
27 #define GOOD_COPY_LEN  128
28 
29 static const struct virtio_transport *
30 virtio_transport_get_ops(struct vsock_sock *vsk)
31 {
32 	const struct vsock_transport *t = vsock_core_get_transport(vsk);
33 
34 	if (WARN_ON(!t))
35 		return NULL;
36 
37 	return container_of(t, struct virtio_transport, transport);
38 }
39 
40 /* Returns a new packet on success, otherwise returns NULL.
41  *
42  * If NULL is returned, errp is set to a negative errno.
43  */
44 static struct sk_buff *
45 virtio_transport_alloc_skb(struct virtio_vsock_pkt_info *info,
46 			   size_t len,
47 			   u32 src_cid,
48 			   u32 src_port,
49 			   u32 dst_cid,
50 			   u32 dst_port)
51 {
52 	const size_t skb_len = VIRTIO_VSOCK_SKB_HEADROOM + len;
53 	struct virtio_vsock_hdr *hdr;
54 	struct sk_buff *skb;
55 	void *payload;
56 	int err;
57 
58 	skb = virtio_vsock_alloc_skb(skb_len, GFP_KERNEL);
59 	if (!skb)
60 		return NULL;
61 
62 	hdr = virtio_vsock_hdr(skb);
63 	hdr->type	= cpu_to_le16(info->type);
64 	hdr->op		= cpu_to_le16(info->op);
65 	hdr->src_cid	= cpu_to_le64(src_cid);
66 	hdr->dst_cid	= cpu_to_le64(dst_cid);
67 	hdr->src_port	= cpu_to_le32(src_port);
68 	hdr->dst_port	= cpu_to_le32(dst_port);
69 	hdr->flags	= cpu_to_le32(info->flags);
70 	hdr->len	= cpu_to_le32(len);
71 
72 	if (info->msg && len > 0) {
73 		payload = skb_put(skb, len);
74 		err = memcpy_from_msg(payload, info->msg, len);
75 		if (err)
76 			goto out;
77 
78 		if (msg_data_left(info->msg) == 0 &&
79 		    info->type == VIRTIO_VSOCK_TYPE_SEQPACKET) {
80 			hdr->flags |= cpu_to_le32(VIRTIO_VSOCK_SEQ_EOM);
81 
82 			if (info->msg->msg_flags & MSG_EOR)
83 				hdr->flags |= cpu_to_le32(VIRTIO_VSOCK_SEQ_EOR);
84 		}
85 	}
86 
87 	if (info->reply)
88 		virtio_vsock_skb_set_reply(skb);
89 
90 	trace_virtio_transport_alloc_pkt(src_cid, src_port,
91 					 dst_cid, dst_port,
92 					 len,
93 					 info->type,
94 					 info->op,
95 					 info->flags);
96 
97 	if (info->vsk && !skb_set_owner_sk_safe(skb, sk_vsock(info->vsk))) {
98 		WARN_ONCE(1, "failed to allocate skb on vsock socket with sk_refcnt == 0\n");
99 		goto out;
100 	}
101 
102 	return skb;
103 
104 out:
105 	kfree_skb(skb);
106 	return NULL;
107 }
108 
109 /* Packet capture */
110 static struct sk_buff *virtio_transport_build_skb(void *opaque)
111 {
112 	struct virtio_vsock_hdr *pkt_hdr;
113 	struct sk_buff *pkt = opaque;
114 	struct af_vsockmon_hdr *hdr;
115 	struct sk_buff *skb;
116 	size_t payload_len;
117 	void *payload_buf;
118 
119 	/* A packet could be split to fit the RX buffer, so we can retrieve
120 	 * the payload length from the header and the buffer pointer taking
121 	 * care of the offset in the original packet.
122 	 */
123 	pkt_hdr = virtio_vsock_hdr(pkt);
124 	payload_len = pkt->len;
125 	payload_buf = pkt->data;
126 
127 	skb = alloc_skb(sizeof(*hdr) + sizeof(*pkt_hdr) + payload_len,
128 			GFP_ATOMIC);
129 	if (!skb)
130 		return NULL;
131 
132 	hdr = skb_put(skb, sizeof(*hdr));
133 
134 	/* pkt->hdr is little-endian so no need to byteswap here */
135 	hdr->src_cid = pkt_hdr->src_cid;
136 	hdr->src_port = pkt_hdr->src_port;
137 	hdr->dst_cid = pkt_hdr->dst_cid;
138 	hdr->dst_port = pkt_hdr->dst_port;
139 
140 	hdr->transport = cpu_to_le16(AF_VSOCK_TRANSPORT_VIRTIO);
141 	hdr->len = cpu_to_le16(sizeof(*pkt_hdr));
142 	memset(hdr->reserved, 0, sizeof(hdr->reserved));
143 
144 	switch (le16_to_cpu(pkt_hdr->op)) {
145 	case VIRTIO_VSOCK_OP_REQUEST:
146 	case VIRTIO_VSOCK_OP_RESPONSE:
147 		hdr->op = cpu_to_le16(AF_VSOCK_OP_CONNECT);
148 		break;
149 	case VIRTIO_VSOCK_OP_RST:
150 	case VIRTIO_VSOCK_OP_SHUTDOWN:
151 		hdr->op = cpu_to_le16(AF_VSOCK_OP_DISCONNECT);
152 		break;
153 	case VIRTIO_VSOCK_OP_RW:
154 		hdr->op = cpu_to_le16(AF_VSOCK_OP_PAYLOAD);
155 		break;
156 	case VIRTIO_VSOCK_OP_CREDIT_UPDATE:
157 	case VIRTIO_VSOCK_OP_CREDIT_REQUEST:
158 		hdr->op = cpu_to_le16(AF_VSOCK_OP_CONTROL);
159 		break;
160 	default:
161 		hdr->op = cpu_to_le16(AF_VSOCK_OP_UNKNOWN);
162 		break;
163 	}
164 
165 	skb_put_data(skb, pkt_hdr, sizeof(*pkt_hdr));
166 
167 	if (payload_len) {
168 		skb_put_data(skb, payload_buf, payload_len);
169 	}
170 
171 	return skb;
172 }
173 
174 void virtio_transport_deliver_tap_pkt(struct sk_buff *skb)
175 {
176 	if (virtio_vsock_skb_tap_delivered(skb))
177 		return;
178 
179 	vsock_deliver_tap(virtio_transport_build_skb, skb);
180 	virtio_vsock_skb_set_tap_delivered(skb);
181 }
182 EXPORT_SYMBOL_GPL(virtio_transport_deliver_tap_pkt);
183 
184 static u16 virtio_transport_get_type(struct sock *sk)
185 {
186 	if (sk->sk_type == SOCK_STREAM)
187 		return VIRTIO_VSOCK_TYPE_STREAM;
188 	else
189 		return VIRTIO_VSOCK_TYPE_SEQPACKET;
190 }
191 
192 /* This function can only be used on connecting/connected sockets,
193  * since a socket assigned to a transport is required.
194  *
195  * Do not use on listener sockets!
196  */
197 static int virtio_transport_send_pkt_info(struct vsock_sock *vsk,
198 					  struct virtio_vsock_pkt_info *info)
199 {
200 	u32 src_cid, src_port, dst_cid, dst_port;
201 	const struct virtio_transport *t_ops;
202 	struct virtio_vsock_sock *vvs;
203 	u32 pkt_len = info->pkt_len;
204 	struct sk_buff *skb;
205 
206 	info->type = virtio_transport_get_type(sk_vsock(vsk));
207 
208 	t_ops = virtio_transport_get_ops(vsk);
209 	if (unlikely(!t_ops))
210 		return -EFAULT;
211 
212 	src_cid = t_ops->transport.get_local_cid();
213 	src_port = vsk->local_addr.svm_port;
214 	if (!info->remote_cid) {
215 		dst_cid	= vsk->remote_addr.svm_cid;
216 		dst_port = vsk->remote_addr.svm_port;
217 	} else {
218 		dst_cid = info->remote_cid;
219 		dst_port = info->remote_port;
220 	}
221 
222 	vvs = vsk->trans;
223 
224 	/* we can send less than pkt_len bytes */
225 	if (pkt_len > VIRTIO_VSOCK_MAX_PKT_BUF_SIZE)
226 		pkt_len = VIRTIO_VSOCK_MAX_PKT_BUF_SIZE;
227 
228 	/* virtio_transport_get_credit might return less than pkt_len credit */
229 	pkt_len = virtio_transport_get_credit(vvs, pkt_len);
230 
231 	/* Do not send zero length OP_RW pkt */
232 	if (pkt_len == 0 && info->op == VIRTIO_VSOCK_OP_RW)
233 		return pkt_len;
234 
235 	skb = virtio_transport_alloc_skb(info, pkt_len,
236 					 src_cid, src_port,
237 					 dst_cid, dst_port);
238 	if (!skb) {
239 		virtio_transport_put_credit(vvs, pkt_len);
240 		return -ENOMEM;
241 	}
242 
243 	virtio_transport_inc_tx_pkt(vvs, skb);
244 
245 	return t_ops->send_pkt(skb);
246 }
247 
248 static bool virtio_transport_inc_rx_pkt(struct virtio_vsock_sock *vvs,
249 					u32 len)
250 {
251 	if (vvs->rx_bytes + len > vvs->buf_alloc)
252 		return false;
253 
254 	vvs->rx_bytes += len;
255 	return true;
256 }
257 
258 static void virtio_transport_dec_rx_pkt(struct virtio_vsock_sock *vvs,
259 					u32 len)
260 {
261 	vvs->rx_bytes -= len;
262 	vvs->fwd_cnt += len;
263 }
264 
265 void virtio_transport_inc_tx_pkt(struct virtio_vsock_sock *vvs, struct sk_buff *skb)
266 {
267 	struct virtio_vsock_hdr *hdr = virtio_vsock_hdr(skb);
268 
269 	spin_lock_bh(&vvs->rx_lock);
270 	vvs->last_fwd_cnt = vvs->fwd_cnt;
271 	hdr->fwd_cnt = cpu_to_le32(vvs->fwd_cnt);
272 	hdr->buf_alloc = cpu_to_le32(vvs->buf_alloc);
273 	spin_unlock_bh(&vvs->rx_lock);
274 }
275 EXPORT_SYMBOL_GPL(virtio_transport_inc_tx_pkt);
276 
277 u32 virtio_transport_get_credit(struct virtio_vsock_sock *vvs, u32 credit)
278 {
279 	u32 ret;
280 
281 	spin_lock_bh(&vvs->tx_lock);
282 	ret = vvs->peer_buf_alloc - (vvs->tx_cnt - vvs->peer_fwd_cnt);
283 	if (ret > credit)
284 		ret = credit;
285 	vvs->tx_cnt += ret;
286 	spin_unlock_bh(&vvs->tx_lock);
287 
288 	return ret;
289 }
290 EXPORT_SYMBOL_GPL(virtio_transport_get_credit);
291 
292 void virtio_transport_put_credit(struct virtio_vsock_sock *vvs, u32 credit)
293 {
294 	spin_lock_bh(&vvs->tx_lock);
295 	vvs->tx_cnt -= credit;
296 	spin_unlock_bh(&vvs->tx_lock);
297 }
298 EXPORT_SYMBOL_GPL(virtio_transport_put_credit);
299 
300 static int virtio_transport_send_credit_update(struct vsock_sock *vsk)
301 {
302 	struct virtio_vsock_pkt_info info = {
303 		.op = VIRTIO_VSOCK_OP_CREDIT_UPDATE,
304 		.vsk = vsk,
305 	};
306 
307 	return virtio_transport_send_pkt_info(vsk, &info);
308 }
309 
310 static ssize_t
311 virtio_transport_stream_do_peek(struct vsock_sock *vsk,
312 				struct msghdr *msg,
313 				size_t len)
314 {
315 	struct virtio_vsock_sock *vvs = vsk->trans;
316 	size_t bytes, total = 0, off;
317 	struct sk_buff *skb, *tmp;
318 	int err = -EFAULT;
319 
320 	spin_lock_bh(&vvs->rx_lock);
321 
322 	skb_queue_walk_safe(&vvs->rx_queue, skb,  tmp) {
323 		off = 0;
324 
325 		if (total == len)
326 			break;
327 
328 		while (total < len && off < skb->len) {
329 			bytes = len - total;
330 			if (bytes > skb->len - off)
331 				bytes = skb->len - off;
332 
333 			/* sk_lock is held by caller so no one else can dequeue.
334 			 * Unlock rx_lock since memcpy_to_msg() may sleep.
335 			 */
336 			spin_unlock_bh(&vvs->rx_lock);
337 
338 			err = memcpy_to_msg(msg, skb->data + off, bytes);
339 			if (err)
340 				goto out;
341 
342 			spin_lock_bh(&vvs->rx_lock);
343 
344 			total += bytes;
345 			off += bytes;
346 		}
347 	}
348 
349 	spin_unlock_bh(&vvs->rx_lock);
350 
351 	return total;
352 
353 out:
354 	if (total)
355 		err = total;
356 	return err;
357 }
358 
359 static ssize_t
360 virtio_transport_stream_do_dequeue(struct vsock_sock *vsk,
361 				   struct msghdr *msg,
362 				   size_t len)
363 {
364 	struct virtio_vsock_sock *vvs = vsk->trans;
365 	size_t bytes, total = 0;
366 	struct sk_buff *skb;
367 	int err = -EFAULT;
368 	u32 free_space;
369 
370 	spin_lock_bh(&vvs->rx_lock);
371 
372 	if (WARN_ONCE(skb_queue_empty(&vvs->rx_queue) && vvs->rx_bytes,
373 		      "rx_queue is empty, but rx_bytes is non-zero\n")) {
374 		spin_unlock_bh(&vvs->rx_lock);
375 		return err;
376 	}
377 
378 	while (total < len && !skb_queue_empty(&vvs->rx_queue)) {
379 		skb = skb_peek(&vvs->rx_queue);
380 
381 		bytes = len - total;
382 		if (bytes > skb->len)
383 			bytes = skb->len;
384 
385 		/* sk_lock is held by caller so no one else can dequeue.
386 		 * Unlock rx_lock since memcpy_to_msg() may sleep.
387 		 */
388 		spin_unlock_bh(&vvs->rx_lock);
389 
390 		err = memcpy_to_msg(msg, skb->data, bytes);
391 		if (err)
392 			goto out;
393 
394 		spin_lock_bh(&vvs->rx_lock);
395 
396 		total += bytes;
397 		skb_pull(skb, bytes);
398 
399 		if (skb->len == 0) {
400 			u32 pkt_len = le32_to_cpu(virtio_vsock_hdr(skb)->len);
401 
402 			virtio_transport_dec_rx_pkt(vvs, pkt_len);
403 			__skb_unlink(skb, &vvs->rx_queue);
404 			consume_skb(skb);
405 		}
406 	}
407 
408 	free_space = vvs->buf_alloc - (vvs->fwd_cnt - vvs->last_fwd_cnt);
409 
410 	spin_unlock_bh(&vvs->rx_lock);
411 
412 	/* To reduce the number of credit update messages,
413 	 * don't update credits as long as lots of space is available.
414 	 * Note: the limit chosen here is arbitrary. Setting the limit
415 	 * too high causes extra messages. Too low causes transmitter
416 	 * stalls. As stalls are in theory more expensive than extra
417 	 * messages, we set the limit to a high value. TODO: experiment
418 	 * with different values.
419 	 */
420 	if (free_space < VIRTIO_VSOCK_MAX_PKT_BUF_SIZE)
421 		virtio_transport_send_credit_update(vsk);
422 
423 	return total;
424 
425 out:
426 	if (total)
427 		err = total;
428 	return err;
429 }
430 
431 static int virtio_transport_seqpacket_do_dequeue(struct vsock_sock *vsk,
432 						 struct msghdr *msg,
433 						 int flags)
434 {
435 	struct virtio_vsock_sock *vvs = vsk->trans;
436 	int dequeued_len = 0;
437 	size_t user_buf_len = msg_data_left(msg);
438 	bool msg_ready = false;
439 	struct sk_buff *skb;
440 
441 	spin_lock_bh(&vvs->rx_lock);
442 
443 	if (vvs->msg_count == 0) {
444 		spin_unlock_bh(&vvs->rx_lock);
445 		return 0;
446 	}
447 
448 	while (!msg_ready) {
449 		struct virtio_vsock_hdr *hdr;
450 		size_t pkt_len;
451 
452 		skb = __skb_dequeue(&vvs->rx_queue);
453 		if (!skb)
454 			break;
455 		hdr = virtio_vsock_hdr(skb);
456 		pkt_len = (size_t)le32_to_cpu(hdr->len);
457 
458 		if (dequeued_len >= 0) {
459 			size_t bytes_to_copy;
460 
461 			bytes_to_copy = min(user_buf_len, pkt_len);
462 
463 			if (bytes_to_copy) {
464 				int err;
465 
466 				/* sk_lock is held by caller so no one else can dequeue.
467 				 * Unlock rx_lock since memcpy_to_msg() may sleep.
468 				 */
469 				spin_unlock_bh(&vvs->rx_lock);
470 
471 				err = memcpy_to_msg(msg, skb->data, bytes_to_copy);
472 				if (err) {
473 					/* Copy of message failed. Rest of
474 					 * fragments will be freed without copy.
475 					 */
476 					dequeued_len = err;
477 				} else {
478 					user_buf_len -= bytes_to_copy;
479 				}
480 
481 				spin_lock_bh(&vvs->rx_lock);
482 			}
483 
484 			if (dequeued_len >= 0)
485 				dequeued_len += pkt_len;
486 		}
487 
488 		if (le32_to_cpu(hdr->flags) & VIRTIO_VSOCK_SEQ_EOM) {
489 			msg_ready = true;
490 			vvs->msg_count--;
491 
492 			if (le32_to_cpu(hdr->flags) & VIRTIO_VSOCK_SEQ_EOR)
493 				msg->msg_flags |= MSG_EOR;
494 		}
495 
496 		virtio_transport_dec_rx_pkt(vvs, pkt_len);
497 		kfree_skb(skb);
498 	}
499 
500 	spin_unlock_bh(&vvs->rx_lock);
501 
502 	virtio_transport_send_credit_update(vsk);
503 
504 	return dequeued_len;
505 }
506 
507 ssize_t
508 virtio_transport_stream_dequeue(struct vsock_sock *vsk,
509 				struct msghdr *msg,
510 				size_t len, int flags)
511 {
512 	if (flags & MSG_PEEK)
513 		return virtio_transport_stream_do_peek(vsk, msg, len);
514 	else
515 		return virtio_transport_stream_do_dequeue(vsk, msg, len);
516 }
517 EXPORT_SYMBOL_GPL(virtio_transport_stream_dequeue);
518 
519 ssize_t
520 virtio_transport_seqpacket_dequeue(struct vsock_sock *vsk,
521 				   struct msghdr *msg,
522 				   int flags)
523 {
524 	if (flags & MSG_PEEK)
525 		return -EOPNOTSUPP;
526 
527 	return virtio_transport_seqpacket_do_dequeue(vsk, msg, flags);
528 }
529 EXPORT_SYMBOL_GPL(virtio_transport_seqpacket_dequeue);
530 
531 int
532 virtio_transport_seqpacket_enqueue(struct vsock_sock *vsk,
533 				   struct msghdr *msg,
534 				   size_t len)
535 {
536 	struct virtio_vsock_sock *vvs = vsk->trans;
537 
538 	spin_lock_bh(&vvs->tx_lock);
539 
540 	if (len > vvs->peer_buf_alloc) {
541 		spin_unlock_bh(&vvs->tx_lock);
542 		return -EMSGSIZE;
543 	}
544 
545 	spin_unlock_bh(&vvs->tx_lock);
546 
547 	return virtio_transport_stream_enqueue(vsk, msg, len);
548 }
549 EXPORT_SYMBOL_GPL(virtio_transport_seqpacket_enqueue);
550 
551 int
552 virtio_transport_dgram_dequeue(struct vsock_sock *vsk,
553 			       struct msghdr *msg,
554 			       size_t len, int flags)
555 {
556 	return -EOPNOTSUPP;
557 }
558 EXPORT_SYMBOL_GPL(virtio_transport_dgram_dequeue);
559 
560 s64 virtio_transport_stream_has_data(struct vsock_sock *vsk)
561 {
562 	struct virtio_vsock_sock *vvs = vsk->trans;
563 	s64 bytes;
564 
565 	spin_lock_bh(&vvs->rx_lock);
566 	bytes = vvs->rx_bytes;
567 	spin_unlock_bh(&vvs->rx_lock);
568 
569 	return bytes;
570 }
571 EXPORT_SYMBOL_GPL(virtio_transport_stream_has_data);
572 
573 u32 virtio_transport_seqpacket_has_data(struct vsock_sock *vsk)
574 {
575 	struct virtio_vsock_sock *vvs = vsk->trans;
576 	u32 msg_count;
577 
578 	spin_lock_bh(&vvs->rx_lock);
579 	msg_count = vvs->msg_count;
580 	spin_unlock_bh(&vvs->rx_lock);
581 
582 	return msg_count;
583 }
584 EXPORT_SYMBOL_GPL(virtio_transport_seqpacket_has_data);
585 
586 static s64 virtio_transport_has_space(struct vsock_sock *vsk)
587 {
588 	struct virtio_vsock_sock *vvs = vsk->trans;
589 	s64 bytes;
590 
591 	bytes = vvs->peer_buf_alloc - (vvs->tx_cnt - vvs->peer_fwd_cnt);
592 	if (bytes < 0)
593 		bytes = 0;
594 
595 	return bytes;
596 }
597 
598 s64 virtio_transport_stream_has_space(struct vsock_sock *vsk)
599 {
600 	struct virtio_vsock_sock *vvs = vsk->trans;
601 	s64 bytes;
602 
603 	spin_lock_bh(&vvs->tx_lock);
604 	bytes = virtio_transport_has_space(vsk);
605 	spin_unlock_bh(&vvs->tx_lock);
606 
607 	return bytes;
608 }
609 EXPORT_SYMBOL_GPL(virtio_transport_stream_has_space);
610 
611 int virtio_transport_do_socket_init(struct vsock_sock *vsk,
612 				    struct vsock_sock *psk)
613 {
614 	struct virtio_vsock_sock *vvs;
615 
616 	vvs = kzalloc(sizeof(*vvs), GFP_KERNEL);
617 	if (!vvs)
618 		return -ENOMEM;
619 
620 	vsk->trans = vvs;
621 	vvs->vsk = vsk;
622 	if (psk && psk->trans) {
623 		struct virtio_vsock_sock *ptrans = psk->trans;
624 
625 		vvs->peer_buf_alloc = ptrans->peer_buf_alloc;
626 	}
627 
628 	if (vsk->buffer_size > VIRTIO_VSOCK_MAX_BUF_SIZE)
629 		vsk->buffer_size = VIRTIO_VSOCK_MAX_BUF_SIZE;
630 
631 	vvs->buf_alloc = vsk->buffer_size;
632 
633 	spin_lock_init(&vvs->rx_lock);
634 	spin_lock_init(&vvs->tx_lock);
635 	skb_queue_head_init(&vvs->rx_queue);
636 
637 	return 0;
638 }
639 EXPORT_SYMBOL_GPL(virtio_transport_do_socket_init);
640 
641 /* sk_lock held by the caller */
642 void virtio_transport_notify_buffer_size(struct vsock_sock *vsk, u64 *val)
643 {
644 	struct virtio_vsock_sock *vvs = vsk->trans;
645 
646 	if (*val > VIRTIO_VSOCK_MAX_BUF_SIZE)
647 		*val = VIRTIO_VSOCK_MAX_BUF_SIZE;
648 
649 	vvs->buf_alloc = *val;
650 
651 	virtio_transport_send_credit_update(vsk);
652 }
653 EXPORT_SYMBOL_GPL(virtio_transport_notify_buffer_size);
654 
655 int
656 virtio_transport_notify_poll_in(struct vsock_sock *vsk,
657 				size_t target,
658 				bool *data_ready_now)
659 {
660 	*data_ready_now = vsock_stream_has_data(vsk) >= target;
661 
662 	return 0;
663 }
664 EXPORT_SYMBOL_GPL(virtio_transport_notify_poll_in);
665 
666 int
667 virtio_transport_notify_poll_out(struct vsock_sock *vsk,
668 				 size_t target,
669 				 bool *space_avail_now)
670 {
671 	s64 free_space;
672 
673 	free_space = vsock_stream_has_space(vsk);
674 	if (free_space > 0)
675 		*space_avail_now = true;
676 	else if (free_space == 0)
677 		*space_avail_now = false;
678 
679 	return 0;
680 }
681 EXPORT_SYMBOL_GPL(virtio_transport_notify_poll_out);
682 
683 int virtio_transport_notify_recv_init(struct vsock_sock *vsk,
684 	size_t target, struct vsock_transport_recv_notify_data *data)
685 {
686 	return 0;
687 }
688 EXPORT_SYMBOL_GPL(virtio_transport_notify_recv_init);
689 
690 int virtio_transport_notify_recv_pre_block(struct vsock_sock *vsk,
691 	size_t target, struct vsock_transport_recv_notify_data *data)
692 {
693 	return 0;
694 }
695 EXPORT_SYMBOL_GPL(virtio_transport_notify_recv_pre_block);
696 
697 int virtio_transport_notify_recv_pre_dequeue(struct vsock_sock *vsk,
698 	size_t target, struct vsock_transport_recv_notify_data *data)
699 {
700 	return 0;
701 }
702 EXPORT_SYMBOL_GPL(virtio_transport_notify_recv_pre_dequeue);
703 
704 int virtio_transport_notify_recv_post_dequeue(struct vsock_sock *vsk,
705 	size_t target, ssize_t copied, bool data_read,
706 	struct vsock_transport_recv_notify_data *data)
707 {
708 	return 0;
709 }
710 EXPORT_SYMBOL_GPL(virtio_transport_notify_recv_post_dequeue);
711 
712 int virtio_transport_notify_send_init(struct vsock_sock *vsk,
713 	struct vsock_transport_send_notify_data *data)
714 {
715 	return 0;
716 }
717 EXPORT_SYMBOL_GPL(virtio_transport_notify_send_init);
718 
719 int virtio_transport_notify_send_pre_block(struct vsock_sock *vsk,
720 	struct vsock_transport_send_notify_data *data)
721 {
722 	return 0;
723 }
724 EXPORT_SYMBOL_GPL(virtio_transport_notify_send_pre_block);
725 
726 int virtio_transport_notify_send_pre_enqueue(struct vsock_sock *vsk,
727 	struct vsock_transport_send_notify_data *data)
728 {
729 	return 0;
730 }
731 EXPORT_SYMBOL_GPL(virtio_transport_notify_send_pre_enqueue);
732 
733 int virtio_transport_notify_send_post_enqueue(struct vsock_sock *vsk,
734 	ssize_t written, struct vsock_transport_send_notify_data *data)
735 {
736 	return 0;
737 }
738 EXPORT_SYMBOL_GPL(virtio_transport_notify_send_post_enqueue);
739 
740 u64 virtio_transport_stream_rcvhiwat(struct vsock_sock *vsk)
741 {
742 	return vsk->buffer_size;
743 }
744 EXPORT_SYMBOL_GPL(virtio_transport_stream_rcvhiwat);
745 
746 bool virtio_transport_stream_is_active(struct vsock_sock *vsk)
747 {
748 	return true;
749 }
750 EXPORT_SYMBOL_GPL(virtio_transport_stream_is_active);
751 
752 bool virtio_transport_stream_allow(u32 cid, u32 port)
753 {
754 	return true;
755 }
756 EXPORT_SYMBOL_GPL(virtio_transport_stream_allow);
757 
758 int virtio_transport_dgram_bind(struct vsock_sock *vsk,
759 				struct sockaddr_vm *addr)
760 {
761 	return -EOPNOTSUPP;
762 }
763 EXPORT_SYMBOL_GPL(virtio_transport_dgram_bind);
764 
765 bool virtio_transport_dgram_allow(u32 cid, u32 port)
766 {
767 	return false;
768 }
769 EXPORT_SYMBOL_GPL(virtio_transport_dgram_allow);
770 
771 int virtio_transport_connect(struct vsock_sock *vsk)
772 {
773 	struct virtio_vsock_pkt_info info = {
774 		.op = VIRTIO_VSOCK_OP_REQUEST,
775 		.vsk = vsk,
776 	};
777 
778 	return virtio_transport_send_pkt_info(vsk, &info);
779 }
780 EXPORT_SYMBOL_GPL(virtio_transport_connect);
781 
782 int virtio_transport_shutdown(struct vsock_sock *vsk, int mode)
783 {
784 	struct virtio_vsock_pkt_info info = {
785 		.op = VIRTIO_VSOCK_OP_SHUTDOWN,
786 		.flags = (mode & RCV_SHUTDOWN ?
787 			  VIRTIO_VSOCK_SHUTDOWN_RCV : 0) |
788 			 (mode & SEND_SHUTDOWN ?
789 			  VIRTIO_VSOCK_SHUTDOWN_SEND : 0),
790 		.vsk = vsk,
791 	};
792 
793 	return virtio_transport_send_pkt_info(vsk, &info);
794 }
795 EXPORT_SYMBOL_GPL(virtio_transport_shutdown);
796 
797 int
798 virtio_transport_dgram_enqueue(struct vsock_sock *vsk,
799 			       struct sockaddr_vm *remote_addr,
800 			       struct msghdr *msg,
801 			       size_t dgram_len)
802 {
803 	return -EOPNOTSUPP;
804 }
805 EXPORT_SYMBOL_GPL(virtio_transport_dgram_enqueue);
806 
807 ssize_t
808 virtio_transport_stream_enqueue(struct vsock_sock *vsk,
809 				struct msghdr *msg,
810 				size_t len)
811 {
812 	struct virtio_vsock_pkt_info info = {
813 		.op = VIRTIO_VSOCK_OP_RW,
814 		.msg = msg,
815 		.pkt_len = len,
816 		.vsk = vsk,
817 	};
818 
819 	return virtio_transport_send_pkt_info(vsk, &info);
820 }
821 EXPORT_SYMBOL_GPL(virtio_transport_stream_enqueue);
822 
823 void virtio_transport_destruct(struct vsock_sock *vsk)
824 {
825 	struct virtio_vsock_sock *vvs = vsk->trans;
826 
827 	kfree(vvs);
828 }
829 EXPORT_SYMBOL_GPL(virtio_transport_destruct);
830 
831 static int virtio_transport_reset(struct vsock_sock *vsk,
832 				  struct sk_buff *skb)
833 {
834 	struct virtio_vsock_pkt_info info = {
835 		.op = VIRTIO_VSOCK_OP_RST,
836 		.reply = !!skb,
837 		.vsk = vsk,
838 	};
839 
840 	/* Send RST only if the original pkt is not a RST pkt */
841 	if (skb && le16_to_cpu(virtio_vsock_hdr(skb)->op) == VIRTIO_VSOCK_OP_RST)
842 		return 0;
843 
844 	return virtio_transport_send_pkt_info(vsk, &info);
845 }
846 
847 /* Normally packets are associated with a socket.  There may be no socket if an
848  * attempt was made to connect to a socket that does not exist.
849  */
850 static int virtio_transport_reset_no_sock(const struct virtio_transport *t,
851 					  struct sk_buff *skb)
852 {
853 	struct virtio_vsock_hdr *hdr = virtio_vsock_hdr(skb);
854 	struct virtio_vsock_pkt_info info = {
855 		.op = VIRTIO_VSOCK_OP_RST,
856 		.type = le16_to_cpu(hdr->type),
857 		.reply = true,
858 	};
859 	struct sk_buff *reply;
860 
861 	/* Send RST only if the original pkt is not a RST pkt */
862 	if (le16_to_cpu(hdr->op) == VIRTIO_VSOCK_OP_RST)
863 		return 0;
864 
865 	reply = virtio_transport_alloc_skb(&info, 0,
866 					   le64_to_cpu(hdr->dst_cid),
867 					   le32_to_cpu(hdr->dst_port),
868 					   le64_to_cpu(hdr->src_cid),
869 					   le32_to_cpu(hdr->src_port));
870 	if (!reply)
871 		return -ENOMEM;
872 
873 	if (!t) {
874 		kfree_skb(reply);
875 		return -ENOTCONN;
876 	}
877 
878 	return t->send_pkt(reply);
879 }
880 
881 /* This function should be called with sk_lock held and SOCK_DONE set */
882 static void virtio_transport_remove_sock(struct vsock_sock *vsk)
883 {
884 	struct virtio_vsock_sock *vvs = vsk->trans;
885 
886 	/* We don't need to take rx_lock, as the socket is closing and we are
887 	 * removing it.
888 	 */
889 	__skb_queue_purge(&vvs->rx_queue);
890 	vsock_remove_sock(vsk);
891 }
892 
893 static void virtio_transport_wait_close(struct sock *sk, long timeout)
894 {
895 	if (timeout) {
896 		DEFINE_WAIT_FUNC(wait, woken_wake_function);
897 
898 		add_wait_queue(sk_sleep(sk), &wait);
899 
900 		do {
901 			if (sk_wait_event(sk, &timeout,
902 					  sock_flag(sk, SOCK_DONE), &wait))
903 				break;
904 		} while (!signal_pending(current) && timeout);
905 
906 		remove_wait_queue(sk_sleep(sk), &wait);
907 	}
908 }
909 
910 static void virtio_transport_do_close(struct vsock_sock *vsk,
911 				      bool cancel_timeout)
912 {
913 	struct sock *sk = sk_vsock(vsk);
914 
915 	sock_set_flag(sk, SOCK_DONE);
916 	vsk->peer_shutdown = SHUTDOWN_MASK;
917 	if (vsock_stream_has_data(vsk) <= 0)
918 		sk->sk_state = TCP_CLOSING;
919 	sk->sk_state_change(sk);
920 
921 	if (vsk->close_work_scheduled &&
922 	    (!cancel_timeout || cancel_delayed_work(&vsk->close_work))) {
923 		vsk->close_work_scheduled = false;
924 
925 		virtio_transport_remove_sock(vsk);
926 
927 		/* Release refcnt obtained when we scheduled the timeout */
928 		sock_put(sk);
929 	}
930 }
931 
932 static void virtio_transport_close_timeout(struct work_struct *work)
933 {
934 	struct vsock_sock *vsk =
935 		container_of(work, struct vsock_sock, close_work.work);
936 	struct sock *sk = sk_vsock(vsk);
937 
938 	sock_hold(sk);
939 	lock_sock(sk);
940 
941 	if (!sock_flag(sk, SOCK_DONE)) {
942 		(void)virtio_transport_reset(vsk, NULL);
943 
944 		virtio_transport_do_close(vsk, false);
945 	}
946 
947 	vsk->close_work_scheduled = false;
948 
949 	release_sock(sk);
950 	sock_put(sk);
951 }
952 
953 /* User context, vsk->sk is locked */
954 static bool virtio_transport_close(struct vsock_sock *vsk)
955 {
956 	struct sock *sk = &vsk->sk;
957 
958 	if (!(sk->sk_state == TCP_ESTABLISHED ||
959 	      sk->sk_state == TCP_CLOSING))
960 		return true;
961 
962 	/* Already received SHUTDOWN from peer, reply with RST */
963 	if ((vsk->peer_shutdown & SHUTDOWN_MASK) == SHUTDOWN_MASK) {
964 		(void)virtio_transport_reset(vsk, NULL);
965 		return true;
966 	}
967 
968 	if ((sk->sk_shutdown & SHUTDOWN_MASK) != SHUTDOWN_MASK)
969 		(void)virtio_transport_shutdown(vsk, SHUTDOWN_MASK);
970 
971 	if (sock_flag(sk, SOCK_LINGER) && !(current->flags & PF_EXITING))
972 		virtio_transport_wait_close(sk, sk->sk_lingertime);
973 
974 	if (sock_flag(sk, SOCK_DONE)) {
975 		return true;
976 	}
977 
978 	sock_hold(sk);
979 	INIT_DELAYED_WORK(&vsk->close_work,
980 			  virtio_transport_close_timeout);
981 	vsk->close_work_scheduled = true;
982 	schedule_delayed_work(&vsk->close_work, VSOCK_CLOSE_TIMEOUT);
983 	return false;
984 }
985 
986 void virtio_transport_release(struct vsock_sock *vsk)
987 {
988 	struct sock *sk = &vsk->sk;
989 	bool remove_sock = true;
990 
991 	if (sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET)
992 		remove_sock = virtio_transport_close(vsk);
993 
994 	if (remove_sock) {
995 		sock_set_flag(sk, SOCK_DONE);
996 		virtio_transport_remove_sock(vsk);
997 	}
998 }
999 EXPORT_SYMBOL_GPL(virtio_transport_release);
1000 
1001 static int
1002 virtio_transport_recv_connecting(struct sock *sk,
1003 				 struct sk_buff *skb)
1004 {
1005 	struct virtio_vsock_hdr *hdr = virtio_vsock_hdr(skb);
1006 	struct vsock_sock *vsk = vsock_sk(sk);
1007 	int skerr;
1008 	int err;
1009 
1010 	switch (le16_to_cpu(hdr->op)) {
1011 	case VIRTIO_VSOCK_OP_RESPONSE:
1012 		sk->sk_state = TCP_ESTABLISHED;
1013 		sk->sk_socket->state = SS_CONNECTED;
1014 		vsock_insert_connected(vsk);
1015 		sk->sk_state_change(sk);
1016 		break;
1017 	case VIRTIO_VSOCK_OP_INVALID:
1018 		break;
1019 	case VIRTIO_VSOCK_OP_RST:
1020 		skerr = ECONNRESET;
1021 		err = 0;
1022 		goto destroy;
1023 	default:
1024 		skerr = EPROTO;
1025 		err = -EINVAL;
1026 		goto destroy;
1027 	}
1028 	return 0;
1029 
1030 destroy:
1031 	virtio_transport_reset(vsk, skb);
1032 	sk->sk_state = TCP_CLOSE;
1033 	sk->sk_err = skerr;
1034 	sk_error_report(sk);
1035 	return err;
1036 }
1037 
1038 static void
1039 virtio_transport_recv_enqueue(struct vsock_sock *vsk,
1040 			      struct sk_buff *skb)
1041 {
1042 	struct virtio_vsock_sock *vvs = vsk->trans;
1043 	bool can_enqueue, free_pkt = false;
1044 	struct virtio_vsock_hdr *hdr;
1045 	u32 len;
1046 
1047 	hdr = virtio_vsock_hdr(skb);
1048 	len = le32_to_cpu(hdr->len);
1049 
1050 	spin_lock_bh(&vvs->rx_lock);
1051 
1052 	can_enqueue = virtio_transport_inc_rx_pkt(vvs, len);
1053 	if (!can_enqueue) {
1054 		free_pkt = true;
1055 		goto out;
1056 	}
1057 
1058 	if (le32_to_cpu(hdr->flags) & VIRTIO_VSOCK_SEQ_EOM)
1059 		vvs->msg_count++;
1060 
1061 	/* Try to copy small packets into the buffer of last packet queued,
1062 	 * to avoid wasting memory queueing the entire buffer with a small
1063 	 * payload.
1064 	 */
1065 	if (len <= GOOD_COPY_LEN && !skb_queue_empty(&vvs->rx_queue)) {
1066 		struct virtio_vsock_hdr *last_hdr;
1067 		struct sk_buff *last_skb;
1068 
1069 		last_skb = skb_peek_tail(&vvs->rx_queue);
1070 		last_hdr = virtio_vsock_hdr(last_skb);
1071 
1072 		/* If there is space in the last packet queued, we copy the
1073 		 * new packet in its buffer. We avoid this if the last packet
1074 		 * queued has VIRTIO_VSOCK_SEQ_EOM set, because this is
1075 		 * delimiter of SEQPACKET message, so 'pkt' is the first packet
1076 		 * of a new message.
1077 		 */
1078 		if (skb->len < skb_tailroom(last_skb) &&
1079 		    !(le32_to_cpu(last_hdr->flags) & VIRTIO_VSOCK_SEQ_EOM)) {
1080 			memcpy(skb_put(last_skb, skb->len), skb->data, skb->len);
1081 			free_pkt = true;
1082 			last_hdr->flags |= hdr->flags;
1083 			le32_add_cpu(&last_hdr->len, len);
1084 			goto out;
1085 		}
1086 	}
1087 
1088 	__skb_queue_tail(&vvs->rx_queue, skb);
1089 
1090 out:
1091 	spin_unlock_bh(&vvs->rx_lock);
1092 	if (free_pkt)
1093 		kfree_skb(skb);
1094 }
1095 
1096 static int
1097 virtio_transport_recv_connected(struct sock *sk,
1098 				struct sk_buff *skb)
1099 {
1100 	struct virtio_vsock_hdr *hdr = virtio_vsock_hdr(skb);
1101 	struct vsock_sock *vsk = vsock_sk(sk);
1102 	int err = 0;
1103 
1104 	switch (le16_to_cpu(hdr->op)) {
1105 	case VIRTIO_VSOCK_OP_RW:
1106 		virtio_transport_recv_enqueue(vsk, skb);
1107 		vsock_data_ready(sk);
1108 		return err;
1109 	case VIRTIO_VSOCK_OP_CREDIT_REQUEST:
1110 		virtio_transport_send_credit_update(vsk);
1111 		break;
1112 	case VIRTIO_VSOCK_OP_CREDIT_UPDATE:
1113 		sk->sk_write_space(sk);
1114 		break;
1115 	case VIRTIO_VSOCK_OP_SHUTDOWN:
1116 		if (le32_to_cpu(hdr->flags) & VIRTIO_VSOCK_SHUTDOWN_RCV)
1117 			vsk->peer_shutdown |= RCV_SHUTDOWN;
1118 		if (le32_to_cpu(hdr->flags) & VIRTIO_VSOCK_SHUTDOWN_SEND)
1119 			vsk->peer_shutdown |= SEND_SHUTDOWN;
1120 		if (vsk->peer_shutdown == SHUTDOWN_MASK &&
1121 		    vsock_stream_has_data(vsk) <= 0 &&
1122 		    !sock_flag(sk, SOCK_DONE)) {
1123 			(void)virtio_transport_reset(vsk, NULL);
1124 			virtio_transport_do_close(vsk, true);
1125 		}
1126 		if (le32_to_cpu(virtio_vsock_hdr(skb)->flags))
1127 			sk->sk_state_change(sk);
1128 		break;
1129 	case VIRTIO_VSOCK_OP_RST:
1130 		virtio_transport_do_close(vsk, true);
1131 		break;
1132 	default:
1133 		err = -EINVAL;
1134 		break;
1135 	}
1136 
1137 	kfree_skb(skb);
1138 	return err;
1139 }
1140 
1141 static void
1142 virtio_transport_recv_disconnecting(struct sock *sk,
1143 				    struct sk_buff *skb)
1144 {
1145 	struct virtio_vsock_hdr *hdr = virtio_vsock_hdr(skb);
1146 	struct vsock_sock *vsk = vsock_sk(sk);
1147 
1148 	if (le16_to_cpu(hdr->op) == VIRTIO_VSOCK_OP_RST)
1149 		virtio_transport_do_close(vsk, true);
1150 }
1151 
1152 static int
1153 virtio_transport_send_response(struct vsock_sock *vsk,
1154 			       struct sk_buff *skb)
1155 {
1156 	struct virtio_vsock_hdr *hdr = virtio_vsock_hdr(skb);
1157 	struct virtio_vsock_pkt_info info = {
1158 		.op = VIRTIO_VSOCK_OP_RESPONSE,
1159 		.remote_cid = le64_to_cpu(hdr->src_cid),
1160 		.remote_port = le32_to_cpu(hdr->src_port),
1161 		.reply = true,
1162 		.vsk = vsk,
1163 	};
1164 
1165 	return virtio_transport_send_pkt_info(vsk, &info);
1166 }
1167 
1168 static bool virtio_transport_space_update(struct sock *sk,
1169 					  struct sk_buff *skb)
1170 {
1171 	struct virtio_vsock_hdr *hdr = virtio_vsock_hdr(skb);
1172 	struct vsock_sock *vsk = vsock_sk(sk);
1173 	struct virtio_vsock_sock *vvs = vsk->trans;
1174 	bool space_available;
1175 
1176 	/* Listener sockets are not associated with any transport, so we are
1177 	 * not able to take the state to see if there is space available in the
1178 	 * remote peer, but since they are only used to receive requests, we
1179 	 * can assume that there is always space available in the other peer.
1180 	 */
1181 	if (!vvs)
1182 		return true;
1183 
1184 	/* buf_alloc and fwd_cnt is always included in the hdr */
1185 	spin_lock_bh(&vvs->tx_lock);
1186 	vvs->peer_buf_alloc = le32_to_cpu(hdr->buf_alloc);
1187 	vvs->peer_fwd_cnt = le32_to_cpu(hdr->fwd_cnt);
1188 	space_available = virtio_transport_has_space(vsk);
1189 	spin_unlock_bh(&vvs->tx_lock);
1190 	return space_available;
1191 }
1192 
1193 /* Handle server socket */
1194 static int
1195 virtio_transport_recv_listen(struct sock *sk, struct sk_buff *skb,
1196 			     struct virtio_transport *t)
1197 {
1198 	struct virtio_vsock_hdr *hdr = virtio_vsock_hdr(skb);
1199 	struct vsock_sock *vsk = vsock_sk(sk);
1200 	struct vsock_sock *vchild;
1201 	struct sock *child;
1202 	int ret;
1203 
1204 	if (le16_to_cpu(hdr->op) != VIRTIO_VSOCK_OP_REQUEST) {
1205 		virtio_transport_reset_no_sock(t, skb);
1206 		return -EINVAL;
1207 	}
1208 
1209 	if (sk_acceptq_is_full(sk)) {
1210 		virtio_transport_reset_no_sock(t, skb);
1211 		return -ENOMEM;
1212 	}
1213 
1214 	child = vsock_create_connected(sk);
1215 	if (!child) {
1216 		virtio_transport_reset_no_sock(t, skb);
1217 		return -ENOMEM;
1218 	}
1219 
1220 	sk_acceptq_added(sk);
1221 
1222 	lock_sock_nested(child, SINGLE_DEPTH_NESTING);
1223 
1224 	child->sk_state = TCP_ESTABLISHED;
1225 
1226 	vchild = vsock_sk(child);
1227 	vsock_addr_init(&vchild->local_addr, le64_to_cpu(hdr->dst_cid),
1228 			le32_to_cpu(hdr->dst_port));
1229 	vsock_addr_init(&vchild->remote_addr, le64_to_cpu(hdr->src_cid),
1230 			le32_to_cpu(hdr->src_port));
1231 
1232 	ret = vsock_assign_transport(vchild, vsk);
1233 	/* Transport assigned (looking at remote_addr) must be the same
1234 	 * where we received the request.
1235 	 */
1236 	if (ret || vchild->transport != &t->transport) {
1237 		release_sock(child);
1238 		virtio_transport_reset_no_sock(t, skb);
1239 		sock_put(child);
1240 		return ret;
1241 	}
1242 
1243 	if (virtio_transport_space_update(child, skb))
1244 		child->sk_write_space(child);
1245 
1246 	vsock_insert_connected(vchild);
1247 	vsock_enqueue_accept(sk, child);
1248 	virtio_transport_send_response(vchild, skb);
1249 
1250 	release_sock(child);
1251 
1252 	sk->sk_data_ready(sk);
1253 	return 0;
1254 }
1255 
1256 static bool virtio_transport_valid_type(u16 type)
1257 {
1258 	return (type == VIRTIO_VSOCK_TYPE_STREAM) ||
1259 	       (type == VIRTIO_VSOCK_TYPE_SEQPACKET);
1260 }
1261 
1262 /* We are under the virtio-vsock's vsock->rx_lock or vhost-vsock's vq->mutex
1263  * lock.
1264  */
1265 void virtio_transport_recv_pkt(struct virtio_transport *t,
1266 			       struct sk_buff *skb)
1267 {
1268 	struct virtio_vsock_hdr *hdr = virtio_vsock_hdr(skb);
1269 	struct sockaddr_vm src, dst;
1270 	struct vsock_sock *vsk;
1271 	struct sock *sk;
1272 	bool space_available;
1273 
1274 	vsock_addr_init(&src, le64_to_cpu(hdr->src_cid),
1275 			le32_to_cpu(hdr->src_port));
1276 	vsock_addr_init(&dst, le64_to_cpu(hdr->dst_cid),
1277 			le32_to_cpu(hdr->dst_port));
1278 
1279 	trace_virtio_transport_recv_pkt(src.svm_cid, src.svm_port,
1280 					dst.svm_cid, dst.svm_port,
1281 					le32_to_cpu(hdr->len),
1282 					le16_to_cpu(hdr->type),
1283 					le16_to_cpu(hdr->op),
1284 					le32_to_cpu(hdr->flags),
1285 					le32_to_cpu(hdr->buf_alloc),
1286 					le32_to_cpu(hdr->fwd_cnt));
1287 
1288 	if (!virtio_transport_valid_type(le16_to_cpu(hdr->type))) {
1289 		(void)virtio_transport_reset_no_sock(t, skb);
1290 		goto free_pkt;
1291 	}
1292 
1293 	/* The socket must be in connected or bound table
1294 	 * otherwise send reset back
1295 	 */
1296 	sk = vsock_find_connected_socket(&src, &dst);
1297 	if (!sk) {
1298 		sk = vsock_find_bound_socket(&dst);
1299 		if (!sk) {
1300 			(void)virtio_transport_reset_no_sock(t, skb);
1301 			goto free_pkt;
1302 		}
1303 	}
1304 
1305 	if (virtio_transport_get_type(sk) != le16_to_cpu(hdr->type)) {
1306 		(void)virtio_transport_reset_no_sock(t, skb);
1307 		sock_put(sk);
1308 		goto free_pkt;
1309 	}
1310 
1311 	if (!skb_set_owner_sk_safe(skb, sk)) {
1312 		WARN_ONCE(1, "receiving vsock socket has sk_refcnt == 0\n");
1313 		goto free_pkt;
1314 	}
1315 
1316 	vsk = vsock_sk(sk);
1317 
1318 	lock_sock(sk);
1319 
1320 	/* Check if sk has been closed before lock_sock */
1321 	if (sock_flag(sk, SOCK_DONE)) {
1322 		(void)virtio_transport_reset_no_sock(t, skb);
1323 		release_sock(sk);
1324 		sock_put(sk);
1325 		goto free_pkt;
1326 	}
1327 
1328 	space_available = virtio_transport_space_update(sk, skb);
1329 
1330 	/* Update CID in case it has changed after a transport reset event */
1331 	if (vsk->local_addr.svm_cid != VMADDR_CID_ANY)
1332 		vsk->local_addr.svm_cid = dst.svm_cid;
1333 
1334 	if (space_available)
1335 		sk->sk_write_space(sk);
1336 
1337 	switch (sk->sk_state) {
1338 	case TCP_LISTEN:
1339 		virtio_transport_recv_listen(sk, skb, t);
1340 		kfree_skb(skb);
1341 		break;
1342 	case TCP_SYN_SENT:
1343 		virtio_transport_recv_connecting(sk, skb);
1344 		kfree_skb(skb);
1345 		break;
1346 	case TCP_ESTABLISHED:
1347 		virtio_transport_recv_connected(sk, skb);
1348 		break;
1349 	case TCP_CLOSING:
1350 		virtio_transport_recv_disconnecting(sk, skb);
1351 		kfree_skb(skb);
1352 		break;
1353 	default:
1354 		(void)virtio_transport_reset_no_sock(t, skb);
1355 		kfree_skb(skb);
1356 		break;
1357 	}
1358 
1359 	release_sock(sk);
1360 
1361 	/* Release refcnt obtained when we fetched this socket out of the
1362 	 * bound or connected list.
1363 	 */
1364 	sock_put(sk);
1365 	return;
1366 
1367 free_pkt:
1368 	kfree_skb(skb);
1369 }
1370 EXPORT_SYMBOL_GPL(virtio_transport_recv_pkt);
1371 
1372 /* Remove skbs found in a queue that have a vsk that matches.
1373  *
1374  * Each skb is freed.
1375  *
1376  * Returns the count of skbs that were reply packets.
1377  */
1378 int virtio_transport_purge_skbs(void *vsk, struct sk_buff_head *queue)
1379 {
1380 	struct sk_buff_head freeme;
1381 	struct sk_buff *skb, *tmp;
1382 	int cnt = 0;
1383 
1384 	skb_queue_head_init(&freeme);
1385 
1386 	spin_lock_bh(&queue->lock);
1387 	skb_queue_walk_safe(queue, skb, tmp) {
1388 		if (vsock_sk(skb->sk) != vsk)
1389 			continue;
1390 
1391 		__skb_unlink(skb, queue);
1392 		__skb_queue_tail(&freeme, skb);
1393 
1394 		if (virtio_vsock_skb_reply(skb))
1395 			cnt++;
1396 	}
1397 	spin_unlock_bh(&queue->lock);
1398 
1399 	__skb_queue_purge(&freeme);
1400 
1401 	return cnt;
1402 }
1403 EXPORT_SYMBOL_GPL(virtio_transport_purge_skbs);
1404 
1405 MODULE_LICENSE("GPL v2");
1406 MODULE_AUTHOR("Asias He");
1407 MODULE_DESCRIPTION("common code for virtio vsock");
1408