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 <[email protected]>
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/inet_dscp.h>
23 #include <net/sock.h>
24 #include <net/request_sock.h>
25 #include <net/netns/hash.h>
26 #include <net/tcp_states.h>
27 #include <net/l3mdev.h>
28
29 /** struct ip_options - IP Options
30 *
31 * @faddr - Saved first hop address
32 * @nexthop - Saved nexthop address in LSRR and SSRR
33 * @is_strictroute - Strict source route
34 * @srr_is_hit - Packet destination addr was our one
35 * @is_changed - IP checksum more not valid
36 * @rr_needaddr - Need to record addr of outgoing dev
37 * @ts_needtime - Need to record timestamp
38 * @ts_needaddr - Need to record addr of outgoing dev
39 */
40 struct ip_options {
41 __be32 faddr;
42 __be32 nexthop;
43 unsigned char optlen;
44 unsigned char srr;
45 unsigned char rr;
46 unsigned char ts;
47 unsigned char is_strictroute:1,
48 srr_is_hit:1,
49 is_changed:1,
50 rr_needaddr:1,
51 ts_needtime:1,
52 ts_needaddr:1;
53 unsigned char router_alert;
54 unsigned char cipso;
55 unsigned char __pad2;
56 unsigned char __data[];
57 };
58
59 struct ip_options_rcu {
60 struct rcu_head rcu;
61 struct ip_options opt;
62 };
63
64 struct ip_options_data {
65 struct ip_options_rcu opt;
66 char data[40];
67 };
68
69 struct inet_request_sock {
70 struct request_sock req;
71 #define ir_loc_addr req.__req_common.skc_rcv_saddr
72 #define ir_rmt_addr req.__req_common.skc_daddr
73 #define ir_num req.__req_common.skc_num
74 #define ir_rmt_port req.__req_common.skc_dport
75 #define ir_v6_rmt_addr req.__req_common.skc_v6_daddr
76 #define ir_v6_loc_addr req.__req_common.skc_v6_rcv_saddr
77 #define ir_iif req.__req_common.skc_bound_dev_if
78 #define ir_cookie req.__req_common.skc_cookie
79 #define ireq_net req.__req_common.skc_net
80 #define ireq_state req.__req_common.skc_state
81 #define ireq_family req.__req_common.skc_family
82
83 u16 snd_wscale : 4,
84 rcv_wscale : 4,
85 tstamp_ok : 1,
86 sack_ok : 1,
87 wscale_ok : 1,
88 ecn_ok : 1,
89 acked : 1,
90 no_srccheck: 1,
91 smc_ok : 1;
92 u32 ir_mark;
93 union {
94 struct ip_options_rcu __rcu *ireq_opt;
95 #if IS_ENABLED(CONFIG_IPV6)
96 struct {
97 struct ipv6_txoptions *ipv6_opt;
98 struct sk_buff *pktopts;
99 };
100 #endif
101 };
102 };
103
inet_rsk(const struct request_sock * sk)104 static inline struct inet_request_sock *inet_rsk(const struct request_sock *sk)
105 {
106 return (struct inet_request_sock *)sk;
107 }
108
inet_request_mark(const struct sock * sk,struct sk_buff * skb)109 static inline u32 inet_request_mark(const struct sock *sk, struct sk_buff *skb)
110 {
111 u32 mark = READ_ONCE(sk->sk_mark);
112
113 if (!mark && READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_fwmark_accept))
114 return skb->mark;
115
116 return mark;
117 }
118
inet_request_bound_dev_if(const struct sock * sk,struct sk_buff * skb)119 static inline int inet_request_bound_dev_if(const struct sock *sk,
120 struct sk_buff *skb)
121 {
122 int bound_dev_if = READ_ONCE(sk->sk_bound_dev_if);
123 #ifdef CONFIG_NET_L3_MASTER_DEV
124 struct net *net = sock_net(sk);
125
126 if (!bound_dev_if && READ_ONCE(net->ipv4.sysctl_tcp_l3mdev_accept))
127 return l3mdev_master_ifindex_by_index(net, skb->skb_iif);
128 #endif
129
130 return bound_dev_if;
131 }
132
inet_sk_bound_l3mdev(const struct sock * sk)133 static inline int inet_sk_bound_l3mdev(const struct sock *sk)
134 {
135 #ifdef CONFIG_NET_L3_MASTER_DEV
136 struct net *net = sock_net(sk);
137
138 if (!READ_ONCE(net->ipv4.sysctl_tcp_l3mdev_accept))
139 return l3mdev_master_ifindex_by_index(net,
140 sk->sk_bound_dev_if);
141 #endif
142
143 return 0;
144 }
145
inet_bound_dev_eq(bool l3mdev_accept,int bound_dev_if,int dif,int sdif)146 static inline bool inet_bound_dev_eq(bool l3mdev_accept, int bound_dev_if,
147 int dif, int sdif)
148 {
149 if (!bound_dev_if)
150 return !sdif || l3mdev_accept;
151 return bound_dev_if == dif || bound_dev_if == sdif;
152 }
153
inet_sk_bound_dev_eq(const struct net * net,int bound_dev_if,int dif,int sdif)154 static inline bool inet_sk_bound_dev_eq(const struct net *net,
155 int bound_dev_if,
156 int dif, int sdif)
157 {
158 #if IS_ENABLED(CONFIG_NET_L3_MASTER_DEV)
159 return inet_bound_dev_eq(!!READ_ONCE(net->ipv4.sysctl_tcp_l3mdev_accept),
160 bound_dev_if, dif, sdif);
161 #else
162 return inet_bound_dev_eq(true, bound_dev_if, dif, sdif);
163 #endif
164 }
165
166 struct inet_cork {
167 unsigned int flags;
168 __be32 addr;
169 struct ip_options *opt;
170 unsigned int fragsize;
171 int length; /* Total length of all frames */
172 struct dst_entry *dst;
173 u8 tx_flags;
174 __u8 ttl;
175 __s16 tos;
176 u32 priority;
177 __u16 gso_size;
178 u32 ts_opt_id;
179 u64 transmit_time;
180 u32 mark;
181 };
182
183 struct inet_cork_full {
184 struct inet_cork base;
185 struct flowi fl;
186 };
187
188 struct ip_mc_socklist;
189 struct ipv6_pinfo;
190 struct rtable;
191
192 /** struct inet_sock - representation of INET sockets
193 *
194 * @sk - ancestor class
195 * @pinet6 - pointer to IPv6 control block
196 * @inet_daddr - Foreign IPv4 addr
197 * @inet_rcv_saddr - Bound local IPv4 addr
198 * @inet_dport - Destination port
199 * @inet_num - Local port
200 * @inet_flags - various atomic flags
201 * @inet_saddr - Sending source
202 * @uc_ttl - Unicast TTL
203 * @inet_sport - Source port
204 * @inet_id - ID counter for DF pkts
205 * @tos - TOS
206 * @mc_ttl - Multicasting TTL
207 * @uc_index - Unicast outgoing device index
208 * @mc_index - Multicast device index
209 * @mc_list - Group array
210 * @cork - info to build ip hdr on each ip frag while socket is corked
211 */
212 struct inet_sock {
213 /* sk and pinet6 has to be the first two members of inet_sock */
214 struct sock sk;
215 #if IS_ENABLED(CONFIG_IPV6)
216 struct ipv6_pinfo *pinet6;
217 #endif
218 /* Socket demultiplex comparisons on incoming packets. */
219 #define inet_daddr sk.__sk_common.skc_daddr
220 #define inet_rcv_saddr sk.__sk_common.skc_rcv_saddr
221 #define inet_dport sk.__sk_common.skc_dport
222 #define inet_num sk.__sk_common.skc_num
223
224 unsigned long inet_flags;
225 __be32 inet_saddr;
226 __s16 uc_ttl;
227 __be16 inet_sport;
228 struct ip_options_rcu __rcu *inet_opt;
229 atomic_t inet_id;
230
231 __u8 tos;
232 __u8 min_ttl;
233 __u8 mc_ttl;
234 __u8 pmtudisc;
235 __u8 rcv_tos;
236 __u8 convert_csum;
237 int uc_index;
238 int mc_index;
239 __be32 mc_addr;
240 u32 local_port_range; /* high << 16 | low */
241
242 struct ip_mc_socklist __rcu *mc_list;
243 struct inet_cork_full cork;
244 };
245
246 #define IPCORK_OPT 1 /* ip-options has been held in ipcork.opt */
247 #define IPCORK_TS_OPT_ID 2 /* ts_opt_id field is valid, overriding sk_tskey */
248
249 enum {
250 INET_FLAGS_PKTINFO = 0,
251 INET_FLAGS_TTL = 1,
252 INET_FLAGS_TOS = 2,
253 INET_FLAGS_RECVOPTS = 3,
254 INET_FLAGS_RETOPTS = 4,
255 INET_FLAGS_PASSSEC = 5,
256 INET_FLAGS_ORIGDSTADDR = 6,
257 INET_FLAGS_CHECKSUM = 7,
258 INET_FLAGS_RECVFRAGSIZE = 8,
259
260 INET_FLAGS_RECVERR = 9,
261 INET_FLAGS_RECVERR_RFC4884 = 10,
262 INET_FLAGS_FREEBIND = 11,
263 INET_FLAGS_HDRINCL = 12,
264 INET_FLAGS_MC_LOOP = 13,
265 INET_FLAGS_MC_ALL = 14,
266 INET_FLAGS_TRANSPARENT = 15,
267 INET_FLAGS_IS_ICSK = 16,
268 INET_FLAGS_NODEFRAG = 17,
269 INET_FLAGS_BIND_ADDRESS_NO_PORT = 18,
270 INET_FLAGS_DEFER_CONNECT = 19,
271 INET_FLAGS_MC6_LOOP = 20,
272 INET_FLAGS_RECVERR6_RFC4884 = 21,
273 INET_FLAGS_MC6_ALL = 22,
274 INET_FLAGS_AUTOFLOWLABEL_SET = 23,
275 INET_FLAGS_AUTOFLOWLABEL = 24,
276 INET_FLAGS_DONTFRAG = 25,
277 INET_FLAGS_RECVERR6 = 26,
278 INET_FLAGS_REPFLOW = 27,
279 INET_FLAGS_RTALERT_ISOLATE = 28,
280 INET_FLAGS_SNDFLOW = 29,
281 INET_FLAGS_RTALERT = 30,
282 };
283
284 /* cmsg flags for inet */
285 #define IP_CMSG_PKTINFO BIT(INET_FLAGS_PKTINFO)
286 #define IP_CMSG_TTL BIT(INET_FLAGS_TTL)
287 #define IP_CMSG_TOS BIT(INET_FLAGS_TOS)
288 #define IP_CMSG_RECVOPTS BIT(INET_FLAGS_RECVOPTS)
289 #define IP_CMSG_RETOPTS BIT(INET_FLAGS_RETOPTS)
290 #define IP_CMSG_PASSSEC BIT(INET_FLAGS_PASSSEC)
291 #define IP_CMSG_ORIGDSTADDR BIT(INET_FLAGS_ORIGDSTADDR)
292 #define IP_CMSG_CHECKSUM BIT(INET_FLAGS_CHECKSUM)
293 #define IP_CMSG_RECVFRAGSIZE BIT(INET_FLAGS_RECVFRAGSIZE)
294
295 #define IP_CMSG_ALL (IP_CMSG_PKTINFO | IP_CMSG_TTL | \
296 IP_CMSG_TOS | IP_CMSG_RECVOPTS | \
297 IP_CMSG_RETOPTS | IP_CMSG_PASSSEC | \
298 IP_CMSG_ORIGDSTADDR | IP_CMSG_CHECKSUM | \
299 IP_CMSG_RECVFRAGSIZE)
300
inet_cmsg_flags(const struct inet_sock * inet)301 static inline unsigned long inet_cmsg_flags(const struct inet_sock *inet)
302 {
303 return READ_ONCE(inet->inet_flags) & IP_CMSG_ALL;
304 }
305
inet_sk_dscp(const struct inet_sock * inet)306 static inline dscp_t inet_sk_dscp(const struct inet_sock *inet)
307 {
308 return inet_dsfield_to_dscp(READ_ONCE(inet->tos));
309 }
310
311 #define inet_test_bit(nr, sk) \
312 test_bit(INET_FLAGS_##nr, &inet_sk(sk)->inet_flags)
313 #define inet_set_bit(nr, sk) \
314 set_bit(INET_FLAGS_##nr, &inet_sk(sk)->inet_flags)
315 #define inet_clear_bit(nr, sk) \
316 clear_bit(INET_FLAGS_##nr, &inet_sk(sk)->inet_flags)
317 #define inet_assign_bit(nr, sk, val) \
318 assign_bit(INET_FLAGS_##nr, &inet_sk(sk)->inet_flags, val)
319
320 /**
321 * sk_to_full_sk - Access to a full socket
322 * @sk: pointer to a socket
323 *
324 * SYNACK messages might be attached to request sockets.
325 * Some places want to reach the listener in this case.
326 */
sk_to_full_sk(struct sock * sk)327 static inline struct sock *sk_to_full_sk(struct sock *sk)
328 {
329 #ifdef CONFIG_INET
330 if (sk && READ_ONCE(sk->sk_state) == TCP_NEW_SYN_RECV)
331 sk = inet_reqsk(sk)->rsk_listener;
332 if (sk && READ_ONCE(sk->sk_state) == TCP_TIME_WAIT)
333 sk = NULL;
334 #endif
335 return sk;
336 }
337
338 /* sk_to_full_sk() variant with a const argument */
sk_const_to_full_sk(const struct sock * sk)339 static inline const struct sock *sk_const_to_full_sk(const struct sock *sk)
340 {
341 #ifdef CONFIG_INET
342 if (sk && READ_ONCE(sk->sk_state) == TCP_NEW_SYN_RECV)
343 sk = ((const struct request_sock *)sk)->rsk_listener;
344 if (sk && READ_ONCE(sk->sk_state) == TCP_TIME_WAIT)
345 sk = NULL;
346 #endif
347 return sk;
348 }
349
skb_to_full_sk(const struct sk_buff * skb)350 static inline struct sock *skb_to_full_sk(const struct sk_buff *skb)
351 {
352 return sk_to_full_sk(skb->sk);
353 }
354
355 #define inet_sk(ptr) container_of_const(ptr, struct inet_sock, sk)
356
__inet_sk_copy_descendant(struct sock * sk_to,const struct sock * sk_from,const int ancestor_size)357 static inline void __inet_sk_copy_descendant(struct sock *sk_to,
358 const struct sock *sk_from,
359 const int ancestor_size)
360 {
361 memcpy(inet_sk(sk_to) + 1, inet_sk(sk_from) + 1,
362 sk_from->sk_prot->obj_size - ancestor_size);
363 }
364
365 int inet_sk_rebuild_header(struct sock *sk);
366
367 /**
368 * inet_sk_state_load - read sk->sk_state for lockless contexts
369 * @sk: socket pointer
370 *
371 * Paired with inet_sk_state_store(). Used in places we don't hold socket lock:
372 * tcp_diag_get_info(), tcp_get_info(), tcp_poll(), get_tcp4_sock() ...
373 */
inet_sk_state_load(const struct sock * sk)374 static inline int inet_sk_state_load(const struct sock *sk)
375 {
376 /* state change might impact lockless readers. */
377 return smp_load_acquire(&sk->sk_state);
378 }
379
380 /**
381 * inet_sk_state_store - update sk->sk_state
382 * @sk: socket pointer
383 * @newstate: new state
384 *
385 * Paired with inet_sk_state_load(). Should be used in contexts where
386 * state change might impact lockless readers.
387 */
388 void inet_sk_state_store(struct sock *sk, int newstate);
389
390 void inet_sk_set_state(struct sock *sk, int state);
391
__inet_ehashfn(const __be32 laddr,const __u16 lport,const __be32 faddr,const __be16 fport,u32 initval)392 static inline unsigned int __inet_ehashfn(const __be32 laddr,
393 const __u16 lport,
394 const __be32 faddr,
395 const __be16 fport,
396 u32 initval)
397 {
398 return jhash_3words((__force __u32) laddr,
399 (__force __u32) faddr,
400 ((__u32) lport) << 16 | (__force __u32)fport,
401 initval);
402 }
403
404 struct request_sock *inet_reqsk_alloc(const struct request_sock_ops *ops,
405 struct sock *sk_listener,
406 bool attach_listener);
407
inet_sk_flowi_flags(const struct sock * sk)408 static inline __u8 inet_sk_flowi_flags(const struct sock *sk)
409 {
410 __u8 flags = 0;
411
412 if (inet_test_bit(TRANSPARENT, sk) || inet_test_bit(HDRINCL, sk))
413 flags |= FLOWI_FLAG_ANYSRC;
414 return flags;
415 }
416
inet_inc_convert_csum(struct sock * sk)417 static inline void inet_inc_convert_csum(struct sock *sk)
418 {
419 inet_sk(sk)->convert_csum++;
420 }
421
inet_dec_convert_csum(struct sock * sk)422 static inline void inet_dec_convert_csum(struct sock *sk)
423 {
424 if (inet_sk(sk)->convert_csum > 0)
425 inet_sk(sk)->convert_csum--;
426 }
427
inet_get_convert_csum(struct sock * sk)428 static inline bool inet_get_convert_csum(struct sock *sk)
429 {
430 return !!inet_sk(sk)->convert_csum;
431 }
432
433
inet_can_nonlocal_bind(struct net * net,struct inet_sock * inet)434 static inline bool inet_can_nonlocal_bind(struct net *net,
435 struct inet_sock *inet)
436 {
437 return READ_ONCE(net->ipv4.sysctl_ip_nonlocal_bind) ||
438 test_bit(INET_FLAGS_FREEBIND, &inet->inet_flags) ||
439 test_bit(INET_FLAGS_TRANSPARENT, &inet->inet_flags);
440 }
441
inet_addr_valid_or_nonlocal(struct net * net,struct inet_sock * inet,__be32 addr,int addr_type)442 static inline bool inet_addr_valid_or_nonlocal(struct net *net,
443 struct inet_sock *inet,
444 __be32 addr,
445 int addr_type)
446 {
447 return inet_can_nonlocal_bind(net, inet) ||
448 addr == htonl(INADDR_ANY) ||
449 addr_type == RTN_LOCAL ||
450 addr_type == RTN_MULTICAST ||
451 addr_type == RTN_BROADCAST;
452 }
453
454 #endif /* _INET_SOCK_H */
455