xref: /freebsd/sys/dev/hyperv/hvsock/hv_sock.c (revision 7fdf597e96a02165cfe22ff357b857d5fa15ed8a)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
3  *
4  * Copyright (c) 2020 Microsoft Corp.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice unmodified, this list of conditions, and the following
12  *    disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
18  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
19  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
21  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
22  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
26  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27  */
28 
29 #include <sys/param.h>
30 #include <sys/bus.h>
31 #include <sys/domain.h>
32 #include <sys/lock.h>
33 #include <sys/kernel.h>
34 #include <sys/types.h>
35 #include <sys/malloc.h>
36 #include <sys/module.h>
37 #include <sys/mutex.h>
38 #include <sys/proc.h>
39 #include <sys/protosw.h>
40 #include <sys/socket.h>
41 #include <sys/sysctl.h>
42 #include <sys/sysproto.h>
43 #include <sys/systm.h>
44 #include <sys/sockbuf.h>
45 #include <sys/sx.h>
46 #include <sys/uio.h>
47 
48 #include <net/vnet.h>
49 
50 #include <dev/hyperv/vmbus/vmbus_reg.h>
51 
52 #include "hv_sock.h"
53 
54 #define HVSOCK_DBG_NONE			0x0
55 #define HVSOCK_DBG_INFO			0x1
56 #define HVSOCK_DBG_ERR			0x2
57 #define HVSOCK_DBG_VERBOSE		0x3
58 
59 
60 SYSCTL_NODE(_net, OID_AUTO, hvsock, CTLFLAG_RD, 0, "HyperV socket");
61 
62 static int hvs_dbg_level;
63 SYSCTL_INT(_net_hvsock, OID_AUTO, hvs_dbg_level, CTLFLAG_RWTUN, &hvs_dbg_level,
64     0, "hyperv socket debug level: 0 = none, 1 = info, 2 = error, 3 = verbose");
65 
66 
67 #define HVSOCK_DBG(level, ...) do {					\
68 	if (hvs_dbg_level >= (level))					\
69 		printf(__VA_ARGS__);					\
70 	} while (0)
71 
72 MALLOC_DEFINE(M_HVSOCK, "hyperv_socket", "hyperv socket control structures");
73 
74 static int hvs_dom_probe(void);
75 
76 /* The MTU is 16KB per host side's design */
77 #define HVSOCK_MTU_SIZE		(1024 * 16)
78 #define HVSOCK_SEND_BUF_SZ	(PAGE_SIZE - sizeof(struct vmpipe_proto_header))
79 
80 #define HVSOCK_HEADER_LEN	(sizeof(struct hvs_pkt_header))
81 
82 #define HVSOCK_PKT_LEN(payload_len)	(HVSOCK_HEADER_LEN + \
83 					 roundup2(payload_len, 8) + \
84 					 sizeof(uint64_t))
85 
86 /*
87  * HyperV Transport sockets
88  */
89 static struct protosw hv_socket_protosw = {
90 	.pr_type =		SOCK_STREAM,
91 	.pr_protocol =		HYPERV_SOCK_PROTO_TRANS,
92 	.pr_flags =		PR_CONNREQUIRED,
93 	.pr_attach =		hvs_trans_attach,
94 	.pr_bind =		hvs_trans_bind,
95 	.pr_listen =		hvs_trans_listen,
96 	.pr_accept =		hvs_trans_accept,
97 	.pr_connect =		hvs_trans_connect,
98 	.pr_peeraddr =		hvs_trans_peeraddr,
99 	.pr_sockaddr =		hvs_trans_sockaddr,
100 	.pr_soreceive =		hvs_trans_soreceive,
101 	.pr_sosend =		hvs_trans_sosend,
102 	.pr_disconnect =	hvs_trans_disconnect,
103 	.pr_close =		hvs_trans_close,
104 	.pr_detach =		hvs_trans_detach,
105 	.pr_shutdown =		hvs_trans_shutdown,
106 	.pr_abort =		hvs_trans_abort,
107 };
108 
109 static struct domain		hv_socket_domain = {
110 	.dom_family =		AF_HYPERV,
111 	.dom_name =		"hyperv",
112 	.dom_probe =		hvs_dom_probe,
113 	.dom_nprotosw =		1,
114 	.dom_protosw =		{ &hv_socket_protosw },
115 };
116 
117 DOMAIN_SET(hv_socket_);
118 
119 #define MAX_PORT			((uint32_t)0xFFFFFFFF)
120 #define MIN_PORT			((uint32_t)0x0)
121 
122 /* 00000000-facb-11e6-bd58-64006a7986d3 */
123 static const struct hyperv_guid srv_id_template = {
124 	.hv_guid = {
125 	    0x00, 0x00, 0x00, 0x00, 0xcb, 0xfa, 0xe6, 0x11,
126 	    0xbd, 0x58, 0x64, 0x00, 0x6a, 0x79, 0x86, 0xd3 }
127 };
128 
129 static int		hvsock_br_callback(void *, int, void *);
130 static uint32_t		hvsock_canread_check(struct hvs_pcb *);
131 static uint32_t		hvsock_canwrite_check(struct hvs_pcb *);
132 static int		hvsock_send_data(struct vmbus_channel *chan,
133     struct uio *uio, uint32_t to_write, struct sockbuf *sb);
134 
135 
136 
137 /* Globals */
138 static struct sx		hvs_trans_socks_sx;
139 static struct mtx		hvs_trans_socks_mtx;
140 static LIST_HEAD(, hvs_pcb)	hvs_trans_bound_socks;
141 static LIST_HEAD(, hvs_pcb)	hvs_trans_connected_socks;
142 static uint32_t			previous_auto_bound_port;
143 
144 static void
145 hvsock_print_guid(struct hyperv_guid *guid)
146 {
147 	unsigned char *p = (unsigned char *)guid;
148 
149 	HVSOCK_DBG(HVSOCK_DBG_INFO,
150 	    "0x%x-0x%x-0x%x-0x%x-0x%x-0x%x-0x%x-0x%x-0x%x-0x%x-0x%x\n",
151 	    *(unsigned int *)p,
152 	    *((unsigned short *) &p[4]),
153 	    *((unsigned short *) &p[6]),
154 	    p[8], p[9], p[10], p[11], p[12], p[13], p[14], p[15]);
155 }
156 
157 static bool
158 is_valid_srv_id(const struct hyperv_guid *id)
159 {
160 	return !memcmp(&id->hv_guid[4],
161 	    &srv_id_template.hv_guid[4], sizeof(struct hyperv_guid) - 4);
162 }
163 
164 static unsigned int
165 get_port_by_srv_id(const struct hyperv_guid *srv_id)
166 {
167 	return *((const unsigned int *)srv_id);
168 }
169 
170 static void
171 set_port_by_srv_id(struct hyperv_guid *srv_id, unsigned int port)
172 {
173 	*((unsigned int *)srv_id) = port;
174 }
175 
176 
177 static void
178 __hvs_remove_pcb_from_list(struct hvs_pcb *pcb, unsigned char list)
179 {
180 	struct hvs_pcb *p = NULL;
181 
182 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE, "%s: pcb is %p\n", __func__, pcb);
183 
184 	if (!pcb)
185 		return;
186 
187 	if (list & HVS_LIST_BOUND) {
188 		LIST_FOREACH(p, &hvs_trans_bound_socks, bound_next)
189 			if  (p == pcb)
190 				LIST_REMOVE(p, bound_next);
191 	}
192 
193 	if (list & HVS_LIST_CONNECTED) {
194 		LIST_FOREACH(p, &hvs_trans_connected_socks, connected_next)
195 			if (p == pcb)
196 				LIST_REMOVE(pcb, connected_next);
197 	}
198 }
199 
200 static void
201 __hvs_remove_socket_from_list(struct socket *so, unsigned char list)
202 {
203 	struct hvs_pcb *pcb = so2hvspcb(so);
204 
205 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE, "%s: pcb is %p\n", __func__, pcb);
206 
207 	__hvs_remove_pcb_from_list(pcb, list);
208 }
209 
210 static void
211 __hvs_insert_socket_on_list(struct socket *so, unsigned char list)
212 {
213 	struct hvs_pcb *pcb = so2hvspcb(so);
214 
215 	if (list & HVS_LIST_BOUND)
216 		LIST_INSERT_HEAD(&hvs_trans_bound_socks,
217 		   pcb, bound_next);
218 
219 	if (list & HVS_LIST_CONNECTED)
220 		LIST_INSERT_HEAD(&hvs_trans_connected_socks,
221 		   pcb, connected_next);
222 }
223 
224 void
225 hvs_remove_socket_from_list(struct socket *so, unsigned char list)
226 {
227 	if (!so || !so->so_pcb) {
228 		HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
229 		    "%s: socket or so_pcb is null\n", __func__);
230 		return;
231 	}
232 
233 	mtx_lock(&hvs_trans_socks_mtx);
234 	__hvs_remove_socket_from_list(so, list);
235 	mtx_unlock(&hvs_trans_socks_mtx);
236 }
237 
238 static void
239 hvs_insert_socket_on_list(struct socket *so, unsigned char list)
240 {
241 	if (!so || !so->so_pcb) {
242 		HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
243 		    "%s: socket or so_pcb is null\n", __func__);
244 		return;
245 	}
246 
247 	mtx_lock(&hvs_trans_socks_mtx);
248 	__hvs_insert_socket_on_list(so, list);
249 	mtx_unlock(&hvs_trans_socks_mtx);
250 }
251 
252 static struct socket *
253 __hvs_find_socket_on_list(struct sockaddr_hvs *addr, unsigned char list)
254 {
255 	struct hvs_pcb *p = NULL;
256 
257 	if (list & HVS_LIST_BOUND)
258 		LIST_FOREACH(p, &hvs_trans_bound_socks, bound_next)
259 			if (p->so != NULL &&
260 			    addr->hvs_port == p->local_addr.hvs_port)
261 				return p->so;
262 
263 	if (list & HVS_LIST_CONNECTED)
264 		LIST_FOREACH(p, &hvs_trans_connected_socks, connected_next)
265 			if (p->so != NULL &&
266 			    addr->hvs_port == p->local_addr.hvs_port)
267 				return p->so;
268 
269 	return NULL;
270 }
271 
272 static struct socket *
273 hvs_find_socket_on_list(struct sockaddr_hvs *addr, unsigned char list)
274 {
275 	struct socket *s = NULL;
276 
277 	mtx_lock(&hvs_trans_socks_mtx);
278 	s = __hvs_find_socket_on_list(addr, list);
279 	mtx_unlock(&hvs_trans_socks_mtx);
280 
281 	return s;
282 }
283 
284 static inline void
285 hvs_addr_set(struct sockaddr_hvs *addr, unsigned int port)
286 {
287 	memset(addr, 0, sizeof(*addr));
288 	addr->sa_family = AF_HYPERV;
289 	addr->sa_len = sizeof(*addr);
290 	addr->hvs_port = port;
291 }
292 
293 void
294 hvs_addr_init(struct sockaddr_hvs *addr, const struct hyperv_guid *svr_id)
295 {
296 	hvs_addr_set(addr, get_port_by_srv_id(svr_id));
297 }
298 
299 int
300 hvs_trans_lock(void)
301 {
302 	sx_xlock(&hvs_trans_socks_sx);
303 	return (0);
304 }
305 
306 void
307 hvs_trans_unlock(void)
308 {
309 	sx_xunlock(&hvs_trans_socks_sx);
310 }
311 
312 static int
313 hvs_dom_probe(void)
314 {
315 
316 	/* Don't even give us a chance to attach on non-HyperV. */
317 	if (vm_guest != VM_GUEST_HV)
318 		return (ENXIO);
319 	return (0);
320 }
321 
322 static void
323 hvs_trans_init(void *arg __unused)
324 {
325 
326 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
327 	    "%s: HyperV Socket hvs_trans_init called\n", __func__);
328 
329 	/* Initialize Globals */
330 	previous_auto_bound_port = MAX_PORT;
331 	sx_init(&hvs_trans_socks_sx, "hvs_trans_sock_sx");
332 	mtx_init(&hvs_trans_socks_mtx,
333 	    "hvs_trans_socks_mtx", NULL, MTX_DEF);
334 	LIST_INIT(&hvs_trans_bound_socks);
335 	LIST_INIT(&hvs_trans_connected_socks);
336 }
337 SYSINIT(hvs_trans_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
338     hvs_trans_init, NULL);
339 
340 /*
341  * Called in two cases:
342  * 1) When user calls socket();
343  * 2) When we accept new incoming conneciton and call sonewconn().
344  */
345 int
346 hvs_trans_attach(struct socket *so, int proto, struct thread *td)
347 {
348 	struct hvs_pcb *pcb = so2hvspcb(so);
349 
350 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
351 	    "%s: HyperV Socket hvs_trans_attach called\n", __func__);
352 
353 	if (so->so_type != SOCK_STREAM)
354 		return (ESOCKTNOSUPPORT);
355 
356 	if (proto != 0 && proto != HYPERV_SOCK_PROTO_TRANS)
357 		return (EPROTONOSUPPORT);
358 
359 	if (pcb != NULL)
360 		return (EISCONN);
361 	pcb = malloc(sizeof(struct hvs_pcb), M_HVSOCK, M_NOWAIT | M_ZERO);
362 	if (pcb == NULL)
363 		return (ENOMEM);
364 
365 	pcb->so = so;
366 	so->so_pcb = (void *)pcb;
367 
368 	return (0);
369 }
370 
371 void
372 hvs_trans_detach(struct socket *so)
373 {
374 	struct hvs_pcb *pcb;
375 
376 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
377 	    "%s: HyperV Socket hvs_trans_detach called\n", __func__);
378 
379 	(void) hvs_trans_lock();
380 	pcb = so2hvspcb(so);
381 	if (pcb == NULL) {
382 		hvs_trans_unlock();
383 		return;
384 	}
385 
386 	if (SOLISTENING(so)) {
387 		bzero(pcb, sizeof(*pcb));
388 		free(pcb, M_HVSOCK);
389 	}
390 
391 	so->so_pcb = NULL;
392 
393 	hvs_trans_unlock();
394 }
395 
396 int
397 hvs_trans_bind(struct socket *so, struct sockaddr *addr, struct thread *td)
398 {
399 	struct hvs_pcb *pcb = so2hvspcb(so);
400 	struct sockaddr_hvs *sa = (struct sockaddr_hvs *) addr;
401 	int error = 0;
402 
403 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
404 	    "%s: HyperV Socket hvs_trans_bind called\n", __func__);
405 
406 	if (sa == NULL) {
407 		return (EINVAL);
408 	}
409 
410 	if (pcb == NULL) {
411 		return (EINVAL);
412 	}
413 
414 	if (sa->sa_family != AF_HYPERV) {
415 		HVSOCK_DBG(HVSOCK_DBG_ERR,
416 		    "%s: Not supported, sa_family is %u\n",
417 		    __func__, sa->sa_family);
418 		return (EAFNOSUPPORT);
419 	}
420 	if (sa->sa_len != sizeof(*sa)) {
421 		HVSOCK_DBG(HVSOCK_DBG_ERR,
422 		    "%s: Not supported, sa_len is %u\n",
423 		    __func__, sa->sa_len);
424 		return (EINVAL);
425 	}
426 
427 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
428 	    "%s: binding port = 0x%x\n", __func__, sa->hvs_port);
429 
430 	mtx_lock(&hvs_trans_socks_mtx);
431 	if (__hvs_find_socket_on_list(sa,
432 	    HVS_LIST_BOUND | HVS_LIST_CONNECTED)) {
433 		error = EADDRINUSE;
434 	} else {
435 		/*
436 		 * The address is available for us to bind.
437 		 * Add socket to the bound list.
438 		 */
439 		hvs_addr_set(&pcb->local_addr, sa->hvs_port);
440 		hvs_addr_set(&pcb->remote_addr, HVADDR_PORT_ANY);
441 		__hvs_insert_socket_on_list(so, HVS_LIST_BOUND);
442 	}
443 	mtx_unlock(&hvs_trans_socks_mtx);
444 
445 	return (error);
446 }
447 
448 int
449 hvs_trans_listen(struct socket *so, int backlog, struct thread *td)
450 {
451 	struct hvs_pcb *pcb = so2hvspcb(so);
452 	struct socket *bound_so;
453 	int error;
454 
455 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
456 	    "%s: HyperV Socket hvs_trans_listen called\n", __func__);
457 
458 	if (pcb == NULL)
459 		return (EINVAL);
460 
461 	/* Check if the address is already bound and it was by us. */
462 	bound_so = hvs_find_socket_on_list(&pcb->local_addr, HVS_LIST_BOUND);
463 	if (bound_so == NULL || bound_so != so) {
464 		HVSOCK_DBG(HVSOCK_DBG_ERR,
465 		    "%s: Address not bound or not by us.\n", __func__);
466 		return (EADDRNOTAVAIL);
467 	}
468 
469 	SOCK_LOCK(so);
470 	error = solisten_proto_check(so);
471 	if (error == 0)
472 		solisten_proto(so, backlog);
473 	SOCK_UNLOCK(so);
474 
475 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
476 	    "%s: HyperV Socket listen error = %d\n", __func__, error);
477 	return (error);
478 }
479 
480 int
481 hvs_trans_accept(struct socket *so, struct sockaddr *sa)
482 {
483 	struct hvs_pcb *pcb = so2hvspcb(so);
484 
485 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
486 	    "%s: HyperV Socket hvs_trans_accept called\n", __func__);
487 
488 	if (pcb == NULL)
489 		return (EINVAL);
490 
491 	memcpy(sa, &pcb->remote_addr, pcb->remote_addr.sa_len);
492 
493 	return (0);
494 }
495 
496 int
497 hvs_trans_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
498 {
499 	struct hvs_pcb *pcb = so2hvspcb(so);
500 	struct sockaddr_hvs *raddr = (struct sockaddr_hvs *)nam;
501 	bool found_auto_bound_port = false;
502 	int i, error = 0;
503 
504 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
505 	    "%s: HyperV Socket hvs_trans_connect called, remote port is %x\n",
506 	    __func__, raddr->hvs_port);
507 
508 	if (pcb == NULL)
509 		return (EINVAL);
510 
511 	/* Verify the remote address */
512 	if (raddr == NULL)
513 		return (EINVAL);
514 	if (raddr->sa_family != AF_HYPERV)
515 		return (EAFNOSUPPORT);
516 	if (raddr->sa_len != sizeof(*raddr))
517 		return (EINVAL);
518 
519 	mtx_lock(&hvs_trans_socks_mtx);
520 	if (so->so_state &
521 	    (SS_ISCONNECTED|SS_ISDISCONNECTING|SS_ISCONNECTING)) {
522 			HVSOCK_DBG(HVSOCK_DBG_ERR,
523 			    "%s: socket connect in progress\n",
524 			    __func__);
525 			error = EINPROGRESS;
526 			goto out;
527 	}
528 
529 	/*
530 	 * Find an available port for us to auto bind the local
531 	 * address.
532 	 */
533 	hvs_addr_set(&pcb->local_addr, 0);
534 
535 	for (i = previous_auto_bound_port - 1;
536 	    i != previous_auto_bound_port; i --) {
537 		if (i == MIN_PORT)
538 			i = MAX_PORT;
539 
540 		pcb->local_addr.hvs_port = i;
541 
542 		if (__hvs_find_socket_on_list(&pcb->local_addr,
543 		    HVS_LIST_BOUND | HVS_LIST_CONNECTED) == NULL) {
544 			found_auto_bound_port = true;
545 			previous_auto_bound_port = i;
546 			HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
547 			    "%s: found local bound port is %x\n",
548 			    __func__, pcb->local_addr.hvs_port);
549 			break;
550 		}
551 	}
552 
553 	if (found_auto_bound_port == true) {
554 		/* Found available port for auto bound, put on list */
555 		__hvs_insert_socket_on_list(so, HVS_LIST_BOUND);
556 		/* Set VM service ID */
557 		pcb->vm_srv_id = srv_id_template;
558 		set_port_by_srv_id(&pcb->vm_srv_id, pcb->local_addr.hvs_port);
559 		/* Set host service ID and remote port */
560 		pcb->host_srv_id = srv_id_template;
561 		set_port_by_srv_id(&pcb->host_srv_id, raddr->hvs_port);
562 		hvs_addr_set(&pcb->remote_addr, raddr->hvs_port);
563 
564 		/* Change the socket state to SS_ISCONNECTING */
565 		soisconnecting(so);
566 	} else {
567 		HVSOCK_DBG(HVSOCK_DBG_ERR,
568 		    "%s: No local port available for auto bound\n",
569 		    __func__);
570 		error = EADDRINUSE;
571 	}
572 
573 	HVSOCK_DBG(HVSOCK_DBG_INFO, "Connect vm_srv_id is ");
574 	hvsock_print_guid(&pcb->vm_srv_id);
575 	HVSOCK_DBG(HVSOCK_DBG_INFO, "Connect host_srv_id is ");
576 	hvsock_print_guid(&pcb->host_srv_id);
577 
578 out:
579 	mtx_unlock(&hvs_trans_socks_mtx);
580 
581 	if (found_auto_bound_port == true)
582 		 vmbus_req_tl_connect(&pcb->vm_srv_id, &pcb->host_srv_id);
583 
584 	return (error);
585 }
586 
587 int
588 hvs_trans_disconnect(struct socket *so)
589 {
590 	struct hvs_pcb *pcb;
591 
592 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
593 	    "%s: HyperV Socket hvs_trans_disconnect called\n", __func__);
594 
595 	(void) hvs_trans_lock();
596 	pcb = so2hvspcb(so);
597 	if (pcb == NULL) {
598 		hvs_trans_unlock();
599 		return (EINVAL);
600 	}
601 
602 	/* If socket is already disconnected, skip this */
603 	if ((so->so_state & SS_ISDISCONNECTED) == 0)
604 		soisdisconnecting(so);
605 
606 	hvs_trans_unlock();
607 
608 	return (0);
609 }
610 
611 struct hvs_callback_arg {
612 	struct uio *uio;
613 	struct sockbuf *sb;
614 };
615 
616 int
617 hvs_trans_soreceive(struct socket *so, struct sockaddr **paddr,
618     struct uio *uio, struct mbuf **mp0, struct mbuf **controlp, int *flagsp)
619 {
620 	struct hvs_pcb *pcb = so2hvspcb(so);
621 	struct sockbuf *sb;
622 	ssize_t orig_resid;
623 	uint32_t canread, to_read;
624 	int flags, error = 0;
625 	struct hvs_callback_arg cbarg;
626 
627 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
628 	    "%s: HyperV Socket hvs_trans_soreceive called\n", __func__);
629 
630 	if (so->so_type != SOCK_STREAM)
631 		return (EINVAL);
632 	if (pcb == NULL)
633 		return (EINVAL);
634 
635 	if (flagsp != NULL)
636 		flags = *flagsp &~ MSG_EOR;
637 	else
638 		flags = 0;
639 
640 	if (flags & MSG_PEEK)
641 		return (EOPNOTSUPP);
642 
643 	/* If no space to copy out anything */
644 	if (uio->uio_resid == 0 || uio->uio_rw != UIO_READ)
645 		return (EINVAL);
646 
647 	orig_resid = uio->uio_resid;
648 
649 	/* Prevent other readers from entering the socket. */
650 	error = SOCK_IO_RECV_LOCK(so, SBLOCKWAIT(flags));
651 	if (error) {
652 		HVSOCK_DBG(HVSOCK_DBG_ERR,
653 		    "%s: soiolock returned error = %d\n", __func__, error);
654 		return (error);
655 	}
656 
657 	sb = &so->so_rcv;
658 	SOCKBUF_LOCK(sb);
659 
660 	cbarg.uio = uio;
661 	cbarg.sb = sb;
662 	/*
663 	 * If the socket is closing, there might still be some data
664 	 * in rx br to read. However we need to make sure
665 	 * the channel is still open.
666 	 */
667 	if ((sb->sb_state & SBS_CANTRCVMORE) &&
668 	    (so->so_state & SS_ISDISCONNECTED)) {
669 		/* Other thread already closed the channel */
670 		error = EPIPE;
671 		goto out;
672 	}
673 
674 	while (true) {
675 		while (uio->uio_resid > 0 &&
676 		    (canread = hvsock_canread_check(pcb)) > 0) {
677 			to_read = MIN(canread, uio->uio_resid);
678 			HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
679 			    "%s: to_read = %u, skip = %u\n", __func__, to_read,
680 			    (unsigned int)(sizeof(struct hvs_pkt_header) +
681 			    pcb->recv_data_off));
682 
683 			error = vmbus_chan_recv_peek_call(pcb->chan, to_read,
684 			    sizeof(struct hvs_pkt_header) + pcb->recv_data_off,
685 			    hvsock_br_callback, (void *)&cbarg);
686 			/*
687 			 * It is possible socket is disconnected becasue
688 			 * we released lock in hvsock_br_callback. So we
689 			 * need to check the state to make sure it is not
690 			 * disconnected.
691 			 */
692 			if (error || so->so_state & SS_ISDISCONNECTED) {
693 				break;
694 			}
695 
696 			pcb->recv_data_len -= to_read;
697 			pcb->recv_data_off += to_read;
698 		}
699 
700 		if (error)
701 			break;
702 
703 		/* Abort if socket has reported problems. */
704 		if (so->so_error) {
705 			if (so->so_error == ESHUTDOWN &&
706 			    orig_resid > uio->uio_resid) {
707 				/*
708 				 * Although we got a FIN, we also received
709 				 * some data in this round. Delivery it
710 				 * to user.
711 				 */
712 				error = 0;
713 			} else {
714 				if (so->so_error != ESHUTDOWN)
715 					error = so->so_error;
716 			}
717 
718 			break;
719 		}
720 
721 		/* Cannot received more. */
722 		if (sb->sb_state & SBS_CANTRCVMORE)
723 			break;
724 
725 		/* We are done if buffer has been filled */
726 		if (uio->uio_resid == 0)
727 			break;
728 
729 		if (!(flags & MSG_WAITALL) && orig_resid > uio->uio_resid)
730 			break;
731 
732 		/* Buffer ring is empty and we shall not block */
733 		if ((so->so_state & SS_NBIO) ||
734 		    (flags & (MSG_DONTWAIT|MSG_NBIO))) {
735 			if (orig_resid == uio->uio_resid) {
736 				/* We have not read anything */
737 				error = EAGAIN;
738 			}
739 			HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
740 			    "%s: non blocked read return, error %d.\n",
741 			    __func__, error);
742 			break;
743 		}
744 
745 		/*
746 		 * Wait and block until (more) data comes in.
747 		 * Note: Drops the sockbuf lock during wait.
748 		 */
749 		error = sbwait(so, SO_RCV);
750 
751 		if (error)
752 			break;
753 
754 		HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
755 		    "%s: wake up from sbwait, read available is %u\n",
756 		    __func__, vmbus_chan_read_available(pcb->chan));
757 	}
758 
759 out:
760 	SOCKBUF_UNLOCK(sb);
761 	SOCK_IO_RECV_UNLOCK(so);
762 
763 	/* We received a FIN in this call */
764 	if (so->so_error == ESHUTDOWN) {
765 		if (so->so_snd.sb_state & SBS_CANTSENDMORE) {
766 			/* Send has already closed */
767 			soisdisconnecting(so);
768 		} else {
769 			/* Just close the receive side */
770 			socantrcvmore(so);
771 		}
772 	}
773 
774 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
775 	    "%s: returning error = %d, so_error = %d\n",
776 	    __func__, error, so->so_error);
777 
778 	return (error);
779 }
780 
781 int
782 hvs_trans_sosend(struct socket *so, struct sockaddr *addr, struct uio *uio,
783     struct mbuf *top, struct mbuf *controlp, int flags, struct thread *td)
784 {
785 	struct hvs_pcb *pcb = so2hvspcb(so);
786 	struct sockbuf *sb;
787 	ssize_t orig_resid;
788 	uint32_t canwrite, to_write;
789 	int error = 0;
790 
791 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
792 	    "%s: HyperV Socket hvs_trans_sosend called, uio_resid = %zd\n",
793 	    __func__, uio->uio_resid);
794 
795 	if (so->so_type != SOCK_STREAM)
796 		return (EINVAL);
797 	if (pcb == NULL)
798 		return (EINVAL);
799 
800 	/* If nothing to send */
801 	if (uio->uio_resid == 0 || uio->uio_rw != UIO_WRITE)
802 		return (EINVAL);
803 
804 	orig_resid = uio->uio_resid;
805 
806 	/* Prevent other writers from entering the socket. */
807 	error = SOCK_IO_SEND_LOCK(so, SBLOCKWAIT(flags));
808 	if (error) {
809 		HVSOCK_DBG(HVSOCK_DBG_ERR,
810 		    "%s: soiolocak returned error = %d\n", __func__, error);
811 		return (error);
812 	}
813 
814 	sb = &so->so_snd;
815 	SOCKBUF_LOCK(sb);
816 
817 	if ((sb->sb_state & SBS_CANTSENDMORE) ||
818 	    so->so_error == ESHUTDOWN) {
819 		error = EPIPE;
820 		goto out;
821 	}
822 
823 	while (uio->uio_resid > 0) {
824 		canwrite = hvsock_canwrite_check(pcb);
825 		if (canwrite == 0) {
826 			/* We have sent some data */
827 			if (orig_resid > uio->uio_resid)
828 				break;
829 			/*
830 			 * We have not sent any data and it is
831 			 * non-blocked io
832 			 */
833 			if (so->so_state & SS_NBIO ||
834 			    (flags & (MSG_NBIO | MSG_DONTWAIT)) != 0) {
835 				error = EWOULDBLOCK;
836 				break;
837 			} else {
838 				/*
839 				 * We are here because there is no space on
840 				 * send buffer ring. Signal the other side
841 				 * to read and free more space.
842 				 * Sleep wait until space avaiable to send
843 				 * Note: Drops the sockbuf lock during wait.
844 				 */
845 				error = sbwait(so, SO_SND);
846 
847 				if (error)
848 					break;
849 
850 				HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
851 				    "%s: wake up from sbwait, space avail on "
852 				    "tx ring is %u\n",
853 				    __func__,
854 				    vmbus_chan_write_available(pcb->chan));
855 
856 				continue;
857 			}
858 		}
859 		to_write = MIN(canwrite, uio->uio_resid);
860 		to_write = MIN(to_write, HVSOCK_SEND_BUF_SZ);
861 
862 		HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
863 		    "%s: canwrite is %u, to_write = %u\n", __func__,
864 		    canwrite, to_write);
865 		error = hvsock_send_data(pcb->chan, uio, to_write, sb);
866 
867 		if (error)
868 			break;
869 	}
870 
871 out:
872 	SOCKBUF_UNLOCK(sb);
873 	SOCK_IO_SEND_UNLOCK(so);
874 
875 	return (error);
876 }
877 
878 int
879 hvs_trans_peeraddr(struct socket *so, struct sockaddr *sa)
880 {
881 	struct hvs_pcb *pcb = so2hvspcb(so);
882 
883 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
884 	    "%s: HyperV Socket hvs_trans_peeraddr called\n", __func__);
885 
886 	if (pcb == NULL)
887 		return (EINVAL);
888 
889 	memcpy(sa, &pcb->remote_addr, pcb->remote_addr.sa_len);
890 
891 	return (0);
892 }
893 
894 int
895 hvs_trans_sockaddr(struct socket *so, struct sockaddr *sa)
896 {
897 	struct hvs_pcb *pcb = so2hvspcb(so);
898 
899 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
900 	    "%s: HyperV Socket hvs_trans_sockaddr called\n", __func__);
901 
902 	if (pcb == NULL)
903 		return (EINVAL);
904 
905 	memcpy(sa, &pcb->local_addr, pcb->local_addr.sa_len);
906 
907 	return (0);
908 }
909 
910 void
911 hvs_trans_close(struct socket *so)
912 {
913 	struct hvs_pcb *pcb;
914 
915 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
916 	    "%s: HyperV Socket hvs_trans_close called\n", __func__);
917 
918 	(void) hvs_trans_lock();
919 	pcb = so2hvspcb(so);
920 	if (!pcb) {
921 		hvs_trans_unlock();
922 		return;
923 	}
924 
925 	if (so->so_state & SS_ISCONNECTED) {
926 		/* Send a FIN to peer */
927 		HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
928 		    "%s: hvs_trans_close sending a FIN to host\n", __func__);
929 		(void) hvsock_send_data(pcb->chan, NULL, 0, NULL);
930 	}
931 
932 	if (so->so_state &
933 	    (SS_ISCONNECTED|SS_ISCONNECTING|SS_ISDISCONNECTING))
934 		soisdisconnected(so);
935 
936 	pcb->chan = NULL;
937 	pcb->so = NULL;
938 
939 	if (SOLISTENING(so)) {
940 		mtx_lock(&hvs_trans_socks_mtx);
941 		/* Remove from bound list */
942 		__hvs_remove_socket_from_list(so, HVS_LIST_BOUND);
943 		mtx_unlock(&hvs_trans_socks_mtx);
944 	}
945 
946 	hvs_trans_unlock();
947 
948 	return;
949 }
950 
951 void
952 hvs_trans_abort(struct socket *so)
953 {
954 	struct hvs_pcb *pcb = so2hvspcb(so);
955 
956 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
957 	    "%s: HyperV Socket hvs_trans_abort called\n", __func__);
958 
959 	(void) hvs_trans_lock();
960 	if (pcb == NULL) {
961 		hvs_trans_unlock();
962 		return;
963 	}
964 
965 	if (SOLISTENING(so)) {
966 		mtx_lock(&hvs_trans_socks_mtx);
967 		/* Remove from bound list */
968 		__hvs_remove_socket_from_list(so, HVS_LIST_BOUND);
969 		mtx_unlock(&hvs_trans_socks_mtx);
970 	}
971 
972 	if (so->so_state & SS_ISCONNECTED) {
973 		(void) sodisconnect(so);
974 	}
975 	hvs_trans_unlock();
976 
977 	return;
978 }
979 
980 int
981 hvs_trans_shutdown(struct socket *so, enum shutdown_how how)
982 {
983 	struct hvs_pcb *pcb = so2hvspcb(so);
984 
985 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
986 	    "%s: HyperV Socket hvs_trans_shutdown called\n", __func__);
987 
988 	SOCK_LOCK(so);
989 	if ((so->so_state &
990 	    (SS_ISCONNECTED | SS_ISCONNECTING | SS_ISDISCONNECTING)) == 0) {
991 		SOCK_UNLOCK(so);
992 		return (ENOTCONN);
993 	}
994 	SOCK_UNLOCK(so);
995 
996 	if (pcb == NULL)
997 		return (EINVAL);
998 
999 	switch (how) {
1000 	case SHUT_RD:
1001 		socantrcvmore(so);
1002 		break;
1003 	case SHUT_RDWR:
1004 		socantrcvmore(so);
1005 		if (so->so_state & SS_ISCONNECTED) {
1006 			/* Send a FIN to peer */
1007 			SOCK_SENDBUF_LOCK(so);
1008 			(void) hvsock_send_data(pcb->chan, NULL, 0,
1009 			    &so->so_snd);
1010 			SOCK_SENDBUF_UNLOCK(so);
1011 			soisdisconnecting(so);
1012 		}
1013 		/* FALLTHROUGH */
1014 	case SHUT_WR:
1015 		socantsendmore(so);
1016 	}
1017 	wakeup(&so->so_timeo);
1018 
1019 	return (0);
1020 }
1021 
1022 /* In the VM, we support Hyper-V Sockets with AF_HYPERV, and the endpoint is
1023  * <port> (see struct sockaddr_hvs).
1024  *
1025  * On the host, Hyper-V Sockets are supported by Winsock AF_HYPERV:
1026  * https://docs.microsoft.com/en-us/virtualization/hyper-v-on-windows/user-
1027  * guide/make-integration-service, and the endpoint is <VmID, ServiceId> with
1028  * the below sockaddr:
1029  *
1030  * struct SOCKADDR_HV
1031  * {
1032  *    ADDRESS_FAMILY Family;
1033  *    USHORT Reserved;
1034  *    GUID VmId;
1035  *    GUID ServiceId;
1036  * };
1037  * Note: VmID is not used by FreeBSD VM and actually it isn't transmitted via
1038  * VMBus, because here it's obvious the host and the VM can easily identify
1039  * each other. Though the VmID is useful on the host, especially in the case
1040  * of Windows container, FreeBSD VM doesn't need it at all.
1041  *
1042  * To be compatible with similar infrastructure in Linux VMs, we have
1043  * to limit the available GUID space of SOCKADDR_HV so that we can create
1044  * a mapping between FreeBSD AF_HYPERV port and SOCKADDR_HV Service GUID.
1045  * The rule of writing Hyper-V Sockets apps on the host and in FreeBSD VM is:
1046  *
1047  ****************************************************************************
1048  * The only valid Service GUIDs, from the perspectives of both the host and *
1049  * FreeBSD VM, that can be connected by the other end, must conform to this *
1050  * format: <port>-facb-11e6-bd58-64006a7986d3.                              *
1051  ****************************************************************************
1052  *
1053  * When we write apps on the host to connect(), the GUID ServiceID is used.
1054  * When we write apps in FreeBSD VM to connect(), we only need to specify the
1055  * port and the driver will form the GUID and use that to request the host.
1056  *
1057  * From the perspective of FreeBSD VM, the remote ephemeral port (i.e. the
1058  * auto-generated remote port for a connect request initiated by the host's
1059  * connect()) is set to HVADDR_PORT_UNKNOWN, which is not realy used on the
1060  * FreeBSD guest.
1061  */
1062 
1063 /*
1064  * Older HyperV hosts (vmbus version 'VMBUS_VERSION_WIN10' or before)
1065  * restricts HyperV socket ring buffer size to six 4K pages. Newer
1066  * HyperV hosts doen't have this limit.
1067  */
1068 #define HVS_RINGBUF_RCV_SIZE	(PAGE_SIZE * 6)
1069 #define HVS_RINGBUF_SND_SIZE	(PAGE_SIZE * 6)
1070 #define HVS_RINGBUF_MAX_SIZE	(PAGE_SIZE * 64)
1071 
1072 struct hvsock_sc {
1073 	device_t		dev;
1074 	struct hvs_pcb		*pcb;
1075 	struct vmbus_channel	*channel;
1076 };
1077 
1078 static bool
1079 hvsock_chan_readable(struct vmbus_channel *chan)
1080 {
1081 	uint32_t readable = vmbus_chan_read_available(chan);
1082 
1083 	return (readable >= HVSOCK_PKT_LEN(0));
1084 }
1085 
1086 static void
1087 hvsock_chan_cb(struct vmbus_channel *chan, void *context)
1088 {
1089 	struct hvs_pcb *pcb = (struct hvs_pcb *) context;
1090 	struct socket *so;
1091 	uint32_t canwrite;
1092 
1093 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
1094 	    "%s: host send us a wakeup on rb data, pcb = %p\n",
1095 	    __func__, pcb);
1096 
1097 	/*
1098 	 * Check if the socket is still attached and valid.
1099 	 * Here we know channel is still open. Need to make
1100 	 * sure the socket has not been closed or freed.
1101 	 */
1102 	(void) hvs_trans_lock();
1103 	so = hsvpcb2so(pcb);
1104 
1105 	if (pcb->chan != NULL && so != NULL) {
1106 		/*
1107 		 * Wake up reader if there are data to read.
1108 		 */
1109 		SOCKBUF_LOCK(&(so)->so_rcv);
1110 
1111 		HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
1112 		    "%s: read available = %u\n", __func__,
1113 		    vmbus_chan_read_available(pcb->chan));
1114 
1115 		if (hvsock_chan_readable(pcb->chan))
1116 			sorwakeup_locked(so);
1117 		else
1118 			SOCKBUF_UNLOCK(&(so)->so_rcv);
1119 
1120 		/*
1121 		 * Wake up sender if space becomes available to write.
1122 		 */
1123 		SOCKBUF_LOCK(&(so)->so_snd);
1124 		canwrite = hvsock_canwrite_check(pcb);
1125 
1126 		HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
1127 		    "%s: canwrite = %u\n", __func__, canwrite);
1128 
1129 		if (canwrite > 0) {
1130 			sowwakeup_locked(so);
1131 		} else {
1132 			SOCKBUF_UNLOCK(&(so)->so_snd);
1133 		}
1134 	}
1135 
1136 	hvs_trans_unlock();
1137 
1138 	return;
1139 }
1140 
1141 static int
1142 hvsock_br_callback(void *datap, int cplen, void *cbarg)
1143 {
1144 	struct hvs_callback_arg *arg = (struct hvs_callback_arg *)cbarg;
1145 	struct uio *uio = arg->uio;
1146 	struct sockbuf *sb = arg->sb;
1147 	int error = 0;
1148 
1149 	if (cbarg == NULL || datap == NULL)
1150 		return (EINVAL);
1151 
1152 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
1153 	    "%s: called, uio_rw = %s, uio_resid = %zd, cplen = %u, "
1154 	    "datap = %p\n",
1155 	    __func__, (uio->uio_rw == UIO_READ) ? "read from br":"write to br",
1156 	    uio->uio_resid, cplen, datap);
1157 
1158 	if (sb)
1159 		SOCKBUF_UNLOCK(sb);
1160 
1161 	error = uiomove(datap, cplen, uio);
1162 
1163 	if (sb)
1164 		SOCKBUF_LOCK(sb);
1165 
1166 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
1167 	    "%s: after uiomove, uio_resid = %zd, error = %d\n",
1168 	    __func__, uio->uio_resid, error);
1169 
1170 	return (error);
1171 }
1172 
1173 static int
1174 hvsock_send_data(struct vmbus_channel *chan, struct uio *uio,
1175     uint32_t to_write, struct sockbuf *sb)
1176 {
1177 	struct hvs_pkt_header hvs_pkt;
1178 	int hvs_pkthlen, hvs_pktlen, pad_pktlen, hlen, error = 0;
1179 	uint64_t pad = 0;
1180 	struct iovec iov[3];
1181 	struct hvs_callback_arg cbarg;
1182 
1183 	if (chan == NULL)
1184 		return (ENOTCONN);
1185 
1186 	hlen = sizeof(struct vmbus_chanpkt_hdr);
1187 	hvs_pkthlen = sizeof(struct hvs_pkt_header);
1188 	hvs_pktlen = hvs_pkthlen + to_write;
1189 	pad_pktlen = VMBUS_CHANPKT_TOTLEN(hvs_pktlen);
1190 
1191 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
1192 	    "%s: hlen = %u, hvs_pkthlen = %u, hvs_pktlen = %u, "
1193 	    "pad_pktlen = %u, data_len = %u\n",
1194 	    __func__, hlen, hvs_pkthlen, hvs_pktlen, pad_pktlen, to_write);
1195 
1196 	hvs_pkt.chan_pkt_hdr.cph_type = VMBUS_CHANPKT_TYPE_INBAND;
1197 	hvs_pkt.chan_pkt_hdr.cph_flags = 0;
1198 	VMBUS_CHANPKT_SETLEN(hvs_pkt.chan_pkt_hdr.cph_hlen, hlen);
1199 	VMBUS_CHANPKT_SETLEN(hvs_pkt.chan_pkt_hdr.cph_tlen, pad_pktlen);
1200 	hvs_pkt.chan_pkt_hdr.cph_xactid = 0;
1201 
1202 	hvs_pkt.vmpipe_pkt_hdr.vmpipe_pkt_type = 1;
1203 	hvs_pkt.vmpipe_pkt_hdr.vmpipe_data_size = to_write;
1204 
1205 	cbarg.uio = uio;
1206 	cbarg.sb = sb;
1207 
1208 	if (uio && to_write > 0) {
1209 		iov[0].iov_base = &hvs_pkt;
1210 		iov[0].iov_len = hvs_pkthlen;
1211 		iov[1].iov_base = NULL;
1212 		iov[1].iov_len = to_write;
1213 		iov[2].iov_base = &pad;
1214 		iov[2].iov_len = pad_pktlen - hvs_pktlen;
1215 
1216 		error = vmbus_chan_iov_send(chan, iov, 3,
1217 		    hvsock_br_callback, &cbarg);
1218 	} else {
1219 		if (to_write == 0) {
1220 			iov[0].iov_base = &hvs_pkt;
1221 			iov[0].iov_len = hvs_pkthlen;
1222 			iov[1].iov_base = &pad;
1223 			iov[1].iov_len = pad_pktlen - hvs_pktlen;
1224 			error = vmbus_chan_iov_send(chan, iov, 2, NULL, NULL);
1225 		}
1226 	}
1227 
1228 	if (error) {
1229 		HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
1230 		    "%s: error = %d\n", __func__, error);
1231 	}
1232 
1233 	return (error);
1234 }
1235 
1236 /*
1237  * Check if we have data on current ring buffer to read
1238  * or not. If not, advance the ring buffer read index to
1239  * next packet. Update the recev_data_len and recev_data_off
1240  * to new value.
1241  * Return the number of bytes can read.
1242  */
1243 static uint32_t
1244 hvsock_canread_check(struct hvs_pcb *pcb)
1245 {
1246 	uint32_t advance;
1247 	uint32_t tlen, hlen, dlen;
1248 	uint32_t bytes_canread = 0;
1249 	int error;
1250 
1251 	if (pcb == NULL || pcb->chan == NULL) {
1252 		pcb->so->so_error = EIO;
1253 		return (0);
1254 	}
1255 
1256 	/* Still have data not read yet on current packet */
1257 	if (pcb->recv_data_len > 0)
1258 		return (pcb->recv_data_len);
1259 
1260 	if (pcb->rb_init)
1261 		advance =
1262 		    VMBUS_CHANPKT_GETLEN(pcb->hvs_pkt.chan_pkt_hdr.cph_tlen);
1263 	else
1264 		advance = 0;
1265 
1266 	bytes_canread = vmbus_chan_read_available(pcb->chan);
1267 
1268 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
1269 	    "%s: bytes_canread on br = %u, advance = %u\n",
1270 	    __func__, bytes_canread, advance);
1271 
1272 	if (pcb->rb_init && bytes_canread == (advance + sizeof(uint64_t))) {
1273 		/*
1274 		 * Nothing to read. Need to advance the rindex before
1275 		 * calling sbwait, so host knows to wake us up when data
1276 		 * is available to read on rb.
1277 		 */
1278 		error = vmbus_chan_recv_idxadv(pcb->chan, advance);
1279 		if (error) {
1280 			HVSOCK_DBG(HVSOCK_DBG_ERR,
1281 			    "%s: after calling vmbus_chan_recv_idxadv, "
1282 			    "got error = %d\n",  __func__, error);
1283 			return (0);
1284 		} else {
1285 			pcb->rb_init = false;
1286 			pcb->recv_data_len = 0;
1287 			pcb->recv_data_off = 0;
1288 			bytes_canread = vmbus_chan_read_available(pcb->chan);
1289 
1290 			HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
1291 			    "%s: advanced %u bytes, "
1292 			    " bytes_canread on br now = %u\n",
1293 			    __func__, advance, bytes_canread);
1294 
1295 			if (bytes_canread == 0)
1296 				return (0);
1297 			else
1298 				advance = 0;
1299 		}
1300 	}
1301 
1302 	if (bytes_canread <
1303 	    advance + (sizeof(struct hvs_pkt_header) + sizeof(uint64_t)))
1304 		return (0);
1305 
1306 	error = vmbus_chan_recv_peek(pcb->chan, &pcb->hvs_pkt,
1307 	    sizeof(struct hvs_pkt_header), advance);
1308 
1309 	/* Don't have anything to read */
1310 	if (error) {
1311 		HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
1312 		    "%s: after calling vmbus_chan_recv_peek, got error = %d\n",
1313 		    __func__, error);
1314 		return (0);
1315 	}
1316 
1317 	/*
1318 	 * We just read in a new packet header. Do some sanity checks.
1319 	 */
1320 	tlen = VMBUS_CHANPKT_GETLEN(pcb->hvs_pkt.chan_pkt_hdr.cph_tlen);
1321 	hlen = VMBUS_CHANPKT_GETLEN(pcb->hvs_pkt.chan_pkt_hdr.cph_hlen);
1322 	dlen = pcb->hvs_pkt.vmpipe_pkt_hdr.vmpipe_data_size;
1323 	if (__predict_false(hlen < sizeof(struct vmbus_chanpkt_hdr)) ||
1324 	    __predict_false(hlen > tlen) ||
1325 	    __predict_false(tlen < dlen + sizeof(struct hvs_pkt_header))) {
1326 		HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
1327 		    "invalid tlen(%u), hlen(%u) or dlen(%u)\n",
1328 		    tlen, hlen, dlen);
1329 		pcb->so->so_error = EIO;
1330 		return (0);
1331 	}
1332 	if (pcb->rb_init == false)
1333 		pcb->rb_init = true;
1334 
1335 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
1336 	    "Got new pkt tlen(%u), hlen(%u) or dlen(%u)\n",
1337 	    tlen, hlen, dlen);
1338 
1339 	/* The other side has sent a close FIN */
1340 	if (dlen == 0) {
1341 		HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
1342 		    "%s: Received FIN from other side\n", __func__);
1343 		/* inform the caller by seting so_error to ESHUTDOWN */
1344 		pcb->so->so_error = ESHUTDOWN;
1345 	}
1346 
1347 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
1348 	    "%s: canread on receive ring is %u \n", __func__, dlen);
1349 
1350 	pcb->recv_data_len = dlen;
1351 	pcb->recv_data_off = 0;
1352 
1353 	return (pcb->recv_data_len);
1354 }
1355 
1356 static uint32_t
1357 hvsock_canwrite_check(struct hvs_pcb *pcb)
1358 {
1359 	uint32_t writeable;
1360 	uint32_t ret;
1361 
1362 	if (pcb == NULL || pcb->chan == NULL)
1363 		return (0);
1364 
1365 	writeable = vmbus_chan_write_available(pcb->chan);
1366 
1367 	/*
1368 	 * We must always reserve a 0-length-payload packet for the FIN.
1369 	 */
1370 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
1371 	    "%s: writeable is %u, should be greater than %ju\n",
1372 	    __func__, writeable,
1373 	    (uintmax_t)(HVSOCK_PKT_LEN(1) + HVSOCK_PKT_LEN(0)));
1374 
1375 	if (writeable < HVSOCK_PKT_LEN(1) + HVSOCK_PKT_LEN(0)) {
1376 		/*
1377 		 * The Tx ring seems full.
1378 		 */
1379 		return (0);
1380 	}
1381 
1382 	ret = writeable - HVSOCK_PKT_LEN(0) - HVSOCK_PKT_LEN(0);
1383 
1384 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
1385 	    "%s: available size is %u\n", __func__, rounddown2(ret, 8));
1386 
1387 	return (rounddown2(ret, 8));
1388 }
1389 
1390 static void
1391 hvsock_set_chan_pending_send_size(struct vmbus_channel *chan)
1392 {
1393 	vmbus_chan_set_pending_send_size(chan,
1394 	    HVSOCK_PKT_LEN(HVSOCK_SEND_BUF_SZ));
1395 }
1396 
1397 static int
1398 hvsock_open_channel(struct vmbus_channel *chan, struct socket *so)
1399 {
1400 	unsigned int rcvbuf, sndbuf;
1401 	struct hvs_pcb *pcb = so2hvspcb(so);
1402 	int ret;
1403 
1404 	if (vmbus_current_version < VMBUS_VERSION_WIN10_V5) {
1405 		sndbuf = HVS_RINGBUF_SND_SIZE;
1406 		rcvbuf = HVS_RINGBUF_RCV_SIZE;
1407 	} else {
1408 		sndbuf = MAX(so->so_snd.sb_hiwat, HVS_RINGBUF_SND_SIZE);
1409 		sndbuf = MIN(sndbuf, HVS_RINGBUF_MAX_SIZE);
1410 		sndbuf = rounddown2(sndbuf, PAGE_SIZE);
1411 		rcvbuf = MAX(so->so_rcv.sb_hiwat, HVS_RINGBUF_RCV_SIZE);
1412 		rcvbuf = MIN(rcvbuf, HVS_RINGBUF_MAX_SIZE);
1413 		rcvbuf = rounddown2(rcvbuf, PAGE_SIZE);
1414 	}
1415 
1416 	/*
1417 	 * Can only read whatever user provided size of data
1418 	 * from ring buffer. Turn off batched reading.
1419 	 */
1420 	vmbus_chan_set_readbatch(chan, false);
1421 
1422 	ret = vmbus_chan_open(chan, sndbuf, rcvbuf, NULL, 0,
1423 	    hvsock_chan_cb, pcb);
1424 
1425 	if (ret != 0) {
1426 		HVSOCK_DBG(HVSOCK_DBG_ERR,
1427 		    "%s: failed to open hvsock channel, sndbuf = %u, "
1428 		    "rcvbuf = %u\n", __func__, sndbuf, rcvbuf);
1429 	} else {
1430 		HVSOCK_DBG(HVSOCK_DBG_INFO,
1431 		    "%s: hvsock channel opened, sndbuf = %u, i"
1432 		    "rcvbuf = %u\n", __func__, sndbuf, rcvbuf);
1433 		/*
1434 		 * Se the pending send size so to receive wakeup
1435 		 * signals from host when there is enough space on
1436 		 * rx buffer ring to write.
1437 		 */
1438 		hvsock_set_chan_pending_send_size(chan);
1439 	}
1440 
1441 	return ret;
1442 }
1443 
1444 /*
1445  * Guest is listening passively on the socket. Open channel and
1446  * create a new socket for the conneciton.
1447  */
1448 static void
1449 hvsock_open_conn_passive(struct vmbus_channel *chan, struct socket *so,
1450     struct hvsock_sc *sc)
1451 {
1452 	struct socket *new_so;
1453 	struct hvs_pcb *new_pcb, *pcb;
1454 	int error;
1455 
1456 	/* Do nothing if socket is not listening */
1457 	if (!SOLISTENING(so)) {
1458 		HVSOCK_DBG(HVSOCK_DBG_ERR,
1459 		    "%s: socket is not a listening one\n", __func__);
1460 		return;
1461 	}
1462 
1463 	/*
1464 	 * Create a new socket. This will call pru_attach to complete
1465 	 * the socket initialization and put the new socket onto
1466 	 * listening socket's sol_incomp list, waiting to be promoted
1467 	 * to sol_comp list.
1468 	 * The new socket created has ref count 0. There is no other
1469 	 * thread that changes the state of this new one at the
1470 	 * moment, so we don't need to hold its lock while opening
1471 	 * channel and filling out its pcb information.
1472 	 */
1473 	new_so = sonewconn(so, 0);
1474 	if (!new_so)
1475 		HVSOCK_DBG(HVSOCK_DBG_ERR,
1476 		    "%s: creating new socket failed\n", __func__);
1477 
1478 	/*
1479 	 * Now open the vmbus channel. If it fails, the socket will be
1480 	 * on the listening socket's sol_incomp queue until it is
1481 	 * replaced and aborted.
1482 	 */
1483 	error = hvsock_open_channel(chan, new_so);
1484 	if (error) {
1485 		new_so->so_error = error;
1486 		return;
1487 	}
1488 
1489 	pcb = so->so_pcb;
1490 	new_pcb = new_so->so_pcb;
1491 
1492 	hvs_addr_set(&(new_pcb->local_addr), pcb->local_addr.hvs_port);
1493 	/* Remote port is unknown to guest in this type of conneciton */
1494 	hvs_addr_set(&(new_pcb->remote_addr), HVADDR_PORT_UNKNOWN);
1495 	new_pcb->chan = chan;
1496 	new_pcb->recv_data_len = 0;
1497 	new_pcb->recv_data_off = 0;
1498 	new_pcb->rb_init = false;
1499 
1500 	new_pcb->vm_srv_id = *vmbus_chan_guid_type(chan);
1501 	new_pcb->host_srv_id = *vmbus_chan_guid_inst(chan);
1502 
1503 	hvs_insert_socket_on_list(new_so, HVS_LIST_CONNECTED);
1504 
1505 	sc->pcb = new_pcb;
1506 
1507 	/*
1508 	 * Change the socket state to SS_ISCONNECTED. This will promote
1509 	 * the socket to sol_comp queue and wake up the thread which
1510 	 * is accepting connection.
1511 	 */
1512 	soisconnected(new_so);
1513 }
1514 
1515 
1516 /*
1517  * Guest is actively connecting to host.
1518  */
1519 static void
1520 hvsock_open_conn_active(struct vmbus_channel *chan, struct socket *so)
1521 {
1522 	struct hvs_pcb *pcb;
1523 	int error;
1524 
1525 	error = hvsock_open_channel(chan, so);
1526 	if (error) {
1527 		so->so_error = error;
1528 		return;
1529 	}
1530 
1531 	pcb = so->so_pcb;
1532 	pcb->chan = chan;
1533 	pcb->recv_data_len = 0;
1534 	pcb->recv_data_off = 0;
1535 	pcb->rb_init = false;
1536 
1537 	mtx_lock(&hvs_trans_socks_mtx);
1538 	__hvs_remove_socket_from_list(so, HVS_LIST_BOUND);
1539 	__hvs_insert_socket_on_list(so, HVS_LIST_CONNECTED);
1540 	mtx_unlock(&hvs_trans_socks_mtx);
1541 
1542 	/*
1543 	 * Change the socket state to SS_ISCONNECTED. This will wake up
1544 	 * the thread sleeping in connect call.
1545 	 */
1546 	soisconnected(so);
1547 }
1548 
1549 static void
1550 hvsock_open_connection(struct vmbus_channel *chan, struct hvsock_sc *sc)
1551 {
1552 	struct hyperv_guid *inst_guid, *type_guid;
1553 	bool conn_from_host;
1554 	struct sockaddr_hvs addr;
1555 	struct socket *so;
1556 	struct hvs_pcb *pcb;
1557 
1558 	type_guid = (struct hyperv_guid *) vmbus_chan_guid_type(chan);
1559 	inst_guid = (struct hyperv_guid *) vmbus_chan_guid_inst(chan);
1560 	conn_from_host = vmbus_chan_is_hvs_conn_from_host(chan);
1561 
1562 	HVSOCK_DBG(HVSOCK_DBG_INFO, "type_guid is ");
1563 	hvsock_print_guid(type_guid);
1564 	HVSOCK_DBG(HVSOCK_DBG_INFO, "inst_guid is ");
1565 	hvsock_print_guid(inst_guid);
1566 	HVSOCK_DBG(HVSOCK_DBG_INFO, "connection %s host\n",
1567 	    (conn_from_host == true ) ? "from" : "to");
1568 
1569 	/*
1570 	 * The listening port should be in [0, MAX_LISTEN_PORT]
1571 	 */
1572 	if (!is_valid_srv_id(type_guid))
1573 		return;
1574 
1575 	/*
1576 	 * There should be a bound socket already created no matter
1577 	 * it is a passive or active connection.
1578 	 * For host initiated connection (passive on guest side),
1579 	 * the  type_guid contains the port which guest is bound and
1580 	 * listening.
1581 	 * For the guest initiated connection (active on guest side),
1582 	 * the inst_guid contains the port that guest has auto bound
1583 	 * to.
1584 	 */
1585 	hvs_addr_init(&addr, conn_from_host ? type_guid : inst_guid);
1586 	so = hvs_find_socket_on_list(&addr, HVS_LIST_BOUND);
1587 	if (!so) {
1588 		HVSOCK_DBG(HVSOCK_DBG_ERR,
1589 		    "%s: no bound socket found for port %u\n",
1590 		    __func__, addr.hvs_port);
1591 		return;
1592 	}
1593 
1594 	if (conn_from_host) {
1595 		hvsock_open_conn_passive(chan, so, sc);
1596 	} else {
1597 		(void) hvs_trans_lock();
1598 		pcb = so->so_pcb;
1599 		if (pcb && pcb->so) {
1600 			sc->pcb = so2hvspcb(so);
1601 			hvsock_open_conn_active(chan, so);
1602 		} else {
1603 			HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
1604 			    "%s: channel detached before open\n", __func__);
1605 		}
1606 		hvs_trans_unlock();
1607 	}
1608 
1609 }
1610 
1611 static int
1612 hvsock_probe(device_t dev)
1613 {
1614 	struct vmbus_channel *channel = vmbus_get_channel(dev);
1615 
1616 	if (!channel || !vmbus_chan_is_hvs(channel)) {
1617 		HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
1618 		    "hvsock_probe called but not a hvsock channel id %u\n",
1619 		    vmbus_chan_id(channel));
1620 
1621 		return ENXIO;
1622 	} else {
1623 		HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
1624 		    "hvsock_probe got a hvsock channel id %u\n",
1625 		    vmbus_chan_id(channel));
1626 
1627 		return BUS_PROBE_DEFAULT;
1628 	}
1629 }
1630 
1631 static int
1632 hvsock_attach(device_t dev)
1633 {
1634 	struct vmbus_channel *channel = vmbus_get_channel(dev);
1635 	struct hvsock_sc *sc = (struct hvsock_sc *)device_get_softc(dev);
1636 
1637 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE, "hvsock_attach called.\n");
1638 
1639 	hvsock_open_connection(channel, sc);
1640 
1641 	/*
1642 	 * Always return success. On error the host will rescind the device
1643 	 * in 30 seconds and we can do cleanup at that time in
1644 	 * vmbus_chan_msgproc_chrescind().
1645 	 */
1646 	return (0);
1647 }
1648 
1649 static int
1650 hvsock_detach(device_t dev)
1651 {
1652 	struct hvsock_sc *sc = (struct hvsock_sc *)device_get_softc(dev);
1653 	struct socket *so;
1654 	int retry;
1655 
1656 	if (bootverbose)
1657 		device_printf(dev, "hvsock_detach called.\n");
1658 
1659 	HVSOCK_DBG(HVSOCK_DBG_VERBOSE, "hvsock_detach called.\n");
1660 
1661 	if (sc->pcb != NULL) {
1662 		(void) hvs_trans_lock();
1663 
1664 		so = hsvpcb2so(sc->pcb);
1665 		if (so) {
1666 			/* Close the connection */
1667 			if (so->so_state &
1668 			    (SS_ISCONNECTED|SS_ISCONNECTING|SS_ISDISCONNECTING))
1669 				soisdisconnected(so);
1670 		}
1671 
1672 		mtx_lock(&hvs_trans_socks_mtx);
1673 		__hvs_remove_pcb_from_list(sc->pcb,
1674 		    HVS_LIST_BOUND | HVS_LIST_CONNECTED);
1675 		mtx_unlock(&hvs_trans_socks_mtx);
1676 
1677 		/*
1678 		 * Close channel while no reader and sender are working
1679 		 * on the buffer rings.
1680 		 */
1681 		if (so) {
1682 			retry = 0;
1683 			while (SOCK_IO_RECV_LOCK(so, 0) == EWOULDBLOCK) {
1684 				/*
1685 				 * Someone is reading, rx br is busy
1686 				 */
1687 				soisdisconnected(so);
1688 				DELAY(500);
1689 				HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
1690 				    "waiting for rx reader to exit, "
1691 				    "retry = %d\n", retry++);
1692 			}
1693 			retry = 0;
1694 			while (SOCK_IO_SEND_LOCK(so, 0) == EWOULDBLOCK) {
1695 				/*
1696 				 * Someone is sending, tx br is busy
1697 				 */
1698 				soisdisconnected(so);
1699 				DELAY(500);
1700 				HVSOCK_DBG(HVSOCK_DBG_VERBOSE,
1701 				    "waiting for tx sender to exit, "
1702 				    "retry = %d\n", retry++);
1703 			}
1704 		}
1705 
1706 
1707 		bzero(sc->pcb, sizeof(struct hvs_pcb));
1708 		free(sc->pcb, M_HVSOCK);
1709 		sc->pcb = NULL;
1710 
1711 		if (so) {
1712 			SOCK_IO_RECV_UNLOCK(so);
1713 			SOCK_IO_SEND_UNLOCK(so);
1714 			so->so_pcb = NULL;
1715 		}
1716 
1717 		hvs_trans_unlock();
1718 	}
1719 
1720 	vmbus_chan_close(vmbus_get_channel(dev));
1721 
1722 	return (0);
1723 }
1724 
1725 static device_method_t hvsock_methods[] = {
1726 	/* Device interface */
1727 	DEVMETHOD(device_probe, hvsock_probe),
1728 	DEVMETHOD(device_attach, hvsock_attach),
1729 	DEVMETHOD(device_detach, hvsock_detach),
1730 	DEVMETHOD_END
1731 };
1732 
1733 static driver_t hvsock_driver = {
1734 	"hv_sock",
1735 	hvsock_methods,
1736 	sizeof(struct hvsock_sc)
1737 };
1738 
1739 DRIVER_MODULE(hvsock, vmbus, hvsock_driver, NULL, NULL);
1740 MODULE_VERSION(hvsock, 1);
1741 MODULE_DEPEND(hvsock, vmbus, 1, 1, 1);
1742