include/linux/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
  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 #define PROTO_CMSG_DATA_ONLY    0x0001
  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 struct vm_area_struct;
 131 struct page;
 132 struct sockaddr;
 133 struct msghdr;
 134 struct module;
 135 struct sk_buff;
 136 typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *,
 137                                unsigned int, size_t);
 138
 139 struct proto_ops {
 140         int             family;
 141         unsigned int    flags;
 142         struct module   *owner;
 143         int             (*release)   (struct socket *sock);
 144         int             (*bind)      (struct socket *sock,
 145                                       struct sockaddr *myaddr,
 146                                       int sockaddr_len);
 147         int             (*connect)   (struct socket *sock,
 148                                       struct sockaddr *vaddr,
 149                                       int sockaddr_len, int flags);
 150         int             (*socketpair)(struct socket *sock1,
 151                                       struct socket *sock2);
 152         int             (*accept)    (struct socket *sock,
 153                                       struct socket *newsock, int flags, bool kern);
 154         int             (*getname)   (struct socket *sock,
 155                                       struct sockaddr *addr,
 156                                       int peer);
 157         __poll_t        (*poll)      (struct file *file, struct socket *sock,
 158                                       struct poll_table_struct *wait);
 159         int             (*ioctl)     (struct socket *sock, unsigned int cmd,
 160                                       unsigned long arg);
 161 #ifdef CONFIG_COMPAT
 162         int             (*compat_ioctl) (struct socket *sock, unsigned int cmd,
 163                                       unsigned long arg);
 164 #endif
 165         int             (*gettstamp) (struct socket *sock, void __user *userstamp,
 166                                       bool timeval, bool time32);
 167         int             (*listen)    (struct socket *sock, int len);
 168         int             (*shutdown)  (struct socket *sock, int flags);
 169         int             (*setsockopt)(struct socket *sock, int level,
 170                                       int optname, sockptr_t optval,
 171                                       unsigned int optlen);
 172         int             (*getsockopt)(struct socket *sock, int level,
 173                                       int optname, char __user *optval, int __user *optlen);
 174         void            (*show_fdinfo)(struct seq_file *m, struct socket *sock);
 175         int             (*sendmsg)   (struct socket *sock, struct msghdr *m,
 176                                       size_t total_len);
 177         /* Notes for implementing recvmsg:
 178          * ===============================
 179          * msg->msg_namelen should get updated by the recvmsg handlers
 180          * iff msg_name != NULL. It is by default 0 to prevent
 181          * returning uninitialized memory to user space.  The recvfrom
 182          * handlers can assume that msg.msg_name is either NULL or has
 183          * a minimum size of sizeof(struct sockaddr_storage).
 184          */
 185         int             (*recvmsg)   (struct socket *sock, struct msghdr *m,
 186                                       size_t total_len, int flags);
 187         int             (*mmap)      (struct file *file, struct socket *sock,
 188                                       struct vm_area_struct * vma);
 189         ssize_t         (*sendpage)  (struct socket *sock, struct page *page,
 190                                       int offset, size_t size, int flags);
 191         ssize_t         (*splice_read)(struct socket *sock,  loff_t *ppos,
 192                                        struct pipe_inode_info *pipe, size_t len, unsigned int flags);
 193         int             (*set_peek_off)(struct sock *sk, int val);
 194         int             (*peek_len)(struct socket *sock);
 195
 196         /* The following functions are called internally by kernel with
 197          * sock lock already held.
 198          */
 199         int             (*read_sock)(struct sock *sk, read_descriptor_t *desc,
 200                                      sk_read_actor_t recv_actor);
 201         int             (*sendpage_locked)(struct sock *sk, struct page *page,
 202                                            int offset, size_t size, int flags);
 203         int             (*sendmsg_locked)(struct sock *sk, struct msghdr *msg,
 204                                           size_t size);
 205         int             (*set_rcvlowat)(struct sock *sk, int val);
 206 };
 207
 208 #define DECLARE_SOCKADDR(type, dst, src)        \
 209         type dst = ({ __sockaddr_check_size(sizeof(*dst)); (type) src; })
 210
 211 struct net_proto_family {
 212         int             family;
 213         int             (*create)(struct net *net, struct socket *sock,
 214                                   int protocol, int kern);
 215         struct module   *owner;
 216 };
 217
 218 struct iovec;
 219 struct kvec;
 220
 221 enum {
 222         SOCK_WAKE_IO,
 223         SOCK_WAKE_WAITD,
 224         SOCK_WAKE_SPACE,
 225         SOCK_WAKE_URG,
 226 };
 227
 228 int sock_wake_async(struct socket_wq *sk_wq, int how, int band);
 229 int sock_register(const struct net_proto_family *fam);
 230 void sock_unregister(int family);
 231 bool sock_is_registered(int family);
 232 int __sock_create(struct net *net, int family, int type, int proto,
 233                   struct socket **res, int kern);
 234 int sock_create(int family, int type, int proto, struct socket **res);
 235 int sock_create_kern(struct net *net, int family, int type, int proto, struct socket **res);
 236 int sock_create_lite(int family, int type, int proto, struct socket **res);
 237 struct socket *sock_alloc(void);
 238 void sock_release(struct socket *sock);
 239 int sock_sendmsg(struct socket *sock, struct msghdr *msg);
 240 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags);
 241 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname);
 242 struct socket *sockfd_lookup(int fd, int *err);
 243 struct socket *sock_from_file(struct file *file);
 244 #define              sockfd_put(sock) fput(sock->file)
 245 int net_ratelimit(void);
 246
 247 #define net_ratelimited_function(function, ...)                 \
 248 do {                                                            \
 249         if (net_ratelimit())                                    \
 250                 function(__VA_ARGS__);                          \
 251 } while (0)
 252
 253 #define net_emerg_ratelimited(fmt, ...)                         \
 254         net_ratelimited_function(pr_emerg, fmt, ##__VA_ARGS__)
 255 #define net_alert_ratelimited(fmt, ...)                         \
 256         net_ratelimited_function(pr_alert, fmt, ##__VA_ARGS__)
 257 #define net_crit_ratelimited(fmt, ...)                          \
 258         net_ratelimited_function(pr_crit, fmt, ##__VA_ARGS__)
 259 #define net_err_ratelimited(fmt, ...)                           \
 260         net_ratelimited_function(pr_err, fmt, ##__VA_ARGS__)
 261 #define net_notice_ratelimited(fmt, ...)                        \
 262         net_ratelimited_function(pr_notice, fmt, ##__VA_ARGS__)
 263 #define net_warn_ratelimited(fmt, ...)                          \
 264         net_ratelimited_function(pr_warn, fmt, ##__VA_ARGS__)
 265 #define net_info_ratelimited(fmt, ...)                          \
 266         net_ratelimited_function(pr_info, fmt, ##__VA_ARGS__)
 267 #if defined(CONFIG_DYNAMIC_DEBUG) || \
 268         (defined(CONFIG_DYNAMIC_DEBUG_CORE) && defined(DYNAMIC_DEBUG_MODULE))
 269 #define net_dbg_ratelimited(fmt, ...)                                   \
 270 do {                                                                    \
 271         DEFINE_DYNAMIC_DEBUG_METADATA(descriptor, fmt);                 \
 272         if (DYNAMIC_DEBUG_BRANCH(descriptor) &&                         \
 273             net_ratelimit())                                            \
 274                 __dynamic_pr_debug(&descriptor, pr_fmt(fmt),            \
 275                                    ##__VA_ARGS__);                      \
 276 } while (0)
 277 #elif defined(DEBUG)
 278 #define net_dbg_ratelimited(fmt, ...)                           \
 279         net_ratelimited_function(pr_debug, fmt, ##__VA_ARGS__)
 280 #else
 281 #define net_dbg_ratelimited(fmt, ...)                           \
 282         do {                                                    \
 283                 if (0)                                          \
 284                         no_printk(KERN_DEBUG pr_fmt(fmt), ##__VA_ARGS__); \
 285         } while (0)
 286 #endif
 287
 288 #define net_get_random_once(buf, nbytes)                        \
 289         get_random_once((buf), (nbytes))
 290 #define net_get_random_once_wait(buf, nbytes)                   \
 291         get_random_once_wait((buf), (nbytes))
 292
 293 /*
 294  * E.g. XFS meta- & log-data is in slab pages, or bcache meta
 295  * data pages, or other high order pages allocated by
 296  * __get_free_pages() without __GFP_COMP, which have a page_count
 297  * of 0 and/or have PageSlab() set. We cannot use send_page for
 298  * those, as that does get_page(); put_page(); and would cause
 299  * either a VM_BUG directly, or __page_cache_release a page that
 300  * would actually still be referenced by someone, leading to some
 301  * obscure delayed Oops somewhere else.
 302  */
 303 static inline bool sendpage_ok(struct page *page)
 304 {
 305         return !PageSlab(page) && page_count(page) >= 1;
 306 }
 307
 308 int kernel_sendmsg(struct socket *sock, struct msghdr *msg, struct kvec *vec,
 309                    size_t num, size_t len);
 310 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
 311                           struct kvec *vec, size_t num, size_t len);
 312 int kernel_recvmsg(struct socket *sock, struct msghdr *msg, struct kvec *vec,
 313                    size_t num, size_t len, int flags);
 314
 315 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen);
 316 int kernel_listen(struct socket *sock, int backlog);
 317 int kernel_accept(struct socket *sock, struct socket **newsock, int flags);
 318 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
 319                    int flags);
 320 int kernel_getsockname(struct socket *sock, struct sockaddr *addr);
 321 int kernel_getpeername(struct socket *sock, struct sockaddr *addr);
 322 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
 323                     size_t size, int flags);
 324 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
 325                            size_t size, int flags);
 326 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how);
 327
 328 /* Routine returns the IP overhead imposed by a (caller-protected) socket. */
 329 u32 kernel_sock_ip_overhead(struct sock *sk);
 330
 331 #define MODULE_ALIAS_NETPROTO(proto) \
 332         MODULE_ALIAS("net-pf-" __stringify(proto))
 333
 334 #define MODULE_ALIAS_NET_PF_PROTO(pf, proto) \
 335         MODULE_ALIAS("net-pf-" __stringify(pf) "-proto-" __stringify(proto))
 336
 337 #define MODULE_ALIAS_NET_PF_PROTO_TYPE(pf, proto, type) \
 338         MODULE_ALIAS("net-pf-" __stringify(pf) "-proto-" __stringify(proto) \
 339                      "-type-" __stringify(type))
 340
 341 #define MODULE_ALIAS_NET_PF_PROTO_NAME(pf, proto, name) \
 342         MODULE_ALIAS("net-pf-" __stringify(pf) "-proto-" __stringify(proto) \
 343                      name)
 344 #endif  /* _LINUX_NET_H */