]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * INET An implementation of the TCP/IP protocol suite for the LINUX | |
3 | * operating system. INET is implemented using the BSD Socket | |
4 | * interface as the means of communication with the user level. | |
5 | * | |
6 | * Definitions for the AF_INET socket handler. | |
7 | * | |
8 | * Version: @(#)sock.h 1.0.4 05/13/93 | |
9 | * | |
02c30a84 | 10 | * Authors: Ross Biro |
1da177e4 LT |
11 | * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> |
12 | * Corey Minyard <wf-rch!minyard@relay.EU.net> | |
13 | * Florian La Roche <flla@stud.uni-sb.de> | |
14 | * | |
15 | * Fixes: | |
16 | * Alan Cox : Volatiles in skbuff pointers. See | |
17 | * skbuff comments. May be overdone, | |
18 | * better to prove they can be removed | |
19 | * than the reverse. | |
20 | * Alan Cox : Added a zapped field for tcp to note | |
21 | * a socket is reset and must stay shut up | |
22 | * Alan Cox : New fields for options | |
23 | * Pauline Middelink : identd support | |
24 | * Alan Cox : Eliminate low level recv/recvfrom | |
25 | * David S. Miller : New socket lookup architecture. | |
26 | * Steve Whitehouse: Default routines for sock_ops | |
27 | * Arnaldo C. Melo : removed net_pinfo, tp_pinfo and made | |
28 | * protinfo be just a void pointer, as the | |
29 | * protocol specific parts were moved to | |
30 | * respective headers and ipv4/v6, etc now | |
31 | * use private slabcaches for its socks | |
32 | * Pedro Hortas : New flags field for socket options | |
33 | * | |
34 | * | |
35 | * This program is free software; you can redistribute it and/or | |
36 | * modify it under the terms of the GNU General Public License | |
37 | * as published by the Free Software Foundation; either version | |
38 | * 2 of the License, or (at your option) any later version. | |
39 | */ | |
40 | #ifndef _SOCK_H | |
41 | #define _SOCK_H | |
42 | ||
a6b7a407 | 43 | #include <linux/hardirq.h> |
172589cc | 44 | #include <linux/kernel.h> |
1da177e4 | 45 | #include <linux/list.h> |
88ab1932 | 46 | #include <linux/list_nulls.h> |
1da177e4 LT |
47 | #include <linux/timer.h> |
48 | #include <linux/cache.h> | |
3f134619 | 49 | #include <linux/bitops.h> |
a5b5bb9a | 50 | #include <linux/lockdep.h> |
1da177e4 LT |
51 | #include <linux/netdevice.h> |
52 | #include <linux/skbuff.h> /* struct sk_buff */ | |
d7fe0f24 | 53 | #include <linux/mm.h> |
1da177e4 | 54 | #include <linux/security.h> |
5a0e3ad6 | 55 | #include <linux/slab.h> |
c6e1a0d1 | 56 | #include <linux/uaccess.h> |
3e32cb2e | 57 | #include <linux/page_counter.h> |
180d8cd9 | 58 | #include <linux/memcontrol.h> |
c5905afb | 59 | #include <linux/static_key.h> |
40401530 | 60 | #include <linux/sched.h> |
1ce0bf50 | 61 | #include <linux/wait.h> |
2a56a1fe | 62 | #include <linux/cgroup-defs.h> |
1da177e4 LT |
63 | |
64 | #include <linux/filter.h> | |
88ab1932 | 65 | #include <linux/rculist_nulls.h> |
a57de0b4 | 66 | #include <linux/poll.h> |
1da177e4 | 67 | |
c31504dc | 68 | #include <linux/atomic.h> |
1da177e4 LT |
69 | #include <net/dst.h> |
70 | #include <net/checksum.h> | |
1d0ab253 | 71 | #include <net/tcp_states.h> |
b9f40e21 | 72 | #include <linux/net_tstamp.h> |
1da177e4 | 73 | |
9f048bfb ED |
74 | struct cgroup; |
75 | struct cgroup_subsys; | |
c607b2ed | 76 | #ifdef CONFIG_NET |
1d62e436 GC |
77 | int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss); |
78 | void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg); | |
c607b2ed GC |
79 | #else |
80 | static inline | |
1d62e436 | 81 | int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss) |
c607b2ed GC |
82 | { |
83 | return 0; | |
84 | } | |
85 | static inline | |
1d62e436 | 86 | void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg) |
c607b2ed GC |
87 | { |
88 | } | |
89 | #endif | |
1da177e4 LT |
90 | /* |
91 | * This structure really needs to be cleaned up. | |
92 | * Most of it is for TCP, and not used by any of | |
93 | * the other protocols. | |
94 | */ | |
95 | ||
96 | /* Define this to get the SOCK_DBG debugging facility. */ | |
97 | #define SOCK_DEBUGGING | |
98 | #ifdef SOCK_DEBUGGING | |
99 | #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \ | |
100 | printk(KERN_DEBUG msg); } while (0) | |
101 | #else | |
4cd9029d | 102 | /* Validate arguments and do nothing */ |
b9075fa9 | 103 | static inline __printf(2, 3) |
dc6b9b78 | 104 | void SOCK_DEBUG(const struct sock *sk, const char *msg, ...) |
4cd9029d SH |
105 | { |
106 | } | |
1da177e4 LT |
107 | #endif |
108 | ||
109 | /* This is the per-socket lock. The spinlock provides a synchronization | |
110 | * between user contexts and software interrupt processing, whereas the | |
111 | * mini-semaphore synchronizes multiple users amongst themselves. | |
112 | */ | |
1da177e4 LT |
113 | typedef struct { |
114 | spinlock_t slock; | |
d2e9117c | 115 | int owned; |
1da177e4 | 116 | wait_queue_head_t wq; |
a5b5bb9a IM |
117 | /* |
118 | * We express the mutex-alike socket_lock semantics | |
119 | * to the lock validator by explicitly managing | |
120 | * the slock as a lock variant (in addition to | |
121 | * the slock itself): | |
122 | */ | |
123 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
124 | struct lockdep_map dep_map; | |
125 | #endif | |
1da177e4 LT |
126 | } socket_lock_t; |
127 | ||
1da177e4 | 128 | struct sock; |
8feaf0c0 | 129 | struct proto; |
0eeb8ffc | 130 | struct net; |
1da177e4 | 131 | |
077b393d ED |
132 | typedef __u32 __bitwise __portpair; |
133 | typedef __u64 __bitwise __addrpair; | |
134 | ||
1da177e4 | 135 | /** |
4dc3b16b | 136 | * struct sock_common - minimal network layer representation of sockets |
68835aba ED |
137 | * @skc_daddr: Foreign IPv4 addr |
138 | * @skc_rcv_saddr: Bound local IPv4 addr | |
4dc6dc71 | 139 | * @skc_hash: hash value used with various protocol lookup tables |
d4cada4a | 140 | * @skc_u16hashes: two u16 hash values used by UDP lookup tables |
ce43b03e ED |
141 | * @skc_dport: placeholder for inet_dport/tw_dport |
142 | * @skc_num: placeholder for inet_num/tw_num | |
4dc3b16b PP |
143 | * @skc_family: network address family |
144 | * @skc_state: Connection state | |
145 | * @skc_reuse: %SO_REUSEADDR setting | |
055dc21a | 146 | * @skc_reuseport: %SO_REUSEPORT setting |
4dc3b16b | 147 | * @skc_bound_dev_if: bound device index if != 0 |
4dc3b16b | 148 | * @skc_bind_node: bind hash linkage for various protocol lookup tables |
512615b6 | 149 | * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol |
8feaf0c0 | 150 | * @skc_prot: protocol handlers inside a network family |
07feaebf | 151 | * @skc_net: reference to the network namespace of this socket |
68835aba ED |
152 | * @skc_node: main hash linkage for various protocol lookup tables |
153 | * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol | |
154 | * @skc_tx_queue_mapping: tx queue number for this connection | |
8e5eb54d ED |
155 | * @skc_flags: place holder for sk_flags |
156 | * %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, | |
157 | * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings | |
70da268b | 158 | * @skc_incoming_cpu: record/match cpu processing incoming packets |
68835aba | 159 | * @skc_refcnt: reference count |
4dc3b16b PP |
160 | * |
161 | * This is the minimal network layer representation of sockets, the header | |
8feaf0c0 ACM |
162 | * for struct sock and struct inet_timewait_sock. |
163 | */ | |
1da177e4 | 164 | struct sock_common { |
ce43b03e | 165 | /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned |
05dbc7b5 | 166 | * address on 64bit arches : cf INET_MATCH() |
4dc6dc71 | 167 | */ |
ce43b03e | 168 | union { |
077b393d | 169 | __addrpair skc_addrpair; |
ce43b03e ED |
170 | struct { |
171 | __be32 skc_daddr; | |
172 | __be32 skc_rcv_saddr; | |
173 | }; | |
174 | }; | |
d4cada4a ED |
175 | union { |
176 | unsigned int skc_hash; | |
177 | __u16 skc_u16hashes[2]; | |
178 | }; | |
ce43b03e ED |
179 | /* skc_dport && skc_num must be grouped as well */ |
180 | union { | |
077b393d | 181 | __portpair skc_portpair; |
ce43b03e ED |
182 | struct { |
183 | __be16 skc_dport; | |
184 | __u16 skc_num; | |
185 | }; | |
186 | }; | |
187 | ||
4dc6dc71 ED |
188 | unsigned short skc_family; |
189 | volatile unsigned char skc_state; | |
055dc21a | 190 | unsigned char skc_reuse:4; |
9fe516ba ED |
191 | unsigned char skc_reuseport:1; |
192 | unsigned char skc_ipv6only:1; | |
26abe143 | 193 | unsigned char skc_net_refcnt:1; |
4dc6dc71 | 194 | int skc_bound_dev_if; |
512615b6 ED |
195 | union { |
196 | struct hlist_node skc_bind_node; | |
197 | struct hlist_nulls_node skc_portaddr_node; | |
198 | }; | |
8feaf0c0 | 199 | struct proto *skc_prot; |
0c5c9fb5 | 200 | possible_net_t skc_net; |
efe4208f ED |
201 | |
202 | #if IS_ENABLED(CONFIG_IPV6) | |
203 | struct in6_addr skc_v6_daddr; | |
204 | struct in6_addr skc_v6_rcv_saddr; | |
205 | #endif | |
206 | ||
33cf7c90 ED |
207 | atomic64_t skc_cookie; |
208 | ||
8e5eb54d ED |
209 | /* following fields are padding to force |
210 | * offset(struct sock, sk_refcnt) == 128 on 64bit arches | |
211 | * assuming IPV6 is enabled. We use this padding differently | |
212 | * for different kind of 'sockets' | |
213 | */ | |
214 | union { | |
215 | unsigned long skc_flags; | |
216 | struct sock *skc_listener; /* request_sock */ | |
217 | struct inet_timewait_death_row *skc_tw_dr; /* inet_timewait_sock */ | |
218 | }; | |
68835aba ED |
219 | /* |
220 | * fields between dontcopy_begin/dontcopy_end | |
221 | * are not copied in sock_copy() | |
222 | */ | |
928c41e7 | 223 | /* private: */ |
68835aba | 224 | int skc_dontcopy_begin[0]; |
928c41e7 | 225 | /* public: */ |
68835aba ED |
226 | union { |
227 | struct hlist_node skc_node; | |
228 | struct hlist_nulls_node skc_nulls_node; | |
229 | }; | |
230 | int skc_tx_queue_mapping; | |
ed53d0ab ED |
231 | union { |
232 | int skc_incoming_cpu; | |
233 | u32 skc_rcv_wnd; | |
d475f090 | 234 | u32 skc_tw_rcv_nxt; /* struct tcp_timewait_sock */ |
ed53d0ab | 235 | }; |
70da268b | 236 | |
68835aba | 237 | atomic_t skc_refcnt; |
928c41e7 | 238 | /* private: */ |
68835aba | 239 | int skc_dontcopy_end[0]; |
ed53d0ab ED |
240 | union { |
241 | u32 skc_rxhash; | |
242 | u32 skc_window_clamp; | |
d475f090 | 243 | u32 skc_tw_snd_nxt; /* struct tcp_timewait_sock */ |
ed53d0ab | 244 | }; |
928c41e7 | 245 | /* public: */ |
1da177e4 LT |
246 | }; |
247 | ||
e1aab161 | 248 | struct cg_proto; |
1da177e4 LT |
249 | /** |
250 | * struct sock - network layer representation of sockets | |
8feaf0c0 | 251 | * @__sk_common: shared layout with inet_timewait_sock |
4dc3b16b PP |
252 | * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN |
253 | * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings | |
254 | * @sk_lock: synchronizer | |
255 | * @sk_rcvbuf: size of receive buffer in bytes | |
43815482 | 256 | * @sk_wq: sock wait queue and async head |
421b3885 | 257 | * @sk_rx_dst: receive input route used by early demux |
4dc3b16b | 258 | * @sk_dst_cache: destination cache |
4dc3b16b | 259 | * @sk_policy: flow policy |
4dc3b16b PP |
260 | * @sk_receive_queue: incoming packets |
261 | * @sk_wmem_alloc: transmit queue bytes committed | |
262 | * @sk_write_queue: Packet sending queue | |
263 | * @sk_omem_alloc: "o" is "option" or "other" | |
264 | * @sk_wmem_queued: persistent queue size | |
265 | * @sk_forward_alloc: space allocated forward | |
06021292 | 266 | * @sk_napi_id: id of the last napi context to receive data for sk |
dafcc438 | 267 | * @sk_ll_usec: usecs to busypoll when there is no data |
4dc3b16b | 268 | * @sk_allocation: allocation mode |
95bd09eb | 269 | * @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler) |
c3f40d7c | 270 | * @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE) |
4dc3b16b | 271 | * @sk_sndbuf: size of send buffer in bytes |
28448b80 TH |
272 | * @sk_no_check_tx: %SO_NO_CHECK setting, set checksum in TX packets |
273 | * @sk_no_check_rx: allow zero checksum in RX packets | |
4dc3b16b | 274 | * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO) |
a465419b | 275 | * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK) |
bcd76111 | 276 | * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4) |
82cc1a7a | 277 | * @sk_gso_max_size: Maximum GSO segment size to build |
1485348d | 278 | * @sk_gso_max_segs: Maximum number of GSO segments |
4dc3b16b | 279 | * @sk_lingertime: %SO_LINGER l_linger setting |
4dc3b16b PP |
280 | * @sk_backlog: always used with the per-socket spinlock held |
281 | * @sk_callback_lock: used with the callbacks in the end of this struct | |
282 | * @sk_error_queue: rarely used | |
33c732c3 WC |
283 | * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, |
284 | * IPV6_ADDRFORM for instance) | |
4dc3b16b | 285 | * @sk_err: last error |
33c732c3 WC |
286 | * @sk_err_soft: errors that don't cause failure but are the cause of a |
287 | * persistent failure not just 'timed out' | |
cb61cb9b | 288 | * @sk_drops: raw/udp drops counter |
4dc3b16b PP |
289 | * @sk_ack_backlog: current listen backlog |
290 | * @sk_max_ack_backlog: listen backlog set in listen() | |
291 | * @sk_priority: %SO_PRIORITY setting | |
292 | * @sk_type: socket type (%SOCK_STREAM, etc) | |
293 | * @sk_protocol: which protocol this socket belongs in this network family | |
53c3fa20 RD |
294 | * @sk_peer_pid: &struct pid for this socket's peer |
295 | * @sk_peer_cred: %SO_PEERCRED setting | |
4dc3b16b PP |
296 | * @sk_rcvlowat: %SO_RCVLOWAT setting |
297 | * @sk_rcvtimeo: %SO_RCVTIMEO setting | |
298 | * @sk_sndtimeo: %SO_SNDTIMEO setting | |
b73c3d0e | 299 | * @sk_txhash: computed flow hash for use on transmit |
4dc3b16b | 300 | * @sk_filter: socket filtering instructions |
4dc3b16b PP |
301 | * @sk_timer: sock cleanup timer |
302 | * @sk_stamp: time stamp of last packet received | |
b9f40e21 | 303 | * @sk_tsflags: SO_TIMESTAMPING socket options |
09c2d251 | 304 | * @sk_tskey: counter to disambiguate concurrent tstamp requests |
4dc3b16b PP |
305 | * @sk_socket: Identd and reporting IO signals |
306 | * @sk_user_data: RPC layer private data | |
5640f768 | 307 | * @sk_frag: cached page frag |
d3d4f0a0 | 308 | * @sk_peek_off: current peek_offset value |
4dc3b16b | 309 | * @sk_send_head: front of stuff to transmit |
67be2dd1 | 310 | * @sk_security: used by security modules |
31729363 | 311 | * @sk_mark: generic packet mark |
2a56a1fe | 312 | * @sk_cgrp_data: cgroup data for this cgroup |
e1aab161 | 313 | * @sk_cgrp: this socket's cgroup-specific proto data |
4dc3b16b PP |
314 | * @sk_write_pending: a write to stream socket waits to start |
315 | * @sk_state_change: callback to indicate change in the state of the sock | |
316 | * @sk_data_ready: callback to indicate there is data to be processed | |
317 | * @sk_write_space: callback to indicate there is bf sending space available | |
318 | * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE) | |
319 | * @sk_backlog_rcv: callback to process the backlog | |
320 | * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0 | |
ef456144 | 321 | * @sk_reuseport_cb: reuseport group container |
1da177e4 LT |
322 | */ |
323 | struct sock { | |
324 | /* | |
8feaf0c0 | 325 | * Now struct inet_timewait_sock also uses sock_common, so please just |
1da177e4 LT |
326 | * don't add nothing before this first member (__sk_common) --acme |
327 | */ | |
328 | struct sock_common __sk_common; | |
4dc6dc71 ED |
329 | #define sk_node __sk_common.skc_node |
330 | #define sk_nulls_node __sk_common.skc_nulls_node | |
331 | #define sk_refcnt __sk_common.skc_refcnt | |
e022f0b4 | 332 | #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping |
4dc6dc71 | 333 | |
68835aba ED |
334 | #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin |
335 | #define sk_dontcopy_end __sk_common.skc_dontcopy_end | |
4dc6dc71 | 336 | #define sk_hash __sk_common.skc_hash |
50805466 | 337 | #define sk_portpair __sk_common.skc_portpair |
05dbc7b5 ED |
338 | #define sk_num __sk_common.skc_num |
339 | #define sk_dport __sk_common.skc_dport | |
50805466 ED |
340 | #define sk_addrpair __sk_common.skc_addrpair |
341 | #define sk_daddr __sk_common.skc_daddr | |
342 | #define sk_rcv_saddr __sk_common.skc_rcv_saddr | |
1da177e4 LT |
343 | #define sk_family __sk_common.skc_family |
344 | #define sk_state __sk_common.skc_state | |
345 | #define sk_reuse __sk_common.skc_reuse | |
055dc21a | 346 | #define sk_reuseport __sk_common.skc_reuseport |
9fe516ba | 347 | #define sk_ipv6only __sk_common.skc_ipv6only |
26abe143 | 348 | #define sk_net_refcnt __sk_common.skc_net_refcnt |
1da177e4 | 349 | #define sk_bound_dev_if __sk_common.skc_bound_dev_if |
1da177e4 | 350 | #define sk_bind_node __sk_common.skc_bind_node |
8feaf0c0 | 351 | #define sk_prot __sk_common.skc_prot |
07feaebf | 352 | #define sk_net __sk_common.skc_net |
efe4208f ED |
353 | #define sk_v6_daddr __sk_common.skc_v6_daddr |
354 | #define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr | |
33cf7c90 | 355 | #define sk_cookie __sk_common.skc_cookie |
70da268b | 356 | #define sk_incoming_cpu __sk_common.skc_incoming_cpu |
8e5eb54d | 357 | #define sk_flags __sk_common.skc_flags |
ed53d0ab | 358 | #define sk_rxhash __sk_common.skc_rxhash |
efe4208f | 359 | |
1da177e4 | 360 | socket_lock_t sk_lock; |
b178bb3d | 361 | struct sk_buff_head sk_receive_queue; |
fa438ccf ED |
362 | /* |
363 | * The backlog queue is special, it is always used with | |
364 | * the per-socket spinlock held and requires low latency | |
365 | * access. Therefore we special case it's implementation. | |
b178bb3d ED |
366 | * Note : rmem_alloc is in this structure to fill a hole |
367 | * on 64bit arches, not because its logically part of | |
368 | * backlog. | |
fa438ccf ED |
369 | */ |
370 | struct { | |
b178bb3d ED |
371 | atomic_t rmem_alloc; |
372 | int len; | |
373 | struct sk_buff *head; | |
374 | struct sk_buff *tail; | |
fa438ccf | 375 | } sk_backlog; |
b178bb3d ED |
376 | #define sk_rmem_alloc sk_backlog.rmem_alloc |
377 | int sk_forward_alloc; | |
2c8c56e1 | 378 | |
b73c3d0e | 379 | __u32 sk_txhash; |
e0d1095a | 380 | #ifdef CONFIG_NET_RX_BUSY_POLL |
06021292 | 381 | unsigned int sk_napi_id; |
dafcc438 | 382 | unsigned int sk_ll_usec; |
b178bb3d ED |
383 | #endif |
384 | atomic_t sk_drops; | |
385 | int sk_rcvbuf; | |
386 | ||
387 | struct sk_filter __rcu *sk_filter; | |
ceb5d58b ED |
388 | union { |
389 | struct socket_wq __rcu *sk_wq; | |
390 | struct socket_wq *sk_wq_raw; | |
391 | }; | |
def8b4fa | 392 | #ifdef CONFIG_XFRM |
d188ba86 | 393 | struct xfrm_policy __rcu *sk_policy[2]; |
def8b4fa | 394 | #endif |
deaa5854 | 395 | struct dst_entry *sk_rx_dst; |
0e36cbb3 | 396 | struct dst_entry __rcu *sk_dst_cache; |
6bd4f355 | 397 | /* Note: 32bit hole on 64bit arches */ |
1da177e4 LT |
398 | atomic_t sk_wmem_alloc; |
399 | atomic_t sk_omem_alloc; | |
4e07a91c | 400 | int sk_sndbuf; |
1da177e4 | 401 | struct sk_buff_head sk_write_queue; |
b178bb3d ED |
402 | kmemcheck_bitfield_begin(flags); |
403 | unsigned int sk_shutdown : 2, | |
28448b80 TH |
404 | sk_no_check_tx : 1, |
405 | sk_no_check_rx : 1, | |
b178bb3d ED |
406 | sk_userlocks : 4, |
407 | sk_protocol : 8, | |
408 | sk_type : 16; | |
7bbadd2d | 409 | #define SK_PROTOCOL_MAX U8_MAX |
b178bb3d | 410 | kmemcheck_bitfield_end(flags); |
1da177e4 | 411 | int sk_wmem_queued; |
7d877f3b | 412 | gfp_t sk_allocation; |
95bd09eb | 413 | u32 sk_pacing_rate; /* bytes per second */ |
62748f32 | 414 | u32 sk_max_pacing_rate; |
c8f44aff MM |
415 | netdev_features_t sk_route_caps; |
416 | netdev_features_t sk_route_nocaps; | |
bcd76111 | 417 | int sk_gso_type; |
82cc1a7a | 418 | unsigned int sk_gso_max_size; |
1485348d | 419 | u16 sk_gso_max_segs; |
9932cf95 | 420 | int sk_rcvlowat; |
1da177e4 | 421 | unsigned long sk_lingertime; |
1da177e4 | 422 | struct sk_buff_head sk_error_queue; |
476e19cf | 423 | struct proto *sk_prot_creator; |
1da177e4 LT |
424 | rwlock_t sk_callback_lock; |
425 | int sk_err, | |
426 | sk_err_soft; | |
becb74f0 ED |
427 | u32 sk_ack_backlog; |
428 | u32 sk_max_ack_backlog; | |
1da177e4 | 429 | __u32 sk_priority; |
297dbde1 | 430 | __u32 sk_mark; |
109f6e39 EB |
431 | struct pid *sk_peer_pid; |
432 | const struct cred *sk_peer_cred; | |
1da177e4 LT |
433 | long sk_rcvtimeo; |
434 | long sk_sndtimeo; | |
1da177e4 | 435 | struct timer_list sk_timer; |
b7aa0bf7 | 436 | ktime_t sk_stamp; |
b9f40e21 | 437 | u16 sk_tsflags; |
09c2d251 | 438 | u32 sk_tskey; |
1da177e4 LT |
439 | struct socket *sk_socket; |
440 | void *sk_user_data; | |
5640f768 | 441 | struct page_frag sk_frag; |
1da177e4 | 442 | struct sk_buff *sk_send_head; |
ef64a54f | 443 | __s32 sk_peek_off; |
1da177e4 | 444 | int sk_write_pending; |
d5f64238 | 445 | #ifdef CONFIG_SECURITY |
1da177e4 | 446 | void *sk_security; |
d5f64238 | 447 | #endif |
2a56a1fe | 448 | struct sock_cgroup_data sk_cgrp_data; |
e1aab161 | 449 | struct cg_proto *sk_cgrp; |
1da177e4 | 450 | void (*sk_state_change)(struct sock *sk); |
676d2369 | 451 | void (*sk_data_ready)(struct sock *sk); |
1da177e4 LT |
452 | void (*sk_write_space)(struct sock *sk); |
453 | void (*sk_error_report)(struct sock *sk); | |
dc6b9b78 ED |
454 | int (*sk_backlog_rcv)(struct sock *sk, |
455 | struct sk_buff *skb); | |
1da177e4 | 456 | void (*sk_destruct)(struct sock *sk); |
ef456144 | 457 | struct sock_reuseport __rcu *sk_reuseport_cb; |
1da177e4 LT |
458 | }; |
459 | ||
559835ea PS |
460 | #define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data))) |
461 | ||
462 | #define rcu_dereference_sk_user_data(sk) rcu_dereference(__sk_user_data((sk))) | |
463 | #define rcu_assign_sk_user_data(sk, ptr) rcu_assign_pointer(__sk_user_data((sk)), ptr) | |
464 | ||
4a17fd52 PE |
465 | /* |
466 | * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK | |
467 | * or not whether his port will be reused by someone else. SK_FORCE_REUSE | |
468 | * on a socket means that the socket will reuse everybody else's port | |
469 | * without looking at the other's sk_reuse value. | |
470 | */ | |
471 | ||
472 | #define SK_NO_REUSE 0 | |
473 | #define SK_CAN_REUSE 1 | |
474 | #define SK_FORCE_REUSE 2 | |
475 | ||
ef64a54f PE |
476 | static inline int sk_peek_offset(struct sock *sk, int flags) |
477 | { | |
478 | if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0)) | |
479 | return sk->sk_peek_off; | |
480 | else | |
481 | return 0; | |
482 | } | |
483 | ||
484 | static inline void sk_peek_offset_bwd(struct sock *sk, int val) | |
485 | { | |
486 | if (sk->sk_peek_off >= 0) { | |
487 | if (sk->sk_peek_off >= val) | |
488 | sk->sk_peek_off -= val; | |
489 | else | |
490 | sk->sk_peek_off = 0; | |
491 | } | |
492 | } | |
493 | ||
494 | static inline void sk_peek_offset_fwd(struct sock *sk, int val) | |
495 | { | |
496 | if (sk->sk_peek_off >= 0) | |
497 | sk->sk_peek_off += val; | |
498 | } | |
499 | ||
1da177e4 LT |
500 | /* |
501 | * Hashed lists helper routines | |
502 | */ | |
c4146644 LZ |
503 | static inline struct sock *sk_entry(const struct hlist_node *node) |
504 | { | |
505 | return hlist_entry(node, struct sock, sk_node); | |
506 | } | |
507 | ||
e48c414e | 508 | static inline struct sock *__sk_head(const struct hlist_head *head) |
1da177e4 LT |
509 | { |
510 | return hlist_entry(head->first, struct sock, sk_node); | |
511 | } | |
512 | ||
e48c414e | 513 | static inline struct sock *sk_head(const struct hlist_head *head) |
1da177e4 LT |
514 | { |
515 | return hlist_empty(head) ? NULL : __sk_head(head); | |
516 | } | |
517 | ||
88ab1932 ED |
518 | static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head) |
519 | { | |
520 | return hlist_nulls_entry(head->first, struct sock, sk_nulls_node); | |
521 | } | |
522 | ||
523 | static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head) | |
524 | { | |
525 | return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head); | |
526 | } | |
527 | ||
e48c414e | 528 | static inline struct sock *sk_next(const struct sock *sk) |
1da177e4 LT |
529 | { |
530 | return sk->sk_node.next ? | |
531 | hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL; | |
532 | } | |
533 | ||
88ab1932 ED |
534 | static inline struct sock *sk_nulls_next(const struct sock *sk) |
535 | { | |
536 | return (!is_a_nulls(sk->sk_nulls_node.next)) ? | |
537 | hlist_nulls_entry(sk->sk_nulls_node.next, | |
538 | struct sock, sk_nulls_node) : | |
539 | NULL; | |
540 | } | |
541 | ||
dc6b9b78 | 542 | static inline bool sk_unhashed(const struct sock *sk) |
1da177e4 LT |
543 | { |
544 | return hlist_unhashed(&sk->sk_node); | |
545 | } | |
546 | ||
dc6b9b78 | 547 | static inline bool sk_hashed(const struct sock *sk) |
1da177e4 | 548 | { |
da753bea | 549 | return !sk_unhashed(sk); |
1da177e4 LT |
550 | } |
551 | ||
dc6b9b78 | 552 | static inline void sk_node_init(struct hlist_node *node) |
1da177e4 LT |
553 | { |
554 | node->pprev = NULL; | |
555 | } | |
556 | ||
dc6b9b78 | 557 | static inline void sk_nulls_node_init(struct hlist_nulls_node *node) |
88ab1932 ED |
558 | { |
559 | node->pprev = NULL; | |
560 | } | |
561 | ||
dc6b9b78 | 562 | static inline void __sk_del_node(struct sock *sk) |
1da177e4 LT |
563 | { |
564 | __hlist_del(&sk->sk_node); | |
565 | } | |
566 | ||
808f5114 | 567 | /* NB: equivalent to hlist_del_init_rcu */ |
dc6b9b78 | 568 | static inline bool __sk_del_node_init(struct sock *sk) |
1da177e4 LT |
569 | { |
570 | if (sk_hashed(sk)) { | |
571 | __sk_del_node(sk); | |
572 | sk_node_init(&sk->sk_node); | |
dc6b9b78 | 573 | return true; |
1da177e4 | 574 | } |
dc6b9b78 | 575 | return false; |
1da177e4 LT |
576 | } |
577 | ||
578 | /* Grab socket reference count. This operation is valid only | |
579 | when sk is ALREADY grabbed f.e. it is found in hash table | |
580 | or a list and the lookup is made under lock preventing hash table | |
581 | modifications. | |
582 | */ | |
583 | ||
584 | static inline void sock_hold(struct sock *sk) | |
585 | { | |
586 | atomic_inc(&sk->sk_refcnt); | |
587 | } | |
588 | ||
589 | /* Ungrab socket in the context, which assumes that socket refcnt | |
590 | cannot hit zero, f.e. it is true in context of any socketcall. | |
591 | */ | |
592 | static inline void __sock_put(struct sock *sk) | |
593 | { | |
594 | atomic_dec(&sk->sk_refcnt); | |
595 | } | |
596 | ||
dc6b9b78 | 597 | static inline bool sk_del_node_init(struct sock *sk) |
1da177e4 | 598 | { |
dc6b9b78 | 599 | bool rc = __sk_del_node_init(sk); |
1da177e4 LT |
600 | |
601 | if (rc) { | |
602 | /* paranoid for a while -acme */ | |
603 | WARN_ON(atomic_read(&sk->sk_refcnt) == 1); | |
604 | __sock_put(sk); | |
605 | } | |
606 | return rc; | |
607 | } | |
808f5114 | 608 | #define sk_del_node_init_rcu(sk) sk_del_node_init(sk) |
1da177e4 | 609 | |
dc6b9b78 | 610 | static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk) |
271b72c7 ED |
611 | { |
612 | if (sk_hashed(sk)) { | |
88ab1932 | 613 | hlist_nulls_del_init_rcu(&sk->sk_nulls_node); |
dc6b9b78 | 614 | return true; |
271b72c7 | 615 | } |
dc6b9b78 | 616 | return false; |
271b72c7 ED |
617 | } |
618 | ||
dc6b9b78 | 619 | static inline bool sk_nulls_del_node_init_rcu(struct sock *sk) |
271b72c7 | 620 | { |
dc6b9b78 | 621 | bool rc = __sk_nulls_del_node_init_rcu(sk); |
271b72c7 ED |
622 | |
623 | if (rc) { | |
624 | /* paranoid for a while -acme */ | |
625 | WARN_ON(atomic_read(&sk->sk_refcnt) == 1); | |
626 | __sock_put(sk); | |
627 | } | |
628 | return rc; | |
629 | } | |
630 | ||
dc6b9b78 | 631 | static inline void __sk_add_node(struct sock *sk, struct hlist_head *list) |
1da177e4 LT |
632 | { |
633 | hlist_add_head(&sk->sk_node, list); | |
634 | } | |
635 | ||
dc6b9b78 | 636 | static inline void sk_add_node(struct sock *sk, struct hlist_head *list) |
1da177e4 LT |
637 | { |
638 | sock_hold(sk); | |
639 | __sk_add_node(sk, list); | |
640 | } | |
641 | ||
dc6b9b78 | 642 | static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list) |
808f5114 | 643 | { |
644 | sock_hold(sk); | |
645 | hlist_add_head_rcu(&sk->sk_node, list); | |
646 | } | |
647 | ||
dc6b9b78 | 648 | static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list) |
271b72c7 | 649 | { |
88ab1932 | 650 | hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list); |
271b72c7 ED |
651 | } |
652 | ||
dc6b9b78 | 653 | static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list) |
271b72c7 ED |
654 | { |
655 | sock_hold(sk); | |
88ab1932 | 656 | __sk_nulls_add_node_rcu(sk, list); |
271b72c7 ED |
657 | } |
658 | ||
dc6b9b78 | 659 | static inline void __sk_del_bind_node(struct sock *sk) |
1da177e4 LT |
660 | { |
661 | __hlist_del(&sk->sk_bind_node); | |
662 | } | |
663 | ||
dc6b9b78 | 664 | static inline void sk_add_bind_node(struct sock *sk, |
1da177e4 LT |
665 | struct hlist_head *list) |
666 | { | |
667 | hlist_add_head(&sk->sk_bind_node, list); | |
668 | } | |
669 | ||
b67bfe0d SL |
670 | #define sk_for_each(__sk, list) \ |
671 | hlist_for_each_entry(__sk, list, sk_node) | |
672 | #define sk_for_each_rcu(__sk, list) \ | |
673 | hlist_for_each_entry_rcu(__sk, list, sk_node) | |
88ab1932 ED |
674 | #define sk_nulls_for_each(__sk, node, list) \ |
675 | hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node) | |
676 | #define sk_nulls_for_each_rcu(__sk, node, list) \ | |
677 | hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node) | |
b67bfe0d SL |
678 | #define sk_for_each_from(__sk) \ |
679 | hlist_for_each_entry_from(__sk, sk_node) | |
88ab1932 ED |
680 | #define sk_nulls_for_each_from(__sk, node) \ |
681 | if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \ | |
682 | hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node) | |
b67bfe0d SL |
683 | #define sk_for_each_safe(__sk, tmp, list) \ |
684 | hlist_for_each_entry_safe(__sk, tmp, list, sk_node) | |
685 | #define sk_for_each_bound(__sk, list) \ | |
686 | hlist_for_each_entry(__sk, list, sk_bind_node) | |
1da177e4 | 687 | |
2dc41cff DH |
688 | /** |
689 | * sk_nulls_for_each_entry_offset - iterate over a list at a given struct offset | |
690 | * @tpos: the type * to use as a loop cursor. | |
691 | * @pos: the &struct hlist_node to use as a loop cursor. | |
692 | * @head: the head for your list. | |
693 | * @offset: offset of hlist_node within the struct. | |
694 | * | |
695 | */ | |
696 | #define sk_nulls_for_each_entry_offset(tpos, pos, head, offset) \ | |
697 | for (pos = (head)->first; \ | |
698 | (!is_a_nulls(pos)) && \ | |
699 | ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;}); \ | |
700 | pos = pos->next) | |
701 | ||
c336d148 EB |
702 | static inline struct user_namespace *sk_user_ns(struct sock *sk) |
703 | { | |
704 | /* Careful only use this in a context where these parameters | |
705 | * can not change and must all be valid, such as recvmsg from | |
706 | * userspace. | |
707 | */ | |
708 | return sk->sk_socket->file->f_cred->user_ns; | |
709 | } | |
710 | ||
1da177e4 LT |
711 | /* Sock flags */ |
712 | enum sock_flags { | |
713 | SOCK_DEAD, | |
714 | SOCK_DONE, | |
715 | SOCK_URGINLINE, | |
716 | SOCK_KEEPOPEN, | |
717 | SOCK_LINGER, | |
718 | SOCK_DESTROY, | |
719 | SOCK_BROADCAST, | |
720 | SOCK_TIMESTAMP, | |
721 | SOCK_ZAPPED, | |
722 | SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */ | |
723 | SOCK_DBG, /* %SO_DEBUG setting */ | |
724 | SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */ | |
92f37fd2 | 725 | SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */ |
1da177e4 LT |
726 | SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */ |
727 | SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */ | |
7cb02404 | 728 | SOCK_MEMALLOC, /* VM depends on this socket for swapping */ |
20d49473 | 729 | SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */ |
bcdce719 | 730 | SOCK_FASYNC, /* fasync() active */ |
3b885787 | 731 | SOCK_RXQ_OVFL, |
1cdebb42 | 732 | SOCK_ZEROCOPY, /* buffers from userspace */ |
6e3e939f | 733 | SOCK_WIFI_STATUS, /* push wifi status to userspace */ |
3bdc0eba BG |
734 | SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS. |
735 | * Will use last 4 bytes of packet sent from | |
736 | * user-space instead. | |
737 | */ | |
d59577b6 | 738 | SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */ |
7d4c04fc | 739 | SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */ |
1da177e4 LT |
740 | }; |
741 | ||
01ce63c9 MRL |
742 | #define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE)) |
743 | ||
53b924b3 RB |
744 | static inline void sock_copy_flags(struct sock *nsk, struct sock *osk) |
745 | { | |
746 | nsk->sk_flags = osk->sk_flags; | |
747 | } | |
748 | ||
1da177e4 LT |
749 | static inline void sock_set_flag(struct sock *sk, enum sock_flags flag) |
750 | { | |
751 | __set_bit(flag, &sk->sk_flags); | |
752 | } | |
753 | ||
754 | static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag) | |
755 | { | |
756 | __clear_bit(flag, &sk->sk_flags); | |
757 | } | |
758 | ||
1b23a5df | 759 | static inline bool sock_flag(const struct sock *sk, enum sock_flags flag) |
1da177e4 LT |
760 | { |
761 | return test_bit(flag, &sk->sk_flags); | |
762 | } | |
763 | ||
c93bdd0e MG |
764 | #ifdef CONFIG_NET |
765 | extern struct static_key memalloc_socks; | |
766 | static inline int sk_memalloc_socks(void) | |
767 | { | |
768 | return static_key_false(&memalloc_socks); | |
769 | } | |
770 | #else | |
771 | ||
772 | static inline int sk_memalloc_socks(void) | |
773 | { | |
774 | return 0; | |
775 | } | |
776 | ||
777 | #endif | |
778 | ||
7450aaf6 | 779 | static inline gfp_t sk_gfp_mask(const struct sock *sk, gfp_t gfp_mask) |
99a1dec7 | 780 | { |
7450aaf6 | 781 | return gfp_mask | (sk->sk_allocation & __GFP_MEMALLOC); |
99a1dec7 MG |
782 | } |
783 | ||
1da177e4 LT |
784 | static inline void sk_acceptq_removed(struct sock *sk) |
785 | { | |
786 | sk->sk_ack_backlog--; | |
787 | } | |
788 | ||
789 | static inline void sk_acceptq_added(struct sock *sk) | |
790 | { | |
791 | sk->sk_ack_backlog++; | |
792 | } | |
793 | ||
dc6b9b78 | 794 | static inline bool sk_acceptq_is_full(const struct sock *sk) |
1da177e4 | 795 | { |
64a14651 | 796 | return sk->sk_ack_backlog > sk->sk_max_ack_backlog; |
1da177e4 LT |
797 | } |
798 | ||
799 | /* | |
800 | * Compute minimal free write space needed to queue new packets. | |
801 | */ | |
dc6b9b78 | 802 | static inline int sk_stream_min_wspace(const struct sock *sk) |
1da177e4 | 803 | { |
8df09ea3 | 804 | return sk->sk_wmem_queued >> 1; |
1da177e4 LT |
805 | } |
806 | ||
dc6b9b78 | 807 | static inline int sk_stream_wspace(const struct sock *sk) |
1da177e4 LT |
808 | { |
809 | return sk->sk_sndbuf - sk->sk_wmem_queued; | |
810 | } | |
811 | ||
69336bd2 | 812 | void sk_stream_write_space(struct sock *sk); |
1da177e4 | 813 | |
8eae939f | 814 | /* OOB backlog add */ |
a3a858ff | 815 | static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb) |
9ee6b535 | 816 | { |
7fee226a | 817 | /* dont let skb dst not refcounted, we are going to leave rcu lock */ |
5037e9ef | 818 | skb_dst_force_safe(skb); |
7fee226a ED |
819 | |
820 | if (!sk->sk_backlog.tail) | |
821 | sk->sk_backlog.head = skb; | |
822 | else | |
9ee6b535 | 823 | sk->sk_backlog.tail->next = skb; |
7fee226a ED |
824 | |
825 | sk->sk_backlog.tail = skb; | |
9ee6b535 SH |
826 | skb->next = NULL; |
827 | } | |
1da177e4 | 828 | |
c377411f ED |
829 | /* |
830 | * Take into account size of receive queue and backlog queue | |
0fd7bac6 ED |
831 | * Do not take into account this skb truesize, |
832 | * to allow even a single big packet to come. | |
c377411f | 833 | */ |
274f482d | 834 | static inline bool sk_rcvqueues_full(const struct sock *sk, unsigned int limit) |
c377411f ED |
835 | { |
836 | unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc); | |
837 | ||
f545a38f | 838 | return qsize > limit; |
c377411f ED |
839 | } |
840 | ||
8eae939f | 841 | /* The per-socket spinlock must be held here. */ |
f545a38f ED |
842 | static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb, |
843 | unsigned int limit) | |
8eae939f | 844 | { |
274f482d | 845 | if (sk_rcvqueues_full(sk, limit)) |
8eae939f ZY |
846 | return -ENOBUFS; |
847 | ||
c7c49b8f ED |
848 | /* |
849 | * If the skb was allocated from pfmemalloc reserves, only | |
850 | * allow SOCK_MEMALLOC sockets to use it as this socket is | |
851 | * helping free memory | |
852 | */ | |
853 | if (skb_pfmemalloc(skb) && !sock_flag(sk, SOCK_MEMALLOC)) | |
854 | return -ENOMEM; | |
855 | ||
a3a858ff | 856 | __sk_add_backlog(sk, skb); |
8eae939f ZY |
857 | sk->sk_backlog.len += skb->truesize; |
858 | return 0; | |
859 | } | |
860 | ||
69336bd2 | 861 | int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb); |
b4b9e355 | 862 | |
c57943a1 PZ |
863 | static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb) |
864 | { | |
b4b9e355 MG |
865 | if (sk_memalloc_socks() && skb_pfmemalloc(skb)) |
866 | return __sk_backlog_rcv(sk, skb); | |
867 | ||
c57943a1 PZ |
868 | return sk->sk_backlog_rcv(sk, skb); |
869 | } | |
870 | ||
2c8c56e1 ED |
871 | static inline void sk_incoming_cpu_update(struct sock *sk) |
872 | { | |
873 | sk->sk_incoming_cpu = raw_smp_processor_id(); | |
874 | } | |
875 | ||
fe477558 | 876 | static inline void sock_rps_record_flow_hash(__u32 hash) |
c58dc01b DM |
877 | { |
878 | #ifdef CONFIG_RPS | |
879 | struct rps_sock_flow_table *sock_flow_table; | |
880 | ||
881 | rcu_read_lock(); | |
882 | sock_flow_table = rcu_dereference(rps_sock_flow_table); | |
fe477558 | 883 | rps_record_sock_flow(sock_flow_table, hash); |
c58dc01b DM |
884 | rcu_read_unlock(); |
885 | #endif | |
886 | } | |
887 | ||
fe477558 TH |
888 | static inline void sock_rps_record_flow(const struct sock *sk) |
889 | { | |
c9d8ca04 | 890 | #ifdef CONFIG_RPS |
fe477558 | 891 | sock_rps_record_flow_hash(sk->sk_rxhash); |
c9d8ca04 | 892 | #endif |
fe477558 TH |
893 | } |
894 | ||
bdeab991 TH |
895 | static inline void sock_rps_save_rxhash(struct sock *sk, |
896 | const struct sk_buff *skb) | |
c58dc01b DM |
897 | { |
898 | #ifdef CONFIG_RPS | |
567e4b79 | 899 | if (unlikely(sk->sk_rxhash != skb->hash)) |
61b905da | 900 | sk->sk_rxhash = skb->hash; |
c58dc01b DM |
901 | #endif |
902 | } | |
903 | ||
bdeab991 TH |
904 | static inline void sock_rps_reset_rxhash(struct sock *sk) |
905 | { | |
906 | #ifdef CONFIG_RPS | |
bdeab991 TH |
907 | sk->sk_rxhash = 0; |
908 | #endif | |
909 | } | |
910 | ||
cfcabdcc SH |
911 | #define sk_wait_event(__sk, __timeo, __condition) \ |
912 | ({ int __rc; \ | |
913 | release_sock(__sk); \ | |
914 | __rc = __condition; \ | |
915 | if (!__rc) { \ | |
916 | *(__timeo) = schedule_timeout(*(__timeo)); \ | |
917 | } \ | |
26cabd31 | 918 | sched_annotate_sleep(); \ |
cfcabdcc SH |
919 | lock_sock(__sk); \ |
920 | __rc = __condition; \ | |
921 | __rc; \ | |
922 | }) | |
1da177e4 | 923 | |
69336bd2 JP |
924 | int sk_stream_wait_connect(struct sock *sk, long *timeo_p); |
925 | int sk_stream_wait_memory(struct sock *sk, long *timeo_p); | |
926 | void sk_stream_wait_close(struct sock *sk, long timeo_p); | |
927 | int sk_stream_error(struct sock *sk, int flags, int err); | |
928 | void sk_stream_kill_queues(struct sock *sk); | |
929 | void sk_set_memalloc(struct sock *sk); | |
930 | void sk_clear_memalloc(struct sock *sk); | |
1da177e4 | 931 | |
dfbafc99 | 932 | int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb); |
1da177e4 | 933 | |
60236fdd | 934 | struct request_sock_ops; |
6d6ee43e | 935 | struct timewait_sock_ops; |
ab1e0a13 | 936 | struct inet_hashinfo; |
fc8717ba | 937 | struct raw_hashinfo; |
de477254 | 938 | struct module; |
2e6599cb | 939 | |
f77d6021 ED |
940 | /* |
941 | * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes | |
942 | * un-modified. Special care is taken when initializing object to zero. | |
943 | */ | |
944 | static inline void sk_prot_clear_nulls(struct sock *sk, int size) | |
945 | { | |
946 | if (offsetof(struct sock, sk_node.next) != 0) | |
947 | memset(sk, 0, offsetof(struct sock, sk_node.next)); | |
948 | memset(&sk->sk_node.pprev, 0, | |
949 | size - offsetof(struct sock, sk_node.pprev)); | |
950 | } | |
951 | ||
1da177e4 LT |
952 | /* Networking protocol blocks we attach to sockets. |
953 | * socket layer -> transport layer interface | |
1da177e4 LT |
954 | */ |
955 | struct proto { | |
dc6b9b78 | 956 | void (*close)(struct sock *sk, |
1da177e4 LT |
957 | long timeout); |
958 | int (*connect)(struct sock *sk, | |
dc6b9b78 | 959 | struct sockaddr *uaddr, |
1da177e4 LT |
960 | int addr_len); |
961 | int (*disconnect)(struct sock *sk, int flags); | |
962 | ||
dc6b9b78 | 963 | struct sock * (*accept)(struct sock *sk, int flags, int *err); |
1da177e4 LT |
964 | |
965 | int (*ioctl)(struct sock *sk, int cmd, | |
966 | unsigned long arg); | |
967 | int (*init)(struct sock *sk); | |
7d06b2e0 | 968 | void (*destroy)(struct sock *sk); |
1da177e4 | 969 | void (*shutdown)(struct sock *sk, int how); |
dc6b9b78 | 970 | int (*setsockopt)(struct sock *sk, int level, |
1da177e4 | 971 | int optname, char __user *optval, |
b7058842 | 972 | unsigned int optlen); |
dc6b9b78 ED |
973 | int (*getsockopt)(struct sock *sk, int level, |
974 | int optname, char __user *optval, | |
975 | int __user *option); | |
af01d537 | 976 | #ifdef CONFIG_COMPAT |
3fdadf7d DM |
977 | int (*compat_setsockopt)(struct sock *sk, |
978 | int level, | |
979 | int optname, char __user *optval, | |
b7058842 | 980 | unsigned int optlen); |
3fdadf7d DM |
981 | int (*compat_getsockopt)(struct sock *sk, |
982 | int level, | |
983 | int optname, char __user *optval, | |
984 | int __user *option); | |
709b46e8 EB |
985 | int (*compat_ioctl)(struct sock *sk, |
986 | unsigned int cmd, unsigned long arg); | |
af01d537 | 987 | #endif |
1b784140 YX |
988 | int (*sendmsg)(struct sock *sk, struct msghdr *msg, |
989 | size_t len); | |
990 | int (*recvmsg)(struct sock *sk, struct msghdr *msg, | |
dc6b9b78 ED |
991 | size_t len, int noblock, int flags, |
992 | int *addr_len); | |
1da177e4 LT |
993 | int (*sendpage)(struct sock *sk, struct page *page, |
994 | int offset, size_t size, int flags); | |
dc6b9b78 | 995 | int (*bind)(struct sock *sk, |
1da177e4 LT |
996 | struct sockaddr *uaddr, int addr_len); |
997 | ||
dc6b9b78 | 998 | int (*backlog_rcv) (struct sock *sk, |
1da177e4 LT |
999 | struct sk_buff *skb); |
1000 | ||
46d3ceab ED |
1001 | void (*release_cb)(struct sock *sk); |
1002 | ||
1da177e4 LT |
1003 | /* Keeping track of sk's, looking them up, and port selection methods. */ |
1004 | void (*hash)(struct sock *sk); | |
1005 | void (*unhash)(struct sock *sk); | |
719f8358 | 1006 | void (*rehash)(struct sock *sk); |
1da177e4 | 1007 | int (*get_port)(struct sock *sk, unsigned short snum); |
fcbdf09d | 1008 | void (*clear_sk)(struct sock *sk, int size); |
1da177e4 | 1009 | |
286ab3d4 | 1010 | /* Keeping track of sockets in use */ |
65f76517 | 1011 | #ifdef CONFIG_PROC_FS |
13ff3d6f | 1012 | unsigned int inuse_idx; |
65f76517 | 1013 | #endif |
ebb53d75 | 1014 | |
c9bee3b7 | 1015 | bool (*stream_memory_free)(const struct sock *sk); |
1da177e4 | 1016 | /* Memory pressure */ |
5c52ba17 | 1017 | void (*enter_memory_pressure)(struct sock *sk); |
8d987e5c | 1018 | atomic_long_t *memory_allocated; /* Current allocated memory. */ |
1748376b | 1019 | struct percpu_counter *sockets_allocated; /* Current number of sockets. */ |
1da177e4 LT |
1020 | /* |
1021 | * Pressure flag: try to collapse. | |
1022 | * Technical note: it is used by multiple contexts non atomically. | |
3ab224be | 1023 | * All the __sk_mem_schedule() is of this nature: accounting |
1da177e4 LT |
1024 | * is strict, actions are advisory and have some latency. |
1025 | */ | |
1026 | int *memory_pressure; | |
8d987e5c | 1027 | long *sysctl_mem; |
1da177e4 LT |
1028 | int *sysctl_wmem; |
1029 | int *sysctl_rmem; | |
1030 | int max_header; | |
7ba42910 | 1031 | bool no_autobind; |
1da177e4 | 1032 | |
271b72c7 | 1033 | struct kmem_cache *slab; |
1da177e4 | 1034 | unsigned int obj_size; |
271b72c7 | 1035 | int slab_flags; |
1da177e4 | 1036 | |
dd24c001 | 1037 | struct percpu_counter *orphan_count; |
8feaf0c0 | 1038 | |
60236fdd | 1039 | struct request_sock_ops *rsk_prot; |
6d6ee43e | 1040 | struct timewait_sock_ops *twsk_prot; |
2e6599cb | 1041 | |
39d8cda7 PE |
1042 | union { |
1043 | struct inet_hashinfo *hashinfo; | |
645ca708 | 1044 | struct udp_table *udp_table; |
fc8717ba | 1045 | struct raw_hashinfo *raw_hash; |
39d8cda7 | 1046 | } h; |
ab1e0a13 | 1047 | |
1da177e4 LT |
1048 | struct module *owner; |
1049 | ||
1050 | char name[32]; | |
1051 | ||
1052 | struct list_head node; | |
e6848976 ACM |
1053 | #ifdef SOCK_REFCNT_DEBUG |
1054 | atomic_t socks; | |
1055 | #endif | |
c255a458 | 1056 | #ifdef CONFIG_MEMCG_KMEM |
e1aab161 GC |
1057 | /* |
1058 | * cgroup specific init/deinit functions. Called once for all | |
1059 | * protocols that implement it, from cgroups populate function. | |
1060 | * This function has to setup any files the protocol want to | |
1061 | * appear in the kmem cgroup filesystem. | |
1062 | */ | |
1d62e436 | 1063 | int (*init_cgroup)(struct mem_cgroup *memcg, |
e1aab161 | 1064 | struct cgroup_subsys *ss); |
1d62e436 | 1065 | void (*destroy_cgroup)(struct mem_cgroup *memcg); |
e1aab161 GC |
1066 | struct cg_proto *(*proto_cgroup)(struct mem_cgroup *memcg); |
1067 | #endif | |
64be0aed | 1068 | int (*diag_destroy)(struct sock *sk, int err); |
e1aab161 GC |
1069 | }; |
1070 | ||
69336bd2 JP |
1071 | int proto_register(struct proto *prot, int alloc_slab); |
1072 | void proto_unregister(struct proto *prot); | |
1da177e4 | 1073 | |
e6848976 ACM |
1074 | #ifdef SOCK_REFCNT_DEBUG |
1075 | static inline void sk_refcnt_debug_inc(struct sock *sk) | |
1076 | { | |
1077 | atomic_inc(&sk->sk_prot->socks); | |
1078 | } | |
1079 | ||
1080 | static inline void sk_refcnt_debug_dec(struct sock *sk) | |
1081 | { | |
1082 | atomic_dec(&sk->sk_prot->socks); | |
1083 | printk(KERN_DEBUG "%s socket %p released, %d are still alive\n", | |
1084 | sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks)); | |
1085 | } | |
1086 | ||
dec34fb0 | 1087 | static inline void sk_refcnt_debug_release(const struct sock *sk) |
e6848976 ACM |
1088 | { |
1089 | if (atomic_read(&sk->sk_refcnt) != 1) | |
1090 | printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n", | |
1091 | sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt)); | |
1092 | } | |
1093 | #else /* SOCK_REFCNT_DEBUG */ | |
1094 | #define sk_refcnt_debug_inc(sk) do { } while (0) | |
1095 | #define sk_refcnt_debug_dec(sk) do { } while (0) | |
1096 | #define sk_refcnt_debug_release(sk) do { } while (0) | |
1097 | #endif /* SOCK_REFCNT_DEBUG */ | |
1098 | ||
c255a458 | 1099 | #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_NET) |
c5905afb | 1100 | extern struct static_key memcg_socket_limit_enabled; |
c5905afb | 1101 | #define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled) |
e1aab161 GC |
1102 | #else |
1103 | #define mem_cgroup_sockets_enabled 0 | |
e1aab161 GC |
1104 | #endif |
1105 | ||
c9bee3b7 ED |
1106 | static inline bool sk_stream_memory_free(const struct sock *sk) |
1107 | { | |
1108 | if (sk->sk_wmem_queued >= sk->sk_sndbuf) | |
1109 | return false; | |
1110 | ||
1111 | return sk->sk_prot->stream_memory_free ? | |
1112 | sk->sk_prot->stream_memory_free(sk) : true; | |
1113 | } | |
1114 | ||
64dc6130 ED |
1115 | static inline bool sk_stream_is_writeable(const struct sock *sk) |
1116 | { | |
c9bee3b7 ED |
1117 | return sk_stream_wspace(sk) >= sk_stream_min_wspace(sk) && |
1118 | sk_stream_memory_free(sk); | |
64dc6130 | 1119 | } |
e1aab161 | 1120 | |
c9bee3b7 | 1121 | |
180d8cd9 GC |
1122 | static inline bool sk_has_memory_pressure(const struct sock *sk) |
1123 | { | |
1124 | return sk->sk_prot->memory_pressure != NULL; | |
1125 | } | |
1126 | ||
1127 | static inline bool sk_under_memory_pressure(const struct sock *sk) | |
1128 | { | |
1129 | if (!sk->sk_prot->memory_pressure) | |
1130 | return false; | |
e1aab161 | 1131 | |
e805605c JW |
1132 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp && |
1133 | mem_cgroup_under_socket_pressure(sk->sk_cgrp)) | |
1134 | return true; | |
e1aab161 | 1135 | |
35b87f6c | 1136 | return !!*sk->sk_prot->memory_pressure; |
180d8cd9 GC |
1137 | } |
1138 | ||
1139 | static inline void sk_leave_memory_pressure(struct sock *sk) | |
1140 | { | |
1141 | int *memory_pressure = sk->sk_prot->memory_pressure; | |
1142 | ||
e1aab161 GC |
1143 | if (!memory_pressure) |
1144 | return; | |
1145 | ||
1146 | if (*memory_pressure) | |
180d8cd9 GC |
1147 | *memory_pressure = 0; |
1148 | } | |
1149 | ||
1150 | static inline void sk_enter_memory_pressure(struct sock *sk) | |
1151 | { | |
e1aab161 GC |
1152 | if (!sk->sk_prot->enter_memory_pressure) |
1153 | return; | |
1154 | ||
e1aab161 | 1155 | sk->sk_prot->enter_memory_pressure(sk); |
180d8cd9 GC |
1156 | } |
1157 | ||
1158 | static inline long sk_prot_mem_limits(const struct sock *sk, int index) | |
1159 | { | |
e805605c | 1160 | return sk->sk_prot->sysctl_mem[index]; |
e1aab161 GC |
1161 | } |
1162 | ||
180d8cd9 GC |
1163 | static inline long |
1164 | sk_memory_allocated(const struct sock *sk) | |
1165 | { | |
e805605c | 1166 | return atomic_long_read(sk->sk_prot->memory_allocated); |
180d8cd9 GC |
1167 | } |
1168 | ||
1169 | static inline long | |
e805605c | 1170 | sk_memory_allocated_add(struct sock *sk, int amt) |
180d8cd9 | 1171 | { |
e805605c | 1172 | return atomic_long_add_return(amt, sk->sk_prot->memory_allocated); |
180d8cd9 GC |
1173 | } |
1174 | ||
1175 | static inline void | |
0e90b31f | 1176 | sk_memory_allocated_sub(struct sock *sk, int amt) |
180d8cd9 | 1177 | { |
e805605c | 1178 | atomic_long_sub(amt, sk->sk_prot->memory_allocated); |
180d8cd9 GC |
1179 | } |
1180 | ||
1181 | static inline void sk_sockets_allocated_dec(struct sock *sk) | |
1182 | { | |
af95d7df | 1183 | percpu_counter_dec(sk->sk_prot->sockets_allocated); |
180d8cd9 GC |
1184 | } |
1185 | ||
1186 | static inline void sk_sockets_allocated_inc(struct sock *sk) | |
1187 | { | |
af95d7df | 1188 | percpu_counter_inc(sk->sk_prot->sockets_allocated); |
180d8cd9 GC |
1189 | } |
1190 | ||
1191 | static inline int | |
1192 | sk_sockets_allocated_read_positive(struct sock *sk) | |
1193 | { | |
af95d7df | 1194 | return percpu_counter_read_positive(sk->sk_prot->sockets_allocated); |
180d8cd9 GC |
1195 | } |
1196 | ||
1197 | static inline int | |
1198 | proto_sockets_allocated_sum_positive(struct proto *prot) | |
1199 | { | |
1200 | return percpu_counter_sum_positive(prot->sockets_allocated); | |
1201 | } | |
1202 | ||
1203 | static inline long | |
1204 | proto_memory_allocated(struct proto *prot) | |
1205 | { | |
1206 | return atomic_long_read(prot->memory_allocated); | |
1207 | } | |
1208 | ||
1209 | static inline bool | |
1210 | proto_memory_pressure(struct proto *prot) | |
1211 | { | |
1212 | if (!prot->memory_pressure) | |
1213 | return false; | |
1214 | return !!*prot->memory_pressure; | |
1215 | } | |
1216 | ||
65f76517 ED |
1217 | |
1218 | #ifdef CONFIG_PROC_FS | |
1da177e4 | 1219 | /* Called with local bh disabled */ |
69336bd2 JP |
1220 | void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc); |
1221 | int sock_prot_inuse_get(struct net *net, struct proto *proto); | |
65f76517 | 1222 | #else |
dc6b9b78 | 1223 | static inline void sock_prot_inuse_add(struct net *net, struct proto *prot, |
c29a0bc4 | 1224 | int inc) |
65f76517 ED |
1225 | { |
1226 | } | |
65f76517 ED |
1227 | #endif |
1228 | ||
1da177e4 | 1229 | |
614c6cb4 ACM |
1230 | /* With per-bucket locks this operation is not-atomic, so that |
1231 | * this version is not worse. | |
1232 | */ | |
1233 | static inline void __sk_prot_rehash(struct sock *sk) | |
1234 | { | |
1235 | sk->sk_prot->unhash(sk); | |
1236 | sk->sk_prot->hash(sk); | |
1237 | } | |
1238 | ||
fcbdf09d OP |
1239 | void sk_prot_clear_portaddr_nulls(struct sock *sk, int size); |
1240 | ||
1da177e4 LT |
1241 | /* About 10 seconds */ |
1242 | #define SOCK_DESTROY_TIME (10*HZ) | |
1243 | ||
1244 | /* Sockets 0-1023 can't be bound to unless you are superuser */ | |
1245 | #define PROT_SOCK 1024 | |
1246 | ||
1247 | #define SHUTDOWN_MASK 3 | |
1248 | #define RCV_SHUTDOWN 1 | |
1249 | #define SEND_SHUTDOWN 2 | |
1250 | ||
1251 | #define SOCK_SNDBUF_LOCK 1 | |
1252 | #define SOCK_RCVBUF_LOCK 2 | |
1253 | #define SOCK_BINDADDR_LOCK 4 | |
1254 | #define SOCK_BINDPORT_LOCK 8 | |
1255 | ||
1da177e4 LT |
1256 | struct socket_alloc { |
1257 | struct socket socket; | |
1258 | struct inode vfs_inode; | |
1259 | }; | |
1260 | ||
1261 | static inline struct socket *SOCKET_I(struct inode *inode) | |
1262 | { | |
1263 | return &container_of(inode, struct socket_alloc, vfs_inode)->socket; | |
1264 | } | |
1265 | ||
1266 | static inline struct inode *SOCK_INODE(struct socket *socket) | |
1267 | { | |
1268 | return &container_of(socket, struct socket_alloc, socket)->vfs_inode; | |
1269 | } | |
1270 | ||
3ab224be HA |
1271 | /* |
1272 | * Functions for memory accounting | |
1273 | */ | |
69336bd2 | 1274 | int __sk_mem_schedule(struct sock *sk, int size, int kind); |
1a24e04e | 1275 | void __sk_mem_reclaim(struct sock *sk, int amount); |
1da177e4 | 1276 | |
3ab224be HA |
1277 | #define SK_MEM_QUANTUM ((int)PAGE_SIZE) |
1278 | #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM) | |
1279 | #define SK_MEM_SEND 0 | |
1280 | #define SK_MEM_RECV 1 | |
1da177e4 | 1281 | |
3ab224be | 1282 | static inline int sk_mem_pages(int amt) |
1da177e4 | 1283 | { |
3ab224be | 1284 | return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT; |
1da177e4 LT |
1285 | } |
1286 | ||
dc6b9b78 | 1287 | static inline bool sk_has_account(struct sock *sk) |
1da177e4 | 1288 | { |
3ab224be HA |
1289 | /* return true if protocol supports memory accounting */ |
1290 | return !!sk->sk_prot->memory_allocated; | |
1da177e4 LT |
1291 | } |
1292 | ||
dc6b9b78 | 1293 | static inline bool sk_wmem_schedule(struct sock *sk, int size) |
1da177e4 | 1294 | { |
3ab224be | 1295 | if (!sk_has_account(sk)) |
dc6b9b78 | 1296 | return true; |
3ab224be HA |
1297 | return size <= sk->sk_forward_alloc || |
1298 | __sk_mem_schedule(sk, size, SK_MEM_SEND); | |
1da177e4 LT |
1299 | } |
1300 | ||
c76562b6 | 1301 | static inline bool |
35c448a8 | 1302 | sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size) |
d80d99d6 | 1303 | { |
3ab224be | 1304 | if (!sk_has_account(sk)) |
dc6b9b78 | 1305 | return true; |
c76562b6 MG |
1306 | return size<= sk->sk_forward_alloc || |
1307 | __sk_mem_schedule(sk, size, SK_MEM_RECV) || | |
1308 | skb_pfmemalloc(skb); | |
3ab224be HA |
1309 | } |
1310 | ||
1311 | static inline void sk_mem_reclaim(struct sock *sk) | |
1312 | { | |
1313 | if (!sk_has_account(sk)) | |
1314 | return; | |
1315 | if (sk->sk_forward_alloc >= SK_MEM_QUANTUM) | |
1a24e04e | 1316 | __sk_mem_reclaim(sk, sk->sk_forward_alloc); |
3ab224be HA |
1317 | } |
1318 | ||
9993e7d3 DM |
1319 | static inline void sk_mem_reclaim_partial(struct sock *sk) |
1320 | { | |
1321 | if (!sk_has_account(sk)) | |
1322 | return; | |
1323 | if (sk->sk_forward_alloc > SK_MEM_QUANTUM) | |
1a24e04e | 1324 | __sk_mem_reclaim(sk, sk->sk_forward_alloc - 1); |
9993e7d3 DM |
1325 | } |
1326 | ||
3ab224be HA |
1327 | static inline void sk_mem_charge(struct sock *sk, int size) |
1328 | { | |
1329 | if (!sk_has_account(sk)) | |
1330 | return; | |
1331 | sk->sk_forward_alloc -= size; | |
1332 | } | |
1333 | ||
1334 | static inline void sk_mem_uncharge(struct sock *sk, int size) | |
1335 | { | |
1336 | if (!sk_has_account(sk)) | |
1337 | return; | |
1338 | sk->sk_forward_alloc += size; | |
1339 | } | |
1340 | ||
1341 | static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb) | |
1342 | { | |
3ab224be HA |
1343 | sock_set_flag(sk, SOCK_QUEUE_SHRUNK); |
1344 | sk->sk_wmem_queued -= skb->truesize; | |
1345 | sk_mem_uncharge(sk, skb->truesize); | |
1346 | __kfree_skb(skb); | |
d80d99d6 HX |
1347 | } |
1348 | ||
1da177e4 LT |
1349 | /* Used by processes to "lock" a socket state, so that |
1350 | * interrupts and bottom half handlers won't change it | |
1351 | * from under us. It essentially blocks any incoming | |
1352 | * packets, so that we won't get any new data or any | |
1353 | * packets that change the state of the socket. | |
1354 | * | |
1355 | * While locked, BH processing will add new packets to | |
1356 | * the backlog queue. This queue is processed by the | |
1357 | * owner of the socket lock right before it is released. | |
1358 | * | |
1359 | * Since ~2.3.5 it is also exclusive sleep lock serializing | |
1360 | * accesses from user process context. | |
1361 | */ | |
d2e9117c | 1362 | #define sock_owned_by_user(sk) ((sk)->sk_lock.owned) |
1da177e4 | 1363 | |
c3f9b018 ED |
1364 | static inline void sock_release_ownership(struct sock *sk) |
1365 | { | |
1366 | sk->sk_lock.owned = 0; | |
1367 | } | |
1368 | ||
ed07536e PZ |
1369 | /* |
1370 | * Macro so as to not evaluate some arguments when | |
1371 | * lockdep is not enabled. | |
1372 | * | |
1373 | * Mark both the sk_lock and the sk_lock.slock as a | |
1374 | * per-address-family lock class. | |
1375 | */ | |
dc6b9b78 | 1376 | #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \ |
ed07536e | 1377 | do { \ |
e8f6fbf6 | 1378 | sk->sk_lock.owned = 0; \ |
ed07536e PZ |
1379 | init_waitqueue_head(&sk->sk_lock.wq); \ |
1380 | spin_lock_init(&(sk)->sk_lock.slock); \ | |
1381 | debug_check_no_locks_freed((void *)&(sk)->sk_lock, \ | |
1382 | sizeof((sk)->sk_lock)); \ | |
1383 | lockdep_set_class_and_name(&(sk)->sk_lock.slock, \ | |
dc6b9b78 | 1384 | (skey), (sname)); \ |
ed07536e PZ |
1385 | lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \ |
1386 | } while (0) | |
1387 | ||
69336bd2 | 1388 | void lock_sock_nested(struct sock *sk, int subclass); |
fcc70d5f PZ |
1389 | |
1390 | static inline void lock_sock(struct sock *sk) | |
1391 | { | |
1392 | lock_sock_nested(sk, 0); | |
1393 | } | |
1394 | ||
69336bd2 | 1395 | void release_sock(struct sock *sk); |
1da177e4 LT |
1396 | |
1397 | /* BH context may only use the following locking interface. */ | |
1398 | #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock)) | |
c6366184 IM |
1399 | #define bh_lock_sock_nested(__sk) \ |
1400 | spin_lock_nested(&((__sk)->sk_lock.slock), \ | |
1401 | SINGLE_DEPTH_NESTING) | |
1da177e4 LT |
1402 | #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock)) |
1403 | ||
69336bd2 | 1404 | bool lock_sock_fast(struct sock *sk); |
8a74ad60 ED |
1405 | /** |
1406 | * unlock_sock_fast - complement of lock_sock_fast | |
1407 | * @sk: socket | |
1408 | * @slow: slow mode | |
1409 | * | |
1410 | * fast unlock socket for user context. | |
1411 | * If slow mode is on, we call regular release_sock() | |
1412 | */ | |
1413 | static inline void unlock_sock_fast(struct sock *sk, bool slow) | |
4b0b72f7 | 1414 | { |
8a74ad60 ED |
1415 | if (slow) |
1416 | release_sock(sk); | |
1417 | else | |
1418 | spin_unlock_bh(&sk->sk_lock.slock); | |
4b0b72f7 ED |
1419 | } |
1420 | ||
4b0b72f7 | 1421 | |
69336bd2 | 1422 | struct sock *sk_alloc(struct net *net, int family, gfp_t priority, |
11aa9c28 | 1423 | struct proto *prot, int kern); |
69336bd2 | 1424 | void sk_free(struct sock *sk); |
eb4cb008 | 1425 | void sk_destruct(struct sock *sk); |
69336bd2 JP |
1426 | struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority); |
1427 | ||
1428 | struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force, | |
1429 | gfp_t priority); | |
69336bd2 JP |
1430 | void sock_wfree(struct sk_buff *skb); |
1431 | void skb_orphan_partial(struct sk_buff *skb); | |
1432 | void sock_rfree(struct sk_buff *skb); | |
62bccb8c | 1433 | void sock_efree(struct sk_buff *skb); |
82eabd9e | 1434 | #ifdef CONFIG_INET |
69336bd2 | 1435 | void sock_edemux(struct sk_buff *skb); |
82eabd9e AD |
1436 | #else |
1437 | #define sock_edemux(skb) sock_efree(skb) | |
1438 | #endif | |
69336bd2 JP |
1439 | |
1440 | int sock_setsockopt(struct socket *sock, int level, int op, | |
1441 | char __user *optval, unsigned int optlen); | |
1442 | ||
1443 | int sock_getsockopt(struct socket *sock, int level, int op, | |
1444 | char __user *optval, int __user *optlen); | |
1445 | struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size, | |
1446 | int noblock, int *errcode); | |
1447 | struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len, | |
1448 | unsigned long data_len, int noblock, | |
1449 | int *errcode, int max_page_order); | |
1450 | void *sock_kmalloc(struct sock *sk, int size, gfp_t priority); | |
1451 | void sock_kfree_s(struct sock *sk, void *mem, int size); | |
79e88659 | 1452 | void sock_kzfree_s(struct sock *sk, void *mem, int size); |
69336bd2 | 1453 | void sk_send_sigurg(struct sock *sk); |
1da177e4 | 1454 | |
f28ea365 EJ |
1455 | struct sockcm_cookie { |
1456 | u32 mark; | |
1457 | }; | |
1458 | ||
1459 | int sock_cmsg_send(struct sock *sk, struct msghdr *msg, | |
1460 | struct sockcm_cookie *sockc); | |
1461 | ||
1da177e4 LT |
1462 | /* |
1463 | * Functions to fill in entries in struct proto_ops when a protocol | |
1464 | * does not implement a particular function. | |
1465 | */ | |
69336bd2 JP |
1466 | int sock_no_bind(struct socket *, struct sockaddr *, int); |
1467 | int sock_no_connect(struct socket *, struct sockaddr *, int, int); | |
1468 | int sock_no_socketpair(struct socket *, struct socket *); | |
1469 | int sock_no_accept(struct socket *, struct socket *, int); | |
1470 | int sock_no_getname(struct socket *, struct sockaddr *, int *, int); | |
1471 | unsigned int sock_no_poll(struct file *, struct socket *, | |
1472 | struct poll_table_struct *); | |
1473 | int sock_no_ioctl(struct socket *, unsigned int, unsigned long); | |
1474 | int sock_no_listen(struct socket *, int); | |
1475 | int sock_no_shutdown(struct socket *, int); | |
1476 | int sock_no_getsockopt(struct socket *, int , int, char __user *, int __user *); | |
1477 | int sock_no_setsockopt(struct socket *, int, int, char __user *, unsigned int); | |
1b784140 YX |
1478 | int sock_no_sendmsg(struct socket *, struct msghdr *, size_t); |
1479 | int sock_no_recvmsg(struct socket *, struct msghdr *, size_t, int); | |
69336bd2 JP |
1480 | int sock_no_mmap(struct file *file, struct socket *sock, |
1481 | struct vm_area_struct *vma); | |
1482 | ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, | |
1483 | size_t size, int flags); | |
1da177e4 LT |
1484 | |
1485 | /* | |
1486 | * Functions to fill in entries in struct proto_ops when a protocol | |
1487 | * uses the inet style. | |
1488 | */ | |
69336bd2 | 1489 | int sock_common_getsockopt(struct socket *sock, int level, int optname, |
1da177e4 | 1490 | char __user *optval, int __user *optlen); |
1b784140 YX |
1491 | int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size, |
1492 | int flags); | |
69336bd2 | 1493 | int sock_common_setsockopt(struct socket *sock, int level, int optname, |
b7058842 | 1494 | char __user *optval, unsigned int optlen); |
69336bd2 | 1495 | int compat_sock_common_getsockopt(struct socket *sock, int level, |
3fdadf7d | 1496 | int optname, char __user *optval, int __user *optlen); |
69336bd2 | 1497 | int compat_sock_common_setsockopt(struct socket *sock, int level, |
b7058842 | 1498 | int optname, char __user *optval, unsigned int optlen); |
1da177e4 | 1499 | |
69336bd2 | 1500 | void sk_common_release(struct sock *sk); |
1da177e4 LT |
1501 | |
1502 | /* | |
1503 | * Default socket callbacks and setup code | |
1504 | */ | |
dc6b9b78 | 1505 | |
1da177e4 | 1506 | /* Initialise core socket variables */ |
69336bd2 | 1507 | void sock_init_data(struct socket *sock, struct sock *sk); |
1da177e4 | 1508 | |
1da177e4 LT |
1509 | /* |
1510 | * Socket reference counting postulates. | |
1511 | * | |
1512 | * * Each user of socket SHOULD hold a reference count. | |
1513 | * * Each access point to socket (an hash table bucket, reference from a list, | |
1514 | * running timer, skb in flight MUST hold a reference count. | |
1515 | * * When reference count hits 0, it means it will never increase back. | |
1516 | * * When reference count hits 0, it means that no references from | |
1517 | * outside exist to this socket and current process on current CPU | |
1518 | * is last user and may/should destroy this socket. | |
1519 | * * sk_free is called from any context: process, BH, IRQ. When | |
1520 | * it is called, socket has no references from outside -> sk_free | |
1521 | * may release descendant resources allocated by the socket, but | |
1522 | * to the time when it is called, socket is NOT referenced by any | |
1523 | * hash tables, lists etc. | |
1524 | * * Packets, delivered from outside (from network or from another process) | |
1525 | * and enqueued on receive/error queues SHOULD NOT grab reference count, | |
1526 | * when they sit in queue. Otherwise, packets will leak to hole, when | |
1527 | * socket is looked up by one cpu and unhasing is made by another CPU. | |
1528 | * It is true for udp/raw, netlink (leak to receive and error queues), tcp | |
1529 | * (leak to backlog). Packet socket does all the processing inside | |
1530 | * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets | |
1531 | * use separate SMP lock, so that they are prone too. | |
1532 | */ | |
1533 | ||
1534 | /* Ungrab socket and destroy it, if it was the last reference. */ | |
1535 | static inline void sock_put(struct sock *sk) | |
1536 | { | |
1537 | if (atomic_dec_and_test(&sk->sk_refcnt)) | |
1538 | sk_free(sk); | |
1539 | } | |
05dbc7b5 | 1540 | /* Generic version of sock_put(), dealing with all sockets |
41b822c5 | 1541 | * (TCP_TIMEWAIT, TCP_NEW_SYN_RECV, ESTABLISHED...) |
05dbc7b5 ED |
1542 | */ |
1543 | void sock_gen_put(struct sock *sk); | |
1da177e4 | 1544 | |
69336bd2 | 1545 | int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested); |
25995ff5 | 1546 | |
e022f0b4 KK |
1547 | static inline void sk_tx_queue_set(struct sock *sk, int tx_queue) |
1548 | { | |
1549 | sk->sk_tx_queue_mapping = tx_queue; | |
1550 | } | |
1551 | ||
1552 | static inline void sk_tx_queue_clear(struct sock *sk) | |
1553 | { | |
1554 | sk->sk_tx_queue_mapping = -1; | |
1555 | } | |
1556 | ||
1557 | static inline int sk_tx_queue_get(const struct sock *sk) | |
1558 | { | |
b0f77d0e | 1559 | return sk ? sk->sk_tx_queue_mapping : -1; |
e022f0b4 KK |
1560 | } |
1561 | ||
972692e0 DM |
1562 | static inline void sk_set_socket(struct sock *sk, struct socket *sock) |
1563 | { | |
e022f0b4 | 1564 | sk_tx_queue_clear(sk); |
972692e0 DM |
1565 | sk->sk_socket = sock; |
1566 | } | |
1567 | ||
aa395145 ED |
1568 | static inline wait_queue_head_t *sk_sleep(struct sock *sk) |
1569 | { | |
eaefd110 ED |
1570 | BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0); |
1571 | return &rcu_dereference_raw(sk->sk_wq)->wait; | |
aa395145 | 1572 | } |
1da177e4 LT |
1573 | /* Detach socket from process context. |
1574 | * Announce socket dead, detach it from wait queue and inode. | |
1575 | * Note that parent inode held reference count on this struct sock, | |
1576 | * we do not release it in this function, because protocol | |
1577 | * probably wants some additional cleanups or even continuing | |
1578 | * to work with this socket (TCP). | |
1579 | */ | |
1580 | static inline void sock_orphan(struct sock *sk) | |
1581 | { | |
1582 | write_lock_bh(&sk->sk_callback_lock); | |
1583 | sock_set_flag(sk, SOCK_DEAD); | |
972692e0 | 1584 | sk_set_socket(sk, NULL); |
43815482 | 1585 | sk->sk_wq = NULL; |
1da177e4 LT |
1586 | write_unlock_bh(&sk->sk_callback_lock); |
1587 | } | |
1588 | ||
1589 | static inline void sock_graft(struct sock *sk, struct socket *parent) | |
1590 | { | |
1591 | write_lock_bh(&sk->sk_callback_lock); | |
eaefd110 | 1592 | sk->sk_wq = parent->wq; |
1da177e4 | 1593 | parent->sk = sk; |
972692e0 | 1594 | sk_set_socket(sk, parent); |
4237c75c | 1595 | security_sock_graft(sk, parent); |
1da177e4 LT |
1596 | write_unlock_bh(&sk->sk_callback_lock); |
1597 | } | |
1598 | ||
69336bd2 JP |
1599 | kuid_t sock_i_uid(struct sock *sk); |
1600 | unsigned long sock_i_ino(struct sock *sk); | |
1da177e4 | 1601 | |
58d607d3 | 1602 | static inline u32 net_tx_rndhash(void) |
877d1f62 | 1603 | { |
58d607d3 ED |
1604 | u32 v = prandom_u32(); |
1605 | ||
1606 | return v ?: 1; | |
1607 | } | |
877d1f62 | 1608 | |
58d607d3 ED |
1609 | static inline void sk_set_txhash(struct sock *sk) |
1610 | { | |
1611 | sk->sk_txhash = net_tx_rndhash(); | |
877d1f62 TH |
1612 | } |
1613 | ||
265f94ff TH |
1614 | static inline void sk_rethink_txhash(struct sock *sk) |
1615 | { | |
1616 | if (sk->sk_txhash) | |
1617 | sk_set_txhash(sk); | |
1618 | } | |
1619 | ||
1da177e4 LT |
1620 | static inline struct dst_entry * |
1621 | __sk_dst_get(struct sock *sk) | |
1622 | { | |
d8bf4ca9 | 1623 | return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) || |
f68c224f | 1624 | lockdep_is_held(&sk->sk_lock.slock)); |
1da177e4 LT |
1625 | } |
1626 | ||
1627 | static inline struct dst_entry * | |
1628 | sk_dst_get(struct sock *sk) | |
1629 | { | |
1630 | struct dst_entry *dst; | |
1631 | ||
b6c6712a ED |
1632 | rcu_read_lock(); |
1633 | dst = rcu_dereference(sk->sk_dst_cache); | |
f8864972 ED |
1634 | if (dst && !atomic_inc_not_zero(&dst->__refcnt)) |
1635 | dst = NULL; | |
b6c6712a | 1636 | rcu_read_unlock(); |
1da177e4 LT |
1637 | return dst; |
1638 | } | |
1639 | ||
b6c6712a ED |
1640 | static inline void dst_negative_advice(struct sock *sk) |
1641 | { | |
1642 | struct dst_entry *ndst, *dst = __sk_dst_get(sk); | |
1643 | ||
265f94ff TH |
1644 | sk_rethink_txhash(sk); |
1645 | ||
b6c6712a ED |
1646 | if (dst && dst->ops->negative_advice) { |
1647 | ndst = dst->ops->negative_advice(dst); | |
1648 | ||
1649 | if (ndst != dst) { | |
1650 | rcu_assign_pointer(sk->sk_dst_cache, ndst); | |
0a6957e7 | 1651 | sk_tx_queue_clear(sk); |
b6c6712a ED |
1652 | } |
1653 | } | |
1654 | } | |
1655 | ||
1da177e4 LT |
1656 | static inline void |
1657 | __sk_dst_set(struct sock *sk, struct dst_entry *dst) | |
1658 | { | |
1659 | struct dst_entry *old_dst; | |
1660 | ||
e022f0b4 | 1661 | sk_tx_queue_clear(sk); |
0b53ff2e ED |
1662 | /* |
1663 | * This can be called while sk is owned by the caller only, | |
1664 | * with no state that can be checked in a rcu_dereference_check() cond | |
1665 | */ | |
1666 | old_dst = rcu_dereference_raw(sk->sk_dst_cache); | |
b6c6712a | 1667 | rcu_assign_pointer(sk->sk_dst_cache, dst); |
1da177e4 LT |
1668 | dst_release(old_dst); |
1669 | } | |
1670 | ||
1671 | static inline void | |
1672 | sk_dst_set(struct sock *sk, struct dst_entry *dst) | |
1673 | { | |
7f502361 ED |
1674 | struct dst_entry *old_dst; |
1675 | ||
1676 | sk_tx_queue_clear(sk); | |
5925a055 | 1677 | old_dst = xchg((__force struct dst_entry **)&sk->sk_dst_cache, dst); |
7f502361 | 1678 | dst_release(old_dst); |
1da177e4 LT |
1679 | } |
1680 | ||
1681 | static inline void | |
1682 | __sk_dst_reset(struct sock *sk) | |
1683 | { | |
b6c6712a | 1684 | __sk_dst_set(sk, NULL); |
1da177e4 LT |
1685 | } |
1686 | ||
1687 | static inline void | |
1688 | sk_dst_reset(struct sock *sk) | |
1689 | { | |
7f502361 | 1690 | sk_dst_set(sk, NULL); |
1da177e4 LT |
1691 | } |
1692 | ||
69336bd2 | 1693 | struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie); |
1da177e4 | 1694 | |
69336bd2 | 1695 | struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie); |
1da177e4 | 1696 | |
f60e5990 | 1697 | bool sk_mc_loop(struct sock *sk); |
1698 | ||
dc6b9b78 | 1699 | static inline bool sk_can_gso(const struct sock *sk) |
bcd76111 HX |
1700 | { |
1701 | return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type); | |
1702 | } | |
1703 | ||
69336bd2 | 1704 | void sk_setup_caps(struct sock *sk, struct dst_entry *dst); |
6cbb0df7 | 1705 | |
c8f44aff | 1706 | static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags) |
a465419b ED |
1707 | { |
1708 | sk->sk_route_nocaps |= flags; | |
1709 | sk->sk_route_caps &= ~flags; | |
1710 | } | |
1711 | ||
9a49850d TH |
1712 | static inline bool sk_check_csum_caps(struct sock *sk) |
1713 | { | |
1714 | return (sk->sk_route_caps & NETIF_F_HW_CSUM) || | |
1715 | (sk->sk_family == PF_INET && | |
1716 | (sk->sk_route_caps & NETIF_F_IP_CSUM)) || | |
1717 | (sk->sk_family == PF_INET6 && | |
1718 | (sk->sk_route_caps & NETIF_F_IPV6_CSUM)); | |
1719 | } | |
1720 | ||
c6e1a0d1 | 1721 | static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb, |
57be5bda | 1722 | struct iov_iter *from, char *to, |
912d398d | 1723 | int copy, int offset) |
c6e1a0d1 TH |
1724 | { |
1725 | if (skb->ip_summed == CHECKSUM_NONE) { | |
57be5bda AV |
1726 | __wsum csum = 0; |
1727 | if (csum_and_copy_from_iter(to, copy, &csum, from) != copy) | |
1728 | return -EFAULT; | |
912d398d | 1729 | skb->csum = csum_block_add(skb->csum, csum, offset); |
c6e1a0d1 | 1730 | } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) { |
57be5bda | 1731 | if (copy_from_iter_nocache(to, copy, from) != copy) |
c6e1a0d1 | 1732 | return -EFAULT; |
57be5bda | 1733 | } else if (copy_from_iter(to, copy, from) != copy) |
c6e1a0d1 TH |
1734 | return -EFAULT; |
1735 | ||
1736 | return 0; | |
1737 | } | |
1738 | ||
1739 | static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb, | |
57be5bda | 1740 | struct iov_iter *from, int copy) |
c6e1a0d1 | 1741 | { |
912d398d | 1742 | int err, offset = skb->len; |
c6e1a0d1 | 1743 | |
912d398d WY |
1744 | err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy), |
1745 | copy, offset); | |
c6e1a0d1 | 1746 | if (err) |
912d398d | 1747 | __skb_trim(skb, offset); |
c6e1a0d1 TH |
1748 | |
1749 | return err; | |
1750 | } | |
1751 | ||
57be5bda | 1752 | static inline int skb_copy_to_page_nocache(struct sock *sk, struct iov_iter *from, |
c6e1a0d1 TH |
1753 | struct sk_buff *skb, |
1754 | struct page *page, | |
1755 | int off, int copy) | |
1756 | { | |
1757 | int err; | |
1758 | ||
912d398d WY |
1759 | err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off, |
1760 | copy, skb->len); | |
c6e1a0d1 TH |
1761 | if (err) |
1762 | return err; | |
1763 | ||
1764 | skb->len += copy; | |
1765 | skb->data_len += copy; | |
1766 | skb->truesize += copy; | |
1767 | sk->sk_wmem_queued += copy; | |
1768 | sk_mem_charge(sk, copy); | |
1769 | return 0; | |
1770 | } | |
1771 | ||
c564039f ED |
1772 | /** |
1773 | * sk_wmem_alloc_get - returns write allocations | |
1774 | * @sk: socket | |
1775 | * | |
1776 | * Returns sk_wmem_alloc minus initial offset of one | |
1777 | */ | |
1778 | static inline int sk_wmem_alloc_get(const struct sock *sk) | |
1779 | { | |
1780 | return atomic_read(&sk->sk_wmem_alloc) - 1; | |
1781 | } | |
1782 | ||
1783 | /** | |
1784 | * sk_rmem_alloc_get - returns read allocations | |
1785 | * @sk: socket | |
1786 | * | |
1787 | * Returns sk_rmem_alloc | |
1788 | */ | |
1789 | static inline int sk_rmem_alloc_get(const struct sock *sk) | |
1790 | { | |
1791 | return atomic_read(&sk->sk_rmem_alloc); | |
1792 | } | |
1793 | ||
1794 | /** | |
1795 | * sk_has_allocations - check if allocations are outstanding | |
1796 | * @sk: socket | |
1797 | * | |
1798 | * Returns true if socket has write or read allocations | |
1799 | */ | |
dc6b9b78 | 1800 | static inline bool sk_has_allocations(const struct sock *sk) |
c564039f ED |
1801 | { |
1802 | return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk); | |
1803 | } | |
1804 | ||
a57de0b4 | 1805 | /** |
1ce0bf50 | 1806 | * skwq_has_sleeper - check if there are any waiting processes |
acfbe96a | 1807 | * @wq: struct socket_wq |
a57de0b4 | 1808 | * |
43815482 | 1809 | * Returns true if socket_wq has waiting processes |
a57de0b4 | 1810 | * |
1ce0bf50 | 1811 | * The purpose of the skwq_has_sleeper and sock_poll_wait is to wrap the memory |
a57de0b4 JO |
1812 | * barrier call. They were added due to the race found within the tcp code. |
1813 | * | |
1814 | * Consider following tcp code paths: | |
1815 | * | |
1816 | * CPU1 CPU2 | |
1817 | * | |
1818 | * sys_select receive packet | |
1819 | * ... ... | |
1820 | * __add_wait_queue update tp->rcv_nxt | |
1821 | * ... ... | |
1822 | * tp->rcv_nxt check sock_def_readable | |
1823 | * ... { | |
43815482 ED |
1824 | * schedule rcu_read_lock(); |
1825 | * wq = rcu_dereference(sk->sk_wq); | |
1826 | * if (wq && waitqueue_active(&wq->wait)) | |
1827 | * wake_up_interruptible(&wq->wait) | |
a57de0b4 JO |
1828 | * ... |
1829 | * } | |
1830 | * | |
1831 | * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay | |
1832 | * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1 | |
1833 | * could then endup calling schedule and sleep forever if there are no more | |
1834 | * data on the socket. | |
ad462769 | 1835 | * |
a57de0b4 | 1836 | */ |
1ce0bf50 | 1837 | static inline bool skwq_has_sleeper(struct socket_wq *wq) |
a57de0b4 | 1838 | { |
1ce0bf50 | 1839 | return wq && wq_has_sleeper(&wq->wait); |
a57de0b4 JO |
1840 | } |
1841 | ||
1842 | /** | |
1843 | * sock_poll_wait - place memory barrier behind the poll_wait call. | |
1844 | * @filp: file | |
1845 | * @wait_address: socket wait queue | |
1846 | * @p: poll_table | |
1847 | * | |
43815482 | 1848 | * See the comments in the wq_has_sleeper function. |
a57de0b4 JO |
1849 | */ |
1850 | static inline void sock_poll_wait(struct file *filp, | |
1851 | wait_queue_head_t *wait_address, poll_table *p) | |
1852 | { | |
626cf236 | 1853 | if (!poll_does_not_wait(p) && wait_address) { |
a57de0b4 | 1854 | poll_wait(filp, wait_address, p); |
dc6b9b78 | 1855 | /* We need to be sure we are in sync with the |
a57de0b4 JO |
1856 | * socket flags modification. |
1857 | * | |
43815482 | 1858 | * This memory barrier is paired in the wq_has_sleeper. |
dc6b9b78 | 1859 | */ |
a57de0b4 JO |
1860 | smp_mb(); |
1861 | } | |
1862 | } | |
1863 | ||
b73c3d0e TH |
1864 | static inline void skb_set_hash_from_sk(struct sk_buff *skb, struct sock *sk) |
1865 | { | |
1866 | if (sk->sk_txhash) { | |
1867 | skb->l4_hash = 1; | |
1868 | skb->hash = sk->sk_txhash; | |
1869 | } | |
1870 | } | |
1871 | ||
9e17f8a4 ED |
1872 | void skb_set_owner_w(struct sk_buff *skb, struct sock *sk); |
1873 | ||
1da177e4 | 1874 | /* |
dc6b9b78 | 1875 | * Queue a received datagram if it will fit. Stream and sequenced |
1da177e4 LT |
1876 | * protocols can't normally use this as they need to fit buffers in |
1877 | * and play with them. | |
1878 | * | |
dc6b9b78 | 1879 | * Inlined as it's very short and called for pretty much every |
1da177e4 LT |
1880 | * packet ever received. |
1881 | */ | |
1da177e4 LT |
1882 | static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk) |
1883 | { | |
d55d87fd | 1884 | skb_orphan(skb); |
1da177e4 LT |
1885 | skb->sk = sk; |
1886 | skb->destructor = sock_rfree; | |
1887 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
3ab224be | 1888 | sk_mem_charge(sk, skb->truesize); |
1da177e4 LT |
1889 | } |
1890 | ||
69336bd2 JP |
1891 | void sk_reset_timer(struct sock *sk, struct timer_list *timer, |
1892 | unsigned long expires); | |
1da177e4 | 1893 | |
69336bd2 | 1894 | void sk_stop_timer(struct sock *sk, struct timer_list *timer); |
1da177e4 | 1895 | |
69336bd2 | 1896 | int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb); |
1da177e4 | 1897 | |
69336bd2 | 1898 | int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb); |
364a9e93 | 1899 | struct sk_buff *sock_dequeue_err_skb(struct sock *sk); |
1da177e4 LT |
1900 | |
1901 | /* | |
1902 | * Recover an error report and clear atomically | |
1903 | */ | |
dc6b9b78 | 1904 | |
1da177e4 LT |
1905 | static inline int sock_error(struct sock *sk) |
1906 | { | |
c1cbe4b7 BL |
1907 | int err; |
1908 | if (likely(!sk->sk_err)) | |
1909 | return 0; | |
1910 | err = xchg(&sk->sk_err, 0); | |
1da177e4 LT |
1911 | return -err; |
1912 | } | |
1913 | ||
1914 | static inline unsigned long sock_wspace(struct sock *sk) | |
1915 | { | |
1916 | int amt = 0; | |
1917 | ||
1918 | if (!(sk->sk_shutdown & SEND_SHUTDOWN)) { | |
1919 | amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc); | |
dc6b9b78 | 1920 | if (amt < 0) |
1da177e4 LT |
1921 | amt = 0; |
1922 | } | |
1923 | return amt; | |
1924 | } | |
1925 | ||
ceb5d58b ED |
1926 | /* Note: |
1927 | * We use sk->sk_wq_raw, from contexts knowing this | |
1928 | * pointer is not NULL and cannot disappear/change. | |
1929 | */ | |
9cd3e072 | 1930 | static inline void sk_set_bit(int nr, struct sock *sk) |
1da177e4 | 1931 | { |
ceb5d58b | 1932 | set_bit(nr, &sk->sk_wq_raw->flags); |
9cd3e072 ED |
1933 | } |
1934 | ||
1935 | static inline void sk_clear_bit(int nr, struct sock *sk) | |
1936 | { | |
ceb5d58b | 1937 | clear_bit(nr, &sk->sk_wq_raw->flags); |
9cd3e072 ED |
1938 | } |
1939 | ||
ceb5d58b | 1940 | static inline void sk_wake_async(const struct sock *sk, int how, int band) |
1da177e4 | 1941 | { |
ceb5d58b ED |
1942 | if (sock_flag(sk, SOCK_FASYNC)) { |
1943 | rcu_read_lock(); | |
1944 | sock_wake_async(rcu_dereference(sk->sk_wq), how, band); | |
1945 | rcu_read_unlock(); | |
1946 | } | |
1da177e4 LT |
1947 | } |
1948 | ||
eea86af6 DB |
1949 | /* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might |
1950 | * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak. | |
1951 | * Note: for send buffers, TCP works better if we can build two skbs at | |
1952 | * minimum. | |
7a91b434 | 1953 | */ |
9eb5bf83 | 1954 | #define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff))) |
eea86af6 DB |
1955 | |
1956 | #define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2) | |
1957 | #define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE | |
1da177e4 LT |
1958 | |
1959 | static inline void sk_stream_moderate_sndbuf(struct sock *sk) | |
1960 | { | |
1961 | if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) { | |
8df09ea3 | 1962 | sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1); |
eea86af6 | 1963 | sk->sk_sndbuf = max_t(u32, sk->sk_sndbuf, SOCK_MIN_SNDBUF); |
1da177e4 LT |
1964 | } |
1965 | } | |
1966 | ||
eb934478 ED |
1967 | struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp, |
1968 | bool force_schedule); | |
1da177e4 | 1969 | |
5640f768 ED |
1970 | /** |
1971 | * sk_page_frag - return an appropriate page_frag | |
1972 | * @sk: socket | |
1973 | * | |
1974 | * If socket allocation mode allows current thread to sleep, it means its | |
1975 | * safe to use the per task page_frag instead of the per socket one. | |
1976 | */ | |
1977 | static inline struct page_frag *sk_page_frag(struct sock *sk) | |
1da177e4 | 1978 | { |
d0164adc | 1979 | if (gfpflags_allow_blocking(sk->sk_allocation)) |
5640f768 | 1980 | return ¤t->task_frag; |
1da177e4 | 1981 | |
5640f768 | 1982 | return &sk->sk_frag; |
1da177e4 LT |
1983 | } |
1984 | ||
69336bd2 | 1985 | bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag); |
5640f768 | 1986 | |
1da177e4 LT |
1987 | /* |
1988 | * Default write policy as shown to user space via poll/select/SIGIO | |
1989 | */ | |
dc6b9b78 | 1990 | static inline bool sock_writeable(const struct sock *sk) |
1da177e4 | 1991 | { |
8df09ea3 | 1992 | return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1); |
1da177e4 LT |
1993 | } |
1994 | ||
dd0fc66f | 1995 | static inline gfp_t gfp_any(void) |
1da177e4 | 1996 | { |
99709372 | 1997 | return in_softirq() ? GFP_ATOMIC : GFP_KERNEL; |
1da177e4 LT |
1998 | } |
1999 | ||
dc6b9b78 | 2000 | static inline long sock_rcvtimeo(const struct sock *sk, bool noblock) |
1da177e4 LT |
2001 | { |
2002 | return noblock ? 0 : sk->sk_rcvtimeo; | |
2003 | } | |
2004 | ||
dc6b9b78 | 2005 | static inline long sock_sndtimeo(const struct sock *sk, bool noblock) |
1da177e4 LT |
2006 | { |
2007 | return noblock ? 0 : sk->sk_sndtimeo; | |
2008 | } | |
2009 | ||
2010 | static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len) | |
2011 | { | |
2012 | return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1; | |
2013 | } | |
2014 | ||
2015 | /* Alas, with timeout socket operations are not restartable. | |
2016 | * Compare this to poll(). | |
2017 | */ | |
2018 | static inline int sock_intr_errno(long timeo) | |
2019 | { | |
2020 | return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR; | |
2021 | } | |
2022 | ||
744d5a3e EB |
2023 | struct sock_skb_cb { |
2024 | u32 dropcount; | |
2025 | }; | |
2026 | ||
2027 | /* Store sock_skb_cb at the end of skb->cb[] so protocol families | |
2028 | * using skb->cb[] would keep using it directly and utilize its | |
2029 | * alignement guarantee. | |
2030 | */ | |
2031 | #define SOCK_SKB_CB_OFFSET ((FIELD_SIZEOF(struct sk_buff, cb) - \ | |
2032 | sizeof(struct sock_skb_cb))) | |
2033 | ||
2034 | #define SOCK_SKB_CB(__skb) ((struct sock_skb_cb *)((__skb)->cb + \ | |
2035 | SOCK_SKB_CB_OFFSET)) | |
2036 | ||
b4772ef8 | 2037 | #define sock_skb_cb_check_size(size) \ |
744d5a3e | 2038 | BUILD_BUG_ON((size) > SOCK_SKB_CB_OFFSET) |
b4772ef8 | 2039 | |
3bc3b96f EB |
2040 | static inline void |
2041 | sock_skb_set_dropcount(const struct sock *sk, struct sk_buff *skb) | |
2042 | { | |
744d5a3e | 2043 | SOCK_SKB_CB(skb)->dropcount = atomic_read(&sk->sk_drops); |
3bc3b96f EB |
2044 | } |
2045 | ||
69336bd2 JP |
2046 | void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, |
2047 | struct sk_buff *skb); | |
2048 | void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk, | |
2049 | struct sk_buff *skb); | |
92f37fd2 | 2050 | |
dc6b9b78 | 2051 | static inline void |
1da177e4 LT |
2052 | sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb) |
2053 | { | |
b7aa0bf7 | 2054 | ktime_t kt = skb->tstamp; |
20d49473 | 2055 | struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb); |
a61bbcf2 | 2056 | |
20d49473 PO |
2057 | /* |
2058 | * generate control messages if | |
b9f40e21 | 2059 | * - receive time stamping in software requested |
20d49473 | 2060 | * - software time stamp available and wanted |
20d49473 | 2061 | * - hardware time stamps available and wanted |
20d49473 PO |
2062 | */ |
2063 | if (sock_flag(sk, SOCK_RCVTSTAMP) || | |
b9f40e21 | 2064 | (sk->sk_tsflags & SOF_TIMESTAMPING_RX_SOFTWARE) || |
c199105d | 2065 | (kt.tv64 && sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) || |
20d49473 | 2066 | (hwtstamps->hwtstamp.tv64 && |
b9f40e21 | 2067 | (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE))) |
92f37fd2 ED |
2068 | __sock_recv_timestamp(msg, sk, skb); |
2069 | else | |
b7aa0bf7 | 2070 | sk->sk_stamp = kt; |
6e3e939f JB |
2071 | |
2072 | if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid) | |
2073 | __sock_recv_wifi_status(msg, sk, skb); | |
1da177e4 LT |
2074 | } |
2075 | ||
69336bd2 JP |
2076 | void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, |
2077 | struct sk_buff *skb); | |
767dd033 ED |
2078 | |
2079 | static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, | |
2080 | struct sk_buff *skb) | |
2081 | { | |
2082 | #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \ | |
b9f40e21 WB |
2083 | (1UL << SOCK_RCVTSTAMP)) |
2084 | #define TSFLAGS_ANY (SOF_TIMESTAMPING_SOFTWARE | \ | |
2085 | SOF_TIMESTAMPING_RAW_HARDWARE) | |
767dd033 | 2086 | |
b9f40e21 | 2087 | if (sk->sk_flags & FLAGS_TS_OR_DROPS || sk->sk_tsflags & TSFLAGS_ANY) |
767dd033 ED |
2088 | __sock_recv_ts_and_drops(msg, sk, skb); |
2089 | else | |
2090 | sk->sk_stamp = skb->tstamp; | |
2091 | } | |
3b885787 | 2092 | |
67cc0d40 WB |
2093 | void __sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags); |
2094 | ||
20d49473 PO |
2095 | /** |
2096 | * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped | |
20d49473 | 2097 | * @sk: socket sending this packet |
140c55d4 ED |
2098 | * @tx_flags: completed with instructions for time stamping |
2099 | * | |
2100 | * Note : callers should take care of initial *tx_flags value (usually 0) | |
20d49473 | 2101 | */ |
67cc0d40 WB |
2102 | static inline void sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags) |
2103 | { | |
2104 | if (unlikely(sk->sk_tsflags)) | |
2105 | __sock_tx_timestamp(sk, tx_flags); | |
2106 | if (unlikely(sock_flag(sk, SOCK_WIFI_STATUS))) | |
2107 | *tx_flags |= SKBTX_WIFI_STATUS; | |
2108 | } | |
20d49473 | 2109 | |
1da177e4 LT |
2110 | /** |
2111 | * sk_eat_skb - Release a skb if it is no longer needed | |
4dc3b16b PP |
2112 | * @sk: socket to eat this skb from |
2113 | * @skb: socket buffer to eat | |
1da177e4 LT |
2114 | * |
2115 | * This routine must be called with interrupts disabled or with the socket | |
2116 | * locked so that the sk_buff queue operation is ok. | |
2117 | */ | |
7bced397 | 2118 | static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
2119 | { |
2120 | __skb_unlink(skb, &sk->sk_receive_queue); | |
2121 | __kfree_skb(skb); | |
2122 | } | |
2123 | ||
3b1e0a65 YH |
2124 | static inline |
2125 | struct net *sock_net(const struct sock *sk) | |
2126 | { | |
c2d9ba9b | 2127 | return read_pnet(&sk->sk_net); |
3b1e0a65 YH |
2128 | } |
2129 | ||
2130 | static inline | |
f5aa23fd | 2131 | void sock_net_set(struct sock *sk, struct net *net) |
3b1e0a65 | 2132 | { |
c2d9ba9b | 2133 | write_pnet(&sk->sk_net, net); |
3b1e0a65 YH |
2134 | } |
2135 | ||
23542618 KK |
2136 | static inline struct sock *skb_steal_sock(struct sk_buff *skb) |
2137 | { | |
efc27f8c | 2138 | if (skb->sk) { |
23542618 KK |
2139 | struct sock *sk = skb->sk; |
2140 | ||
2141 | skb->destructor = NULL; | |
2142 | skb->sk = NULL; | |
2143 | return sk; | |
2144 | } | |
2145 | return NULL; | |
2146 | } | |
2147 | ||
1d0ab253 ED |
2148 | /* This helper checks if a socket is a full socket, |
2149 | * ie _not_ a timewait or request socket. | |
2150 | */ | |
2151 | static inline bool sk_fullsock(const struct sock *sk) | |
2152 | { | |
2153 | return (1 << sk->sk_state) & ~(TCPF_TIME_WAIT | TCPF_NEW_SYN_RECV); | |
2154 | } | |
2155 | ||
e446f9df ED |
2156 | /* This helper checks if a socket is a LISTEN or NEW_SYN_RECV |
2157 | * SYNACK messages can be attached to either ones (depending on SYNCOOKIE) | |
2158 | */ | |
2159 | static inline bool sk_listener(const struct sock *sk) | |
2160 | { | |
2161 | return (1 << sk->sk_state) & (TCPF_LISTEN | TCPF_NEW_SYN_RECV); | |
2162 | } | |
2163 | ||
00fd38d9 ED |
2164 | /** |
2165 | * sk_state_load - read sk->sk_state for lockless contexts | |
2166 | * @sk: socket pointer | |
2167 | * | |
2168 | * Paired with sk_state_store(). Used in places we do not hold socket lock : | |
2169 | * tcp_diag_get_info(), tcp_get_info(), tcp_poll(), get_tcp4_sock() ... | |
2170 | */ | |
2171 | static inline int sk_state_load(const struct sock *sk) | |
2172 | { | |
2173 | return smp_load_acquire(&sk->sk_state); | |
2174 | } | |
2175 | ||
2176 | /** | |
2177 | * sk_state_store - update sk->sk_state | |
2178 | * @sk: socket pointer | |
2179 | * @newstate: new state | |
2180 | * | |
2181 | * Paired with sk_state_load(). Should be used in contexts where | |
2182 | * state change might impact lockless readers. | |
2183 | */ | |
2184 | static inline void sk_state_store(struct sock *sk, int newstate) | |
2185 | { | |
2186 | smp_store_release(&sk->sk_state, newstate); | |
2187 | } | |
2188 | ||
69336bd2 JP |
2189 | void sock_enable_timestamp(struct sock *sk, int flag); |
2190 | int sock_get_timestamp(struct sock *, struct timeval __user *); | |
2191 | int sock_get_timestampns(struct sock *, struct timespec __user *); | |
2192 | int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len, int level, | |
2193 | int type); | |
1da177e4 | 2194 | |
a3b299da EB |
2195 | bool sk_ns_capable(const struct sock *sk, |
2196 | struct user_namespace *user_ns, int cap); | |
2197 | bool sk_capable(const struct sock *sk, int cap); | |
2198 | bool sk_net_capable(const struct sock *sk, int cap); | |
2199 | ||
1da177e4 LT |
2200 | extern __u32 sysctl_wmem_max; |
2201 | extern __u32 sysctl_rmem_max; | |
2202 | ||
b245be1f | 2203 | extern int sysctl_tstamp_allow_data; |
6baf1f41 DM |
2204 | extern int sysctl_optmem_max; |
2205 | ||
20380731 ACM |
2206 | extern __u32 sysctl_wmem_default; |
2207 | extern __u32 sysctl_rmem_default; | |
20380731 | 2208 | |
1da177e4 | 2209 | #endif /* _SOCK_H */ |