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 #include <linux/uio.h> 27 28 #include <uapi/linux/net.h> 29 30 /** 31 * struct sockopt - socket option value container 32 * @iter_in: iov_iter for reading optval with the content from the caller. 33 * Use copy_from_iter() given this iov direction is ITER_SOURCE 34 * @iter_out: iov_iter for protocols to update optval data to userspace 35 * Use _copy_to_iter() given iov direction is ITER_DEST 36 * @optlen: serves as both input (buffer size) and output (returned data size). 37 * 38 * Type-safe wrapper for socket option data that works with both 39 * user and kernel buffers. 40 * 41 * The optlen field allows callbacks to return a specific length value 42 * independent of the bytes written via copy_to_iter(). 43 */ 44 typedef struct sockopt { 45 struct iov_iter iter_in; 46 struct iov_iter iter_out; 47 int optlen; 48 } sockopt_t; 49 50 int sockptr_to_sockopt(sockopt_t *opt, sockptr_t optval, sockptr_t optlen, 51 struct kvec *kvec); 52 53 /* 54 * Initialize a user-backed sockopt_t from the (optval, optlen) __user pair of 55 * a getsockopt() callback. Used by transitional __user getsockopt wrappers 56 * while the proto-layer callbacks are converted to take a sockopt_t; the 57 * caller writes opt->optlen back to the user optlen after the callback. 58 */ 59 static inline int sockopt_init_user(sockopt_t *opt, char __user *optval, 60 int __user *optlen) 61 { 62 int len; 63 64 if (get_user(len, optlen)) 65 return -EFAULT; 66 if (len < 0) 67 return -EINVAL; 68 69 iov_iter_ubuf(&opt->iter_out, ITER_DEST, optval, len); 70 iov_iter_ubuf(&opt->iter_in, ITER_SOURCE, optval, len); 71 opt->optlen = len; 72 73 return 0; 74 } 75 76 /* 77 * Grow optval to @size, for the options whose reply is sized by a count the 78 * caller left in optval rather than by optlen. Those write past optlen today 79 * and userspace relies on it. 80 * 81 * Call it before writing through opt->iter_out: it re-anchors the iterator at 82 * the head of optval. Only a user buffer can be longer than the optlen the 83 * caller declared, so a kernel-backed optval is refused with -EINVAL. 84 */ 85 static inline int sockopt_expand_out(sockopt_t *opt, size_t size) 86 { 87 if (size <= (size_t)opt->optlen) 88 return 0; 89 90 if (size > INT_MAX) 91 return -EINVAL; 92 93 /* Re-anchoring reads iter_out.ubuf, so the iterator has to be a user 94 * buffer that nothing has written through yet. 95 */ 96 if (WARN_ON_ONCE(!iter_is_ubuf(&opt->iter_out) || 97 iov_iter_count(&opt->iter_out) != (size_t)opt->optlen)) 98 return -EINVAL; 99 100 iov_iter_ubuf(&opt->iter_out, ITER_DEST, opt->iter_out.ubuf, size); 101 102 return 0; 103 } 104 105 struct poll_table_struct; 106 struct pipe_inode_info; 107 struct inode; 108 struct file; 109 struct net; 110 111 /* Historically, SOCKWQ_ASYNC_NOSPACE & SOCKWQ_ASYNC_WAITDATA were located 112 * in sock->flags, but moved into sk->sk_wq->flags to be RCU protected. 113 * Eventually all flags will be in sk->sk_wq->flags. 114 */ 115 enum socket_flags { 116 SOCKWQ_ASYNC_NOSPACE, 117 SOCKWQ_ASYNC_WAITDATA, 118 SOCK_NOSPACE, 119 SOCK_SUPPORT_ZC, 120 SOCK_CUSTOM_SOCKOPT, 121 }; 122 123 #ifndef ARCH_HAS_SOCKET_TYPES 124 /** 125 * enum sock_type - Socket types 126 * @SOCK_STREAM: stream (connection) socket 127 * @SOCK_DGRAM: datagram (conn.less) socket 128 * @SOCK_RAW: raw socket 129 * @SOCK_RDM: reliably-delivered message 130 * @SOCK_SEQPACKET: sequential packet socket 131 * @SOCK_DCCP: Datagram Congestion Control Protocol socket 132 * @SOCK_PACKET: linux specific way of getting packets at the dev level. 133 * For writing rarp and other similar things on the user level. 134 * 135 * When adding some new socket type please 136 * grep ARCH_HAS_SOCKET_TYPE include/asm-* /socket.h, at least MIPS 137 * overrides this enum for binary compat reasons. 138 */ 139 enum sock_type { 140 SOCK_STREAM = 1, 141 SOCK_DGRAM = 2, 142 SOCK_RAW = 3, 143 SOCK_RDM = 4, 144 SOCK_SEQPACKET = 5, 145 SOCK_DCCP = 6, 146 SOCK_PACKET = 10, 147 }; 148 #endif /* ARCH_HAS_SOCKET_TYPES */ 149 150 #define SOCK_MAX (SOCK_PACKET + 1) 151 /* Mask which covers at least up to SOCK_MASK-1. The 152 * remaining bits are used as flags. */ 153 #define SOCK_TYPE_MASK 0xf 154 155 /* Flags for socket, socketpair, accept4 */ 156 #define SOCK_CLOEXEC O_CLOEXEC 157 #ifndef SOCK_NONBLOCK 158 #define SOCK_NONBLOCK O_NONBLOCK 159 #endif 160 #define SOCK_COREDUMP O_NOCTTY 161 162 /** 163 * enum sock_shutdown_cmd - Shutdown types 164 * @SHUT_RD: shutdown receptions 165 * @SHUT_WR: shutdown transmissions 166 * @SHUT_RDWR: shutdown receptions/transmissions 167 */ 168 enum sock_shutdown_cmd { 169 SHUT_RD, 170 SHUT_WR, 171 SHUT_RDWR, 172 }; 173 174 struct socket_wq { 175 /* Note: wait MUST be first field of socket_wq */ 176 wait_queue_head_t wait; 177 struct fasync_struct *fasync_list; 178 unsigned long flags; /* %SOCKWQ_ASYNC_NOSPACE, etc */ 179 struct rcu_head rcu; 180 } ____cacheline_aligned_in_smp; 181 182 /** 183 * struct socket - general BSD socket 184 * @state: socket state (%SS_CONNECTED, etc) 185 * @type: socket type (%SOCK_STREAM, etc) 186 * @flags: socket flags (%SOCK_NOSPACE, etc) 187 * @ops: protocol specific socket operations 188 * @file: File back pointer for gc 189 * @sk: internal networking protocol agnostic socket representation 190 * @wq: wait queue for several uses 191 */ 192 struct socket { 193 socket_state state; 194 195 short type; 196 197 unsigned long flags; 198 199 struct file *file; 200 struct sock *sk; 201 const struct proto_ops *ops; /* Might change with MPTCP. */ 202 203 struct socket_wq wq; 204 }; 205 206 /* 207 * "descriptor" for what we're up to with a read. 208 * This allows us to use the same read code yet 209 * have multiple different users of the data that 210 * we read from a file. 211 * 212 * The simplest case just copies the data to user 213 * mode. 214 */ 215 typedef struct { 216 size_t written; 217 size_t count; 218 union { 219 char __user *buf; 220 void *data; 221 } arg; 222 int error; 223 } read_descriptor_t; 224 225 struct vm_area_struct; 226 struct page; 227 struct msghdr; 228 struct module; 229 struct sk_buff; 230 struct proto_accept_arg; 231 typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *, 232 unsigned int, size_t); 233 typedef int (*skb_read_actor_t)(struct sock *, struct sk_buff *); 234 235 236 struct proto_ops { 237 int family; 238 struct module *owner; 239 int (*release) (struct socket *sock); 240 int (*bind) (struct socket *sock, 241 struct sockaddr_unsized *myaddr, 242 int sockaddr_len); 243 int (*connect) (struct socket *sock, 244 struct sockaddr_unsized *vaddr, 245 int sockaddr_len, int flags); 246 int (*socketpair)(struct socket *sock1, 247 struct socket *sock2); 248 int (*accept) (struct socket *sock, 249 struct socket *newsock, 250 struct proto_accept_arg *arg); 251 int (*getname) (struct socket *sock, 252 struct sockaddr *addr, 253 int peer); 254 __poll_t (*poll) (struct file *file, struct socket *sock, 255 struct poll_table_struct *wait); 256 int (*ioctl) (struct socket *sock, unsigned int cmd, 257 unsigned long arg); 258 #ifdef CONFIG_COMPAT 259 int (*compat_ioctl) (struct socket *sock, unsigned int cmd, 260 unsigned long arg); 261 #endif 262 int (*gettstamp) (struct socket *sock, void __user *userstamp, 263 bool timeval, bool time32); 264 int (*listen) (struct socket *sock, int len); 265 int (*shutdown) (struct socket *sock, int flags); 266 int (*setsockopt)(struct socket *sock, int level, 267 int optname, sockptr_t optval, 268 unsigned int optlen); 269 int (*getsockopt)(struct socket *sock, int level, 270 int optname, char __user *optval, int __user *optlen); 271 int (*getsockopt_iter)(struct socket *sock, int level, 272 int optname, sockopt_t *opt); 273 void (*show_fdinfo)(struct seq_file *m, struct socket *sock); 274 int (*sendmsg) (struct socket *sock, struct msghdr *m, 275 size_t total_len); 276 /* Notes for implementing recvmsg: 277 * =============================== 278 * msg->msg_namelen should get updated by the recvmsg handlers 279 * iff msg_name != NULL. It is by default 0 to prevent 280 * returning uninitialized memory to user space. The recvfrom 281 * handlers can assume that msg.msg_name is either NULL or has 282 * a minimum size of sizeof(struct sockaddr_storage). 283 */ 284 int (*recvmsg) (struct socket *sock, struct msghdr *m, 285 size_t total_len, int flags); 286 int (*mmap) (struct file *file, struct socket *sock, 287 struct vm_area_struct * vma); 288 ssize_t (*splice_read)(struct socket *sock, loff_t *ppos, 289 struct pipe_inode_info *pipe, size_t len, unsigned int flags); 290 void (*splice_eof)(struct socket *sock); 291 int (*set_peek_off)(struct sock *sk, int val); 292 int (*peek_len)(struct socket *sock); 293 294 /* The following functions are called internally by kernel with 295 * sock lock already held. 296 */ 297 int (*read_sock)(struct sock *sk, read_descriptor_t *desc, 298 sk_read_actor_t recv_actor); 299 /* This is different from read_sock(), it reads an entire skb at a time. */ 300 int (*read_skb)(struct sock *sk, skb_read_actor_t recv_actor); 301 int (*sendmsg_locked)(struct sock *sk, struct msghdr *msg, 302 size_t size); 303 int (*set_rcvlowat)(struct sock *sk, int val); 304 void (*set_rcvbuf)(struct sock *sk, int val); 305 }; 306 307 #define DECLARE_SOCKADDR(type, dst, src) \ 308 type dst = ({ __sockaddr_check_size(sizeof(*dst)); (type) src; }) 309 310 struct net_proto_family { 311 int family; 312 int (*create)(struct net *net, struct socket *sock, 313 int protocol, int kern); 314 struct module *owner; 315 }; 316 317 struct iovec; 318 struct kvec; 319 320 enum { 321 SOCK_WAKE_IO, 322 SOCK_WAKE_WAITD, 323 SOCK_WAKE_SPACE, 324 SOCK_WAKE_URG, 325 }; 326 327 int sock_wake_async(struct socket_wq *sk_wq, int how, int band); 328 int sock_register(const struct net_proto_family *fam); 329 void sock_unregister(int family); 330 bool sock_is_registered(int family); 331 int __sock_create(struct net *net, int family, int type, int proto, 332 struct socket **res, int kern); 333 int sock_create(int family, int type, int proto, struct socket **res); 334 int sock_create_kern(struct net *net, int family, int type, int proto, struct socket **res); 335 int sock_create_lite(int family, int type, int proto, struct socket **res); 336 struct socket *sock_alloc(void); 337 void sock_release(struct socket *sock); 338 int sock_sendmsg(struct socket *sock, struct msghdr *msg); 339 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags); 340 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname); 341 struct socket *sockfd_lookup(int fd, int *err); 342 struct socket *sock_from_file(struct file *file); 343 int sock_read_xattr(struct socket *sock, const char *name, void *value, size_t size); 344 #define sockfd_put(sock) fput(sock->file) 345 int net_ratelimit(void); 346 347 #define net_ratelimited_function(function, ...) \ 348 do { \ 349 if (net_ratelimit()) \ 350 function(__VA_ARGS__); \ 351 } while (0) 352 353 #define net_emerg_ratelimited(fmt, ...) \ 354 net_ratelimited_function(pr_emerg, fmt, ##__VA_ARGS__) 355 #define net_alert_ratelimited(fmt, ...) \ 356 net_ratelimited_function(pr_alert, fmt, ##__VA_ARGS__) 357 #define net_crit_ratelimited(fmt, ...) \ 358 net_ratelimited_function(pr_crit, fmt, ##__VA_ARGS__) 359 #define net_err_ratelimited(fmt, ...) \ 360 net_ratelimited_function(pr_err, fmt, ##__VA_ARGS__) 361 #define net_notice_ratelimited(fmt, ...) \ 362 net_ratelimited_function(pr_notice, fmt, ##__VA_ARGS__) 363 #define net_warn_ratelimited(fmt, ...) \ 364 net_ratelimited_function(pr_warn, fmt, ##__VA_ARGS__) 365 #define net_info_ratelimited(fmt, ...) \ 366 net_ratelimited_function(pr_info, fmt, ##__VA_ARGS__) 367 #if defined(CONFIG_DYNAMIC_DEBUG) || \ 368 (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE)) 369 #define net_dbg_ratelimited(fmt, ...) \ 370 do { \ 371 DEFINE_DYNAMIC_DEBUG_METADATA(descriptor, fmt); \ 372 if (DYNAMIC_DEBUG_BRANCH(descriptor) && \ 373 net_ratelimit()) \ 374 __dynamic_pr_debug(&descriptor, pr_fmt(fmt), \ 375 ##__VA_ARGS__); \ 376 } while (0) 377 #elif defined(DEBUG) 378 #define net_dbg_ratelimited(fmt, ...) \ 379 net_ratelimited_function(pr_debug, fmt, ##__VA_ARGS__) 380 #else 381 #define net_dbg_ratelimited(fmt, ...) \ 382 no_printk(KERN_DEBUG pr_fmt(fmt), ##__VA_ARGS__) 383 #endif 384 385 #define net_get_random_once(buf, nbytes) \ 386 get_random_once((buf), (nbytes)) 387 #define net_get_random_sleepable_once(buf, nbytes) \ 388 get_random_sleepable_once((buf), (nbytes)) 389 390 /* 391 * E.g. XFS meta- & log-data is in slab pages, or bcache meta 392 * data pages, or other high order pages allocated by 393 * __get_free_pages() without __GFP_COMP, which have a page_count 394 * of 0 and/or have PageSlab() set. We cannot use send_page for 395 * those, as that does get_page(); put_page(); and would cause 396 * either a VM_BUG directly, or __page_cache_release a page that 397 * would actually still be referenced by someone, leading to some 398 * obscure delayed Oops somewhere else. 399 */ 400 static inline bool sendpage_ok(struct page *page) 401 { 402 return !PageSlab(page) && page_count(page) >= 1; 403 } 404 405 /* 406 * Check sendpage_ok on contiguous pages. 407 */ 408 static inline bool sendpages_ok(struct page *page, size_t len, size_t offset) 409 { 410 struct page *p = page + (offset >> PAGE_SHIFT); 411 size_t count = 0; 412 413 while (count < len) { 414 if (!sendpage_ok(p)) 415 return false; 416 417 p++; 418 count += PAGE_SIZE; 419 } 420 421 return true; 422 } 423 424 int kernel_sendmsg(struct socket *sock, struct msghdr *msg, struct kvec *vec, 425 size_t num, size_t len); 426 int kernel_recvmsg(struct socket *sock, struct msghdr *msg, struct kvec *vec, 427 size_t num, size_t len, int flags); 428 429 int kernel_bind(struct socket *sock, struct sockaddr_unsized *addr, int addrlen); 430 int kernel_listen(struct socket *sock, int backlog); 431 int kernel_accept(struct socket *sock, struct socket **newsock, int flags); 432 int kernel_connect(struct socket *sock, struct sockaddr_unsized *addr, int addrlen, 433 int flags); 434 int kernel_getsockname(struct socket *sock, struct sockaddr *addr); 435 int kernel_getpeername(struct socket *sock, struct sockaddr *addr); 436 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how); 437 438 /* Routine returns the IP overhead imposed by a (caller-protected) socket. */ 439 u32 kernel_sock_ip_overhead(struct sock *sk); 440 441 #define MODULE_ALIAS_NETPROTO(proto) \ 442 MODULE_ALIAS("net-pf-" __stringify(proto)) 443 444 #define MODULE_ALIAS_NET_PF_PROTO(pf, proto) \ 445 MODULE_ALIAS("net-pf-" __stringify(pf) "-proto-" __stringify(proto)) 446 447 #define MODULE_ALIAS_NET_PF_PROTO_TYPE(pf, proto, type) \ 448 MODULE_ALIAS("net-pf-" __stringify(pf) "-proto-" __stringify(proto) \ 449 "-type-" __stringify(type)) 450 451 #define MODULE_ALIAS_NET_PF_PROTO_NAME(pf, proto, name) \ 452 MODULE_ALIAS("net-pf-" __stringify(pf) "-proto-" __stringify(proto) \ 453 name) 454 #endif /* _LINUX_NET_H */ 455