]>
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 AV |
60 | #include <linux/aio.h> |
61 | #include <linux/sched.h> | |
1da177e4 LT |
62 | |
63 | #include <linux/filter.h> | |
88ab1932 | 64 | #include <linux/rculist_nulls.h> |
a57de0b4 | 65 | #include <linux/poll.h> |
1da177e4 | 66 | |
c31504dc | 67 | #include <linux/atomic.h> |
1da177e4 LT |
68 | #include <net/dst.h> |
69 | #include <net/checksum.h> | |
b9f40e21 | 70 | #include <linux/net_tstamp.h> |
1da177e4 | 71 | |
9f048bfb ED |
72 | struct cgroup; |
73 | struct cgroup_subsys; | |
c607b2ed | 74 | #ifdef CONFIG_NET |
1d62e436 GC |
75 | int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss); |
76 | void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg); | |
c607b2ed GC |
77 | #else |
78 | static inline | |
1d62e436 | 79 | int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss) |
c607b2ed GC |
80 | { |
81 | return 0; | |
82 | } | |
83 | static inline | |
1d62e436 | 84 | void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg) |
c607b2ed GC |
85 | { |
86 | } | |
87 | #endif | |
1da177e4 LT |
88 | /* |
89 | * This structure really needs to be cleaned up. | |
90 | * Most of it is for TCP, and not used by any of | |
91 | * the other protocols. | |
92 | */ | |
93 | ||
94 | /* Define this to get the SOCK_DBG debugging facility. */ | |
95 | #define SOCK_DEBUGGING | |
96 | #ifdef SOCK_DEBUGGING | |
97 | #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \ | |
98 | printk(KERN_DEBUG msg); } while (0) | |
99 | #else | |
4cd9029d | 100 | /* Validate arguments and do nothing */ |
b9075fa9 | 101 | static inline __printf(2, 3) |
dc6b9b78 | 102 | void SOCK_DEBUG(const struct sock *sk, const char *msg, ...) |
4cd9029d SH |
103 | { |
104 | } | |
1da177e4 LT |
105 | #endif |
106 | ||
107 | /* This is the per-socket lock. The spinlock provides a synchronization | |
108 | * between user contexts and software interrupt processing, whereas the | |
109 | * mini-semaphore synchronizes multiple users amongst themselves. | |
110 | */ | |
1da177e4 LT |
111 | typedef struct { |
112 | spinlock_t slock; | |
d2e9117c | 113 | int owned; |
1da177e4 | 114 | wait_queue_head_t wq; |
a5b5bb9a IM |
115 | /* |
116 | * We express the mutex-alike socket_lock semantics | |
117 | * to the lock validator by explicitly managing | |
118 | * the slock as a lock variant (in addition to | |
119 | * the slock itself): | |
120 | */ | |
121 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
122 | struct lockdep_map dep_map; | |
123 | #endif | |
1da177e4 LT |
124 | } socket_lock_t; |
125 | ||
1da177e4 | 126 | struct sock; |
8feaf0c0 | 127 | struct proto; |
0eeb8ffc | 128 | struct net; |
1da177e4 | 129 | |
077b393d ED |
130 | typedef __u32 __bitwise __portpair; |
131 | typedef __u64 __bitwise __addrpair; | |
132 | ||
1da177e4 | 133 | /** |
4dc3b16b | 134 | * struct sock_common - minimal network layer representation of sockets |
68835aba ED |
135 | * @skc_daddr: Foreign IPv4 addr |
136 | * @skc_rcv_saddr: Bound local IPv4 addr | |
4dc6dc71 | 137 | * @skc_hash: hash value used with various protocol lookup tables |
d4cada4a | 138 | * @skc_u16hashes: two u16 hash values used by UDP lookup tables |
ce43b03e ED |
139 | * @skc_dport: placeholder for inet_dport/tw_dport |
140 | * @skc_num: placeholder for inet_num/tw_num | |
4dc3b16b PP |
141 | * @skc_family: network address family |
142 | * @skc_state: Connection state | |
143 | * @skc_reuse: %SO_REUSEADDR setting | |
055dc21a | 144 | * @skc_reuseport: %SO_REUSEPORT setting |
4dc3b16b | 145 | * @skc_bound_dev_if: bound device index if != 0 |
4dc3b16b | 146 | * @skc_bind_node: bind hash linkage for various protocol lookup tables |
512615b6 | 147 | * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol |
8feaf0c0 | 148 | * @skc_prot: protocol handlers inside a network family |
07feaebf | 149 | * @skc_net: reference to the network namespace of this socket |
68835aba ED |
150 | * @skc_node: main hash linkage for various protocol lookup tables |
151 | * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol | |
152 | * @skc_tx_queue_mapping: tx queue number for this connection | |
153 | * @skc_refcnt: reference count | |
4dc3b16b PP |
154 | * |
155 | * This is the minimal network layer representation of sockets, the header | |
8feaf0c0 ACM |
156 | * for struct sock and struct inet_timewait_sock. |
157 | */ | |
1da177e4 | 158 | struct sock_common { |
ce43b03e | 159 | /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned |
05dbc7b5 | 160 | * address on 64bit arches : cf INET_MATCH() |
4dc6dc71 | 161 | */ |
ce43b03e | 162 | union { |
077b393d | 163 | __addrpair skc_addrpair; |
ce43b03e ED |
164 | struct { |
165 | __be32 skc_daddr; | |
166 | __be32 skc_rcv_saddr; | |
167 | }; | |
168 | }; | |
d4cada4a ED |
169 | union { |
170 | unsigned int skc_hash; | |
171 | __u16 skc_u16hashes[2]; | |
172 | }; | |
ce43b03e ED |
173 | /* skc_dport && skc_num must be grouped as well */ |
174 | union { | |
077b393d | 175 | __portpair skc_portpair; |
ce43b03e ED |
176 | struct { |
177 | __be16 skc_dport; | |
178 | __u16 skc_num; | |
179 | }; | |
180 | }; | |
181 | ||
4dc6dc71 ED |
182 | unsigned short skc_family; |
183 | volatile unsigned char skc_state; | |
055dc21a | 184 | unsigned char skc_reuse:4; |
9fe516ba ED |
185 | unsigned char skc_reuseport:1; |
186 | unsigned char skc_ipv6only:1; | |
4dc6dc71 | 187 | int skc_bound_dev_if; |
512615b6 ED |
188 | union { |
189 | struct hlist_node skc_bind_node; | |
190 | struct hlist_nulls_node skc_portaddr_node; | |
191 | }; | |
8feaf0c0 | 192 | struct proto *skc_prot; |
3b1e0a65 | 193 | #ifdef CONFIG_NET_NS |
07feaebf | 194 | struct net *skc_net; |
3b1e0a65 | 195 | #endif |
efe4208f ED |
196 | |
197 | #if IS_ENABLED(CONFIG_IPV6) | |
198 | struct in6_addr skc_v6_daddr; | |
199 | struct in6_addr skc_v6_rcv_saddr; | |
200 | #endif | |
201 | ||
68835aba ED |
202 | /* |
203 | * fields between dontcopy_begin/dontcopy_end | |
204 | * are not copied in sock_copy() | |
205 | */ | |
928c41e7 | 206 | /* private: */ |
68835aba | 207 | int skc_dontcopy_begin[0]; |
928c41e7 | 208 | /* public: */ |
68835aba ED |
209 | union { |
210 | struct hlist_node skc_node; | |
211 | struct hlist_nulls_node skc_nulls_node; | |
212 | }; | |
213 | int skc_tx_queue_mapping; | |
214 | atomic_t skc_refcnt; | |
928c41e7 | 215 | /* private: */ |
68835aba | 216 | int skc_dontcopy_end[0]; |
928c41e7 | 217 | /* public: */ |
1da177e4 LT |
218 | }; |
219 | ||
e1aab161 | 220 | struct cg_proto; |
1da177e4 LT |
221 | /** |
222 | * struct sock - network layer representation of sockets | |
8feaf0c0 | 223 | * @__sk_common: shared layout with inet_timewait_sock |
4dc3b16b PP |
224 | * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN |
225 | * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings | |
226 | * @sk_lock: synchronizer | |
227 | * @sk_rcvbuf: size of receive buffer in bytes | |
43815482 | 228 | * @sk_wq: sock wait queue and async head |
421b3885 | 229 | * @sk_rx_dst: receive input route used by early demux |
4dc3b16b PP |
230 | * @sk_dst_cache: destination cache |
231 | * @sk_dst_lock: destination cache lock | |
232 | * @sk_policy: flow policy | |
4dc3b16b PP |
233 | * @sk_receive_queue: incoming packets |
234 | * @sk_wmem_alloc: transmit queue bytes committed | |
235 | * @sk_write_queue: Packet sending queue | |
236 | * @sk_omem_alloc: "o" is "option" or "other" | |
237 | * @sk_wmem_queued: persistent queue size | |
238 | * @sk_forward_alloc: space allocated forward | |
06021292 | 239 | * @sk_napi_id: id of the last napi context to receive data for sk |
dafcc438 | 240 | * @sk_ll_usec: usecs to busypoll when there is no data |
4dc3b16b | 241 | * @sk_allocation: allocation mode |
95bd09eb | 242 | * @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler) |
c3f40d7c | 243 | * @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE) |
4dc3b16b | 244 | * @sk_sndbuf: size of send buffer in bytes |
33c732c3 | 245 | * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, |
20d49473 | 246 | * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings |
28448b80 TH |
247 | * @sk_no_check_tx: %SO_NO_CHECK setting, set checksum in TX packets |
248 | * @sk_no_check_rx: allow zero checksum in RX packets | |
4dc3b16b | 249 | * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO) |
a465419b | 250 | * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK) |
bcd76111 | 251 | * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4) |
82cc1a7a | 252 | * @sk_gso_max_size: Maximum GSO segment size to build |
1485348d | 253 | * @sk_gso_max_segs: Maximum number of GSO segments |
4dc3b16b | 254 | * @sk_lingertime: %SO_LINGER l_linger setting |
4dc3b16b PP |
255 | * @sk_backlog: always used with the per-socket spinlock held |
256 | * @sk_callback_lock: used with the callbacks in the end of this struct | |
257 | * @sk_error_queue: rarely used | |
33c732c3 WC |
258 | * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, |
259 | * IPV6_ADDRFORM for instance) | |
4dc3b16b | 260 | * @sk_err: last error |
33c732c3 WC |
261 | * @sk_err_soft: errors that don't cause failure but are the cause of a |
262 | * persistent failure not just 'timed out' | |
cb61cb9b | 263 | * @sk_drops: raw/udp drops counter |
4dc3b16b PP |
264 | * @sk_ack_backlog: current listen backlog |
265 | * @sk_max_ack_backlog: listen backlog set in listen() | |
266 | * @sk_priority: %SO_PRIORITY setting | |
1a3bc369 | 267 | * @sk_cgrp_prioidx: socket group's priority map index |
4dc3b16b PP |
268 | * @sk_type: socket type (%SOCK_STREAM, etc) |
269 | * @sk_protocol: which protocol this socket belongs in this network family | |
53c3fa20 RD |
270 | * @sk_peer_pid: &struct pid for this socket's peer |
271 | * @sk_peer_cred: %SO_PEERCRED setting | |
4dc3b16b PP |
272 | * @sk_rcvlowat: %SO_RCVLOWAT setting |
273 | * @sk_rcvtimeo: %SO_RCVTIMEO setting | |
274 | * @sk_sndtimeo: %SO_SNDTIMEO setting | |
c58dc01b | 275 | * @sk_rxhash: flow hash received from netif layer |
2c8c56e1 | 276 | * @sk_incoming_cpu: record cpu processing incoming packets |
b73c3d0e | 277 | * @sk_txhash: computed flow hash for use on transmit |
4dc3b16b PP |
278 | * @sk_filter: socket filtering instructions |
279 | * @sk_protinfo: private area, net family specific, when not using slab | |
280 | * @sk_timer: sock cleanup timer | |
281 | * @sk_stamp: time stamp of last packet received | |
b9f40e21 | 282 | * @sk_tsflags: SO_TIMESTAMPING socket options |
09c2d251 | 283 | * @sk_tskey: counter to disambiguate concurrent tstamp requests |
4dc3b16b PP |
284 | * @sk_socket: Identd and reporting IO signals |
285 | * @sk_user_data: RPC layer private data | |
5640f768 | 286 | * @sk_frag: cached page frag |
d3d4f0a0 | 287 | * @sk_peek_off: current peek_offset value |
4dc3b16b | 288 | * @sk_send_head: front of stuff to transmit |
67be2dd1 | 289 | * @sk_security: used by security modules |
31729363 | 290 | * @sk_mark: generic packet mark |
53c3fa20 | 291 | * @sk_classid: this socket's cgroup classid |
e1aab161 | 292 | * @sk_cgrp: this socket's cgroup-specific proto data |
4dc3b16b PP |
293 | * @sk_write_pending: a write to stream socket waits to start |
294 | * @sk_state_change: callback to indicate change in the state of the sock | |
295 | * @sk_data_ready: callback to indicate there is data to be processed | |
296 | * @sk_write_space: callback to indicate there is bf sending space available | |
297 | * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE) | |
298 | * @sk_backlog_rcv: callback to process the backlog | |
299 | * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0 | |
1da177e4 LT |
300 | */ |
301 | struct sock { | |
302 | /* | |
8feaf0c0 | 303 | * Now struct inet_timewait_sock also uses sock_common, so please just |
1da177e4 LT |
304 | * don't add nothing before this first member (__sk_common) --acme |
305 | */ | |
306 | struct sock_common __sk_common; | |
4dc6dc71 ED |
307 | #define sk_node __sk_common.skc_node |
308 | #define sk_nulls_node __sk_common.skc_nulls_node | |
309 | #define sk_refcnt __sk_common.skc_refcnt | |
e022f0b4 | 310 | #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping |
4dc6dc71 | 311 | |
68835aba ED |
312 | #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin |
313 | #define sk_dontcopy_end __sk_common.skc_dontcopy_end | |
4dc6dc71 | 314 | #define sk_hash __sk_common.skc_hash |
50805466 | 315 | #define sk_portpair __sk_common.skc_portpair |
05dbc7b5 ED |
316 | #define sk_num __sk_common.skc_num |
317 | #define sk_dport __sk_common.skc_dport | |
50805466 ED |
318 | #define sk_addrpair __sk_common.skc_addrpair |
319 | #define sk_daddr __sk_common.skc_daddr | |
320 | #define sk_rcv_saddr __sk_common.skc_rcv_saddr | |
1da177e4 LT |
321 | #define sk_family __sk_common.skc_family |
322 | #define sk_state __sk_common.skc_state | |
323 | #define sk_reuse __sk_common.skc_reuse | |
055dc21a | 324 | #define sk_reuseport __sk_common.skc_reuseport |
9fe516ba | 325 | #define sk_ipv6only __sk_common.skc_ipv6only |
1da177e4 | 326 | #define sk_bound_dev_if __sk_common.skc_bound_dev_if |
1da177e4 | 327 | #define sk_bind_node __sk_common.skc_bind_node |
8feaf0c0 | 328 | #define sk_prot __sk_common.skc_prot |
07feaebf | 329 | #define sk_net __sk_common.skc_net |
efe4208f ED |
330 | #define sk_v6_daddr __sk_common.skc_v6_daddr |
331 | #define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr | |
332 | ||
1da177e4 | 333 | socket_lock_t sk_lock; |
b178bb3d | 334 | struct sk_buff_head sk_receive_queue; |
fa438ccf ED |
335 | /* |
336 | * The backlog queue is special, it is always used with | |
337 | * the per-socket spinlock held and requires low latency | |
338 | * access. Therefore we special case it's implementation. | |
b178bb3d ED |
339 | * Note : rmem_alloc is in this structure to fill a hole |
340 | * on 64bit arches, not because its logically part of | |
341 | * backlog. | |
fa438ccf ED |
342 | */ |
343 | struct { | |
b178bb3d ED |
344 | atomic_t rmem_alloc; |
345 | int len; | |
346 | struct sk_buff *head; | |
347 | struct sk_buff *tail; | |
fa438ccf | 348 | } sk_backlog; |
b178bb3d ED |
349 | #define sk_rmem_alloc sk_backlog.rmem_alloc |
350 | int sk_forward_alloc; | |
351 | #ifdef CONFIG_RPS | |
352 | __u32 sk_rxhash; | |
06021292 | 353 | #endif |
2c8c56e1 ED |
354 | u16 sk_incoming_cpu; |
355 | /* 16bit hole | |
356 | * Warned : sk_incoming_cpu can be set from softirq, | |
357 | * Do not use this hole without fully understanding possible issues. | |
358 | */ | |
359 | ||
b73c3d0e | 360 | __u32 sk_txhash; |
e0d1095a | 361 | #ifdef CONFIG_NET_RX_BUSY_POLL |
06021292 | 362 | unsigned int sk_napi_id; |
dafcc438 | 363 | unsigned int sk_ll_usec; |
b178bb3d ED |
364 | #endif |
365 | atomic_t sk_drops; | |
366 | int sk_rcvbuf; | |
367 | ||
368 | struct sk_filter __rcu *sk_filter; | |
eaefd110 | 369 | struct socket_wq __rcu *sk_wq; |
b178bb3d | 370 | |
def8b4fa | 371 | #ifdef CONFIG_XFRM |
1da177e4 | 372 | struct xfrm_policy *sk_policy[2]; |
def8b4fa | 373 | #endif |
b178bb3d | 374 | unsigned long sk_flags; |
deaa5854 | 375 | struct dst_entry *sk_rx_dst; |
0e36cbb3 | 376 | struct dst_entry __rcu *sk_dst_cache; |
b6c6712a | 377 | spinlock_t sk_dst_lock; |
1da177e4 LT |
378 | atomic_t sk_wmem_alloc; |
379 | atomic_t sk_omem_alloc; | |
4e07a91c | 380 | int sk_sndbuf; |
1da177e4 | 381 | struct sk_buff_head sk_write_queue; |
b178bb3d ED |
382 | kmemcheck_bitfield_begin(flags); |
383 | unsigned int sk_shutdown : 2, | |
28448b80 TH |
384 | sk_no_check_tx : 1, |
385 | sk_no_check_rx : 1, | |
b178bb3d ED |
386 | sk_userlocks : 4, |
387 | sk_protocol : 8, | |
388 | sk_type : 16; | |
389 | kmemcheck_bitfield_end(flags); | |
1da177e4 | 390 | int sk_wmem_queued; |
7d877f3b | 391 | gfp_t sk_allocation; |
95bd09eb | 392 | u32 sk_pacing_rate; /* bytes per second */ |
62748f32 | 393 | u32 sk_max_pacing_rate; |
c8f44aff MM |
394 | netdev_features_t sk_route_caps; |
395 | netdev_features_t sk_route_nocaps; | |
bcd76111 | 396 | int sk_gso_type; |
82cc1a7a | 397 | unsigned int sk_gso_max_size; |
1485348d | 398 | u16 sk_gso_max_segs; |
9932cf95 | 399 | int sk_rcvlowat; |
1da177e4 | 400 | unsigned long sk_lingertime; |
1da177e4 | 401 | struct sk_buff_head sk_error_queue; |
476e19cf | 402 | struct proto *sk_prot_creator; |
1da177e4 LT |
403 | rwlock_t sk_callback_lock; |
404 | int sk_err, | |
405 | sk_err_soft; | |
406 | unsigned short sk_ack_backlog; | |
407 | unsigned short sk_max_ack_backlog; | |
408 | __u32 sk_priority; | |
86f8515f | 409 | #if IS_ENABLED(CONFIG_CGROUP_NET_PRIO) |
5bc1421e NH |
410 | __u32 sk_cgrp_prioidx; |
411 | #endif | |
109f6e39 EB |
412 | struct pid *sk_peer_pid; |
413 | const struct cred *sk_peer_cred; | |
1da177e4 LT |
414 | long sk_rcvtimeo; |
415 | long sk_sndtimeo; | |
1da177e4 LT |
416 | void *sk_protinfo; |
417 | struct timer_list sk_timer; | |
b7aa0bf7 | 418 | ktime_t sk_stamp; |
b9f40e21 | 419 | u16 sk_tsflags; |
09c2d251 | 420 | u32 sk_tskey; |
1da177e4 LT |
421 | struct socket *sk_socket; |
422 | void *sk_user_data; | |
5640f768 | 423 | struct page_frag sk_frag; |
1da177e4 | 424 | struct sk_buff *sk_send_head; |
ef64a54f | 425 | __s32 sk_peek_off; |
1da177e4 | 426 | int sk_write_pending; |
d5f64238 | 427 | #ifdef CONFIG_SECURITY |
1da177e4 | 428 | void *sk_security; |
d5f64238 | 429 | #endif |
4a19ec58 | 430 | __u32 sk_mark; |
f8451725 | 431 | u32 sk_classid; |
e1aab161 | 432 | struct cg_proto *sk_cgrp; |
1da177e4 | 433 | void (*sk_state_change)(struct sock *sk); |
676d2369 | 434 | void (*sk_data_ready)(struct sock *sk); |
1da177e4 LT |
435 | void (*sk_write_space)(struct sock *sk); |
436 | void (*sk_error_report)(struct sock *sk); | |
dc6b9b78 ED |
437 | int (*sk_backlog_rcv)(struct sock *sk, |
438 | struct sk_buff *skb); | |
1da177e4 LT |
439 | void (*sk_destruct)(struct sock *sk); |
440 | }; | |
441 | ||
559835ea PS |
442 | #define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data))) |
443 | ||
444 | #define rcu_dereference_sk_user_data(sk) rcu_dereference(__sk_user_data((sk))) | |
445 | #define rcu_assign_sk_user_data(sk, ptr) rcu_assign_pointer(__sk_user_data((sk)), ptr) | |
446 | ||
4a17fd52 PE |
447 | /* |
448 | * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK | |
449 | * or not whether his port will be reused by someone else. SK_FORCE_REUSE | |
450 | * on a socket means that the socket will reuse everybody else's port | |
451 | * without looking at the other's sk_reuse value. | |
452 | */ | |
453 | ||
454 | #define SK_NO_REUSE 0 | |
455 | #define SK_CAN_REUSE 1 | |
456 | #define SK_FORCE_REUSE 2 | |
457 | ||
ef64a54f PE |
458 | static inline int sk_peek_offset(struct sock *sk, int flags) |
459 | { | |
460 | if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0)) | |
461 | return sk->sk_peek_off; | |
462 | else | |
463 | return 0; | |
464 | } | |
465 | ||
466 | static inline void sk_peek_offset_bwd(struct sock *sk, int val) | |
467 | { | |
468 | if (sk->sk_peek_off >= 0) { | |
469 | if (sk->sk_peek_off >= val) | |
470 | sk->sk_peek_off -= val; | |
471 | else | |
472 | sk->sk_peek_off = 0; | |
473 | } | |
474 | } | |
475 | ||
476 | static inline void sk_peek_offset_fwd(struct sock *sk, int val) | |
477 | { | |
478 | if (sk->sk_peek_off >= 0) | |
479 | sk->sk_peek_off += val; | |
480 | } | |
481 | ||
1da177e4 LT |
482 | /* |
483 | * Hashed lists helper routines | |
484 | */ | |
c4146644 LZ |
485 | static inline struct sock *sk_entry(const struct hlist_node *node) |
486 | { | |
487 | return hlist_entry(node, struct sock, sk_node); | |
488 | } | |
489 | ||
e48c414e | 490 | static inline struct sock *__sk_head(const struct hlist_head *head) |
1da177e4 LT |
491 | { |
492 | return hlist_entry(head->first, struct sock, sk_node); | |
493 | } | |
494 | ||
e48c414e | 495 | static inline struct sock *sk_head(const struct hlist_head *head) |
1da177e4 LT |
496 | { |
497 | return hlist_empty(head) ? NULL : __sk_head(head); | |
498 | } | |
499 | ||
88ab1932 ED |
500 | static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head) |
501 | { | |
502 | return hlist_nulls_entry(head->first, struct sock, sk_nulls_node); | |
503 | } | |
504 | ||
505 | static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head) | |
506 | { | |
507 | return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head); | |
508 | } | |
509 | ||
e48c414e | 510 | static inline struct sock *sk_next(const struct sock *sk) |
1da177e4 LT |
511 | { |
512 | return sk->sk_node.next ? | |
513 | hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL; | |
514 | } | |
515 | ||
88ab1932 ED |
516 | static inline struct sock *sk_nulls_next(const struct sock *sk) |
517 | { | |
518 | return (!is_a_nulls(sk->sk_nulls_node.next)) ? | |
519 | hlist_nulls_entry(sk->sk_nulls_node.next, | |
520 | struct sock, sk_nulls_node) : | |
521 | NULL; | |
522 | } | |
523 | ||
dc6b9b78 | 524 | static inline bool sk_unhashed(const struct sock *sk) |
1da177e4 LT |
525 | { |
526 | return hlist_unhashed(&sk->sk_node); | |
527 | } | |
528 | ||
dc6b9b78 | 529 | static inline bool sk_hashed(const struct sock *sk) |
1da177e4 | 530 | { |
da753bea | 531 | return !sk_unhashed(sk); |
1da177e4 LT |
532 | } |
533 | ||
dc6b9b78 | 534 | static inline void sk_node_init(struct hlist_node *node) |
1da177e4 LT |
535 | { |
536 | node->pprev = NULL; | |
537 | } | |
538 | ||
dc6b9b78 | 539 | static inline void sk_nulls_node_init(struct hlist_nulls_node *node) |
88ab1932 ED |
540 | { |
541 | node->pprev = NULL; | |
542 | } | |
543 | ||
dc6b9b78 | 544 | static inline void __sk_del_node(struct sock *sk) |
1da177e4 LT |
545 | { |
546 | __hlist_del(&sk->sk_node); | |
547 | } | |
548 | ||
808f5114 | 549 | /* NB: equivalent to hlist_del_init_rcu */ |
dc6b9b78 | 550 | static inline bool __sk_del_node_init(struct sock *sk) |
1da177e4 LT |
551 | { |
552 | if (sk_hashed(sk)) { | |
553 | __sk_del_node(sk); | |
554 | sk_node_init(&sk->sk_node); | |
dc6b9b78 | 555 | return true; |
1da177e4 | 556 | } |
dc6b9b78 | 557 | return false; |
1da177e4 LT |
558 | } |
559 | ||
560 | /* Grab socket reference count. This operation is valid only | |
561 | when sk is ALREADY grabbed f.e. it is found in hash table | |
562 | or a list and the lookup is made under lock preventing hash table | |
563 | modifications. | |
564 | */ | |
565 | ||
566 | static inline void sock_hold(struct sock *sk) | |
567 | { | |
568 | atomic_inc(&sk->sk_refcnt); | |
569 | } | |
570 | ||
571 | /* Ungrab socket in the context, which assumes that socket refcnt | |
572 | cannot hit zero, f.e. it is true in context of any socketcall. | |
573 | */ | |
574 | static inline void __sock_put(struct sock *sk) | |
575 | { | |
576 | atomic_dec(&sk->sk_refcnt); | |
577 | } | |
578 | ||
dc6b9b78 | 579 | static inline bool sk_del_node_init(struct sock *sk) |
1da177e4 | 580 | { |
dc6b9b78 | 581 | bool rc = __sk_del_node_init(sk); |
1da177e4 LT |
582 | |
583 | if (rc) { | |
584 | /* paranoid for a while -acme */ | |
585 | WARN_ON(atomic_read(&sk->sk_refcnt) == 1); | |
586 | __sock_put(sk); | |
587 | } | |
588 | return rc; | |
589 | } | |
808f5114 | 590 | #define sk_del_node_init_rcu(sk) sk_del_node_init(sk) |
1da177e4 | 591 | |
dc6b9b78 | 592 | static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk) |
271b72c7 ED |
593 | { |
594 | if (sk_hashed(sk)) { | |
88ab1932 | 595 | hlist_nulls_del_init_rcu(&sk->sk_nulls_node); |
dc6b9b78 | 596 | return true; |
271b72c7 | 597 | } |
dc6b9b78 | 598 | return false; |
271b72c7 ED |
599 | } |
600 | ||
dc6b9b78 | 601 | static inline bool sk_nulls_del_node_init_rcu(struct sock *sk) |
271b72c7 | 602 | { |
dc6b9b78 | 603 | bool rc = __sk_nulls_del_node_init_rcu(sk); |
271b72c7 ED |
604 | |
605 | if (rc) { | |
606 | /* paranoid for a while -acme */ | |
607 | WARN_ON(atomic_read(&sk->sk_refcnt) == 1); | |
608 | __sock_put(sk); | |
609 | } | |
610 | return rc; | |
611 | } | |
612 | ||
dc6b9b78 | 613 | static inline void __sk_add_node(struct sock *sk, struct hlist_head *list) |
1da177e4 LT |
614 | { |
615 | hlist_add_head(&sk->sk_node, list); | |
616 | } | |
617 | ||
dc6b9b78 | 618 | static inline void sk_add_node(struct sock *sk, struct hlist_head *list) |
1da177e4 LT |
619 | { |
620 | sock_hold(sk); | |
621 | __sk_add_node(sk, list); | |
622 | } | |
623 | ||
dc6b9b78 | 624 | static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list) |
808f5114 | 625 | { |
626 | sock_hold(sk); | |
627 | hlist_add_head_rcu(&sk->sk_node, list); | |
628 | } | |
629 | ||
dc6b9b78 | 630 | static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list) |
271b72c7 | 631 | { |
88ab1932 | 632 | hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list); |
271b72c7 ED |
633 | } |
634 | ||
dc6b9b78 | 635 | static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list) |
271b72c7 ED |
636 | { |
637 | sock_hold(sk); | |
88ab1932 | 638 | __sk_nulls_add_node_rcu(sk, list); |
271b72c7 ED |
639 | } |
640 | ||
dc6b9b78 | 641 | static inline void __sk_del_bind_node(struct sock *sk) |
1da177e4 LT |
642 | { |
643 | __hlist_del(&sk->sk_bind_node); | |
644 | } | |
645 | ||
dc6b9b78 | 646 | static inline void sk_add_bind_node(struct sock *sk, |
1da177e4 LT |
647 | struct hlist_head *list) |
648 | { | |
649 | hlist_add_head(&sk->sk_bind_node, list); | |
650 | } | |
651 | ||
b67bfe0d SL |
652 | #define sk_for_each(__sk, list) \ |
653 | hlist_for_each_entry(__sk, list, sk_node) | |
654 | #define sk_for_each_rcu(__sk, list) \ | |
655 | hlist_for_each_entry_rcu(__sk, list, sk_node) | |
88ab1932 ED |
656 | #define sk_nulls_for_each(__sk, node, list) \ |
657 | hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node) | |
658 | #define sk_nulls_for_each_rcu(__sk, node, list) \ | |
659 | hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node) | |
b67bfe0d SL |
660 | #define sk_for_each_from(__sk) \ |
661 | hlist_for_each_entry_from(__sk, sk_node) | |
88ab1932 ED |
662 | #define sk_nulls_for_each_from(__sk, node) \ |
663 | if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \ | |
664 | hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node) | |
b67bfe0d SL |
665 | #define sk_for_each_safe(__sk, tmp, list) \ |
666 | hlist_for_each_entry_safe(__sk, tmp, list, sk_node) | |
667 | #define sk_for_each_bound(__sk, list) \ | |
668 | hlist_for_each_entry(__sk, list, sk_bind_node) | |
1da177e4 | 669 | |
2dc41cff DH |
670 | /** |
671 | * sk_nulls_for_each_entry_offset - iterate over a list at a given struct offset | |
672 | * @tpos: the type * to use as a loop cursor. | |
673 | * @pos: the &struct hlist_node to use as a loop cursor. | |
674 | * @head: the head for your list. | |
675 | * @offset: offset of hlist_node within the struct. | |
676 | * | |
677 | */ | |
678 | #define sk_nulls_for_each_entry_offset(tpos, pos, head, offset) \ | |
679 | for (pos = (head)->first; \ | |
680 | (!is_a_nulls(pos)) && \ | |
681 | ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;}); \ | |
682 | pos = pos->next) | |
683 | ||
c336d148 EB |
684 | static inline struct user_namespace *sk_user_ns(struct sock *sk) |
685 | { | |
686 | /* Careful only use this in a context where these parameters | |
687 | * can not change and must all be valid, such as recvmsg from | |
688 | * userspace. | |
689 | */ | |
690 | return sk->sk_socket->file->f_cred->user_ns; | |
691 | } | |
692 | ||
1da177e4 LT |
693 | /* Sock flags */ |
694 | enum sock_flags { | |
695 | SOCK_DEAD, | |
696 | SOCK_DONE, | |
697 | SOCK_URGINLINE, | |
698 | SOCK_KEEPOPEN, | |
699 | SOCK_LINGER, | |
700 | SOCK_DESTROY, | |
701 | SOCK_BROADCAST, | |
702 | SOCK_TIMESTAMP, | |
703 | SOCK_ZAPPED, | |
704 | SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */ | |
705 | SOCK_DBG, /* %SO_DEBUG setting */ | |
706 | SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */ | |
92f37fd2 | 707 | SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */ |
1da177e4 LT |
708 | SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */ |
709 | SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */ | |
7cb02404 | 710 | SOCK_MEMALLOC, /* VM depends on this socket for swapping */ |
20d49473 | 711 | SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */ |
bcdce719 | 712 | SOCK_FASYNC, /* fasync() active */ |
3b885787 | 713 | SOCK_RXQ_OVFL, |
1cdebb42 | 714 | SOCK_ZEROCOPY, /* buffers from userspace */ |
6e3e939f | 715 | SOCK_WIFI_STATUS, /* push wifi status to userspace */ |
3bdc0eba BG |
716 | SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS. |
717 | * Will use last 4 bytes of packet sent from | |
718 | * user-space instead. | |
719 | */ | |
d59577b6 | 720 | SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */ |
7d4c04fc | 721 | SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */ |
1da177e4 LT |
722 | }; |
723 | ||
53b924b3 RB |
724 | static inline void sock_copy_flags(struct sock *nsk, struct sock *osk) |
725 | { | |
726 | nsk->sk_flags = osk->sk_flags; | |
727 | } | |
728 | ||
1da177e4 LT |
729 | static inline void sock_set_flag(struct sock *sk, enum sock_flags flag) |
730 | { | |
731 | __set_bit(flag, &sk->sk_flags); | |
732 | } | |
733 | ||
734 | static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag) | |
735 | { | |
736 | __clear_bit(flag, &sk->sk_flags); | |
737 | } | |
738 | ||
1b23a5df | 739 | static inline bool sock_flag(const struct sock *sk, enum sock_flags flag) |
1da177e4 LT |
740 | { |
741 | return test_bit(flag, &sk->sk_flags); | |
742 | } | |
743 | ||
c93bdd0e MG |
744 | #ifdef CONFIG_NET |
745 | extern struct static_key memalloc_socks; | |
746 | static inline int sk_memalloc_socks(void) | |
747 | { | |
748 | return static_key_false(&memalloc_socks); | |
749 | } | |
750 | #else | |
751 | ||
752 | static inline int sk_memalloc_socks(void) | |
753 | { | |
754 | return 0; | |
755 | } | |
756 | ||
757 | #endif | |
758 | ||
99a1dec7 MG |
759 | static inline gfp_t sk_gfp_atomic(struct sock *sk, gfp_t gfp_mask) |
760 | { | |
7cb02404 | 761 | return GFP_ATOMIC | (sk->sk_allocation & __GFP_MEMALLOC); |
99a1dec7 MG |
762 | } |
763 | ||
1da177e4 LT |
764 | static inline void sk_acceptq_removed(struct sock *sk) |
765 | { | |
766 | sk->sk_ack_backlog--; | |
767 | } | |
768 | ||
769 | static inline void sk_acceptq_added(struct sock *sk) | |
770 | { | |
771 | sk->sk_ack_backlog++; | |
772 | } | |
773 | ||
dc6b9b78 | 774 | static inline bool sk_acceptq_is_full(const struct sock *sk) |
1da177e4 | 775 | { |
64a14651 | 776 | return sk->sk_ack_backlog > sk->sk_max_ack_backlog; |
1da177e4 LT |
777 | } |
778 | ||
779 | /* | |
780 | * Compute minimal free write space needed to queue new packets. | |
781 | */ | |
dc6b9b78 | 782 | static inline int sk_stream_min_wspace(const struct sock *sk) |
1da177e4 | 783 | { |
8df09ea3 | 784 | return sk->sk_wmem_queued >> 1; |
1da177e4 LT |
785 | } |
786 | ||
dc6b9b78 | 787 | static inline int sk_stream_wspace(const struct sock *sk) |
1da177e4 LT |
788 | { |
789 | return sk->sk_sndbuf - sk->sk_wmem_queued; | |
790 | } | |
791 | ||
69336bd2 | 792 | void sk_stream_write_space(struct sock *sk); |
1da177e4 | 793 | |
8eae939f | 794 | /* OOB backlog add */ |
a3a858ff | 795 | static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb) |
9ee6b535 | 796 | { |
7fee226a ED |
797 | /* dont let skb dst not refcounted, we are going to leave rcu lock */ |
798 | skb_dst_force(skb); | |
799 | ||
800 | if (!sk->sk_backlog.tail) | |
801 | sk->sk_backlog.head = skb; | |
802 | else | |
9ee6b535 | 803 | sk->sk_backlog.tail->next = skb; |
7fee226a ED |
804 | |
805 | sk->sk_backlog.tail = skb; | |
9ee6b535 SH |
806 | skb->next = NULL; |
807 | } | |
1da177e4 | 808 | |
c377411f ED |
809 | /* |
810 | * Take into account size of receive queue and backlog queue | |
0fd7bac6 ED |
811 | * Do not take into account this skb truesize, |
812 | * to allow even a single big packet to come. | |
c377411f | 813 | */ |
274f482d | 814 | static inline bool sk_rcvqueues_full(const struct sock *sk, unsigned int limit) |
c377411f ED |
815 | { |
816 | unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc); | |
817 | ||
f545a38f | 818 | return qsize > limit; |
c377411f ED |
819 | } |
820 | ||
8eae939f | 821 | /* The per-socket spinlock must be held here. */ |
f545a38f ED |
822 | static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb, |
823 | unsigned int limit) | |
8eae939f | 824 | { |
274f482d | 825 | if (sk_rcvqueues_full(sk, limit)) |
8eae939f ZY |
826 | return -ENOBUFS; |
827 | ||
a3a858ff | 828 | __sk_add_backlog(sk, skb); |
8eae939f ZY |
829 | sk->sk_backlog.len += skb->truesize; |
830 | return 0; | |
831 | } | |
832 | ||
69336bd2 | 833 | int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb); |
b4b9e355 | 834 | |
c57943a1 PZ |
835 | static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb) |
836 | { | |
b4b9e355 MG |
837 | if (sk_memalloc_socks() && skb_pfmemalloc(skb)) |
838 | return __sk_backlog_rcv(sk, skb); | |
839 | ||
c57943a1 PZ |
840 | return sk->sk_backlog_rcv(sk, skb); |
841 | } | |
842 | ||
2c8c56e1 ED |
843 | static inline void sk_incoming_cpu_update(struct sock *sk) |
844 | { | |
845 | sk->sk_incoming_cpu = raw_smp_processor_id(); | |
846 | } | |
847 | ||
fe477558 | 848 | static inline void sock_rps_record_flow_hash(__u32 hash) |
c58dc01b DM |
849 | { |
850 | #ifdef CONFIG_RPS | |
851 | struct rps_sock_flow_table *sock_flow_table; | |
852 | ||
853 | rcu_read_lock(); | |
854 | sock_flow_table = rcu_dereference(rps_sock_flow_table); | |
fe477558 | 855 | rps_record_sock_flow(sock_flow_table, hash); |
c58dc01b DM |
856 | rcu_read_unlock(); |
857 | #endif | |
858 | } | |
859 | ||
fe477558 | 860 | static inline void sock_rps_reset_flow_hash(__u32 hash) |
c58dc01b DM |
861 | { |
862 | #ifdef CONFIG_RPS | |
863 | struct rps_sock_flow_table *sock_flow_table; | |
864 | ||
865 | rcu_read_lock(); | |
866 | sock_flow_table = rcu_dereference(rps_sock_flow_table); | |
fe477558 | 867 | rps_reset_sock_flow(sock_flow_table, hash); |
c58dc01b DM |
868 | rcu_read_unlock(); |
869 | #endif | |
870 | } | |
871 | ||
fe477558 TH |
872 | static inline void sock_rps_record_flow(const struct sock *sk) |
873 | { | |
c9d8ca04 | 874 | #ifdef CONFIG_RPS |
fe477558 | 875 | sock_rps_record_flow_hash(sk->sk_rxhash); |
c9d8ca04 | 876 | #endif |
fe477558 TH |
877 | } |
878 | ||
879 | static inline void sock_rps_reset_flow(const struct sock *sk) | |
880 | { | |
c9d8ca04 | 881 | #ifdef CONFIG_RPS |
fe477558 | 882 | sock_rps_reset_flow_hash(sk->sk_rxhash); |
c9d8ca04 | 883 | #endif |
fe477558 TH |
884 | } |
885 | ||
bdeab991 TH |
886 | static inline void sock_rps_save_rxhash(struct sock *sk, |
887 | const struct sk_buff *skb) | |
c58dc01b DM |
888 | { |
889 | #ifdef CONFIG_RPS | |
61b905da | 890 | if (unlikely(sk->sk_rxhash != skb->hash)) { |
c58dc01b | 891 | sock_rps_reset_flow(sk); |
61b905da | 892 | sk->sk_rxhash = skb->hash; |
c58dc01b DM |
893 | } |
894 | #endif | |
895 | } | |
896 | ||
bdeab991 TH |
897 | static inline void sock_rps_reset_rxhash(struct sock *sk) |
898 | { | |
899 | #ifdef CONFIG_RPS | |
900 | sock_rps_reset_flow(sk); | |
901 | sk->sk_rxhash = 0; | |
902 | #endif | |
903 | } | |
904 | ||
cfcabdcc SH |
905 | #define sk_wait_event(__sk, __timeo, __condition) \ |
906 | ({ int __rc; \ | |
907 | release_sock(__sk); \ | |
908 | __rc = __condition; \ | |
909 | if (!__rc) { \ | |
910 | *(__timeo) = schedule_timeout(*(__timeo)); \ | |
911 | } \ | |
26cabd31 | 912 | sched_annotate_sleep(); \ |
cfcabdcc SH |
913 | lock_sock(__sk); \ |
914 | __rc = __condition; \ | |
915 | __rc; \ | |
916 | }) | |
1da177e4 | 917 | |
69336bd2 JP |
918 | int sk_stream_wait_connect(struct sock *sk, long *timeo_p); |
919 | int sk_stream_wait_memory(struct sock *sk, long *timeo_p); | |
920 | void sk_stream_wait_close(struct sock *sk, long timeo_p); | |
921 | int sk_stream_error(struct sock *sk, int flags, int err); | |
922 | void sk_stream_kill_queues(struct sock *sk); | |
923 | void sk_set_memalloc(struct sock *sk); | |
924 | void sk_clear_memalloc(struct sock *sk); | |
1da177e4 | 925 | |
69336bd2 | 926 | int sk_wait_data(struct sock *sk, long *timeo); |
1da177e4 | 927 | |
60236fdd | 928 | struct request_sock_ops; |
6d6ee43e | 929 | struct timewait_sock_ops; |
ab1e0a13 | 930 | struct inet_hashinfo; |
fc8717ba | 931 | struct raw_hashinfo; |
de477254 | 932 | struct module; |
2e6599cb | 933 | |
f77d6021 ED |
934 | /* |
935 | * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes | |
936 | * un-modified. Special care is taken when initializing object to zero. | |
937 | */ | |
938 | static inline void sk_prot_clear_nulls(struct sock *sk, int size) | |
939 | { | |
940 | if (offsetof(struct sock, sk_node.next) != 0) | |
941 | memset(sk, 0, offsetof(struct sock, sk_node.next)); | |
942 | memset(&sk->sk_node.pprev, 0, | |
943 | size - offsetof(struct sock, sk_node.pprev)); | |
944 | } | |
945 | ||
1da177e4 LT |
946 | /* Networking protocol blocks we attach to sockets. |
947 | * socket layer -> transport layer interface | |
948 | * transport -> network interface is defined by struct inet_proto | |
949 | */ | |
950 | struct proto { | |
dc6b9b78 | 951 | void (*close)(struct sock *sk, |
1da177e4 LT |
952 | long timeout); |
953 | int (*connect)(struct sock *sk, | |
dc6b9b78 | 954 | struct sockaddr *uaddr, |
1da177e4 LT |
955 | int addr_len); |
956 | int (*disconnect)(struct sock *sk, int flags); | |
957 | ||
dc6b9b78 | 958 | struct sock * (*accept)(struct sock *sk, int flags, int *err); |
1da177e4 LT |
959 | |
960 | int (*ioctl)(struct sock *sk, int cmd, | |
961 | unsigned long arg); | |
962 | int (*init)(struct sock *sk); | |
7d06b2e0 | 963 | void (*destroy)(struct sock *sk); |
1da177e4 | 964 | void (*shutdown)(struct sock *sk, int how); |
dc6b9b78 | 965 | int (*setsockopt)(struct sock *sk, int level, |
1da177e4 | 966 | int optname, char __user *optval, |
b7058842 | 967 | unsigned int optlen); |
dc6b9b78 ED |
968 | int (*getsockopt)(struct sock *sk, int level, |
969 | int optname, char __user *optval, | |
970 | int __user *option); | |
af01d537 | 971 | #ifdef CONFIG_COMPAT |
3fdadf7d DM |
972 | int (*compat_setsockopt)(struct sock *sk, |
973 | int level, | |
974 | int optname, char __user *optval, | |
b7058842 | 975 | unsigned int optlen); |
3fdadf7d DM |
976 | int (*compat_getsockopt)(struct sock *sk, |
977 | int level, | |
978 | int optname, char __user *optval, | |
979 | int __user *option); | |
709b46e8 EB |
980 | int (*compat_ioctl)(struct sock *sk, |
981 | unsigned int cmd, unsigned long arg); | |
af01d537 | 982 | #endif |
1da177e4 LT |
983 | int (*sendmsg)(struct kiocb *iocb, struct sock *sk, |
984 | struct msghdr *msg, size_t len); | |
985 | int (*recvmsg)(struct kiocb *iocb, struct sock *sk, | |
986 | struct msghdr *msg, | |
dc6b9b78 ED |
987 | size_t len, int noblock, int flags, |
988 | int *addr_len); | |
1da177e4 LT |
989 | int (*sendpage)(struct sock *sk, struct page *page, |
990 | int offset, size_t size, int flags); | |
dc6b9b78 | 991 | int (*bind)(struct sock *sk, |
1da177e4 LT |
992 | struct sockaddr *uaddr, int addr_len); |
993 | ||
dc6b9b78 | 994 | int (*backlog_rcv) (struct sock *sk, |
1da177e4 LT |
995 | struct sk_buff *skb); |
996 | ||
46d3ceab ED |
997 | void (*release_cb)(struct sock *sk); |
998 | ||
1da177e4 LT |
999 | /* Keeping track of sk's, looking them up, and port selection methods. */ |
1000 | void (*hash)(struct sock *sk); | |
1001 | void (*unhash)(struct sock *sk); | |
719f8358 | 1002 | void (*rehash)(struct sock *sk); |
1da177e4 | 1003 | int (*get_port)(struct sock *sk, unsigned short snum); |
fcbdf09d | 1004 | void (*clear_sk)(struct sock *sk, int size); |
1da177e4 | 1005 | |
286ab3d4 | 1006 | /* Keeping track of sockets in use */ |
65f76517 | 1007 | #ifdef CONFIG_PROC_FS |
13ff3d6f | 1008 | unsigned int inuse_idx; |
65f76517 | 1009 | #endif |
ebb53d75 | 1010 | |
c9bee3b7 | 1011 | bool (*stream_memory_free)(const struct sock *sk); |
1da177e4 | 1012 | /* Memory pressure */ |
5c52ba17 | 1013 | void (*enter_memory_pressure)(struct sock *sk); |
8d987e5c | 1014 | atomic_long_t *memory_allocated; /* Current allocated memory. */ |
1748376b | 1015 | struct percpu_counter *sockets_allocated; /* Current number of sockets. */ |
1da177e4 LT |
1016 | /* |
1017 | * Pressure flag: try to collapse. | |
1018 | * Technical note: it is used by multiple contexts non atomically. | |
3ab224be | 1019 | * All the __sk_mem_schedule() is of this nature: accounting |
1da177e4 LT |
1020 | * is strict, actions are advisory and have some latency. |
1021 | */ | |
1022 | int *memory_pressure; | |
8d987e5c | 1023 | long *sysctl_mem; |
1da177e4 LT |
1024 | int *sysctl_wmem; |
1025 | int *sysctl_rmem; | |
1026 | int max_header; | |
7ba42910 | 1027 | bool no_autobind; |
1da177e4 | 1028 | |
271b72c7 | 1029 | struct kmem_cache *slab; |
1da177e4 | 1030 | unsigned int obj_size; |
271b72c7 | 1031 | int slab_flags; |
1da177e4 | 1032 | |
dd24c001 | 1033 | struct percpu_counter *orphan_count; |
8feaf0c0 | 1034 | |
60236fdd | 1035 | struct request_sock_ops *rsk_prot; |
6d6ee43e | 1036 | struct timewait_sock_ops *twsk_prot; |
2e6599cb | 1037 | |
39d8cda7 PE |
1038 | union { |
1039 | struct inet_hashinfo *hashinfo; | |
645ca708 | 1040 | struct udp_table *udp_table; |
fc8717ba | 1041 | struct raw_hashinfo *raw_hash; |
39d8cda7 | 1042 | } h; |
ab1e0a13 | 1043 | |
1da177e4 LT |
1044 | struct module *owner; |
1045 | ||
1046 | char name[32]; | |
1047 | ||
1048 | struct list_head node; | |
e6848976 ACM |
1049 | #ifdef SOCK_REFCNT_DEBUG |
1050 | atomic_t socks; | |
1051 | #endif | |
c255a458 | 1052 | #ifdef CONFIG_MEMCG_KMEM |
e1aab161 GC |
1053 | /* |
1054 | * cgroup specific init/deinit functions. Called once for all | |
1055 | * protocols that implement it, from cgroups populate function. | |
1056 | * This function has to setup any files the protocol want to | |
1057 | * appear in the kmem cgroup filesystem. | |
1058 | */ | |
1d62e436 | 1059 | int (*init_cgroup)(struct mem_cgroup *memcg, |
e1aab161 | 1060 | struct cgroup_subsys *ss); |
1d62e436 | 1061 | void (*destroy_cgroup)(struct mem_cgroup *memcg); |
e1aab161 GC |
1062 | struct cg_proto *(*proto_cgroup)(struct mem_cgroup *memcg); |
1063 | #endif | |
1064 | }; | |
1065 | ||
3f134619 GC |
1066 | /* |
1067 | * Bits in struct cg_proto.flags | |
1068 | */ | |
1069 | enum cg_proto_flags { | |
1070 | /* Currently active and new sockets should be assigned to cgroups */ | |
1071 | MEMCG_SOCK_ACTIVE, | |
1072 | /* It was ever activated; we must disarm static keys on destruction */ | |
1073 | MEMCG_SOCK_ACTIVATED, | |
1074 | }; | |
1075 | ||
e1aab161 | 1076 | struct cg_proto { |
3e32cb2e | 1077 | struct page_counter memory_allocated; /* Current allocated memory. */ |
2e685cad EB |
1078 | struct percpu_counter sockets_allocated; /* Current number of sockets. */ |
1079 | int memory_pressure; | |
1080 | long sysctl_mem[3]; | |
3f134619 | 1081 | unsigned long flags; |
e1aab161 GC |
1082 | /* |
1083 | * memcg field is used to find which memcg we belong directly | |
1084 | * Each memcg struct can hold more than one cg_proto, so container_of | |
1085 | * won't really cut. | |
1086 | * | |
1087 | * The elegant solution would be having an inverse function to | |
1088 | * proto_cgroup in struct proto, but that means polluting the structure | |
1089 | * for everybody, instead of just for memcg users. | |
1090 | */ | |
1091 | struct mem_cgroup *memcg; | |
1da177e4 LT |
1092 | }; |
1093 | ||
69336bd2 JP |
1094 | int proto_register(struct proto *prot, int alloc_slab); |
1095 | void proto_unregister(struct proto *prot); | |
1da177e4 | 1096 | |
3f134619 GC |
1097 | static inline bool memcg_proto_active(struct cg_proto *cg_proto) |
1098 | { | |
1099 | return test_bit(MEMCG_SOCK_ACTIVE, &cg_proto->flags); | |
1100 | } | |
1101 | ||
1102 | static inline bool memcg_proto_activated(struct cg_proto *cg_proto) | |
1103 | { | |
1104 | return test_bit(MEMCG_SOCK_ACTIVATED, &cg_proto->flags); | |
1105 | } | |
1106 | ||
e6848976 ACM |
1107 | #ifdef SOCK_REFCNT_DEBUG |
1108 | static inline void sk_refcnt_debug_inc(struct sock *sk) | |
1109 | { | |
1110 | atomic_inc(&sk->sk_prot->socks); | |
1111 | } | |
1112 | ||
1113 | static inline void sk_refcnt_debug_dec(struct sock *sk) | |
1114 | { | |
1115 | atomic_dec(&sk->sk_prot->socks); | |
1116 | printk(KERN_DEBUG "%s socket %p released, %d are still alive\n", | |
1117 | sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks)); | |
1118 | } | |
1119 | ||
dec34fb0 | 1120 | static inline void sk_refcnt_debug_release(const struct sock *sk) |
e6848976 ACM |
1121 | { |
1122 | if (atomic_read(&sk->sk_refcnt) != 1) | |
1123 | printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n", | |
1124 | sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt)); | |
1125 | } | |
1126 | #else /* SOCK_REFCNT_DEBUG */ | |
1127 | #define sk_refcnt_debug_inc(sk) do { } while (0) | |
1128 | #define sk_refcnt_debug_dec(sk) do { } while (0) | |
1129 | #define sk_refcnt_debug_release(sk) do { } while (0) | |
1130 | #endif /* SOCK_REFCNT_DEBUG */ | |
1131 | ||
c255a458 | 1132 | #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_NET) |
c5905afb | 1133 | extern struct static_key memcg_socket_limit_enabled; |
e1aab161 GC |
1134 | static inline struct cg_proto *parent_cg_proto(struct proto *proto, |
1135 | struct cg_proto *cg_proto) | |
1136 | { | |
1137 | return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg)); | |
1138 | } | |
c5905afb | 1139 | #define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled) |
e1aab161 GC |
1140 | #else |
1141 | #define mem_cgroup_sockets_enabled 0 | |
1142 | static inline struct cg_proto *parent_cg_proto(struct proto *proto, | |
1143 | struct cg_proto *cg_proto) | |
1144 | { | |
1145 | return NULL; | |
1146 | } | |
1147 | #endif | |
1148 | ||
c9bee3b7 ED |
1149 | static inline bool sk_stream_memory_free(const struct sock *sk) |
1150 | { | |
1151 | if (sk->sk_wmem_queued >= sk->sk_sndbuf) | |
1152 | return false; | |
1153 | ||
1154 | return sk->sk_prot->stream_memory_free ? | |
1155 | sk->sk_prot->stream_memory_free(sk) : true; | |
1156 | } | |
1157 | ||
64dc6130 ED |
1158 | static inline bool sk_stream_is_writeable(const struct sock *sk) |
1159 | { | |
c9bee3b7 ED |
1160 | return sk_stream_wspace(sk) >= sk_stream_min_wspace(sk) && |
1161 | sk_stream_memory_free(sk); | |
64dc6130 | 1162 | } |
e1aab161 | 1163 | |
c9bee3b7 | 1164 | |
180d8cd9 GC |
1165 | static inline bool sk_has_memory_pressure(const struct sock *sk) |
1166 | { | |
1167 | return sk->sk_prot->memory_pressure != NULL; | |
1168 | } | |
1169 | ||
1170 | static inline bool sk_under_memory_pressure(const struct sock *sk) | |
1171 | { | |
1172 | if (!sk->sk_prot->memory_pressure) | |
1173 | return false; | |
e1aab161 GC |
1174 | |
1175 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) | |
2e685cad | 1176 | return !!sk->sk_cgrp->memory_pressure; |
e1aab161 | 1177 | |
35b87f6c | 1178 | return !!*sk->sk_prot->memory_pressure; |
180d8cd9 GC |
1179 | } |
1180 | ||
1181 | static inline void sk_leave_memory_pressure(struct sock *sk) | |
1182 | { | |
1183 | int *memory_pressure = sk->sk_prot->memory_pressure; | |
1184 | ||
e1aab161 GC |
1185 | if (!memory_pressure) |
1186 | return; | |
1187 | ||
1188 | if (*memory_pressure) | |
180d8cd9 | 1189 | *memory_pressure = 0; |
e1aab161 GC |
1190 | |
1191 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) { | |
1192 | struct cg_proto *cg_proto = sk->sk_cgrp; | |
1193 | struct proto *prot = sk->sk_prot; | |
1194 | ||
1195 | for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto)) | |
7f2cbdc2 | 1196 | cg_proto->memory_pressure = 0; |
e1aab161 GC |
1197 | } |
1198 | ||
180d8cd9 GC |
1199 | } |
1200 | ||
1201 | static inline void sk_enter_memory_pressure(struct sock *sk) | |
1202 | { | |
e1aab161 GC |
1203 | if (!sk->sk_prot->enter_memory_pressure) |
1204 | return; | |
1205 | ||
1206 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) { | |
1207 | struct cg_proto *cg_proto = sk->sk_cgrp; | |
1208 | struct proto *prot = sk->sk_prot; | |
1209 | ||
1210 | for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto)) | |
7f2cbdc2 | 1211 | cg_proto->memory_pressure = 1; |
e1aab161 GC |
1212 | } |
1213 | ||
1214 | sk->sk_prot->enter_memory_pressure(sk); | |
180d8cd9 GC |
1215 | } |
1216 | ||
1217 | static inline long sk_prot_mem_limits(const struct sock *sk, int index) | |
1218 | { | |
1219 | long *prot = sk->sk_prot->sysctl_mem; | |
e1aab161 GC |
1220 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) |
1221 | prot = sk->sk_cgrp->sysctl_mem; | |
180d8cd9 GC |
1222 | return prot[index]; |
1223 | } | |
1224 | ||
e1aab161 GC |
1225 | static inline void memcg_memory_allocated_add(struct cg_proto *prot, |
1226 | unsigned long amt, | |
1227 | int *parent_status) | |
1228 | { | |
3e32cb2e | 1229 | page_counter_charge(&prot->memory_allocated, amt); |
e1aab161 | 1230 | |
3e32cb2e JW |
1231 | if (page_counter_read(&prot->memory_allocated) > |
1232 | prot->memory_allocated.limit) | |
e1aab161 GC |
1233 | *parent_status = OVER_LIMIT; |
1234 | } | |
1235 | ||
1236 | static inline void memcg_memory_allocated_sub(struct cg_proto *prot, | |
1237 | unsigned long amt) | |
1238 | { | |
3e32cb2e | 1239 | page_counter_uncharge(&prot->memory_allocated, amt); |
e1aab161 GC |
1240 | } |
1241 | ||
180d8cd9 GC |
1242 | static inline long |
1243 | sk_memory_allocated(const struct sock *sk) | |
1244 | { | |
1245 | struct proto *prot = sk->sk_prot; | |
3e32cb2e | 1246 | |
e1aab161 | 1247 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) |
3e32cb2e | 1248 | return page_counter_read(&sk->sk_cgrp->memory_allocated); |
e1aab161 | 1249 | |
180d8cd9 GC |
1250 | return atomic_long_read(prot->memory_allocated); |
1251 | } | |
1252 | ||
1253 | static inline long | |
e1aab161 | 1254 | sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status) |
180d8cd9 GC |
1255 | { |
1256 | struct proto *prot = sk->sk_prot; | |
e1aab161 GC |
1257 | |
1258 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) { | |
1259 | memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status); | |
1260 | /* update the root cgroup regardless */ | |
1261 | atomic_long_add_return(amt, prot->memory_allocated); | |
3e32cb2e | 1262 | return page_counter_read(&sk->sk_cgrp->memory_allocated); |
e1aab161 GC |
1263 | } |
1264 | ||
180d8cd9 GC |
1265 | return atomic_long_add_return(amt, prot->memory_allocated); |
1266 | } | |
1267 | ||
1268 | static inline void | |
0e90b31f | 1269 | sk_memory_allocated_sub(struct sock *sk, int amt) |
180d8cd9 GC |
1270 | { |
1271 | struct proto *prot = sk->sk_prot; | |
e1aab161 | 1272 | |
0e90b31f | 1273 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) |
e1aab161 GC |
1274 | memcg_memory_allocated_sub(sk->sk_cgrp, amt); |
1275 | ||
180d8cd9 GC |
1276 | atomic_long_sub(amt, prot->memory_allocated); |
1277 | } | |
1278 | ||
1279 | static inline void sk_sockets_allocated_dec(struct sock *sk) | |
1280 | { | |
1281 | struct proto *prot = sk->sk_prot; | |
e1aab161 GC |
1282 | |
1283 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) { | |
1284 | struct cg_proto *cg_proto = sk->sk_cgrp; | |
1285 | ||
1286 | for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto)) | |
2e685cad | 1287 | percpu_counter_dec(&cg_proto->sockets_allocated); |
e1aab161 GC |
1288 | } |
1289 | ||
180d8cd9 GC |
1290 | percpu_counter_dec(prot->sockets_allocated); |
1291 | } | |
1292 | ||
1293 | static inline void sk_sockets_allocated_inc(struct sock *sk) | |
1294 | { | |
1295 | struct proto *prot = sk->sk_prot; | |
e1aab161 GC |
1296 | |
1297 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) { | |
1298 | struct cg_proto *cg_proto = sk->sk_cgrp; | |
1299 | ||
1300 | for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto)) | |
2e685cad | 1301 | percpu_counter_inc(&cg_proto->sockets_allocated); |
e1aab161 GC |
1302 | } |
1303 | ||
180d8cd9 GC |
1304 | percpu_counter_inc(prot->sockets_allocated); |
1305 | } | |
1306 | ||
1307 | static inline int | |
1308 | sk_sockets_allocated_read_positive(struct sock *sk) | |
1309 | { | |
1310 | struct proto *prot = sk->sk_prot; | |
1311 | ||
e1aab161 | 1312 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) |
2e685cad | 1313 | return percpu_counter_read_positive(&sk->sk_cgrp->sockets_allocated); |
e1aab161 | 1314 | |
518fbf9c | 1315 | return percpu_counter_read_positive(prot->sockets_allocated); |
180d8cd9 GC |
1316 | } |
1317 | ||
1318 | static inline int | |
1319 | proto_sockets_allocated_sum_positive(struct proto *prot) | |
1320 | { | |
1321 | return percpu_counter_sum_positive(prot->sockets_allocated); | |
1322 | } | |
1323 | ||
1324 | static inline long | |
1325 | proto_memory_allocated(struct proto *prot) | |
1326 | { | |
1327 | return atomic_long_read(prot->memory_allocated); | |
1328 | } | |
1329 | ||
1330 | static inline bool | |
1331 | proto_memory_pressure(struct proto *prot) | |
1332 | { | |
1333 | if (!prot->memory_pressure) | |
1334 | return false; | |
1335 | return !!*prot->memory_pressure; | |
1336 | } | |
1337 | ||
65f76517 ED |
1338 | |
1339 | #ifdef CONFIG_PROC_FS | |
1da177e4 | 1340 | /* Called with local bh disabled */ |
69336bd2 JP |
1341 | void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc); |
1342 | int sock_prot_inuse_get(struct net *net, struct proto *proto); | |
65f76517 | 1343 | #else |
dc6b9b78 | 1344 | static inline void sock_prot_inuse_add(struct net *net, struct proto *prot, |
c29a0bc4 | 1345 | int inc) |
65f76517 ED |
1346 | { |
1347 | } | |
65f76517 ED |
1348 | #endif |
1349 | ||
1da177e4 | 1350 | |
614c6cb4 ACM |
1351 | /* With per-bucket locks this operation is not-atomic, so that |
1352 | * this version is not worse. | |
1353 | */ | |
1354 | static inline void __sk_prot_rehash(struct sock *sk) | |
1355 | { | |
1356 | sk->sk_prot->unhash(sk); | |
1357 | sk->sk_prot->hash(sk); | |
1358 | } | |
1359 | ||
fcbdf09d OP |
1360 | void sk_prot_clear_portaddr_nulls(struct sock *sk, int size); |
1361 | ||
1da177e4 LT |
1362 | /* About 10 seconds */ |
1363 | #define SOCK_DESTROY_TIME (10*HZ) | |
1364 | ||
1365 | /* Sockets 0-1023 can't be bound to unless you are superuser */ | |
1366 | #define PROT_SOCK 1024 | |
1367 | ||
1368 | #define SHUTDOWN_MASK 3 | |
1369 | #define RCV_SHUTDOWN 1 | |
1370 | #define SEND_SHUTDOWN 2 | |
1371 | ||
1372 | #define SOCK_SNDBUF_LOCK 1 | |
1373 | #define SOCK_RCVBUF_LOCK 2 | |
1374 | #define SOCK_BINDADDR_LOCK 4 | |
1375 | #define SOCK_BINDPORT_LOCK 8 | |
1376 | ||
1da177e4 LT |
1377 | struct socket_alloc { |
1378 | struct socket socket; | |
1379 | struct inode vfs_inode; | |
1380 | }; | |
1381 | ||
1382 | static inline struct socket *SOCKET_I(struct inode *inode) | |
1383 | { | |
1384 | return &container_of(inode, struct socket_alloc, vfs_inode)->socket; | |
1385 | } | |
1386 | ||
1387 | static inline struct inode *SOCK_INODE(struct socket *socket) | |
1388 | { | |
1389 | return &container_of(socket, struct socket_alloc, socket)->vfs_inode; | |
1390 | } | |
1391 | ||
3ab224be HA |
1392 | /* |
1393 | * Functions for memory accounting | |
1394 | */ | |
69336bd2 JP |
1395 | int __sk_mem_schedule(struct sock *sk, int size, int kind); |
1396 | void __sk_mem_reclaim(struct sock *sk); | |
1da177e4 | 1397 | |
3ab224be HA |
1398 | #define SK_MEM_QUANTUM ((int)PAGE_SIZE) |
1399 | #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM) | |
1400 | #define SK_MEM_SEND 0 | |
1401 | #define SK_MEM_RECV 1 | |
1da177e4 | 1402 | |
3ab224be | 1403 | static inline int sk_mem_pages(int amt) |
1da177e4 | 1404 | { |
3ab224be | 1405 | return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT; |
1da177e4 LT |
1406 | } |
1407 | ||
dc6b9b78 | 1408 | static inline bool sk_has_account(struct sock *sk) |
1da177e4 | 1409 | { |
3ab224be HA |
1410 | /* return true if protocol supports memory accounting */ |
1411 | return !!sk->sk_prot->memory_allocated; | |
1da177e4 LT |
1412 | } |
1413 | ||
dc6b9b78 | 1414 | static inline bool sk_wmem_schedule(struct sock *sk, int size) |
1da177e4 | 1415 | { |
3ab224be | 1416 | if (!sk_has_account(sk)) |
dc6b9b78 | 1417 | return true; |
3ab224be HA |
1418 | return size <= sk->sk_forward_alloc || |
1419 | __sk_mem_schedule(sk, size, SK_MEM_SEND); | |
1da177e4 LT |
1420 | } |
1421 | ||
c76562b6 | 1422 | static inline bool |
35c448a8 | 1423 | sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size) |
d80d99d6 | 1424 | { |
3ab224be | 1425 | if (!sk_has_account(sk)) |
dc6b9b78 | 1426 | return true; |
c76562b6 MG |
1427 | return size<= sk->sk_forward_alloc || |
1428 | __sk_mem_schedule(sk, size, SK_MEM_RECV) || | |
1429 | skb_pfmemalloc(skb); | |
3ab224be HA |
1430 | } |
1431 | ||
1432 | static inline void sk_mem_reclaim(struct sock *sk) | |
1433 | { | |
1434 | if (!sk_has_account(sk)) | |
1435 | return; | |
1436 | if (sk->sk_forward_alloc >= SK_MEM_QUANTUM) | |
1437 | __sk_mem_reclaim(sk); | |
1438 | } | |
1439 | ||
9993e7d3 DM |
1440 | static inline void sk_mem_reclaim_partial(struct sock *sk) |
1441 | { | |
1442 | if (!sk_has_account(sk)) | |
1443 | return; | |
1444 | if (sk->sk_forward_alloc > SK_MEM_QUANTUM) | |
1445 | __sk_mem_reclaim(sk); | |
1446 | } | |
1447 | ||
3ab224be HA |
1448 | static inline void sk_mem_charge(struct sock *sk, int size) |
1449 | { | |
1450 | if (!sk_has_account(sk)) | |
1451 | return; | |
1452 | sk->sk_forward_alloc -= size; | |
1453 | } | |
1454 | ||
1455 | static inline void sk_mem_uncharge(struct sock *sk, int size) | |
1456 | { | |
1457 | if (!sk_has_account(sk)) | |
1458 | return; | |
1459 | sk->sk_forward_alloc += size; | |
1460 | } | |
1461 | ||
1462 | static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb) | |
1463 | { | |
3ab224be HA |
1464 | sock_set_flag(sk, SOCK_QUEUE_SHRUNK); |
1465 | sk->sk_wmem_queued -= skb->truesize; | |
1466 | sk_mem_uncharge(sk, skb->truesize); | |
1467 | __kfree_skb(skb); | |
d80d99d6 HX |
1468 | } |
1469 | ||
1da177e4 LT |
1470 | /* Used by processes to "lock" a socket state, so that |
1471 | * interrupts and bottom half handlers won't change it | |
1472 | * from under us. It essentially blocks any incoming | |
1473 | * packets, so that we won't get any new data or any | |
1474 | * packets that change the state of the socket. | |
1475 | * | |
1476 | * While locked, BH processing will add new packets to | |
1477 | * the backlog queue. This queue is processed by the | |
1478 | * owner of the socket lock right before it is released. | |
1479 | * | |
1480 | * Since ~2.3.5 it is also exclusive sleep lock serializing | |
1481 | * accesses from user process context. | |
1482 | */ | |
d2e9117c | 1483 | #define sock_owned_by_user(sk) ((sk)->sk_lock.owned) |
1da177e4 | 1484 | |
c3f9b018 ED |
1485 | static inline void sock_release_ownership(struct sock *sk) |
1486 | { | |
1487 | sk->sk_lock.owned = 0; | |
1488 | } | |
1489 | ||
ed07536e PZ |
1490 | /* |
1491 | * Macro so as to not evaluate some arguments when | |
1492 | * lockdep is not enabled. | |
1493 | * | |
1494 | * Mark both the sk_lock and the sk_lock.slock as a | |
1495 | * per-address-family lock class. | |
1496 | */ | |
dc6b9b78 | 1497 | #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \ |
ed07536e | 1498 | do { \ |
e8f6fbf6 | 1499 | sk->sk_lock.owned = 0; \ |
ed07536e PZ |
1500 | init_waitqueue_head(&sk->sk_lock.wq); \ |
1501 | spin_lock_init(&(sk)->sk_lock.slock); \ | |
1502 | debug_check_no_locks_freed((void *)&(sk)->sk_lock, \ | |
1503 | sizeof((sk)->sk_lock)); \ | |
1504 | lockdep_set_class_and_name(&(sk)->sk_lock.slock, \ | |
dc6b9b78 | 1505 | (skey), (sname)); \ |
ed07536e PZ |
1506 | lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \ |
1507 | } while (0) | |
1508 | ||
69336bd2 | 1509 | void lock_sock_nested(struct sock *sk, int subclass); |
fcc70d5f PZ |
1510 | |
1511 | static inline void lock_sock(struct sock *sk) | |
1512 | { | |
1513 | lock_sock_nested(sk, 0); | |
1514 | } | |
1515 | ||
69336bd2 | 1516 | void release_sock(struct sock *sk); |
1da177e4 LT |
1517 | |
1518 | /* BH context may only use the following locking interface. */ | |
1519 | #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock)) | |
c6366184 IM |
1520 | #define bh_lock_sock_nested(__sk) \ |
1521 | spin_lock_nested(&((__sk)->sk_lock.slock), \ | |
1522 | SINGLE_DEPTH_NESTING) | |
1da177e4 LT |
1523 | #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock)) |
1524 | ||
69336bd2 | 1525 | bool lock_sock_fast(struct sock *sk); |
8a74ad60 ED |
1526 | /** |
1527 | * unlock_sock_fast - complement of lock_sock_fast | |
1528 | * @sk: socket | |
1529 | * @slow: slow mode | |
1530 | * | |
1531 | * fast unlock socket for user context. | |
1532 | * If slow mode is on, we call regular release_sock() | |
1533 | */ | |
1534 | static inline void unlock_sock_fast(struct sock *sk, bool slow) | |
4b0b72f7 | 1535 | { |
8a74ad60 ED |
1536 | if (slow) |
1537 | release_sock(sk); | |
1538 | else | |
1539 | spin_unlock_bh(&sk->sk_lock.slock); | |
4b0b72f7 ED |
1540 | } |
1541 | ||
4b0b72f7 | 1542 | |
69336bd2 JP |
1543 | struct sock *sk_alloc(struct net *net, int family, gfp_t priority, |
1544 | struct proto *prot); | |
1545 | void sk_free(struct sock *sk); | |
1546 | void sk_release_kernel(struct sock *sk); | |
1547 | struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority); | |
1548 | ||
1549 | struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force, | |
1550 | gfp_t priority); | |
69336bd2 JP |
1551 | void sock_wfree(struct sk_buff *skb); |
1552 | void skb_orphan_partial(struct sk_buff *skb); | |
1553 | void sock_rfree(struct sk_buff *skb); | |
62bccb8c | 1554 | void sock_efree(struct sk_buff *skb); |
82eabd9e | 1555 | #ifdef CONFIG_INET |
69336bd2 | 1556 | void sock_edemux(struct sk_buff *skb); |
82eabd9e AD |
1557 | #else |
1558 | #define sock_edemux(skb) sock_efree(skb) | |
1559 | #endif | |
69336bd2 JP |
1560 | |
1561 | int sock_setsockopt(struct socket *sock, int level, int op, | |
1562 | char __user *optval, unsigned int optlen); | |
1563 | ||
1564 | int sock_getsockopt(struct socket *sock, int level, int op, | |
1565 | char __user *optval, int __user *optlen); | |
1566 | struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size, | |
1567 | int noblock, int *errcode); | |
1568 | struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len, | |
1569 | unsigned long data_len, int noblock, | |
1570 | int *errcode, int max_page_order); | |
1571 | void *sock_kmalloc(struct sock *sk, int size, gfp_t priority); | |
1572 | void sock_kfree_s(struct sock *sk, void *mem, int size); | |
79e88659 | 1573 | void sock_kzfree_s(struct sock *sk, void *mem, int size); |
69336bd2 | 1574 | void sk_send_sigurg(struct sock *sk); |
1da177e4 LT |
1575 | |
1576 | /* | |
1577 | * Functions to fill in entries in struct proto_ops when a protocol | |
1578 | * does not implement a particular function. | |
1579 | */ | |
69336bd2 JP |
1580 | int sock_no_bind(struct socket *, struct sockaddr *, int); |
1581 | int sock_no_connect(struct socket *, struct sockaddr *, int, int); | |
1582 | int sock_no_socketpair(struct socket *, struct socket *); | |
1583 | int sock_no_accept(struct socket *, struct socket *, int); | |
1584 | int sock_no_getname(struct socket *, struct sockaddr *, int *, int); | |
1585 | unsigned int sock_no_poll(struct file *, struct socket *, | |
1586 | struct poll_table_struct *); | |
1587 | int sock_no_ioctl(struct socket *, unsigned int, unsigned long); | |
1588 | int sock_no_listen(struct socket *, int); | |
1589 | int sock_no_shutdown(struct socket *, int); | |
1590 | int sock_no_getsockopt(struct socket *, int , int, char __user *, int __user *); | |
1591 | int sock_no_setsockopt(struct socket *, int, int, char __user *, unsigned int); | |
1592 | int sock_no_sendmsg(struct kiocb *, struct socket *, struct msghdr *, size_t); | |
1593 | int sock_no_recvmsg(struct kiocb *, struct socket *, struct msghdr *, size_t, | |
1594 | int); | |
1595 | int sock_no_mmap(struct file *file, struct socket *sock, | |
1596 | struct vm_area_struct *vma); | |
1597 | ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, | |
1598 | size_t size, int flags); | |
1da177e4 LT |
1599 | |
1600 | /* | |
1601 | * Functions to fill in entries in struct proto_ops when a protocol | |
1602 | * uses the inet style. | |
1603 | */ | |
69336bd2 | 1604 | int sock_common_getsockopt(struct socket *sock, int level, int optname, |
1da177e4 | 1605 | char __user *optval, int __user *optlen); |
69336bd2 | 1606 | int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock, |
1da177e4 | 1607 | struct msghdr *msg, size_t size, int flags); |
69336bd2 | 1608 | int sock_common_setsockopt(struct socket *sock, int level, int optname, |
b7058842 | 1609 | char __user *optval, unsigned int optlen); |
69336bd2 | 1610 | int compat_sock_common_getsockopt(struct socket *sock, int level, |
3fdadf7d | 1611 | int optname, char __user *optval, int __user *optlen); |
69336bd2 | 1612 | int compat_sock_common_setsockopt(struct socket *sock, int level, |
b7058842 | 1613 | int optname, char __user *optval, unsigned int optlen); |
1da177e4 | 1614 | |
69336bd2 | 1615 | void sk_common_release(struct sock *sk); |
1da177e4 LT |
1616 | |
1617 | /* | |
1618 | * Default socket callbacks and setup code | |
1619 | */ | |
dc6b9b78 | 1620 | |
1da177e4 | 1621 | /* Initialise core socket variables */ |
69336bd2 | 1622 | void sock_init_data(struct socket *sock, struct sock *sk); |
1da177e4 | 1623 | |
1da177e4 LT |
1624 | /* |
1625 | * Socket reference counting postulates. | |
1626 | * | |
1627 | * * Each user of socket SHOULD hold a reference count. | |
1628 | * * Each access point to socket (an hash table bucket, reference from a list, | |
1629 | * running timer, skb in flight MUST hold a reference count. | |
1630 | * * When reference count hits 0, it means it will never increase back. | |
1631 | * * When reference count hits 0, it means that no references from | |
1632 | * outside exist to this socket and current process on current CPU | |
1633 | * is last user and may/should destroy this socket. | |
1634 | * * sk_free is called from any context: process, BH, IRQ. When | |
1635 | * it is called, socket has no references from outside -> sk_free | |
1636 | * may release descendant resources allocated by the socket, but | |
1637 | * to the time when it is called, socket is NOT referenced by any | |
1638 | * hash tables, lists etc. | |
1639 | * * Packets, delivered from outside (from network or from another process) | |
1640 | * and enqueued on receive/error queues SHOULD NOT grab reference count, | |
1641 | * when they sit in queue. Otherwise, packets will leak to hole, when | |
1642 | * socket is looked up by one cpu and unhasing is made by another CPU. | |
1643 | * It is true for udp/raw, netlink (leak to receive and error queues), tcp | |
1644 | * (leak to backlog). Packet socket does all the processing inside | |
1645 | * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets | |
1646 | * use separate SMP lock, so that they are prone too. | |
1647 | */ | |
1648 | ||
1649 | /* Ungrab socket and destroy it, if it was the last reference. */ | |
1650 | static inline void sock_put(struct sock *sk) | |
1651 | { | |
1652 | if (atomic_dec_and_test(&sk->sk_refcnt)) | |
1653 | sk_free(sk); | |
1654 | } | |
05dbc7b5 ED |
1655 | /* Generic version of sock_put(), dealing with all sockets |
1656 | * (TCP_TIMEWAIT, ESTABLISHED...) | |
1657 | */ | |
1658 | void sock_gen_put(struct sock *sk); | |
1da177e4 | 1659 | |
69336bd2 | 1660 | int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested); |
25995ff5 | 1661 | |
e022f0b4 KK |
1662 | static inline void sk_tx_queue_set(struct sock *sk, int tx_queue) |
1663 | { | |
1664 | sk->sk_tx_queue_mapping = tx_queue; | |
1665 | } | |
1666 | ||
1667 | static inline void sk_tx_queue_clear(struct sock *sk) | |
1668 | { | |
1669 | sk->sk_tx_queue_mapping = -1; | |
1670 | } | |
1671 | ||
1672 | static inline int sk_tx_queue_get(const struct sock *sk) | |
1673 | { | |
b0f77d0e | 1674 | return sk ? sk->sk_tx_queue_mapping : -1; |
e022f0b4 KK |
1675 | } |
1676 | ||
972692e0 DM |
1677 | static inline void sk_set_socket(struct sock *sk, struct socket *sock) |
1678 | { | |
e022f0b4 | 1679 | sk_tx_queue_clear(sk); |
972692e0 DM |
1680 | sk->sk_socket = sock; |
1681 | } | |
1682 | ||
aa395145 ED |
1683 | static inline wait_queue_head_t *sk_sleep(struct sock *sk) |
1684 | { | |
eaefd110 ED |
1685 | BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0); |
1686 | return &rcu_dereference_raw(sk->sk_wq)->wait; | |
aa395145 | 1687 | } |
1da177e4 LT |
1688 | /* Detach socket from process context. |
1689 | * Announce socket dead, detach it from wait queue and inode. | |
1690 | * Note that parent inode held reference count on this struct sock, | |
1691 | * we do not release it in this function, because protocol | |
1692 | * probably wants some additional cleanups or even continuing | |
1693 | * to work with this socket (TCP). | |
1694 | */ | |
1695 | static inline void sock_orphan(struct sock *sk) | |
1696 | { | |
1697 | write_lock_bh(&sk->sk_callback_lock); | |
1698 | sock_set_flag(sk, SOCK_DEAD); | |
972692e0 | 1699 | sk_set_socket(sk, NULL); |
43815482 | 1700 | sk->sk_wq = NULL; |
1da177e4 LT |
1701 | write_unlock_bh(&sk->sk_callback_lock); |
1702 | } | |
1703 | ||
1704 | static inline void sock_graft(struct sock *sk, struct socket *parent) | |
1705 | { | |
1706 | write_lock_bh(&sk->sk_callback_lock); | |
eaefd110 | 1707 | sk->sk_wq = parent->wq; |
1da177e4 | 1708 | parent->sk = sk; |
972692e0 | 1709 | sk_set_socket(sk, parent); |
4237c75c | 1710 | security_sock_graft(sk, parent); |
1da177e4 LT |
1711 | write_unlock_bh(&sk->sk_callback_lock); |
1712 | } | |
1713 | ||
69336bd2 JP |
1714 | kuid_t sock_i_uid(struct sock *sk); |
1715 | unsigned long sock_i_ino(struct sock *sk); | |
1da177e4 LT |
1716 | |
1717 | static inline struct dst_entry * | |
1718 | __sk_dst_get(struct sock *sk) | |
1719 | { | |
d8bf4ca9 | 1720 | return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) || |
f68c224f | 1721 | lockdep_is_held(&sk->sk_lock.slock)); |
1da177e4 LT |
1722 | } |
1723 | ||
1724 | static inline struct dst_entry * | |
1725 | sk_dst_get(struct sock *sk) | |
1726 | { | |
1727 | struct dst_entry *dst; | |
1728 | ||
b6c6712a ED |
1729 | rcu_read_lock(); |
1730 | dst = rcu_dereference(sk->sk_dst_cache); | |
f8864972 ED |
1731 | if (dst && !atomic_inc_not_zero(&dst->__refcnt)) |
1732 | dst = NULL; | |
b6c6712a | 1733 | rcu_read_unlock(); |
1da177e4 LT |
1734 | return dst; |
1735 | } | |
1736 | ||
b6c6712a ED |
1737 | static inline void dst_negative_advice(struct sock *sk) |
1738 | { | |
1739 | struct dst_entry *ndst, *dst = __sk_dst_get(sk); | |
1740 | ||
1741 | if (dst && dst->ops->negative_advice) { | |
1742 | ndst = dst->ops->negative_advice(dst); | |
1743 | ||
1744 | if (ndst != dst) { | |
1745 | rcu_assign_pointer(sk->sk_dst_cache, ndst); | |
0a6957e7 | 1746 | sk_tx_queue_clear(sk); |
b6c6712a ED |
1747 | } |
1748 | } | |
1749 | } | |
1750 | ||
1da177e4 LT |
1751 | static inline void |
1752 | __sk_dst_set(struct sock *sk, struct dst_entry *dst) | |
1753 | { | |
1754 | struct dst_entry *old_dst; | |
1755 | ||
e022f0b4 | 1756 | sk_tx_queue_clear(sk); |
0b53ff2e ED |
1757 | /* |
1758 | * This can be called while sk is owned by the caller only, | |
1759 | * with no state that can be checked in a rcu_dereference_check() cond | |
1760 | */ | |
1761 | old_dst = rcu_dereference_raw(sk->sk_dst_cache); | |
b6c6712a | 1762 | rcu_assign_pointer(sk->sk_dst_cache, dst); |
1da177e4 LT |
1763 | dst_release(old_dst); |
1764 | } | |
1765 | ||
1766 | static inline void | |
1767 | sk_dst_set(struct sock *sk, struct dst_entry *dst) | |
1768 | { | |
7f502361 ED |
1769 | struct dst_entry *old_dst; |
1770 | ||
1771 | sk_tx_queue_clear(sk); | |
5925a055 | 1772 | old_dst = xchg((__force struct dst_entry **)&sk->sk_dst_cache, dst); |
7f502361 | 1773 | dst_release(old_dst); |
1da177e4 LT |
1774 | } |
1775 | ||
1776 | static inline void | |
1777 | __sk_dst_reset(struct sock *sk) | |
1778 | { | |
b6c6712a | 1779 | __sk_dst_set(sk, NULL); |
1da177e4 LT |
1780 | } |
1781 | ||
1782 | static inline void | |
1783 | sk_dst_reset(struct sock *sk) | |
1784 | { | |
7f502361 | 1785 | sk_dst_set(sk, NULL); |
1da177e4 LT |
1786 | } |
1787 | ||
69336bd2 | 1788 | struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie); |
1da177e4 | 1789 | |
69336bd2 | 1790 | struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie); |
1da177e4 | 1791 | |
dc6b9b78 | 1792 | static inline bool sk_can_gso(const struct sock *sk) |
bcd76111 HX |
1793 | { |
1794 | return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type); | |
1795 | } | |
1796 | ||
69336bd2 | 1797 | void sk_setup_caps(struct sock *sk, struct dst_entry *dst); |
6cbb0df7 | 1798 | |
c8f44aff | 1799 | static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags) |
a465419b ED |
1800 | { |
1801 | sk->sk_route_nocaps |= flags; | |
1802 | sk->sk_route_caps &= ~flags; | |
1803 | } | |
1804 | ||
c6e1a0d1 TH |
1805 | static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb, |
1806 | char __user *from, char *to, | |
912d398d | 1807 | int copy, int offset) |
c6e1a0d1 TH |
1808 | { |
1809 | if (skb->ip_summed == CHECKSUM_NONE) { | |
1810 | int err = 0; | |
1811 | __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err); | |
1812 | if (err) | |
1813 | return err; | |
912d398d | 1814 | skb->csum = csum_block_add(skb->csum, csum, offset); |
c6e1a0d1 TH |
1815 | } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) { |
1816 | if (!access_ok(VERIFY_READ, from, copy) || | |
1817 | __copy_from_user_nocache(to, from, copy)) | |
1818 | return -EFAULT; | |
1819 | } else if (copy_from_user(to, from, copy)) | |
1820 | return -EFAULT; | |
1821 | ||
1822 | return 0; | |
1823 | } | |
1824 | ||
1825 | static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb, | |
1826 | char __user *from, int copy) | |
1827 | { | |
912d398d | 1828 | int err, offset = skb->len; |
c6e1a0d1 | 1829 | |
912d398d WY |
1830 | err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy), |
1831 | copy, offset); | |
c6e1a0d1 | 1832 | if (err) |
912d398d | 1833 | __skb_trim(skb, offset); |
c6e1a0d1 TH |
1834 | |
1835 | return err; | |
1836 | } | |
1837 | ||
1838 | static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from, | |
1839 | struct sk_buff *skb, | |
1840 | struct page *page, | |
1841 | int off, int copy) | |
1842 | { | |
1843 | int err; | |
1844 | ||
912d398d WY |
1845 | err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off, |
1846 | copy, skb->len); | |
c6e1a0d1 TH |
1847 | if (err) |
1848 | return err; | |
1849 | ||
1850 | skb->len += copy; | |
1851 | skb->data_len += copy; | |
1852 | skb->truesize += copy; | |
1853 | sk->sk_wmem_queued += copy; | |
1854 | sk_mem_charge(sk, copy); | |
1855 | return 0; | |
1856 | } | |
1857 | ||
c564039f ED |
1858 | /** |
1859 | * sk_wmem_alloc_get - returns write allocations | |
1860 | * @sk: socket | |
1861 | * | |
1862 | * Returns sk_wmem_alloc minus initial offset of one | |
1863 | */ | |
1864 | static inline int sk_wmem_alloc_get(const struct sock *sk) | |
1865 | { | |
1866 | return atomic_read(&sk->sk_wmem_alloc) - 1; | |
1867 | } | |
1868 | ||
1869 | /** | |
1870 | * sk_rmem_alloc_get - returns read allocations | |
1871 | * @sk: socket | |
1872 | * | |
1873 | * Returns sk_rmem_alloc | |
1874 | */ | |
1875 | static inline int sk_rmem_alloc_get(const struct sock *sk) | |
1876 | { | |
1877 | return atomic_read(&sk->sk_rmem_alloc); | |
1878 | } | |
1879 | ||
1880 | /** | |
1881 | * sk_has_allocations - check if allocations are outstanding | |
1882 | * @sk: socket | |
1883 | * | |
1884 | * Returns true if socket has write or read allocations | |
1885 | */ | |
dc6b9b78 | 1886 | static inline bool sk_has_allocations(const struct sock *sk) |
c564039f ED |
1887 | { |
1888 | return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk); | |
1889 | } | |
1890 | ||
a57de0b4 | 1891 | /** |
43815482 | 1892 | * wq_has_sleeper - check if there are any waiting processes |
acfbe96a | 1893 | * @wq: struct socket_wq |
a57de0b4 | 1894 | * |
43815482 | 1895 | * Returns true if socket_wq has waiting processes |
a57de0b4 | 1896 | * |
43815482 | 1897 | * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory |
a57de0b4 JO |
1898 | * barrier call. They were added due to the race found within the tcp code. |
1899 | * | |
1900 | * Consider following tcp code paths: | |
1901 | * | |
1902 | * CPU1 CPU2 | |
1903 | * | |
1904 | * sys_select receive packet | |
1905 | * ... ... | |
1906 | * __add_wait_queue update tp->rcv_nxt | |
1907 | * ... ... | |
1908 | * tp->rcv_nxt check sock_def_readable | |
1909 | * ... { | |
43815482 ED |
1910 | * schedule rcu_read_lock(); |
1911 | * wq = rcu_dereference(sk->sk_wq); | |
1912 | * if (wq && waitqueue_active(&wq->wait)) | |
1913 | * wake_up_interruptible(&wq->wait) | |
a57de0b4 JO |
1914 | * ... |
1915 | * } | |
1916 | * | |
1917 | * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay | |
1918 | * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1 | |
1919 | * could then endup calling schedule and sleep forever if there are no more | |
1920 | * data on the socket. | |
ad462769 | 1921 | * |
a57de0b4 | 1922 | */ |
43815482 | 1923 | static inline bool wq_has_sleeper(struct socket_wq *wq) |
a57de0b4 | 1924 | { |
dc6b9b78 | 1925 | /* We need to be sure we are in sync with the |
a57de0b4 JO |
1926 | * add_wait_queue modifications to the wait queue. |
1927 | * | |
1928 | * This memory barrier is paired in the sock_poll_wait. | |
1929 | */ | |
43815482 ED |
1930 | smp_mb(); |
1931 | return wq && waitqueue_active(&wq->wait); | |
a57de0b4 JO |
1932 | } |
1933 | ||
1934 | /** | |
1935 | * sock_poll_wait - place memory barrier behind the poll_wait call. | |
1936 | * @filp: file | |
1937 | * @wait_address: socket wait queue | |
1938 | * @p: poll_table | |
1939 | * | |
43815482 | 1940 | * See the comments in the wq_has_sleeper function. |
a57de0b4 JO |
1941 | */ |
1942 | static inline void sock_poll_wait(struct file *filp, | |
1943 | wait_queue_head_t *wait_address, poll_table *p) | |
1944 | { | |
626cf236 | 1945 | if (!poll_does_not_wait(p) && wait_address) { |
a57de0b4 | 1946 | poll_wait(filp, wait_address, p); |
dc6b9b78 | 1947 | /* We need to be sure we are in sync with the |
a57de0b4 JO |
1948 | * socket flags modification. |
1949 | * | |
43815482 | 1950 | * This memory barrier is paired in the wq_has_sleeper. |
dc6b9b78 | 1951 | */ |
a57de0b4 JO |
1952 | smp_mb(); |
1953 | } | |
1954 | } | |
1955 | ||
b73c3d0e TH |
1956 | static inline void skb_set_hash_from_sk(struct sk_buff *skb, struct sock *sk) |
1957 | { | |
1958 | if (sk->sk_txhash) { | |
1959 | skb->l4_hash = 1; | |
1960 | skb->hash = sk->sk_txhash; | |
1961 | } | |
1962 | } | |
1963 | ||
1da177e4 | 1964 | /* |
dc6b9b78 | 1965 | * Queue a received datagram if it will fit. Stream and sequenced |
1da177e4 LT |
1966 | * protocols can't normally use this as they need to fit buffers in |
1967 | * and play with them. | |
1968 | * | |
dc6b9b78 | 1969 | * Inlined as it's very short and called for pretty much every |
1da177e4 LT |
1970 | * packet ever received. |
1971 | */ | |
1972 | ||
1973 | static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk) | |
1974 | { | |
d55d87fd | 1975 | skb_orphan(skb); |
1da177e4 LT |
1976 | skb->sk = sk; |
1977 | skb->destructor = sock_wfree; | |
b73c3d0e | 1978 | skb_set_hash_from_sk(skb, sk); |
2b85a34e ED |
1979 | /* |
1980 | * We used to take a refcount on sk, but following operation | |
1981 | * is enough to guarantee sk_free() wont free this sock until | |
1982 | * all in-flight packets are completed | |
1983 | */ | |
1da177e4 LT |
1984 | atomic_add(skb->truesize, &sk->sk_wmem_alloc); |
1985 | } | |
1986 | ||
1987 | static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk) | |
1988 | { | |
d55d87fd | 1989 | skb_orphan(skb); |
1da177e4 LT |
1990 | skb->sk = sk; |
1991 | skb->destructor = sock_rfree; | |
1992 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
3ab224be | 1993 | sk_mem_charge(sk, skb->truesize); |
1da177e4 LT |
1994 | } |
1995 | ||
69336bd2 JP |
1996 | void sk_reset_timer(struct sock *sk, struct timer_list *timer, |
1997 | unsigned long expires); | |
1da177e4 | 1998 | |
69336bd2 | 1999 | void sk_stop_timer(struct sock *sk, struct timer_list *timer); |
1da177e4 | 2000 | |
69336bd2 | 2001 | int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb); |
1da177e4 | 2002 | |
69336bd2 | 2003 | int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb); |
364a9e93 | 2004 | struct sk_buff *sock_dequeue_err_skb(struct sock *sk); |
1da177e4 LT |
2005 | |
2006 | /* | |
2007 | * Recover an error report and clear atomically | |
2008 | */ | |
dc6b9b78 | 2009 | |
1da177e4 LT |
2010 | static inline int sock_error(struct sock *sk) |
2011 | { | |
c1cbe4b7 BL |
2012 | int err; |
2013 | if (likely(!sk->sk_err)) | |
2014 | return 0; | |
2015 | err = xchg(&sk->sk_err, 0); | |
1da177e4 LT |
2016 | return -err; |
2017 | } | |
2018 | ||
2019 | static inline unsigned long sock_wspace(struct sock *sk) | |
2020 | { | |
2021 | int amt = 0; | |
2022 | ||
2023 | if (!(sk->sk_shutdown & SEND_SHUTDOWN)) { | |
2024 | amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc); | |
dc6b9b78 | 2025 | if (amt < 0) |
1da177e4 LT |
2026 | amt = 0; |
2027 | } | |
2028 | return amt; | |
2029 | } | |
2030 | ||
2031 | static inline void sk_wake_async(struct sock *sk, int how, int band) | |
2032 | { | |
bcdce719 | 2033 | if (sock_flag(sk, SOCK_FASYNC)) |
1da177e4 LT |
2034 | sock_wake_async(sk->sk_socket, how, band); |
2035 | } | |
2036 | ||
eea86af6 DB |
2037 | /* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might |
2038 | * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak. | |
2039 | * Note: for send buffers, TCP works better if we can build two skbs at | |
2040 | * minimum. | |
7a91b434 | 2041 | */ |
9eb5bf83 | 2042 | #define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff))) |
eea86af6 DB |
2043 | |
2044 | #define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2) | |
2045 | #define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE | |
1da177e4 LT |
2046 | |
2047 | static inline void sk_stream_moderate_sndbuf(struct sock *sk) | |
2048 | { | |
2049 | if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) { | |
8df09ea3 | 2050 | sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1); |
eea86af6 | 2051 | sk->sk_sndbuf = max_t(u32, sk->sk_sndbuf, SOCK_MIN_SNDBUF); |
1da177e4 LT |
2052 | } |
2053 | } | |
2054 | ||
df97c708 | 2055 | struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp); |
1da177e4 | 2056 | |
5640f768 ED |
2057 | /** |
2058 | * sk_page_frag - return an appropriate page_frag | |
2059 | * @sk: socket | |
2060 | * | |
2061 | * If socket allocation mode allows current thread to sleep, it means its | |
2062 | * safe to use the per task page_frag instead of the per socket one. | |
2063 | */ | |
2064 | static inline struct page_frag *sk_page_frag(struct sock *sk) | |
1da177e4 | 2065 | { |
5640f768 ED |
2066 | if (sk->sk_allocation & __GFP_WAIT) |
2067 | return ¤t->task_frag; | |
1da177e4 | 2068 | |
5640f768 | 2069 | return &sk->sk_frag; |
1da177e4 LT |
2070 | } |
2071 | ||
69336bd2 | 2072 | bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag); |
5640f768 | 2073 | |
1da177e4 LT |
2074 | /* |
2075 | * Default write policy as shown to user space via poll/select/SIGIO | |
2076 | */ | |
dc6b9b78 | 2077 | static inline bool sock_writeable(const struct sock *sk) |
1da177e4 | 2078 | { |
8df09ea3 | 2079 | return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1); |
1da177e4 LT |
2080 | } |
2081 | ||
dd0fc66f | 2082 | static inline gfp_t gfp_any(void) |
1da177e4 | 2083 | { |
99709372 | 2084 | return in_softirq() ? GFP_ATOMIC : GFP_KERNEL; |
1da177e4 LT |
2085 | } |
2086 | ||
dc6b9b78 | 2087 | static inline long sock_rcvtimeo(const struct sock *sk, bool noblock) |
1da177e4 LT |
2088 | { |
2089 | return noblock ? 0 : sk->sk_rcvtimeo; | |
2090 | } | |
2091 | ||
dc6b9b78 | 2092 | static inline long sock_sndtimeo(const struct sock *sk, bool noblock) |
1da177e4 LT |
2093 | { |
2094 | return noblock ? 0 : sk->sk_sndtimeo; | |
2095 | } | |
2096 | ||
2097 | static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len) | |
2098 | { | |
2099 | return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1; | |
2100 | } | |
2101 | ||
2102 | /* Alas, with timeout socket operations are not restartable. | |
2103 | * Compare this to poll(). | |
2104 | */ | |
2105 | static inline int sock_intr_errno(long timeo) | |
2106 | { | |
2107 | return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR; | |
2108 | } | |
2109 | ||
69336bd2 JP |
2110 | void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, |
2111 | struct sk_buff *skb); | |
2112 | void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk, | |
2113 | struct sk_buff *skb); | |
92f37fd2 | 2114 | |
dc6b9b78 | 2115 | static inline void |
1da177e4 LT |
2116 | sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb) |
2117 | { | |
b7aa0bf7 | 2118 | ktime_t kt = skb->tstamp; |
20d49473 | 2119 | struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb); |
a61bbcf2 | 2120 | |
20d49473 PO |
2121 | /* |
2122 | * generate control messages if | |
b9f40e21 | 2123 | * - receive time stamping in software requested |
20d49473 | 2124 | * - software time stamp available and wanted |
20d49473 | 2125 | * - hardware time stamps available and wanted |
20d49473 PO |
2126 | */ |
2127 | if (sock_flag(sk, SOCK_RCVTSTAMP) || | |
b9f40e21 | 2128 | (sk->sk_tsflags & SOF_TIMESTAMPING_RX_SOFTWARE) || |
c199105d | 2129 | (kt.tv64 && sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) || |
20d49473 | 2130 | (hwtstamps->hwtstamp.tv64 && |
b9f40e21 | 2131 | (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE))) |
92f37fd2 ED |
2132 | __sock_recv_timestamp(msg, sk, skb); |
2133 | else | |
b7aa0bf7 | 2134 | sk->sk_stamp = kt; |
6e3e939f JB |
2135 | |
2136 | if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid) | |
2137 | __sock_recv_wifi_status(msg, sk, skb); | |
1da177e4 LT |
2138 | } |
2139 | ||
69336bd2 JP |
2140 | void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, |
2141 | struct sk_buff *skb); | |
767dd033 ED |
2142 | |
2143 | static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, | |
2144 | struct sk_buff *skb) | |
2145 | { | |
2146 | #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \ | |
b9f40e21 WB |
2147 | (1UL << SOCK_RCVTSTAMP)) |
2148 | #define TSFLAGS_ANY (SOF_TIMESTAMPING_SOFTWARE | \ | |
2149 | SOF_TIMESTAMPING_RAW_HARDWARE) | |
767dd033 | 2150 | |
b9f40e21 | 2151 | if (sk->sk_flags & FLAGS_TS_OR_DROPS || sk->sk_tsflags & TSFLAGS_ANY) |
767dd033 ED |
2152 | __sock_recv_ts_and_drops(msg, sk, skb); |
2153 | else | |
2154 | sk->sk_stamp = skb->tstamp; | |
2155 | } | |
3b885787 | 2156 | |
67cc0d40 WB |
2157 | void __sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags); |
2158 | ||
20d49473 PO |
2159 | /** |
2160 | * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped | |
20d49473 | 2161 | * @sk: socket sending this packet |
140c55d4 ED |
2162 | * @tx_flags: completed with instructions for time stamping |
2163 | * | |
2164 | * Note : callers should take care of initial *tx_flags value (usually 0) | |
20d49473 | 2165 | */ |
67cc0d40 WB |
2166 | static inline void sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags) |
2167 | { | |
2168 | if (unlikely(sk->sk_tsflags)) | |
2169 | __sock_tx_timestamp(sk, tx_flags); | |
2170 | if (unlikely(sock_flag(sk, SOCK_WIFI_STATUS))) | |
2171 | *tx_flags |= SKBTX_WIFI_STATUS; | |
2172 | } | |
20d49473 | 2173 | |
1da177e4 LT |
2174 | /** |
2175 | * sk_eat_skb - Release a skb if it is no longer needed | |
4dc3b16b PP |
2176 | * @sk: socket to eat this skb from |
2177 | * @skb: socket buffer to eat | |
1da177e4 LT |
2178 | * |
2179 | * This routine must be called with interrupts disabled or with the socket | |
2180 | * locked so that the sk_buff queue operation is ok. | |
2181 | */ | |
7bced397 | 2182 | static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
2183 | { |
2184 | __skb_unlink(skb, &sk->sk_receive_queue); | |
2185 | __kfree_skb(skb); | |
2186 | } | |
2187 | ||
3b1e0a65 YH |
2188 | static inline |
2189 | struct net *sock_net(const struct sock *sk) | |
2190 | { | |
c2d9ba9b | 2191 | return read_pnet(&sk->sk_net); |
3b1e0a65 YH |
2192 | } |
2193 | ||
2194 | static inline | |
f5aa23fd | 2195 | void sock_net_set(struct sock *sk, struct net *net) |
3b1e0a65 | 2196 | { |
c2d9ba9b | 2197 | write_pnet(&sk->sk_net, net); |
3b1e0a65 YH |
2198 | } |
2199 | ||
edf02087 DL |
2200 | /* |
2201 | * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace. | |
25985edc | 2202 | * They should not hold a reference to a namespace in order to allow |
edf02087 DL |
2203 | * to stop it. |
2204 | * Sockets after sk_change_net should be released using sk_release_kernel | |
2205 | */ | |
2206 | static inline void sk_change_net(struct sock *sk, struct net *net) | |
2207 | { | |
5812521b GZ |
2208 | struct net *current_net = sock_net(sk); |
2209 | ||
2210 | if (!net_eq(current_net, net)) { | |
2211 | put_net(current_net); | |
2212 | sock_net_set(sk, hold_net(net)); | |
2213 | } | |
edf02087 DL |
2214 | } |
2215 | ||
23542618 KK |
2216 | static inline struct sock *skb_steal_sock(struct sk_buff *skb) |
2217 | { | |
efc27f8c | 2218 | if (skb->sk) { |
23542618 KK |
2219 | struct sock *sk = skb->sk; |
2220 | ||
2221 | skb->destructor = NULL; | |
2222 | skb->sk = NULL; | |
2223 | return sk; | |
2224 | } | |
2225 | return NULL; | |
2226 | } | |
2227 | ||
69336bd2 JP |
2228 | void sock_enable_timestamp(struct sock *sk, int flag); |
2229 | int sock_get_timestamp(struct sock *, struct timeval __user *); | |
2230 | int sock_get_timestampns(struct sock *, struct timespec __user *); | |
2231 | int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len, int level, | |
2232 | int type); | |
1da177e4 | 2233 | |
a3b299da EB |
2234 | bool sk_ns_capable(const struct sock *sk, |
2235 | struct user_namespace *user_ns, int cap); | |
2236 | bool sk_capable(const struct sock *sk, int cap); | |
2237 | bool sk_net_capable(const struct sock *sk, int cap); | |
2238 | ||
1da177e4 LT |
2239 | extern __u32 sysctl_wmem_max; |
2240 | extern __u32 sysctl_rmem_max; | |
2241 | ||
6baf1f41 DM |
2242 | extern int sysctl_optmem_max; |
2243 | ||
20380731 ACM |
2244 | extern __u32 sysctl_wmem_default; |
2245 | extern __u32 sysctl_rmem_default; | |
20380731 | 2246 | |
1da177e4 | 2247 | #endif /* _SOCK_H */ |