1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * INET An implementation of the TCP/IP protocol suite for the LINUX 4 * operating system. INET is implemented using the BSD Socket 5 * interface as the means of communication with the user level. 6 * 7 * Definitions for inet_sock 8 * 9 * Authors: Many, reorganised here by 10 * Arnaldo Carvalho de Melo <acme@mandriva.com> 11 */ 12 #ifndef _INET_SOCK_H 13 #define _INET_SOCK_H 14 15 #include <linux/bitops.h> 16 #include <linux/string.h> 17 #include <linux/types.h> 18 #include <linux/jhash.h> 19 #include <linux/netdevice.h> 20 21 #include <net/flow.h> 22 #include <net/sock.h> 23 #include <net/request_sock.h> 24 #include <net/netns/hash.h> 25 #include <net/tcp_states.h> 26 #include <net/l3mdev.h> 27 28 /** struct ip_options - IP Options 29 * 30 * @faddr - Saved first hop address 31 * @nexthop - Saved nexthop address in LSRR and SSRR 32 * @is_strictroute - Strict source route 33 * @srr_is_hit - Packet destination addr was our one 34 * @is_changed - IP checksum more not valid 35 * @rr_needaddr - Need to record addr of outgoing dev 36 * @ts_needtime - Need to record timestamp 37 * @ts_needaddr - Need to record addr of outgoing dev 38 */ 39 struct ip_options { 40 __be32 faddr; 41 __be32 nexthop; 42 unsigned char optlen; 43 unsigned char srr; 44 unsigned char rr; 45 unsigned char ts; 46 unsigned char is_strictroute:1, 47 srr_is_hit:1, 48 is_changed:1, 49 rr_needaddr:1, 50 ts_needtime:1, 51 ts_needaddr:1; 52 unsigned char router_alert; 53 unsigned char cipso; 54 unsigned char __pad2; 55 unsigned char __data[]; 56 }; 57 58 struct ip_options_rcu { 59 struct rcu_head rcu; 60 struct ip_options opt; 61 }; 62 63 struct ip_options_data { 64 struct ip_options_rcu opt; 65 char data[40]; 66 }; 67 68 struct inet_request_sock { 69 struct request_sock req; 70 #define ir_loc_addr req.__req_common.skc_rcv_saddr 71 #define ir_rmt_addr req.__req_common.skc_daddr 72 #define ir_num req.__req_common.skc_num 73 #define ir_rmt_port req.__req_common.skc_dport 74 #define ir_v6_rmt_addr req.__req_common.skc_v6_daddr 75 #define ir_v6_loc_addr req.__req_common.skc_v6_rcv_saddr 76 #define ir_iif req.__req_common.skc_bound_dev_if 77 #define ir_cookie req.__req_common.skc_cookie 78 #define ireq_net req.__req_common.skc_net 79 #define ireq_state req.__req_common.skc_state 80 #define ireq_family req.__req_common.skc_family 81 82 u16 snd_wscale : 4, 83 rcv_wscale : 4, 84 tstamp_ok : 1, 85 sack_ok : 1, 86 wscale_ok : 1, 87 ecn_ok : 1, 88 acked : 1, 89 no_srccheck: 1, 90 smc_ok : 1; 91 u32 ir_mark; 92 union { 93 struct ip_options_rcu __rcu *ireq_opt; 94 #if IS_ENABLED(CONFIG_IPV6) 95 struct { 96 struct ipv6_txoptions *ipv6_opt; 97 struct sk_buff *pktopts; 98 }; 99 #endif 100 }; 101 }; 102 103 static inline struct inet_request_sock *inet_rsk(const struct request_sock *sk) 104 { 105 return (struct inet_request_sock *)sk; 106 } 107 108 static inline u32 inet_request_mark(const struct sock *sk, struct sk_buff *skb) 109 { 110 if (!sk->sk_mark && 111 READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_fwmark_accept)) 112 return skb->mark; 113 114 return sk->sk_mark; 115 } 116 117 static inline int inet_request_bound_dev_if(const struct sock *sk, 118 struct sk_buff *skb) 119 { 120 int bound_dev_if = READ_ONCE(sk->sk_bound_dev_if); 121 #ifdef CONFIG_NET_L3_MASTER_DEV 122 struct net *net = sock_net(sk); 123 124 if (!bound_dev_if && READ_ONCE(net->ipv4.sysctl_tcp_l3mdev_accept)) 125 return l3mdev_master_ifindex_by_index(net, skb->skb_iif); 126 #endif 127 128 return bound_dev_if; 129 } 130 131 static inline int inet_sk_bound_l3mdev(const struct sock *sk) 132 { 133 #ifdef CONFIG_NET_L3_MASTER_DEV 134 struct net *net = sock_net(sk); 135 136 if (!READ_ONCE(net->ipv4.sysctl_tcp_l3mdev_accept)) 137 return l3mdev_master_ifindex_by_index(net, 138 sk->sk_bound_dev_if); 139 #endif 140 141 return 0; 142 } 143 144 static inline bool inet_bound_dev_eq(bool l3mdev_accept, int bound_dev_if, 145 int dif, int sdif) 146 { 147 if (!bound_dev_if) 148 return !sdif || l3mdev_accept; 149 return bound_dev_if == dif || bound_dev_if == sdif; 150 } 151 152 static inline bool inet_sk_bound_dev_eq(struct net *net, int bound_dev_if, 153 int dif, int sdif) 154 { 155 #if IS_ENABLED(CONFIG_NET_L3_MASTER_DEV) 156 return inet_bound_dev_eq(!!READ_ONCE(net->ipv4.sysctl_tcp_l3mdev_accept), 157 bound_dev_if, dif, sdif); 158 #else 159 return inet_bound_dev_eq(true, bound_dev_if, dif, sdif); 160 #endif 161 } 162 163 struct inet_cork { 164 unsigned int flags; 165 __be32 addr; 166 struct ip_options *opt; 167 unsigned int fragsize; 168 int length; /* Total length of all frames */ 169 struct dst_entry *dst; 170 u8 tx_flags; 171 __u8 ttl; 172 __s16 tos; 173 char priority; 174 __u16 gso_size; 175 u64 transmit_time; 176 u32 mark; 177 }; 178 179 struct inet_cork_full { 180 struct inet_cork base; 181 struct flowi fl; 182 }; 183 184 struct ip_mc_socklist; 185 struct ipv6_pinfo; 186 struct rtable; 187 188 /** struct inet_sock - representation of INET sockets 189 * 190 * @sk - ancestor class 191 * @pinet6 - pointer to IPv6 control block 192 * @inet_daddr - Foreign IPv4 addr 193 * @inet_rcv_saddr - Bound local IPv4 addr 194 * @inet_dport - Destination port 195 * @inet_num - Local port 196 * @inet_saddr - Sending source 197 * @uc_ttl - Unicast TTL 198 * @inet_sport - Source port 199 * @inet_id - ID counter for DF pkts 200 * @tos - TOS 201 * @mc_ttl - Multicasting TTL 202 * @is_icsk - is this an inet_connection_sock? 203 * @uc_index - Unicast outgoing device index 204 * @mc_index - Multicast device index 205 * @mc_list - Group array 206 * @cork - info to build ip hdr on each ip frag while socket is corked 207 */ 208 struct inet_sock { 209 /* sk and pinet6 has to be the first two members of inet_sock */ 210 struct sock sk; 211 #if IS_ENABLED(CONFIG_IPV6) 212 struct ipv6_pinfo *pinet6; 213 #endif 214 /* Socket demultiplex comparisons on incoming packets. */ 215 #define inet_daddr sk.__sk_common.skc_daddr 216 #define inet_rcv_saddr sk.__sk_common.skc_rcv_saddr 217 #define inet_dport sk.__sk_common.skc_dport 218 #define inet_num sk.__sk_common.skc_num 219 220 __be32 inet_saddr; 221 __s16 uc_ttl; 222 __u16 cmsg_flags; 223 struct ip_options_rcu __rcu *inet_opt; 224 __be16 inet_sport; 225 __u16 inet_id; 226 227 __u8 tos; 228 __u8 min_ttl; 229 __u8 mc_ttl; 230 __u8 pmtudisc; 231 __u8 recverr:1, 232 is_icsk:1, 233 freebind:1, 234 hdrincl:1, 235 mc_loop:1, 236 transparent:1, 237 mc_all:1, 238 nodefrag:1; 239 __u8 bind_address_no_port:1, 240 recverr_rfc4884:1, 241 defer_connect:1; /* Indicates that fastopen_connect is set 242 * and cookie exists so we defer connect 243 * until first data frame is written 244 */ 245 __u8 rcv_tos; 246 __u8 convert_csum; 247 int uc_index; 248 int mc_index; 249 __be32 mc_addr; 250 struct ip_mc_socklist __rcu *mc_list; 251 struct inet_cork_full cork; 252 struct { 253 __u16 lo; 254 __u16 hi; 255 } local_port_range; 256 }; 257 258 #define IPCORK_OPT 1 /* ip-options has been held in ipcork.opt */ 259 #define IPCORK_ALLFRAG 2 /* always fragment (for ipv6 for now) */ 260 261 /* cmsg flags for inet */ 262 #define IP_CMSG_PKTINFO BIT(0) 263 #define IP_CMSG_TTL BIT(1) 264 #define IP_CMSG_TOS BIT(2) 265 #define IP_CMSG_RECVOPTS BIT(3) 266 #define IP_CMSG_RETOPTS BIT(4) 267 #define IP_CMSG_PASSSEC BIT(5) 268 #define IP_CMSG_ORIGDSTADDR BIT(6) 269 #define IP_CMSG_CHECKSUM BIT(7) 270 #define IP_CMSG_RECVFRAGSIZE BIT(8) 271 272 static inline bool sk_is_inet(struct sock *sk) 273 { 274 return sk->sk_family == AF_INET || sk->sk_family == AF_INET6; 275 } 276 277 /** 278 * sk_to_full_sk - Access to a full socket 279 * @sk: pointer to a socket 280 * 281 * SYNACK messages might be attached to request sockets. 282 * Some places want to reach the listener in this case. 283 */ 284 static inline struct sock *sk_to_full_sk(struct sock *sk) 285 { 286 #ifdef CONFIG_INET 287 if (sk && sk->sk_state == TCP_NEW_SYN_RECV) 288 sk = inet_reqsk(sk)->rsk_listener; 289 #endif 290 return sk; 291 } 292 293 /* sk_to_full_sk() variant with a const argument */ 294 static inline const struct sock *sk_const_to_full_sk(const struct sock *sk) 295 { 296 #ifdef CONFIG_INET 297 if (sk && sk->sk_state == TCP_NEW_SYN_RECV) 298 sk = ((const struct request_sock *)sk)->rsk_listener; 299 #endif 300 return sk; 301 } 302 303 static inline struct sock *skb_to_full_sk(const struct sk_buff *skb) 304 { 305 return sk_to_full_sk(skb->sk); 306 } 307 308 #define inet_sk(ptr) container_of_const(ptr, struct inet_sock, sk) 309 310 static inline void __inet_sk_copy_descendant(struct sock *sk_to, 311 const struct sock *sk_from, 312 const int ancestor_size) 313 { 314 memcpy(inet_sk(sk_to) + 1, inet_sk(sk_from) + 1, 315 sk_from->sk_prot->obj_size - ancestor_size); 316 } 317 318 int inet_sk_rebuild_header(struct sock *sk); 319 320 /** 321 * inet_sk_state_load - read sk->sk_state for lockless contexts 322 * @sk: socket pointer 323 * 324 * Paired with inet_sk_state_store(). Used in places we don't hold socket lock: 325 * tcp_diag_get_info(), tcp_get_info(), tcp_poll(), get_tcp4_sock() ... 326 */ 327 static inline int inet_sk_state_load(const struct sock *sk) 328 { 329 /* state change might impact lockless readers. */ 330 return smp_load_acquire(&sk->sk_state); 331 } 332 333 /** 334 * inet_sk_state_store - update sk->sk_state 335 * @sk: socket pointer 336 * @newstate: new state 337 * 338 * Paired with inet_sk_state_load(). Should be used in contexts where 339 * state change might impact lockless readers. 340 */ 341 void inet_sk_state_store(struct sock *sk, int newstate); 342 343 void inet_sk_set_state(struct sock *sk, int state); 344 345 static inline unsigned int __inet_ehashfn(const __be32 laddr, 346 const __u16 lport, 347 const __be32 faddr, 348 const __be16 fport, 349 u32 initval) 350 { 351 return jhash_3words((__force __u32) laddr, 352 (__force __u32) faddr, 353 ((__u32) lport) << 16 | (__force __u32)fport, 354 initval); 355 } 356 357 struct request_sock *inet_reqsk_alloc(const struct request_sock_ops *ops, 358 struct sock *sk_listener, 359 bool attach_listener); 360 361 static inline __u8 inet_sk_flowi_flags(const struct sock *sk) 362 { 363 __u8 flags = 0; 364 365 if (inet_sk(sk)->transparent || inet_sk(sk)->hdrincl) 366 flags |= FLOWI_FLAG_ANYSRC; 367 return flags; 368 } 369 370 static inline void inet_inc_convert_csum(struct sock *sk) 371 { 372 inet_sk(sk)->convert_csum++; 373 } 374 375 static inline void inet_dec_convert_csum(struct sock *sk) 376 { 377 if (inet_sk(sk)->convert_csum > 0) 378 inet_sk(sk)->convert_csum--; 379 } 380 381 static inline bool inet_get_convert_csum(struct sock *sk) 382 { 383 return !!inet_sk(sk)->convert_csum; 384 } 385 386 387 static inline bool inet_can_nonlocal_bind(struct net *net, 388 struct inet_sock *inet) 389 { 390 return READ_ONCE(net->ipv4.sysctl_ip_nonlocal_bind) || 391 inet->freebind || inet->transparent; 392 } 393 394 static inline bool inet_addr_valid_or_nonlocal(struct net *net, 395 struct inet_sock *inet, 396 __be32 addr, 397 int addr_type) 398 { 399 return inet_can_nonlocal_bind(net, inet) || 400 addr == htonl(INADDR_ANY) || 401 addr_type == RTN_LOCAL || 402 addr_type == RTN_MULTICAST || 403 addr_type == RTN_BROADCAST; 404 } 405 406 #endif /* _INET_SOCK_H */ 407