xref: /linux-6.15/include/linux/net.h (revision 2bfc6685)
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, <[email protected]>
11  *		Ross Biro
12  *		Fred N. van Kempen, <[email protected]>
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 #define SOCK_SUPPORT_ZC		5
45 #define SOCK_CUSTOM_SOCKOPT	6
46 
47 #ifndef ARCH_HAS_SOCKET_TYPES
48 /**
49  * enum sock_type - Socket types
50  * @SOCK_STREAM: stream (connection) socket
51  * @SOCK_DGRAM: datagram (conn.less) socket
52  * @SOCK_RAW: raw socket
53  * @SOCK_RDM: reliably-delivered message
54  * @SOCK_SEQPACKET: sequential packet socket
55  * @SOCK_DCCP: Datagram Congestion Control Protocol socket
56  * @SOCK_PACKET: linux specific way of getting packets at the dev level.
57  *		  For writing rarp and other similar things on the user level.
58  *
59  * When adding some new socket type please
60  * grep ARCH_HAS_SOCKET_TYPE include/asm-* /socket.h, at least MIPS
61  * overrides this enum for binary compat reasons.
62  */
63 enum sock_type {
64 	SOCK_STREAM	= 1,
65 	SOCK_DGRAM	= 2,
66 	SOCK_RAW	= 3,
67 	SOCK_RDM	= 4,
68 	SOCK_SEQPACKET	= 5,
69 	SOCK_DCCP	= 6,
70 	SOCK_PACKET	= 10,
71 };
72 
73 #define SOCK_MAX (SOCK_PACKET + 1)
74 /* Mask which covers at least up to SOCK_MASK-1.  The
75  * remaining bits are used as flags. */
76 #define SOCK_TYPE_MASK 0xf
77 
78 /* Flags for socket, socketpair, accept4 */
79 #define SOCK_CLOEXEC	O_CLOEXEC
80 #ifndef SOCK_NONBLOCK
81 #define SOCK_NONBLOCK	O_NONBLOCK
82 #endif
83 
84 #endif /* ARCH_HAS_SOCKET_TYPES */
85 
86 /**
87  * enum sock_shutdown_cmd - Shutdown types
88  * @SHUT_RD: shutdown receptions
89  * @SHUT_WR: shutdown transmissions
90  * @SHUT_RDWR: shutdown receptions/transmissions
91  */
92 enum sock_shutdown_cmd {
93 	SHUT_RD,
94 	SHUT_WR,
95 	SHUT_RDWR,
96 };
97 
98 struct socket_wq {
99 	/* Note: wait MUST be first field of socket_wq */
100 	wait_queue_head_t	wait;
101 	struct fasync_struct	*fasync_list;
102 	unsigned long		flags; /* %SOCKWQ_ASYNC_NOSPACE, etc */
103 	struct rcu_head		rcu;
104 } ____cacheline_aligned_in_smp;
105 
106 /**
107  *  struct socket - general BSD socket
108  *  @state: socket state (%SS_CONNECTED, etc)
109  *  @type: socket type (%SOCK_STREAM, etc)
110  *  @flags: socket flags (%SOCK_NOSPACE, etc)
111  *  @ops: protocol specific socket operations
112  *  @file: File back pointer for gc
113  *  @sk: internal networking protocol agnostic socket representation
114  *  @wq: wait queue for several uses
115  */
116 struct socket {
117 	socket_state		state;
118 
119 	short			type;
120 
121 	unsigned long		flags;
122 
123 	struct file		*file;
124 	struct sock		*sk;
125 	const struct proto_ops	*ops;
126 
127 	struct socket_wq	wq;
128 };
129 
130 /*
131  * "descriptor" for what we're up to with a read.
132  * This allows us to use the same read code yet
133  * have multiple different users of the data that
134  * we read from a file.
135  *
136  * The simplest case just copies the data to user
137  * mode.
138  */
139 typedef struct {
140 	size_t written;
141 	size_t count;
142 	union {
143 		char __user *buf;
144 		void *data;
145 	} arg;
146 	int error;
147 } read_descriptor_t;
148 
149 struct vm_area_struct;
150 struct page;
151 struct sockaddr;
152 struct msghdr;
153 struct module;
154 struct sk_buff;
155 typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *,
156 			       unsigned int, size_t);
157 typedef int (*skb_read_actor_t)(struct sock *, struct sk_buff *);
158 
159 
160 struct proto_ops {
161 	int		family;
162 	struct module	*owner;
163 	int		(*release)   (struct socket *sock);
164 	int		(*bind)	     (struct socket *sock,
165 				      struct sockaddr *myaddr,
166 				      int sockaddr_len);
167 	int		(*connect)   (struct socket *sock,
168 				      struct sockaddr *vaddr,
169 				      int sockaddr_len, int flags);
170 	int		(*socketpair)(struct socket *sock1,
171 				      struct socket *sock2);
172 	int		(*accept)    (struct socket *sock,
173 				      struct socket *newsock, int flags, bool kern);
174 	int		(*getname)   (struct socket *sock,
175 				      struct sockaddr *addr,
176 				      int peer);
177 	__poll_t	(*poll)	     (struct file *file, struct socket *sock,
178 				      struct poll_table_struct *wait);
179 	int		(*ioctl)     (struct socket *sock, unsigned int cmd,
180 				      unsigned long arg);
181 #ifdef CONFIG_COMPAT
182 	int	 	(*compat_ioctl) (struct socket *sock, unsigned int cmd,
183 				      unsigned long arg);
184 #endif
185 	int		(*gettstamp) (struct socket *sock, void __user *userstamp,
186 				      bool timeval, bool time32);
187 	int		(*listen)    (struct socket *sock, int len);
188 	int		(*shutdown)  (struct socket *sock, int flags);
189 	int		(*setsockopt)(struct socket *sock, int level,
190 				      int optname, sockptr_t optval,
191 				      unsigned int optlen);
192 	int		(*getsockopt)(struct socket *sock, int level,
193 				      int optname, char __user *optval, int __user *optlen);
194 	void		(*show_fdinfo)(struct seq_file *m, struct socket *sock);
195 	int		(*sendmsg)   (struct socket *sock, struct msghdr *m,
196 				      size_t total_len);
197 	/* Notes for implementing recvmsg:
198 	 * ===============================
199 	 * msg->msg_namelen should get updated by the recvmsg handlers
200 	 * iff msg_name != NULL. It is by default 0 to prevent
201 	 * returning uninitialized memory to user space.  The recvfrom
202 	 * handlers can assume that msg.msg_name is either NULL or has
203 	 * a minimum size of sizeof(struct sockaddr_storage).
204 	 */
205 	int		(*recvmsg)   (struct socket *sock, struct msghdr *m,
206 				      size_t total_len, int flags);
207 	int		(*mmap)	     (struct file *file, struct socket *sock,
208 				      struct vm_area_struct * vma);
209 	ssize_t		(*sendpage)  (struct socket *sock, struct page *page,
210 				      int offset, size_t size, int flags);
211 	ssize_t 	(*splice_read)(struct socket *sock,  loff_t *ppos,
212 				       struct pipe_inode_info *pipe, size_t len, unsigned int flags);
213 	void		(*splice_eof)(struct socket *sock);
214 	int		(*set_peek_off)(struct sock *sk, int val);
215 	int		(*peek_len)(struct socket *sock);
216 
217 	/* The following functions are called internally by kernel with
218 	 * sock lock already held.
219 	 */
220 	int		(*read_sock)(struct sock *sk, read_descriptor_t *desc,
221 				     sk_read_actor_t recv_actor);
222 	/* This is different from read_sock(), it reads an entire skb at a time. */
223 	int		(*read_skb)(struct sock *sk, skb_read_actor_t recv_actor);
224 	int		(*sendpage_locked)(struct sock *sk, struct page *page,
225 					   int offset, size_t size, int flags);
226 	int		(*sendmsg_locked)(struct sock *sk, struct msghdr *msg,
227 					  size_t size);
228 	int		(*set_rcvlowat)(struct sock *sk, int val);
229 };
230 
231 #define DECLARE_SOCKADDR(type, dst, src)	\
232 	type dst = ({ __sockaddr_check_size(sizeof(*dst)); (type) src; })
233 
234 struct net_proto_family {
235 	int		family;
236 	int		(*create)(struct net *net, struct socket *sock,
237 				  int protocol, int kern);
238 	struct module	*owner;
239 };
240 
241 struct iovec;
242 struct kvec;
243 
244 enum {
245 	SOCK_WAKE_IO,
246 	SOCK_WAKE_WAITD,
247 	SOCK_WAKE_SPACE,
248 	SOCK_WAKE_URG,
249 };
250 
251 int sock_wake_async(struct socket_wq *sk_wq, int how, int band);
252 int sock_register(const struct net_proto_family *fam);
253 void sock_unregister(int family);
254 bool sock_is_registered(int family);
255 int __sock_create(struct net *net, int family, int type, int proto,
256 		  struct socket **res, int kern);
257 int sock_create(int family, int type, int proto, struct socket **res);
258 int sock_create_kern(struct net *net, int family, int type, int proto, struct socket **res);
259 int sock_create_lite(int family, int type, int proto, struct socket **res);
260 struct socket *sock_alloc(void);
261 void sock_release(struct socket *sock);
262 int sock_sendmsg(struct socket *sock, struct msghdr *msg);
263 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags);
264 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname);
265 struct socket *sockfd_lookup(int fd, int *err);
266 struct socket *sock_from_file(struct file *file);
267 #define		     sockfd_put(sock) fput(sock->file)
268 int net_ratelimit(void);
269 
270 #define net_ratelimited_function(function, ...)			\
271 do {								\
272 	if (net_ratelimit())					\
273 		function(__VA_ARGS__);				\
274 } while (0)
275 
276 #define net_emerg_ratelimited(fmt, ...)				\
277 	net_ratelimited_function(pr_emerg, fmt, ##__VA_ARGS__)
278 #define net_alert_ratelimited(fmt, ...)				\
279 	net_ratelimited_function(pr_alert, fmt, ##__VA_ARGS__)
280 #define net_crit_ratelimited(fmt, ...)				\
281 	net_ratelimited_function(pr_crit, fmt, ##__VA_ARGS__)
282 #define net_err_ratelimited(fmt, ...)				\
283 	net_ratelimited_function(pr_err, fmt, ##__VA_ARGS__)
284 #define net_notice_ratelimited(fmt, ...)			\
285 	net_ratelimited_function(pr_notice, fmt, ##__VA_ARGS__)
286 #define net_warn_ratelimited(fmt, ...)				\
287 	net_ratelimited_function(pr_warn, fmt, ##__VA_ARGS__)
288 #define net_info_ratelimited(fmt, ...)				\
289 	net_ratelimited_function(pr_info, fmt, ##__VA_ARGS__)
290 #if defined(CONFIG_DYNAMIC_DEBUG) || \
291 	(defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
292 #define net_dbg_ratelimited(fmt, ...)					\
293 do {									\
294 	DEFINE_DYNAMIC_DEBUG_METADATA(descriptor, fmt);			\
295 	if (DYNAMIC_DEBUG_BRANCH(descriptor) &&				\
296 	    net_ratelimit())						\
297 		__dynamic_pr_debug(&descriptor, pr_fmt(fmt),		\
298 		                   ##__VA_ARGS__);			\
299 } while (0)
300 #elif defined(DEBUG)
301 #define net_dbg_ratelimited(fmt, ...)				\
302 	net_ratelimited_function(pr_debug, fmt, ##__VA_ARGS__)
303 #else
304 #define net_dbg_ratelimited(fmt, ...)				\
305 	do {							\
306 		if (0)						\
307 			no_printk(KERN_DEBUG pr_fmt(fmt), ##__VA_ARGS__); \
308 	} while (0)
309 #endif
310 
311 #define net_get_random_once(buf, nbytes)			\
312 	get_random_once((buf), (nbytes))
313 
314 /*
315  * E.g. XFS meta- & log-data is in slab pages, or bcache meta
316  * data pages, or other high order pages allocated by
317  * __get_free_pages() without __GFP_COMP, which have a page_count
318  * of 0 and/or have PageSlab() set. We cannot use send_page for
319  * those, as that does get_page(); put_page(); and would cause
320  * either a VM_BUG directly, or __page_cache_release a page that
321  * would actually still be referenced by someone, leading to some
322  * obscure delayed Oops somewhere else.
323  */
324 static inline bool sendpage_ok(struct page *page)
325 {
326 	return !PageSlab(page) && page_count(page) >= 1;
327 }
328 
329 int kernel_sendmsg(struct socket *sock, struct msghdr *msg, struct kvec *vec,
330 		   size_t num, size_t len);
331 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
332 			  struct kvec *vec, size_t num, size_t len);
333 int kernel_recvmsg(struct socket *sock, struct msghdr *msg, struct kvec *vec,
334 		   size_t num, size_t len, int flags);
335 
336 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen);
337 int kernel_listen(struct socket *sock, int backlog);
338 int kernel_accept(struct socket *sock, struct socket **newsock, int flags);
339 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
340 		   int flags);
341 int kernel_getsockname(struct socket *sock, struct sockaddr *addr);
342 int kernel_getpeername(struct socket *sock, struct sockaddr *addr);
343 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
344 		    size_t size, int flags);
345 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
346 			   size_t size, int flags);
347 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how);
348 
349 /* Routine returns the IP overhead imposed by a (caller-protected) socket. */
350 u32 kernel_sock_ip_overhead(struct sock *sk);
351 
352 #define MODULE_ALIAS_NETPROTO(proto) \
353 	MODULE_ALIAS("net-pf-" __stringify(proto))
354 
355 #define MODULE_ALIAS_NET_PF_PROTO(pf, proto) \
356 	MODULE_ALIAS("net-pf-" __stringify(pf) "-proto-" __stringify(proto))
357 
358 #define MODULE_ALIAS_NET_PF_PROTO_TYPE(pf, proto, type) \
359 	MODULE_ALIAS("net-pf-" __stringify(pf) "-proto-" __stringify(proto) \
360 		     "-type-" __stringify(type))
361 
362 #define MODULE_ALIAS_NET_PF_PROTO_NAME(pf, proto, name) \
363 	MODULE_ALIAS("net-pf-" __stringify(pf) "-proto-" __stringify(proto) \
364 		     name)
365 #endif	/* _LINUX_NET_H */
366