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> |
180d8cd9 | 57 | #include <linux/memcontrol.h> |
e1aab161 | 58 | #include <linux/res_counter.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> | |
70 | ||
9f048bfb ED |
71 | struct cgroup; |
72 | struct cgroup_subsys; | |
c607b2ed | 73 | #ifdef CONFIG_NET |
1d62e436 GC |
74 | int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss); |
75 | void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg); | |
c607b2ed GC |
76 | #else |
77 | static inline | |
1d62e436 | 78 | int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss) |
c607b2ed GC |
79 | { |
80 | return 0; | |
81 | } | |
82 | static inline | |
1d62e436 | 83 | void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg) |
c607b2ed GC |
84 | { |
85 | } | |
86 | #endif | |
1da177e4 LT |
87 | /* |
88 | * This structure really needs to be cleaned up. | |
89 | * Most of it is for TCP, and not used by any of | |
90 | * the other protocols. | |
91 | */ | |
92 | ||
93 | /* Define this to get the SOCK_DBG debugging facility. */ | |
94 | #define SOCK_DEBUGGING | |
95 | #ifdef SOCK_DEBUGGING | |
96 | #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \ | |
97 | printk(KERN_DEBUG msg); } while (0) | |
98 | #else | |
4cd9029d | 99 | /* Validate arguments and do nothing */ |
b9075fa9 | 100 | static inline __printf(2, 3) |
dc6b9b78 | 101 | void SOCK_DEBUG(const struct sock *sk, const char *msg, ...) |
4cd9029d SH |
102 | { |
103 | } | |
1da177e4 LT |
104 | #endif |
105 | ||
106 | /* This is the per-socket lock. The spinlock provides a synchronization | |
107 | * between user contexts and software interrupt processing, whereas the | |
108 | * mini-semaphore synchronizes multiple users amongst themselves. | |
109 | */ | |
1da177e4 LT |
110 | typedef struct { |
111 | spinlock_t slock; | |
d2e9117c | 112 | int owned; |
1da177e4 | 113 | wait_queue_head_t wq; |
a5b5bb9a IM |
114 | /* |
115 | * We express the mutex-alike socket_lock semantics | |
116 | * to the lock validator by explicitly managing | |
117 | * the slock as a lock variant (in addition to | |
118 | * the slock itself): | |
119 | */ | |
120 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
121 | struct lockdep_map dep_map; | |
122 | #endif | |
1da177e4 LT |
123 | } socket_lock_t; |
124 | ||
1da177e4 | 125 | struct sock; |
8feaf0c0 | 126 | struct proto; |
0eeb8ffc | 127 | struct net; |
1da177e4 LT |
128 | |
129 | /** | |
4dc3b16b | 130 | * struct sock_common - minimal network layer representation of sockets |
68835aba ED |
131 | * @skc_daddr: Foreign IPv4 addr |
132 | * @skc_rcv_saddr: Bound local IPv4 addr | |
4dc6dc71 | 133 | * @skc_hash: hash value used with various protocol lookup tables |
d4cada4a | 134 | * @skc_u16hashes: two u16 hash values used by UDP lookup tables |
4dc3b16b PP |
135 | * @skc_family: network address family |
136 | * @skc_state: Connection state | |
137 | * @skc_reuse: %SO_REUSEADDR setting | |
138 | * @skc_bound_dev_if: bound device index if != 0 | |
4dc3b16b | 139 | * @skc_bind_node: bind hash linkage for various protocol lookup tables |
512615b6 | 140 | * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol |
8feaf0c0 | 141 | * @skc_prot: protocol handlers inside a network family |
07feaebf | 142 | * @skc_net: reference to the network namespace of this socket |
68835aba ED |
143 | * @skc_node: main hash linkage for various protocol lookup tables |
144 | * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol | |
145 | * @skc_tx_queue_mapping: tx queue number for this connection | |
146 | * @skc_refcnt: reference count | |
4dc3b16b PP |
147 | * |
148 | * This is the minimal network layer representation of sockets, the header | |
8feaf0c0 ACM |
149 | * for struct sock and struct inet_timewait_sock. |
150 | */ | |
1da177e4 | 151 | struct sock_common { |
68835aba ED |
152 | /* skc_daddr and skc_rcv_saddr must be grouped : |
153 | * cf INET_MATCH() and INET_TW_MATCH() | |
4dc6dc71 | 154 | */ |
68835aba ED |
155 | __be32 skc_daddr; |
156 | __be32 skc_rcv_saddr; | |
4dc6dc71 | 157 | |
d4cada4a ED |
158 | union { |
159 | unsigned int skc_hash; | |
160 | __u16 skc_u16hashes[2]; | |
161 | }; | |
4dc6dc71 ED |
162 | unsigned short skc_family; |
163 | volatile unsigned char skc_state; | |
164 | unsigned char skc_reuse; | |
165 | int skc_bound_dev_if; | |
512615b6 ED |
166 | union { |
167 | struct hlist_node skc_bind_node; | |
168 | struct hlist_nulls_node skc_portaddr_node; | |
169 | }; | |
8feaf0c0 | 170 | struct proto *skc_prot; |
3b1e0a65 | 171 | #ifdef CONFIG_NET_NS |
07feaebf | 172 | struct net *skc_net; |
3b1e0a65 | 173 | #endif |
68835aba ED |
174 | /* |
175 | * fields between dontcopy_begin/dontcopy_end | |
176 | * are not copied in sock_copy() | |
177 | */ | |
928c41e7 | 178 | /* private: */ |
68835aba | 179 | int skc_dontcopy_begin[0]; |
928c41e7 | 180 | /* public: */ |
68835aba ED |
181 | union { |
182 | struct hlist_node skc_node; | |
183 | struct hlist_nulls_node skc_nulls_node; | |
184 | }; | |
185 | int skc_tx_queue_mapping; | |
186 | atomic_t skc_refcnt; | |
928c41e7 | 187 | /* private: */ |
68835aba | 188 | int skc_dontcopy_end[0]; |
928c41e7 | 189 | /* public: */ |
1da177e4 LT |
190 | }; |
191 | ||
e1aab161 | 192 | struct cg_proto; |
1da177e4 LT |
193 | /** |
194 | * struct sock - network layer representation of sockets | |
8feaf0c0 | 195 | * @__sk_common: shared layout with inet_timewait_sock |
4dc3b16b PP |
196 | * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN |
197 | * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings | |
198 | * @sk_lock: synchronizer | |
199 | * @sk_rcvbuf: size of receive buffer in bytes | |
43815482 | 200 | * @sk_wq: sock wait queue and async head |
deaa5854 | 201 | * @sk_rx_dst: receive input route used by early tcp demux |
4dc3b16b PP |
202 | * @sk_dst_cache: destination cache |
203 | * @sk_dst_lock: destination cache lock | |
204 | * @sk_policy: flow policy | |
4dc3b16b PP |
205 | * @sk_receive_queue: incoming packets |
206 | * @sk_wmem_alloc: transmit queue bytes committed | |
207 | * @sk_write_queue: Packet sending queue | |
97fc2f08 | 208 | * @sk_async_wait_queue: DMA copied packets |
4dc3b16b PP |
209 | * @sk_omem_alloc: "o" is "option" or "other" |
210 | * @sk_wmem_queued: persistent queue size | |
211 | * @sk_forward_alloc: space allocated forward | |
212 | * @sk_allocation: allocation mode | |
213 | * @sk_sndbuf: size of send buffer in bytes | |
33c732c3 | 214 | * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, |
20d49473 | 215 | * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings |
4dc3b16b PP |
216 | * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets |
217 | * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO) | |
a465419b | 218 | * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK) |
bcd76111 | 219 | * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4) |
82cc1a7a | 220 | * @sk_gso_max_size: Maximum GSO segment size to build |
1485348d | 221 | * @sk_gso_max_segs: Maximum number of GSO segments |
4dc3b16b | 222 | * @sk_lingertime: %SO_LINGER l_linger setting |
4dc3b16b PP |
223 | * @sk_backlog: always used with the per-socket spinlock held |
224 | * @sk_callback_lock: used with the callbacks in the end of this struct | |
225 | * @sk_error_queue: rarely used | |
33c732c3 WC |
226 | * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, |
227 | * IPV6_ADDRFORM for instance) | |
4dc3b16b | 228 | * @sk_err: last error |
33c732c3 WC |
229 | * @sk_err_soft: errors that don't cause failure but are the cause of a |
230 | * persistent failure not just 'timed out' | |
cb61cb9b | 231 | * @sk_drops: raw/udp drops counter |
4dc3b16b PP |
232 | * @sk_ack_backlog: current listen backlog |
233 | * @sk_max_ack_backlog: listen backlog set in listen() | |
234 | * @sk_priority: %SO_PRIORITY setting | |
1a3bc369 | 235 | * @sk_cgrp_prioidx: socket group's priority map index |
4dc3b16b PP |
236 | * @sk_type: socket type (%SOCK_STREAM, etc) |
237 | * @sk_protocol: which protocol this socket belongs in this network family | |
53c3fa20 RD |
238 | * @sk_peer_pid: &struct pid for this socket's peer |
239 | * @sk_peer_cred: %SO_PEERCRED setting | |
4dc3b16b PP |
240 | * @sk_rcvlowat: %SO_RCVLOWAT setting |
241 | * @sk_rcvtimeo: %SO_RCVTIMEO setting | |
242 | * @sk_sndtimeo: %SO_SNDTIMEO setting | |
c58dc01b | 243 | * @sk_rxhash: flow hash received from netif layer |
4dc3b16b PP |
244 | * @sk_filter: socket filtering instructions |
245 | * @sk_protinfo: private area, net family specific, when not using slab | |
246 | * @sk_timer: sock cleanup timer | |
247 | * @sk_stamp: time stamp of last packet received | |
248 | * @sk_socket: Identd and reporting IO signals | |
249 | * @sk_user_data: RPC layer private data | |
250 | * @sk_sndmsg_page: cached page for sendmsg | |
251 | * @sk_sndmsg_off: cached offset for sendmsg | |
d3d4f0a0 | 252 | * @sk_peek_off: current peek_offset value |
4dc3b16b | 253 | * @sk_send_head: front of stuff to transmit |
67be2dd1 | 254 | * @sk_security: used by security modules |
31729363 | 255 | * @sk_mark: generic packet mark |
53c3fa20 | 256 | * @sk_classid: this socket's cgroup classid |
e1aab161 | 257 | * @sk_cgrp: this socket's cgroup-specific proto data |
4dc3b16b PP |
258 | * @sk_write_pending: a write to stream socket waits to start |
259 | * @sk_state_change: callback to indicate change in the state of the sock | |
260 | * @sk_data_ready: callback to indicate there is data to be processed | |
261 | * @sk_write_space: callback to indicate there is bf sending space available | |
262 | * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE) | |
263 | * @sk_backlog_rcv: callback to process the backlog | |
264 | * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0 | |
1da177e4 LT |
265 | */ |
266 | struct sock { | |
267 | /* | |
8feaf0c0 | 268 | * Now struct inet_timewait_sock also uses sock_common, so please just |
1da177e4 LT |
269 | * don't add nothing before this first member (__sk_common) --acme |
270 | */ | |
271 | struct sock_common __sk_common; | |
4dc6dc71 ED |
272 | #define sk_node __sk_common.skc_node |
273 | #define sk_nulls_node __sk_common.skc_nulls_node | |
274 | #define sk_refcnt __sk_common.skc_refcnt | |
e022f0b4 | 275 | #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping |
4dc6dc71 | 276 | |
68835aba ED |
277 | #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin |
278 | #define sk_dontcopy_end __sk_common.skc_dontcopy_end | |
4dc6dc71 | 279 | #define sk_hash __sk_common.skc_hash |
1da177e4 LT |
280 | #define sk_family __sk_common.skc_family |
281 | #define sk_state __sk_common.skc_state | |
282 | #define sk_reuse __sk_common.skc_reuse | |
283 | #define sk_bound_dev_if __sk_common.skc_bound_dev_if | |
1da177e4 | 284 | #define sk_bind_node __sk_common.skc_bind_node |
8feaf0c0 | 285 | #define sk_prot __sk_common.skc_prot |
07feaebf | 286 | #define sk_net __sk_common.skc_net |
1da177e4 | 287 | socket_lock_t sk_lock; |
b178bb3d | 288 | struct sk_buff_head sk_receive_queue; |
fa438ccf ED |
289 | /* |
290 | * The backlog queue is special, it is always used with | |
291 | * the per-socket spinlock held and requires low latency | |
292 | * access. Therefore we special case it's implementation. | |
b178bb3d ED |
293 | * Note : rmem_alloc is in this structure to fill a hole |
294 | * on 64bit arches, not because its logically part of | |
295 | * backlog. | |
fa438ccf ED |
296 | */ |
297 | struct { | |
b178bb3d ED |
298 | atomic_t rmem_alloc; |
299 | int len; | |
300 | struct sk_buff *head; | |
301 | struct sk_buff *tail; | |
fa438ccf | 302 | } sk_backlog; |
b178bb3d ED |
303 | #define sk_rmem_alloc sk_backlog.rmem_alloc |
304 | int sk_forward_alloc; | |
305 | #ifdef CONFIG_RPS | |
306 | __u32 sk_rxhash; | |
307 | #endif | |
308 | atomic_t sk_drops; | |
309 | int sk_rcvbuf; | |
310 | ||
311 | struct sk_filter __rcu *sk_filter; | |
eaefd110 | 312 | struct socket_wq __rcu *sk_wq; |
b178bb3d ED |
313 | |
314 | #ifdef CONFIG_NET_DMA | |
315 | struct sk_buff_head sk_async_wait_queue; | |
316 | #endif | |
317 | ||
def8b4fa | 318 | #ifdef CONFIG_XFRM |
1da177e4 | 319 | struct xfrm_policy *sk_policy[2]; |
def8b4fa | 320 | #endif |
b178bb3d | 321 | unsigned long sk_flags; |
deaa5854 | 322 | struct dst_entry *sk_rx_dst; |
b178bb3d | 323 | struct dst_entry *sk_dst_cache; |
b6c6712a | 324 | spinlock_t sk_dst_lock; |
1da177e4 LT |
325 | atomic_t sk_wmem_alloc; |
326 | atomic_t sk_omem_alloc; | |
4e07a91c | 327 | int sk_sndbuf; |
1da177e4 | 328 | struct sk_buff_head sk_write_queue; |
b178bb3d ED |
329 | kmemcheck_bitfield_begin(flags); |
330 | unsigned int sk_shutdown : 2, | |
331 | sk_no_check : 2, | |
332 | sk_userlocks : 4, | |
333 | sk_protocol : 8, | |
334 | sk_type : 16; | |
335 | kmemcheck_bitfield_end(flags); | |
1da177e4 | 336 | int sk_wmem_queued; |
7d877f3b | 337 | gfp_t sk_allocation; |
c8f44aff MM |
338 | netdev_features_t sk_route_caps; |
339 | netdev_features_t sk_route_nocaps; | |
bcd76111 | 340 | int sk_gso_type; |
82cc1a7a | 341 | unsigned int sk_gso_max_size; |
1485348d | 342 | u16 sk_gso_max_segs; |
9932cf95 | 343 | int sk_rcvlowat; |
1da177e4 | 344 | unsigned long sk_lingertime; |
1da177e4 | 345 | struct sk_buff_head sk_error_queue; |
476e19cf | 346 | struct proto *sk_prot_creator; |
1da177e4 LT |
347 | rwlock_t sk_callback_lock; |
348 | int sk_err, | |
349 | sk_err_soft; | |
350 | unsigned short sk_ack_backlog; | |
351 | unsigned short sk_max_ack_backlog; | |
352 | __u32 sk_priority; | |
5bc1421e NH |
353 | #ifdef CONFIG_CGROUPS |
354 | __u32 sk_cgrp_prioidx; | |
355 | #endif | |
109f6e39 EB |
356 | struct pid *sk_peer_pid; |
357 | const struct cred *sk_peer_cred; | |
1da177e4 LT |
358 | long sk_rcvtimeo; |
359 | long sk_sndtimeo; | |
1da177e4 LT |
360 | void *sk_protinfo; |
361 | struct timer_list sk_timer; | |
b7aa0bf7 | 362 | ktime_t sk_stamp; |
1da177e4 LT |
363 | struct socket *sk_socket; |
364 | void *sk_user_data; | |
365 | struct page *sk_sndmsg_page; | |
366 | struct sk_buff *sk_send_head; | |
367 | __u32 sk_sndmsg_off; | |
ef64a54f | 368 | __s32 sk_peek_off; |
1da177e4 | 369 | int sk_write_pending; |
d5f64238 | 370 | #ifdef CONFIG_SECURITY |
1da177e4 | 371 | void *sk_security; |
d5f64238 | 372 | #endif |
4a19ec58 | 373 | __u32 sk_mark; |
f8451725 | 374 | u32 sk_classid; |
e1aab161 | 375 | struct cg_proto *sk_cgrp; |
1da177e4 LT |
376 | void (*sk_state_change)(struct sock *sk); |
377 | void (*sk_data_ready)(struct sock *sk, int bytes); | |
378 | void (*sk_write_space)(struct sock *sk); | |
379 | void (*sk_error_report)(struct sock *sk); | |
dc6b9b78 ED |
380 | int (*sk_backlog_rcv)(struct sock *sk, |
381 | struct sk_buff *skb); | |
1da177e4 LT |
382 | void (*sk_destruct)(struct sock *sk); |
383 | }; | |
384 | ||
4a17fd52 PE |
385 | /* |
386 | * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK | |
387 | * or not whether his port will be reused by someone else. SK_FORCE_REUSE | |
388 | * on a socket means that the socket will reuse everybody else's port | |
389 | * without looking at the other's sk_reuse value. | |
390 | */ | |
391 | ||
392 | #define SK_NO_REUSE 0 | |
393 | #define SK_CAN_REUSE 1 | |
394 | #define SK_FORCE_REUSE 2 | |
395 | ||
ef64a54f PE |
396 | static inline int sk_peek_offset(struct sock *sk, int flags) |
397 | { | |
398 | if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0)) | |
399 | return sk->sk_peek_off; | |
400 | else | |
401 | return 0; | |
402 | } | |
403 | ||
404 | static inline void sk_peek_offset_bwd(struct sock *sk, int val) | |
405 | { | |
406 | if (sk->sk_peek_off >= 0) { | |
407 | if (sk->sk_peek_off >= val) | |
408 | sk->sk_peek_off -= val; | |
409 | else | |
410 | sk->sk_peek_off = 0; | |
411 | } | |
412 | } | |
413 | ||
414 | static inline void sk_peek_offset_fwd(struct sock *sk, int val) | |
415 | { | |
416 | if (sk->sk_peek_off >= 0) | |
417 | sk->sk_peek_off += val; | |
418 | } | |
419 | ||
1da177e4 LT |
420 | /* |
421 | * Hashed lists helper routines | |
422 | */ | |
c4146644 LZ |
423 | static inline struct sock *sk_entry(const struct hlist_node *node) |
424 | { | |
425 | return hlist_entry(node, struct sock, sk_node); | |
426 | } | |
427 | ||
e48c414e | 428 | static inline struct sock *__sk_head(const struct hlist_head *head) |
1da177e4 LT |
429 | { |
430 | return hlist_entry(head->first, struct sock, sk_node); | |
431 | } | |
432 | ||
e48c414e | 433 | static inline struct sock *sk_head(const struct hlist_head *head) |
1da177e4 LT |
434 | { |
435 | return hlist_empty(head) ? NULL : __sk_head(head); | |
436 | } | |
437 | ||
88ab1932 ED |
438 | static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head) |
439 | { | |
440 | return hlist_nulls_entry(head->first, struct sock, sk_nulls_node); | |
441 | } | |
442 | ||
443 | static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head) | |
444 | { | |
445 | return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head); | |
446 | } | |
447 | ||
e48c414e | 448 | static inline struct sock *sk_next(const struct sock *sk) |
1da177e4 LT |
449 | { |
450 | return sk->sk_node.next ? | |
451 | hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL; | |
452 | } | |
453 | ||
88ab1932 ED |
454 | static inline struct sock *sk_nulls_next(const struct sock *sk) |
455 | { | |
456 | return (!is_a_nulls(sk->sk_nulls_node.next)) ? | |
457 | hlist_nulls_entry(sk->sk_nulls_node.next, | |
458 | struct sock, sk_nulls_node) : | |
459 | NULL; | |
460 | } | |
461 | ||
dc6b9b78 | 462 | static inline bool sk_unhashed(const struct sock *sk) |
1da177e4 LT |
463 | { |
464 | return hlist_unhashed(&sk->sk_node); | |
465 | } | |
466 | ||
dc6b9b78 | 467 | static inline bool sk_hashed(const struct sock *sk) |
1da177e4 | 468 | { |
da753bea | 469 | return !sk_unhashed(sk); |
1da177e4 LT |
470 | } |
471 | ||
dc6b9b78 | 472 | static inline void sk_node_init(struct hlist_node *node) |
1da177e4 LT |
473 | { |
474 | node->pprev = NULL; | |
475 | } | |
476 | ||
dc6b9b78 | 477 | static inline void sk_nulls_node_init(struct hlist_nulls_node *node) |
88ab1932 ED |
478 | { |
479 | node->pprev = NULL; | |
480 | } | |
481 | ||
dc6b9b78 | 482 | static inline void __sk_del_node(struct sock *sk) |
1da177e4 LT |
483 | { |
484 | __hlist_del(&sk->sk_node); | |
485 | } | |
486 | ||
808f5114 | 487 | /* NB: equivalent to hlist_del_init_rcu */ |
dc6b9b78 | 488 | static inline bool __sk_del_node_init(struct sock *sk) |
1da177e4 LT |
489 | { |
490 | if (sk_hashed(sk)) { | |
491 | __sk_del_node(sk); | |
492 | sk_node_init(&sk->sk_node); | |
dc6b9b78 | 493 | return true; |
1da177e4 | 494 | } |
dc6b9b78 | 495 | return false; |
1da177e4 LT |
496 | } |
497 | ||
498 | /* Grab socket reference count. This operation is valid only | |
499 | when sk is ALREADY grabbed f.e. it is found in hash table | |
500 | or a list and the lookup is made under lock preventing hash table | |
501 | modifications. | |
502 | */ | |
503 | ||
504 | static inline void sock_hold(struct sock *sk) | |
505 | { | |
506 | atomic_inc(&sk->sk_refcnt); | |
507 | } | |
508 | ||
509 | /* Ungrab socket in the context, which assumes that socket refcnt | |
510 | cannot hit zero, f.e. it is true in context of any socketcall. | |
511 | */ | |
512 | static inline void __sock_put(struct sock *sk) | |
513 | { | |
514 | atomic_dec(&sk->sk_refcnt); | |
515 | } | |
516 | ||
dc6b9b78 | 517 | static inline bool sk_del_node_init(struct sock *sk) |
1da177e4 | 518 | { |
dc6b9b78 | 519 | bool rc = __sk_del_node_init(sk); |
1da177e4 LT |
520 | |
521 | if (rc) { | |
522 | /* paranoid for a while -acme */ | |
523 | WARN_ON(atomic_read(&sk->sk_refcnt) == 1); | |
524 | __sock_put(sk); | |
525 | } | |
526 | return rc; | |
527 | } | |
808f5114 | 528 | #define sk_del_node_init_rcu(sk) sk_del_node_init(sk) |
1da177e4 | 529 | |
dc6b9b78 | 530 | static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk) |
271b72c7 ED |
531 | { |
532 | if (sk_hashed(sk)) { | |
88ab1932 | 533 | hlist_nulls_del_init_rcu(&sk->sk_nulls_node); |
dc6b9b78 | 534 | return true; |
271b72c7 | 535 | } |
dc6b9b78 | 536 | return false; |
271b72c7 ED |
537 | } |
538 | ||
dc6b9b78 | 539 | static inline bool sk_nulls_del_node_init_rcu(struct sock *sk) |
271b72c7 | 540 | { |
dc6b9b78 | 541 | bool rc = __sk_nulls_del_node_init_rcu(sk); |
271b72c7 ED |
542 | |
543 | if (rc) { | |
544 | /* paranoid for a while -acme */ | |
545 | WARN_ON(atomic_read(&sk->sk_refcnt) == 1); | |
546 | __sock_put(sk); | |
547 | } | |
548 | return rc; | |
549 | } | |
550 | ||
dc6b9b78 | 551 | static inline void __sk_add_node(struct sock *sk, struct hlist_head *list) |
1da177e4 LT |
552 | { |
553 | hlist_add_head(&sk->sk_node, list); | |
554 | } | |
555 | ||
dc6b9b78 | 556 | static inline void sk_add_node(struct sock *sk, struct hlist_head *list) |
1da177e4 LT |
557 | { |
558 | sock_hold(sk); | |
559 | __sk_add_node(sk, list); | |
560 | } | |
561 | ||
dc6b9b78 | 562 | static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list) |
808f5114 | 563 | { |
564 | sock_hold(sk); | |
565 | hlist_add_head_rcu(&sk->sk_node, list); | |
566 | } | |
567 | ||
dc6b9b78 | 568 | static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list) |
271b72c7 | 569 | { |
88ab1932 | 570 | hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list); |
271b72c7 ED |
571 | } |
572 | ||
dc6b9b78 | 573 | static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list) |
271b72c7 ED |
574 | { |
575 | sock_hold(sk); | |
88ab1932 | 576 | __sk_nulls_add_node_rcu(sk, list); |
271b72c7 ED |
577 | } |
578 | ||
dc6b9b78 | 579 | static inline void __sk_del_bind_node(struct sock *sk) |
1da177e4 LT |
580 | { |
581 | __hlist_del(&sk->sk_bind_node); | |
582 | } | |
583 | ||
dc6b9b78 | 584 | static inline void sk_add_bind_node(struct sock *sk, |
1da177e4 LT |
585 | struct hlist_head *list) |
586 | { | |
587 | hlist_add_head(&sk->sk_bind_node, list); | |
588 | } | |
589 | ||
590 | #define sk_for_each(__sk, node, list) \ | |
591 | hlist_for_each_entry(__sk, node, list, sk_node) | |
808f5114 | 592 | #define sk_for_each_rcu(__sk, node, list) \ |
593 | hlist_for_each_entry_rcu(__sk, node, list, sk_node) | |
88ab1932 ED |
594 | #define sk_nulls_for_each(__sk, node, list) \ |
595 | hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node) | |
596 | #define sk_nulls_for_each_rcu(__sk, node, list) \ | |
597 | hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node) | |
1da177e4 LT |
598 | #define sk_for_each_from(__sk, node) \ |
599 | if (__sk && ({ node = &(__sk)->sk_node; 1; })) \ | |
600 | hlist_for_each_entry_from(__sk, node, sk_node) | |
88ab1932 ED |
601 | #define sk_nulls_for_each_from(__sk, node) \ |
602 | if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \ | |
603 | hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node) | |
1da177e4 LT |
604 | #define sk_for_each_safe(__sk, node, tmp, list) \ |
605 | hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node) | |
606 | #define sk_for_each_bound(__sk, node, list) \ | |
607 | hlist_for_each_entry(__sk, node, list, sk_bind_node) | |
608 | ||
609 | /* Sock flags */ | |
610 | enum sock_flags { | |
611 | SOCK_DEAD, | |
612 | SOCK_DONE, | |
613 | SOCK_URGINLINE, | |
614 | SOCK_KEEPOPEN, | |
615 | SOCK_LINGER, | |
616 | SOCK_DESTROY, | |
617 | SOCK_BROADCAST, | |
618 | SOCK_TIMESTAMP, | |
619 | SOCK_ZAPPED, | |
620 | SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */ | |
621 | SOCK_DBG, /* %SO_DEBUG setting */ | |
622 | SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */ | |
92f37fd2 | 623 | SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */ |
1da177e4 LT |
624 | SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */ |
625 | SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */ | |
7cb02404 | 626 | SOCK_MEMALLOC, /* VM depends on this socket for swapping */ |
20d49473 PO |
627 | SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */ |
628 | SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */ | |
629 | SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */ | |
630 | SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */ | |
631 | SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */ | |
632 | SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */ | |
633 | SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */ | |
bcdce719 | 634 | SOCK_FASYNC, /* fasync() active */ |
3b885787 | 635 | SOCK_RXQ_OVFL, |
1cdebb42 | 636 | SOCK_ZEROCOPY, /* buffers from userspace */ |
6e3e939f | 637 | SOCK_WIFI_STATUS, /* push wifi status to userspace */ |
3bdc0eba BG |
638 | SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS. |
639 | * Will use last 4 bytes of packet sent from | |
640 | * user-space instead. | |
641 | */ | |
1da177e4 LT |
642 | }; |
643 | ||
53b924b3 RB |
644 | static inline void sock_copy_flags(struct sock *nsk, struct sock *osk) |
645 | { | |
646 | nsk->sk_flags = osk->sk_flags; | |
647 | } | |
648 | ||
1da177e4 LT |
649 | static inline void sock_set_flag(struct sock *sk, enum sock_flags flag) |
650 | { | |
651 | __set_bit(flag, &sk->sk_flags); | |
652 | } | |
653 | ||
654 | static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag) | |
655 | { | |
656 | __clear_bit(flag, &sk->sk_flags); | |
657 | } | |
658 | ||
1b23a5df | 659 | static inline bool sock_flag(const struct sock *sk, enum sock_flags flag) |
1da177e4 LT |
660 | { |
661 | return test_bit(flag, &sk->sk_flags); | |
662 | } | |
663 | ||
c93bdd0e MG |
664 | #ifdef CONFIG_NET |
665 | extern struct static_key memalloc_socks; | |
666 | static inline int sk_memalloc_socks(void) | |
667 | { | |
668 | return static_key_false(&memalloc_socks); | |
669 | } | |
670 | #else | |
671 | ||
672 | static inline int sk_memalloc_socks(void) | |
673 | { | |
674 | return 0; | |
675 | } | |
676 | ||
677 | #endif | |
678 | ||
99a1dec7 MG |
679 | static inline gfp_t sk_gfp_atomic(struct sock *sk, gfp_t gfp_mask) |
680 | { | |
7cb02404 | 681 | return GFP_ATOMIC | (sk->sk_allocation & __GFP_MEMALLOC); |
99a1dec7 MG |
682 | } |
683 | ||
1da177e4 LT |
684 | static inline void sk_acceptq_removed(struct sock *sk) |
685 | { | |
686 | sk->sk_ack_backlog--; | |
687 | } | |
688 | ||
689 | static inline void sk_acceptq_added(struct sock *sk) | |
690 | { | |
691 | sk->sk_ack_backlog++; | |
692 | } | |
693 | ||
dc6b9b78 | 694 | static inline bool sk_acceptq_is_full(const struct sock *sk) |
1da177e4 | 695 | { |
64a14651 | 696 | return sk->sk_ack_backlog > sk->sk_max_ack_backlog; |
1da177e4 LT |
697 | } |
698 | ||
699 | /* | |
700 | * Compute minimal free write space needed to queue new packets. | |
701 | */ | |
dc6b9b78 | 702 | static inline int sk_stream_min_wspace(const struct sock *sk) |
1da177e4 | 703 | { |
8df09ea3 | 704 | return sk->sk_wmem_queued >> 1; |
1da177e4 LT |
705 | } |
706 | ||
dc6b9b78 | 707 | static inline int sk_stream_wspace(const struct sock *sk) |
1da177e4 LT |
708 | { |
709 | return sk->sk_sndbuf - sk->sk_wmem_queued; | |
710 | } | |
711 | ||
712 | extern void sk_stream_write_space(struct sock *sk); | |
713 | ||
dc6b9b78 | 714 | static inline bool sk_stream_memory_free(const struct sock *sk) |
1da177e4 LT |
715 | { |
716 | return sk->sk_wmem_queued < sk->sk_sndbuf; | |
717 | } | |
718 | ||
8eae939f | 719 | /* OOB backlog add */ |
a3a858ff | 720 | static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb) |
9ee6b535 | 721 | { |
7fee226a ED |
722 | /* dont let skb dst not refcounted, we are going to leave rcu lock */ |
723 | skb_dst_force(skb); | |
724 | ||
725 | if (!sk->sk_backlog.tail) | |
726 | sk->sk_backlog.head = skb; | |
727 | else | |
9ee6b535 | 728 | sk->sk_backlog.tail->next = skb; |
7fee226a ED |
729 | |
730 | sk->sk_backlog.tail = skb; | |
9ee6b535 SH |
731 | skb->next = NULL; |
732 | } | |
1da177e4 | 733 | |
c377411f ED |
734 | /* |
735 | * Take into account size of receive queue and backlog queue | |
0fd7bac6 ED |
736 | * Do not take into account this skb truesize, |
737 | * to allow even a single big packet to come. | |
c377411f | 738 | */ |
f545a38f ED |
739 | static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb, |
740 | unsigned int limit) | |
c377411f ED |
741 | { |
742 | unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc); | |
743 | ||
f545a38f | 744 | return qsize > limit; |
c377411f ED |
745 | } |
746 | ||
8eae939f | 747 | /* The per-socket spinlock must be held here. */ |
f545a38f ED |
748 | static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb, |
749 | unsigned int limit) | |
8eae939f | 750 | { |
f545a38f | 751 | if (sk_rcvqueues_full(sk, skb, limit)) |
8eae939f ZY |
752 | return -ENOBUFS; |
753 | ||
a3a858ff | 754 | __sk_add_backlog(sk, skb); |
8eae939f ZY |
755 | sk->sk_backlog.len += skb->truesize; |
756 | return 0; | |
757 | } | |
758 | ||
b4b9e355 MG |
759 | extern int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb); |
760 | ||
c57943a1 PZ |
761 | static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb) |
762 | { | |
b4b9e355 MG |
763 | if (sk_memalloc_socks() && skb_pfmemalloc(skb)) |
764 | return __sk_backlog_rcv(sk, skb); | |
765 | ||
c57943a1 PZ |
766 | return sk->sk_backlog_rcv(sk, skb); |
767 | } | |
768 | ||
c58dc01b DM |
769 | static inline void sock_rps_record_flow(const struct sock *sk) |
770 | { | |
771 | #ifdef CONFIG_RPS | |
772 | struct rps_sock_flow_table *sock_flow_table; | |
773 | ||
774 | rcu_read_lock(); | |
775 | sock_flow_table = rcu_dereference(rps_sock_flow_table); | |
776 | rps_record_sock_flow(sock_flow_table, sk->sk_rxhash); | |
777 | rcu_read_unlock(); | |
778 | #endif | |
779 | } | |
780 | ||
781 | static inline void sock_rps_reset_flow(const struct sock *sk) | |
782 | { | |
783 | #ifdef CONFIG_RPS | |
784 | struct rps_sock_flow_table *sock_flow_table; | |
785 | ||
786 | rcu_read_lock(); | |
787 | sock_flow_table = rcu_dereference(rps_sock_flow_table); | |
788 | rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash); | |
789 | rcu_read_unlock(); | |
790 | #endif | |
791 | } | |
792 | ||
bdeab991 TH |
793 | static inline void sock_rps_save_rxhash(struct sock *sk, |
794 | const struct sk_buff *skb) | |
c58dc01b DM |
795 | { |
796 | #ifdef CONFIG_RPS | |
bdeab991 | 797 | if (unlikely(sk->sk_rxhash != skb->rxhash)) { |
c58dc01b | 798 | sock_rps_reset_flow(sk); |
bdeab991 | 799 | sk->sk_rxhash = skb->rxhash; |
c58dc01b DM |
800 | } |
801 | #endif | |
802 | } | |
803 | ||
bdeab991 TH |
804 | static inline void sock_rps_reset_rxhash(struct sock *sk) |
805 | { | |
806 | #ifdef CONFIG_RPS | |
807 | sock_rps_reset_flow(sk); | |
808 | sk->sk_rxhash = 0; | |
809 | #endif | |
810 | } | |
811 | ||
cfcabdcc SH |
812 | #define sk_wait_event(__sk, __timeo, __condition) \ |
813 | ({ int __rc; \ | |
814 | release_sock(__sk); \ | |
815 | __rc = __condition; \ | |
816 | if (!__rc) { \ | |
817 | *(__timeo) = schedule_timeout(*(__timeo)); \ | |
818 | } \ | |
819 | lock_sock(__sk); \ | |
820 | __rc = __condition; \ | |
821 | __rc; \ | |
822 | }) | |
1da177e4 LT |
823 | |
824 | extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p); | |
825 | extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p); | |
826 | extern void sk_stream_wait_close(struct sock *sk, long timeo_p); | |
827 | extern int sk_stream_error(struct sock *sk, int flags, int err); | |
828 | extern void sk_stream_kill_queues(struct sock *sk); | |
7cb02404 MG |
829 | extern void sk_set_memalloc(struct sock *sk); |
830 | extern void sk_clear_memalloc(struct sock *sk); | |
1da177e4 LT |
831 | |
832 | extern int sk_wait_data(struct sock *sk, long *timeo); | |
833 | ||
60236fdd | 834 | struct request_sock_ops; |
6d6ee43e | 835 | struct timewait_sock_ops; |
ab1e0a13 | 836 | struct inet_hashinfo; |
fc8717ba | 837 | struct raw_hashinfo; |
de477254 | 838 | struct module; |
2e6599cb | 839 | |
1da177e4 LT |
840 | /* Networking protocol blocks we attach to sockets. |
841 | * socket layer -> transport layer interface | |
842 | * transport -> network interface is defined by struct inet_proto | |
843 | */ | |
844 | struct proto { | |
dc6b9b78 | 845 | void (*close)(struct sock *sk, |
1da177e4 LT |
846 | long timeout); |
847 | int (*connect)(struct sock *sk, | |
dc6b9b78 | 848 | struct sockaddr *uaddr, |
1da177e4 LT |
849 | int addr_len); |
850 | int (*disconnect)(struct sock *sk, int flags); | |
851 | ||
dc6b9b78 | 852 | struct sock * (*accept)(struct sock *sk, int flags, int *err); |
1da177e4 LT |
853 | |
854 | int (*ioctl)(struct sock *sk, int cmd, | |
855 | unsigned long arg); | |
856 | int (*init)(struct sock *sk); | |
7d06b2e0 | 857 | void (*destroy)(struct sock *sk); |
1da177e4 | 858 | void (*shutdown)(struct sock *sk, int how); |
dc6b9b78 | 859 | int (*setsockopt)(struct sock *sk, int level, |
1da177e4 | 860 | int optname, char __user *optval, |
b7058842 | 861 | unsigned int optlen); |
dc6b9b78 ED |
862 | int (*getsockopt)(struct sock *sk, int level, |
863 | int optname, char __user *optval, | |
864 | int __user *option); | |
af01d537 | 865 | #ifdef CONFIG_COMPAT |
3fdadf7d DM |
866 | int (*compat_setsockopt)(struct sock *sk, |
867 | int level, | |
868 | int optname, char __user *optval, | |
b7058842 | 869 | unsigned int optlen); |
3fdadf7d DM |
870 | int (*compat_getsockopt)(struct sock *sk, |
871 | int level, | |
872 | int optname, char __user *optval, | |
873 | int __user *option); | |
709b46e8 EB |
874 | int (*compat_ioctl)(struct sock *sk, |
875 | unsigned int cmd, unsigned long arg); | |
af01d537 | 876 | #endif |
1da177e4 LT |
877 | int (*sendmsg)(struct kiocb *iocb, struct sock *sk, |
878 | struct msghdr *msg, size_t len); | |
879 | int (*recvmsg)(struct kiocb *iocb, struct sock *sk, | |
880 | struct msghdr *msg, | |
dc6b9b78 ED |
881 | size_t len, int noblock, int flags, |
882 | int *addr_len); | |
1da177e4 LT |
883 | int (*sendpage)(struct sock *sk, struct page *page, |
884 | int offset, size_t size, int flags); | |
dc6b9b78 | 885 | int (*bind)(struct sock *sk, |
1da177e4 LT |
886 | struct sockaddr *uaddr, int addr_len); |
887 | ||
dc6b9b78 | 888 | int (*backlog_rcv) (struct sock *sk, |
1da177e4 LT |
889 | struct sk_buff *skb); |
890 | ||
46d3ceab | 891 | void (*release_cb)(struct sock *sk); |
563d34d0 | 892 | void (*mtu_reduced)(struct sock *sk); |
46d3ceab | 893 | |
1da177e4 LT |
894 | /* Keeping track of sk's, looking them up, and port selection methods. */ |
895 | void (*hash)(struct sock *sk); | |
896 | void (*unhash)(struct sock *sk); | |
719f8358 | 897 | void (*rehash)(struct sock *sk); |
1da177e4 | 898 | int (*get_port)(struct sock *sk, unsigned short snum); |
fcbdf09d | 899 | void (*clear_sk)(struct sock *sk, int size); |
1da177e4 | 900 | |
286ab3d4 | 901 | /* Keeping track of sockets in use */ |
65f76517 | 902 | #ifdef CONFIG_PROC_FS |
13ff3d6f | 903 | unsigned int inuse_idx; |
65f76517 | 904 | #endif |
ebb53d75 | 905 | |
1da177e4 | 906 | /* Memory pressure */ |
5c52ba17 | 907 | void (*enter_memory_pressure)(struct sock *sk); |
8d987e5c | 908 | atomic_long_t *memory_allocated; /* Current allocated memory. */ |
1748376b | 909 | struct percpu_counter *sockets_allocated; /* Current number of sockets. */ |
1da177e4 LT |
910 | /* |
911 | * Pressure flag: try to collapse. | |
912 | * Technical note: it is used by multiple contexts non atomically. | |
3ab224be | 913 | * All the __sk_mem_schedule() is of this nature: accounting |
1da177e4 LT |
914 | * is strict, actions are advisory and have some latency. |
915 | */ | |
916 | int *memory_pressure; | |
8d987e5c | 917 | long *sysctl_mem; |
1da177e4 LT |
918 | int *sysctl_wmem; |
919 | int *sysctl_rmem; | |
920 | int max_header; | |
7ba42910 | 921 | bool no_autobind; |
1da177e4 | 922 | |
271b72c7 | 923 | struct kmem_cache *slab; |
1da177e4 | 924 | unsigned int obj_size; |
271b72c7 | 925 | int slab_flags; |
1da177e4 | 926 | |
dd24c001 | 927 | struct percpu_counter *orphan_count; |
8feaf0c0 | 928 | |
60236fdd | 929 | struct request_sock_ops *rsk_prot; |
6d6ee43e | 930 | struct timewait_sock_ops *twsk_prot; |
2e6599cb | 931 | |
39d8cda7 PE |
932 | union { |
933 | struct inet_hashinfo *hashinfo; | |
645ca708 | 934 | struct udp_table *udp_table; |
fc8717ba | 935 | struct raw_hashinfo *raw_hash; |
39d8cda7 | 936 | } h; |
ab1e0a13 | 937 | |
1da177e4 LT |
938 | struct module *owner; |
939 | ||
940 | char name[32]; | |
941 | ||
942 | struct list_head node; | |
e6848976 ACM |
943 | #ifdef SOCK_REFCNT_DEBUG |
944 | atomic_t socks; | |
945 | #endif | |
c255a458 | 946 | #ifdef CONFIG_MEMCG_KMEM |
e1aab161 GC |
947 | /* |
948 | * cgroup specific init/deinit functions. Called once for all | |
949 | * protocols that implement it, from cgroups populate function. | |
950 | * This function has to setup any files the protocol want to | |
951 | * appear in the kmem cgroup filesystem. | |
952 | */ | |
1d62e436 | 953 | int (*init_cgroup)(struct mem_cgroup *memcg, |
e1aab161 | 954 | struct cgroup_subsys *ss); |
1d62e436 | 955 | void (*destroy_cgroup)(struct mem_cgroup *memcg); |
e1aab161 GC |
956 | struct cg_proto *(*proto_cgroup)(struct mem_cgroup *memcg); |
957 | #endif | |
958 | }; | |
959 | ||
3f134619 GC |
960 | /* |
961 | * Bits in struct cg_proto.flags | |
962 | */ | |
963 | enum cg_proto_flags { | |
964 | /* Currently active and new sockets should be assigned to cgroups */ | |
965 | MEMCG_SOCK_ACTIVE, | |
966 | /* It was ever activated; we must disarm static keys on destruction */ | |
967 | MEMCG_SOCK_ACTIVATED, | |
968 | }; | |
969 | ||
e1aab161 GC |
970 | struct cg_proto { |
971 | void (*enter_memory_pressure)(struct sock *sk); | |
972 | struct res_counter *memory_allocated; /* Current allocated memory. */ | |
973 | struct percpu_counter *sockets_allocated; /* Current number of sockets. */ | |
974 | int *memory_pressure; | |
975 | long *sysctl_mem; | |
3f134619 | 976 | unsigned long flags; |
e1aab161 GC |
977 | /* |
978 | * memcg field is used to find which memcg we belong directly | |
979 | * Each memcg struct can hold more than one cg_proto, so container_of | |
980 | * won't really cut. | |
981 | * | |
982 | * The elegant solution would be having an inverse function to | |
983 | * proto_cgroup in struct proto, but that means polluting the structure | |
984 | * for everybody, instead of just for memcg users. | |
985 | */ | |
986 | struct mem_cgroup *memcg; | |
1da177e4 LT |
987 | }; |
988 | ||
989 | extern int proto_register(struct proto *prot, int alloc_slab); | |
990 | extern void proto_unregister(struct proto *prot); | |
991 | ||
3f134619 GC |
992 | static inline bool memcg_proto_active(struct cg_proto *cg_proto) |
993 | { | |
994 | return test_bit(MEMCG_SOCK_ACTIVE, &cg_proto->flags); | |
995 | } | |
996 | ||
997 | static inline bool memcg_proto_activated(struct cg_proto *cg_proto) | |
998 | { | |
999 | return test_bit(MEMCG_SOCK_ACTIVATED, &cg_proto->flags); | |
1000 | } | |
1001 | ||
e6848976 ACM |
1002 | #ifdef SOCK_REFCNT_DEBUG |
1003 | static inline void sk_refcnt_debug_inc(struct sock *sk) | |
1004 | { | |
1005 | atomic_inc(&sk->sk_prot->socks); | |
1006 | } | |
1007 | ||
1008 | static inline void sk_refcnt_debug_dec(struct sock *sk) | |
1009 | { | |
1010 | atomic_dec(&sk->sk_prot->socks); | |
1011 | printk(KERN_DEBUG "%s socket %p released, %d are still alive\n", | |
1012 | sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks)); | |
1013 | } | |
1014 | ||
e1aab161 | 1015 | inline void sk_refcnt_debug_release(const struct sock *sk) |
e6848976 ACM |
1016 | { |
1017 | if (atomic_read(&sk->sk_refcnt) != 1) | |
1018 | printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n", | |
1019 | sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt)); | |
1020 | } | |
1021 | #else /* SOCK_REFCNT_DEBUG */ | |
1022 | #define sk_refcnt_debug_inc(sk) do { } while (0) | |
1023 | #define sk_refcnt_debug_dec(sk) do { } while (0) | |
1024 | #define sk_refcnt_debug_release(sk) do { } while (0) | |
1025 | #endif /* SOCK_REFCNT_DEBUG */ | |
1026 | ||
c255a458 | 1027 | #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_NET) |
c5905afb | 1028 | extern struct static_key memcg_socket_limit_enabled; |
e1aab161 GC |
1029 | static inline struct cg_proto *parent_cg_proto(struct proto *proto, |
1030 | struct cg_proto *cg_proto) | |
1031 | { | |
1032 | return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg)); | |
1033 | } | |
c5905afb | 1034 | #define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled) |
e1aab161 GC |
1035 | #else |
1036 | #define mem_cgroup_sockets_enabled 0 | |
1037 | static inline struct cg_proto *parent_cg_proto(struct proto *proto, | |
1038 | struct cg_proto *cg_proto) | |
1039 | { | |
1040 | return NULL; | |
1041 | } | |
1042 | #endif | |
1043 | ||
1044 | ||
180d8cd9 GC |
1045 | static inline bool sk_has_memory_pressure(const struct sock *sk) |
1046 | { | |
1047 | return sk->sk_prot->memory_pressure != NULL; | |
1048 | } | |
1049 | ||
1050 | static inline bool sk_under_memory_pressure(const struct sock *sk) | |
1051 | { | |
1052 | if (!sk->sk_prot->memory_pressure) | |
1053 | return false; | |
e1aab161 GC |
1054 | |
1055 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) | |
1056 | return !!*sk->sk_cgrp->memory_pressure; | |
1057 | ||
180d8cd9 GC |
1058 | return !!*sk->sk_prot->memory_pressure; |
1059 | } | |
1060 | ||
1061 | static inline void sk_leave_memory_pressure(struct sock *sk) | |
1062 | { | |
1063 | int *memory_pressure = sk->sk_prot->memory_pressure; | |
1064 | ||
e1aab161 GC |
1065 | if (!memory_pressure) |
1066 | return; | |
1067 | ||
1068 | if (*memory_pressure) | |
180d8cd9 | 1069 | *memory_pressure = 0; |
e1aab161 GC |
1070 | |
1071 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) { | |
1072 | struct cg_proto *cg_proto = sk->sk_cgrp; | |
1073 | struct proto *prot = sk->sk_prot; | |
1074 | ||
1075 | for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto)) | |
1076 | if (*cg_proto->memory_pressure) | |
1077 | *cg_proto->memory_pressure = 0; | |
1078 | } | |
1079 | ||
180d8cd9 GC |
1080 | } |
1081 | ||
1082 | static inline void sk_enter_memory_pressure(struct sock *sk) | |
1083 | { | |
e1aab161 GC |
1084 | if (!sk->sk_prot->enter_memory_pressure) |
1085 | return; | |
1086 | ||
1087 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) { | |
1088 | struct cg_proto *cg_proto = sk->sk_cgrp; | |
1089 | struct proto *prot = sk->sk_prot; | |
1090 | ||
1091 | for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto)) | |
1092 | cg_proto->enter_memory_pressure(sk); | |
1093 | } | |
1094 | ||
1095 | sk->sk_prot->enter_memory_pressure(sk); | |
180d8cd9 GC |
1096 | } |
1097 | ||
1098 | static inline long sk_prot_mem_limits(const struct sock *sk, int index) | |
1099 | { | |
1100 | long *prot = sk->sk_prot->sysctl_mem; | |
e1aab161 GC |
1101 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) |
1102 | prot = sk->sk_cgrp->sysctl_mem; | |
180d8cd9 GC |
1103 | return prot[index]; |
1104 | } | |
1105 | ||
e1aab161 GC |
1106 | static inline void memcg_memory_allocated_add(struct cg_proto *prot, |
1107 | unsigned long amt, | |
1108 | int *parent_status) | |
1109 | { | |
1110 | struct res_counter *fail; | |
1111 | int ret; | |
1112 | ||
0e90b31f GC |
1113 | ret = res_counter_charge_nofail(prot->memory_allocated, |
1114 | amt << PAGE_SHIFT, &fail); | |
e1aab161 GC |
1115 | if (ret < 0) |
1116 | *parent_status = OVER_LIMIT; | |
1117 | } | |
1118 | ||
1119 | static inline void memcg_memory_allocated_sub(struct cg_proto *prot, | |
1120 | unsigned long amt) | |
1121 | { | |
1122 | res_counter_uncharge(prot->memory_allocated, amt << PAGE_SHIFT); | |
1123 | } | |
1124 | ||
1125 | static inline u64 memcg_memory_allocated_read(struct cg_proto *prot) | |
1126 | { | |
1127 | u64 ret; | |
1128 | ret = res_counter_read_u64(prot->memory_allocated, RES_USAGE); | |
1129 | return ret >> PAGE_SHIFT; | |
1130 | } | |
1131 | ||
180d8cd9 GC |
1132 | static inline long |
1133 | sk_memory_allocated(const struct sock *sk) | |
1134 | { | |
1135 | struct proto *prot = sk->sk_prot; | |
e1aab161 GC |
1136 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) |
1137 | return memcg_memory_allocated_read(sk->sk_cgrp); | |
1138 | ||
180d8cd9 GC |
1139 | return atomic_long_read(prot->memory_allocated); |
1140 | } | |
1141 | ||
1142 | static inline long | |
e1aab161 | 1143 | sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status) |
180d8cd9 GC |
1144 | { |
1145 | struct proto *prot = sk->sk_prot; | |
e1aab161 GC |
1146 | |
1147 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) { | |
1148 | memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status); | |
1149 | /* update the root cgroup regardless */ | |
1150 | atomic_long_add_return(amt, prot->memory_allocated); | |
1151 | return memcg_memory_allocated_read(sk->sk_cgrp); | |
1152 | } | |
1153 | ||
180d8cd9 GC |
1154 | return atomic_long_add_return(amt, prot->memory_allocated); |
1155 | } | |
1156 | ||
1157 | static inline void | |
0e90b31f | 1158 | sk_memory_allocated_sub(struct sock *sk, int amt) |
180d8cd9 GC |
1159 | { |
1160 | struct proto *prot = sk->sk_prot; | |
e1aab161 | 1161 | |
0e90b31f | 1162 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) |
e1aab161 GC |
1163 | memcg_memory_allocated_sub(sk->sk_cgrp, amt); |
1164 | ||
180d8cd9 GC |
1165 | atomic_long_sub(amt, prot->memory_allocated); |
1166 | } | |
1167 | ||
1168 | static inline void sk_sockets_allocated_dec(struct sock *sk) | |
1169 | { | |
1170 | struct proto *prot = sk->sk_prot; | |
e1aab161 GC |
1171 | |
1172 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) { | |
1173 | struct cg_proto *cg_proto = sk->sk_cgrp; | |
1174 | ||
1175 | for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto)) | |
1176 | percpu_counter_dec(cg_proto->sockets_allocated); | |
1177 | } | |
1178 | ||
180d8cd9 GC |
1179 | percpu_counter_dec(prot->sockets_allocated); |
1180 | } | |
1181 | ||
1182 | static inline void sk_sockets_allocated_inc(struct sock *sk) | |
1183 | { | |
1184 | struct proto *prot = sk->sk_prot; | |
e1aab161 GC |
1185 | |
1186 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) { | |
1187 | struct cg_proto *cg_proto = sk->sk_cgrp; | |
1188 | ||
1189 | for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto)) | |
1190 | percpu_counter_inc(cg_proto->sockets_allocated); | |
1191 | } | |
1192 | ||
180d8cd9 GC |
1193 | percpu_counter_inc(prot->sockets_allocated); |
1194 | } | |
1195 | ||
1196 | static inline int | |
1197 | sk_sockets_allocated_read_positive(struct sock *sk) | |
1198 | { | |
1199 | struct proto *prot = sk->sk_prot; | |
1200 | ||
e1aab161 | 1201 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) |
518fbf9c | 1202 | return percpu_counter_read_positive(sk->sk_cgrp->sockets_allocated); |
e1aab161 | 1203 | |
518fbf9c | 1204 | return percpu_counter_read_positive(prot->sockets_allocated); |
180d8cd9 GC |
1205 | } |
1206 | ||
1207 | static inline int | |
1208 | proto_sockets_allocated_sum_positive(struct proto *prot) | |
1209 | { | |
1210 | return percpu_counter_sum_positive(prot->sockets_allocated); | |
1211 | } | |
1212 | ||
1213 | static inline long | |
1214 | proto_memory_allocated(struct proto *prot) | |
1215 | { | |
1216 | return atomic_long_read(prot->memory_allocated); | |
1217 | } | |
1218 | ||
1219 | static inline bool | |
1220 | proto_memory_pressure(struct proto *prot) | |
1221 | { | |
1222 | if (!prot->memory_pressure) | |
1223 | return false; | |
1224 | return !!*prot->memory_pressure; | |
1225 | } | |
1226 | ||
65f76517 ED |
1227 | |
1228 | #ifdef CONFIG_PROC_FS | |
1da177e4 | 1229 | /* Called with local bh disabled */ |
c29a0bc4 PE |
1230 | extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc); |
1231 | extern int sock_prot_inuse_get(struct net *net, struct proto *proto); | |
65f76517 | 1232 | #else |
dc6b9b78 | 1233 | static inline void sock_prot_inuse_add(struct net *net, struct proto *prot, |
c29a0bc4 | 1234 | int inc) |
65f76517 ED |
1235 | { |
1236 | } | |
65f76517 ED |
1237 | #endif |
1238 | ||
1da177e4 | 1239 | |
614c6cb4 ACM |
1240 | /* With per-bucket locks this operation is not-atomic, so that |
1241 | * this version is not worse. | |
1242 | */ | |
1243 | static inline void __sk_prot_rehash(struct sock *sk) | |
1244 | { | |
1245 | sk->sk_prot->unhash(sk); | |
1246 | sk->sk_prot->hash(sk); | |
1247 | } | |
1248 | ||
fcbdf09d OP |
1249 | void sk_prot_clear_portaddr_nulls(struct sock *sk, int size); |
1250 | ||
1da177e4 LT |
1251 | /* About 10 seconds */ |
1252 | #define SOCK_DESTROY_TIME (10*HZ) | |
1253 | ||
1254 | /* Sockets 0-1023 can't be bound to unless you are superuser */ | |
1255 | #define PROT_SOCK 1024 | |
1256 | ||
1257 | #define SHUTDOWN_MASK 3 | |
1258 | #define RCV_SHUTDOWN 1 | |
1259 | #define SEND_SHUTDOWN 2 | |
1260 | ||
1261 | #define SOCK_SNDBUF_LOCK 1 | |
1262 | #define SOCK_RCVBUF_LOCK 2 | |
1263 | #define SOCK_BINDADDR_LOCK 4 | |
1264 | #define SOCK_BINDPORT_LOCK 8 | |
1265 | ||
1266 | /* sock_iocb: used to kick off async processing of socket ios */ | |
1267 | struct sock_iocb { | |
1268 | struct list_head list; | |
1269 | ||
1270 | int flags; | |
1271 | int size; | |
1272 | struct socket *sock; | |
1273 | struct sock *sk; | |
1274 | struct scm_cookie *scm; | |
1275 | struct msghdr *msg, async_msg; | |
1da177e4 LT |
1276 | struct kiocb *kiocb; |
1277 | }; | |
1278 | ||
1279 | static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb) | |
1280 | { | |
1281 | return (struct sock_iocb *)iocb->private; | |
1282 | } | |
1283 | ||
1284 | static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si) | |
1285 | { | |
1286 | return si->kiocb; | |
1287 | } | |
1288 | ||
1289 | struct socket_alloc { | |
1290 | struct socket socket; | |
1291 | struct inode vfs_inode; | |
1292 | }; | |
1293 | ||
1294 | static inline struct socket *SOCKET_I(struct inode *inode) | |
1295 | { | |
1296 | return &container_of(inode, struct socket_alloc, vfs_inode)->socket; | |
1297 | } | |
1298 | ||
1299 | static inline struct inode *SOCK_INODE(struct socket *socket) | |
1300 | { | |
1301 | return &container_of(socket, struct socket_alloc, socket)->vfs_inode; | |
1302 | } | |
1303 | ||
3ab224be HA |
1304 | /* |
1305 | * Functions for memory accounting | |
1306 | */ | |
1307 | extern int __sk_mem_schedule(struct sock *sk, int size, int kind); | |
1308 | extern void __sk_mem_reclaim(struct sock *sk); | |
1da177e4 | 1309 | |
3ab224be HA |
1310 | #define SK_MEM_QUANTUM ((int)PAGE_SIZE) |
1311 | #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM) | |
1312 | #define SK_MEM_SEND 0 | |
1313 | #define SK_MEM_RECV 1 | |
1da177e4 | 1314 | |
3ab224be | 1315 | static inline int sk_mem_pages(int amt) |
1da177e4 | 1316 | { |
3ab224be | 1317 | return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT; |
1da177e4 LT |
1318 | } |
1319 | ||
dc6b9b78 | 1320 | static inline bool sk_has_account(struct sock *sk) |
1da177e4 | 1321 | { |
3ab224be HA |
1322 | /* return true if protocol supports memory accounting */ |
1323 | return !!sk->sk_prot->memory_allocated; | |
1da177e4 LT |
1324 | } |
1325 | ||
dc6b9b78 | 1326 | static inline bool sk_wmem_schedule(struct sock *sk, int size) |
1da177e4 | 1327 | { |
3ab224be | 1328 | if (!sk_has_account(sk)) |
dc6b9b78 | 1329 | return true; |
3ab224be HA |
1330 | return size <= sk->sk_forward_alloc || |
1331 | __sk_mem_schedule(sk, size, SK_MEM_SEND); | |
1da177e4 LT |
1332 | } |
1333 | ||
c76562b6 | 1334 | static inline bool |
35c448a8 | 1335 | sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size) |
d80d99d6 | 1336 | { |
3ab224be | 1337 | if (!sk_has_account(sk)) |
dc6b9b78 | 1338 | return true; |
c76562b6 MG |
1339 | return size<= sk->sk_forward_alloc || |
1340 | __sk_mem_schedule(sk, size, SK_MEM_RECV) || | |
1341 | skb_pfmemalloc(skb); | |
3ab224be HA |
1342 | } |
1343 | ||
1344 | static inline void sk_mem_reclaim(struct sock *sk) | |
1345 | { | |
1346 | if (!sk_has_account(sk)) | |
1347 | return; | |
1348 | if (sk->sk_forward_alloc >= SK_MEM_QUANTUM) | |
1349 | __sk_mem_reclaim(sk); | |
1350 | } | |
1351 | ||
9993e7d3 DM |
1352 | static inline void sk_mem_reclaim_partial(struct sock *sk) |
1353 | { | |
1354 | if (!sk_has_account(sk)) | |
1355 | return; | |
1356 | if (sk->sk_forward_alloc > SK_MEM_QUANTUM) | |
1357 | __sk_mem_reclaim(sk); | |
1358 | } | |
1359 | ||
3ab224be HA |
1360 | static inline void sk_mem_charge(struct sock *sk, int size) |
1361 | { | |
1362 | if (!sk_has_account(sk)) | |
1363 | return; | |
1364 | sk->sk_forward_alloc -= size; | |
1365 | } | |
1366 | ||
1367 | static inline void sk_mem_uncharge(struct sock *sk, int size) | |
1368 | { | |
1369 | if (!sk_has_account(sk)) | |
1370 | return; | |
1371 | sk->sk_forward_alloc += size; | |
1372 | } | |
1373 | ||
1374 | static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb) | |
1375 | { | |
3ab224be HA |
1376 | sock_set_flag(sk, SOCK_QUEUE_SHRUNK); |
1377 | sk->sk_wmem_queued -= skb->truesize; | |
1378 | sk_mem_uncharge(sk, skb->truesize); | |
1379 | __kfree_skb(skb); | |
d80d99d6 HX |
1380 | } |
1381 | ||
1da177e4 LT |
1382 | /* Used by processes to "lock" a socket state, so that |
1383 | * interrupts and bottom half handlers won't change it | |
1384 | * from under us. It essentially blocks any incoming | |
1385 | * packets, so that we won't get any new data or any | |
1386 | * packets that change the state of the socket. | |
1387 | * | |
1388 | * While locked, BH processing will add new packets to | |
1389 | * the backlog queue. This queue is processed by the | |
1390 | * owner of the socket lock right before it is released. | |
1391 | * | |
1392 | * Since ~2.3.5 it is also exclusive sleep lock serializing | |
1393 | * accesses from user process context. | |
1394 | */ | |
d2e9117c | 1395 | #define sock_owned_by_user(sk) ((sk)->sk_lock.owned) |
1da177e4 | 1396 | |
ed07536e PZ |
1397 | /* |
1398 | * Macro so as to not evaluate some arguments when | |
1399 | * lockdep is not enabled. | |
1400 | * | |
1401 | * Mark both the sk_lock and the sk_lock.slock as a | |
1402 | * per-address-family lock class. | |
1403 | */ | |
dc6b9b78 | 1404 | #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \ |
ed07536e | 1405 | do { \ |
e8f6fbf6 | 1406 | sk->sk_lock.owned = 0; \ |
ed07536e PZ |
1407 | init_waitqueue_head(&sk->sk_lock.wq); \ |
1408 | spin_lock_init(&(sk)->sk_lock.slock); \ | |
1409 | debug_check_no_locks_freed((void *)&(sk)->sk_lock, \ | |
1410 | sizeof((sk)->sk_lock)); \ | |
1411 | lockdep_set_class_and_name(&(sk)->sk_lock.slock, \ | |
dc6b9b78 | 1412 | (skey), (sname)); \ |
ed07536e PZ |
1413 | lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \ |
1414 | } while (0) | |
1415 | ||
41380930 | 1416 | extern void lock_sock_nested(struct sock *sk, int subclass); |
fcc70d5f PZ |
1417 | |
1418 | static inline void lock_sock(struct sock *sk) | |
1419 | { | |
1420 | lock_sock_nested(sk, 0); | |
1421 | } | |
1422 | ||
41380930 | 1423 | extern void release_sock(struct sock *sk); |
1da177e4 LT |
1424 | |
1425 | /* BH context may only use the following locking interface. */ | |
1426 | #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock)) | |
c6366184 IM |
1427 | #define bh_lock_sock_nested(__sk) \ |
1428 | spin_lock_nested(&((__sk)->sk_lock.slock), \ | |
1429 | SINGLE_DEPTH_NESTING) | |
1da177e4 LT |
1430 | #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock)) |
1431 | ||
8a74ad60 ED |
1432 | extern bool lock_sock_fast(struct sock *sk); |
1433 | /** | |
1434 | * unlock_sock_fast - complement of lock_sock_fast | |
1435 | * @sk: socket | |
1436 | * @slow: slow mode | |
1437 | * | |
1438 | * fast unlock socket for user context. | |
1439 | * If slow mode is on, we call regular release_sock() | |
1440 | */ | |
1441 | static inline void unlock_sock_fast(struct sock *sk, bool slow) | |
4b0b72f7 | 1442 | { |
8a74ad60 ED |
1443 | if (slow) |
1444 | release_sock(sk); | |
1445 | else | |
1446 | spin_unlock_bh(&sk->sk_lock.slock); | |
4b0b72f7 ED |
1447 | } |
1448 | ||
4b0b72f7 | 1449 | |
1b8d7ae4 | 1450 | extern struct sock *sk_alloc(struct net *net, int family, |
dd0fc66f | 1451 | gfp_t priority, |
6257ff21 | 1452 | struct proto *prot); |
1da177e4 | 1453 | extern void sk_free(struct sock *sk); |
edf02087 | 1454 | extern void sk_release_kernel(struct sock *sk); |
e56c57d0 ED |
1455 | extern struct sock *sk_clone_lock(const struct sock *sk, |
1456 | const gfp_t priority); | |
1da177e4 LT |
1457 | |
1458 | extern struct sk_buff *sock_wmalloc(struct sock *sk, | |
1459 | unsigned long size, int force, | |
dd0fc66f | 1460 | gfp_t priority); |
1da177e4 LT |
1461 | extern struct sk_buff *sock_rmalloc(struct sock *sk, |
1462 | unsigned long size, int force, | |
dd0fc66f | 1463 | gfp_t priority); |
1da177e4 LT |
1464 | extern void sock_wfree(struct sk_buff *skb); |
1465 | extern void sock_rfree(struct sk_buff *skb); | |
41063e9d | 1466 | extern void sock_edemux(struct sk_buff *skb); |
1da177e4 LT |
1467 | |
1468 | extern int sock_setsockopt(struct socket *sock, int level, | |
1469 | int op, char __user *optval, | |
b7058842 | 1470 | unsigned int optlen); |
1da177e4 LT |
1471 | |
1472 | extern int sock_getsockopt(struct socket *sock, int level, | |
dc6b9b78 | 1473 | int op, char __user *optval, |
1da177e4 | 1474 | int __user *optlen); |
dc6b9b78 | 1475 | extern struct sk_buff *sock_alloc_send_skb(struct sock *sk, |
1da177e4 LT |
1476 | unsigned long size, |
1477 | int noblock, | |
1478 | int *errcode); | |
dc6b9b78 | 1479 | extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk, |
4cc7f68d HX |
1480 | unsigned long header_len, |
1481 | unsigned long data_len, | |
1482 | int noblock, | |
1483 | int *errcode); | |
86a76caf | 1484 | extern void *sock_kmalloc(struct sock *sk, int size, |
dd0fc66f | 1485 | gfp_t priority); |
1da177e4 LT |
1486 | extern void sock_kfree_s(struct sock *sk, void *mem, int size); |
1487 | extern void sk_send_sigurg(struct sock *sk); | |
1488 | ||
f8451725 HX |
1489 | #ifdef CONFIG_CGROUPS |
1490 | extern void sock_update_classid(struct sock *sk); | |
1491 | #else | |
1492 | static inline void sock_update_classid(struct sock *sk) | |
1493 | { | |
1494 | } | |
1495 | #endif | |
1496 | ||
1da177e4 LT |
1497 | /* |
1498 | * Functions to fill in entries in struct proto_ops when a protocol | |
1499 | * does not implement a particular function. | |
1500 | */ | |
dc6b9b78 | 1501 | extern int sock_no_bind(struct socket *, |
1da177e4 LT |
1502 | struct sockaddr *, int); |
1503 | extern int sock_no_connect(struct socket *, | |
1504 | struct sockaddr *, int, int); | |
1505 | extern int sock_no_socketpair(struct socket *, | |
1506 | struct socket *); | |
1507 | extern int sock_no_accept(struct socket *, | |
1508 | struct socket *, int); | |
1509 | extern int sock_no_getname(struct socket *, | |
1510 | struct sockaddr *, int *, int); | |
1511 | extern unsigned int sock_no_poll(struct file *, struct socket *, | |
1512 | struct poll_table_struct *); | |
1513 | extern int sock_no_ioctl(struct socket *, unsigned int, | |
1514 | unsigned long); | |
1515 | extern int sock_no_listen(struct socket *, int); | |
1516 | extern int sock_no_shutdown(struct socket *, int); | |
1517 | extern int sock_no_getsockopt(struct socket *, int , int, | |
1518 | char __user *, int __user *); | |
1519 | extern int sock_no_setsockopt(struct socket *, int, int, | |
b7058842 | 1520 | char __user *, unsigned int); |
1da177e4 LT |
1521 | extern int sock_no_sendmsg(struct kiocb *, struct socket *, |
1522 | struct msghdr *, size_t); | |
1523 | extern int sock_no_recvmsg(struct kiocb *, struct socket *, | |
1524 | struct msghdr *, size_t, int); | |
1525 | extern int sock_no_mmap(struct file *file, | |
1526 | struct socket *sock, | |
1527 | struct vm_area_struct *vma); | |
1528 | extern ssize_t sock_no_sendpage(struct socket *sock, | |
1529 | struct page *page, | |
dc6b9b78 | 1530 | int offset, size_t size, |
1da177e4 LT |
1531 | int flags); |
1532 | ||
1533 | /* | |
1534 | * Functions to fill in entries in struct proto_ops when a protocol | |
1535 | * uses the inet style. | |
1536 | */ | |
1537 | extern int sock_common_getsockopt(struct socket *sock, int level, int optname, | |
1538 | char __user *optval, int __user *optlen); | |
1539 | extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock, | |
1540 | struct msghdr *msg, size_t size, int flags); | |
1541 | extern int sock_common_setsockopt(struct socket *sock, int level, int optname, | |
b7058842 | 1542 | char __user *optval, unsigned int optlen); |
3fdadf7d DM |
1543 | extern int compat_sock_common_getsockopt(struct socket *sock, int level, |
1544 | int optname, char __user *optval, int __user *optlen); | |
1545 | extern int compat_sock_common_setsockopt(struct socket *sock, int level, | |
b7058842 | 1546 | int optname, char __user *optval, unsigned int optlen); |
1da177e4 LT |
1547 | |
1548 | extern void sk_common_release(struct sock *sk); | |
1549 | ||
1550 | /* | |
1551 | * Default socket callbacks and setup code | |
1552 | */ | |
dc6b9b78 | 1553 | |
1da177e4 LT |
1554 | /* Initialise core socket variables */ |
1555 | extern void sock_init_data(struct socket *sock, struct sock *sk); | |
1556 | ||
46bcf14f ED |
1557 | extern void sk_filter_release_rcu(struct rcu_head *rcu); |
1558 | ||
dc9b3346 | 1559 | /** |
1a5778aa | 1560 | * sk_filter_release - release a socket filter |
dc9b3346 PB |
1561 | * @fp: filter to remove |
1562 | * | |
1563 | * Remove a filter from a socket and release its resources. | |
1564 | */ | |
1565 | ||
309dd5fc PE |
1566 | static inline void sk_filter_release(struct sk_filter *fp) |
1567 | { | |
1568 | if (atomic_dec_and_test(&fp->refcnt)) | |
80f8f102 | 1569 | call_rcu(&fp->rcu, sk_filter_release_rcu); |
309dd5fc PE |
1570 | } |
1571 | ||
1572 | static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp) | |
1da177e4 LT |
1573 | { |
1574 | unsigned int size = sk_filter_len(fp); | |
1575 | ||
1576 | atomic_sub(size, &sk->sk_omem_alloc); | |
309dd5fc | 1577 | sk_filter_release(fp); |
1da177e4 LT |
1578 | } |
1579 | ||
1580 | static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp) | |
1581 | { | |
1582 | atomic_inc(&fp->refcnt); | |
1583 | atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc); | |
1584 | } | |
1585 | ||
1586 | /* | |
1587 | * Socket reference counting postulates. | |
1588 | * | |
1589 | * * Each user of socket SHOULD hold a reference count. | |
1590 | * * Each access point to socket (an hash table bucket, reference from a list, | |
1591 | * running timer, skb in flight MUST hold a reference count. | |
1592 | * * When reference count hits 0, it means it will never increase back. | |
1593 | * * When reference count hits 0, it means that no references from | |
1594 | * outside exist to this socket and current process on current CPU | |
1595 | * is last user and may/should destroy this socket. | |
1596 | * * sk_free is called from any context: process, BH, IRQ. When | |
1597 | * it is called, socket has no references from outside -> sk_free | |
1598 | * may release descendant resources allocated by the socket, but | |
1599 | * to the time when it is called, socket is NOT referenced by any | |
1600 | * hash tables, lists etc. | |
1601 | * * Packets, delivered from outside (from network or from another process) | |
1602 | * and enqueued on receive/error queues SHOULD NOT grab reference count, | |
1603 | * when they sit in queue. Otherwise, packets will leak to hole, when | |
1604 | * socket is looked up by one cpu and unhasing is made by another CPU. | |
1605 | * It is true for udp/raw, netlink (leak to receive and error queues), tcp | |
1606 | * (leak to backlog). Packet socket does all the processing inside | |
1607 | * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets | |
1608 | * use separate SMP lock, so that they are prone too. | |
1609 | */ | |
1610 | ||
1611 | /* Ungrab socket and destroy it, if it was the last reference. */ | |
1612 | static inline void sock_put(struct sock *sk) | |
1613 | { | |
1614 | if (atomic_dec_and_test(&sk->sk_refcnt)) | |
1615 | sk_free(sk); | |
1616 | } | |
1617 | ||
58a5a7b9 ACM |
1618 | extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb, |
1619 | const int nested); | |
25995ff5 | 1620 | |
e022f0b4 KK |
1621 | static inline void sk_tx_queue_set(struct sock *sk, int tx_queue) |
1622 | { | |
1623 | sk->sk_tx_queue_mapping = tx_queue; | |
1624 | } | |
1625 | ||
1626 | static inline void sk_tx_queue_clear(struct sock *sk) | |
1627 | { | |
1628 | sk->sk_tx_queue_mapping = -1; | |
1629 | } | |
1630 | ||
1631 | static inline int sk_tx_queue_get(const struct sock *sk) | |
1632 | { | |
b0f77d0e | 1633 | return sk ? sk->sk_tx_queue_mapping : -1; |
e022f0b4 KK |
1634 | } |
1635 | ||
972692e0 DM |
1636 | static inline void sk_set_socket(struct sock *sk, struct socket *sock) |
1637 | { | |
e022f0b4 | 1638 | sk_tx_queue_clear(sk); |
972692e0 DM |
1639 | sk->sk_socket = sock; |
1640 | } | |
1641 | ||
aa395145 ED |
1642 | static inline wait_queue_head_t *sk_sleep(struct sock *sk) |
1643 | { | |
eaefd110 ED |
1644 | BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0); |
1645 | return &rcu_dereference_raw(sk->sk_wq)->wait; | |
aa395145 | 1646 | } |
1da177e4 LT |
1647 | /* Detach socket from process context. |
1648 | * Announce socket dead, detach it from wait queue and inode. | |
1649 | * Note that parent inode held reference count on this struct sock, | |
1650 | * we do not release it in this function, because protocol | |
1651 | * probably wants some additional cleanups or even continuing | |
1652 | * to work with this socket (TCP). | |
1653 | */ | |
1654 | static inline void sock_orphan(struct sock *sk) | |
1655 | { | |
1656 | write_lock_bh(&sk->sk_callback_lock); | |
1657 | sock_set_flag(sk, SOCK_DEAD); | |
972692e0 | 1658 | sk_set_socket(sk, NULL); |
43815482 | 1659 | sk->sk_wq = NULL; |
1da177e4 LT |
1660 | write_unlock_bh(&sk->sk_callback_lock); |
1661 | } | |
1662 | ||
1663 | static inline void sock_graft(struct sock *sk, struct socket *parent) | |
1664 | { | |
1665 | write_lock_bh(&sk->sk_callback_lock); | |
eaefd110 | 1666 | sk->sk_wq = parent->wq; |
1da177e4 | 1667 | parent->sk = sk; |
972692e0 | 1668 | sk_set_socket(sk, parent); |
4237c75c | 1669 | security_sock_graft(sk, parent); |
1da177e4 LT |
1670 | write_unlock_bh(&sk->sk_callback_lock); |
1671 | } | |
1672 | ||
1673 | extern int sock_i_uid(struct sock *sk); | |
1674 | extern unsigned long sock_i_ino(struct sock *sk); | |
1675 | ||
1676 | static inline struct dst_entry * | |
1677 | __sk_dst_get(struct sock *sk) | |
1678 | { | |
d8bf4ca9 | 1679 | return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) || |
f68c224f | 1680 | lockdep_is_held(&sk->sk_lock.slock)); |
1da177e4 LT |
1681 | } |
1682 | ||
1683 | static inline struct dst_entry * | |
1684 | sk_dst_get(struct sock *sk) | |
1685 | { | |
1686 | struct dst_entry *dst; | |
1687 | ||
b6c6712a ED |
1688 | rcu_read_lock(); |
1689 | dst = rcu_dereference(sk->sk_dst_cache); | |
1da177e4 LT |
1690 | if (dst) |
1691 | dst_hold(dst); | |
b6c6712a | 1692 | rcu_read_unlock(); |
1da177e4 LT |
1693 | return dst; |
1694 | } | |
1695 | ||
b6c6712a ED |
1696 | extern void sk_reset_txq(struct sock *sk); |
1697 | ||
1698 | static inline void dst_negative_advice(struct sock *sk) | |
1699 | { | |
1700 | struct dst_entry *ndst, *dst = __sk_dst_get(sk); | |
1701 | ||
1702 | if (dst && dst->ops->negative_advice) { | |
1703 | ndst = dst->ops->negative_advice(dst); | |
1704 | ||
1705 | if (ndst != dst) { | |
1706 | rcu_assign_pointer(sk->sk_dst_cache, ndst); | |
1707 | sk_reset_txq(sk); | |
1708 | } | |
1709 | } | |
1710 | } | |
1711 | ||
1da177e4 LT |
1712 | static inline void |
1713 | __sk_dst_set(struct sock *sk, struct dst_entry *dst) | |
1714 | { | |
1715 | struct dst_entry *old_dst; | |
1716 | ||
e022f0b4 | 1717 | sk_tx_queue_clear(sk); |
0b53ff2e ED |
1718 | /* |
1719 | * This can be called while sk is owned by the caller only, | |
1720 | * with no state that can be checked in a rcu_dereference_check() cond | |
1721 | */ | |
1722 | old_dst = rcu_dereference_raw(sk->sk_dst_cache); | |
b6c6712a | 1723 | rcu_assign_pointer(sk->sk_dst_cache, dst); |
1da177e4 LT |
1724 | dst_release(old_dst); |
1725 | } | |
1726 | ||
1727 | static inline void | |
1728 | sk_dst_set(struct sock *sk, struct dst_entry *dst) | |
1729 | { | |
b6c6712a | 1730 | spin_lock(&sk->sk_dst_lock); |
1da177e4 | 1731 | __sk_dst_set(sk, dst); |
b6c6712a | 1732 | spin_unlock(&sk->sk_dst_lock); |
1da177e4 LT |
1733 | } |
1734 | ||
1735 | static inline void | |
1736 | __sk_dst_reset(struct sock *sk) | |
1737 | { | |
b6c6712a | 1738 | __sk_dst_set(sk, NULL); |
1da177e4 LT |
1739 | } |
1740 | ||
1741 | static inline void | |
1742 | sk_dst_reset(struct sock *sk) | |
1743 | { | |
b6c6712a | 1744 | spin_lock(&sk->sk_dst_lock); |
1da177e4 | 1745 | __sk_dst_reset(sk); |
b6c6712a | 1746 | spin_unlock(&sk->sk_dst_lock); |
1da177e4 LT |
1747 | } |
1748 | ||
f0088a50 | 1749 | extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie); |
1da177e4 | 1750 | |
f0088a50 | 1751 | extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie); |
1da177e4 | 1752 | |
dc6b9b78 | 1753 | static inline bool sk_can_gso(const struct sock *sk) |
bcd76111 HX |
1754 | { |
1755 | return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type); | |
1756 | } | |
1757 | ||
9958089a | 1758 | extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst); |
6cbb0df7 | 1759 | |
c8f44aff | 1760 | static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags) |
a465419b ED |
1761 | { |
1762 | sk->sk_route_nocaps |= flags; | |
1763 | sk->sk_route_caps &= ~flags; | |
1764 | } | |
1765 | ||
c6e1a0d1 TH |
1766 | static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb, |
1767 | char __user *from, char *to, | |
912d398d | 1768 | int copy, int offset) |
c6e1a0d1 TH |
1769 | { |
1770 | if (skb->ip_summed == CHECKSUM_NONE) { | |
1771 | int err = 0; | |
1772 | __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err); | |
1773 | if (err) | |
1774 | return err; | |
912d398d | 1775 | skb->csum = csum_block_add(skb->csum, csum, offset); |
c6e1a0d1 TH |
1776 | } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) { |
1777 | if (!access_ok(VERIFY_READ, from, copy) || | |
1778 | __copy_from_user_nocache(to, from, copy)) | |
1779 | return -EFAULT; | |
1780 | } else if (copy_from_user(to, from, copy)) | |
1781 | return -EFAULT; | |
1782 | ||
1783 | return 0; | |
1784 | } | |
1785 | ||
1786 | static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb, | |
1787 | char __user *from, int copy) | |
1788 | { | |
912d398d | 1789 | int err, offset = skb->len; |
c6e1a0d1 | 1790 | |
912d398d WY |
1791 | err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy), |
1792 | copy, offset); | |
c6e1a0d1 | 1793 | if (err) |
912d398d | 1794 | __skb_trim(skb, offset); |
c6e1a0d1 TH |
1795 | |
1796 | return err; | |
1797 | } | |
1798 | ||
1799 | static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from, | |
1800 | struct sk_buff *skb, | |
1801 | struct page *page, | |
1802 | int off, int copy) | |
1803 | { | |
1804 | int err; | |
1805 | ||
912d398d WY |
1806 | err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off, |
1807 | copy, skb->len); | |
c6e1a0d1 TH |
1808 | if (err) |
1809 | return err; | |
1810 | ||
1811 | skb->len += copy; | |
1812 | skb->data_len += copy; | |
1813 | skb->truesize += copy; | |
1814 | sk->sk_wmem_queued += copy; | |
1815 | sk_mem_charge(sk, copy); | |
1816 | return 0; | |
1817 | } | |
1818 | ||
1da177e4 LT |
1819 | static inline int skb_copy_to_page(struct sock *sk, char __user *from, |
1820 | struct sk_buff *skb, struct page *page, | |
1821 | int off, int copy) | |
1822 | { | |
1823 | if (skb->ip_summed == CHECKSUM_NONE) { | |
1824 | int err = 0; | |
5084205f | 1825 | __wsum csum = csum_and_copy_from_user(from, |
1da177e4 LT |
1826 | page_address(page) + off, |
1827 | copy, 0, &err); | |
1828 | if (err) | |
1829 | return err; | |
1830 | skb->csum = csum_block_add(skb->csum, csum, skb->len); | |
1831 | } else if (copy_from_user(page_address(page) + off, from, copy)) | |
1832 | return -EFAULT; | |
1833 | ||
1834 | skb->len += copy; | |
1835 | skb->data_len += copy; | |
1836 | skb->truesize += copy; | |
1837 | sk->sk_wmem_queued += copy; | |
3ab224be | 1838 | sk_mem_charge(sk, copy); |
1da177e4 LT |
1839 | return 0; |
1840 | } | |
1841 | ||
c564039f ED |
1842 | /** |
1843 | * sk_wmem_alloc_get - returns write allocations | |
1844 | * @sk: socket | |
1845 | * | |
1846 | * Returns sk_wmem_alloc minus initial offset of one | |
1847 | */ | |
1848 | static inline int sk_wmem_alloc_get(const struct sock *sk) | |
1849 | { | |
1850 | return atomic_read(&sk->sk_wmem_alloc) - 1; | |
1851 | } | |
1852 | ||
1853 | /** | |
1854 | * sk_rmem_alloc_get - returns read allocations | |
1855 | * @sk: socket | |
1856 | * | |
1857 | * Returns sk_rmem_alloc | |
1858 | */ | |
1859 | static inline int sk_rmem_alloc_get(const struct sock *sk) | |
1860 | { | |
1861 | return atomic_read(&sk->sk_rmem_alloc); | |
1862 | } | |
1863 | ||
1864 | /** | |
1865 | * sk_has_allocations - check if allocations are outstanding | |
1866 | * @sk: socket | |
1867 | * | |
1868 | * Returns true if socket has write or read allocations | |
1869 | */ | |
dc6b9b78 | 1870 | static inline bool sk_has_allocations(const struct sock *sk) |
c564039f ED |
1871 | { |
1872 | return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk); | |
1873 | } | |
1874 | ||
a57de0b4 | 1875 | /** |
43815482 | 1876 | * wq_has_sleeper - check if there are any waiting processes |
acfbe96a | 1877 | * @wq: struct socket_wq |
a57de0b4 | 1878 | * |
43815482 | 1879 | * Returns true if socket_wq has waiting processes |
a57de0b4 | 1880 | * |
43815482 | 1881 | * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory |
a57de0b4 JO |
1882 | * barrier call. They were added due to the race found within the tcp code. |
1883 | * | |
1884 | * Consider following tcp code paths: | |
1885 | * | |
1886 | * CPU1 CPU2 | |
1887 | * | |
1888 | * sys_select receive packet | |
1889 | * ... ... | |
1890 | * __add_wait_queue update tp->rcv_nxt | |
1891 | * ... ... | |
1892 | * tp->rcv_nxt check sock_def_readable | |
1893 | * ... { | |
43815482 ED |
1894 | * schedule rcu_read_lock(); |
1895 | * wq = rcu_dereference(sk->sk_wq); | |
1896 | * if (wq && waitqueue_active(&wq->wait)) | |
1897 | * wake_up_interruptible(&wq->wait) | |
a57de0b4 JO |
1898 | * ... |
1899 | * } | |
1900 | * | |
1901 | * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay | |
1902 | * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1 | |
1903 | * could then endup calling schedule and sleep forever if there are no more | |
1904 | * data on the socket. | |
ad462769 | 1905 | * |
a57de0b4 | 1906 | */ |
43815482 | 1907 | static inline bool wq_has_sleeper(struct socket_wq *wq) |
a57de0b4 | 1908 | { |
dc6b9b78 | 1909 | /* We need to be sure we are in sync with the |
a57de0b4 JO |
1910 | * add_wait_queue modifications to the wait queue. |
1911 | * | |
1912 | * This memory barrier is paired in the sock_poll_wait. | |
1913 | */ | |
43815482 ED |
1914 | smp_mb(); |
1915 | return wq && waitqueue_active(&wq->wait); | |
a57de0b4 JO |
1916 | } |
1917 | ||
1918 | /** | |
1919 | * sock_poll_wait - place memory barrier behind the poll_wait call. | |
1920 | * @filp: file | |
1921 | * @wait_address: socket wait queue | |
1922 | * @p: poll_table | |
1923 | * | |
43815482 | 1924 | * See the comments in the wq_has_sleeper function. |
a57de0b4 JO |
1925 | */ |
1926 | static inline void sock_poll_wait(struct file *filp, | |
1927 | wait_queue_head_t *wait_address, poll_table *p) | |
1928 | { | |
626cf236 | 1929 | if (!poll_does_not_wait(p) && wait_address) { |
a57de0b4 | 1930 | poll_wait(filp, wait_address, p); |
dc6b9b78 | 1931 | /* We need to be sure we are in sync with the |
a57de0b4 JO |
1932 | * socket flags modification. |
1933 | * | |
43815482 | 1934 | * This memory barrier is paired in the wq_has_sleeper. |
dc6b9b78 | 1935 | */ |
a57de0b4 JO |
1936 | smp_mb(); |
1937 | } | |
1938 | } | |
1939 | ||
1da177e4 | 1940 | /* |
dc6b9b78 | 1941 | * Queue a received datagram if it will fit. Stream and sequenced |
1da177e4 LT |
1942 | * protocols can't normally use this as they need to fit buffers in |
1943 | * and play with them. | |
1944 | * | |
dc6b9b78 | 1945 | * Inlined as it's very short and called for pretty much every |
1da177e4 LT |
1946 | * packet ever received. |
1947 | */ | |
1948 | ||
1949 | static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk) | |
1950 | { | |
d55d87fd | 1951 | skb_orphan(skb); |
1da177e4 LT |
1952 | skb->sk = sk; |
1953 | skb->destructor = sock_wfree; | |
2b85a34e ED |
1954 | /* |
1955 | * We used to take a refcount on sk, but following operation | |
1956 | * is enough to guarantee sk_free() wont free this sock until | |
1957 | * all in-flight packets are completed | |
1958 | */ | |
1da177e4 LT |
1959 | atomic_add(skb->truesize, &sk->sk_wmem_alloc); |
1960 | } | |
1961 | ||
1962 | static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk) | |
1963 | { | |
d55d87fd | 1964 | skb_orphan(skb); |
1da177e4 LT |
1965 | skb->sk = sk; |
1966 | skb->destructor = sock_rfree; | |
1967 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
3ab224be | 1968 | sk_mem_charge(sk, skb->truesize); |
1da177e4 LT |
1969 | } |
1970 | ||
dc6b9b78 | 1971 | extern void sk_reset_timer(struct sock *sk, struct timer_list *timer, |
1da177e4 LT |
1972 | unsigned long expires); |
1973 | ||
dc6b9b78 | 1974 | extern void sk_stop_timer(struct sock *sk, struct timer_list *timer); |
1da177e4 | 1975 | |
f0088a50 | 1976 | extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb); |
1da177e4 | 1977 | |
b1faf566 | 1978 | extern int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb); |
1da177e4 LT |
1979 | |
1980 | /* | |
1981 | * Recover an error report and clear atomically | |
1982 | */ | |
dc6b9b78 | 1983 | |
1da177e4 LT |
1984 | static inline int sock_error(struct sock *sk) |
1985 | { | |
c1cbe4b7 BL |
1986 | int err; |
1987 | if (likely(!sk->sk_err)) | |
1988 | return 0; | |
1989 | err = xchg(&sk->sk_err, 0); | |
1da177e4 LT |
1990 | return -err; |
1991 | } | |
1992 | ||
1993 | static inline unsigned long sock_wspace(struct sock *sk) | |
1994 | { | |
1995 | int amt = 0; | |
1996 | ||
1997 | if (!(sk->sk_shutdown & SEND_SHUTDOWN)) { | |
1998 | amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc); | |
dc6b9b78 | 1999 | if (amt < 0) |
1da177e4 LT |
2000 | amt = 0; |
2001 | } | |
2002 | return amt; | |
2003 | } | |
2004 | ||
2005 | static inline void sk_wake_async(struct sock *sk, int how, int band) | |
2006 | { | |
bcdce719 | 2007 | if (sock_flag(sk, SOCK_FASYNC)) |
1da177e4 LT |
2008 | sock_wake_async(sk->sk_socket, how, band); |
2009 | } | |
2010 | ||
2011 | #define SOCK_MIN_SNDBUF 2048 | |
7a91b434 ED |
2012 | /* |
2013 | * Since sk_rmem_alloc sums skb->truesize, even a small frame might need | |
2014 | * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak | |
2015 | */ | |
2016 | #define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff)) | |
1da177e4 LT |
2017 | |
2018 | static inline void sk_stream_moderate_sndbuf(struct sock *sk) | |
2019 | { | |
2020 | if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) { | |
8df09ea3 | 2021 | sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1); |
1da177e4 LT |
2022 | sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF); |
2023 | } | |
2024 | } | |
2025 | ||
df97c708 | 2026 | struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp); |
1da177e4 LT |
2027 | |
2028 | static inline struct page *sk_stream_alloc_page(struct sock *sk) | |
2029 | { | |
2030 | struct page *page = NULL; | |
2031 | ||
ef015786 HX |
2032 | page = alloc_pages(sk->sk_allocation, 0); |
2033 | if (!page) { | |
180d8cd9 | 2034 | sk_enter_memory_pressure(sk); |
1da177e4 LT |
2035 | sk_stream_moderate_sndbuf(sk); |
2036 | } | |
2037 | return page; | |
2038 | } | |
2039 | ||
1da177e4 LT |
2040 | /* |
2041 | * Default write policy as shown to user space via poll/select/SIGIO | |
2042 | */ | |
dc6b9b78 | 2043 | static inline bool sock_writeable(const struct sock *sk) |
1da177e4 | 2044 | { |
8df09ea3 | 2045 | return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1); |
1da177e4 LT |
2046 | } |
2047 | ||
dd0fc66f | 2048 | static inline gfp_t gfp_any(void) |
1da177e4 | 2049 | { |
99709372 | 2050 | return in_softirq() ? GFP_ATOMIC : GFP_KERNEL; |
1da177e4 LT |
2051 | } |
2052 | ||
dc6b9b78 | 2053 | static inline long sock_rcvtimeo(const struct sock *sk, bool noblock) |
1da177e4 LT |
2054 | { |
2055 | return noblock ? 0 : sk->sk_rcvtimeo; | |
2056 | } | |
2057 | ||
dc6b9b78 | 2058 | static inline long sock_sndtimeo(const struct sock *sk, bool noblock) |
1da177e4 LT |
2059 | { |
2060 | return noblock ? 0 : sk->sk_sndtimeo; | |
2061 | } | |
2062 | ||
2063 | static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len) | |
2064 | { | |
2065 | return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1; | |
2066 | } | |
2067 | ||
2068 | /* Alas, with timeout socket operations are not restartable. | |
2069 | * Compare this to poll(). | |
2070 | */ | |
2071 | static inline int sock_intr_errno(long timeo) | |
2072 | { | |
2073 | return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR; | |
2074 | } | |
2075 | ||
92f37fd2 ED |
2076 | extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, |
2077 | struct sk_buff *skb); | |
6e3e939f JB |
2078 | extern void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk, |
2079 | struct sk_buff *skb); | |
92f37fd2 | 2080 | |
dc6b9b78 | 2081 | static inline void |
1da177e4 LT |
2082 | sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb) |
2083 | { | |
b7aa0bf7 | 2084 | ktime_t kt = skb->tstamp; |
20d49473 | 2085 | struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb); |
a61bbcf2 | 2086 | |
20d49473 PO |
2087 | /* |
2088 | * generate control messages if | |
2089 | * - receive time stamping in software requested (SOCK_RCVTSTAMP | |
2090 | * or SOCK_TIMESTAMPING_RX_SOFTWARE) | |
2091 | * - software time stamp available and wanted | |
2092 | * (SOCK_TIMESTAMPING_SOFTWARE) | |
2093 | * - hardware time stamps available and wanted | |
2094 | * (SOCK_TIMESTAMPING_SYS_HARDWARE or | |
2095 | * SOCK_TIMESTAMPING_RAW_HARDWARE) | |
2096 | */ | |
2097 | if (sock_flag(sk, SOCK_RCVTSTAMP) || | |
2098 | sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) || | |
2099 | (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) || | |
2100 | (hwtstamps->hwtstamp.tv64 && | |
2101 | sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) || | |
2102 | (hwtstamps->syststamp.tv64 && | |
2103 | sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE))) | |
92f37fd2 ED |
2104 | __sock_recv_timestamp(msg, sk, skb); |
2105 | else | |
b7aa0bf7 | 2106 | sk->sk_stamp = kt; |
6e3e939f JB |
2107 | |
2108 | if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid) | |
2109 | __sock_recv_wifi_status(msg, sk, skb); | |
1da177e4 LT |
2110 | } |
2111 | ||
767dd033 ED |
2112 | extern void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, |
2113 | struct sk_buff *skb); | |
2114 | ||
2115 | static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, | |
2116 | struct sk_buff *skb) | |
2117 | { | |
2118 | #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \ | |
2119 | (1UL << SOCK_RCVTSTAMP) | \ | |
2120 | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \ | |
2121 | (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \ | |
dc6b9b78 | 2122 | (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \ |
767dd033 ED |
2123 | (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE)) |
2124 | ||
2125 | if (sk->sk_flags & FLAGS_TS_OR_DROPS) | |
2126 | __sock_recv_ts_and_drops(msg, sk, skb); | |
2127 | else | |
2128 | sk->sk_stamp = skb->tstamp; | |
2129 | } | |
3b885787 | 2130 | |
20d49473 PO |
2131 | /** |
2132 | * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped | |
20d49473 | 2133 | * @sk: socket sending this packet |
2244d07b | 2134 | * @tx_flags: filled with instructions for time stamping |
20d49473 PO |
2135 | * |
2136 | * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if | |
2137 | * parameters are invalid. | |
2138 | */ | |
2244d07b | 2139 | extern int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags); |
20d49473 | 2140 | |
1da177e4 LT |
2141 | /** |
2142 | * sk_eat_skb - Release a skb if it is no longer needed | |
4dc3b16b PP |
2143 | * @sk: socket to eat this skb from |
2144 | * @skb: socket buffer to eat | |
f4b8ea78 | 2145 | * @copied_early: flag indicating whether DMA operations copied this data early |
1da177e4 LT |
2146 | * |
2147 | * This routine must be called with interrupts disabled or with the socket | |
2148 | * locked so that the sk_buff queue operation is ok. | |
2149 | */ | |
624d1164 | 2150 | #ifdef CONFIG_NET_DMA |
dc6b9b78 | 2151 | static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early) |
624d1164 CL |
2152 | { |
2153 | __skb_unlink(skb, &sk->sk_receive_queue); | |
2154 | if (!copied_early) | |
2155 | __kfree_skb(skb); | |
2156 | else | |
2157 | __skb_queue_tail(&sk->sk_async_wait_queue, skb); | |
2158 | } | |
2159 | #else | |
dc6b9b78 | 2160 | static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early) |
1da177e4 LT |
2161 | { |
2162 | __skb_unlink(skb, &sk->sk_receive_queue); | |
2163 | __kfree_skb(skb); | |
2164 | } | |
624d1164 | 2165 | #endif |
1da177e4 | 2166 | |
3b1e0a65 YH |
2167 | static inline |
2168 | struct net *sock_net(const struct sock *sk) | |
2169 | { | |
c2d9ba9b | 2170 | return read_pnet(&sk->sk_net); |
3b1e0a65 YH |
2171 | } |
2172 | ||
2173 | static inline | |
f5aa23fd | 2174 | void sock_net_set(struct sock *sk, struct net *net) |
3b1e0a65 | 2175 | { |
c2d9ba9b | 2176 | write_pnet(&sk->sk_net, net); |
3b1e0a65 YH |
2177 | } |
2178 | ||
edf02087 DL |
2179 | /* |
2180 | * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace. | |
25985edc | 2181 | * They should not hold a reference to a namespace in order to allow |
edf02087 DL |
2182 | * to stop it. |
2183 | * Sockets after sk_change_net should be released using sk_release_kernel | |
2184 | */ | |
2185 | static inline void sk_change_net(struct sock *sk, struct net *net) | |
2186 | { | |
3b1e0a65 | 2187 | put_net(sock_net(sk)); |
65a18ec5 | 2188 | sock_net_set(sk, hold_net(net)); |
edf02087 DL |
2189 | } |
2190 | ||
23542618 KK |
2191 | static inline struct sock *skb_steal_sock(struct sk_buff *skb) |
2192 | { | |
efc27f8c | 2193 | if (skb->sk) { |
23542618 KK |
2194 | struct sock *sk = skb->sk; |
2195 | ||
2196 | skb->destructor = NULL; | |
2197 | skb->sk = NULL; | |
2198 | return sk; | |
2199 | } | |
2200 | return NULL; | |
2201 | } | |
2202 | ||
20d49473 | 2203 | extern void sock_enable_timestamp(struct sock *sk, int flag); |
1da177e4 | 2204 | extern int sock_get_timestamp(struct sock *, struct timeval __user *); |
ae40eb1e | 2205 | extern int sock_get_timestampns(struct sock *, struct timespec __user *); |
1da177e4 | 2206 | |
dc6b9b78 ED |
2207 | /* |
2208 | * Enable debug/info messages | |
1da177e4 | 2209 | */ |
a2a316fd SH |
2210 | extern int net_msg_warn; |
2211 | #define NETDEBUG(fmt, args...) \ | |
2212 | do { if (net_msg_warn) printk(fmt,##args); } while (0) | |
1da177e4 | 2213 | |
a2a316fd SH |
2214 | #define LIMIT_NETDEBUG(fmt, args...) \ |
2215 | do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0) | |
1da177e4 | 2216 | |
1da177e4 LT |
2217 | extern __u32 sysctl_wmem_max; |
2218 | extern __u32 sysctl_rmem_max; | |
2219 | ||
20380731 ACM |
2220 | extern void sk_init(void); |
2221 | ||
6baf1f41 DM |
2222 | extern int sysctl_optmem_max; |
2223 | ||
20380731 ACM |
2224 | extern __u32 sysctl_wmem_default; |
2225 | extern __u32 sysctl_rmem_default; | |
20380731 | 2226 | |
1da177e4 | 2227 | #endif /* _SOCK_H */ |