1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * Hyper-V transport for vsock
4  *
5  * Hyper-V Sockets supplies a byte-stream based communication mechanism
6  * between the host and the VM. This driver implements the necessary
7  * support in the VM by introducing the new vsock transport.
8  *
9  * Copyright (c) 2017, Microsoft Corporation.
10  */
11 #include <linux/module.h>
12 #include <linux/vmalloc.h>
13 #include <linux/hyperv.h>
14 #include <net/sock.h>
15 #include <net/af_vsock.h>
16 
17 /* The host side's design of the feature requires 6 exact 4KB pages for
18  * recv/send rings respectively -- this is suboptimal considering memory
19  * consumption, however unluckily we have to live with it, before the
20  * host comes up with a better design in the future.
21  */
22 #define PAGE_SIZE_4K		4096
23 #define RINGBUFFER_HVS_RCV_SIZE (PAGE_SIZE_4K * 6)
24 #define RINGBUFFER_HVS_SND_SIZE (PAGE_SIZE_4K * 6)
25 
26 /* The MTU is 16KB per the host side's design */
27 #define HVS_MTU_SIZE		(1024 * 16)
28 
29 /* How long to wait for graceful shutdown of a connection */
30 #define HVS_CLOSE_TIMEOUT (8 * HZ)
31 
32 struct vmpipe_proto_header {
33 	u32 pkt_type;
34 	u32 data_size;
35 };
36 
37 /* For recv, we use the VMBus in-place packet iterator APIs to directly copy
38  * data from the ringbuffer into the userspace buffer.
39  */
40 struct hvs_recv_buf {
41 	/* The header before the payload data */
42 	struct vmpipe_proto_header hdr;
43 
44 	/* The payload */
45 	u8 data[HVS_MTU_SIZE];
46 };
47 
48 /* We can send up to HVS_MTU_SIZE bytes of payload to the host, but let's use
49  * a small size, i.e. HVS_SEND_BUF_SIZE, to minimize the dynamically-allocated
50  * buffer, because tests show there is no significant performance difference.
51  *
52  * Note: the buffer can be eliminated in the future when we add new VMBus
53  * ringbuffer APIs that allow us to directly copy data from userspace buffer
54  * to VMBus ringbuffer.
55  */
56 #define HVS_SEND_BUF_SIZE (PAGE_SIZE_4K - sizeof(struct vmpipe_proto_header))
57 
58 struct hvs_send_buf {
59 	/* The header before the payload data */
60 	struct vmpipe_proto_header hdr;
61 
62 	/* The payload */
63 	u8 data[HVS_SEND_BUF_SIZE];
64 };
65 
66 #define HVS_HEADER_LEN	(sizeof(struct vmpacket_descriptor) + \
67 			 sizeof(struct vmpipe_proto_header))
68 
69 /* See 'prev_indices' in hv_ringbuffer_read(), hv_ringbuffer_write(), and
70  * __hv_pkt_iter_next().
71  */
72 #define VMBUS_PKT_TRAILER_SIZE	(sizeof(u64))
73 
74 #define HVS_PKT_LEN(payload_len)	(HVS_HEADER_LEN + \
75 					 ALIGN((payload_len), 8) + \
76 					 VMBUS_PKT_TRAILER_SIZE)
77 
78 union hvs_service_id {
79 	uuid_le	srv_id;
80 
81 	struct {
82 		unsigned int svm_port;
83 		unsigned char b[sizeof(uuid_le) - sizeof(unsigned int)];
84 	};
85 };
86 
87 /* Per-socket state (accessed via vsk->trans) */
88 struct hvsock {
89 	struct vsock_sock *vsk;
90 
91 	uuid_le vm_srv_id;
92 	uuid_le host_srv_id;
93 
94 	struct vmbus_channel *chan;
95 	struct vmpacket_descriptor *recv_desc;
96 
97 	/* The length of the payload not delivered to userland yet */
98 	u32 recv_data_len;
99 	/* The offset of the payload */
100 	u32 recv_data_off;
101 
102 	/* Have we sent the zero-length packet (FIN)? */
103 	bool fin_sent;
104 };
105 
106 /* In the VM, we support Hyper-V Sockets with AF_VSOCK, and the endpoint is
107  * <cid, port> (see struct sockaddr_vm). Note: cid is not really used here:
108  * when we write apps to connect to the host, we can only use VMADDR_CID_ANY
109  * or VMADDR_CID_HOST (both are equivalent) as the remote cid, and when we
110  * write apps to bind() & listen() in the VM, we can only use VMADDR_CID_ANY
111  * as the local cid.
112  *
113  * On the host, Hyper-V Sockets are supported by Winsock AF_HYPERV:
114  * https://docs.microsoft.com/en-us/virtualization/hyper-v-on-windows/user-
115  * guide/make-integration-service, and the endpoint is <VmID, ServiceId> with
116  * the below sockaddr:
117  *
118  * struct SOCKADDR_HV
119  * {
120  *    ADDRESS_FAMILY Family;
121  *    USHORT Reserved;
122  *    GUID VmId;
123  *    GUID ServiceId;
124  * };
125  * Note: VmID is not used by Linux VM and actually it isn't transmitted via
126  * VMBus, because here it's obvious the host and the VM can easily identify
127  * each other. Though the VmID is useful on the host, especially in the case
128  * of Windows container, Linux VM doesn't need it at all.
129  *
130  * To make use of the AF_VSOCK infrastructure in Linux VM, we have to limit
131  * the available GUID space of SOCKADDR_HV so that we can create a mapping
132  * between AF_VSOCK port and SOCKADDR_HV Service GUID. The rule of writing
133  * Hyper-V Sockets apps on the host and in Linux VM is:
134  *
135  ****************************************************************************
136  * The only valid Service GUIDs, from the perspectives of both the host and *
137  * Linux VM, that can be connected by the other end, must conform to this   *
138  * format: <port>-facb-11e6-bd58-64006a7986d3, and the "port" must be in    *
139  * this range [0, 0x7FFFFFFF].                                              *
140  ****************************************************************************
141  *
142  * When we write apps on the host to connect(), the GUID ServiceID is used.
143  * When we write apps in Linux VM to connect(), we only need to specify the
144  * port and the driver will form the GUID and use that to request the host.
145  *
146  * From the perspective of Linux VM:
147  * 1. the local ephemeral port (i.e. the local auto-bound port when we call
148  * connect() without explicit bind()) is generated by __vsock_bind_stream(),
149  * and the range is [1024, 0xFFFFFFFF).
150  * 2. the remote ephemeral port (i.e. the auto-generated remote port for
151  * a connect request initiated by the host's connect()) is generated by
152  * hvs_remote_addr_init() and the range is [0x80000000, 0xFFFFFFFF).
153  */
154 
155 #define MAX_LISTEN_PORT			((u32)0x7FFFFFFF)
156 #define MAX_VM_LISTEN_PORT		MAX_LISTEN_PORT
157 #define MAX_HOST_LISTEN_PORT		MAX_LISTEN_PORT
158 #define MIN_HOST_EPHEMERAL_PORT		(MAX_HOST_LISTEN_PORT + 1)
159 
160 /* 00000000-facb-11e6-bd58-64006a7986d3 */
161 static const uuid_le srv_id_template =
162 	UUID_LE(0x00000000, 0xfacb, 0x11e6, 0xbd, 0x58,
163 		0x64, 0x00, 0x6a, 0x79, 0x86, 0xd3);
164 
165 static bool is_valid_srv_id(const uuid_le *id)
166 {
167 	return !memcmp(&id->b[4], &srv_id_template.b[4], sizeof(uuid_le) - 4);
168 }
169 
170 static unsigned int get_port_by_srv_id(const uuid_le *svr_id)
171 {
172 	return *((unsigned int *)svr_id);
173 }
174 
175 static void hvs_addr_init(struct sockaddr_vm *addr, const uuid_le *svr_id)
176 {
177 	unsigned int port = get_port_by_srv_id(svr_id);
178 
179 	vsock_addr_init(addr, VMADDR_CID_ANY, port);
180 }
181 
182 static void hvs_remote_addr_init(struct sockaddr_vm *remote,
183 				 struct sockaddr_vm *local)
184 {
185 	static u32 host_ephemeral_port = MIN_HOST_EPHEMERAL_PORT;
186 	struct sock *sk;
187 
188 	vsock_addr_init(remote, VMADDR_CID_ANY, VMADDR_PORT_ANY);
189 
190 	while (1) {
191 		/* Wrap around ? */
192 		if (host_ephemeral_port < MIN_HOST_EPHEMERAL_PORT ||
193 		    host_ephemeral_port == VMADDR_PORT_ANY)
194 			host_ephemeral_port = MIN_HOST_EPHEMERAL_PORT;
195 
196 		remote->svm_port = host_ephemeral_port++;
197 
198 		sk = vsock_find_connected_socket(remote, local);
199 		if (!sk) {
200 			/* Found an available ephemeral port */
201 			return;
202 		}
203 
204 		/* Release refcnt got in vsock_find_connected_socket */
205 		sock_put(sk);
206 	}
207 }
208 
209 static void hvs_set_channel_pending_send_size(struct vmbus_channel *chan)
210 {
211 	set_channel_pending_send_size(chan,
212 				      HVS_PKT_LEN(HVS_SEND_BUF_SIZE));
213 
214 	virt_mb();
215 }
216 
217 static bool hvs_channel_readable(struct vmbus_channel *chan)
218 {
219 	u32 readable = hv_get_bytes_to_read(&chan->inbound);
220 
221 	/* 0-size payload means FIN */
222 	return readable >= HVS_PKT_LEN(0);
223 }
224 
225 static int hvs_channel_readable_payload(struct vmbus_channel *chan)
226 {
227 	u32 readable = hv_get_bytes_to_read(&chan->inbound);
228 
229 	if (readable > HVS_PKT_LEN(0)) {
230 		/* At least we have 1 byte to read. We don't need to return
231 		 * the exact readable bytes: see vsock_stream_recvmsg() ->
232 		 * vsock_stream_has_data().
233 		 */
234 		return 1;
235 	}
236 
237 	if (readable == HVS_PKT_LEN(0)) {
238 		/* 0-size payload means FIN */
239 		return 0;
240 	}
241 
242 	/* No payload or FIN */
243 	return -1;
244 }
245 
246 static size_t hvs_channel_writable_bytes(struct vmbus_channel *chan)
247 {
248 	u32 writeable = hv_get_bytes_to_write(&chan->outbound);
249 	size_t ret;
250 
251 	/* The ringbuffer mustn't be 100% full, and we should reserve a
252 	 * zero-length-payload packet for the FIN: see hv_ringbuffer_write()
253 	 * and hvs_shutdown().
254 	 */
255 	if (writeable <= HVS_PKT_LEN(1) + HVS_PKT_LEN(0))
256 		return 0;
257 
258 	ret = writeable - HVS_PKT_LEN(1) - HVS_PKT_LEN(0);
259 
260 	return round_down(ret, 8);
261 }
262 
263 static int hvs_send_data(struct vmbus_channel *chan,
264 			 struct hvs_send_buf *send_buf, size_t to_write)
265 {
266 	send_buf->hdr.pkt_type = 1;
267 	send_buf->hdr.data_size = to_write;
268 	return vmbus_sendpacket(chan, &send_buf->hdr,
269 				sizeof(send_buf->hdr) + to_write,
270 				0, VM_PKT_DATA_INBAND, 0);
271 }
272 
273 static void hvs_channel_cb(void *ctx)
274 {
275 	struct sock *sk = (struct sock *)ctx;
276 	struct vsock_sock *vsk = vsock_sk(sk);
277 	struct hvsock *hvs = vsk->trans;
278 	struct vmbus_channel *chan = hvs->chan;
279 
280 	if (hvs_channel_readable(chan))
281 		sk->sk_data_ready(sk);
282 
283 	if (hv_get_bytes_to_write(&chan->outbound) > 0)
284 		sk->sk_write_space(sk);
285 }
286 
287 static void hvs_do_close_lock_held(struct vsock_sock *vsk,
288 				   bool cancel_timeout)
289 {
290 	struct sock *sk = sk_vsock(vsk);
291 
292 	sock_set_flag(sk, SOCK_DONE);
293 	vsk->peer_shutdown = SHUTDOWN_MASK;
294 	if (vsock_stream_has_data(vsk) <= 0)
295 		sk->sk_state = TCP_CLOSING;
296 	sk->sk_state_change(sk);
297 	if (vsk->close_work_scheduled &&
298 	    (!cancel_timeout || cancel_delayed_work(&vsk->close_work))) {
299 		vsk->close_work_scheduled = false;
300 		vsock_remove_sock(vsk);
301 
302 		/* Release the reference taken while scheduling the timeout */
303 		sock_put(sk);
304 	}
305 }
306 
307 static void hvs_close_connection(struct vmbus_channel *chan)
308 {
309 	struct sock *sk = get_per_channel_state(chan);
310 
311 	lock_sock(sk);
312 	hvs_do_close_lock_held(vsock_sk(sk), true);
313 	release_sock(sk);
314 }
315 
316 static void hvs_open_connection(struct vmbus_channel *chan)
317 {
318 	uuid_le *if_instance, *if_type;
319 	unsigned char conn_from_host;
320 
321 	struct sockaddr_vm addr;
322 	struct sock *sk, *new = NULL;
323 	struct vsock_sock *vnew = NULL;
324 	struct hvsock *hvs, *hvs_new = NULL;
325 	int ret;
326 
327 	if_type = &chan->offermsg.offer.if_type;
328 	if_instance = &chan->offermsg.offer.if_instance;
329 	conn_from_host = chan->offermsg.offer.u.pipe.user_def[0];
330 
331 	/* The host or the VM should only listen on a port in
332 	 * [0, MAX_LISTEN_PORT]
333 	 */
334 	if (!is_valid_srv_id(if_type) ||
335 	    get_port_by_srv_id(if_type) > MAX_LISTEN_PORT)
336 		return;
337 
338 	hvs_addr_init(&addr, conn_from_host ? if_type : if_instance);
339 	sk = vsock_find_bound_socket(&addr);
340 	if (!sk)
341 		return;
342 
343 	lock_sock(sk);
344 	if ((conn_from_host && sk->sk_state != TCP_LISTEN) ||
345 	    (!conn_from_host && sk->sk_state != TCP_SYN_SENT))
346 		goto out;
347 
348 	if (conn_from_host) {
349 		if (sk->sk_ack_backlog >= sk->sk_max_ack_backlog)
350 			goto out;
351 
352 		new = __vsock_create(sock_net(sk), NULL, sk, GFP_KERNEL,
353 				     sk->sk_type, 0);
354 		if (!new)
355 			goto out;
356 
357 		new->sk_state = TCP_SYN_SENT;
358 		vnew = vsock_sk(new);
359 		hvs_new = vnew->trans;
360 		hvs_new->chan = chan;
361 	} else {
362 		hvs = vsock_sk(sk)->trans;
363 		hvs->chan = chan;
364 	}
365 
366 	set_channel_read_mode(chan, HV_CALL_DIRECT);
367 	ret = vmbus_open(chan, RINGBUFFER_HVS_SND_SIZE,
368 			 RINGBUFFER_HVS_RCV_SIZE, NULL, 0,
369 			 hvs_channel_cb, conn_from_host ? new : sk);
370 	if (ret != 0) {
371 		if (conn_from_host) {
372 			hvs_new->chan = NULL;
373 			sock_put(new);
374 		} else {
375 			hvs->chan = NULL;
376 		}
377 		goto out;
378 	}
379 
380 	set_per_channel_state(chan, conn_from_host ? new : sk);
381 	vmbus_set_chn_rescind_callback(chan, hvs_close_connection);
382 
383 	/* Set the pending send size to max packet size to always get
384 	 * notifications from the host when there is enough writable space.
385 	 * The host is optimized to send notifications only when the pending
386 	 * size boundary is crossed, and not always.
387 	 */
388 	hvs_set_channel_pending_send_size(chan);
389 
390 	if (conn_from_host) {
391 		new->sk_state = TCP_ESTABLISHED;
392 		sk->sk_ack_backlog++;
393 
394 		hvs_addr_init(&vnew->local_addr, if_type);
395 		hvs_remote_addr_init(&vnew->remote_addr, &vnew->local_addr);
396 
397 		hvs_new->vm_srv_id = *if_type;
398 		hvs_new->host_srv_id = *if_instance;
399 
400 		vsock_insert_connected(vnew);
401 
402 		vsock_enqueue_accept(sk, new);
403 	} else {
404 		sk->sk_state = TCP_ESTABLISHED;
405 		sk->sk_socket->state = SS_CONNECTED;
406 
407 		vsock_insert_connected(vsock_sk(sk));
408 	}
409 
410 	sk->sk_state_change(sk);
411 
412 out:
413 	/* Release refcnt obtained when we called vsock_find_bound_socket() */
414 	sock_put(sk);
415 
416 	release_sock(sk);
417 }
418 
419 static u32 hvs_get_local_cid(void)
420 {
421 	return VMADDR_CID_ANY;
422 }
423 
424 static int hvs_sock_init(struct vsock_sock *vsk, struct vsock_sock *psk)
425 {
426 	struct hvsock *hvs;
427 
428 	hvs = kzalloc(sizeof(*hvs), GFP_KERNEL);
429 	if (!hvs)
430 		return -ENOMEM;
431 
432 	vsk->trans = hvs;
433 	hvs->vsk = vsk;
434 
435 	return 0;
436 }
437 
438 static int hvs_connect(struct vsock_sock *vsk)
439 {
440 	union hvs_service_id vm, host;
441 	struct hvsock *h = vsk->trans;
442 
443 	vm.srv_id = srv_id_template;
444 	vm.svm_port = vsk->local_addr.svm_port;
445 	h->vm_srv_id = vm.srv_id;
446 
447 	host.srv_id = srv_id_template;
448 	host.svm_port = vsk->remote_addr.svm_port;
449 	h->host_srv_id = host.srv_id;
450 
451 	return vmbus_send_tl_connect_request(&h->vm_srv_id, &h->host_srv_id);
452 }
453 
454 static void hvs_shutdown_lock_held(struct hvsock *hvs, int mode)
455 {
456 	struct vmpipe_proto_header hdr;
457 
458 	if (hvs->fin_sent || !hvs->chan)
459 		return;
460 
461 	/* It can't fail: see hvs_channel_writable_bytes(). */
462 	(void)hvs_send_data(hvs->chan, (struct hvs_send_buf *)&hdr, 0);
463 	hvs->fin_sent = true;
464 }
465 
466 static int hvs_shutdown(struct vsock_sock *vsk, int mode)
467 {
468 	struct sock *sk = sk_vsock(vsk);
469 
470 	if (!(mode & SEND_SHUTDOWN))
471 		return 0;
472 
473 	lock_sock(sk);
474 	hvs_shutdown_lock_held(vsk->trans, mode);
475 	release_sock(sk);
476 	return 0;
477 }
478 
479 static void hvs_close_timeout(struct work_struct *work)
480 {
481 	struct vsock_sock *vsk =
482 		container_of(work, struct vsock_sock, close_work.work);
483 	struct sock *sk = sk_vsock(vsk);
484 
485 	sock_hold(sk);
486 	lock_sock(sk);
487 	if (!sock_flag(sk, SOCK_DONE))
488 		hvs_do_close_lock_held(vsk, false);
489 
490 	vsk->close_work_scheduled = false;
491 	release_sock(sk);
492 	sock_put(sk);
493 }
494 
495 /* Returns true, if it is safe to remove socket; false otherwise */
496 static bool hvs_close_lock_held(struct vsock_sock *vsk)
497 {
498 	struct sock *sk = sk_vsock(vsk);
499 
500 	if (!(sk->sk_state == TCP_ESTABLISHED ||
501 	      sk->sk_state == TCP_CLOSING))
502 		return true;
503 
504 	if ((sk->sk_shutdown & SHUTDOWN_MASK) != SHUTDOWN_MASK)
505 		hvs_shutdown_lock_held(vsk->trans, SHUTDOWN_MASK);
506 
507 	if (sock_flag(sk, SOCK_DONE))
508 		return true;
509 
510 	/* This reference will be dropped by the delayed close routine */
511 	sock_hold(sk);
512 	INIT_DELAYED_WORK(&vsk->close_work, hvs_close_timeout);
513 	vsk->close_work_scheduled = true;
514 	schedule_delayed_work(&vsk->close_work, HVS_CLOSE_TIMEOUT);
515 	return false;
516 }
517 
518 static void hvs_release(struct vsock_sock *vsk)
519 {
520 	struct sock *sk = sk_vsock(vsk);
521 	bool remove_sock;
522 
523 	lock_sock(sk);
524 	remove_sock = hvs_close_lock_held(vsk);
525 	release_sock(sk);
526 	if (remove_sock)
527 		vsock_remove_sock(vsk);
528 }
529 
530 static void hvs_destruct(struct vsock_sock *vsk)
531 {
532 	struct hvsock *hvs = vsk->trans;
533 	struct vmbus_channel *chan = hvs->chan;
534 
535 	if (chan)
536 		vmbus_hvsock_device_unregister(chan);
537 
538 	kfree(hvs);
539 }
540 
541 static int hvs_dgram_bind(struct vsock_sock *vsk, struct sockaddr_vm *addr)
542 {
543 	return -EOPNOTSUPP;
544 }
545 
546 static int hvs_dgram_dequeue(struct vsock_sock *vsk, struct msghdr *msg,
547 			     size_t len, int flags)
548 {
549 	return -EOPNOTSUPP;
550 }
551 
552 static int hvs_dgram_enqueue(struct vsock_sock *vsk,
553 			     struct sockaddr_vm *remote, struct msghdr *msg,
554 			     size_t dgram_len)
555 {
556 	return -EOPNOTSUPP;
557 }
558 
559 static bool hvs_dgram_allow(u32 cid, u32 port)
560 {
561 	return false;
562 }
563 
564 static int hvs_update_recv_data(struct hvsock *hvs)
565 {
566 	struct hvs_recv_buf *recv_buf;
567 	u32 payload_len;
568 
569 	recv_buf = (struct hvs_recv_buf *)(hvs->recv_desc + 1);
570 	payload_len = recv_buf->hdr.data_size;
571 
572 	if (payload_len > HVS_MTU_SIZE)
573 		return -EIO;
574 
575 	if (payload_len == 0)
576 		hvs->vsk->peer_shutdown |= SEND_SHUTDOWN;
577 
578 	hvs->recv_data_len = payload_len;
579 	hvs->recv_data_off = 0;
580 
581 	return 0;
582 }
583 
584 static ssize_t hvs_stream_dequeue(struct vsock_sock *vsk, struct msghdr *msg,
585 				  size_t len, int flags)
586 {
587 	struct hvsock *hvs = vsk->trans;
588 	bool need_refill = !hvs->recv_desc;
589 	struct hvs_recv_buf *recv_buf;
590 	u32 to_read;
591 	int ret;
592 
593 	if (flags & MSG_PEEK)
594 		return -EOPNOTSUPP;
595 
596 	if (need_refill) {
597 		hvs->recv_desc = hv_pkt_iter_first(hvs->chan);
598 		ret = hvs_update_recv_data(hvs);
599 		if (ret)
600 			return ret;
601 	}
602 
603 	recv_buf = (struct hvs_recv_buf *)(hvs->recv_desc + 1);
604 	to_read = min_t(u32, len, hvs->recv_data_len);
605 	ret = memcpy_to_msg(msg, recv_buf->data + hvs->recv_data_off, to_read);
606 	if (ret != 0)
607 		return ret;
608 
609 	hvs->recv_data_len -= to_read;
610 	if (hvs->recv_data_len == 0) {
611 		hvs->recv_desc = hv_pkt_iter_next(hvs->chan, hvs->recv_desc);
612 		if (hvs->recv_desc) {
613 			ret = hvs_update_recv_data(hvs);
614 			if (ret)
615 				return ret;
616 		}
617 	} else {
618 		hvs->recv_data_off += to_read;
619 	}
620 
621 	return to_read;
622 }
623 
624 static ssize_t hvs_stream_enqueue(struct vsock_sock *vsk, struct msghdr *msg,
625 				  size_t len)
626 {
627 	struct hvsock *hvs = vsk->trans;
628 	struct vmbus_channel *chan = hvs->chan;
629 	struct hvs_send_buf *send_buf;
630 	ssize_t to_write, max_writable, ret;
631 
632 	BUILD_BUG_ON(sizeof(*send_buf) != PAGE_SIZE_4K);
633 
634 	send_buf = kmalloc(sizeof(*send_buf), GFP_KERNEL);
635 	if (!send_buf)
636 		return -ENOMEM;
637 
638 	max_writable = hvs_channel_writable_bytes(chan);
639 	to_write = min_t(ssize_t, len, max_writable);
640 	to_write = min_t(ssize_t, to_write, HVS_SEND_BUF_SIZE);
641 
642 	ret = memcpy_from_msg(send_buf->data, msg, to_write);
643 	if (ret < 0)
644 		goto out;
645 
646 	ret = hvs_send_data(hvs->chan, send_buf, to_write);
647 	if (ret < 0)
648 		goto out;
649 
650 	ret = to_write;
651 out:
652 	kfree(send_buf);
653 	return ret;
654 }
655 
656 static s64 hvs_stream_has_data(struct vsock_sock *vsk)
657 {
658 	struct hvsock *hvs = vsk->trans;
659 	s64 ret;
660 
661 	if (hvs->recv_data_len > 0)
662 		return 1;
663 
664 	switch (hvs_channel_readable_payload(hvs->chan)) {
665 	case 1:
666 		ret = 1;
667 		break;
668 	case 0:
669 		vsk->peer_shutdown |= SEND_SHUTDOWN;
670 		ret = 0;
671 		break;
672 	default: /* -1 */
673 		ret = 0;
674 		break;
675 	}
676 
677 	return ret;
678 }
679 
680 static s64 hvs_stream_has_space(struct vsock_sock *vsk)
681 {
682 	struct hvsock *hvs = vsk->trans;
683 
684 	return hvs_channel_writable_bytes(hvs->chan);
685 }
686 
687 static u64 hvs_stream_rcvhiwat(struct vsock_sock *vsk)
688 {
689 	return HVS_MTU_SIZE + 1;
690 }
691 
692 static bool hvs_stream_is_active(struct vsock_sock *vsk)
693 {
694 	struct hvsock *hvs = vsk->trans;
695 
696 	return hvs->chan != NULL;
697 }
698 
699 static bool hvs_stream_allow(u32 cid, u32 port)
700 {
701 	/* The host's port range [MIN_HOST_EPHEMERAL_PORT, 0xFFFFFFFF) is
702 	 * reserved as ephemeral ports, which are used as the host's ports
703 	 * when the host initiates connections.
704 	 *
705 	 * Perform this check in the guest so an immediate error is produced
706 	 * instead of a timeout.
707 	 */
708 	if (port > MAX_HOST_LISTEN_PORT)
709 		return false;
710 
711 	if (cid == VMADDR_CID_HOST)
712 		return true;
713 
714 	return false;
715 }
716 
717 static
718 int hvs_notify_poll_in(struct vsock_sock *vsk, size_t target, bool *readable)
719 {
720 	struct hvsock *hvs = vsk->trans;
721 
722 	*readable = hvs_channel_readable(hvs->chan);
723 	return 0;
724 }
725 
726 static
727 int hvs_notify_poll_out(struct vsock_sock *vsk, size_t target, bool *writable)
728 {
729 	*writable = hvs_stream_has_space(vsk) > 0;
730 
731 	return 0;
732 }
733 
734 static
735 int hvs_notify_recv_init(struct vsock_sock *vsk, size_t target,
736 			 struct vsock_transport_recv_notify_data *d)
737 {
738 	return 0;
739 }
740 
741 static
742 int hvs_notify_recv_pre_block(struct vsock_sock *vsk, size_t target,
743 			      struct vsock_transport_recv_notify_data *d)
744 {
745 	return 0;
746 }
747 
748 static
749 int hvs_notify_recv_pre_dequeue(struct vsock_sock *vsk, size_t target,
750 				struct vsock_transport_recv_notify_data *d)
751 {
752 	return 0;
753 }
754 
755 static
756 int hvs_notify_recv_post_dequeue(struct vsock_sock *vsk, size_t target,
757 				 ssize_t copied, bool data_read,
758 				 struct vsock_transport_recv_notify_data *d)
759 {
760 	return 0;
761 }
762 
763 static
764 int hvs_notify_send_init(struct vsock_sock *vsk,
765 			 struct vsock_transport_send_notify_data *d)
766 {
767 	return 0;
768 }
769 
770 static
771 int hvs_notify_send_pre_block(struct vsock_sock *vsk,
772 			      struct vsock_transport_send_notify_data *d)
773 {
774 	return 0;
775 }
776 
777 static
778 int hvs_notify_send_pre_enqueue(struct vsock_sock *vsk,
779 				struct vsock_transport_send_notify_data *d)
780 {
781 	return 0;
782 }
783 
784 static
785 int hvs_notify_send_post_enqueue(struct vsock_sock *vsk, ssize_t written,
786 				 struct vsock_transport_send_notify_data *d)
787 {
788 	return 0;
789 }
790 
791 static void hvs_set_buffer_size(struct vsock_sock *vsk, u64 val)
792 {
793 	/* Ignored. */
794 }
795 
796 static void hvs_set_min_buffer_size(struct vsock_sock *vsk, u64 val)
797 {
798 	/* Ignored. */
799 }
800 
801 static void hvs_set_max_buffer_size(struct vsock_sock *vsk, u64 val)
802 {
803 	/* Ignored. */
804 }
805 
806 static u64 hvs_get_buffer_size(struct vsock_sock *vsk)
807 {
808 	return -ENOPROTOOPT;
809 }
810 
811 static u64 hvs_get_min_buffer_size(struct vsock_sock *vsk)
812 {
813 	return -ENOPROTOOPT;
814 }
815 
816 static u64 hvs_get_max_buffer_size(struct vsock_sock *vsk)
817 {
818 	return -ENOPROTOOPT;
819 }
820 
821 static struct vsock_transport hvs_transport = {
822 	.get_local_cid            = hvs_get_local_cid,
823 
824 	.init                     = hvs_sock_init,
825 	.destruct                 = hvs_destruct,
826 	.release                  = hvs_release,
827 	.connect                  = hvs_connect,
828 	.shutdown                 = hvs_shutdown,
829 
830 	.dgram_bind               = hvs_dgram_bind,
831 	.dgram_dequeue            = hvs_dgram_dequeue,
832 	.dgram_enqueue            = hvs_dgram_enqueue,
833 	.dgram_allow              = hvs_dgram_allow,
834 
835 	.stream_dequeue           = hvs_stream_dequeue,
836 	.stream_enqueue           = hvs_stream_enqueue,
837 	.stream_has_data          = hvs_stream_has_data,
838 	.stream_has_space         = hvs_stream_has_space,
839 	.stream_rcvhiwat          = hvs_stream_rcvhiwat,
840 	.stream_is_active         = hvs_stream_is_active,
841 	.stream_allow             = hvs_stream_allow,
842 
843 	.notify_poll_in           = hvs_notify_poll_in,
844 	.notify_poll_out          = hvs_notify_poll_out,
845 	.notify_recv_init         = hvs_notify_recv_init,
846 	.notify_recv_pre_block    = hvs_notify_recv_pre_block,
847 	.notify_recv_pre_dequeue  = hvs_notify_recv_pre_dequeue,
848 	.notify_recv_post_dequeue = hvs_notify_recv_post_dequeue,
849 	.notify_send_init         = hvs_notify_send_init,
850 	.notify_send_pre_block    = hvs_notify_send_pre_block,
851 	.notify_send_pre_enqueue  = hvs_notify_send_pre_enqueue,
852 	.notify_send_post_enqueue = hvs_notify_send_post_enqueue,
853 
854 	.set_buffer_size          = hvs_set_buffer_size,
855 	.set_min_buffer_size      = hvs_set_min_buffer_size,
856 	.set_max_buffer_size      = hvs_set_max_buffer_size,
857 	.get_buffer_size          = hvs_get_buffer_size,
858 	.get_min_buffer_size      = hvs_get_min_buffer_size,
859 	.get_max_buffer_size      = hvs_get_max_buffer_size,
860 };
861 
862 static int hvs_probe(struct hv_device *hdev,
863 		     const struct hv_vmbus_device_id *dev_id)
864 {
865 	struct vmbus_channel *chan = hdev->channel;
866 
867 	hvs_open_connection(chan);
868 
869 	/* Always return success to suppress the unnecessary error message
870 	 * in vmbus_probe(): on error the host will rescind the device in
871 	 * 30 seconds and we can do cleanup at that time in
872 	 * vmbus_onoffer_rescind().
873 	 */
874 	return 0;
875 }
876 
877 static int hvs_remove(struct hv_device *hdev)
878 {
879 	struct vmbus_channel *chan = hdev->channel;
880 
881 	vmbus_close(chan);
882 
883 	return 0;
884 }
885 
886 /* This isn't really used. See vmbus_match() and vmbus_probe() */
887 static const struct hv_vmbus_device_id id_table[] = {
888 	{},
889 };
890 
891 static struct hv_driver hvs_drv = {
892 	.name		= "hv_sock",
893 	.hvsock		= true,
894 	.id_table	= id_table,
895 	.probe		= hvs_probe,
896 	.remove		= hvs_remove,
897 };
898 
899 static int __init hvs_init(void)
900 {
901 	int ret;
902 
903 	if (vmbus_proto_version < VERSION_WIN10)
904 		return -ENODEV;
905 
906 	ret = vmbus_driver_register(&hvs_drv);
907 	if (ret != 0)
908 		return ret;
909 
910 	ret = vsock_core_init(&hvs_transport);
911 	if (ret) {
912 		vmbus_driver_unregister(&hvs_drv);
913 		return ret;
914 	}
915 
916 	return 0;
917 }
918 
919 static void __exit hvs_exit(void)
920 {
921 	vsock_core_exit();
922 	vmbus_driver_unregister(&hvs_drv);
923 }
924 
925 module_init(hvs_init);
926 module_exit(hvs_exit);
927 
928 MODULE_DESCRIPTION("Hyper-V Sockets");
929 MODULE_VERSION("1.0.0");
930 MODULE_LICENSE("GPL");
931 MODULE_ALIAS_NETPROTO(PF_VSOCK);
932