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 struct inet_cork { 153 unsigned int flags; 154 __be32 addr; 155 struct ip_options *opt; 156 unsigned int fragsize; 157 int length; /* Total length of all frames */ 158 struct dst_entry *dst; 159 u8 tx_flags; 160 __u8 ttl; 161 __s16 tos; 162 char priority; 163 __u16 gso_size; 164 u64 transmit_time; 165 u32 mark; 166 }; 167 168 struct inet_cork_full { 169 struct inet_cork base; 170 struct flowi fl; 171 }; 172 173 struct ip_mc_socklist; 174 struct ipv6_pinfo; 175 struct rtable; 176 177 /** struct inet_sock - representation of INET sockets 178 * 179 * @sk - ancestor class 180 * @pinet6 - pointer to IPv6 control block 181 * @inet_daddr - Foreign IPv4 addr 182 * @inet_rcv_saddr - Bound local IPv4 addr 183 * @inet_dport - Destination port 184 * @inet_num - Local port 185 * @inet_saddr - Sending source 186 * @uc_ttl - Unicast TTL 187 * @inet_sport - Source port 188 * @inet_id - ID counter for DF pkts 189 * @tos - TOS 190 * @mc_ttl - Multicasting TTL 191 * @is_icsk - is this an inet_connection_sock? 192 * @uc_index - Unicast outgoing device index 193 * @mc_index - Multicast device index 194 * @mc_list - Group array 195 * @cork - info to build ip hdr on each ip frag while socket is corked 196 */ 197 struct inet_sock { 198 /* sk and pinet6 has to be the first two members of inet_sock */ 199 struct sock sk; 200 #if IS_ENABLED(CONFIG_IPV6) 201 struct ipv6_pinfo *pinet6; 202 #endif 203 /* Socket demultiplex comparisons on incoming packets. */ 204 #define inet_daddr sk.__sk_common.skc_daddr 205 #define inet_rcv_saddr sk.__sk_common.skc_rcv_saddr 206 #define inet_dport sk.__sk_common.skc_dport 207 #define inet_num sk.__sk_common.skc_num 208 209 __be32 inet_saddr; 210 __s16 uc_ttl; 211 __u16 cmsg_flags; 212 struct ip_options_rcu __rcu *inet_opt; 213 __be16 inet_sport; 214 __u16 inet_id; 215 216 __u8 tos; 217 __u8 min_ttl; 218 __u8 mc_ttl; 219 __u8 pmtudisc; 220 __u8 recverr:1, 221 is_icsk:1, 222 freebind:1, 223 hdrincl:1, 224 mc_loop:1, 225 transparent:1, 226 mc_all:1, 227 nodefrag:1; 228 __u8 bind_address_no_port:1, 229 recverr_rfc4884:1, 230 defer_connect:1; /* Indicates that fastopen_connect is set 231 * and cookie exists so we defer connect 232 * until first data frame is written 233 */ 234 __u8 rcv_tos; 235 __u8 convert_csum; 236 int uc_index; 237 int mc_index; 238 __be32 mc_addr; 239 struct ip_mc_socklist __rcu *mc_list; 240 struct inet_cork_full cork; 241 }; 242 243 #define IPCORK_OPT 1 /* ip-options has been held in ipcork.opt */ 244 #define IPCORK_ALLFRAG 2 /* always fragment (for ipv6 for now) */ 245 246 /* cmsg flags for inet */ 247 #define IP_CMSG_PKTINFO BIT(0) 248 #define IP_CMSG_TTL BIT(1) 249 #define IP_CMSG_TOS BIT(2) 250 #define IP_CMSG_RECVOPTS BIT(3) 251 #define IP_CMSG_RETOPTS BIT(4) 252 #define IP_CMSG_PASSSEC BIT(5) 253 #define IP_CMSG_ORIGDSTADDR BIT(6) 254 #define IP_CMSG_CHECKSUM BIT(7) 255 #define IP_CMSG_RECVFRAGSIZE BIT(8) 256 257 static inline bool sk_is_inet(struct sock *sk) 258 { 259 return sk->sk_family == AF_INET || sk->sk_family == AF_INET6; 260 } 261 262 /** 263 * sk_to_full_sk - Access to a full socket 264 * @sk: pointer to a socket 265 * 266 * SYNACK messages might be attached to request sockets. 267 * Some places want to reach the listener in this case. 268 */ 269 static inline struct sock *sk_to_full_sk(struct sock *sk) 270 { 271 #ifdef CONFIG_INET 272 if (sk && sk->sk_state == TCP_NEW_SYN_RECV) 273 sk = inet_reqsk(sk)->rsk_listener; 274 #endif 275 return sk; 276 } 277 278 /* sk_to_full_sk() variant with a const argument */ 279 static inline const struct sock *sk_const_to_full_sk(const struct sock *sk) 280 { 281 #ifdef CONFIG_INET 282 if (sk && sk->sk_state == TCP_NEW_SYN_RECV) 283 sk = ((const struct request_sock *)sk)->rsk_listener; 284 #endif 285 return sk; 286 } 287 288 static inline struct sock *skb_to_full_sk(const struct sk_buff *skb) 289 { 290 return sk_to_full_sk(skb->sk); 291 } 292 293 static inline struct inet_sock *inet_sk(const struct sock *sk) 294 { 295 return (struct inet_sock *)sk; 296 } 297 298 static inline void __inet_sk_copy_descendant(struct sock *sk_to, 299 const struct sock *sk_from, 300 const int ancestor_size) 301 { 302 memcpy(inet_sk(sk_to) + 1, inet_sk(sk_from) + 1, 303 sk_from->sk_prot->obj_size - ancestor_size); 304 } 305 306 int inet_sk_rebuild_header(struct sock *sk); 307 308 /** 309 * inet_sk_state_load - read sk->sk_state for lockless contexts 310 * @sk: socket pointer 311 * 312 * Paired with inet_sk_state_store(). Used in places we don't hold socket lock: 313 * tcp_diag_get_info(), tcp_get_info(), tcp_poll(), get_tcp4_sock() ... 314 */ 315 static inline int inet_sk_state_load(const struct sock *sk) 316 { 317 /* state change might impact lockless readers. */ 318 return smp_load_acquire(&sk->sk_state); 319 } 320 321 /** 322 * inet_sk_state_store - update sk->sk_state 323 * @sk: socket pointer 324 * @newstate: new state 325 * 326 * Paired with inet_sk_state_load(). Should be used in contexts where 327 * state change might impact lockless readers. 328 */ 329 void inet_sk_state_store(struct sock *sk, int newstate); 330 331 void inet_sk_set_state(struct sock *sk, int state); 332 333 static inline unsigned int __inet_ehashfn(const __be32 laddr, 334 const __u16 lport, 335 const __be32 faddr, 336 const __be16 fport, 337 u32 initval) 338 { 339 return jhash_3words((__force __u32) laddr, 340 (__force __u32) faddr, 341 ((__u32) lport) << 16 | (__force __u32)fport, 342 initval); 343 } 344 345 struct request_sock *inet_reqsk_alloc(const struct request_sock_ops *ops, 346 struct sock *sk_listener, 347 bool attach_listener); 348 349 static inline __u8 inet_sk_flowi_flags(const struct sock *sk) 350 { 351 __u8 flags = 0; 352 353 if (inet_sk(sk)->transparent || inet_sk(sk)->hdrincl) 354 flags |= FLOWI_FLAG_ANYSRC; 355 return flags; 356 } 357 358 static inline void inet_inc_convert_csum(struct sock *sk) 359 { 360 inet_sk(sk)->convert_csum++; 361 } 362 363 static inline void inet_dec_convert_csum(struct sock *sk) 364 { 365 if (inet_sk(sk)->convert_csum > 0) 366 inet_sk(sk)->convert_csum--; 367 } 368 369 static inline bool inet_get_convert_csum(struct sock *sk) 370 { 371 return !!inet_sk(sk)->convert_csum; 372 } 373 374 375 static inline bool inet_can_nonlocal_bind(struct net *net, 376 struct inet_sock *inet) 377 { 378 return READ_ONCE(net->ipv4.sysctl_ip_nonlocal_bind) || 379 inet->freebind || inet->transparent; 380 } 381 382 static inline bool inet_addr_valid_or_nonlocal(struct net *net, 383 struct inet_sock *inet, 384 __be32 addr, 385 int addr_type) 386 { 387 return inet_can_nonlocal_bind(net, inet) || 388 addr == htonl(INADDR_ANY) || 389 addr_type == RTN_LOCAL || 390 addr_type == RTN_MULTICAST || 391 addr_type == RTN_BROADCAST; 392 } 393 394 #endif /* _INET_SOCK_H */ 395