1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /*
3 * NET An implementation of the SOCKET network access protocol.
4 * This is the master header file for the Linux NET layer,
5 * or, in plain English: the networking handling part of the
6 * kernel.
7 *
8 * Version: @(#)net.h 1.0.3 05/25/93
9 *
10 * Authors: Orest Zborowski, <obz@Kodak.COM>
11 * Ross Biro
12 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
13 */
14 #ifndef _LINUX_NET_H
15 #define _LINUX_NET_H
16
17 #include <linux/stringify.h>
18 #include <linux/random.h>
19 #include <linux/wait.h>
20 #include <linux/fcntl.h> /* For O_CLOEXEC and O_NONBLOCK */
21 #include <linux/rcupdate.h>
22 #include <linux/once.h>
23 #include <linux/fs.h>
24 #include <linux/mm.h>
25 #include <linux/sockptr.h>
26
27 #include <uapi/linux/net.h>
28
29 struct poll_table_struct;
30 struct pipe_inode_info;
31 struct inode;
32 struct file;
33 struct net;
34
35 /* Historically, SOCKWQ_ASYNC_NOSPACE & SOCKWQ_ASYNC_WAITDATA were located
36 * in sock->flags, but moved into sk->sk_wq->flags to be RCU protected.
37 * Eventually all flags will be in sk->sk_wq->flags.
38 */
39 #define SOCKWQ_ASYNC_NOSPACE 0
40 #define SOCKWQ_ASYNC_WAITDATA 1
41 #define SOCK_NOSPACE 2
42 #define SOCK_PASSCRED 3
43 #define SOCK_PASSSEC 4
44
45 #ifndef ARCH_HAS_SOCKET_TYPES
46 /**
47 * enum sock_type - Socket types
48 * @SOCK_STREAM: stream (connection) socket
49 * @SOCK_DGRAM: datagram (conn.less) socket
50 * @SOCK_RAW: raw socket
51 * @SOCK_RDM: reliably-delivered message
52 * @SOCK_SEQPACKET: sequential packet socket
53 * @SOCK_DCCP: Datagram Congestion Control Protocol socket
54 * @SOCK_PACKET: linux specific way of getting packets at the dev level.
55 * For writing rarp and other similar things on the user level.
56 *
57 * When adding some new socket type please
58 * grep ARCH_HAS_SOCKET_TYPE include/asm-* /socket.h, at least MIPS
59 * overrides this enum for binary compat reasons.
60 */
61 enum sock_type {
62 SOCK_STREAM = 1,
63 SOCK_DGRAM = 2,
64 SOCK_RAW = 3,
65 SOCK_RDM = 4,
66 SOCK_SEQPACKET = 5,
67 SOCK_DCCP = 6,
68 SOCK_PACKET = 10,
69 };
70
71 #define SOCK_MAX (SOCK_PACKET + 1)
72 /* Mask which covers at least up to SOCK_MASK-1. The
73 * remaining bits are used as flags. */
74 #define SOCK_TYPE_MASK 0xf
75
76 /* Flags for socket, socketpair, accept4 */
77 #define SOCK_CLOEXEC O_CLOEXEC
78 #ifndef SOCK_NONBLOCK
79 #define SOCK_NONBLOCK O_NONBLOCK
80 #endif
81
82 #endif /* ARCH_HAS_SOCKET_TYPES */
83
84 /**
85 * enum sock_shutdown_cmd - Shutdown types
86 * @SHUT_RD: shutdown receptions
87 * @SHUT_WR: shutdown transmissions
88 * @SHUT_RDWR: shutdown receptions/transmissions
89 */
90 enum sock_shutdown_cmd {
91 SHUT_RD,
92 SHUT_WR,
93 SHUT_RDWR,
94 };
95
96 struct socket_wq {
97 /* Note: wait MUST be first field of socket_wq */
98 wait_queue_head_t wait;
99 struct fasync_struct *fasync_list;
100 unsigned long flags; /* %SOCKWQ_ASYNC_NOSPACE, etc */
101 struct rcu_head rcu;
102 } ____cacheline_aligned_in_smp;
103
104 /**
105 * struct socket - general BSD socket
106 * @state: socket state (%SS_CONNECTED, etc)
107 * @type: socket type (%SOCK_STREAM, etc)
108 * @flags: socket flags (%SOCK_NOSPACE, etc)
109 * @ops: protocol specific socket operations
110 * @file: File back pointer for gc
111 * @sk: internal networking protocol agnostic socket representation
112 * @wq: wait queue for several uses
113 */
114 struct socket {
115 socket_state state;
116
117 short type;
118
119 unsigned long flags;
120
121 struct file *file;
122 struct sock *sk;
123 const struct proto_ops *ops;
124
125 struct socket_wq wq;
126 };
127
128 struct vm_area_struct;
129 struct page;
130 struct sockaddr;
131 struct msghdr;
132 struct module;
133 struct sk_buff;
134 typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *,
135 unsigned int, size_t);
136
137 struct proto_ops {
138 int family;
139 struct module *owner;
140 int (*release) (struct socket *sock);
141 int (*bind) (struct socket *sock,
142 struct sockaddr *myaddr,
143 int sockaddr_len);
144 int (*connect) (struct socket *sock,
145 struct sockaddr *vaddr,
146 int sockaddr_len, int flags);
147 int (*socketpair)(struct socket *sock1,
148 struct socket *sock2);
149 int (*accept) (struct socket *sock,
150 struct socket *newsock, int flags, bool kern);
151 int (*getname) (struct socket *sock,
152 struct sockaddr *addr,
153 int peer);
154 __poll_t (*poll) (struct file *file, struct socket *sock,
155 struct poll_table_struct *wait);
156 int (*ioctl) (struct socket *sock, unsigned int cmd,
157 unsigned long arg);
158 #ifdef CONFIG_COMPAT
159 int (*compat_ioctl) (struct socket *sock, unsigned int cmd,
160 unsigned long arg);
161 #endif
162 int (*gettstamp) (struct socket *sock, void __user *userstamp,
163 bool timeval, bool time32);
164 int (*listen) (struct socket *sock, int len);
165 int (*shutdown) (struct socket *sock, int flags);
166 int (*setsockopt)(struct socket *sock, int level,
167 int optname, sockptr_t optval,
168 unsigned int optlen);
169 int (*getsockopt)(struct socket *sock, int level,
170 int optname, char __user *optval, int __user *optlen);
171 void (*show_fdinfo)(struct seq_file *m, struct socket *sock);
172 int (*sendmsg) (struct socket *sock, struct msghdr *m,
173 size_t total_len);
174 /* Notes for implementing recvmsg:
175 * ===============================
176 * msg->msg_namelen should get updated by the recvmsg handlers
177 * iff msg_name != NULL. It is by default 0 to prevent
178 * returning uninitialized memory to user space. The recvfrom
179 * handlers can assume that msg.msg_name is either NULL or has
180 * a minimum size of sizeof(struct sockaddr_storage).
181 */
182 int (*recvmsg) (struct socket *sock, struct msghdr *m,
183 size_t total_len, int flags);
184 int (*mmap) (struct file *file, struct socket *sock,
185 struct vm_area_struct * vma);
186 ssize_t (*sendpage) (struct socket *sock, struct page *page,
187 int offset, size_t size, int flags);
188 ssize_t (*splice_read)(struct socket *sock, loff_t *ppos,
189 struct pipe_inode_info *pipe, size_t len, unsigned int flags);
190 int (*set_peek_off)(struct sock *sk, int val);
191 int (*peek_len)(struct socket *sock);
192
193 /* The following functions are called internally by kernel with
194 * sock lock already held.
195 */
196 int (*read_sock)(struct sock *sk, read_descriptor_t *desc,
197 sk_read_actor_t recv_actor);
198 int (*sendpage_locked)(struct sock *sk, struct page *page,
199 int offset, size_t size, int flags);
200 int (*sendmsg_locked)(struct sock *sk, struct msghdr *msg,
201 size_t size);
202 int (*set_rcvlowat)(struct sock *sk, int val);
203 };
204
205 #define DECLARE_SOCKADDR(type, dst, src) \
206 type dst = ({ __sockaddr_check_size(sizeof(*dst)); (type) src; })
207
208 struct net_proto_family {
209 int family;
210 int (*create)(struct net *net, struct socket *sock,
211 int protocol, int kern);
212 struct module *owner;
213 };
214
215 struct iovec;
216 struct kvec;
217
218 enum {
219 SOCK_WAKE_IO,
220 SOCK_WAKE_WAITD,
221 SOCK_WAKE_SPACE,
222 SOCK_WAKE_URG,
223 };
224
225 int sock_wake_async(struct socket_wq *sk_wq, int how, int band);
226 int sock_register(const struct net_proto_family *fam);
227 void sock_unregister(int family);
228 bool sock_is_registered(int family);
229 int __sock_create(struct net *net, int family, int type, int proto,
230 struct socket **res, int kern);
231 int sock_create(int family, int type, int proto, struct socket **res);
232 int sock_create_kern(struct net *net, int family, int type, int proto, struct socket **res);
233 int sock_create_lite(int family, int type, int proto, struct socket **res);
234 struct socket *sock_alloc(void);
235 void sock_release(struct socket *sock);
236 int sock_sendmsg(struct socket *sock, struct msghdr *msg);
237 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags);
238 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname);
239 struct socket *sockfd_lookup(int fd, int *err);
240 struct socket *sock_from_file(struct file *file);
241 #define sockfd_put(sock) fput(sock->file)
242 int net_ratelimit(void);
243
244 #define net_ratelimited_function(function, ...) \
245 do { \
246 if (net_ratelimit()) \
247 function(__VA_ARGS__); \
248 } while (0)
249
250 #define net_emerg_ratelimited(fmt, ...) \
251 net_ratelimited_function(pr_emerg, fmt, ##__VA_ARGS__)
252 #define net_alert_ratelimited(fmt, ...) \
253 net_ratelimited_function(pr_alert, fmt, ##__VA_ARGS__)
254 #define net_crit_ratelimited(fmt, ...) \
255 net_ratelimited_function(pr_crit, fmt, ##__VA_ARGS__)
256 #define net_err_ratelimited(fmt, ...) \
257 net_ratelimited_function(pr_err, fmt, ##__VA_ARGS__)
258 #define net_notice_ratelimited(fmt, ...) \
259 net_ratelimited_function(pr_notice, fmt, ##__VA_ARGS__)
260 #define net_warn_ratelimited(fmt, ...) \
261 net_ratelimited_function(pr_warn, fmt, ##__VA_ARGS__)
262 #define net_info_ratelimited(fmt, ...) \
263 net_ratelimited_function(pr_info, fmt, ##__VA_ARGS__)
264 #if defined(CONFIG_DYNAMIC_DEBUG) || \
265 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
266 #define net_dbg_ratelimited(fmt, ...) \
267 do { \
268 DEFINE_DYNAMIC_DEBUG_METADATA(descriptor, fmt); \
269 if (DYNAMIC_DEBUG_BRANCH(descriptor) && \
270 net_ratelimit()) \
271 __dynamic_pr_debug(&descriptor, pr_fmt(fmt), \
272 ##__VA_ARGS__); \
273 } while (0)
274 #elif defined(DEBUG)
275 #define net_dbg_ratelimited(fmt, ...) \
276 net_ratelimited_function(pr_debug, fmt, ##__VA_ARGS__)
277 #else
278 #define net_dbg_ratelimited(fmt, ...) \
279 do { \
280 if (0) \
281 no_printk(KERN_DEBUG pr_fmt(fmt), ##__VA_ARGS__); \
282 } while (0)
283 #endif
284
285 #define net_get_random_once(buf, nbytes) \
286 get_random_once((buf), (nbytes))
287 #define net_get_random_once_wait(buf, nbytes) \
288 get_random_once_wait((buf), (nbytes))
289
290 /*
291 * E.g. XFS meta- & log-data is in slab pages, or bcache meta
292 * data pages, or other high order pages allocated by
293 * __get_free_pages() without __GFP_COMP, which have a page_count
294 * of 0 and/or have PageSlab() set. We cannot use send_page for
295 * those, as that does get_page(); put_page(); and would cause
296 * either a VM_BUG directly, or __page_cache_release a page that
297 * would actually still be referenced by someone, leading to some
298 * obscure delayed Oops somewhere else.
299 */
sendpage_ok(struct page * page)300 static inline bool sendpage_ok(struct page *page)
301 {
302 return !PageSlab(page) && page_count(page) >= 1;
303 }
304
305 int kernel_sendmsg(struct socket *sock, struct msghdr *msg, struct kvec *vec,
306 size_t num, size_t len);
307 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
308 struct kvec *vec, size_t num, size_t len);
309 int kernel_recvmsg(struct socket *sock, struct msghdr *msg, struct kvec *vec,
310 size_t num, size_t len, int flags);
311
312 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen);
313 int kernel_listen(struct socket *sock, int backlog);
314 int kernel_accept(struct socket *sock, struct socket **newsock, int flags);
315 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
316 int flags);
317 int kernel_getsockname(struct socket *sock, struct sockaddr *addr);
318 int kernel_getpeername(struct socket *sock, struct sockaddr *addr);
319 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
320 size_t size, int flags);
321 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
322 size_t size, int flags);
323 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how);
324
325 /* Routine returns the IP overhead imposed by a (caller-protected) socket. */
326 u32 kernel_sock_ip_overhead(struct sock *sk);
327
328 #define MODULE_ALIAS_NETPROTO(proto) \
329 MODULE_ALIAS("net-pf-" __stringify(proto))
330
331 #define MODULE_ALIAS_NET_PF_PROTO(pf, proto) \
332 MODULE_ALIAS("net-pf-" __stringify(pf) "-proto-" __stringify(proto))
333
334 #define MODULE_ALIAS_NET_PF_PROTO_TYPE(pf, proto, type) \
335 MODULE_ALIAS("net-pf-" __stringify(pf) "-proto-" __stringify(proto) \
336 "-type-" __stringify(type))
337
338 #define MODULE_ALIAS_NET_PF_PROTO_NAME(pf, proto, name) \
339 MODULE_ALIAS("net-pf-" __stringify(pf) "-proto-" __stringify(proto) \
340 name)
341 #endif /* _LINUX_NET_H */
342