xref: /linux/include/net/af_vsock.h (revision d7bb20b91122fd1e8ecbe6ff9fd13b8eefff8240)
1 /* SPDX-License-Identifier: GPL-2.0-only */
2 /*
3  * VMware vSockets Driver
4  *
5  * Copyright (C) 2007-2013 VMware, Inc. All rights reserved.
6  */
7 
8 #ifndef __AF_VSOCK_H__
9 #define __AF_VSOCK_H__
10 
11 #include <linux/kernel.h>
12 #include <linux/workqueue.h>
13 #include <net/netns/vsock.h>
14 #include <net/sock.h>
15 #include <uapi/linux/vm_sockets.h>
16 
17 #include "vsock_addr.h"
18 
19 #define LAST_RESERVED_PORT 1023
20 
21 #define VSOCK_HASH_SIZE         251
22 extern struct list_head vsock_bind_table[VSOCK_HASH_SIZE + 1];
23 extern struct list_head vsock_connected_table[VSOCK_HASH_SIZE];
24 extern spinlock_t vsock_table_lock;
25 
26 #define vsock_sk(__sk)    ((struct vsock_sock *)__sk)
27 #define sk_vsock(__vsk)   (&(__vsk)->sk)
28 
29 struct vsock_sock {
30 	/* sk must be the first member. */
31 	struct sock sk;
32 	const struct vsock_transport *transport;
33 	struct sockaddr_vm local_addr;
34 	struct sockaddr_vm remote_addr;
35 	/* Links for the global tables of bound and connected sockets. */
36 	struct list_head bound_table;
37 	struct list_head connected_table;
38 	/* Accessed without the socket lock held. This means it can never be
39 	 * modified outsided of socket create or destruct.
40 	 */
41 	bool trusted;
42 	bool cached_peer_allow_dgram;	/* Dgram communication allowed to
43 					 * cached peer?
44 					 */
45 	u32 cached_peer;  /* Context ID of last dgram destination check. */
46 	const struct cred *owner;
47 	/* Rest are SOCK_STREAM only. */
48 	long connect_timeout;
49 	/* Listening socket that this came from. */
50 	struct sock *listener;
51 	/* Used for pending list and accept queue during connection handshake.
52 	 * The listening socket is the head for both lists.  Sockets created
53 	 * for connection requests are placed in the pending list until they
54 	 * are connected, at which point they are put in the accept queue list
55 	 * so they can be accepted in accept().
56 	 */
57 	struct list_head pending_links;
58 	struct list_head accept_queue;
59 	struct delayed_work connect_work;
60 	struct delayed_work pending_work;
61 	struct delayed_work close_work;
62 	bool close_work_scheduled;
63 	u32 peer_shutdown;
64 	bool sent_request;
65 	bool ignore_connecting_rst;
66 
67 	/* Protected by lock_sock(sk) */
68 	u64 buffer_size;
69 	u64 buffer_min_size;
70 	u64 buffer_max_size;
71 	bool recvmsg_inq;
72 
73 	/* Private to transport. */
74 	void *trans;
75 };
76 
77 s64 vsock_connectible_has_data(struct vsock_sock *vsk);
78 s64 vsock_stream_has_data(struct vsock_sock *vsk);
79 s64 vsock_stream_has_space(struct vsock_sock *vsk);
80 struct sock *vsock_create_connected(struct sock *parent);
81 void vsock_data_ready(struct sock *sk);
82 
83 /**** TRANSPORT ****/
84 
85 struct vsock_transport_recv_notify_data {
86 	u64 data1; /* Transport-defined. */
87 	u64 data2; /* Transport-defined. */
88 	bool notify_on_block;
89 };
90 
91 struct vsock_transport_send_notify_data {
92 	u64 data1; /* Transport-defined. */
93 	u64 data2; /* Transport-defined. */
94 };
95 
96 /* Transport features flags */
97 /* Transport provides host->guest communication */
98 #define VSOCK_TRANSPORT_F_H2G		0x00000001
99 /* Transport provides guest->host communication */
100 #define VSOCK_TRANSPORT_F_G2H		0x00000002
101 /* Transport provides DGRAM communication */
102 #define VSOCK_TRANSPORT_F_DGRAM		0x00000004
103 /* Transport provides local (loopback) communication */
104 #define VSOCK_TRANSPORT_F_LOCAL		0x00000008
105 
106 struct vsock_transport {
107 	struct module *module;
108 
109 	/* Initialize/tear-down socket. */
110 	int (*init)(struct vsock_sock *, struct vsock_sock *);
111 	void (*destruct)(struct vsock_sock *);
112 	void (*release)(struct vsock_sock *);
113 
114 	/* Cancel all pending packets sent on vsock. */
115 	int (*cancel_pkt)(struct vsock_sock *vsk);
116 
117 	/* Connections. */
118 	int (*connect)(struct vsock_sock *);
119 
120 	/* DGRAM. */
121 	int (*dgram_bind)(struct vsock_sock *, struct sockaddr_vm *);
122 	int (*dgram_dequeue)(struct vsock_sock *vsk, struct msghdr *msg,
123 			     size_t len, int flags);
124 	int (*dgram_enqueue)(struct vsock_sock *, struct sockaddr_vm *,
125 			     struct msghdr *, size_t len);
126 	bool (*dgram_allow)(struct vsock_sock *vsk, u32 cid, u32 port);
127 
128 	/* STREAM. */
129 	/* TODO: stream_bind() */
130 	ssize_t (*stream_dequeue)(struct vsock_sock *, struct msghdr *,
131 				  size_t len, int flags);
132 	ssize_t (*stream_enqueue)(struct vsock_sock *, struct msghdr *,
133 				  size_t len);
134 	s64 (*stream_has_data)(struct vsock_sock *);
135 	s64 (*stream_has_space)(struct vsock_sock *);
136 	u64 (*stream_rcvhiwat)(struct vsock_sock *);
137 	bool (*stream_is_active)(struct vsock_sock *);
138 	bool (*stream_allow)(struct vsock_sock *vsk, u32 cid, u32 port);
139 
140 	/* SEQ_PACKET. */
141 	ssize_t (*seqpacket_dequeue)(struct vsock_sock *vsk, struct msghdr *msg,
142 				     int flags);
143 	int (*seqpacket_enqueue)(struct vsock_sock *vsk, struct msghdr *msg,
144 				 size_t len);
145 	bool (*seqpacket_allow)(struct vsock_sock *vsk, u32 remote_cid);
146 	u32 (*seqpacket_has_data)(struct vsock_sock *vsk);
147 
148 	/* Notification. */
149 	int (*notify_poll_in)(struct vsock_sock *, size_t, bool *);
150 	int (*notify_poll_out)(struct vsock_sock *, size_t, bool *);
151 	int (*notify_recv_init)(struct vsock_sock *, size_t,
152 		struct vsock_transport_recv_notify_data *);
153 	int (*notify_recv_pre_block)(struct vsock_sock *, size_t,
154 		struct vsock_transport_recv_notify_data *);
155 	int (*notify_recv_pre_dequeue)(struct vsock_sock *, size_t,
156 		struct vsock_transport_recv_notify_data *);
157 	int (*notify_recv_post_dequeue)(struct vsock_sock *, size_t,
158 		ssize_t, bool, struct vsock_transport_recv_notify_data *);
159 	int (*notify_send_init)(struct vsock_sock *,
160 		struct vsock_transport_send_notify_data *);
161 	int (*notify_send_pre_block)(struct vsock_sock *,
162 		struct vsock_transport_send_notify_data *);
163 	int (*notify_send_pre_enqueue)(struct vsock_sock *,
164 		struct vsock_transport_send_notify_data *);
165 	int (*notify_send_post_enqueue)(struct vsock_sock *, ssize_t,
166 		struct vsock_transport_send_notify_data *);
167 	/* sk_lock held by the caller */
168 	void (*notify_buffer_size)(struct vsock_sock *, u64 *);
169 	int (*notify_set_rcvlowat)(struct vsock_sock *vsk, int val);
170 
171 	/* SIOCOUTQ ioctl */
172 	ssize_t (*unsent_bytes)(struct vsock_sock *vsk);
173 
174 	/* Shutdown. */
175 	int (*shutdown)(struct vsock_sock *, int);
176 
177 	/* Addressing. */
178 	u32 (*get_local_cid)(void);
179 
180 	/* Check if this transport serves a specific remote CID.
181 	 * For H2G transports: return true if the CID belongs to a registered
182 	 * guest. If not implemented, all CIDs > VMADDR_CID_HOST go to H2G.
183 	 * For G2H transports: return true if the transport can reach arbitrary
184 	 * CIDs via the hypervisor (i.e. supports the fallback overlay). VMCI
185 	 * does not implement this as it only serves CIDs 0 and 2.
186 	 */
187 	bool (*has_remote_cid)(struct vsock_sock *vsk, u32 remote_cid);
188 
189 	/* Read a single skb */
190 	int (*read_skb)(struct vsock_sock *, skb_read_actor_t);
191 
192 	/* Zero-copy. */
193 	bool (*msgzerocopy_allow)(void);
194 };
195 
196 /**** CORE ****/
197 
198 int vsock_core_register(const struct vsock_transport *t, int features);
199 void vsock_core_unregister(const struct vsock_transport *t);
200 
201 /* The transport may downcast this to access transport-specific functions */
202 const struct vsock_transport *vsock_core_get_transport(struct vsock_sock *vsk);
203 
204 /**** UTILS ****/
205 
206 /* vsock_table_lock must be held */
207 static inline bool __vsock_in_bound_table(struct vsock_sock *vsk)
208 {
209 	return !list_empty(&vsk->bound_table);
210 }
211 
212 /* vsock_table_lock must be held */
213 static inline bool __vsock_in_connected_table(struct vsock_sock *vsk)
214 {
215 	return !list_empty(&vsk->connected_table);
216 }
217 
218 void vsock_add_pending(struct sock *listener, struct sock *pending);
219 void vsock_remove_pending(struct sock *listener, struct sock *pending);
220 void vsock_enqueue_accept(struct sock *listener, struct sock *connected);
221 void vsock_pending_to_accept(struct sock *listener, struct sock *pending);
222 void vsock_insert_connected(struct vsock_sock *vsk);
223 void vsock_remove_bound(struct vsock_sock *vsk);
224 void vsock_remove_connected(struct vsock_sock *vsk);
225 struct sock *vsock_find_bound_socket(struct sockaddr_vm *addr);
226 struct sock *vsock_find_connected_socket(struct sockaddr_vm *src,
227 					 struct sockaddr_vm *dst);
228 struct sock *vsock_find_bound_socket_net(struct sockaddr_vm *addr,
229 					 struct net *net);
230 struct sock *vsock_find_connected_socket_net(struct sockaddr_vm *src,
231 					     struct sockaddr_vm *dst,
232 					     struct net *net);
233 bool vsock_check_source(const struct vsock_sock *vsk,
234 			const struct vsock_transport *transport,
235 			const struct sockaddr_vm *src);
236 void vsock_remove_sock(struct vsock_sock *vsk);
237 void vsock_for_each_connected_socket(struct vsock_transport *transport,
238 				     void (*fn)(struct sock *sk));
239 int vsock_assign_transport(struct vsock_sock *vsk, struct vsock_sock *psk);
240 bool vsock_find_cid(unsigned int cid);
241 void vsock_linger(struct sock *sk);
242 
243 /**** TAP ****/
244 
245 struct vsock_tap {
246 	struct net_device *dev;
247 	struct module *module;
248 	struct list_head list;
249 };
250 
251 int vsock_add_tap(struct vsock_tap *vt);
252 int vsock_remove_tap(struct vsock_tap *vt);
253 void vsock_deliver_tap(struct sk_buff *build_skb(void *opaque), void *opaque);
254 int __vsock_connectible_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
255 				int flags);
256 int vsock_connectible_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
257 			      int flags);
258 int __vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg,
259 			  size_t len, int flags);
260 int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg,
261 			size_t len, int flags);
262 
263 extern struct proto vsock_proto;
264 #ifdef CONFIG_BPF_SYSCALL
265 int vsock_bpf_update_proto(struct sock *sk, struct sk_psock *psock, bool restore);
266 void __init vsock_bpf_build_proto(void);
267 #else
268 static inline void __init vsock_bpf_build_proto(void)
269 {}
270 #endif
271 
272 static inline bool vsock_msgzerocopy_allow(const struct vsock_transport *t)
273 {
274 	return t->msgzerocopy_allow && t->msgzerocopy_allow();
275 }
276 
277 static inline enum vsock_net_mode vsock_net_mode(struct net *net)
278 {
279 	if (!net)
280 		return VSOCK_NET_MODE_GLOBAL;
281 
282 	return READ_ONCE(net->vsock.mode);
283 }
284 
285 static inline bool vsock_net_mode_global(struct vsock_sock *vsk)
286 {
287 	return vsock_net_mode(sock_net(sk_vsock(vsk))) == VSOCK_NET_MODE_GLOBAL;
288 }
289 
290 static inline bool vsock_net_set_child_mode(struct net *net,
291 					    enum vsock_net_mode mode)
292 {
293 	int new_locked = mode + 1;
294 	int old_locked = 0; /* unlocked */
295 
296 	if (try_cmpxchg(&net->vsock.child_ns_mode_locked,
297 			&old_locked, new_locked)) {
298 		WRITE_ONCE(net->vsock.child_ns_mode, mode);
299 		return true;
300 	}
301 
302 	return old_locked == new_locked;
303 }
304 
305 static inline enum vsock_net_mode vsock_net_child_mode(struct net *net)
306 {
307 	return READ_ONCE(net->vsock.child_ns_mode);
308 }
309 
310 /* Return true if two namespaces pass the mode rules. Otherwise, return false.
311  *
312  * A NULL namespace is treated as VSOCK_NET_MODE_GLOBAL.
313  *
314  * Read more about modes in the comment header of net/vmw_vsock/af_vsock.c.
315  */
316 static inline bool vsock_net_check_mode(struct net *ns0, struct net *ns1)
317 {
318 	enum vsock_net_mode mode0, mode1;
319 
320 	/* Any vsocks within the same network namespace are always reachable,
321 	 * regardless of the mode.
322 	 */
323 	if (net_eq(ns0, ns1))
324 		return true;
325 
326 	mode0 = vsock_net_mode(ns0);
327 	mode1 = vsock_net_mode(ns1);
328 
329 	/* Different namespaces are only reachable if they are both
330 	 * global mode.
331 	 */
332 	return mode0 == VSOCK_NET_MODE_GLOBAL && mode0 == mode1;
333 }
334 #endif /* __AF_VSOCK_H__ */
335