1 /*
2  * VMware vSockets Driver
3  *
4  * Copyright (C) 2007-2013 VMware, Inc. All rights reserved.
5  *
6  * This program is free software; you can redistribute it and/or modify it
7  * under the terms of the GNU General Public License as published by the Free
8  * Software Foundation version 2 and no later version.
9  *
10  * This program is distributed in the hope that it will be useful, but WITHOUT
11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13  * more details.
14  */
15 
16 #include <linux/types.h>
17 #include <linux/bitops.h>
18 #include <linux/cred.h>
19 #include <linux/init.h>
20 #include <linux/io.h>
21 #include <linux/kernel.h>
22 #include <linux/kmod.h>
23 #include <linux/list.h>
24 #include <linux/miscdevice.h>
25 #include <linux/module.h>
26 #include <linux/mutex.h>
27 #include <linux/net.h>
28 #include <linux/poll.h>
29 #include <linux/skbuff.h>
30 #include <linux/smp.h>
31 #include <linux/socket.h>
32 #include <linux/stddef.h>
33 #include <linux/unistd.h>
34 #include <linux/wait.h>
35 #include <linux/workqueue.h>
36 #include <net/sock.h>
37 #include <net/af_vsock.h>
38 
39 #include "vmci_transport_notify.h"
40 
41 static int vmci_transport_recv_dgram_cb(void *data, struct vmci_datagram *dg);
42 static int vmci_transport_recv_stream_cb(void *data, struct vmci_datagram *dg);
43 static void vmci_transport_peer_detach_cb(u32 sub_id,
44 					  const struct vmci_event_data *ed,
45 					  void *client_data);
46 static void vmci_transport_recv_pkt_work(struct work_struct *work);
47 static void vmci_transport_cleanup(struct work_struct *work);
48 static int vmci_transport_recv_listen(struct sock *sk,
49 				      struct vmci_transport_packet *pkt);
50 static int vmci_transport_recv_connecting_server(
51 					struct sock *sk,
52 					struct sock *pending,
53 					struct vmci_transport_packet *pkt);
54 static int vmci_transport_recv_connecting_client(
55 					struct sock *sk,
56 					struct vmci_transport_packet *pkt);
57 static int vmci_transport_recv_connecting_client_negotiate(
58 					struct sock *sk,
59 					struct vmci_transport_packet *pkt);
60 static int vmci_transport_recv_connecting_client_invalid(
61 					struct sock *sk,
62 					struct vmci_transport_packet *pkt);
63 static int vmci_transport_recv_connected(struct sock *sk,
64 					 struct vmci_transport_packet *pkt);
65 static bool vmci_transport_old_proto_override(bool *old_pkt_proto);
66 static u16 vmci_transport_new_proto_supported_versions(void);
67 static bool vmci_transport_proto_to_notify_struct(struct sock *sk, u16 *proto,
68 						  bool old_pkt_proto);
69 
70 struct vmci_transport_recv_pkt_info {
71 	struct work_struct work;
72 	struct sock *sk;
73 	struct vmci_transport_packet pkt;
74 };
75 
76 static LIST_HEAD(vmci_transport_cleanup_list);
77 static DEFINE_SPINLOCK(vmci_transport_cleanup_lock);
78 static DECLARE_WORK(vmci_transport_cleanup_work, vmci_transport_cleanup);
79 
80 static struct vmci_handle vmci_transport_stream_handle = { VMCI_INVALID_ID,
81 							   VMCI_INVALID_ID };
82 static u32 vmci_transport_qp_resumed_sub_id = VMCI_INVALID_ID;
83 
84 static int PROTOCOL_OVERRIDE = -1;
85 
86 #define VMCI_TRANSPORT_DEFAULT_QP_SIZE_MIN   128
87 #define VMCI_TRANSPORT_DEFAULT_QP_SIZE       262144
88 #define VMCI_TRANSPORT_DEFAULT_QP_SIZE_MAX   262144
89 
90 /* The default peer timeout indicates how long we will wait for a peer response
91  * to a control message.
92  */
93 #define VSOCK_DEFAULT_CONNECT_TIMEOUT (2 * HZ)
94 
95 /* Helper function to convert from a VMCI error code to a VSock error code. */
96 
97 static s32 vmci_transport_error_to_vsock_error(s32 vmci_error)
98 {
99 	int err;
100 
101 	switch (vmci_error) {
102 	case VMCI_ERROR_NO_MEM:
103 		err = ENOMEM;
104 		break;
105 	case VMCI_ERROR_DUPLICATE_ENTRY:
106 	case VMCI_ERROR_ALREADY_EXISTS:
107 		err = EADDRINUSE;
108 		break;
109 	case VMCI_ERROR_NO_ACCESS:
110 		err = EPERM;
111 		break;
112 	case VMCI_ERROR_NO_RESOURCES:
113 		err = ENOBUFS;
114 		break;
115 	case VMCI_ERROR_INVALID_RESOURCE:
116 		err = EHOSTUNREACH;
117 		break;
118 	case VMCI_ERROR_INVALID_ARGS:
119 	default:
120 		err = EINVAL;
121 	}
122 
123 	return err > 0 ? -err : err;
124 }
125 
126 static u32 vmci_transport_peer_rid(u32 peer_cid)
127 {
128 	if (VMADDR_CID_HYPERVISOR == peer_cid)
129 		return VMCI_TRANSPORT_HYPERVISOR_PACKET_RID;
130 
131 	return VMCI_TRANSPORT_PACKET_RID;
132 }
133 
134 static inline void
135 vmci_transport_packet_init(struct vmci_transport_packet *pkt,
136 			   struct sockaddr_vm *src,
137 			   struct sockaddr_vm *dst,
138 			   u8 type,
139 			   u64 size,
140 			   u64 mode,
141 			   struct vmci_transport_waiting_info *wait,
142 			   u16 proto,
143 			   struct vmci_handle handle)
144 {
145 	/* We register the stream control handler as an any cid handle so we
146 	 * must always send from a source address of VMADDR_CID_ANY
147 	 */
148 	pkt->dg.src = vmci_make_handle(VMADDR_CID_ANY,
149 				       VMCI_TRANSPORT_PACKET_RID);
150 	pkt->dg.dst = vmci_make_handle(dst->svm_cid,
151 				       vmci_transport_peer_rid(dst->svm_cid));
152 	pkt->dg.payload_size = sizeof(*pkt) - sizeof(pkt->dg);
153 	pkt->version = VMCI_TRANSPORT_PACKET_VERSION;
154 	pkt->type = type;
155 	pkt->src_port = src->svm_port;
156 	pkt->dst_port = dst->svm_port;
157 	memset(&pkt->proto, 0, sizeof(pkt->proto));
158 	memset(&pkt->_reserved2, 0, sizeof(pkt->_reserved2));
159 
160 	switch (pkt->type) {
161 	case VMCI_TRANSPORT_PACKET_TYPE_INVALID:
162 		pkt->u.size = 0;
163 		break;
164 
165 	case VMCI_TRANSPORT_PACKET_TYPE_REQUEST:
166 	case VMCI_TRANSPORT_PACKET_TYPE_NEGOTIATE:
167 		pkt->u.size = size;
168 		break;
169 
170 	case VMCI_TRANSPORT_PACKET_TYPE_OFFER:
171 	case VMCI_TRANSPORT_PACKET_TYPE_ATTACH:
172 		pkt->u.handle = handle;
173 		break;
174 
175 	case VMCI_TRANSPORT_PACKET_TYPE_WROTE:
176 	case VMCI_TRANSPORT_PACKET_TYPE_READ:
177 	case VMCI_TRANSPORT_PACKET_TYPE_RST:
178 		pkt->u.size = 0;
179 		break;
180 
181 	case VMCI_TRANSPORT_PACKET_TYPE_SHUTDOWN:
182 		pkt->u.mode = mode;
183 		break;
184 
185 	case VMCI_TRANSPORT_PACKET_TYPE_WAITING_READ:
186 	case VMCI_TRANSPORT_PACKET_TYPE_WAITING_WRITE:
187 		memcpy(&pkt->u.wait, wait, sizeof(pkt->u.wait));
188 		break;
189 
190 	case VMCI_TRANSPORT_PACKET_TYPE_REQUEST2:
191 	case VMCI_TRANSPORT_PACKET_TYPE_NEGOTIATE2:
192 		pkt->u.size = size;
193 		pkt->proto = proto;
194 		break;
195 	}
196 }
197 
198 static inline void
199 vmci_transport_packet_get_addresses(struct vmci_transport_packet *pkt,
200 				    struct sockaddr_vm *local,
201 				    struct sockaddr_vm *remote)
202 {
203 	vsock_addr_init(local, pkt->dg.dst.context, pkt->dst_port);
204 	vsock_addr_init(remote, pkt->dg.src.context, pkt->src_port);
205 }
206 
207 static int
208 __vmci_transport_send_control_pkt(struct vmci_transport_packet *pkt,
209 				  struct sockaddr_vm *src,
210 				  struct sockaddr_vm *dst,
211 				  enum vmci_transport_packet_type type,
212 				  u64 size,
213 				  u64 mode,
214 				  struct vmci_transport_waiting_info *wait,
215 				  u16 proto,
216 				  struct vmci_handle handle,
217 				  bool convert_error)
218 {
219 	int err;
220 
221 	vmci_transport_packet_init(pkt, src, dst, type, size, mode, wait,
222 				   proto, handle);
223 	err = vmci_datagram_send(&pkt->dg);
224 	if (convert_error && (err < 0))
225 		return vmci_transport_error_to_vsock_error(err);
226 
227 	return err;
228 }
229 
230 static int
231 vmci_transport_reply_control_pkt_fast(struct vmci_transport_packet *pkt,
232 				      enum vmci_transport_packet_type type,
233 				      u64 size,
234 				      u64 mode,
235 				      struct vmci_transport_waiting_info *wait,
236 				      struct vmci_handle handle)
237 {
238 	struct vmci_transport_packet reply;
239 	struct sockaddr_vm src, dst;
240 
241 	if (pkt->type == VMCI_TRANSPORT_PACKET_TYPE_RST) {
242 		return 0;
243 	} else {
244 		vmci_transport_packet_get_addresses(pkt, &src, &dst);
245 		return __vmci_transport_send_control_pkt(&reply, &src, &dst,
246 							 type,
247 							 size, mode, wait,
248 							 VSOCK_PROTO_INVALID,
249 							 handle, true);
250 	}
251 }
252 
253 static int
254 vmci_transport_send_control_pkt_bh(struct sockaddr_vm *src,
255 				   struct sockaddr_vm *dst,
256 				   enum vmci_transport_packet_type type,
257 				   u64 size,
258 				   u64 mode,
259 				   struct vmci_transport_waiting_info *wait,
260 				   struct vmci_handle handle)
261 {
262 	/* Note that it is safe to use a single packet across all CPUs since
263 	 * two tasklets of the same type are guaranteed to not ever run
264 	 * simultaneously. If that ever changes, or VMCI stops using tasklets,
265 	 * we can use per-cpu packets.
266 	 */
267 	static struct vmci_transport_packet pkt;
268 
269 	return __vmci_transport_send_control_pkt(&pkt, src, dst, type,
270 						 size, mode, wait,
271 						 VSOCK_PROTO_INVALID, handle,
272 						 false);
273 }
274 
275 static int
276 vmci_transport_send_control_pkt(struct sock *sk,
277 				enum vmci_transport_packet_type type,
278 				u64 size,
279 				u64 mode,
280 				struct vmci_transport_waiting_info *wait,
281 				u16 proto,
282 				struct vmci_handle handle)
283 {
284 	struct vmci_transport_packet *pkt;
285 	struct vsock_sock *vsk;
286 	int err;
287 
288 	vsk = vsock_sk(sk);
289 
290 	if (!vsock_addr_bound(&vsk->local_addr))
291 		return -EINVAL;
292 
293 	if (!vsock_addr_bound(&vsk->remote_addr))
294 		return -EINVAL;
295 
296 	pkt = kmalloc(sizeof(*pkt), GFP_KERNEL);
297 	if (!pkt)
298 		return -ENOMEM;
299 
300 	err = __vmci_transport_send_control_pkt(pkt, &vsk->local_addr,
301 						&vsk->remote_addr, type, size,
302 						mode, wait, proto, handle,
303 						true);
304 	kfree(pkt);
305 
306 	return err;
307 }
308 
309 static int vmci_transport_send_reset_bh(struct sockaddr_vm *dst,
310 					struct sockaddr_vm *src,
311 					struct vmci_transport_packet *pkt)
312 {
313 	if (pkt->type == VMCI_TRANSPORT_PACKET_TYPE_RST)
314 		return 0;
315 	return vmci_transport_send_control_pkt_bh(
316 					dst, src,
317 					VMCI_TRANSPORT_PACKET_TYPE_RST, 0,
318 					0, NULL, VMCI_INVALID_HANDLE);
319 }
320 
321 static int vmci_transport_send_reset(struct sock *sk,
322 				     struct vmci_transport_packet *pkt)
323 {
324 	if (pkt->type == VMCI_TRANSPORT_PACKET_TYPE_RST)
325 		return 0;
326 	return vmci_transport_send_control_pkt(sk,
327 					VMCI_TRANSPORT_PACKET_TYPE_RST,
328 					0, 0, NULL, VSOCK_PROTO_INVALID,
329 					VMCI_INVALID_HANDLE);
330 }
331 
332 static int vmci_transport_send_negotiate(struct sock *sk, size_t size)
333 {
334 	return vmci_transport_send_control_pkt(
335 					sk,
336 					VMCI_TRANSPORT_PACKET_TYPE_NEGOTIATE,
337 					size, 0, NULL,
338 					VSOCK_PROTO_INVALID,
339 					VMCI_INVALID_HANDLE);
340 }
341 
342 static int vmci_transport_send_negotiate2(struct sock *sk, size_t size,
343 					  u16 version)
344 {
345 	return vmci_transport_send_control_pkt(
346 					sk,
347 					VMCI_TRANSPORT_PACKET_TYPE_NEGOTIATE2,
348 					size, 0, NULL, version,
349 					VMCI_INVALID_HANDLE);
350 }
351 
352 static int vmci_transport_send_qp_offer(struct sock *sk,
353 					struct vmci_handle handle)
354 {
355 	return vmci_transport_send_control_pkt(
356 					sk, VMCI_TRANSPORT_PACKET_TYPE_OFFER, 0,
357 					0, NULL,
358 					VSOCK_PROTO_INVALID, handle);
359 }
360 
361 static int vmci_transport_send_attach(struct sock *sk,
362 				      struct vmci_handle handle)
363 {
364 	return vmci_transport_send_control_pkt(
365 					sk, VMCI_TRANSPORT_PACKET_TYPE_ATTACH,
366 					0, 0, NULL, VSOCK_PROTO_INVALID,
367 					handle);
368 }
369 
370 static int vmci_transport_reply_reset(struct vmci_transport_packet *pkt)
371 {
372 	return vmci_transport_reply_control_pkt_fast(
373 						pkt,
374 						VMCI_TRANSPORT_PACKET_TYPE_RST,
375 						0, 0, NULL,
376 						VMCI_INVALID_HANDLE);
377 }
378 
379 static int vmci_transport_send_invalid_bh(struct sockaddr_vm *dst,
380 					  struct sockaddr_vm *src)
381 {
382 	return vmci_transport_send_control_pkt_bh(
383 					dst, src,
384 					VMCI_TRANSPORT_PACKET_TYPE_INVALID,
385 					0, 0, NULL, VMCI_INVALID_HANDLE);
386 }
387 
388 int vmci_transport_send_wrote_bh(struct sockaddr_vm *dst,
389 				 struct sockaddr_vm *src)
390 {
391 	return vmci_transport_send_control_pkt_bh(
392 					dst, src,
393 					VMCI_TRANSPORT_PACKET_TYPE_WROTE, 0,
394 					0, NULL, VMCI_INVALID_HANDLE);
395 }
396 
397 int vmci_transport_send_read_bh(struct sockaddr_vm *dst,
398 				struct sockaddr_vm *src)
399 {
400 	return vmci_transport_send_control_pkt_bh(
401 					dst, src,
402 					VMCI_TRANSPORT_PACKET_TYPE_READ, 0,
403 					0, NULL, VMCI_INVALID_HANDLE);
404 }
405 
406 int vmci_transport_send_wrote(struct sock *sk)
407 {
408 	return vmci_transport_send_control_pkt(
409 					sk, VMCI_TRANSPORT_PACKET_TYPE_WROTE, 0,
410 					0, NULL, VSOCK_PROTO_INVALID,
411 					VMCI_INVALID_HANDLE);
412 }
413 
414 int vmci_transport_send_read(struct sock *sk)
415 {
416 	return vmci_transport_send_control_pkt(
417 					sk, VMCI_TRANSPORT_PACKET_TYPE_READ, 0,
418 					0, NULL, VSOCK_PROTO_INVALID,
419 					VMCI_INVALID_HANDLE);
420 }
421 
422 int vmci_transport_send_waiting_write(struct sock *sk,
423 				      struct vmci_transport_waiting_info *wait)
424 {
425 	return vmci_transport_send_control_pkt(
426 				sk, VMCI_TRANSPORT_PACKET_TYPE_WAITING_WRITE,
427 				0, 0, wait, VSOCK_PROTO_INVALID,
428 				VMCI_INVALID_HANDLE);
429 }
430 
431 int vmci_transport_send_waiting_read(struct sock *sk,
432 				     struct vmci_transport_waiting_info *wait)
433 {
434 	return vmci_transport_send_control_pkt(
435 				sk, VMCI_TRANSPORT_PACKET_TYPE_WAITING_READ,
436 				0, 0, wait, VSOCK_PROTO_INVALID,
437 				VMCI_INVALID_HANDLE);
438 }
439 
440 static int vmci_transport_shutdown(struct vsock_sock *vsk, int mode)
441 {
442 	return vmci_transport_send_control_pkt(
443 					&vsk->sk,
444 					VMCI_TRANSPORT_PACKET_TYPE_SHUTDOWN,
445 					0, mode, NULL,
446 					VSOCK_PROTO_INVALID,
447 					VMCI_INVALID_HANDLE);
448 }
449 
450 static int vmci_transport_send_conn_request(struct sock *sk, size_t size)
451 {
452 	return vmci_transport_send_control_pkt(sk,
453 					VMCI_TRANSPORT_PACKET_TYPE_REQUEST,
454 					size, 0, NULL,
455 					VSOCK_PROTO_INVALID,
456 					VMCI_INVALID_HANDLE);
457 }
458 
459 static int vmci_transport_send_conn_request2(struct sock *sk, size_t size,
460 					     u16 version)
461 {
462 	return vmci_transport_send_control_pkt(
463 					sk, VMCI_TRANSPORT_PACKET_TYPE_REQUEST2,
464 					size, 0, NULL, version,
465 					VMCI_INVALID_HANDLE);
466 }
467 
468 static struct sock *vmci_transport_get_pending(
469 					struct sock *listener,
470 					struct vmci_transport_packet *pkt)
471 {
472 	struct vsock_sock *vlistener;
473 	struct vsock_sock *vpending;
474 	struct sock *pending;
475 	struct sockaddr_vm src;
476 
477 	vsock_addr_init(&src, pkt->dg.src.context, pkt->src_port);
478 
479 	vlistener = vsock_sk(listener);
480 
481 	list_for_each_entry(vpending, &vlistener->pending_links,
482 			    pending_links) {
483 		if (vsock_addr_equals_addr(&src, &vpending->remote_addr) &&
484 		    pkt->dst_port == vpending->local_addr.svm_port) {
485 			pending = sk_vsock(vpending);
486 			sock_hold(pending);
487 			goto found;
488 		}
489 	}
490 
491 	pending = NULL;
492 found:
493 	return pending;
494 
495 }
496 
497 static void vmci_transport_release_pending(struct sock *pending)
498 {
499 	sock_put(pending);
500 }
501 
502 /* We allow two kinds of sockets to communicate with a restricted VM: 1)
503  * trusted sockets 2) sockets from applications running as the same user as the
504  * VM (this is only true for the host side and only when using hosted products)
505  */
506 
507 static bool vmci_transport_is_trusted(struct vsock_sock *vsock, u32 peer_cid)
508 {
509 	return vsock->trusted ||
510 	       vmci_is_context_owner(peer_cid, vsock->owner->uid);
511 }
512 
513 /* We allow sending datagrams to and receiving datagrams from a restricted VM
514  * only if it is trusted as described in vmci_transport_is_trusted.
515  */
516 
517 static bool vmci_transport_allow_dgram(struct vsock_sock *vsock, u32 peer_cid)
518 {
519 	if (VMADDR_CID_HYPERVISOR == peer_cid)
520 		return true;
521 
522 	if (vsock->cached_peer != peer_cid) {
523 		vsock->cached_peer = peer_cid;
524 		if (!vmci_transport_is_trusted(vsock, peer_cid) &&
525 		    (vmci_context_get_priv_flags(peer_cid) &
526 		     VMCI_PRIVILEGE_FLAG_RESTRICTED)) {
527 			vsock->cached_peer_allow_dgram = false;
528 		} else {
529 			vsock->cached_peer_allow_dgram = true;
530 		}
531 	}
532 
533 	return vsock->cached_peer_allow_dgram;
534 }
535 
536 static int
537 vmci_transport_queue_pair_alloc(struct vmci_qp **qpair,
538 				struct vmci_handle *handle,
539 				u64 produce_size,
540 				u64 consume_size,
541 				u32 peer, u32 flags, bool trusted)
542 {
543 	int err = 0;
544 
545 	if (trusted) {
546 		/* Try to allocate our queue pair as trusted. This will only
547 		 * work if vsock is running in the host.
548 		 */
549 
550 		err = vmci_qpair_alloc(qpair, handle, produce_size,
551 				       consume_size,
552 				       peer, flags,
553 				       VMCI_PRIVILEGE_FLAG_TRUSTED);
554 		if (err != VMCI_ERROR_NO_ACCESS)
555 			goto out;
556 
557 	}
558 
559 	err = vmci_qpair_alloc(qpair, handle, produce_size, consume_size,
560 			       peer, flags, VMCI_NO_PRIVILEGE_FLAGS);
561 out:
562 	if (err < 0) {
563 		pr_err("Could not attach to queue pair with %d\n",
564 		       err);
565 		err = vmci_transport_error_to_vsock_error(err);
566 	}
567 
568 	return err;
569 }
570 
571 static int
572 vmci_transport_datagram_create_hnd(u32 resource_id,
573 				   u32 flags,
574 				   vmci_datagram_recv_cb recv_cb,
575 				   void *client_data,
576 				   struct vmci_handle *out_handle)
577 {
578 	int err = 0;
579 
580 	/* Try to allocate our datagram handler as trusted. This will only work
581 	 * if vsock is running in the host.
582 	 */
583 
584 	err = vmci_datagram_create_handle_priv(resource_id, flags,
585 					       VMCI_PRIVILEGE_FLAG_TRUSTED,
586 					       recv_cb,
587 					       client_data, out_handle);
588 
589 	if (err == VMCI_ERROR_NO_ACCESS)
590 		err = vmci_datagram_create_handle(resource_id, flags,
591 						  recv_cb, client_data,
592 						  out_handle);
593 
594 	return err;
595 }
596 
597 /* This is invoked as part of a tasklet that's scheduled when the VMCI
598  * interrupt fires.  This is run in bottom-half context and if it ever needs to
599  * sleep it should defer that work to a work queue.
600  */
601 
602 static int vmci_transport_recv_dgram_cb(void *data, struct vmci_datagram *dg)
603 {
604 	struct sock *sk;
605 	size_t size;
606 	struct sk_buff *skb;
607 	struct vsock_sock *vsk;
608 
609 	sk = (struct sock *)data;
610 
611 	/* This handler is privileged when this module is running on the host.
612 	 * We will get datagrams from all endpoints (even VMs that are in a
613 	 * restricted context). If we get one from a restricted context then
614 	 * the destination socket must be trusted.
615 	 *
616 	 * NOTE: We access the socket struct without holding the lock here.
617 	 * This is ok because the field we are interested is never modified
618 	 * outside of the create and destruct socket functions.
619 	 */
620 	vsk = vsock_sk(sk);
621 	if (!vmci_transport_allow_dgram(vsk, dg->src.context))
622 		return VMCI_ERROR_NO_ACCESS;
623 
624 	size = VMCI_DG_SIZE(dg);
625 
626 	/* Attach the packet to the socket's receive queue as an sk_buff. */
627 	skb = alloc_skb(size, GFP_ATOMIC);
628 	if (!skb)
629 		return VMCI_ERROR_NO_MEM;
630 
631 	/* sk_receive_skb() will do a sock_put(), so hold here. */
632 	sock_hold(sk);
633 	skb_put(skb, size);
634 	memcpy(skb->data, dg, size);
635 	sk_receive_skb(sk, skb, 0);
636 
637 	return VMCI_SUCCESS;
638 }
639 
640 static bool vmci_transport_stream_allow(u32 cid, u32 port)
641 {
642 	static const u32 non_socket_contexts[] = {
643 		VMADDR_CID_RESERVED,
644 	};
645 	int i;
646 
647 	BUILD_BUG_ON(sizeof(cid) != sizeof(*non_socket_contexts));
648 
649 	for (i = 0; i < ARRAY_SIZE(non_socket_contexts); i++) {
650 		if (cid == non_socket_contexts[i])
651 			return false;
652 	}
653 
654 	return true;
655 }
656 
657 /* This is invoked as part of a tasklet that's scheduled when the VMCI
658  * interrupt fires.  This is run in bottom-half context but it defers most of
659  * its work to the packet handling work queue.
660  */
661 
662 static int vmci_transport_recv_stream_cb(void *data, struct vmci_datagram *dg)
663 {
664 	struct sock *sk;
665 	struct sockaddr_vm dst;
666 	struct sockaddr_vm src;
667 	struct vmci_transport_packet *pkt;
668 	struct vsock_sock *vsk;
669 	bool bh_process_pkt;
670 	int err;
671 
672 	sk = NULL;
673 	err = VMCI_SUCCESS;
674 	bh_process_pkt = false;
675 
676 	/* Ignore incoming packets from contexts without sockets, or resources
677 	 * that aren't vsock implementations.
678 	 */
679 
680 	if (!vmci_transport_stream_allow(dg->src.context, -1)
681 	    || vmci_transport_peer_rid(dg->src.context) != dg->src.resource)
682 		return VMCI_ERROR_NO_ACCESS;
683 
684 	if (VMCI_DG_SIZE(dg) < sizeof(*pkt))
685 		/* Drop datagrams that do not contain full VSock packets. */
686 		return VMCI_ERROR_INVALID_ARGS;
687 
688 	pkt = (struct vmci_transport_packet *)dg;
689 
690 	/* Find the socket that should handle this packet.  First we look for a
691 	 * connected socket and if there is none we look for a socket bound to
692 	 * the destintation address.
693 	 */
694 	vsock_addr_init(&src, pkt->dg.src.context, pkt->src_port);
695 	vsock_addr_init(&dst, pkt->dg.dst.context, pkt->dst_port);
696 
697 	sk = vsock_find_connected_socket(&src, &dst);
698 	if (!sk) {
699 		sk = vsock_find_bound_socket(&dst);
700 		if (!sk) {
701 			/* We could not find a socket for this specified
702 			 * address.  If this packet is a RST, we just drop it.
703 			 * If it is another packet, we send a RST.  Note that
704 			 * we do not send a RST reply to RSTs so that we do not
705 			 * continually send RSTs between two endpoints.
706 			 *
707 			 * Note that since this is a reply, dst is src and src
708 			 * is dst.
709 			 */
710 			if (vmci_transport_send_reset_bh(&dst, &src, pkt) < 0)
711 				pr_err("unable to send reset\n");
712 
713 			err = VMCI_ERROR_NOT_FOUND;
714 			goto out;
715 		}
716 	}
717 
718 	/* If the received packet type is beyond all types known to this
719 	 * implementation, reply with an invalid message.  Hopefully this will
720 	 * help when implementing backwards compatibility in the future.
721 	 */
722 	if (pkt->type >= VMCI_TRANSPORT_PACKET_TYPE_MAX) {
723 		vmci_transport_send_invalid_bh(&dst, &src);
724 		err = VMCI_ERROR_INVALID_ARGS;
725 		goto out;
726 	}
727 
728 	/* This handler is privileged when this module is running on the host.
729 	 * We will get datagram connect requests from all endpoints (even VMs
730 	 * that are in a restricted context). If we get one from a restricted
731 	 * context then the destination socket must be trusted.
732 	 *
733 	 * NOTE: We access the socket struct without holding the lock here.
734 	 * This is ok because the field we are interested is never modified
735 	 * outside of the create and destruct socket functions.
736 	 */
737 	vsk = vsock_sk(sk);
738 	if (!vmci_transport_allow_dgram(vsk, pkt->dg.src.context)) {
739 		err = VMCI_ERROR_NO_ACCESS;
740 		goto out;
741 	}
742 
743 	/* We do most everything in a work queue, but let's fast path the
744 	 * notification of reads and writes to help data transfer performance.
745 	 * We can only do this if there is no process context code executing
746 	 * for this socket since that may change the state.
747 	 */
748 	bh_lock_sock(sk);
749 
750 	if (!sock_owned_by_user(sk)) {
751 		/* The local context ID may be out of date, update it. */
752 		vsk->local_addr.svm_cid = dst.svm_cid;
753 
754 		if (sk->sk_state == SS_CONNECTED)
755 			vmci_trans(vsk)->notify_ops->handle_notify_pkt(
756 					sk, pkt, true, &dst, &src,
757 					&bh_process_pkt);
758 	}
759 
760 	bh_unlock_sock(sk);
761 
762 	if (!bh_process_pkt) {
763 		struct vmci_transport_recv_pkt_info *recv_pkt_info;
764 
765 		recv_pkt_info = kmalloc(sizeof(*recv_pkt_info), GFP_ATOMIC);
766 		if (!recv_pkt_info) {
767 			if (vmci_transport_send_reset_bh(&dst, &src, pkt) < 0)
768 				pr_err("unable to send reset\n");
769 
770 			err = VMCI_ERROR_NO_MEM;
771 			goto out;
772 		}
773 
774 		recv_pkt_info->sk = sk;
775 		memcpy(&recv_pkt_info->pkt, pkt, sizeof(recv_pkt_info->pkt));
776 		INIT_WORK(&recv_pkt_info->work, vmci_transport_recv_pkt_work);
777 
778 		schedule_work(&recv_pkt_info->work);
779 		/* Clear sk so that the reference count incremented by one of
780 		 * the Find functions above is not decremented below.  We need
781 		 * that reference count for the packet handler we've scheduled
782 		 * to run.
783 		 */
784 		sk = NULL;
785 	}
786 
787 out:
788 	if (sk)
789 		sock_put(sk);
790 
791 	return err;
792 }
793 
794 static void vmci_transport_handle_detach(struct sock *sk)
795 {
796 	struct vsock_sock *vsk;
797 
798 	vsk = vsock_sk(sk);
799 	if (!vmci_handle_is_invalid(vmci_trans(vsk)->qp_handle)) {
800 		sock_set_flag(sk, SOCK_DONE);
801 
802 		/* On a detach the peer will not be sending or receiving
803 		 * anymore.
804 		 */
805 		vsk->peer_shutdown = SHUTDOWN_MASK;
806 
807 		/* We should not be sending anymore since the peer won't be
808 		 * there to receive, but we can still receive if there is data
809 		 * left in our consume queue.
810 		 */
811 		if (vsock_stream_has_data(vsk) <= 0) {
812 			if (sk->sk_state == SS_CONNECTING) {
813 				/* The peer may detach from a queue pair while
814 				 * we are still in the connecting state, i.e.,
815 				 * if the peer VM is killed after attaching to
816 				 * a queue pair, but before we complete the
817 				 * handshake. In that case, we treat the detach
818 				 * event like a reset.
819 				 */
820 
821 				sk->sk_state = SS_UNCONNECTED;
822 				sk->sk_err = ECONNRESET;
823 				sk->sk_error_report(sk);
824 				return;
825 			}
826 			sk->sk_state = SS_UNCONNECTED;
827 		}
828 		sk->sk_state_change(sk);
829 	}
830 }
831 
832 static void vmci_transport_peer_detach_cb(u32 sub_id,
833 					  const struct vmci_event_data *e_data,
834 					  void *client_data)
835 {
836 	struct vmci_transport *trans = client_data;
837 	const struct vmci_event_payload_qp *e_payload;
838 
839 	e_payload = vmci_event_data_const_payload(e_data);
840 
841 	/* XXX This is lame, we should provide a way to lookup sockets by
842 	 * qp_handle.
843 	 */
844 	if (vmci_handle_is_invalid(e_payload->handle) ||
845 	    !vmci_handle_is_equal(trans->qp_handle, e_payload->handle))
846 		return;
847 
848 	/* We don't ask for delayed CBs when we subscribe to this event (we
849 	 * pass 0 as flags to vmci_event_subscribe()).  VMCI makes no
850 	 * guarantees in that case about what context we might be running in,
851 	 * so it could be BH or process, blockable or non-blockable.  So we
852 	 * need to account for all possible contexts here.
853 	 */
854 	spin_lock_bh(&trans->lock);
855 	if (!trans->sk)
856 		goto out;
857 
858 	/* Apart from here, trans->lock is only grabbed as part of sk destruct,
859 	 * where trans->sk isn't locked.
860 	 */
861 	bh_lock_sock(trans->sk);
862 
863 	vmci_transport_handle_detach(trans->sk);
864 
865 	bh_unlock_sock(trans->sk);
866  out:
867 	spin_unlock_bh(&trans->lock);
868 }
869 
870 static void vmci_transport_qp_resumed_cb(u32 sub_id,
871 					 const struct vmci_event_data *e_data,
872 					 void *client_data)
873 {
874 	vsock_for_each_connected_socket(vmci_transport_handle_detach);
875 }
876 
877 static void vmci_transport_recv_pkt_work(struct work_struct *work)
878 {
879 	struct vmci_transport_recv_pkt_info *recv_pkt_info;
880 	struct vmci_transport_packet *pkt;
881 	struct sock *sk;
882 
883 	recv_pkt_info =
884 		container_of(work, struct vmci_transport_recv_pkt_info, work);
885 	sk = recv_pkt_info->sk;
886 	pkt = &recv_pkt_info->pkt;
887 
888 	lock_sock(sk);
889 
890 	/* The local context ID may be out of date. */
891 	vsock_sk(sk)->local_addr.svm_cid = pkt->dg.dst.context;
892 
893 	switch (sk->sk_state) {
894 	case VSOCK_SS_LISTEN:
895 		vmci_transport_recv_listen(sk, pkt);
896 		break;
897 	case SS_CONNECTING:
898 		/* Processing of pending connections for servers goes through
899 		 * the listening socket, so see vmci_transport_recv_listen()
900 		 * for that path.
901 		 */
902 		vmci_transport_recv_connecting_client(sk, pkt);
903 		break;
904 	case SS_CONNECTED:
905 		vmci_transport_recv_connected(sk, pkt);
906 		break;
907 	default:
908 		/* Because this function does not run in the same context as
909 		 * vmci_transport_recv_stream_cb it is possible that the
910 		 * socket has closed. We need to let the other side know or it
911 		 * could be sitting in a connect and hang forever. Send a
912 		 * reset to prevent that.
913 		 */
914 		vmci_transport_send_reset(sk, pkt);
915 		break;
916 	}
917 
918 	release_sock(sk);
919 	kfree(recv_pkt_info);
920 	/* Release reference obtained in the stream callback when we fetched
921 	 * this socket out of the bound or connected list.
922 	 */
923 	sock_put(sk);
924 }
925 
926 static int vmci_transport_recv_listen(struct sock *sk,
927 				      struct vmci_transport_packet *pkt)
928 {
929 	struct sock *pending;
930 	struct vsock_sock *vpending;
931 	int err;
932 	u64 qp_size;
933 	bool old_request = false;
934 	bool old_pkt_proto = false;
935 
936 	err = 0;
937 
938 	/* Because we are in the listen state, we could be receiving a packet
939 	 * for ourself or any previous connection requests that we received.
940 	 * If it's the latter, we try to find a socket in our list of pending
941 	 * connections and, if we do, call the appropriate handler for the
942 	 * state that that socket is in.  Otherwise we try to service the
943 	 * connection request.
944 	 */
945 	pending = vmci_transport_get_pending(sk, pkt);
946 	if (pending) {
947 		lock_sock(pending);
948 
949 		/* The local context ID may be out of date. */
950 		vsock_sk(pending)->local_addr.svm_cid = pkt->dg.dst.context;
951 
952 		switch (pending->sk_state) {
953 		case SS_CONNECTING:
954 			err = vmci_transport_recv_connecting_server(sk,
955 								    pending,
956 								    pkt);
957 			break;
958 		default:
959 			vmci_transport_send_reset(pending, pkt);
960 			err = -EINVAL;
961 		}
962 
963 		if (err < 0)
964 			vsock_remove_pending(sk, pending);
965 
966 		release_sock(pending);
967 		vmci_transport_release_pending(pending);
968 
969 		return err;
970 	}
971 
972 	/* The listen state only accepts connection requests.  Reply with a
973 	 * reset unless we received a reset.
974 	 */
975 
976 	if (!(pkt->type == VMCI_TRANSPORT_PACKET_TYPE_REQUEST ||
977 	      pkt->type == VMCI_TRANSPORT_PACKET_TYPE_REQUEST2)) {
978 		vmci_transport_reply_reset(pkt);
979 		return -EINVAL;
980 	}
981 
982 	if (pkt->u.size == 0) {
983 		vmci_transport_reply_reset(pkt);
984 		return -EINVAL;
985 	}
986 
987 	/* If this socket can't accommodate this connection request, we send a
988 	 * reset.  Otherwise we create and initialize a child socket and reply
989 	 * with a connection negotiation.
990 	 */
991 	if (sk->sk_ack_backlog >= sk->sk_max_ack_backlog) {
992 		vmci_transport_reply_reset(pkt);
993 		return -ECONNREFUSED;
994 	}
995 
996 	pending = __vsock_create(sock_net(sk), NULL, sk, GFP_KERNEL,
997 				 sk->sk_type, 0);
998 	if (!pending) {
999 		vmci_transport_send_reset(sk, pkt);
1000 		return -ENOMEM;
1001 	}
1002 
1003 	vpending = vsock_sk(pending);
1004 
1005 	vsock_addr_init(&vpending->local_addr, pkt->dg.dst.context,
1006 			pkt->dst_port);
1007 	vsock_addr_init(&vpending->remote_addr, pkt->dg.src.context,
1008 			pkt->src_port);
1009 
1010 	/* If the proposed size fits within our min/max, accept it. Otherwise
1011 	 * propose our own size.
1012 	 */
1013 	if (pkt->u.size >= vmci_trans(vpending)->queue_pair_min_size &&
1014 	    pkt->u.size <= vmci_trans(vpending)->queue_pair_max_size) {
1015 		qp_size = pkt->u.size;
1016 	} else {
1017 		qp_size = vmci_trans(vpending)->queue_pair_size;
1018 	}
1019 
1020 	/* Figure out if we are using old or new requests based on the
1021 	 * overrides pkt types sent by our peer.
1022 	 */
1023 	if (vmci_transport_old_proto_override(&old_pkt_proto)) {
1024 		old_request = old_pkt_proto;
1025 	} else {
1026 		if (pkt->type == VMCI_TRANSPORT_PACKET_TYPE_REQUEST)
1027 			old_request = true;
1028 		else if (pkt->type == VMCI_TRANSPORT_PACKET_TYPE_REQUEST2)
1029 			old_request = false;
1030 
1031 	}
1032 
1033 	if (old_request) {
1034 		/* Handle a REQUEST (or override) */
1035 		u16 version = VSOCK_PROTO_INVALID;
1036 		if (vmci_transport_proto_to_notify_struct(
1037 			pending, &version, true))
1038 			err = vmci_transport_send_negotiate(pending, qp_size);
1039 		else
1040 			err = -EINVAL;
1041 
1042 	} else {
1043 		/* Handle a REQUEST2 (or override) */
1044 		int proto_int = pkt->proto;
1045 		int pos;
1046 		u16 active_proto_version = 0;
1047 
1048 		/* The list of possible protocols is the intersection of all
1049 		 * protocols the client supports ... plus all the protocols we
1050 		 * support.
1051 		 */
1052 		proto_int &= vmci_transport_new_proto_supported_versions();
1053 
1054 		/* We choose the highest possible protocol version and use that
1055 		 * one.
1056 		 */
1057 		pos = fls(proto_int);
1058 		if (pos) {
1059 			active_proto_version = (1 << (pos - 1));
1060 			if (vmci_transport_proto_to_notify_struct(
1061 				pending, &active_proto_version, false))
1062 				err = vmci_transport_send_negotiate2(pending,
1063 							qp_size,
1064 							active_proto_version);
1065 			else
1066 				err = -EINVAL;
1067 
1068 		} else {
1069 			err = -EINVAL;
1070 		}
1071 	}
1072 
1073 	if (err < 0) {
1074 		vmci_transport_send_reset(sk, pkt);
1075 		sock_put(pending);
1076 		err = vmci_transport_error_to_vsock_error(err);
1077 		goto out;
1078 	}
1079 
1080 	vsock_add_pending(sk, pending);
1081 	sk->sk_ack_backlog++;
1082 
1083 	pending->sk_state = SS_CONNECTING;
1084 	vmci_trans(vpending)->produce_size =
1085 		vmci_trans(vpending)->consume_size = qp_size;
1086 	vmci_trans(vpending)->queue_pair_size = qp_size;
1087 
1088 	vmci_trans(vpending)->notify_ops->process_request(pending);
1089 
1090 	/* We might never receive another message for this socket and it's not
1091 	 * connected to any process, so we have to ensure it gets cleaned up
1092 	 * ourself.  Our delayed work function will take care of that.  Note
1093 	 * that we do not ever cancel this function since we have few
1094 	 * guarantees about its state when calling cancel_delayed_work().
1095 	 * Instead we hold a reference on the socket for that function and make
1096 	 * it capable of handling cases where it needs to do nothing but
1097 	 * release that reference.
1098 	 */
1099 	vpending->listener = sk;
1100 	sock_hold(sk);
1101 	sock_hold(pending);
1102 	INIT_DELAYED_WORK(&vpending->dwork, vsock_pending_work);
1103 	schedule_delayed_work(&vpending->dwork, HZ);
1104 
1105 out:
1106 	return err;
1107 }
1108 
1109 static int
1110 vmci_transport_recv_connecting_server(struct sock *listener,
1111 				      struct sock *pending,
1112 				      struct vmci_transport_packet *pkt)
1113 {
1114 	struct vsock_sock *vpending;
1115 	struct vmci_handle handle;
1116 	struct vmci_qp *qpair;
1117 	bool is_local;
1118 	u32 flags;
1119 	u32 detach_sub_id;
1120 	int err;
1121 	int skerr;
1122 
1123 	vpending = vsock_sk(pending);
1124 	detach_sub_id = VMCI_INVALID_ID;
1125 
1126 	switch (pkt->type) {
1127 	case VMCI_TRANSPORT_PACKET_TYPE_OFFER:
1128 		if (vmci_handle_is_invalid(pkt->u.handle)) {
1129 			vmci_transport_send_reset(pending, pkt);
1130 			skerr = EPROTO;
1131 			err = -EINVAL;
1132 			goto destroy;
1133 		}
1134 		break;
1135 	default:
1136 		/* Close and cleanup the connection. */
1137 		vmci_transport_send_reset(pending, pkt);
1138 		skerr = EPROTO;
1139 		err = pkt->type == VMCI_TRANSPORT_PACKET_TYPE_RST ? 0 : -EINVAL;
1140 		goto destroy;
1141 	}
1142 
1143 	/* In order to complete the connection we need to attach to the offered
1144 	 * queue pair and send an attach notification.  We also subscribe to the
1145 	 * detach event so we know when our peer goes away, and we do that
1146 	 * before attaching so we don't miss an event.  If all this succeeds,
1147 	 * we update our state and wakeup anything waiting in accept() for a
1148 	 * connection.
1149 	 */
1150 
1151 	/* We don't care about attach since we ensure the other side has
1152 	 * attached by specifying the ATTACH_ONLY flag below.
1153 	 */
1154 	err = vmci_event_subscribe(VMCI_EVENT_QP_PEER_DETACH,
1155 				   vmci_transport_peer_detach_cb,
1156 				   vmci_trans(vpending), &detach_sub_id);
1157 	if (err < VMCI_SUCCESS) {
1158 		vmci_transport_send_reset(pending, pkt);
1159 		err = vmci_transport_error_to_vsock_error(err);
1160 		skerr = -err;
1161 		goto destroy;
1162 	}
1163 
1164 	vmci_trans(vpending)->detach_sub_id = detach_sub_id;
1165 
1166 	/* Now attach to the queue pair the client created. */
1167 	handle = pkt->u.handle;
1168 
1169 	/* vpending->local_addr always has a context id so we do not need to
1170 	 * worry about VMADDR_CID_ANY in this case.
1171 	 */
1172 	is_local =
1173 	    vpending->remote_addr.svm_cid == vpending->local_addr.svm_cid;
1174 	flags = VMCI_QPFLAG_ATTACH_ONLY;
1175 	flags |= is_local ? VMCI_QPFLAG_LOCAL : 0;
1176 
1177 	err = vmci_transport_queue_pair_alloc(
1178 					&qpair,
1179 					&handle,
1180 					vmci_trans(vpending)->produce_size,
1181 					vmci_trans(vpending)->consume_size,
1182 					pkt->dg.src.context,
1183 					flags,
1184 					vmci_transport_is_trusted(
1185 						vpending,
1186 						vpending->remote_addr.svm_cid));
1187 	if (err < 0) {
1188 		vmci_transport_send_reset(pending, pkt);
1189 		skerr = -err;
1190 		goto destroy;
1191 	}
1192 
1193 	vmci_trans(vpending)->qp_handle = handle;
1194 	vmci_trans(vpending)->qpair = qpair;
1195 
1196 	/* When we send the attach message, we must be ready to handle incoming
1197 	 * control messages on the newly connected socket. So we move the
1198 	 * pending socket to the connected state before sending the attach
1199 	 * message. Otherwise, an incoming packet triggered by the attach being
1200 	 * received by the peer may be processed concurrently with what happens
1201 	 * below after sending the attach message, and that incoming packet
1202 	 * will find the listening socket instead of the (currently) pending
1203 	 * socket. Note that enqueueing the socket increments the reference
1204 	 * count, so even if a reset comes before the connection is accepted,
1205 	 * the socket will be valid until it is removed from the queue.
1206 	 *
1207 	 * If we fail sending the attach below, we remove the socket from the
1208 	 * connected list and move the socket to SS_UNCONNECTED before
1209 	 * releasing the lock, so a pending slow path processing of an incoming
1210 	 * packet will not see the socket in the connected state in that case.
1211 	 */
1212 	pending->sk_state = SS_CONNECTED;
1213 
1214 	vsock_insert_connected(vpending);
1215 
1216 	/* Notify our peer of our attach. */
1217 	err = vmci_transport_send_attach(pending, handle);
1218 	if (err < 0) {
1219 		vsock_remove_connected(vpending);
1220 		pr_err("Could not send attach\n");
1221 		vmci_transport_send_reset(pending, pkt);
1222 		err = vmci_transport_error_to_vsock_error(err);
1223 		skerr = -err;
1224 		goto destroy;
1225 	}
1226 
1227 	/* We have a connection. Move the now connected socket from the
1228 	 * listener's pending list to the accept queue so callers of accept()
1229 	 * can find it.
1230 	 */
1231 	vsock_remove_pending(listener, pending);
1232 	vsock_enqueue_accept(listener, pending);
1233 
1234 	/* Callers of accept() will be be waiting on the listening socket, not
1235 	 * the pending socket.
1236 	 */
1237 	listener->sk_data_ready(listener);
1238 
1239 	return 0;
1240 
1241 destroy:
1242 	pending->sk_err = skerr;
1243 	pending->sk_state = SS_UNCONNECTED;
1244 	/* As long as we drop our reference, all necessary cleanup will handle
1245 	 * when the cleanup function drops its reference and our destruct
1246 	 * implementation is called.  Note that since the listen handler will
1247 	 * remove pending from the pending list upon our failure, the cleanup
1248 	 * function won't drop the additional reference, which is why we do it
1249 	 * here.
1250 	 */
1251 	sock_put(pending);
1252 
1253 	return err;
1254 }
1255 
1256 static int
1257 vmci_transport_recv_connecting_client(struct sock *sk,
1258 				      struct vmci_transport_packet *pkt)
1259 {
1260 	struct vsock_sock *vsk;
1261 	int err;
1262 	int skerr;
1263 
1264 	vsk = vsock_sk(sk);
1265 
1266 	switch (pkt->type) {
1267 	case VMCI_TRANSPORT_PACKET_TYPE_ATTACH:
1268 		if (vmci_handle_is_invalid(pkt->u.handle) ||
1269 		    !vmci_handle_is_equal(pkt->u.handle,
1270 					  vmci_trans(vsk)->qp_handle)) {
1271 			skerr = EPROTO;
1272 			err = -EINVAL;
1273 			goto destroy;
1274 		}
1275 
1276 		/* Signify the socket is connected and wakeup the waiter in
1277 		 * connect(). Also place the socket in the connected table for
1278 		 * accounting (it can already be found since it's in the bound
1279 		 * table).
1280 		 */
1281 		sk->sk_state = SS_CONNECTED;
1282 		sk->sk_socket->state = SS_CONNECTED;
1283 		vsock_insert_connected(vsk);
1284 		sk->sk_state_change(sk);
1285 
1286 		break;
1287 	case VMCI_TRANSPORT_PACKET_TYPE_NEGOTIATE:
1288 	case VMCI_TRANSPORT_PACKET_TYPE_NEGOTIATE2:
1289 		if (pkt->u.size == 0
1290 		    || pkt->dg.src.context != vsk->remote_addr.svm_cid
1291 		    || pkt->src_port != vsk->remote_addr.svm_port
1292 		    || !vmci_handle_is_invalid(vmci_trans(vsk)->qp_handle)
1293 		    || vmci_trans(vsk)->qpair
1294 		    || vmci_trans(vsk)->produce_size != 0
1295 		    || vmci_trans(vsk)->consume_size != 0
1296 		    || vmci_trans(vsk)->detach_sub_id != VMCI_INVALID_ID) {
1297 			skerr = EPROTO;
1298 			err = -EINVAL;
1299 
1300 			goto destroy;
1301 		}
1302 
1303 		err = vmci_transport_recv_connecting_client_negotiate(sk, pkt);
1304 		if (err) {
1305 			skerr = -err;
1306 			goto destroy;
1307 		}
1308 
1309 		break;
1310 	case VMCI_TRANSPORT_PACKET_TYPE_INVALID:
1311 		err = vmci_transport_recv_connecting_client_invalid(sk, pkt);
1312 		if (err) {
1313 			skerr = -err;
1314 			goto destroy;
1315 		}
1316 
1317 		break;
1318 	case VMCI_TRANSPORT_PACKET_TYPE_RST:
1319 		/* Older versions of the linux code (WS 6.5 / ESX 4.0) used to
1320 		 * continue processing here after they sent an INVALID packet.
1321 		 * This meant that we got a RST after the INVALID. We ignore a
1322 		 * RST after an INVALID. The common code doesn't send the RST
1323 		 * ... so we can hang if an old version of the common code
1324 		 * fails between getting a REQUEST and sending an OFFER back.
1325 		 * Not much we can do about it... except hope that it doesn't
1326 		 * happen.
1327 		 */
1328 		if (vsk->ignore_connecting_rst) {
1329 			vsk->ignore_connecting_rst = false;
1330 		} else {
1331 			skerr = ECONNRESET;
1332 			err = 0;
1333 			goto destroy;
1334 		}
1335 
1336 		break;
1337 	default:
1338 		/* Close and cleanup the connection. */
1339 		skerr = EPROTO;
1340 		err = -EINVAL;
1341 		goto destroy;
1342 	}
1343 
1344 	return 0;
1345 
1346 destroy:
1347 	vmci_transport_send_reset(sk, pkt);
1348 
1349 	sk->sk_state = SS_UNCONNECTED;
1350 	sk->sk_err = skerr;
1351 	sk->sk_error_report(sk);
1352 	return err;
1353 }
1354 
1355 static int vmci_transport_recv_connecting_client_negotiate(
1356 					struct sock *sk,
1357 					struct vmci_transport_packet *pkt)
1358 {
1359 	int err;
1360 	struct vsock_sock *vsk;
1361 	struct vmci_handle handle;
1362 	struct vmci_qp *qpair;
1363 	u32 detach_sub_id;
1364 	bool is_local;
1365 	u32 flags;
1366 	bool old_proto = true;
1367 	bool old_pkt_proto;
1368 	u16 version;
1369 
1370 	vsk = vsock_sk(sk);
1371 	handle = VMCI_INVALID_HANDLE;
1372 	detach_sub_id = VMCI_INVALID_ID;
1373 
1374 	/* If we have gotten here then we should be past the point where old
1375 	 * linux vsock could have sent the bogus rst.
1376 	 */
1377 	vsk->sent_request = false;
1378 	vsk->ignore_connecting_rst = false;
1379 
1380 	/* Verify that we're OK with the proposed queue pair size */
1381 	if (pkt->u.size < vmci_trans(vsk)->queue_pair_min_size ||
1382 	    pkt->u.size > vmci_trans(vsk)->queue_pair_max_size) {
1383 		err = -EINVAL;
1384 		goto destroy;
1385 	}
1386 
1387 	/* At this point we know the CID the peer is using to talk to us. */
1388 
1389 	if (vsk->local_addr.svm_cid == VMADDR_CID_ANY)
1390 		vsk->local_addr.svm_cid = pkt->dg.dst.context;
1391 
1392 	/* Setup the notify ops to be the highest supported version that both
1393 	 * the server and the client support.
1394 	 */
1395 
1396 	if (vmci_transport_old_proto_override(&old_pkt_proto)) {
1397 		old_proto = old_pkt_proto;
1398 	} else {
1399 		if (pkt->type == VMCI_TRANSPORT_PACKET_TYPE_NEGOTIATE)
1400 			old_proto = true;
1401 		else if (pkt->type == VMCI_TRANSPORT_PACKET_TYPE_NEGOTIATE2)
1402 			old_proto = false;
1403 
1404 	}
1405 
1406 	if (old_proto)
1407 		version = VSOCK_PROTO_INVALID;
1408 	else
1409 		version = pkt->proto;
1410 
1411 	if (!vmci_transport_proto_to_notify_struct(sk, &version, old_proto)) {
1412 		err = -EINVAL;
1413 		goto destroy;
1414 	}
1415 
1416 	/* Subscribe to detach events first.
1417 	 *
1418 	 * XXX We attach once for each queue pair created for now so it is easy
1419 	 * to find the socket (it's provided), but later we should only
1420 	 * subscribe once and add a way to lookup sockets by queue pair handle.
1421 	 */
1422 	err = vmci_event_subscribe(VMCI_EVENT_QP_PEER_DETACH,
1423 				   vmci_transport_peer_detach_cb,
1424 				   vmci_trans(vsk), &detach_sub_id);
1425 	if (err < VMCI_SUCCESS) {
1426 		err = vmci_transport_error_to_vsock_error(err);
1427 		goto destroy;
1428 	}
1429 
1430 	/* Make VMCI select the handle for us. */
1431 	handle = VMCI_INVALID_HANDLE;
1432 	is_local = vsk->remote_addr.svm_cid == vsk->local_addr.svm_cid;
1433 	flags = is_local ? VMCI_QPFLAG_LOCAL : 0;
1434 
1435 	err = vmci_transport_queue_pair_alloc(&qpair,
1436 					      &handle,
1437 					      pkt->u.size,
1438 					      pkt->u.size,
1439 					      vsk->remote_addr.svm_cid,
1440 					      flags,
1441 					      vmci_transport_is_trusted(
1442 						  vsk,
1443 						  vsk->
1444 						  remote_addr.svm_cid));
1445 	if (err < 0)
1446 		goto destroy;
1447 
1448 	err = vmci_transport_send_qp_offer(sk, handle);
1449 	if (err < 0) {
1450 		err = vmci_transport_error_to_vsock_error(err);
1451 		goto destroy;
1452 	}
1453 
1454 	vmci_trans(vsk)->qp_handle = handle;
1455 	vmci_trans(vsk)->qpair = qpair;
1456 
1457 	vmci_trans(vsk)->produce_size = vmci_trans(vsk)->consume_size =
1458 		pkt->u.size;
1459 
1460 	vmci_trans(vsk)->detach_sub_id = detach_sub_id;
1461 
1462 	vmci_trans(vsk)->notify_ops->process_negotiate(sk);
1463 
1464 	return 0;
1465 
1466 destroy:
1467 	if (detach_sub_id != VMCI_INVALID_ID)
1468 		vmci_event_unsubscribe(detach_sub_id);
1469 
1470 	if (!vmci_handle_is_invalid(handle))
1471 		vmci_qpair_detach(&qpair);
1472 
1473 	return err;
1474 }
1475 
1476 static int
1477 vmci_transport_recv_connecting_client_invalid(struct sock *sk,
1478 					      struct vmci_transport_packet *pkt)
1479 {
1480 	int err = 0;
1481 	struct vsock_sock *vsk = vsock_sk(sk);
1482 
1483 	if (vsk->sent_request) {
1484 		vsk->sent_request = false;
1485 		vsk->ignore_connecting_rst = true;
1486 
1487 		err = vmci_transport_send_conn_request(
1488 			sk, vmci_trans(vsk)->queue_pair_size);
1489 		if (err < 0)
1490 			err = vmci_transport_error_to_vsock_error(err);
1491 		else
1492 			err = 0;
1493 
1494 	}
1495 
1496 	return err;
1497 }
1498 
1499 static int vmci_transport_recv_connected(struct sock *sk,
1500 					 struct vmci_transport_packet *pkt)
1501 {
1502 	struct vsock_sock *vsk;
1503 	bool pkt_processed = false;
1504 
1505 	/* In cases where we are closing the connection, it's sufficient to
1506 	 * mark the state change (and maybe error) and wake up any waiting
1507 	 * threads. Since this is a connected socket, it's owned by a user
1508 	 * process and will be cleaned up when the failure is passed back on
1509 	 * the current or next system call.  Our system call implementations
1510 	 * must therefore check for error and state changes on entry and when
1511 	 * being awoken.
1512 	 */
1513 	switch (pkt->type) {
1514 	case VMCI_TRANSPORT_PACKET_TYPE_SHUTDOWN:
1515 		if (pkt->u.mode) {
1516 			vsk = vsock_sk(sk);
1517 
1518 			vsk->peer_shutdown |= pkt->u.mode;
1519 			sk->sk_state_change(sk);
1520 		}
1521 		break;
1522 
1523 	case VMCI_TRANSPORT_PACKET_TYPE_RST:
1524 		vsk = vsock_sk(sk);
1525 		/* It is possible that we sent our peer a message (e.g a
1526 		 * WAITING_READ) right before we got notified that the peer had
1527 		 * detached. If that happens then we can get a RST pkt back
1528 		 * from our peer even though there is data available for us to
1529 		 * read. In that case, don't shutdown the socket completely but
1530 		 * instead allow the local client to finish reading data off
1531 		 * the queuepair. Always treat a RST pkt in connected mode like
1532 		 * a clean shutdown.
1533 		 */
1534 		sock_set_flag(sk, SOCK_DONE);
1535 		vsk->peer_shutdown = SHUTDOWN_MASK;
1536 		if (vsock_stream_has_data(vsk) <= 0)
1537 			sk->sk_state = SS_DISCONNECTING;
1538 
1539 		sk->sk_state_change(sk);
1540 		break;
1541 
1542 	default:
1543 		vsk = vsock_sk(sk);
1544 		vmci_trans(vsk)->notify_ops->handle_notify_pkt(
1545 				sk, pkt, false, NULL, NULL,
1546 				&pkt_processed);
1547 		if (!pkt_processed)
1548 			return -EINVAL;
1549 
1550 		break;
1551 	}
1552 
1553 	return 0;
1554 }
1555 
1556 static int vmci_transport_socket_init(struct vsock_sock *vsk,
1557 				      struct vsock_sock *psk)
1558 {
1559 	vsk->trans = kmalloc(sizeof(struct vmci_transport), GFP_KERNEL);
1560 	if (!vsk->trans)
1561 		return -ENOMEM;
1562 
1563 	vmci_trans(vsk)->dg_handle = VMCI_INVALID_HANDLE;
1564 	vmci_trans(vsk)->qp_handle = VMCI_INVALID_HANDLE;
1565 	vmci_trans(vsk)->qpair = NULL;
1566 	vmci_trans(vsk)->produce_size = vmci_trans(vsk)->consume_size = 0;
1567 	vmci_trans(vsk)->detach_sub_id = VMCI_INVALID_ID;
1568 	vmci_trans(vsk)->notify_ops = NULL;
1569 	INIT_LIST_HEAD(&vmci_trans(vsk)->elem);
1570 	vmci_trans(vsk)->sk = &vsk->sk;
1571 	spin_lock_init(&vmci_trans(vsk)->lock);
1572 	if (psk) {
1573 		vmci_trans(vsk)->queue_pair_size =
1574 			vmci_trans(psk)->queue_pair_size;
1575 		vmci_trans(vsk)->queue_pair_min_size =
1576 			vmci_trans(psk)->queue_pair_min_size;
1577 		vmci_trans(vsk)->queue_pair_max_size =
1578 			vmci_trans(psk)->queue_pair_max_size;
1579 	} else {
1580 		vmci_trans(vsk)->queue_pair_size =
1581 			VMCI_TRANSPORT_DEFAULT_QP_SIZE;
1582 		vmci_trans(vsk)->queue_pair_min_size =
1583 			 VMCI_TRANSPORT_DEFAULT_QP_SIZE_MIN;
1584 		vmci_trans(vsk)->queue_pair_max_size =
1585 			VMCI_TRANSPORT_DEFAULT_QP_SIZE_MAX;
1586 	}
1587 
1588 	return 0;
1589 }
1590 
1591 static void vmci_transport_free_resources(struct list_head *transport_list)
1592 {
1593 	while (!list_empty(transport_list)) {
1594 		struct vmci_transport *transport =
1595 		    list_first_entry(transport_list, struct vmci_transport,
1596 				     elem);
1597 		list_del(&transport->elem);
1598 
1599 		if (transport->detach_sub_id != VMCI_INVALID_ID) {
1600 			vmci_event_unsubscribe(transport->detach_sub_id);
1601 			transport->detach_sub_id = VMCI_INVALID_ID;
1602 		}
1603 
1604 		if (!vmci_handle_is_invalid(transport->qp_handle)) {
1605 			vmci_qpair_detach(&transport->qpair);
1606 			transport->qp_handle = VMCI_INVALID_HANDLE;
1607 			transport->produce_size = 0;
1608 			transport->consume_size = 0;
1609 		}
1610 
1611 		kfree(transport);
1612 	}
1613 }
1614 
1615 static void vmci_transport_cleanup(struct work_struct *work)
1616 {
1617 	LIST_HEAD(pending);
1618 
1619 	spin_lock_bh(&vmci_transport_cleanup_lock);
1620 	list_replace_init(&vmci_transport_cleanup_list, &pending);
1621 	spin_unlock_bh(&vmci_transport_cleanup_lock);
1622 	vmci_transport_free_resources(&pending);
1623 }
1624 
1625 static void vmci_transport_destruct(struct vsock_sock *vsk)
1626 {
1627 	/* Ensure that the detach callback doesn't use the sk/vsk
1628 	 * we are about to destruct.
1629 	 */
1630 	spin_lock_bh(&vmci_trans(vsk)->lock);
1631 	vmci_trans(vsk)->sk = NULL;
1632 	spin_unlock_bh(&vmci_trans(vsk)->lock);
1633 
1634 	if (vmci_trans(vsk)->notify_ops)
1635 		vmci_trans(vsk)->notify_ops->socket_destruct(vsk);
1636 
1637 	spin_lock_bh(&vmci_transport_cleanup_lock);
1638 	list_add(&vmci_trans(vsk)->elem, &vmci_transport_cleanup_list);
1639 	spin_unlock_bh(&vmci_transport_cleanup_lock);
1640 	schedule_work(&vmci_transport_cleanup_work);
1641 
1642 	vsk->trans = NULL;
1643 }
1644 
1645 static void vmci_transport_release(struct vsock_sock *vsk)
1646 {
1647 	vsock_remove_sock(vsk);
1648 
1649 	if (!vmci_handle_is_invalid(vmci_trans(vsk)->dg_handle)) {
1650 		vmci_datagram_destroy_handle(vmci_trans(vsk)->dg_handle);
1651 		vmci_trans(vsk)->dg_handle = VMCI_INVALID_HANDLE;
1652 	}
1653 }
1654 
1655 static int vmci_transport_dgram_bind(struct vsock_sock *vsk,
1656 				     struct sockaddr_vm *addr)
1657 {
1658 	u32 port;
1659 	u32 flags;
1660 	int err;
1661 
1662 	/* VMCI will select a resource ID for us if we provide
1663 	 * VMCI_INVALID_ID.
1664 	 */
1665 	port = addr->svm_port == VMADDR_PORT_ANY ?
1666 			VMCI_INVALID_ID : addr->svm_port;
1667 
1668 	if (port <= LAST_RESERVED_PORT && !capable(CAP_NET_BIND_SERVICE))
1669 		return -EACCES;
1670 
1671 	flags = addr->svm_cid == VMADDR_CID_ANY ?
1672 				VMCI_FLAG_ANYCID_DG_HND : 0;
1673 
1674 	err = vmci_transport_datagram_create_hnd(port, flags,
1675 						 vmci_transport_recv_dgram_cb,
1676 						 &vsk->sk,
1677 						 &vmci_trans(vsk)->dg_handle);
1678 	if (err < VMCI_SUCCESS)
1679 		return vmci_transport_error_to_vsock_error(err);
1680 	vsock_addr_init(&vsk->local_addr, addr->svm_cid,
1681 			vmci_trans(vsk)->dg_handle.resource);
1682 
1683 	return 0;
1684 }
1685 
1686 static int vmci_transport_dgram_enqueue(
1687 	struct vsock_sock *vsk,
1688 	struct sockaddr_vm *remote_addr,
1689 	struct msghdr *msg,
1690 	size_t len)
1691 {
1692 	int err;
1693 	struct vmci_datagram *dg;
1694 
1695 	if (len > VMCI_MAX_DG_PAYLOAD_SIZE)
1696 		return -EMSGSIZE;
1697 
1698 	if (!vmci_transport_allow_dgram(vsk, remote_addr->svm_cid))
1699 		return -EPERM;
1700 
1701 	/* Allocate a buffer for the user's message and our packet header. */
1702 	dg = kmalloc(len + sizeof(*dg), GFP_KERNEL);
1703 	if (!dg)
1704 		return -ENOMEM;
1705 
1706 	memcpy_from_msg(VMCI_DG_PAYLOAD(dg), msg, len);
1707 
1708 	dg->dst = vmci_make_handle(remote_addr->svm_cid,
1709 				   remote_addr->svm_port);
1710 	dg->src = vmci_make_handle(vsk->local_addr.svm_cid,
1711 				   vsk->local_addr.svm_port);
1712 	dg->payload_size = len;
1713 
1714 	err = vmci_datagram_send(dg);
1715 	kfree(dg);
1716 	if (err < 0)
1717 		return vmci_transport_error_to_vsock_error(err);
1718 
1719 	return err - sizeof(*dg);
1720 }
1721 
1722 static int vmci_transport_dgram_dequeue(struct vsock_sock *vsk,
1723 					struct msghdr *msg, size_t len,
1724 					int flags)
1725 {
1726 	int err;
1727 	int noblock;
1728 	struct vmci_datagram *dg;
1729 	size_t payload_len;
1730 	struct sk_buff *skb;
1731 
1732 	noblock = flags & MSG_DONTWAIT;
1733 
1734 	if (flags & MSG_OOB || flags & MSG_ERRQUEUE)
1735 		return -EOPNOTSUPP;
1736 
1737 	/* Retrieve the head sk_buff from the socket's receive queue. */
1738 	err = 0;
1739 	skb = skb_recv_datagram(&vsk->sk, flags, noblock, &err);
1740 	if (!skb)
1741 		return err;
1742 
1743 	dg = (struct vmci_datagram *)skb->data;
1744 	if (!dg)
1745 		/* err is 0, meaning we read zero bytes. */
1746 		goto out;
1747 
1748 	payload_len = dg->payload_size;
1749 	/* Ensure the sk_buff matches the payload size claimed in the packet. */
1750 	if (payload_len != skb->len - sizeof(*dg)) {
1751 		err = -EINVAL;
1752 		goto out;
1753 	}
1754 
1755 	if (payload_len > len) {
1756 		payload_len = len;
1757 		msg->msg_flags |= MSG_TRUNC;
1758 	}
1759 
1760 	/* Place the datagram payload in the user's iovec. */
1761 	err = skb_copy_datagram_msg(skb, sizeof(*dg), msg, payload_len);
1762 	if (err)
1763 		goto out;
1764 
1765 	if (msg->msg_name) {
1766 		/* Provide the address of the sender. */
1767 		DECLARE_SOCKADDR(struct sockaddr_vm *, vm_addr, msg->msg_name);
1768 		vsock_addr_init(vm_addr, dg->src.context, dg->src.resource);
1769 		msg->msg_namelen = sizeof(*vm_addr);
1770 	}
1771 	err = payload_len;
1772 
1773 out:
1774 	skb_free_datagram(&vsk->sk, skb);
1775 	return err;
1776 }
1777 
1778 static bool vmci_transport_dgram_allow(u32 cid, u32 port)
1779 {
1780 	if (cid == VMADDR_CID_HYPERVISOR) {
1781 		/* Registrations of PBRPC Servers do not modify VMX/Hypervisor
1782 		 * state and are allowed.
1783 		 */
1784 		return port == VMCI_UNITY_PBRPC_REGISTER;
1785 	}
1786 
1787 	return true;
1788 }
1789 
1790 static int vmci_transport_connect(struct vsock_sock *vsk)
1791 {
1792 	int err;
1793 	bool old_pkt_proto = false;
1794 	struct sock *sk = &vsk->sk;
1795 
1796 	if (vmci_transport_old_proto_override(&old_pkt_proto) &&
1797 		old_pkt_proto) {
1798 		err = vmci_transport_send_conn_request(
1799 			sk, vmci_trans(vsk)->queue_pair_size);
1800 		if (err < 0) {
1801 			sk->sk_state = SS_UNCONNECTED;
1802 			return err;
1803 		}
1804 	} else {
1805 		int supported_proto_versions =
1806 			vmci_transport_new_proto_supported_versions();
1807 		err = vmci_transport_send_conn_request2(
1808 				sk, vmci_trans(vsk)->queue_pair_size,
1809 				supported_proto_versions);
1810 		if (err < 0) {
1811 			sk->sk_state = SS_UNCONNECTED;
1812 			return err;
1813 		}
1814 
1815 		vsk->sent_request = true;
1816 	}
1817 
1818 	return err;
1819 }
1820 
1821 static ssize_t vmci_transport_stream_dequeue(
1822 	struct vsock_sock *vsk,
1823 	struct msghdr *msg,
1824 	size_t len,
1825 	int flags)
1826 {
1827 	if (flags & MSG_PEEK)
1828 		return vmci_qpair_peekv(vmci_trans(vsk)->qpair, msg, len, 0);
1829 	else
1830 		return vmci_qpair_dequev(vmci_trans(vsk)->qpair, msg, len, 0);
1831 }
1832 
1833 static ssize_t vmci_transport_stream_enqueue(
1834 	struct vsock_sock *vsk,
1835 	struct msghdr *msg,
1836 	size_t len)
1837 {
1838 	return vmci_qpair_enquev(vmci_trans(vsk)->qpair, msg, len, 0);
1839 }
1840 
1841 static s64 vmci_transport_stream_has_data(struct vsock_sock *vsk)
1842 {
1843 	return vmci_qpair_consume_buf_ready(vmci_trans(vsk)->qpair);
1844 }
1845 
1846 static s64 vmci_transport_stream_has_space(struct vsock_sock *vsk)
1847 {
1848 	return vmci_qpair_produce_free_space(vmci_trans(vsk)->qpair);
1849 }
1850 
1851 static u64 vmci_transport_stream_rcvhiwat(struct vsock_sock *vsk)
1852 {
1853 	return vmci_trans(vsk)->consume_size;
1854 }
1855 
1856 static bool vmci_transport_stream_is_active(struct vsock_sock *vsk)
1857 {
1858 	return !vmci_handle_is_invalid(vmci_trans(vsk)->qp_handle);
1859 }
1860 
1861 static u64 vmci_transport_get_buffer_size(struct vsock_sock *vsk)
1862 {
1863 	return vmci_trans(vsk)->queue_pair_size;
1864 }
1865 
1866 static u64 vmci_transport_get_min_buffer_size(struct vsock_sock *vsk)
1867 {
1868 	return vmci_trans(vsk)->queue_pair_min_size;
1869 }
1870 
1871 static u64 vmci_transport_get_max_buffer_size(struct vsock_sock *vsk)
1872 {
1873 	return vmci_trans(vsk)->queue_pair_max_size;
1874 }
1875 
1876 static void vmci_transport_set_buffer_size(struct vsock_sock *vsk, u64 val)
1877 {
1878 	if (val < vmci_trans(vsk)->queue_pair_min_size)
1879 		vmci_trans(vsk)->queue_pair_min_size = val;
1880 	if (val > vmci_trans(vsk)->queue_pair_max_size)
1881 		vmci_trans(vsk)->queue_pair_max_size = val;
1882 	vmci_trans(vsk)->queue_pair_size = val;
1883 }
1884 
1885 static void vmci_transport_set_min_buffer_size(struct vsock_sock *vsk,
1886 					       u64 val)
1887 {
1888 	if (val > vmci_trans(vsk)->queue_pair_size)
1889 		vmci_trans(vsk)->queue_pair_size = val;
1890 	vmci_trans(vsk)->queue_pair_min_size = val;
1891 }
1892 
1893 static void vmci_transport_set_max_buffer_size(struct vsock_sock *vsk,
1894 					       u64 val)
1895 {
1896 	if (val < vmci_trans(vsk)->queue_pair_size)
1897 		vmci_trans(vsk)->queue_pair_size = val;
1898 	vmci_trans(vsk)->queue_pair_max_size = val;
1899 }
1900 
1901 static int vmci_transport_notify_poll_in(
1902 	struct vsock_sock *vsk,
1903 	size_t target,
1904 	bool *data_ready_now)
1905 {
1906 	return vmci_trans(vsk)->notify_ops->poll_in(
1907 			&vsk->sk, target, data_ready_now);
1908 }
1909 
1910 static int vmci_transport_notify_poll_out(
1911 	struct vsock_sock *vsk,
1912 	size_t target,
1913 	bool *space_available_now)
1914 {
1915 	return vmci_trans(vsk)->notify_ops->poll_out(
1916 			&vsk->sk, target, space_available_now);
1917 }
1918 
1919 static int vmci_transport_notify_recv_init(
1920 	struct vsock_sock *vsk,
1921 	size_t target,
1922 	struct vsock_transport_recv_notify_data *data)
1923 {
1924 	return vmci_trans(vsk)->notify_ops->recv_init(
1925 			&vsk->sk, target,
1926 			(struct vmci_transport_recv_notify_data *)data);
1927 }
1928 
1929 static int vmci_transport_notify_recv_pre_block(
1930 	struct vsock_sock *vsk,
1931 	size_t target,
1932 	struct vsock_transport_recv_notify_data *data)
1933 {
1934 	return vmci_trans(vsk)->notify_ops->recv_pre_block(
1935 			&vsk->sk, target,
1936 			(struct vmci_transport_recv_notify_data *)data);
1937 }
1938 
1939 static int vmci_transport_notify_recv_pre_dequeue(
1940 	struct vsock_sock *vsk,
1941 	size_t target,
1942 	struct vsock_transport_recv_notify_data *data)
1943 {
1944 	return vmci_trans(vsk)->notify_ops->recv_pre_dequeue(
1945 			&vsk->sk, target,
1946 			(struct vmci_transport_recv_notify_data *)data);
1947 }
1948 
1949 static int vmci_transport_notify_recv_post_dequeue(
1950 	struct vsock_sock *vsk,
1951 	size_t target,
1952 	ssize_t copied,
1953 	bool data_read,
1954 	struct vsock_transport_recv_notify_data *data)
1955 {
1956 	return vmci_trans(vsk)->notify_ops->recv_post_dequeue(
1957 			&vsk->sk, target, copied, data_read,
1958 			(struct vmci_transport_recv_notify_data *)data);
1959 }
1960 
1961 static int vmci_transport_notify_send_init(
1962 	struct vsock_sock *vsk,
1963 	struct vsock_transport_send_notify_data *data)
1964 {
1965 	return vmci_trans(vsk)->notify_ops->send_init(
1966 			&vsk->sk,
1967 			(struct vmci_transport_send_notify_data *)data);
1968 }
1969 
1970 static int vmci_transport_notify_send_pre_block(
1971 	struct vsock_sock *vsk,
1972 	struct vsock_transport_send_notify_data *data)
1973 {
1974 	return vmci_trans(vsk)->notify_ops->send_pre_block(
1975 			&vsk->sk,
1976 			(struct vmci_transport_send_notify_data *)data);
1977 }
1978 
1979 static int vmci_transport_notify_send_pre_enqueue(
1980 	struct vsock_sock *vsk,
1981 	struct vsock_transport_send_notify_data *data)
1982 {
1983 	return vmci_trans(vsk)->notify_ops->send_pre_enqueue(
1984 			&vsk->sk,
1985 			(struct vmci_transport_send_notify_data *)data);
1986 }
1987 
1988 static int vmci_transport_notify_send_post_enqueue(
1989 	struct vsock_sock *vsk,
1990 	ssize_t written,
1991 	struct vsock_transport_send_notify_data *data)
1992 {
1993 	return vmci_trans(vsk)->notify_ops->send_post_enqueue(
1994 			&vsk->sk, written,
1995 			(struct vmci_transport_send_notify_data *)data);
1996 }
1997 
1998 static bool vmci_transport_old_proto_override(bool *old_pkt_proto)
1999 {
2000 	if (PROTOCOL_OVERRIDE != -1) {
2001 		if (PROTOCOL_OVERRIDE == 0)
2002 			*old_pkt_proto = true;
2003 		else
2004 			*old_pkt_proto = false;
2005 
2006 		pr_info("Proto override in use\n");
2007 		return true;
2008 	}
2009 
2010 	return false;
2011 }
2012 
2013 static bool vmci_transport_proto_to_notify_struct(struct sock *sk,
2014 						  u16 *proto,
2015 						  bool old_pkt_proto)
2016 {
2017 	struct vsock_sock *vsk = vsock_sk(sk);
2018 
2019 	if (old_pkt_proto) {
2020 		if (*proto != VSOCK_PROTO_INVALID) {
2021 			pr_err("Can't set both an old and new protocol\n");
2022 			return false;
2023 		}
2024 		vmci_trans(vsk)->notify_ops = &vmci_transport_notify_pkt_ops;
2025 		goto exit;
2026 	}
2027 
2028 	switch (*proto) {
2029 	case VSOCK_PROTO_PKT_ON_NOTIFY:
2030 		vmci_trans(vsk)->notify_ops =
2031 			&vmci_transport_notify_pkt_q_state_ops;
2032 		break;
2033 	default:
2034 		pr_err("Unknown notify protocol version\n");
2035 		return false;
2036 	}
2037 
2038 exit:
2039 	vmci_trans(vsk)->notify_ops->socket_init(sk);
2040 	return true;
2041 }
2042 
2043 static u16 vmci_transport_new_proto_supported_versions(void)
2044 {
2045 	if (PROTOCOL_OVERRIDE != -1)
2046 		return PROTOCOL_OVERRIDE;
2047 
2048 	return VSOCK_PROTO_ALL_SUPPORTED;
2049 }
2050 
2051 static u32 vmci_transport_get_local_cid(void)
2052 {
2053 	return vmci_get_context_id();
2054 }
2055 
2056 static const struct vsock_transport vmci_transport = {
2057 	.init = vmci_transport_socket_init,
2058 	.destruct = vmci_transport_destruct,
2059 	.release = vmci_transport_release,
2060 	.connect = vmci_transport_connect,
2061 	.dgram_bind = vmci_transport_dgram_bind,
2062 	.dgram_dequeue = vmci_transport_dgram_dequeue,
2063 	.dgram_enqueue = vmci_transport_dgram_enqueue,
2064 	.dgram_allow = vmci_transport_dgram_allow,
2065 	.stream_dequeue = vmci_transport_stream_dequeue,
2066 	.stream_enqueue = vmci_transport_stream_enqueue,
2067 	.stream_has_data = vmci_transport_stream_has_data,
2068 	.stream_has_space = vmci_transport_stream_has_space,
2069 	.stream_rcvhiwat = vmci_transport_stream_rcvhiwat,
2070 	.stream_is_active = vmci_transport_stream_is_active,
2071 	.stream_allow = vmci_transport_stream_allow,
2072 	.notify_poll_in = vmci_transport_notify_poll_in,
2073 	.notify_poll_out = vmci_transport_notify_poll_out,
2074 	.notify_recv_init = vmci_transport_notify_recv_init,
2075 	.notify_recv_pre_block = vmci_transport_notify_recv_pre_block,
2076 	.notify_recv_pre_dequeue = vmci_transport_notify_recv_pre_dequeue,
2077 	.notify_recv_post_dequeue = vmci_transport_notify_recv_post_dequeue,
2078 	.notify_send_init = vmci_transport_notify_send_init,
2079 	.notify_send_pre_block = vmci_transport_notify_send_pre_block,
2080 	.notify_send_pre_enqueue = vmci_transport_notify_send_pre_enqueue,
2081 	.notify_send_post_enqueue = vmci_transport_notify_send_post_enqueue,
2082 	.shutdown = vmci_transport_shutdown,
2083 	.set_buffer_size = vmci_transport_set_buffer_size,
2084 	.set_min_buffer_size = vmci_transport_set_min_buffer_size,
2085 	.set_max_buffer_size = vmci_transport_set_max_buffer_size,
2086 	.get_buffer_size = vmci_transport_get_buffer_size,
2087 	.get_min_buffer_size = vmci_transport_get_min_buffer_size,
2088 	.get_max_buffer_size = vmci_transport_get_max_buffer_size,
2089 	.get_local_cid = vmci_transport_get_local_cid,
2090 };
2091 
2092 static int __init vmci_transport_init(void)
2093 {
2094 	int err;
2095 
2096 	/* Create the datagram handle that we will use to send and receive all
2097 	 * VSocket control messages for this context.
2098 	 */
2099 	err = vmci_transport_datagram_create_hnd(VMCI_TRANSPORT_PACKET_RID,
2100 						 VMCI_FLAG_ANYCID_DG_HND,
2101 						 vmci_transport_recv_stream_cb,
2102 						 NULL,
2103 						 &vmci_transport_stream_handle);
2104 	if (err < VMCI_SUCCESS) {
2105 		pr_err("Unable to create datagram handle. (%d)\n", err);
2106 		return vmci_transport_error_to_vsock_error(err);
2107 	}
2108 
2109 	err = vmci_event_subscribe(VMCI_EVENT_QP_RESUMED,
2110 				   vmci_transport_qp_resumed_cb,
2111 				   NULL, &vmci_transport_qp_resumed_sub_id);
2112 	if (err < VMCI_SUCCESS) {
2113 		pr_err("Unable to subscribe to resumed event. (%d)\n", err);
2114 		err = vmci_transport_error_to_vsock_error(err);
2115 		vmci_transport_qp_resumed_sub_id = VMCI_INVALID_ID;
2116 		goto err_destroy_stream_handle;
2117 	}
2118 
2119 	err = vsock_core_init(&vmci_transport);
2120 	if (err < 0)
2121 		goto err_unsubscribe;
2122 
2123 	return 0;
2124 
2125 err_unsubscribe:
2126 	vmci_event_unsubscribe(vmci_transport_qp_resumed_sub_id);
2127 err_destroy_stream_handle:
2128 	vmci_datagram_destroy_handle(vmci_transport_stream_handle);
2129 	return err;
2130 }
2131 module_init(vmci_transport_init);
2132 
2133 static void __exit vmci_transport_exit(void)
2134 {
2135 	cancel_work_sync(&vmci_transport_cleanup_work);
2136 	vmci_transport_free_resources(&vmci_transport_cleanup_list);
2137 
2138 	if (!vmci_handle_is_invalid(vmci_transport_stream_handle)) {
2139 		if (vmci_datagram_destroy_handle(
2140 			vmci_transport_stream_handle) != VMCI_SUCCESS)
2141 			pr_err("Couldn't destroy datagram handle\n");
2142 		vmci_transport_stream_handle = VMCI_INVALID_HANDLE;
2143 	}
2144 
2145 	if (vmci_transport_qp_resumed_sub_id != VMCI_INVALID_ID) {
2146 		vmci_event_unsubscribe(vmci_transport_qp_resumed_sub_id);
2147 		vmci_transport_qp_resumed_sub_id = VMCI_INVALID_ID;
2148 	}
2149 
2150 	vsock_core_exit();
2151 }
2152 module_exit(vmci_transport_exit);
2153 
2154 MODULE_AUTHOR("VMware, Inc.");
2155 MODULE_DESCRIPTION("VMCI transport for Virtual Sockets");
2156 MODULE_VERSION("1.0.4.0-k");
2157 MODULE_LICENSE("GPL v2");
2158 MODULE_ALIAS("vmware_vsock");
2159 MODULE_ALIAS_NETPROTO(PF_VSOCK);
2160