net.h

  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
 29struct poll_table_struct;
 30struct pipe_inode_info;
 31struct inode;
 32struct file;
 33struct 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 */
 39enum socket_flags {
 40	SOCKWQ_ASYNC_NOSPACE,
 41	SOCKWQ_ASYNC_WAITDATA,
 42	SOCK_NOSPACE,
 43	SOCK_SUPPORT_ZC,
 44	SOCK_CUSTOM_SOCKOPT,
 45};
 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 */
 63enum 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#endif /* ARCH_HAS_SOCKET_TYPES */
 73
 74#define SOCK_MAX (SOCK_PACKET + 1)
 75/* Mask which covers at least up to SOCK_MASK-1.  The
 76 * remaining bits are used as flags. */
 77#define SOCK_TYPE_MASK 0xf
 78
 79/* Flags for socket, socketpair, accept4 */
 80#define SOCK_CLOEXEC	O_CLOEXEC
 81#ifndef SOCK_NONBLOCK
 82#define SOCK_NONBLOCK	O_NONBLOCK
 83#endif
 84#define SOCK_COREDUMP	O_NOCTTY
 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 */
 92enum sock_shutdown_cmd {
 93	SHUT_RD,
 94	SHUT_WR,
 95	SHUT_RDWR,
 96};
 97
 98struct 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 */
116struct 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; /* Might change with IPV6_ADDRFORM or MPTCP. */
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 */
139typedef 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
149struct vm_area_struct;
150struct page;
151struct msghdr;
152struct module;
153struct sk_buff;
154struct proto_accept_arg;
155typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *,
156			       unsigned int, size_t);
157typedef int (*skb_read_actor_t)(struct sock *, struct sk_buff *);
158
159
160struct proto_ops {
161	int		family;
162	struct module	*owner;
163	int		(*release)   (struct socket *sock);
164	int		(*bind)	     (struct socket *sock,
165				      struct sockaddr_unsized *myaddr,
166				      int sockaddr_len);
167	int		(*connect)   (struct socket *sock,
168				      struct sockaddr_unsized *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,
174				      struct proto_accept_arg *arg);
175	int		(*getname)   (struct socket *sock,
176				      struct sockaddr *addr,
177				      int peer);
178	__poll_t	(*poll)	     (struct file *file, struct socket *sock,
179				      struct poll_table_struct *wait);
180	int		(*ioctl)     (struct socket *sock, unsigned int cmd,
181				      unsigned long arg);
182#ifdef CONFIG_COMPAT
183	int	 	(*compat_ioctl) (struct socket *sock, unsigned int cmd,
184				      unsigned long arg);
185#endif
186	int		(*gettstamp) (struct socket *sock, void __user *userstamp,
187				      bool timeval, bool time32);
188	int		(*listen)    (struct socket *sock, int len);
189	int		(*shutdown)  (struct socket *sock, int flags);
190	int		(*setsockopt)(struct socket *sock, int level,
191				      int optname, sockptr_t optval,
192				      unsigned int optlen);
193	int		(*getsockopt)(struct socket *sock, int level,
194				      int optname, char __user *optval, int __user *optlen);
195	void		(*show_fdinfo)(struct seq_file *m, struct socket *sock);
196	int		(*sendmsg)   (struct socket *sock, struct msghdr *m,
197				      size_t total_len);
198	/* Notes for implementing recvmsg:
199	 * ===============================
200	 * msg->msg_namelen should get updated by the recvmsg handlers
201	 * iff msg_name != NULL. It is by default 0 to prevent
202	 * returning uninitialized memory to user space.  The recvfrom
203	 * handlers can assume that msg.msg_name is either NULL or has
204	 * a minimum size of sizeof(struct sockaddr_storage).
205	 */
206	int		(*recvmsg)   (struct socket *sock, struct msghdr *m,
207				      size_t total_len, int flags);
208	int		(*mmap)	     (struct file *file, struct socket *sock,
209				      struct vm_area_struct * vma);
210	ssize_t 	(*splice_read)(struct socket *sock,  loff_t *ppos,
211				       struct pipe_inode_info *pipe, size_t len, unsigned int flags);
212	void		(*splice_eof)(struct socket *sock);
213	int		(*set_peek_off)(struct sock *sk, int val);
214	int		(*peek_len)(struct socket *sock);
215
216	/* The following functions are called internally by kernel with
217	 * sock lock already held.
218	 */
219	int		(*read_sock)(struct sock *sk, read_descriptor_t *desc,
220				     sk_read_actor_t recv_actor);
221	/* This is different from read_sock(), it reads an entire skb at a time. */
222	int		(*read_skb)(struct sock *sk, skb_read_actor_t recv_actor);
223	int		(*sendmsg_locked)(struct sock *sk, struct msghdr *msg,
224					  size_t size);
225	int		(*set_rcvlowat)(struct sock *sk, int val);
226};
227
228#define DECLARE_SOCKADDR(type, dst, src)	\
229	type dst = ({ __sockaddr_check_size(sizeof(*dst)); (type) src; })
230
231struct net_proto_family {
232	int		family;
233	int		(*create)(struct net *net, struct socket *sock,
234				  int protocol, int kern);
235	struct module	*owner;
236};
237
238struct iovec;
239struct kvec;
240
241enum {
242	SOCK_WAKE_IO,
243	SOCK_WAKE_WAITD,
244	SOCK_WAKE_SPACE,
245	SOCK_WAKE_URG,
246};
247
248int sock_wake_async(struct socket_wq *sk_wq, int how, int band);
249int sock_register(const struct net_proto_family *fam);
250void sock_unregister(int family);
251bool sock_is_registered(int family);
252int __sock_create(struct net *net, int family, int type, int proto,
253		  struct socket **res, int kern);
254int sock_create(int family, int type, int proto, struct socket **res);
255int sock_create_kern(struct net *net, int family, int type, int proto, struct socket **res);
256int sock_create_lite(int family, int type, int proto, struct socket **res);
257struct socket *sock_alloc(void);
258void sock_release(struct socket *sock);
259int sock_sendmsg(struct socket *sock, struct msghdr *msg);
260int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags);
261struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname);
262struct socket *sockfd_lookup(int fd, int *err);
263struct socket *sock_from_file(struct file *file);
264#define		     sockfd_put(sock) fput(sock->file)
265int net_ratelimit(void);
266
267#define net_ratelimited_function(function, ...)			\
268do {								\
269	if (net_ratelimit())					\
270		function(__VA_ARGS__);				\
271} while (0)
272
273#define net_emerg_ratelimited(fmt, ...)				\
274	net_ratelimited_function(pr_emerg, fmt, ##__VA_ARGS__)
275#define net_alert_ratelimited(fmt, ...)				\
276	net_ratelimited_function(pr_alert, fmt, ##__VA_ARGS__)
277#define net_crit_ratelimited(fmt, ...)				\
278	net_ratelimited_function(pr_crit, fmt, ##__VA_ARGS__)
279#define net_err_ratelimited(fmt, ...)				\
280	net_ratelimited_function(pr_err, fmt, ##__VA_ARGS__)
281#define net_notice_ratelimited(fmt, ...)			\
282	net_ratelimited_function(pr_notice, fmt, ##__VA_ARGS__)
283#define net_warn_ratelimited(fmt, ...)				\
284	net_ratelimited_function(pr_warn, fmt, ##__VA_ARGS__)
285#define net_info_ratelimited(fmt, ...)				\
286	net_ratelimited_function(pr_info, fmt, ##__VA_ARGS__)
287#if defined(CONFIG_DYNAMIC_DEBUG) || \
288	(defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
289#define net_dbg_ratelimited(fmt, ...)					\
290do {									\
291	DEFINE_DYNAMIC_DEBUG_METADATA(descriptor, fmt);			\
292	if (DYNAMIC_DEBUG_BRANCH(descriptor) &&				\
293	    net_ratelimit())						\
294		__dynamic_pr_debug(&descriptor, pr_fmt(fmt),		\
295		                   ##__VA_ARGS__);			\
296} while (0)
297#elif defined(DEBUG)
298#define net_dbg_ratelimited(fmt, ...)				\
299	net_ratelimited_function(pr_debug, fmt, ##__VA_ARGS__)
300#else
301#define net_dbg_ratelimited(fmt, ...)				\
302	no_printk(KERN_DEBUG pr_fmt(fmt), ##__VA_ARGS__)
303#endif
304
305#define net_get_random_once(buf, nbytes)			\
306	get_random_once((buf), (nbytes))
307
308/*
309 * E.g. XFS meta- & log-data is in slab pages, or bcache meta
310 * data pages, or other high order pages allocated by
311 * __get_free_pages() without __GFP_COMP, which have a page_count
312 * of 0 and/or have PageSlab() set. We cannot use send_page for
313 * those, as that does get_page(); put_page(); and would cause
314 * either a VM_BUG directly, or __page_cache_release a page that
315 * would actually still be referenced by someone, leading to some
316 * obscure delayed Oops somewhere else.
317 */
318static inline bool sendpage_ok(struct page *page)
319{
320	return !PageSlab(page) && page_count(page) >= 1;
321}
322
323/*
324 * Check sendpage_ok on contiguous pages.
325 */
326static inline bool sendpages_ok(struct page *page, size_t len, size_t offset)
327{
328	struct page *p = page + (offset >> PAGE_SHIFT);
329	size_t count = 0;
330
331	while (count < len) {
332		if (!sendpage_ok(p))
333			return false;
334
335		p++;
336		count += PAGE_SIZE;
337	}
338
339	return true;
340}
341
342int kernel_sendmsg(struct socket *sock, struct msghdr *msg, struct kvec *vec,
343		   size_t num, size_t len);
344int kernel_recvmsg(struct socket *sock, struct msghdr *msg, struct kvec *vec,
345		   size_t num, size_t len, int flags);
346
347int kernel_bind(struct socket *sock, struct sockaddr_unsized *addr, int addrlen);
348int kernel_listen(struct socket *sock, int backlog);
349int kernel_accept(struct socket *sock, struct socket **newsock, int flags);
350int kernel_connect(struct socket *sock, struct sockaddr_unsized *addr, int addrlen,
351		   int flags);
352int kernel_getsockname(struct socket *sock, struct sockaddr *addr);
353int kernel_getpeername(struct socket *sock, struct sockaddr *addr);
354int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how);
355
356/* Routine returns the IP overhead imposed by a (caller-protected) socket. */
357u32 kernel_sock_ip_overhead(struct sock *sk);
358
359#define MODULE_ALIAS_NETPROTO(proto) \
360	MODULE_ALIAS("net-pf-" __stringify(proto))
361
362#define MODULE_ALIAS_NET_PF_PROTO(pf, proto) \
363	MODULE_ALIAS("net-pf-" __stringify(pf) "-proto-" __stringify(proto))
364
365#define MODULE_ALIAS_NET_PF_PROTO_TYPE(pf, proto, type) \
366	MODULE_ALIAS("net-pf-" __stringify(pf) "-proto-" __stringify(proto) \
367		     "-type-" __stringify(type))
368
369#define MODULE_ALIAS_NET_PF_PROTO_NAME(pf, proto, name) \
370	MODULE_ALIAS("net-pf-" __stringify(pf) "-proto-" __stringify(proto) \
371		     name)
372#endif	/* _LINUX_NET_H */