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/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 #define IP_OPTIONS_DATA_FIXED_SIZE 40
30
31 /** struct ip_options - IP Options
32 *
33 * @faddr - Saved first hop address
34 * @nexthop - Saved nexthop address in LSRR and SSRR
35 * @is_strictroute - Strict source route
36 * @srr_is_hit - Packet destination addr was our one
37 * @is_changed - IP checksum more not valid
38 * @rr_needaddr - Need to record addr of outgoing dev
39 * @ts_needtime - Need to record timestamp
40 * @ts_needaddr - Need to record addr of outgoing dev
41 */
42 struct ip_options {
43 __be32 faddr;
44 __be32 nexthop;
45 unsigned char optlen;
46 unsigned char srr;
47 unsigned char rr;
48 unsigned char ts;
49 unsigned char is_strictroute:1,
50 srr_is_hit:1,
51 is_changed:1,
52 rr_needaddr:1,
53 ts_needtime:1,
54 ts_needaddr:1;
55 unsigned char router_alert;
56 unsigned char cipso;
57 unsigned char __pad2;
58 unsigned char __data[];
59 };
60
61 struct ip_options_rcu {
62 struct rcu_head rcu;
63
64 /* Must be last as it ends in a flexible-array member. */
65 struct ip_options opt;
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 #define inet_rsk(ptr) container_of_const(ptr, struct inet_request_sock, req)
104
inet_request_mark(const struct sock * sk,struct sk_buff * skb)105 static inline u32 inet_request_mark(const struct sock *sk, struct sk_buff *skb)
106 {
107 u32 mark = READ_ONCE(sk->sk_mark);
108
109 if (!mark && READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_fwmark_accept))
110 return skb->mark;
111
112 return mark;
113 }
114
inet_request_bound_dev_if(const struct sock * sk,struct sk_buff * skb)115 static inline int inet_request_bound_dev_if(const struct sock *sk,
116 struct sk_buff *skb)
117 {
118 int bound_dev_if = READ_ONCE(sk->sk_bound_dev_if);
119 #ifdef CONFIG_NET_L3_MASTER_DEV
120 struct net *net = sock_net(sk);
121
122 if (!bound_dev_if && READ_ONCE(net->ipv4.sysctl_tcp_l3mdev_accept))
123 return l3mdev_master_ifindex_by_index(net, skb->skb_iif);
124 #endif
125
126 return bound_dev_if;
127 }
128
inet_sk_bound_l3mdev(const struct sock * sk)129 static inline int inet_sk_bound_l3mdev(const struct sock *sk)
130 {
131 #ifdef CONFIG_NET_L3_MASTER_DEV
132 struct net *net = sock_net(sk);
133
134 if (!READ_ONCE(net->ipv4.sysctl_tcp_l3mdev_accept))
135 return l3mdev_master_ifindex_by_index(net,
136 sk->sk_bound_dev_if);
137 #endif
138
139 return 0;
140 }
141
inet_bound_dev_eq(bool l3mdev_accept,int bound_dev_if,int dif,int sdif)142 static inline bool inet_bound_dev_eq(bool l3mdev_accept, int bound_dev_if,
143 int dif, int sdif)
144 {
145 if (!bound_dev_if)
146 return !sdif || l3mdev_accept;
147 return bound_dev_if == dif || bound_dev_if == sdif;
148 }
149
inet_sk_bound_dev_eq(const struct net * net,int bound_dev_if,int dif,int sdif)150 static inline bool inet_sk_bound_dev_eq(const struct net *net,
151 int bound_dev_if,
152 int dif, int sdif)
153 {
154 #if IS_ENABLED(CONFIG_NET_L3_MASTER_DEV)
155 return inet_bound_dev_eq(!!READ_ONCE(net->ipv4.sysctl_tcp_l3mdev_accept),
156 bound_dev_if, dif, sdif);
157 #else
158 return inet_bound_dev_eq(true, bound_dev_if, dif, sdif);
159 #endif
160 }
161
162 struct inet6_cork {
163 struct ipv6_txoptions *opt;
164 u8 hop_limit;
165 u8 tclass;
166 u8 dontfrag:1;
167 };
168
169 struct inet_cork {
170 unsigned int flags;
171 __be32 addr;
172 struct ip_options *opt;
173 unsigned int fragsize;
174 int length; /* Total length of all frames */
175 struct dst_entry *dst;
176 u8 tx_flags;
177 __u8 ttl;
178 __s16 tos;
179 u32 priority;
180 __u16 gso_size;
181 u32 ts_opt_id;
182 u64 transmit_time;
183 u32 mark;
184 };
185
186 struct inet_cork_full {
187 struct inet_cork base;
188 struct flowi fl;
189 #if IS_ENABLED(CONFIG_IPV6)
190 struct inet6_cork base6;
191 #endif
192 };
193
194 struct ip_mc_socklist;
195 struct ipv6_pinfo;
196 struct rtable;
197
198 /** struct inet_sock - representation of INET sockets
199 *
200 * @sk - ancestor class
201 * @pinet6 - pointer to IPv6 control block
202 * @inet_daddr - Foreign IPv4 addr
203 * @inet_rcv_saddr - Bound local IPv4 addr
204 * @inet_dport - Destination port
205 * @inet_num - Local port
206 * @inet_flags - various atomic flags
207 * @inet_saddr - Sending source
208 * @uc_ttl - Unicast TTL
209 * @inet_sport - Source port
210 * @inet_id - ID counter for DF pkts
211 * @tos - TOS
212 * @mc_ttl - Multicasting TTL
213 * @uc_index - Unicast outgoing device index
214 * @mc_index - Multicast device index
215 * @mc_list - Group array
216 * @cork - info to build ip hdr on each ip frag while socket is corked
217 */
218 struct inet_sock {
219 /* sk and pinet6 has to be the first two members of inet_sock */
220 struct sock sk;
221 #if IS_ENABLED(CONFIG_IPV6)
222 struct ipv6_pinfo *pinet6;
223 struct ipv6_fl_socklist __rcu *ipv6_fl_list;
224 #endif
225 /* Socket demultiplex comparisons on incoming packets. */
226 #define inet_daddr sk.__sk_common.skc_daddr
227 #define inet_rcv_saddr sk.__sk_common.skc_rcv_saddr
228 #define inet_dport sk.__sk_common.skc_dport
229 #define inet_num sk.__sk_common.skc_num
230
231 unsigned long inet_flags;
232 __be32 inet_saddr;
233 __s16 uc_ttl;
234 __be16 inet_sport;
235 struct ip_options_rcu __rcu *inet_opt;
236 atomic_t inet_id;
237
238 __u8 tos;
239 __u8 min_ttl;
240 __u8 mc_ttl;
241 __u8 pmtudisc;
242 __u8 rcv_tos;
243 __u8 convert_csum;
244 int uc_index;
245 int mc_index;
246 __be32 mc_addr;
247 u32 local_port_range; /* high << 16 | low */
248
249 struct ip_mc_socklist __rcu *mc_list;
250 struct inet_cork_full cork;
251 };
252
253 #define IPCORK_OPT 1 /* ip-options has been held in ipcork.opt */
254 #define IPCORK_TS_OPT_ID 2 /* ts_opt_id field is valid, overriding sk_tskey */
255
256 enum {
257 INET_FLAGS_PKTINFO = 0,
258 INET_FLAGS_TTL = 1,
259 INET_FLAGS_TOS = 2,
260 INET_FLAGS_RECVOPTS = 3,
261 INET_FLAGS_RETOPTS = 4,
262 INET_FLAGS_PASSSEC = 5,
263 INET_FLAGS_ORIGDSTADDR = 6,
264 INET_FLAGS_CHECKSUM = 7,
265 INET_FLAGS_RECVFRAGSIZE = 8,
266
267 INET_FLAGS_RECVERR = 9,
268 INET_FLAGS_RECVERR_RFC4884 = 10,
269 INET_FLAGS_FREEBIND = 11,
270 INET_FLAGS_HDRINCL = 12,
271 INET_FLAGS_MC_LOOP = 13,
272 INET_FLAGS_MC_ALL = 14,
273 INET_FLAGS_TRANSPARENT = 15,
274 INET_FLAGS_IS_ICSK = 16,
275 INET_FLAGS_NODEFRAG = 17,
276 INET_FLAGS_BIND_ADDRESS_NO_PORT = 18,
277 INET_FLAGS_DEFER_CONNECT = 19,
278 INET_FLAGS_MC6_LOOP = 20,
279 INET_FLAGS_RECVERR6_RFC4884 = 21,
280 INET_FLAGS_MC6_ALL = 22,
281 INET_FLAGS_AUTOFLOWLABEL_SET = 23,
282 INET_FLAGS_AUTOFLOWLABEL = 24,
283 INET_FLAGS_DONTFRAG = 25,
284 INET_FLAGS_RECVERR6 = 26,
285 INET_FLAGS_REPFLOW = 27,
286 INET_FLAGS_RTALERT_ISOLATE = 28,
287 INET_FLAGS_SNDFLOW = 29,
288 INET_FLAGS_RTALERT = 30,
289 };
290
291 /* cmsg flags for inet */
292 #define IP_CMSG_PKTINFO BIT(INET_FLAGS_PKTINFO)
293 #define IP_CMSG_TTL BIT(INET_FLAGS_TTL)
294 #define IP_CMSG_TOS BIT(INET_FLAGS_TOS)
295 #define IP_CMSG_RECVOPTS BIT(INET_FLAGS_RECVOPTS)
296 #define IP_CMSG_RETOPTS BIT(INET_FLAGS_RETOPTS)
297 #define IP_CMSG_PASSSEC BIT(INET_FLAGS_PASSSEC)
298 #define IP_CMSG_ORIGDSTADDR BIT(INET_FLAGS_ORIGDSTADDR)
299 #define IP_CMSG_CHECKSUM BIT(INET_FLAGS_CHECKSUM)
300 #define IP_CMSG_RECVFRAGSIZE BIT(INET_FLAGS_RECVFRAGSIZE)
301
302 #define IP_CMSG_ALL (IP_CMSG_PKTINFO | IP_CMSG_TTL | \
303 IP_CMSG_TOS | IP_CMSG_RECVOPTS | \
304 IP_CMSG_RETOPTS | IP_CMSG_PASSSEC | \
305 IP_CMSG_ORIGDSTADDR | IP_CMSG_CHECKSUM | \
306 IP_CMSG_RECVFRAGSIZE)
307
inet_cmsg_flags(const struct inet_sock * inet)308 static inline unsigned long inet_cmsg_flags(const struct inet_sock *inet)
309 {
310 return READ_ONCE(inet->inet_flags) & IP_CMSG_ALL;
311 }
312
inet_sk_dscp(const struct inet_sock * inet)313 static inline dscp_t inet_sk_dscp(const struct inet_sock *inet)
314 {
315 return inet_dsfield_to_dscp(READ_ONCE(inet->tos));
316 }
317
318 #define inet_test_bit(nr, sk) \
319 test_bit(INET_FLAGS_##nr, &inet_sk(sk)->inet_flags)
320 #define inet_set_bit(nr, sk) \
321 set_bit(INET_FLAGS_##nr, &inet_sk(sk)->inet_flags)
322 #define inet_clear_bit(nr, sk) \
323 clear_bit(INET_FLAGS_##nr, &inet_sk(sk)->inet_flags)
324 #define inet_assign_bit(nr, sk, val) \
325 assign_bit(INET_FLAGS_##nr, &inet_sk(sk)->inet_flags, val)
326
327 /**
328 * sk_to_full_sk - Access to a full socket
329 * @sk: pointer to a socket
330 *
331 * SYNACK messages might be attached to request sockets.
332 * Some places want to reach the listener in this case.
333 */
sk_to_full_sk(struct sock * sk)334 static inline struct sock *sk_to_full_sk(struct sock *sk)
335 {
336 #ifdef CONFIG_INET
337 if (sk && READ_ONCE(sk->sk_state) == TCP_NEW_SYN_RECV)
338 sk = inet_reqsk(sk)->rsk_listener;
339 if (sk && READ_ONCE(sk->sk_state) == TCP_TIME_WAIT)
340 sk = NULL;
341 #endif
342 return sk;
343 }
344
345 /* sk_to_full_sk() variant with a const argument */
sk_const_to_full_sk(const struct sock * sk)346 static inline const struct sock *sk_const_to_full_sk(const struct sock *sk)
347 {
348 #ifdef CONFIG_INET
349 if (sk && READ_ONCE(sk->sk_state) == TCP_NEW_SYN_RECV)
350 sk = ((const struct request_sock *)sk)->rsk_listener;
351 if (sk && READ_ONCE(sk->sk_state) == TCP_TIME_WAIT)
352 sk = NULL;
353 #endif
354 return sk;
355 }
356
skb_to_full_sk(const struct sk_buff * skb)357 static inline struct sock *skb_to_full_sk(const struct sk_buff *skb)
358 {
359 return sk_to_full_sk(skb->sk);
360 }
361
362 #define inet_sk(ptr) container_of_const(ptr, struct inet_sock, sk)
363
364 int inet_sk_rebuild_header(struct sock *sk);
365
366 /**
367 * inet_sk_state_load - read sk->sk_state for lockless contexts
368 * @sk: socket pointer
369 *
370 * Paired with inet_sk_state_store(). Used in places we don't hold socket lock:
371 * tcp_diag_get_info(), tcp_get_info(), tcp_poll(), get_tcp4_sock() ...
372 */
inet_sk_state_load(const struct sock * sk)373 static inline int inet_sk_state_load(const struct sock *sk)
374 {
375 /* state change might impact lockless readers. */
376 return smp_load_acquire(&sk->sk_state);
377 }
378
379 /**
380 * inet_sk_state_store - update sk->sk_state
381 * @sk: socket pointer
382 * @newstate: new state
383 *
384 * Paired with inet_sk_state_load(). Should be used in contexts where
385 * state change might impact lockless readers.
386 */
387 void inet_sk_state_store(struct sock *sk, int newstate);
388
389 void inet_sk_set_state(struct sock *sk, int state);
390
__inet_ehashfn(const __be32 laddr,const __u16 lport,const __be32 faddr,const __be16 fport,u32 initval)391 static inline unsigned int __inet_ehashfn(const __be32 laddr,
392 const __u16 lport,
393 const __be32 faddr,
394 const __be16 fport,
395 u32 initval)
396 {
397 return jhash_3words((__force __u32) laddr,
398 (__force __u32) faddr,
399 ((__u32) lport) << 16 | (__force __u32)fport,
400 initval);
401 }
402
403 struct request_sock *inet_reqsk_alloc(const struct request_sock_ops *ops,
404 struct sock *sk_listener,
405 bool attach_listener);
406
inet_sk_flowi_flags(const struct sock * sk)407 static inline __u8 inet_sk_flowi_flags(const struct sock *sk)
408 {
409 __u8 flags = 0;
410
411 if (inet_test_bit(TRANSPARENT, sk) || inet_test_bit(HDRINCL, sk))
412 flags |= FLOWI_FLAG_ANYSRC;
413 return flags;
414 }
415
inet_inc_convert_csum(struct sock * sk)416 static inline void inet_inc_convert_csum(struct sock *sk)
417 {
418 inet_sk(sk)->convert_csum++;
419 }
420
inet_dec_convert_csum(struct sock * sk)421 static inline void inet_dec_convert_csum(struct sock *sk)
422 {
423 if (inet_sk(sk)->convert_csum > 0)
424 inet_sk(sk)->convert_csum--;
425 }
426
inet_get_convert_csum(struct sock * sk)427 static inline bool inet_get_convert_csum(struct sock *sk)
428 {
429 return !!inet_sk(sk)->convert_csum;
430 }
431
432
inet_can_nonlocal_bind(struct net * net,struct inet_sock * inet)433 static inline bool inet_can_nonlocal_bind(struct net *net,
434 struct inet_sock *inet)
435 {
436 return READ_ONCE(net->ipv4.sysctl_ip_nonlocal_bind) ||
437 test_bit(INET_FLAGS_FREEBIND, &inet->inet_flags) ||
438 test_bit(INET_FLAGS_TRANSPARENT, &inet->inet_flags);
439 }
440
inet_addr_valid_or_nonlocal(struct net * net,struct inet_sock * inet,__be32 addr,int addr_type)441 static inline bool inet_addr_valid_or_nonlocal(struct net *net,
442 struct inet_sock *inet,
443 __be32 addr,
444 int addr_type)
445 {
446 return inet_can_nonlocal_bind(net, inet) ||
447 addr == htonl(INADDR_ANY) ||
448 addr_type == RTN_LOCAL ||
449 addr_type == RTN_MULTICAST ||
450 addr_type == RTN_BROADCAST;
451 }
452
453 #endif /* _INET_SOCK_H */
454