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 && sock_net(sk)->ipv4.sysctl_tcp_fwmark_accept) 111 return skb->mark; 112 113 return sk->sk_mark; 114 } 115 116 static inline int inet_request_bound_dev_if(const struct sock *sk, 117 struct sk_buff *skb) 118 { 119 int bound_dev_if = READ_ONCE(sk->sk_bound_dev_if); 120 #ifdef CONFIG_NET_L3_MASTER_DEV 121 struct net *net = sock_net(sk); 122 123 if (!bound_dev_if && net->ipv4.sysctl_tcp_l3mdev_accept) 124 return l3mdev_master_ifindex_by_index(net, skb->skb_iif); 125 #endif 126 127 return bound_dev_if; 128 } 129 130 static inline int inet_sk_bound_l3mdev(const struct sock *sk) 131 { 132 #ifdef CONFIG_NET_L3_MASTER_DEV 133 struct net *net = sock_net(sk); 134 135 if (!net->ipv4.sysctl_tcp_l3mdev_accept) 136 return l3mdev_master_ifindex_by_index(net, 137 sk->sk_bound_dev_if); 138 #endif 139 140 return 0; 141 } 142 143 static inline bool inet_bound_dev_eq(bool l3mdev_accept, int bound_dev_if, 144 int dif, int sdif) 145 { 146 if (!bound_dev_if) 147 return !sdif || l3mdev_accept; 148 return bound_dev_if == dif || bound_dev_if == sdif; 149 } 150 151 struct inet_cork { 152 unsigned int flags; 153 __be32 addr; 154 struct ip_options *opt; 155 unsigned int fragsize; 156 int length; /* Total length of all frames */ 157 struct dst_entry *dst; 158 u8 tx_flags; 159 __u8 ttl; 160 __s16 tos; 161 char priority; 162 __u16 gso_size; 163 u64 transmit_time; 164 u32 mark; 165 }; 166 167 struct inet_cork_full { 168 struct inet_cork base; 169 struct flowi fl; 170 }; 171 172 struct ip_mc_socklist; 173 struct ipv6_pinfo; 174 struct rtable; 175 176 /** struct inet_sock - representation of INET sockets 177 * 178 * @sk - ancestor class 179 * @pinet6 - pointer to IPv6 control block 180 * @inet_daddr - Foreign IPv4 addr 181 * @inet_rcv_saddr - Bound local IPv4 addr 182 * @inet_dport - Destination port 183 * @inet_num - Local port 184 * @inet_saddr - Sending source 185 * @uc_ttl - Unicast TTL 186 * @inet_sport - Source port 187 * @inet_id - ID counter for DF pkts 188 * @tos - TOS 189 * @mc_ttl - Multicasting TTL 190 * @is_icsk - is this an inet_connection_sock? 191 * @uc_index - Unicast outgoing device index 192 * @mc_index - Multicast device index 193 * @mc_list - Group array 194 * @cork - info to build ip hdr on each ip frag while socket is corked 195 */ 196 struct inet_sock { 197 /* sk and pinet6 has to be the first two members of inet_sock */ 198 struct sock sk; 199 #if IS_ENABLED(CONFIG_IPV6) 200 struct ipv6_pinfo *pinet6; 201 #endif 202 /* Socket demultiplex comparisons on incoming packets. */ 203 #define inet_daddr sk.__sk_common.skc_daddr 204 #define inet_rcv_saddr sk.__sk_common.skc_rcv_saddr 205 #define inet_dport sk.__sk_common.skc_dport 206 #define inet_num sk.__sk_common.skc_num 207 208 __be32 inet_saddr; 209 __s16 uc_ttl; 210 __u16 cmsg_flags; 211 struct ip_options_rcu __rcu *inet_opt; 212 __be16 inet_sport; 213 __u16 inet_id; 214 215 __u8 tos; 216 __u8 min_ttl; 217 __u8 mc_ttl; 218 __u8 pmtudisc; 219 __u8 recverr:1, 220 is_icsk:1, 221 freebind:1, 222 hdrincl:1, 223 mc_loop:1, 224 transparent:1, 225 mc_all:1, 226 nodefrag:1; 227 __u8 bind_address_no_port:1, 228 recverr_rfc4884:1, 229 defer_connect:1; /* Indicates that fastopen_connect is set 230 * and cookie exists so we defer connect 231 * until first data frame is written 232 */ 233 __u8 rcv_tos; 234 __u8 convert_csum; 235 int uc_index; 236 int mc_index; 237 __be32 mc_addr; 238 struct ip_mc_socklist __rcu *mc_list; 239 struct inet_cork_full cork; 240 }; 241 242 #define IPCORK_OPT 1 /* ip-options has been held in ipcork.opt */ 243 #define IPCORK_ALLFRAG 2 /* always fragment (for ipv6 for now) */ 244 245 /* cmsg flags for inet */ 246 #define IP_CMSG_PKTINFO BIT(0) 247 #define IP_CMSG_TTL BIT(1) 248 #define IP_CMSG_TOS BIT(2) 249 #define IP_CMSG_RECVOPTS BIT(3) 250 #define IP_CMSG_RETOPTS BIT(4) 251 #define IP_CMSG_PASSSEC BIT(5) 252 #define IP_CMSG_ORIGDSTADDR BIT(6) 253 #define IP_CMSG_CHECKSUM BIT(7) 254 #define IP_CMSG_RECVFRAGSIZE BIT(8) 255 256 static inline bool sk_is_inet(struct sock *sk) 257 { 258 return sk->sk_family == AF_INET || sk->sk_family == AF_INET6; 259 } 260 261 /** 262 * sk_to_full_sk - Access to a full socket 263 * @sk: pointer to a socket 264 * 265 * SYNACK messages might be attached to request sockets. 266 * Some places want to reach the listener in this case. 267 */ 268 static inline struct sock *sk_to_full_sk(struct sock *sk) 269 { 270 #ifdef CONFIG_INET 271 if (sk && sk->sk_state == TCP_NEW_SYN_RECV) 272 sk = inet_reqsk(sk)->rsk_listener; 273 #endif 274 return sk; 275 } 276 277 /* sk_to_full_sk() variant with a const argument */ 278 static inline const struct sock *sk_const_to_full_sk(const struct sock *sk) 279 { 280 #ifdef CONFIG_INET 281 if (sk && sk->sk_state == TCP_NEW_SYN_RECV) 282 sk = ((const struct request_sock *)sk)->rsk_listener; 283 #endif 284 return sk; 285 } 286 287 static inline struct sock *skb_to_full_sk(const struct sk_buff *skb) 288 { 289 return sk_to_full_sk(skb->sk); 290 } 291 292 static inline struct inet_sock *inet_sk(const struct sock *sk) 293 { 294 return (struct inet_sock *)sk; 295 } 296 297 static inline void __inet_sk_copy_descendant(struct sock *sk_to, 298 const struct sock *sk_from, 299 const int ancestor_size) 300 { 301 memcpy(inet_sk(sk_to) + 1, inet_sk(sk_from) + 1, 302 sk_from->sk_prot->obj_size - ancestor_size); 303 } 304 305 int inet_sk_rebuild_header(struct sock *sk); 306 307 /** 308 * inet_sk_state_load - read sk->sk_state for lockless contexts 309 * @sk: socket pointer 310 * 311 * Paired with inet_sk_state_store(). Used in places we don't hold socket lock: 312 * tcp_diag_get_info(), tcp_get_info(), tcp_poll(), get_tcp4_sock() ... 313 */ 314 static inline int inet_sk_state_load(const struct sock *sk) 315 { 316 /* state change might impact lockless readers. */ 317 return smp_load_acquire(&sk->sk_state); 318 } 319 320 /** 321 * inet_sk_state_store - update sk->sk_state 322 * @sk: socket pointer 323 * @newstate: new state 324 * 325 * Paired with inet_sk_state_load(). Should be used in contexts where 326 * state change might impact lockless readers. 327 */ 328 void inet_sk_state_store(struct sock *sk, int newstate); 329 330 void inet_sk_set_state(struct sock *sk, int state); 331 332 static inline unsigned int __inet_ehashfn(const __be32 laddr, 333 const __u16 lport, 334 const __be32 faddr, 335 const __be16 fport, 336 u32 initval) 337 { 338 return jhash_3words((__force __u32) laddr, 339 (__force __u32) faddr, 340 ((__u32) lport) << 16 | (__force __u32)fport, 341 initval); 342 } 343 344 struct request_sock *inet_reqsk_alloc(const struct request_sock_ops *ops, 345 struct sock *sk_listener, 346 bool attach_listener); 347 348 static inline __u8 inet_sk_flowi_flags(const struct sock *sk) 349 { 350 __u8 flags = 0; 351 352 if (inet_sk(sk)->transparent || inet_sk(sk)->hdrincl) 353 flags |= FLOWI_FLAG_ANYSRC; 354 return flags; 355 } 356 357 static inline void inet_inc_convert_csum(struct sock *sk) 358 { 359 inet_sk(sk)->convert_csum++; 360 } 361 362 static inline void inet_dec_convert_csum(struct sock *sk) 363 { 364 if (inet_sk(sk)->convert_csum > 0) 365 inet_sk(sk)->convert_csum--; 366 } 367 368 static inline bool inet_get_convert_csum(struct sock *sk) 369 { 370 return !!inet_sk(sk)->convert_csum; 371 } 372 373 374 static inline bool inet_can_nonlocal_bind(struct net *net, 375 struct inet_sock *inet) 376 { 377 return net->ipv4.sysctl_ip_nonlocal_bind || 378 inet->freebind || inet->transparent; 379 } 380 381 static inline bool inet_addr_valid_or_nonlocal(struct net *net, 382 struct inet_sock *inet, 383 __be32 addr, 384 int addr_type) 385 { 386 return inet_can_nonlocal_bind(net, inet) || 387 addr == htonl(INADDR_ANY) || 388 addr_type == RTN_LOCAL || 389 addr_type == RTN_MULTICAST || 390 addr_type == RTN_BROADCAST; 391 } 392 393 #endif /* _INET_SOCK_H */ 394