xref: /linux/include/net/af_vsock.h (revision 91ec2035134982b98fab0609a9fd8480e8217dc1)
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 
72 	/* Private to transport. */
73 	void *trans;
74 };
75 
76 s64 vsock_connectible_has_data(struct vsock_sock *vsk);
77 s64 vsock_stream_has_data(struct vsock_sock *vsk);
78 s64 vsock_stream_has_space(struct vsock_sock *vsk);
79 struct sock *vsock_create_connected(struct sock *parent);
80 void vsock_data_ready(struct sock *sk);
81 
82 /**** TRANSPORT ****/
83 
84 struct vsock_transport_recv_notify_data {
85 	u64 data1; /* Transport-defined. */
86 	u64 data2; /* Transport-defined. */
87 	bool notify_on_block;
88 };
89 
90 struct vsock_transport_send_notify_data {
91 	u64 data1; /* Transport-defined. */
92 	u64 data2; /* Transport-defined. */
93 };
94 
95 /* Transport features flags */
96 /* Transport provides host->guest communication */
97 #define VSOCK_TRANSPORT_F_H2G		0x00000001
98 /* Transport provides guest->host communication */
99 #define VSOCK_TRANSPORT_F_G2H		0x00000002
100 /* Transport provides DGRAM communication */
101 #define VSOCK_TRANSPORT_F_DGRAM		0x00000004
102 /* Transport provides local (loopback) communication */
103 #define VSOCK_TRANSPORT_F_LOCAL		0x00000008
104 
105 struct vsock_transport {
106 	struct module *module;
107 
108 	/* Initialize/tear-down socket. */
109 	int (*init)(struct vsock_sock *, struct vsock_sock *);
110 	void (*destruct)(struct vsock_sock *);
111 	void (*release)(struct vsock_sock *);
112 
113 	/* Cancel all pending packets sent on vsock. */
114 	int (*cancel_pkt)(struct vsock_sock *vsk);
115 
116 	/* Connections. */
117 	int (*connect)(struct vsock_sock *);
118 
119 	/* DGRAM. */
120 	int (*dgram_bind)(struct vsock_sock *, struct sockaddr_vm *);
121 	int (*dgram_dequeue)(struct vsock_sock *vsk, struct msghdr *msg,
122 			     size_t len, int flags);
123 	int (*dgram_enqueue)(struct vsock_sock *, struct sockaddr_vm *,
124 			     struct msghdr *, size_t len);
125 	bool (*dgram_allow)(struct vsock_sock *vsk, u32 cid, u32 port);
126 
127 	/* STREAM. */
128 	/* TODO: stream_bind() */
129 	ssize_t (*stream_dequeue)(struct vsock_sock *, struct msghdr *,
130 				  size_t len, int flags);
131 	ssize_t (*stream_enqueue)(struct vsock_sock *, struct msghdr *,
132 				  size_t len);
133 	s64 (*stream_has_data)(struct vsock_sock *);
134 	s64 (*stream_has_space)(struct vsock_sock *);
135 	u64 (*stream_rcvhiwat)(struct vsock_sock *);
136 	bool (*stream_is_active)(struct vsock_sock *);
137 	bool (*stream_allow)(struct vsock_sock *vsk, u32 cid, u32 port);
138 
139 	/* SEQ_PACKET. */
140 	ssize_t (*seqpacket_dequeue)(struct vsock_sock *vsk, struct msghdr *msg,
141 				     int flags);
142 	int (*seqpacket_enqueue)(struct vsock_sock *vsk, struct msghdr *msg,
143 				 size_t len);
144 	bool (*seqpacket_allow)(struct vsock_sock *vsk, u32 remote_cid);
145 	u32 (*seqpacket_has_data)(struct vsock_sock *vsk);
146 
147 	/* Notification. */
148 	int (*notify_poll_in)(struct vsock_sock *, size_t, bool *);
149 	int (*notify_poll_out)(struct vsock_sock *, size_t, bool *);
150 	int (*notify_recv_init)(struct vsock_sock *, size_t,
151 		struct vsock_transport_recv_notify_data *);
152 	int (*notify_recv_pre_block)(struct vsock_sock *, size_t,
153 		struct vsock_transport_recv_notify_data *);
154 	int (*notify_recv_pre_dequeue)(struct vsock_sock *, size_t,
155 		struct vsock_transport_recv_notify_data *);
156 	int (*notify_recv_post_dequeue)(struct vsock_sock *, size_t,
157 		ssize_t, bool, struct vsock_transport_recv_notify_data *);
158 	int (*notify_send_init)(struct vsock_sock *,
159 		struct vsock_transport_send_notify_data *);
160 	int (*notify_send_pre_block)(struct vsock_sock *,
161 		struct vsock_transport_send_notify_data *);
162 	int (*notify_send_pre_enqueue)(struct vsock_sock *,
163 		struct vsock_transport_send_notify_data *);
164 	int (*notify_send_post_enqueue)(struct vsock_sock *, ssize_t,
165 		struct vsock_transport_send_notify_data *);
166 	/* sk_lock held by the caller */
167 	void (*notify_buffer_size)(struct vsock_sock *, u64 *);
168 	int (*notify_set_rcvlowat)(struct vsock_sock *vsk, int val);
169 
170 	/* SIOCOUTQ ioctl */
171 	ssize_t (*unsent_bytes)(struct vsock_sock *vsk);
172 
173 	/* Shutdown. */
174 	int (*shutdown)(struct vsock_sock *, int);
175 
176 	/* Addressing. */
177 	u32 (*get_local_cid)(void);
178 
179 	/* Check if this transport serves a specific remote CID.
180 	 * For H2G transports: return true if the CID belongs to a registered
181 	 * guest. If not implemented, all CIDs > VMADDR_CID_HOST go to H2G.
182 	 * For G2H transports: return true if the transport can reach arbitrary
183 	 * CIDs via the hypervisor (i.e. supports the fallback overlay). VMCI
184 	 * does not implement this as it only serves CIDs 0 and 2.
185 	 */
186 	bool (*has_remote_cid)(struct vsock_sock *vsk, u32 remote_cid);
187 
188 	/* Read a single skb */
189 	int (*read_skb)(struct vsock_sock *, skb_read_actor_t);
190 
191 	/* Zero-copy. */
192 	bool (*msgzerocopy_allow)(void);
193 };
194 
195 /**** CORE ****/
196 
197 int vsock_core_register(const struct vsock_transport *t, int features);
198 void vsock_core_unregister(const struct vsock_transport *t);
199 
200 /* The transport may downcast this to access transport-specific functions */
201 const struct vsock_transport *vsock_core_get_transport(struct vsock_sock *vsk);
202 
203 /**** UTILS ****/
204 
205 /* vsock_table_lock must be held */
__vsock_in_bound_table(struct vsock_sock * vsk)206 static inline bool __vsock_in_bound_table(struct vsock_sock *vsk)
207 {
208 	return !list_empty(&vsk->bound_table);
209 }
210 
211 /* vsock_table_lock must be held */
__vsock_in_connected_table(struct vsock_sock * vsk)212 static inline bool __vsock_in_connected_table(struct vsock_sock *vsk)
213 {
214 	return !list_empty(&vsk->connected_table);
215 }
216 
217 void vsock_add_pending(struct sock *listener, struct sock *pending);
218 void vsock_remove_pending(struct sock *listener, struct sock *pending);
219 void vsock_enqueue_accept(struct sock *listener, struct sock *connected);
220 void vsock_pending_to_accept(struct sock *listener, struct sock *pending);
221 void vsock_insert_connected(struct vsock_sock *vsk);
222 void vsock_remove_bound(struct vsock_sock *vsk);
223 void vsock_remove_connected(struct vsock_sock *vsk);
224 struct sock *vsock_find_bound_socket(struct sockaddr_vm *addr);
225 struct sock *vsock_find_connected_socket(struct sockaddr_vm *src,
226 					 struct sockaddr_vm *dst);
227 struct sock *vsock_find_bound_socket_net(struct sockaddr_vm *addr,
228 					 struct net *net);
229 struct sock *vsock_find_connected_socket_net(struct sockaddr_vm *src,
230 					     struct sockaddr_vm *dst,
231 					     struct net *net);
232 void vsock_remove_sock(struct vsock_sock *vsk);
233 void vsock_for_each_connected_socket(struct vsock_transport *transport,
234 				     void (*fn)(struct sock *sk));
235 int vsock_assign_transport(struct vsock_sock *vsk, struct vsock_sock *psk);
236 bool vsock_find_cid(unsigned int cid);
237 void vsock_linger(struct sock *sk);
238 
239 /**** TAP ****/
240 
241 struct vsock_tap {
242 	struct net_device *dev;
243 	struct module *module;
244 	struct list_head list;
245 };
246 
247 int vsock_add_tap(struct vsock_tap *vt);
248 int vsock_remove_tap(struct vsock_tap *vt);
249 void vsock_deliver_tap(struct sk_buff *build_skb(void *opaque), void *opaque);
250 int __vsock_connectible_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
251 				int flags);
252 int vsock_connectible_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
253 			      int flags);
254 int __vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg,
255 			  size_t len, int flags);
256 int vsock_dgram_recvmsg(struct socket *sock, struct msghdr *msg,
257 			size_t len, int flags);
258 
259 extern struct proto vsock_proto;
260 #ifdef CONFIG_BPF_SYSCALL
261 int vsock_bpf_update_proto(struct sock *sk, struct sk_psock *psock, bool restore);
262 void __init vsock_bpf_build_proto(void);
263 #else
vsock_bpf_build_proto(void)264 static inline void __init vsock_bpf_build_proto(void)
265 {}
266 #endif
267 
vsock_msgzerocopy_allow(const struct vsock_transport * t)268 static inline bool vsock_msgzerocopy_allow(const struct vsock_transport *t)
269 {
270 	return t->msgzerocopy_allow && t->msgzerocopy_allow();
271 }
272 
vsock_net_mode(struct net * net)273 static inline enum vsock_net_mode vsock_net_mode(struct net *net)
274 {
275 	if (!net)
276 		return VSOCK_NET_MODE_GLOBAL;
277 
278 	return READ_ONCE(net->vsock.mode);
279 }
280 
vsock_net_mode_global(struct vsock_sock * vsk)281 static inline bool vsock_net_mode_global(struct vsock_sock *vsk)
282 {
283 	return vsock_net_mode(sock_net(sk_vsock(vsk))) == VSOCK_NET_MODE_GLOBAL;
284 }
285 
vsock_net_set_child_mode(struct net * net,enum vsock_net_mode mode)286 static inline bool vsock_net_set_child_mode(struct net *net,
287 					    enum vsock_net_mode mode)
288 {
289 	int new_locked = mode + 1;
290 	int old_locked = 0; /* unlocked */
291 
292 	if (try_cmpxchg(&net->vsock.child_ns_mode_locked,
293 			&old_locked, new_locked)) {
294 		WRITE_ONCE(net->vsock.child_ns_mode, mode);
295 		return true;
296 	}
297 
298 	return old_locked == new_locked;
299 }
300 
vsock_net_child_mode(struct net * net)301 static inline enum vsock_net_mode vsock_net_child_mode(struct net *net)
302 {
303 	return READ_ONCE(net->vsock.child_ns_mode);
304 }
305 
306 /* Return true if two namespaces pass the mode rules. Otherwise, return false.
307  *
308  * A NULL namespace is treated as VSOCK_NET_MODE_GLOBAL.
309  *
310  * Read more about modes in the comment header of net/vmw_vsock/af_vsock.c.
311  */
vsock_net_check_mode(struct net * ns0,struct net * ns1)312 static inline bool vsock_net_check_mode(struct net *ns0, struct net *ns1)
313 {
314 	enum vsock_net_mode mode0, mode1;
315 
316 	/* Any vsocks within the same network namespace are always reachable,
317 	 * regardless of the mode.
318 	 */
319 	if (net_eq(ns0, ns1))
320 		return true;
321 
322 	mode0 = vsock_net_mode(ns0);
323 	mode1 = vsock_net_mode(ns1);
324 
325 	/* Different namespaces are only reachable if they are both
326 	 * global mode.
327 	 */
328 	return mode0 == VSOCK_NET_MODE_GLOBAL && mode0 == mode1;
329 }
330 #endif /* __AF_VSOCK_H__ */
331