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