]> Git Repo - linux.git/blame - include/net/sock.h
sock: relax WARN_ON() in sock_owned_by_user()
[linux.git] / include / net / sock.h
CommitLineData
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, <[email protected]>
12 * Corey Minyard <[email protected]>
13 * Florian La Roche <[email protected]>
14 *
15 * Fixes:
16 * Alan Cox : Volatiles in skbuff pointers. See
17 * skbuff comments. May be overdone,
18 * better to prove they can be removed
19 * than the reverse.
20 * Alan Cox : Added a zapped field for tcp to note
21 * a socket is reset and must stay shut up
22 * Alan Cox : New fields for options
23 * Pauline Middelink : identd support
24 * Alan Cox : Eliminate low level recv/recvfrom
25 * David S. Miller : New socket lookup architecture.
26 * Steve Whitehouse: Default routines for sock_ops
27 * Arnaldo C. Melo : removed net_pinfo, tp_pinfo and made
28 * protinfo be just a void pointer, as the
29 * protocol specific parts were moved to
30 * respective headers and ipv4/v6, etc now
31 * use private slabcaches for its socks
32 * Pedro Hortas : New flags field for socket options
33 *
34 *
35 * This program is free software; you can redistribute it and/or
36 * modify it under the terms of the GNU General Public License
37 * as published by the Free Software Foundation; either version
38 * 2 of the License, or (at your option) any later version.
39 */
40#ifndef _SOCK_H
41#define _SOCK_H
42
a6b7a407 43#include <linux/hardirq.h>
172589cc 44#include <linux/kernel.h>
1da177e4 45#include <linux/list.h>
88ab1932 46#include <linux/list_nulls.h>
1da177e4
LT
47#include <linux/timer.h>
48#include <linux/cache.h>
3f134619 49#include <linux/bitops.h>
a5b5bb9a 50#include <linux/lockdep.h>
1da177e4
LT
51#include <linux/netdevice.h>
52#include <linux/skbuff.h> /* struct sk_buff */
d7fe0f24 53#include <linux/mm.h>
1da177e4 54#include <linux/security.h>
5a0e3ad6 55#include <linux/slab.h>
c6e1a0d1 56#include <linux/uaccess.h>
3e32cb2e 57#include <linux/page_counter.h>
180d8cd9 58#include <linux/memcontrol.h>
c5905afb 59#include <linux/static_key.h>
40401530 60#include <linux/sched.h>
1ce0bf50 61#include <linux/wait.h>
2a56a1fe 62#include <linux/cgroup-defs.h>
1da177e4
LT
63
64#include <linux/filter.h>
88ab1932 65#include <linux/rculist_nulls.h>
a57de0b4 66#include <linux/poll.h>
1da177e4 67
c31504dc 68#include <linux/atomic.h>
1da177e4
LT
69#include <net/dst.h>
70#include <net/checksum.h>
1d0ab253 71#include <net/tcp_states.h>
b9f40e21 72#include <linux/net_tstamp.h>
1da177e4
LT
73
74/*
75 * This structure really needs to be cleaned up.
76 * Most of it is for TCP, and not used by any of
77 * the other protocols.
78 */
79
80/* Define this to get the SOCK_DBG debugging facility. */
81#define SOCK_DEBUGGING
82#ifdef SOCK_DEBUGGING
83#define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
84 printk(KERN_DEBUG msg); } while (0)
85#else
4cd9029d 86/* Validate arguments and do nothing */
b9075fa9 87static inline __printf(2, 3)
dc6b9b78 88void SOCK_DEBUG(const struct sock *sk, const char *msg, ...)
4cd9029d
SH
89{
90}
1da177e4
LT
91#endif
92
93/* This is the per-socket lock. The spinlock provides a synchronization
94 * between user contexts and software interrupt processing, whereas the
95 * mini-semaphore synchronizes multiple users amongst themselves.
96 */
1da177e4
LT
97typedef struct {
98 spinlock_t slock;
d2e9117c 99 int owned;
1da177e4 100 wait_queue_head_t wq;
a5b5bb9a
IM
101 /*
102 * We express the mutex-alike socket_lock semantics
103 * to the lock validator by explicitly managing
104 * the slock as a lock variant (in addition to
105 * the slock itself):
106 */
107#ifdef CONFIG_DEBUG_LOCK_ALLOC
108 struct lockdep_map dep_map;
109#endif
1da177e4
LT
110} socket_lock_t;
111
1da177e4 112struct sock;
8feaf0c0 113struct proto;
0eeb8ffc 114struct net;
1da177e4 115
077b393d
ED
116typedef __u32 __bitwise __portpair;
117typedef __u64 __bitwise __addrpair;
118
1da177e4 119/**
4dc3b16b 120 * struct sock_common - minimal network layer representation of sockets
68835aba
ED
121 * @skc_daddr: Foreign IPv4 addr
122 * @skc_rcv_saddr: Bound local IPv4 addr
4dc6dc71 123 * @skc_hash: hash value used with various protocol lookup tables
d4cada4a 124 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
ce43b03e
ED
125 * @skc_dport: placeholder for inet_dport/tw_dport
126 * @skc_num: placeholder for inet_num/tw_num
4dc3b16b
PP
127 * @skc_family: network address family
128 * @skc_state: Connection state
129 * @skc_reuse: %SO_REUSEADDR setting
055dc21a 130 * @skc_reuseport: %SO_REUSEPORT setting
4dc3b16b 131 * @skc_bound_dev_if: bound device index if != 0
4dc3b16b 132 * @skc_bind_node: bind hash linkage for various protocol lookup tables
512615b6 133 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
8feaf0c0 134 * @skc_prot: protocol handlers inside a network family
07feaebf 135 * @skc_net: reference to the network namespace of this socket
68835aba
ED
136 * @skc_node: main hash linkage for various protocol lookup tables
137 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
138 * @skc_tx_queue_mapping: tx queue number for this connection
8e5eb54d
ED
139 * @skc_flags: place holder for sk_flags
140 * %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
141 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
70da268b 142 * @skc_incoming_cpu: record/match cpu processing incoming packets
68835aba 143 * @skc_refcnt: reference count
4dc3b16b
PP
144 *
145 * This is the minimal network layer representation of sockets, the header
8feaf0c0
ACM
146 * for struct sock and struct inet_timewait_sock.
147 */
1da177e4 148struct sock_common {
ce43b03e 149 /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned
05dbc7b5 150 * address on 64bit arches : cf INET_MATCH()
4dc6dc71 151 */
ce43b03e 152 union {
077b393d 153 __addrpair skc_addrpair;
ce43b03e
ED
154 struct {
155 __be32 skc_daddr;
156 __be32 skc_rcv_saddr;
157 };
158 };
d4cada4a
ED
159 union {
160 unsigned int skc_hash;
161 __u16 skc_u16hashes[2];
162 };
ce43b03e
ED
163 /* skc_dport && skc_num must be grouped as well */
164 union {
077b393d 165 __portpair skc_portpair;
ce43b03e
ED
166 struct {
167 __be16 skc_dport;
168 __u16 skc_num;
169 };
170 };
171
4dc6dc71
ED
172 unsigned short skc_family;
173 volatile unsigned char skc_state;
055dc21a 174 unsigned char skc_reuse:4;
9fe516ba
ED
175 unsigned char skc_reuseport:1;
176 unsigned char skc_ipv6only:1;
26abe143 177 unsigned char skc_net_refcnt:1;
4dc6dc71 178 int skc_bound_dev_if;
512615b6
ED
179 union {
180 struct hlist_node skc_bind_node;
ca065d0c 181 struct hlist_node skc_portaddr_node;
512615b6 182 };
8feaf0c0 183 struct proto *skc_prot;
0c5c9fb5 184 possible_net_t skc_net;
efe4208f
ED
185
186#if IS_ENABLED(CONFIG_IPV6)
187 struct in6_addr skc_v6_daddr;
188 struct in6_addr skc_v6_rcv_saddr;
189#endif
190
33cf7c90
ED
191 atomic64_t skc_cookie;
192
8e5eb54d
ED
193 /* following fields are padding to force
194 * offset(struct sock, sk_refcnt) == 128 on 64bit arches
195 * assuming IPV6 is enabled. We use this padding differently
196 * for different kind of 'sockets'
197 */
198 union {
199 unsigned long skc_flags;
200 struct sock *skc_listener; /* request_sock */
201 struct inet_timewait_death_row *skc_tw_dr; /* inet_timewait_sock */
202 };
68835aba
ED
203 /*
204 * fields between dontcopy_begin/dontcopy_end
205 * are not copied in sock_copy()
206 */
928c41e7 207 /* private: */
68835aba 208 int skc_dontcopy_begin[0];
928c41e7 209 /* public: */
68835aba
ED
210 union {
211 struct hlist_node skc_node;
212 struct hlist_nulls_node skc_nulls_node;
213 };
214 int skc_tx_queue_mapping;
ed53d0ab
ED
215 union {
216 int skc_incoming_cpu;
217 u32 skc_rcv_wnd;
d475f090 218 u32 skc_tw_rcv_nxt; /* struct tcp_timewait_sock */
ed53d0ab 219 };
70da268b 220
68835aba 221 atomic_t skc_refcnt;
928c41e7 222 /* private: */
68835aba 223 int skc_dontcopy_end[0];
ed53d0ab
ED
224 union {
225 u32 skc_rxhash;
226 u32 skc_window_clamp;
d475f090 227 u32 skc_tw_snd_nxt; /* struct tcp_timewait_sock */
ed53d0ab 228 };
928c41e7 229 /* public: */
1da177e4
LT
230};
231
232/**
233 * struct sock - network layer representation of sockets
8feaf0c0 234 * @__sk_common: shared layout with inet_timewait_sock
4dc3b16b
PP
235 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
236 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
237 * @sk_lock: synchronizer
238 * @sk_rcvbuf: size of receive buffer in bytes
43815482 239 * @sk_wq: sock wait queue and async head
421b3885 240 * @sk_rx_dst: receive input route used by early demux
4dc3b16b 241 * @sk_dst_cache: destination cache
4dc3b16b 242 * @sk_policy: flow policy
4dc3b16b
PP
243 * @sk_receive_queue: incoming packets
244 * @sk_wmem_alloc: transmit queue bytes committed
245 * @sk_write_queue: Packet sending queue
246 * @sk_omem_alloc: "o" is "option" or "other"
247 * @sk_wmem_queued: persistent queue size
248 * @sk_forward_alloc: space allocated forward
06021292 249 * @sk_napi_id: id of the last napi context to receive data for sk
dafcc438 250 * @sk_ll_usec: usecs to busypoll when there is no data
4dc3b16b 251 * @sk_allocation: allocation mode
95bd09eb 252 * @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
c3f40d7c 253 * @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE)
4dc3b16b 254 * @sk_sndbuf: size of send buffer in bytes
28448b80
TH
255 * @sk_no_check_tx: %SO_NO_CHECK setting, set checksum in TX packets
256 * @sk_no_check_rx: allow zero checksum in RX packets
4dc3b16b 257 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
a465419b 258 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
bcd76111 259 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
82cc1a7a 260 * @sk_gso_max_size: Maximum GSO segment size to build
1485348d 261 * @sk_gso_max_segs: Maximum number of GSO segments
4dc3b16b 262 * @sk_lingertime: %SO_LINGER l_linger setting
4dc3b16b
PP
263 * @sk_backlog: always used with the per-socket spinlock held
264 * @sk_callback_lock: used with the callbacks in the end of this struct
265 * @sk_error_queue: rarely used
33c732c3
WC
266 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
267 * IPV6_ADDRFORM for instance)
4dc3b16b 268 * @sk_err: last error
33c732c3
WC
269 * @sk_err_soft: errors that don't cause failure but are the cause of a
270 * persistent failure not just 'timed out'
cb61cb9b 271 * @sk_drops: raw/udp drops counter
4dc3b16b
PP
272 * @sk_ack_backlog: current listen backlog
273 * @sk_max_ack_backlog: listen backlog set in listen()
274 * @sk_priority: %SO_PRIORITY setting
275 * @sk_type: socket type (%SOCK_STREAM, etc)
276 * @sk_protocol: which protocol this socket belongs in this network family
53c3fa20
RD
277 * @sk_peer_pid: &struct pid for this socket's peer
278 * @sk_peer_cred: %SO_PEERCRED setting
4dc3b16b
PP
279 * @sk_rcvlowat: %SO_RCVLOWAT setting
280 * @sk_rcvtimeo: %SO_RCVTIMEO setting
281 * @sk_sndtimeo: %SO_SNDTIMEO setting
b73c3d0e 282 * @sk_txhash: computed flow hash for use on transmit
4dc3b16b 283 * @sk_filter: socket filtering instructions
4dc3b16b
PP
284 * @sk_timer: sock cleanup timer
285 * @sk_stamp: time stamp of last packet received
b9f40e21 286 * @sk_tsflags: SO_TIMESTAMPING socket options
09c2d251 287 * @sk_tskey: counter to disambiguate concurrent tstamp requests
4dc3b16b
PP
288 * @sk_socket: Identd and reporting IO signals
289 * @sk_user_data: RPC layer private data
5640f768 290 * @sk_frag: cached page frag
d3d4f0a0 291 * @sk_peek_off: current peek_offset value
4dc3b16b 292 * @sk_send_head: front of stuff to transmit
67be2dd1 293 * @sk_security: used by security modules
31729363 294 * @sk_mark: generic packet mark
2a56a1fe 295 * @sk_cgrp_data: cgroup data for this cgroup
baac50bb 296 * @sk_memcg: this socket's memory cgroup association
4dc3b16b
PP
297 * @sk_write_pending: a write to stream socket waits to start
298 * @sk_state_change: callback to indicate change in the state of the sock
299 * @sk_data_ready: callback to indicate there is data to be processed
300 * @sk_write_space: callback to indicate there is bf sending space available
301 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
302 * @sk_backlog_rcv: callback to process the backlog
303 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
ef456144 304 * @sk_reuseport_cb: reuseport group container
1da177e4
LT
305 */
306struct sock {
307 /*
8feaf0c0 308 * Now struct inet_timewait_sock also uses sock_common, so please just
1da177e4
LT
309 * don't add nothing before this first member (__sk_common) --acme
310 */
311 struct sock_common __sk_common;
4dc6dc71
ED
312#define sk_node __sk_common.skc_node
313#define sk_nulls_node __sk_common.skc_nulls_node
314#define sk_refcnt __sk_common.skc_refcnt
e022f0b4 315#define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
4dc6dc71 316
68835aba
ED
317#define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
318#define sk_dontcopy_end __sk_common.skc_dontcopy_end
4dc6dc71 319#define sk_hash __sk_common.skc_hash
50805466 320#define sk_portpair __sk_common.skc_portpair
05dbc7b5
ED
321#define sk_num __sk_common.skc_num
322#define sk_dport __sk_common.skc_dport
50805466
ED
323#define sk_addrpair __sk_common.skc_addrpair
324#define sk_daddr __sk_common.skc_daddr
325#define sk_rcv_saddr __sk_common.skc_rcv_saddr
1da177e4
LT
326#define sk_family __sk_common.skc_family
327#define sk_state __sk_common.skc_state
328#define sk_reuse __sk_common.skc_reuse
055dc21a 329#define sk_reuseport __sk_common.skc_reuseport
9fe516ba 330#define sk_ipv6only __sk_common.skc_ipv6only
26abe143 331#define sk_net_refcnt __sk_common.skc_net_refcnt
1da177e4 332#define sk_bound_dev_if __sk_common.skc_bound_dev_if
1da177e4 333#define sk_bind_node __sk_common.skc_bind_node
8feaf0c0 334#define sk_prot __sk_common.skc_prot
07feaebf 335#define sk_net __sk_common.skc_net
efe4208f
ED
336#define sk_v6_daddr __sk_common.skc_v6_daddr
337#define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr
33cf7c90 338#define sk_cookie __sk_common.skc_cookie
70da268b 339#define sk_incoming_cpu __sk_common.skc_incoming_cpu
8e5eb54d 340#define sk_flags __sk_common.skc_flags
ed53d0ab 341#define sk_rxhash __sk_common.skc_rxhash
efe4208f 342
1da177e4 343 socket_lock_t sk_lock;
b178bb3d 344 struct sk_buff_head sk_receive_queue;
fa438ccf
ED
345 /*
346 * The backlog queue is special, it is always used with
347 * the per-socket spinlock held and requires low latency
348 * access. Therefore we special case it's implementation.
b178bb3d
ED
349 * Note : rmem_alloc is in this structure to fill a hole
350 * on 64bit arches, not because its logically part of
351 * backlog.
fa438ccf
ED
352 */
353 struct {
b178bb3d
ED
354 atomic_t rmem_alloc;
355 int len;
356 struct sk_buff *head;
357 struct sk_buff *tail;
fa438ccf 358 } sk_backlog;
b178bb3d
ED
359#define sk_rmem_alloc sk_backlog.rmem_alloc
360 int sk_forward_alloc;
2c8c56e1 361
b73c3d0e 362 __u32 sk_txhash;
e0d1095a 363#ifdef CONFIG_NET_RX_BUSY_POLL
06021292 364 unsigned int sk_napi_id;
dafcc438 365 unsigned int sk_ll_usec;
b178bb3d
ED
366#endif
367 atomic_t sk_drops;
368 int sk_rcvbuf;
369
370 struct sk_filter __rcu *sk_filter;
ceb5d58b
ED
371 union {
372 struct socket_wq __rcu *sk_wq;
373 struct socket_wq *sk_wq_raw;
374 };
def8b4fa 375#ifdef CONFIG_XFRM
d188ba86 376 struct xfrm_policy __rcu *sk_policy[2];
def8b4fa 377#endif
deaa5854 378 struct dst_entry *sk_rx_dst;
0e36cbb3 379 struct dst_entry __rcu *sk_dst_cache;
6bd4f355 380 /* Note: 32bit hole on 64bit arches */
1da177e4
LT
381 atomic_t sk_wmem_alloc;
382 atomic_t sk_omem_alloc;
4e07a91c 383 int sk_sndbuf;
1da177e4 384 struct sk_buff_head sk_write_queue;
b178bb3d
ED
385 kmemcheck_bitfield_begin(flags);
386 unsigned int sk_shutdown : 2,
28448b80
TH
387 sk_no_check_tx : 1,
388 sk_no_check_rx : 1,
b178bb3d
ED
389 sk_userlocks : 4,
390 sk_protocol : 8,
391 sk_type : 16;
7bbadd2d 392#define SK_PROTOCOL_MAX U8_MAX
b178bb3d 393 kmemcheck_bitfield_end(flags);
1da177e4 394 int sk_wmem_queued;
7d877f3b 395 gfp_t sk_allocation;
95bd09eb 396 u32 sk_pacing_rate; /* bytes per second */
62748f32 397 u32 sk_max_pacing_rate;
c8f44aff
MM
398 netdev_features_t sk_route_caps;
399 netdev_features_t sk_route_nocaps;
bcd76111 400 int sk_gso_type;
82cc1a7a 401 unsigned int sk_gso_max_size;
1485348d 402 u16 sk_gso_max_segs;
9932cf95 403 int sk_rcvlowat;
1da177e4 404 unsigned long sk_lingertime;
1da177e4 405 struct sk_buff_head sk_error_queue;
476e19cf 406 struct proto *sk_prot_creator;
1da177e4
LT
407 rwlock_t sk_callback_lock;
408 int sk_err,
409 sk_err_soft;
becb74f0
ED
410 u32 sk_ack_backlog;
411 u32 sk_max_ack_backlog;
1da177e4 412 __u32 sk_priority;
297dbde1 413 __u32 sk_mark;
109f6e39
EB
414 struct pid *sk_peer_pid;
415 const struct cred *sk_peer_cred;
1da177e4
LT
416 long sk_rcvtimeo;
417 long sk_sndtimeo;
1da177e4 418 struct timer_list sk_timer;
b7aa0bf7 419 ktime_t sk_stamp;
b9f40e21 420 u16 sk_tsflags;
09c2d251 421 u32 sk_tskey;
1da177e4
LT
422 struct socket *sk_socket;
423 void *sk_user_data;
5640f768 424 struct page_frag sk_frag;
1da177e4 425 struct sk_buff *sk_send_head;
ef64a54f 426 __s32 sk_peek_off;
1da177e4 427 int sk_write_pending;
d5f64238 428#ifdef CONFIG_SECURITY
1da177e4 429 void *sk_security;
d5f64238 430#endif
2a56a1fe 431 struct sock_cgroup_data sk_cgrp_data;
baac50bb 432 struct mem_cgroup *sk_memcg;
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 439 void (*sk_destruct)(struct sock *sk);
ef456144 440 struct sock_reuseport __rcu *sk_reuseport_cb;
a4298e45 441 struct rcu_head sk_rcu;
1da177e4
LT
442};
443
559835ea
PS
444#define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
445
446#define rcu_dereference_sk_user_data(sk) rcu_dereference(__sk_user_data((sk)))
447#define rcu_assign_sk_user_data(sk, ptr) rcu_assign_pointer(__sk_user_data((sk)), ptr)
448
4a17fd52
PE
449/*
450 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
451 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
452 * on a socket means that the socket will reuse everybody else's port
453 * without looking at the other's sk_reuse value.
454 */
455
456#define SK_NO_REUSE 0
457#define SK_CAN_REUSE 1
458#define SK_FORCE_REUSE 2
459
627d2d6b 460int sk_set_peek_off(struct sock *sk, int val);
461
ef64a54f
PE
462static inline int sk_peek_offset(struct sock *sk, int flags)
463{
b9bb53f3
WB
464 if (unlikely(flags & MSG_PEEK)) {
465 s32 off = READ_ONCE(sk->sk_peek_off);
466 if (off >= 0)
467 return off;
468 }
469
470 return 0;
ef64a54f
PE
471}
472
473static inline void sk_peek_offset_bwd(struct sock *sk, int val)
474{
b9bb53f3
WB
475 s32 off = READ_ONCE(sk->sk_peek_off);
476
477 if (unlikely(off >= 0)) {
478 off = max_t(s32, off - val, 0);
479 WRITE_ONCE(sk->sk_peek_off, off);
ef64a54f
PE
480 }
481}
482
483static inline void sk_peek_offset_fwd(struct sock *sk, int val)
484{
b9bb53f3 485 sk_peek_offset_bwd(sk, -val);
ef64a54f
PE
486}
487
1da177e4
LT
488/*
489 * Hashed lists helper routines
490 */
c4146644
LZ
491static inline struct sock *sk_entry(const struct hlist_node *node)
492{
493 return hlist_entry(node, struct sock, sk_node);
494}
495
e48c414e 496static inline struct sock *__sk_head(const struct hlist_head *head)
1da177e4
LT
497{
498 return hlist_entry(head->first, struct sock, sk_node);
499}
500
e48c414e 501static inline struct sock *sk_head(const struct hlist_head *head)
1da177e4
LT
502{
503 return hlist_empty(head) ? NULL : __sk_head(head);
504}
505
88ab1932
ED
506static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
507{
508 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
509}
510
511static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
512{
513 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
514}
515
e48c414e 516static inline struct sock *sk_next(const struct sock *sk)
1da177e4
LT
517{
518 return sk->sk_node.next ?
519 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
520}
521
88ab1932
ED
522static inline struct sock *sk_nulls_next(const struct sock *sk)
523{
524 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
525 hlist_nulls_entry(sk->sk_nulls_node.next,
526 struct sock, sk_nulls_node) :
527 NULL;
528}
529
dc6b9b78 530static inline bool sk_unhashed(const struct sock *sk)
1da177e4
LT
531{
532 return hlist_unhashed(&sk->sk_node);
533}
534
dc6b9b78 535static inline bool sk_hashed(const struct sock *sk)
1da177e4 536{
da753bea 537 return !sk_unhashed(sk);
1da177e4
LT
538}
539
dc6b9b78 540static inline void sk_node_init(struct hlist_node *node)
1da177e4
LT
541{
542 node->pprev = NULL;
543}
544
dc6b9b78 545static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
88ab1932
ED
546{
547 node->pprev = NULL;
548}
549
dc6b9b78 550static inline void __sk_del_node(struct sock *sk)
1da177e4
LT
551{
552 __hlist_del(&sk->sk_node);
553}
554
808f5114 555/* NB: equivalent to hlist_del_init_rcu */
dc6b9b78 556static inline bool __sk_del_node_init(struct sock *sk)
1da177e4
LT
557{
558 if (sk_hashed(sk)) {
559 __sk_del_node(sk);
560 sk_node_init(&sk->sk_node);
dc6b9b78 561 return true;
1da177e4 562 }
dc6b9b78 563 return false;
1da177e4
LT
564}
565
566/* Grab socket reference count. This operation is valid only
567 when sk is ALREADY grabbed f.e. it is found in hash table
568 or a list and the lookup is made under lock preventing hash table
569 modifications.
570 */
571
f9a7cbbf 572static __always_inline void sock_hold(struct sock *sk)
1da177e4
LT
573{
574 atomic_inc(&sk->sk_refcnt);
575}
576
577/* Ungrab socket in the context, which assumes that socket refcnt
578 cannot hit zero, f.e. it is true in context of any socketcall.
579 */
f9a7cbbf 580static __always_inline void __sock_put(struct sock *sk)
1da177e4
LT
581{
582 atomic_dec(&sk->sk_refcnt);
583}
584
dc6b9b78 585static inline bool sk_del_node_init(struct sock *sk)
1da177e4 586{
dc6b9b78 587 bool rc = __sk_del_node_init(sk);
1da177e4
LT
588
589 if (rc) {
590 /* paranoid for a while -acme */
591 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
592 __sock_put(sk);
593 }
594 return rc;
595}
808f5114 596#define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
1da177e4 597
dc6b9b78 598static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
271b72c7
ED
599{
600 if (sk_hashed(sk)) {
88ab1932 601 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
dc6b9b78 602 return true;
271b72c7 603 }
dc6b9b78 604 return false;
271b72c7
ED
605}
606
dc6b9b78 607static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
271b72c7 608{
dc6b9b78 609 bool rc = __sk_nulls_del_node_init_rcu(sk);
271b72c7
ED
610
611 if (rc) {
612 /* paranoid for a while -acme */
613 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
614 __sock_put(sk);
615 }
616 return rc;
617}
618
dc6b9b78 619static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
1da177e4
LT
620{
621 hlist_add_head(&sk->sk_node, list);
622}
623
dc6b9b78 624static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
1da177e4
LT
625{
626 sock_hold(sk);
627 __sk_add_node(sk, list);
628}
629
dc6b9b78 630static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
808f5114 631{
632 sock_hold(sk);
633 hlist_add_head_rcu(&sk->sk_node, list);
634}
635
dc6b9b78 636static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
271b72c7 637{
d894ba18
CG
638 if (IS_ENABLED(CONFIG_IPV6) && sk->sk_reuseport &&
639 sk->sk_family == AF_INET6)
640 hlist_nulls_add_tail_rcu(&sk->sk_nulls_node, list);
641 else
642 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
271b72c7
ED
643}
644
dc6b9b78 645static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
271b72c7
ED
646{
647 sock_hold(sk);
88ab1932 648 __sk_nulls_add_node_rcu(sk, list);
271b72c7
ED
649}
650
dc6b9b78 651static inline void __sk_del_bind_node(struct sock *sk)
1da177e4
LT
652{
653 __hlist_del(&sk->sk_bind_node);
654}
655
dc6b9b78 656static inline void sk_add_bind_node(struct sock *sk,
1da177e4
LT
657 struct hlist_head *list)
658{
659 hlist_add_head(&sk->sk_bind_node, list);
660}
661
b67bfe0d
SL
662#define sk_for_each(__sk, list) \
663 hlist_for_each_entry(__sk, list, sk_node)
664#define sk_for_each_rcu(__sk, list) \
665 hlist_for_each_entry_rcu(__sk, list, sk_node)
88ab1932
ED
666#define sk_nulls_for_each(__sk, node, list) \
667 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
668#define sk_nulls_for_each_rcu(__sk, node, list) \
669 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
b67bfe0d
SL
670#define sk_for_each_from(__sk) \
671 hlist_for_each_entry_from(__sk, sk_node)
88ab1932
ED
672#define sk_nulls_for_each_from(__sk, node) \
673 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
674 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
b67bfe0d
SL
675#define sk_for_each_safe(__sk, tmp, list) \
676 hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
677#define sk_for_each_bound(__sk, list) \
678 hlist_for_each_entry(__sk, list, sk_bind_node)
1da177e4 679
2dc41cff 680/**
ca065d0c 681 * sk_for_each_entry_offset_rcu - iterate over a list at a given struct offset
2dc41cff
DH
682 * @tpos: the type * to use as a loop cursor.
683 * @pos: the &struct hlist_node to use as a loop cursor.
684 * @head: the head for your list.
685 * @offset: offset of hlist_node within the struct.
686 *
687 */
ca065d0c
ED
688#define sk_for_each_entry_offset_rcu(tpos, pos, head, offset) \
689 for (pos = rcu_dereference((head)->first); \
690 pos != NULL && \
2dc41cff 691 ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;}); \
ca065d0c 692 pos = rcu_dereference(pos->next))
2dc41cff 693
c336d148
EB
694static inline struct user_namespace *sk_user_ns(struct sock *sk)
695{
696 /* Careful only use this in a context where these parameters
697 * can not change and must all be valid, such as recvmsg from
698 * userspace.
699 */
700 return sk->sk_socket->file->f_cred->user_ns;
701}
702
1da177e4
LT
703/* Sock flags */
704enum sock_flags {
705 SOCK_DEAD,
706 SOCK_DONE,
707 SOCK_URGINLINE,
708 SOCK_KEEPOPEN,
709 SOCK_LINGER,
710 SOCK_DESTROY,
711 SOCK_BROADCAST,
712 SOCK_TIMESTAMP,
713 SOCK_ZAPPED,
714 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
715 SOCK_DBG, /* %SO_DEBUG setting */
716 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
92f37fd2 717 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
1da177e4
LT
718 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
719 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
7cb02404 720 SOCK_MEMALLOC, /* VM depends on this socket for swapping */
20d49473 721 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
bcdce719 722 SOCK_FASYNC, /* fasync() active */
3b885787 723 SOCK_RXQ_OVFL,
1cdebb42 724 SOCK_ZEROCOPY, /* buffers from userspace */
6e3e939f 725 SOCK_WIFI_STATUS, /* push wifi status to userspace */
3bdc0eba
BG
726 SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
727 * Will use last 4 bytes of packet sent from
728 * user-space instead.
729 */
d59577b6 730 SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */
7d4c04fc 731 SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */
a4298e45 732 SOCK_RCU_FREE, /* wait rcu grace period in sk_destruct() */
1da177e4
LT
733};
734
01ce63c9
MRL
735#define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
736
53b924b3
RB
737static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
738{
739 nsk->sk_flags = osk->sk_flags;
740}
741
1da177e4
LT
742static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
743{
744 __set_bit(flag, &sk->sk_flags);
745}
746
747static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
748{
749 __clear_bit(flag, &sk->sk_flags);
750}
751
1b23a5df 752static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
1da177e4
LT
753{
754 return test_bit(flag, &sk->sk_flags);
755}
756
c93bdd0e
MG
757#ifdef CONFIG_NET
758extern struct static_key memalloc_socks;
759static inline int sk_memalloc_socks(void)
760{
761 return static_key_false(&memalloc_socks);
762}
763#else
764
765static inline int sk_memalloc_socks(void)
766{
767 return 0;
768}
769
770#endif
771
7450aaf6 772static inline gfp_t sk_gfp_mask(const struct sock *sk, gfp_t gfp_mask)
99a1dec7 773{
7450aaf6 774 return gfp_mask | (sk->sk_allocation & __GFP_MEMALLOC);
99a1dec7
MG
775}
776
1da177e4
LT
777static inline void sk_acceptq_removed(struct sock *sk)
778{
779 sk->sk_ack_backlog--;
780}
781
782static inline void sk_acceptq_added(struct sock *sk)
783{
784 sk->sk_ack_backlog++;
785}
786
dc6b9b78 787static inline bool sk_acceptq_is_full(const struct sock *sk)
1da177e4 788{
64a14651 789 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
1da177e4
LT
790}
791
792/*
793 * Compute minimal free write space needed to queue new packets.
794 */
dc6b9b78 795static inline int sk_stream_min_wspace(const struct sock *sk)
1da177e4 796{
8df09ea3 797 return sk->sk_wmem_queued >> 1;
1da177e4
LT
798}
799
dc6b9b78 800static inline int sk_stream_wspace(const struct sock *sk)
1da177e4
LT
801{
802 return sk->sk_sndbuf - sk->sk_wmem_queued;
803}
804
69336bd2 805void sk_stream_write_space(struct sock *sk);
1da177e4 806
8eae939f 807/* OOB backlog add */
a3a858ff 808static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
9ee6b535 809{
7fee226a 810 /* dont let skb dst not refcounted, we are going to leave rcu lock */
5037e9ef 811 skb_dst_force_safe(skb);
7fee226a
ED
812
813 if (!sk->sk_backlog.tail)
814 sk->sk_backlog.head = skb;
815 else
9ee6b535 816 sk->sk_backlog.tail->next = skb;
7fee226a
ED
817
818 sk->sk_backlog.tail = skb;
9ee6b535
SH
819 skb->next = NULL;
820}
1da177e4 821
c377411f
ED
822/*
823 * Take into account size of receive queue and backlog queue
0fd7bac6
ED
824 * Do not take into account this skb truesize,
825 * to allow even a single big packet to come.
c377411f 826 */
274f482d 827static inline bool sk_rcvqueues_full(const struct sock *sk, unsigned int limit)
c377411f
ED
828{
829 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
830
f545a38f 831 return qsize > limit;
c377411f
ED
832}
833
8eae939f 834/* The per-socket spinlock must be held here. */
f545a38f
ED
835static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
836 unsigned int limit)
8eae939f 837{
274f482d 838 if (sk_rcvqueues_full(sk, limit))
8eae939f
ZY
839 return -ENOBUFS;
840
c7c49b8f
ED
841 /*
842 * If the skb was allocated from pfmemalloc reserves, only
843 * allow SOCK_MEMALLOC sockets to use it as this socket is
844 * helping free memory
845 */
846 if (skb_pfmemalloc(skb) && !sock_flag(sk, SOCK_MEMALLOC))
847 return -ENOMEM;
848
a3a858ff 849 __sk_add_backlog(sk, skb);
8eae939f
ZY
850 sk->sk_backlog.len += skb->truesize;
851 return 0;
852}
853
69336bd2 854int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
b4b9e355 855
c57943a1
PZ
856static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
857{
b4b9e355
MG
858 if (sk_memalloc_socks() && skb_pfmemalloc(skb))
859 return __sk_backlog_rcv(sk, skb);
860
c57943a1
PZ
861 return sk->sk_backlog_rcv(sk, skb);
862}
863
2c8c56e1
ED
864static inline void sk_incoming_cpu_update(struct sock *sk)
865{
866 sk->sk_incoming_cpu = raw_smp_processor_id();
867}
868
fe477558 869static inline void sock_rps_record_flow_hash(__u32 hash)
c58dc01b
DM
870{
871#ifdef CONFIG_RPS
872 struct rps_sock_flow_table *sock_flow_table;
873
874 rcu_read_lock();
875 sock_flow_table = rcu_dereference(rps_sock_flow_table);
fe477558 876 rps_record_sock_flow(sock_flow_table, hash);
c58dc01b
DM
877 rcu_read_unlock();
878#endif
879}
880
fe477558
TH
881static inline void sock_rps_record_flow(const struct sock *sk)
882{
c9d8ca04 883#ifdef CONFIG_RPS
fe477558 884 sock_rps_record_flow_hash(sk->sk_rxhash);
c9d8ca04 885#endif
fe477558
TH
886}
887
bdeab991
TH
888static inline void sock_rps_save_rxhash(struct sock *sk,
889 const struct sk_buff *skb)
c58dc01b
DM
890{
891#ifdef CONFIG_RPS
567e4b79 892 if (unlikely(sk->sk_rxhash != skb->hash))
61b905da 893 sk->sk_rxhash = skb->hash;
c58dc01b
DM
894#endif
895}
896
bdeab991
TH
897static inline void sock_rps_reset_rxhash(struct sock *sk)
898{
899#ifdef CONFIG_RPS
bdeab991
TH
900 sk->sk_rxhash = 0;
901#endif
902}
903
cfcabdcc
SH
904#define sk_wait_event(__sk, __timeo, __condition) \
905 ({ int __rc; \
906 release_sock(__sk); \
907 __rc = __condition; \
908 if (!__rc) { \
909 *(__timeo) = schedule_timeout(*(__timeo)); \
910 } \
26cabd31 911 sched_annotate_sleep(); \
cfcabdcc
SH
912 lock_sock(__sk); \
913 __rc = __condition; \
914 __rc; \
915 })
1da177e4 916
69336bd2
JP
917int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
918int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
919void sk_stream_wait_close(struct sock *sk, long timeo_p);
920int sk_stream_error(struct sock *sk, int flags, int err);
921void sk_stream_kill_queues(struct sock *sk);
922void sk_set_memalloc(struct sock *sk);
923void sk_clear_memalloc(struct sock *sk);
1da177e4 924
dfbafc99 925int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb);
1da177e4 926
60236fdd 927struct request_sock_ops;
6d6ee43e 928struct timewait_sock_ops;
ab1e0a13 929struct inet_hashinfo;
fc8717ba 930struct raw_hashinfo;
de477254 931struct module;
2e6599cb 932
f77d6021
ED
933/*
934 * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
935 * un-modified. Special care is taken when initializing object to zero.
936 */
937static inline void sk_prot_clear_nulls(struct sock *sk, int size)
938{
939 if (offsetof(struct sock, sk_node.next) != 0)
940 memset(sk, 0, offsetof(struct sock, sk_node.next));
941 memset(&sk->sk_node.pprev, 0,
942 size - offsetof(struct sock, sk_node.pprev));
943}
944
1da177e4
LT
945/* Networking protocol blocks we attach to sockets.
946 * socket layer -> transport layer interface
1da177e4
LT
947 */
948struct proto {
dc6b9b78 949 void (*close)(struct sock *sk,
1da177e4
LT
950 long timeout);
951 int (*connect)(struct sock *sk,
dc6b9b78 952 struct sockaddr *uaddr,
1da177e4
LT
953 int addr_len);
954 int (*disconnect)(struct sock *sk, int flags);
955
dc6b9b78 956 struct sock * (*accept)(struct sock *sk, int flags, int *err);
1da177e4
LT
957
958 int (*ioctl)(struct sock *sk, int cmd,
959 unsigned long arg);
960 int (*init)(struct sock *sk);
7d06b2e0 961 void (*destroy)(struct sock *sk);
1da177e4 962 void (*shutdown)(struct sock *sk, int how);
dc6b9b78 963 int (*setsockopt)(struct sock *sk, int level,
1da177e4 964 int optname, char __user *optval,
b7058842 965 unsigned int optlen);
dc6b9b78
ED
966 int (*getsockopt)(struct sock *sk, int level,
967 int optname, char __user *optval,
968 int __user *option);
af01d537 969#ifdef CONFIG_COMPAT
3fdadf7d
DM
970 int (*compat_setsockopt)(struct sock *sk,
971 int level,
972 int optname, char __user *optval,
b7058842 973 unsigned int optlen);
3fdadf7d
DM
974 int (*compat_getsockopt)(struct sock *sk,
975 int level,
976 int optname, char __user *optval,
977 int __user *option);
709b46e8
EB
978 int (*compat_ioctl)(struct sock *sk,
979 unsigned int cmd, unsigned long arg);
af01d537 980#endif
1b784140
YX
981 int (*sendmsg)(struct sock *sk, struct msghdr *msg,
982 size_t len);
983 int (*recvmsg)(struct sock *sk, struct msghdr *msg,
dc6b9b78
ED
984 size_t len, int noblock, int flags,
985 int *addr_len);
1da177e4
LT
986 int (*sendpage)(struct sock *sk, struct page *page,
987 int offset, size_t size, int flags);
dc6b9b78 988 int (*bind)(struct sock *sk,
1da177e4
LT
989 struct sockaddr *uaddr, int addr_len);
990
dc6b9b78 991 int (*backlog_rcv) (struct sock *sk,
1da177e4
LT
992 struct sk_buff *skb);
993
46d3ceab
ED
994 void (*release_cb)(struct sock *sk);
995
1da177e4 996 /* Keeping track of sk's, looking them up, and port selection methods. */
086c653f 997 int (*hash)(struct sock *sk);
1da177e4 998 void (*unhash)(struct sock *sk);
719f8358 999 void (*rehash)(struct sock *sk);
1da177e4 1000 int (*get_port)(struct sock *sk, unsigned short snum);
fcbdf09d 1001 void (*clear_sk)(struct sock *sk, int size);
1da177e4 1002
286ab3d4 1003 /* Keeping track of sockets in use */
65f76517 1004#ifdef CONFIG_PROC_FS
13ff3d6f 1005 unsigned int inuse_idx;
65f76517 1006#endif
ebb53d75 1007
c9bee3b7 1008 bool (*stream_memory_free)(const struct sock *sk);
1da177e4 1009 /* Memory pressure */
5c52ba17 1010 void (*enter_memory_pressure)(struct sock *sk);
8d987e5c 1011 atomic_long_t *memory_allocated; /* Current allocated memory. */
1748376b 1012 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
1da177e4
LT
1013 /*
1014 * Pressure flag: try to collapse.
1015 * Technical note: it is used by multiple contexts non atomically.
3ab224be 1016 * All the __sk_mem_schedule() is of this nature: accounting
1da177e4
LT
1017 * is strict, actions are advisory and have some latency.
1018 */
1019 int *memory_pressure;
8d987e5c 1020 long *sysctl_mem;
1da177e4
LT
1021 int *sysctl_wmem;
1022 int *sysctl_rmem;
1023 int max_header;
7ba42910 1024 bool no_autobind;
1da177e4 1025
271b72c7 1026 struct kmem_cache *slab;
1da177e4 1027 unsigned int obj_size;
271b72c7 1028 int slab_flags;
1da177e4 1029
dd24c001 1030 struct percpu_counter *orphan_count;
8feaf0c0 1031
60236fdd 1032 struct request_sock_ops *rsk_prot;
6d6ee43e 1033 struct timewait_sock_ops *twsk_prot;
2e6599cb 1034
39d8cda7
PE
1035 union {
1036 struct inet_hashinfo *hashinfo;
645ca708 1037 struct udp_table *udp_table;
fc8717ba 1038 struct raw_hashinfo *raw_hash;
39d8cda7 1039 } h;
ab1e0a13 1040
1da177e4
LT
1041 struct module *owner;
1042
1043 char name[32];
1044
1045 struct list_head node;
e6848976
ACM
1046#ifdef SOCK_REFCNT_DEBUG
1047 atomic_t socks;
e1aab161 1048#endif
64be0aed 1049 int (*diag_destroy)(struct sock *sk, int err);
e1aab161
GC
1050};
1051
69336bd2
JP
1052int proto_register(struct proto *prot, int alloc_slab);
1053void proto_unregister(struct proto *prot);
1da177e4 1054
e6848976
ACM
1055#ifdef SOCK_REFCNT_DEBUG
1056static inline void sk_refcnt_debug_inc(struct sock *sk)
1057{
1058 atomic_inc(&sk->sk_prot->socks);
1059}
1060
1061static inline void sk_refcnt_debug_dec(struct sock *sk)
1062{
1063 atomic_dec(&sk->sk_prot->socks);
1064 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
1065 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
1066}
1067
dec34fb0 1068static inline void sk_refcnt_debug_release(const struct sock *sk)
e6848976
ACM
1069{
1070 if (atomic_read(&sk->sk_refcnt) != 1)
1071 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
1072 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
1073}
1074#else /* SOCK_REFCNT_DEBUG */
1075#define sk_refcnt_debug_inc(sk) do { } while (0)
1076#define sk_refcnt_debug_dec(sk) do { } while (0)
1077#define sk_refcnt_debug_release(sk) do { } while (0)
1078#endif /* SOCK_REFCNT_DEBUG */
1079
c9bee3b7
ED
1080static inline bool sk_stream_memory_free(const struct sock *sk)
1081{
1082 if (sk->sk_wmem_queued >= sk->sk_sndbuf)
1083 return false;
1084
1085 return sk->sk_prot->stream_memory_free ?
1086 sk->sk_prot->stream_memory_free(sk) : true;
1087}
1088
64dc6130
ED
1089static inline bool sk_stream_is_writeable(const struct sock *sk)
1090{
c9bee3b7
ED
1091 return sk_stream_wspace(sk) >= sk_stream_min_wspace(sk) &&
1092 sk_stream_memory_free(sk);
64dc6130 1093}
e1aab161 1094
c9bee3b7 1095
180d8cd9
GC
1096static inline bool sk_has_memory_pressure(const struct sock *sk)
1097{
1098 return sk->sk_prot->memory_pressure != NULL;
1099}
1100
1101static inline bool sk_under_memory_pressure(const struct sock *sk)
1102{
1103 if (!sk->sk_prot->memory_pressure)
1104 return false;
e1aab161 1105
baac50bb
JW
1106 if (mem_cgroup_sockets_enabled && sk->sk_memcg &&
1107 mem_cgroup_under_socket_pressure(sk->sk_memcg))
e805605c 1108 return true;
e1aab161 1109
35b87f6c 1110 return !!*sk->sk_prot->memory_pressure;
180d8cd9
GC
1111}
1112
1113static inline void sk_leave_memory_pressure(struct sock *sk)
1114{
1115 int *memory_pressure = sk->sk_prot->memory_pressure;
1116
e1aab161
GC
1117 if (!memory_pressure)
1118 return;
1119
1120 if (*memory_pressure)
180d8cd9
GC
1121 *memory_pressure = 0;
1122}
1123
1124static inline void sk_enter_memory_pressure(struct sock *sk)
1125{
e1aab161
GC
1126 if (!sk->sk_prot->enter_memory_pressure)
1127 return;
1128
e1aab161 1129 sk->sk_prot->enter_memory_pressure(sk);
180d8cd9
GC
1130}
1131
1132static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1133{
e805605c 1134 return sk->sk_prot->sysctl_mem[index];
e1aab161
GC
1135}
1136
180d8cd9
GC
1137static inline long
1138sk_memory_allocated(const struct sock *sk)
1139{
e805605c 1140 return atomic_long_read(sk->sk_prot->memory_allocated);
180d8cd9
GC
1141}
1142
1143static inline long
e805605c 1144sk_memory_allocated_add(struct sock *sk, int amt)
180d8cd9 1145{
e805605c 1146 return atomic_long_add_return(amt, sk->sk_prot->memory_allocated);
180d8cd9
GC
1147}
1148
1149static inline void
0e90b31f 1150sk_memory_allocated_sub(struct sock *sk, int amt)
180d8cd9 1151{
e805605c 1152 atomic_long_sub(amt, sk->sk_prot->memory_allocated);
180d8cd9
GC
1153}
1154
1155static inline void sk_sockets_allocated_dec(struct sock *sk)
1156{
af95d7df 1157 percpu_counter_dec(sk->sk_prot->sockets_allocated);
180d8cd9
GC
1158}
1159
1160static inline void sk_sockets_allocated_inc(struct sock *sk)
1161{
af95d7df 1162 percpu_counter_inc(sk->sk_prot->sockets_allocated);
180d8cd9
GC
1163}
1164
1165static inline int
1166sk_sockets_allocated_read_positive(struct sock *sk)
1167{
af95d7df 1168 return percpu_counter_read_positive(sk->sk_prot->sockets_allocated);
180d8cd9
GC
1169}
1170
1171static inline int
1172proto_sockets_allocated_sum_positive(struct proto *prot)
1173{
1174 return percpu_counter_sum_positive(prot->sockets_allocated);
1175}
1176
1177static inline long
1178proto_memory_allocated(struct proto *prot)
1179{
1180 return atomic_long_read(prot->memory_allocated);
1181}
1182
1183static inline bool
1184proto_memory_pressure(struct proto *prot)
1185{
1186 if (!prot->memory_pressure)
1187 return false;
1188 return !!*prot->memory_pressure;
1189}
1190
65f76517
ED
1191
1192#ifdef CONFIG_PROC_FS
1da177e4 1193/* Called with local bh disabled */
69336bd2
JP
1194void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1195int sock_prot_inuse_get(struct net *net, struct proto *proto);
65f76517 1196#else
dc6b9b78 1197static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
c29a0bc4 1198 int inc)
65f76517
ED
1199{
1200}
65f76517
ED
1201#endif
1202
1da177e4 1203
614c6cb4
ACM
1204/* With per-bucket locks this operation is not-atomic, so that
1205 * this version is not worse.
1206 */
086c653f 1207static inline int __sk_prot_rehash(struct sock *sk)
614c6cb4
ACM
1208{
1209 sk->sk_prot->unhash(sk);
086c653f 1210 return sk->sk_prot->hash(sk);
614c6cb4
ACM
1211}
1212
fcbdf09d
OP
1213void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
1214
1da177e4
LT
1215/* About 10 seconds */
1216#define SOCK_DESTROY_TIME (10*HZ)
1217
1218/* Sockets 0-1023 can't be bound to unless you are superuser */
1219#define PROT_SOCK 1024
1220
1221#define SHUTDOWN_MASK 3
1222#define RCV_SHUTDOWN 1
1223#define SEND_SHUTDOWN 2
1224
1225#define SOCK_SNDBUF_LOCK 1
1226#define SOCK_RCVBUF_LOCK 2
1227#define SOCK_BINDADDR_LOCK 4
1228#define SOCK_BINDPORT_LOCK 8
1229
1da177e4
LT
1230struct socket_alloc {
1231 struct socket socket;
1232 struct inode vfs_inode;
1233};
1234
1235static inline struct socket *SOCKET_I(struct inode *inode)
1236{
1237 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1238}
1239
1240static inline struct inode *SOCK_INODE(struct socket *socket)
1241{
1242 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1243}
1244
3ab224be
HA
1245/*
1246 * Functions for memory accounting
1247 */
69336bd2 1248int __sk_mem_schedule(struct sock *sk, int size, int kind);
1a24e04e 1249void __sk_mem_reclaim(struct sock *sk, int amount);
1da177e4 1250
3ab224be
HA
1251#define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1252#define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1253#define SK_MEM_SEND 0
1254#define SK_MEM_RECV 1
1da177e4 1255
3ab224be 1256static inline int sk_mem_pages(int amt)
1da177e4 1257{
3ab224be 1258 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
1da177e4
LT
1259}
1260
dc6b9b78 1261static inline bool sk_has_account(struct sock *sk)
1da177e4 1262{
3ab224be
HA
1263 /* return true if protocol supports memory accounting */
1264 return !!sk->sk_prot->memory_allocated;
1da177e4
LT
1265}
1266
dc6b9b78 1267static inline bool sk_wmem_schedule(struct sock *sk, int size)
1da177e4 1268{
3ab224be 1269 if (!sk_has_account(sk))
dc6b9b78 1270 return true;
3ab224be
HA
1271 return size <= sk->sk_forward_alloc ||
1272 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1da177e4
LT
1273}
1274
c76562b6 1275static inline bool
35c448a8 1276sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size)
d80d99d6 1277{
3ab224be 1278 if (!sk_has_account(sk))
dc6b9b78 1279 return true;
c76562b6
MG
1280 return size<= sk->sk_forward_alloc ||
1281 __sk_mem_schedule(sk, size, SK_MEM_RECV) ||
1282 skb_pfmemalloc(skb);
3ab224be
HA
1283}
1284
1285static inline void sk_mem_reclaim(struct sock *sk)
1286{
1287 if (!sk_has_account(sk))
1288 return;
1289 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1a24e04e 1290 __sk_mem_reclaim(sk, sk->sk_forward_alloc);
3ab224be
HA
1291}
1292
9993e7d3
DM
1293static inline void sk_mem_reclaim_partial(struct sock *sk)
1294{
1295 if (!sk_has_account(sk))
1296 return;
1297 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1a24e04e 1298 __sk_mem_reclaim(sk, sk->sk_forward_alloc - 1);
9993e7d3
DM
1299}
1300
3ab224be
HA
1301static inline void sk_mem_charge(struct sock *sk, int size)
1302{
1303 if (!sk_has_account(sk))
1304 return;
1305 sk->sk_forward_alloc -= size;
1306}
1307
1308static inline void sk_mem_uncharge(struct sock *sk, int size)
1309{
1310 if (!sk_has_account(sk))
1311 return;
1312 sk->sk_forward_alloc += size;
1313}
1314
1315static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1316{
3ab224be
HA
1317 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1318 sk->sk_wmem_queued -= skb->truesize;
1319 sk_mem_uncharge(sk, skb->truesize);
1320 __kfree_skb(skb);
d80d99d6
HX
1321}
1322
c3f9b018
ED
1323static inline void sock_release_ownership(struct sock *sk)
1324{
61881cfb
HFS
1325 if (sk->sk_lock.owned) {
1326 sk->sk_lock.owned = 0;
1327
1328 /* The sk_lock has mutex_unlock() semantics: */
1329 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
1330 }
c3f9b018
ED
1331}
1332
ed07536e
PZ
1333/*
1334 * Macro so as to not evaluate some arguments when
1335 * lockdep is not enabled.
1336 *
1337 * Mark both the sk_lock and the sk_lock.slock as a
1338 * per-address-family lock class.
1339 */
dc6b9b78 1340#define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
ed07536e 1341do { \
e8f6fbf6 1342 sk->sk_lock.owned = 0; \
ed07536e
PZ
1343 init_waitqueue_head(&sk->sk_lock.wq); \
1344 spin_lock_init(&(sk)->sk_lock.slock); \
1345 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1346 sizeof((sk)->sk_lock)); \
1347 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
dc6b9b78 1348 (skey), (sname)); \
ed07536e
PZ
1349 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1350} while (0)
1351
b33b0a1b 1352#ifdef CONFIG_LOCKDEP
03be9822 1353static inline bool lockdep_sock_is_held(const struct sock *csk)
1e1d04e6
HFS
1354{
1355 struct sock *sk = (struct sock *)csk;
1356
1357 return lockdep_is_held(&sk->sk_lock) ||
1358 lockdep_is_held(&sk->sk_lock.slock);
1359}
b33b0a1b 1360#endif
1e1d04e6 1361
69336bd2 1362void lock_sock_nested(struct sock *sk, int subclass);
fcc70d5f
PZ
1363
1364static inline void lock_sock(struct sock *sk)
1365{
1366 lock_sock_nested(sk, 0);
1367}
1368
69336bd2 1369void release_sock(struct sock *sk);
1da177e4
LT
1370
1371/* BH context may only use the following locking interface. */
1372#define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
c6366184
IM
1373#define bh_lock_sock_nested(__sk) \
1374 spin_lock_nested(&((__sk)->sk_lock.slock), \
1375 SINGLE_DEPTH_NESTING)
1da177e4
LT
1376#define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1377
69336bd2 1378bool lock_sock_fast(struct sock *sk);
8a74ad60
ED
1379/**
1380 * unlock_sock_fast - complement of lock_sock_fast
1381 * @sk: socket
1382 * @slow: slow mode
1383 *
1384 * fast unlock socket for user context.
1385 * If slow mode is on, we call regular release_sock()
1386 */
1387static inline void unlock_sock_fast(struct sock *sk, bool slow)
4b0b72f7 1388{
8a74ad60
ED
1389 if (slow)
1390 release_sock(sk);
1391 else
1392 spin_unlock_bh(&sk->sk_lock.slock);
4b0b72f7
ED
1393}
1394
fafc4e1e
HFS
1395/* Used by processes to "lock" a socket state, so that
1396 * interrupts and bottom half handlers won't change it
1397 * from under us. It essentially blocks any incoming
1398 * packets, so that we won't get any new data or any
1399 * packets that change the state of the socket.
1400 *
1401 * While locked, BH processing will add new packets to
1402 * the backlog queue. This queue is processed by the
1403 * owner of the socket lock right before it is released.
1404 *
1405 * Since ~2.3.5 it is also exclusive sleep lock serializing
1406 * accesses from user process context.
1407 */
1408
1409static inline bool sock_owned_by_user(const struct sock *sk)
1410{
1411#ifdef CONFIG_LOCKDEP
5e91f6ce 1412 WARN_ON_ONCE(!lockdep_sock_is_held(sk) && debug_locks);
fafc4e1e
HFS
1413#endif
1414 return sk->sk_lock.owned;
1415}
1416
1417/* no reclassification while locks are held */
1418static inline bool sock_allow_reclassification(const struct sock *csk)
1419{
1420 struct sock *sk = (struct sock *)csk;
1421
1422 return !sk->sk_lock.owned && !spin_is_locked(&sk->sk_lock.slock);
1423}
4b0b72f7 1424
69336bd2 1425struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
11aa9c28 1426 struct proto *prot, int kern);
69336bd2 1427void sk_free(struct sock *sk);
eb4cb008 1428void sk_destruct(struct sock *sk);
69336bd2
JP
1429struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority);
1430
1431struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
1432 gfp_t priority);
69336bd2
JP
1433void sock_wfree(struct sk_buff *skb);
1434void skb_orphan_partial(struct sk_buff *skb);
1435void sock_rfree(struct sk_buff *skb);
62bccb8c 1436void sock_efree(struct sk_buff *skb);
82eabd9e 1437#ifdef CONFIG_INET
69336bd2 1438void sock_edemux(struct sk_buff *skb);
82eabd9e
AD
1439#else
1440#define sock_edemux(skb) sock_efree(skb)
1441#endif
69336bd2
JP
1442
1443int sock_setsockopt(struct socket *sock, int level, int op,
1444 char __user *optval, unsigned int optlen);
1445
1446int sock_getsockopt(struct socket *sock, int level, int op,
1447 char __user *optval, int __user *optlen);
1448struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1449 int noblock, int *errcode);
1450struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1451 unsigned long data_len, int noblock,
1452 int *errcode, int max_page_order);
1453void *sock_kmalloc(struct sock *sk, int size, gfp_t priority);
1454void sock_kfree_s(struct sock *sk, void *mem, int size);
79e88659 1455void sock_kzfree_s(struct sock *sk, void *mem, int size);
69336bd2 1456void sk_send_sigurg(struct sock *sk);
1da177e4 1457
f28ea365
EJ
1458struct sockcm_cookie {
1459 u32 mark;
3dd17e63 1460 u16 tsflags;
f28ea365
EJ
1461};
1462
39771b12
WB
1463int __sock_cmsg_send(struct sock *sk, struct msghdr *msg, struct cmsghdr *cmsg,
1464 struct sockcm_cookie *sockc);
f28ea365
EJ
1465int sock_cmsg_send(struct sock *sk, struct msghdr *msg,
1466 struct sockcm_cookie *sockc);
1467
1da177e4
LT
1468/*
1469 * Functions to fill in entries in struct proto_ops when a protocol
1470 * does not implement a particular function.
1471 */
69336bd2
JP
1472int sock_no_bind(struct socket *, struct sockaddr *, int);
1473int sock_no_connect(struct socket *, struct sockaddr *, int, int);
1474int sock_no_socketpair(struct socket *, struct socket *);
1475int sock_no_accept(struct socket *, struct socket *, int);
1476int sock_no_getname(struct socket *, struct sockaddr *, int *, int);
1477unsigned int sock_no_poll(struct file *, struct socket *,
1478 struct poll_table_struct *);
1479int sock_no_ioctl(struct socket *, unsigned int, unsigned long);
1480int sock_no_listen(struct socket *, int);
1481int sock_no_shutdown(struct socket *, int);
1482int sock_no_getsockopt(struct socket *, int , int, char __user *, int __user *);
1483int sock_no_setsockopt(struct socket *, int, int, char __user *, unsigned int);
1b784140
YX
1484int sock_no_sendmsg(struct socket *, struct msghdr *, size_t);
1485int sock_no_recvmsg(struct socket *, struct msghdr *, size_t, int);
69336bd2
JP
1486int sock_no_mmap(struct file *file, struct socket *sock,
1487 struct vm_area_struct *vma);
1488ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset,
1489 size_t size, int flags);
1da177e4
LT
1490
1491/*
1492 * Functions to fill in entries in struct proto_ops when a protocol
1493 * uses the inet style.
1494 */
69336bd2 1495int sock_common_getsockopt(struct socket *sock, int level, int optname,
1da177e4 1496 char __user *optval, int __user *optlen);
1b784140
YX
1497int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
1498 int flags);
69336bd2 1499int sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 1500 char __user *optval, unsigned int optlen);
69336bd2 1501int compat_sock_common_getsockopt(struct socket *sock, int level,
3fdadf7d 1502 int optname, char __user *optval, int __user *optlen);
69336bd2 1503int compat_sock_common_setsockopt(struct socket *sock, int level,
b7058842 1504 int optname, char __user *optval, unsigned int optlen);
1da177e4 1505
69336bd2 1506void sk_common_release(struct sock *sk);
1da177e4
LT
1507
1508/*
1509 * Default socket callbacks and setup code
1510 */
dc6b9b78 1511
1da177e4 1512/* Initialise core socket variables */
69336bd2 1513void sock_init_data(struct socket *sock, struct sock *sk);
1da177e4 1514
1da177e4
LT
1515/*
1516 * Socket reference counting postulates.
1517 *
1518 * * Each user of socket SHOULD hold a reference count.
1519 * * Each access point to socket (an hash table bucket, reference from a list,
1520 * running timer, skb in flight MUST hold a reference count.
1521 * * When reference count hits 0, it means it will never increase back.
1522 * * When reference count hits 0, it means that no references from
1523 * outside exist to this socket and current process on current CPU
1524 * is last user and may/should destroy this socket.
1525 * * sk_free is called from any context: process, BH, IRQ. When
1526 * it is called, socket has no references from outside -> sk_free
1527 * may release descendant resources allocated by the socket, but
1528 * to the time when it is called, socket is NOT referenced by any
1529 * hash tables, lists etc.
1530 * * Packets, delivered from outside (from network or from another process)
1531 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1532 * when they sit in queue. Otherwise, packets will leak to hole, when
1533 * socket is looked up by one cpu and unhasing is made by another CPU.
1534 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1535 * (leak to backlog). Packet socket does all the processing inside
1536 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1537 * use separate SMP lock, so that they are prone too.
1538 */
1539
1540/* Ungrab socket and destroy it, if it was the last reference. */
1541static inline void sock_put(struct sock *sk)
1542{
1543 if (atomic_dec_and_test(&sk->sk_refcnt))
1544 sk_free(sk);
1545}
05dbc7b5 1546/* Generic version of sock_put(), dealing with all sockets
41b822c5 1547 * (TCP_TIMEWAIT, TCP_NEW_SYN_RECV, ESTABLISHED...)
05dbc7b5
ED
1548 */
1549void sock_gen_put(struct sock *sk);
1da177e4 1550
69336bd2 1551int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested);
25995ff5 1552
e022f0b4
KK
1553static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1554{
1555 sk->sk_tx_queue_mapping = tx_queue;
1556}
1557
1558static inline void sk_tx_queue_clear(struct sock *sk)
1559{
1560 sk->sk_tx_queue_mapping = -1;
1561}
1562
1563static inline int sk_tx_queue_get(const struct sock *sk)
1564{
b0f77d0e 1565 return sk ? sk->sk_tx_queue_mapping : -1;
e022f0b4
KK
1566}
1567
972692e0
DM
1568static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1569{
e022f0b4 1570 sk_tx_queue_clear(sk);
972692e0
DM
1571 sk->sk_socket = sock;
1572}
1573
aa395145
ED
1574static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1575{
eaefd110
ED
1576 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1577 return &rcu_dereference_raw(sk->sk_wq)->wait;
aa395145 1578}
1da177e4
LT
1579/* Detach socket from process context.
1580 * Announce socket dead, detach it from wait queue and inode.
1581 * Note that parent inode held reference count on this struct sock,
1582 * we do not release it in this function, because protocol
1583 * probably wants some additional cleanups or even continuing
1584 * to work with this socket (TCP).
1585 */
1586static inline void sock_orphan(struct sock *sk)
1587{
1588 write_lock_bh(&sk->sk_callback_lock);
1589 sock_set_flag(sk, SOCK_DEAD);
972692e0 1590 sk_set_socket(sk, NULL);
43815482 1591 sk->sk_wq = NULL;
1da177e4
LT
1592 write_unlock_bh(&sk->sk_callback_lock);
1593}
1594
1595static inline void sock_graft(struct sock *sk, struct socket *parent)
1596{
1597 write_lock_bh(&sk->sk_callback_lock);
eaefd110 1598 sk->sk_wq = parent->wq;
1da177e4 1599 parent->sk = sk;
972692e0 1600 sk_set_socket(sk, parent);
4237c75c 1601 security_sock_graft(sk, parent);
1da177e4
LT
1602 write_unlock_bh(&sk->sk_callback_lock);
1603}
1604
69336bd2
JP
1605kuid_t sock_i_uid(struct sock *sk);
1606unsigned long sock_i_ino(struct sock *sk);
1da177e4 1607
58d607d3 1608static inline u32 net_tx_rndhash(void)
877d1f62 1609{
58d607d3
ED
1610 u32 v = prandom_u32();
1611
1612 return v ?: 1;
1613}
877d1f62 1614
58d607d3
ED
1615static inline void sk_set_txhash(struct sock *sk)
1616{
1617 sk->sk_txhash = net_tx_rndhash();
877d1f62
TH
1618}
1619
265f94ff
TH
1620static inline void sk_rethink_txhash(struct sock *sk)
1621{
1622 if (sk->sk_txhash)
1623 sk_set_txhash(sk);
1624}
1625
1da177e4
LT
1626static inline struct dst_entry *
1627__sk_dst_get(struct sock *sk)
1628{
1e1d04e6
HFS
1629 return rcu_dereference_check(sk->sk_dst_cache,
1630 lockdep_sock_is_held(sk));
1da177e4
LT
1631}
1632
1633static inline struct dst_entry *
1634sk_dst_get(struct sock *sk)
1635{
1636 struct dst_entry *dst;
1637
b6c6712a
ED
1638 rcu_read_lock();
1639 dst = rcu_dereference(sk->sk_dst_cache);
f8864972
ED
1640 if (dst && !atomic_inc_not_zero(&dst->__refcnt))
1641 dst = NULL;
b6c6712a 1642 rcu_read_unlock();
1da177e4
LT
1643 return dst;
1644}
1645
b6c6712a
ED
1646static inline void dst_negative_advice(struct sock *sk)
1647{
1648 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1649
265f94ff
TH
1650 sk_rethink_txhash(sk);
1651
b6c6712a
ED
1652 if (dst && dst->ops->negative_advice) {
1653 ndst = dst->ops->negative_advice(dst);
1654
1655 if (ndst != dst) {
1656 rcu_assign_pointer(sk->sk_dst_cache, ndst);
0a6957e7 1657 sk_tx_queue_clear(sk);
b6c6712a
ED
1658 }
1659 }
1660}
1661
1da177e4
LT
1662static inline void
1663__sk_dst_set(struct sock *sk, struct dst_entry *dst)
1664{
1665 struct dst_entry *old_dst;
1666
e022f0b4 1667 sk_tx_queue_clear(sk);
0b53ff2e
ED
1668 /*
1669 * This can be called while sk is owned by the caller only,
1670 * with no state that can be checked in a rcu_dereference_check() cond
1671 */
1672 old_dst = rcu_dereference_raw(sk->sk_dst_cache);
b6c6712a 1673 rcu_assign_pointer(sk->sk_dst_cache, dst);
1da177e4
LT
1674 dst_release(old_dst);
1675}
1676
1677static inline void
1678sk_dst_set(struct sock *sk, struct dst_entry *dst)
1679{
7f502361
ED
1680 struct dst_entry *old_dst;
1681
1682 sk_tx_queue_clear(sk);
5925a055 1683 old_dst = xchg((__force struct dst_entry **)&sk->sk_dst_cache, dst);
7f502361 1684 dst_release(old_dst);
1da177e4
LT
1685}
1686
1687static inline void
1688__sk_dst_reset(struct sock *sk)
1689{
b6c6712a 1690 __sk_dst_set(sk, NULL);
1da177e4
LT
1691}
1692
1693static inline void
1694sk_dst_reset(struct sock *sk)
1695{
7f502361 1696 sk_dst_set(sk, NULL);
1da177e4
LT
1697}
1698
69336bd2 1699struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1da177e4 1700
69336bd2 1701struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1da177e4 1702
f60e5990 1703bool sk_mc_loop(struct sock *sk);
1704
dc6b9b78 1705static inline bool sk_can_gso(const struct sock *sk)
bcd76111
HX
1706{
1707 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1708}
1709
69336bd2 1710void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
6cbb0df7 1711
c8f44aff 1712static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
a465419b
ED
1713{
1714 sk->sk_route_nocaps |= flags;
1715 sk->sk_route_caps &= ~flags;
1716}
1717
9a49850d
TH
1718static inline bool sk_check_csum_caps(struct sock *sk)
1719{
1720 return (sk->sk_route_caps & NETIF_F_HW_CSUM) ||
1721 (sk->sk_family == PF_INET &&
1722 (sk->sk_route_caps & NETIF_F_IP_CSUM)) ||
1723 (sk->sk_family == PF_INET6 &&
1724 (sk->sk_route_caps & NETIF_F_IPV6_CSUM));
1725}
1726
c6e1a0d1 1727static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
57be5bda 1728 struct iov_iter *from, char *to,
912d398d 1729 int copy, int offset)
c6e1a0d1
TH
1730{
1731 if (skb->ip_summed == CHECKSUM_NONE) {
57be5bda
AV
1732 __wsum csum = 0;
1733 if (csum_and_copy_from_iter(to, copy, &csum, from) != copy)
1734 return -EFAULT;
912d398d 1735 skb->csum = csum_block_add(skb->csum, csum, offset);
c6e1a0d1 1736 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
57be5bda 1737 if (copy_from_iter_nocache(to, copy, from) != copy)
c6e1a0d1 1738 return -EFAULT;
57be5bda 1739 } else if (copy_from_iter(to, copy, from) != copy)
c6e1a0d1
TH
1740 return -EFAULT;
1741
1742 return 0;
1743}
1744
1745static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
57be5bda 1746 struct iov_iter *from, int copy)
c6e1a0d1 1747{
912d398d 1748 int err, offset = skb->len;
c6e1a0d1 1749
912d398d
WY
1750 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1751 copy, offset);
c6e1a0d1 1752 if (err)
912d398d 1753 __skb_trim(skb, offset);
c6e1a0d1
TH
1754
1755 return err;
1756}
1757
57be5bda 1758static inline int skb_copy_to_page_nocache(struct sock *sk, struct iov_iter *from,
c6e1a0d1
TH
1759 struct sk_buff *skb,
1760 struct page *page,
1761 int off, int copy)
1762{
1763 int err;
1764
912d398d
WY
1765 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1766 copy, skb->len);
c6e1a0d1
TH
1767 if (err)
1768 return err;
1769
1770 skb->len += copy;
1771 skb->data_len += copy;
1772 skb->truesize += copy;
1773 sk->sk_wmem_queued += copy;
1774 sk_mem_charge(sk, copy);
1775 return 0;
1776}
1777
c564039f
ED
1778/**
1779 * sk_wmem_alloc_get - returns write allocations
1780 * @sk: socket
1781 *
1782 * Returns sk_wmem_alloc minus initial offset of one
1783 */
1784static inline int sk_wmem_alloc_get(const struct sock *sk)
1785{
1786 return atomic_read(&sk->sk_wmem_alloc) - 1;
1787}
1788
1789/**
1790 * sk_rmem_alloc_get - returns read allocations
1791 * @sk: socket
1792 *
1793 * Returns sk_rmem_alloc
1794 */
1795static inline int sk_rmem_alloc_get(const struct sock *sk)
1796{
1797 return atomic_read(&sk->sk_rmem_alloc);
1798}
1799
1800/**
1801 * sk_has_allocations - check if allocations are outstanding
1802 * @sk: socket
1803 *
1804 * Returns true if socket has write or read allocations
1805 */
dc6b9b78 1806static inline bool sk_has_allocations(const struct sock *sk)
c564039f
ED
1807{
1808 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1809}
1810
a57de0b4 1811/**
1ce0bf50 1812 * skwq_has_sleeper - check if there are any waiting processes
acfbe96a 1813 * @wq: struct socket_wq
a57de0b4 1814 *
43815482 1815 * Returns true if socket_wq has waiting processes
a57de0b4 1816 *
1ce0bf50 1817 * The purpose of the skwq_has_sleeper and sock_poll_wait is to wrap the memory
a57de0b4
JO
1818 * barrier call. They were added due to the race found within the tcp code.
1819 *
1820 * Consider following tcp code paths:
1821 *
1822 * CPU1 CPU2
1823 *
1824 * sys_select receive packet
1825 * ... ...
1826 * __add_wait_queue update tp->rcv_nxt
1827 * ... ...
1828 * tp->rcv_nxt check sock_def_readable
1829 * ... {
43815482
ED
1830 * schedule rcu_read_lock();
1831 * wq = rcu_dereference(sk->sk_wq);
1832 * if (wq && waitqueue_active(&wq->wait))
1833 * wake_up_interruptible(&wq->wait)
a57de0b4
JO
1834 * ...
1835 * }
1836 *
1837 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1838 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1839 * could then endup calling schedule and sleep forever if there are no more
1840 * data on the socket.
ad462769 1841 *
a57de0b4 1842 */
1ce0bf50 1843static inline bool skwq_has_sleeper(struct socket_wq *wq)
a57de0b4 1844{
1ce0bf50 1845 return wq && wq_has_sleeper(&wq->wait);
a57de0b4
JO
1846}
1847
1848/**
1849 * sock_poll_wait - place memory barrier behind the poll_wait call.
1850 * @filp: file
1851 * @wait_address: socket wait queue
1852 * @p: poll_table
1853 *
43815482 1854 * See the comments in the wq_has_sleeper function.
a57de0b4
JO
1855 */
1856static inline void sock_poll_wait(struct file *filp,
1857 wait_queue_head_t *wait_address, poll_table *p)
1858{
626cf236 1859 if (!poll_does_not_wait(p) && wait_address) {
a57de0b4 1860 poll_wait(filp, wait_address, p);
dc6b9b78 1861 /* We need to be sure we are in sync with the
a57de0b4
JO
1862 * socket flags modification.
1863 *
43815482 1864 * This memory barrier is paired in the wq_has_sleeper.
dc6b9b78 1865 */
a57de0b4
JO
1866 smp_mb();
1867 }
1868}
1869
b73c3d0e
TH
1870static inline void skb_set_hash_from_sk(struct sk_buff *skb, struct sock *sk)
1871{
1872 if (sk->sk_txhash) {
1873 skb->l4_hash = 1;
1874 skb->hash = sk->sk_txhash;
1875 }
1876}
1877
9e17f8a4
ED
1878void skb_set_owner_w(struct sk_buff *skb, struct sock *sk);
1879
1da177e4 1880/*
dc6b9b78 1881 * Queue a received datagram if it will fit. Stream and sequenced
1da177e4
LT
1882 * protocols can't normally use this as they need to fit buffers in
1883 * and play with them.
1884 *
dc6b9b78 1885 * Inlined as it's very short and called for pretty much every
1da177e4
LT
1886 * packet ever received.
1887 */
1da177e4
LT
1888static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1889{
d55d87fd 1890 skb_orphan(skb);
1da177e4
LT
1891 skb->sk = sk;
1892 skb->destructor = sock_rfree;
1893 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
3ab224be 1894 sk_mem_charge(sk, skb->truesize);
1da177e4
LT
1895}
1896
69336bd2
JP
1897void sk_reset_timer(struct sock *sk, struct timer_list *timer,
1898 unsigned long expires);
1da177e4 1899
69336bd2 1900void sk_stop_timer(struct sock *sk, struct timer_list *timer);
1da177e4 1901
e6afc8ac 1902int __sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
69336bd2 1903int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
1da177e4 1904
69336bd2 1905int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
364a9e93 1906struct sk_buff *sock_dequeue_err_skb(struct sock *sk);
1da177e4
LT
1907
1908/*
1909 * Recover an error report and clear atomically
1910 */
dc6b9b78 1911
1da177e4
LT
1912static inline int sock_error(struct sock *sk)
1913{
c1cbe4b7
BL
1914 int err;
1915 if (likely(!sk->sk_err))
1916 return 0;
1917 err = xchg(&sk->sk_err, 0);
1da177e4
LT
1918 return -err;
1919}
1920
1921static inline unsigned long sock_wspace(struct sock *sk)
1922{
1923 int amt = 0;
1924
1925 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1926 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
dc6b9b78 1927 if (amt < 0)
1da177e4
LT
1928 amt = 0;
1929 }
1930 return amt;
1931}
1932
ceb5d58b
ED
1933/* Note:
1934 * We use sk->sk_wq_raw, from contexts knowing this
1935 * pointer is not NULL and cannot disappear/change.
1936 */
9cd3e072 1937static inline void sk_set_bit(int nr, struct sock *sk)
1da177e4 1938{
ceb5d58b 1939 set_bit(nr, &sk->sk_wq_raw->flags);
9cd3e072
ED
1940}
1941
1942static inline void sk_clear_bit(int nr, struct sock *sk)
1943{
ceb5d58b 1944 clear_bit(nr, &sk->sk_wq_raw->flags);
9cd3e072
ED
1945}
1946
ceb5d58b 1947static inline void sk_wake_async(const struct sock *sk, int how, int band)
1da177e4 1948{
ceb5d58b
ED
1949 if (sock_flag(sk, SOCK_FASYNC)) {
1950 rcu_read_lock();
1951 sock_wake_async(rcu_dereference(sk->sk_wq), how, band);
1952 rcu_read_unlock();
1953 }
1da177e4
LT
1954}
1955
eea86af6
DB
1956/* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
1957 * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
1958 * Note: for send buffers, TCP works better if we can build two skbs at
1959 * minimum.
7a91b434 1960 */
9eb5bf83 1961#define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
eea86af6
DB
1962
1963#define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2)
1964#define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE
1da177e4
LT
1965
1966static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1967{
1968 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
8df09ea3 1969 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
eea86af6 1970 sk->sk_sndbuf = max_t(u32, sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1da177e4
LT
1971 }
1972}
1973
eb934478
ED
1974struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp,
1975 bool force_schedule);
1da177e4 1976
5640f768
ED
1977/**
1978 * sk_page_frag - return an appropriate page_frag
1979 * @sk: socket
1980 *
1981 * If socket allocation mode allows current thread to sleep, it means its
1982 * safe to use the per task page_frag instead of the per socket one.
1983 */
1984static inline struct page_frag *sk_page_frag(struct sock *sk)
1da177e4 1985{
d0164adc 1986 if (gfpflags_allow_blocking(sk->sk_allocation))
5640f768 1987 return &current->task_frag;
1da177e4 1988
5640f768 1989 return &sk->sk_frag;
1da177e4
LT
1990}
1991
69336bd2 1992bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag);
5640f768 1993
1da177e4
LT
1994/*
1995 * Default write policy as shown to user space via poll/select/SIGIO
1996 */
dc6b9b78 1997static inline bool sock_writeable(const struct sock *sk)
1da177e4 1998{
8df09ea3 1999 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
1da177e4
LT
2000}
2001
dd0fc66f 2002static inline gfp_t gfp_any(void)
1da177e4 2003{
99709372 2004 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1da177e4
LT
2005}
2006
dc6b9b78 2007static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
1da177e4
LT
2008{
2009 return noblock ? 0 : sk->sk_rcvtimeo;
2010}
2011
dc6b9b78 2012static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
1da177e4
LT
2013{
2014 return noblock ? 0 : sk->sk_sndtimeo;
2015}
2016
2017static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2018{
2019 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
2020}
2021
2022/* Alas, with timeout socket operations are not restartable.
2023 * Compare this to poll().
2024 */
2025static inline int sock_intr_errno(long timeo)
2026{
2027 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2028}
2029
744d5a3e
EB
2030struct sock_skb_cb {
2031 u32 dropcount;
2032};
2033
2034/* Store sock_skb_cb at the end of skb->cb[] so protocol families
2035 * using skb->cb[] would keep using it directly and utilize its
2036 * alignement guarantee.
2037 */
2038#define SOCK_SKB_CB_OFFSET ((FIELD_SIZEOF(struct sk_buff, cb) - \
2039 sizeof(struct sock_skb_cb)))
2040
2041#define SOCK_SKB_CB(__skb) ((struct sock_skb_cb *)((__skb)->cb + \
2042 SOCK_SKB_CB_OFFSET))
2043
b4772ef8 2044#define sock_skb_cb_check_size(size) \
744d5a3e 2045 BUILD_BUG_ON((size) > SOCK_SKB_CB_OFFSET)
b4772ef8 2046
3bc3b96f
EB
2047static inline void
2048sock_skb_set_dropcount(const struct sock *sk, struct sk_buff *skb)
2049{
744d5a3e 2050 SOCK_SKB_CB(skb)->dropcount = atomic_read(&sk->sk_drops);
3bc3b96f
EB
2051}
2052
532182cd
ED
2053static inline void sk_drops_add(struct sock *sk, const struct sk_buff *skb)
2054{
2055 int segs = max_t(u16, 1, skb_shinfo(skb)->gso_segs);
2056
2057 atomic_add(segs, &sk->sk_drops);
2058}
2059
69336bd2
JP
2060void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2061 struct sk_buff *skb);
2062void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2063 struct sk_buff *skb);
92f37fd2 2064
dc6b9b78 2065static inline void
1da177e4
LT
2066sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2067{
b7aa0bf7 2068 ktime_t kt = skb->tstamp;
20d49473 2069 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
a61bbcf2 2070
20d49473
PO
2071 /*
2072 * generate control messages if
b9f40e21 2073 * - receive time stamping in software requested
20d49473 2074 * - software time stamp available and wanted
20d49473 2075 * - hardware time stamps available and wanted
20d49473
PO
2076 */
2077 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
b9f40e21 2078 (sk->sk_tsflags & SOF_TIMESTAMPING_RX_SOFTWARE) ||
c199105d 2079 (kt.tv64 && sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) ||
20d49473 2080 (hwtstamps->hwtstamp.tv64 &&
b9f40e21 2081 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE)))
92f37fd2
ED
2082 __sock_recv_timestamp(msg, sk, skb);
2083 else
b7aa0bf7 2084 sk->sk_stamp = kt;
6e3e939f
JB
2085
2086 if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2087 __sock_recv_wifi_status(msg, sk, skb);
1da177e4
LT
2088}
2089
69336bd2
JP
2090void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2091 struct sk_buff *skb);
767dd033
ED
2092
2093static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2094 struct sk_buff *skb)
2095{
2096#define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
b9f40e21
WB
2097 (1UL << SOCK_RCVTSTAMP))
2098#define TSFLAGS_ANY (SOF_TIMESTAMPING_SOFTWARE | \
2099 SOF_TIMESTAMPING_RAW_HARDWARE)
767dd033 2100
b9f40e21 2101 if (sk->sk_flags & FLAGS_TS_OR_DROPS || sk->sk_tsflags & TSFLAGS_ANY)
767dd033
ED
2102 __sock_recv_ts_and_drops(msg, sk, skb);
2103 else
2104 sk->sk_stamp = skb->tstamp;
2105}
3b885787 2106
c14ac945 2107void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags);
67cc0d40 2108
20d49473
PO
2109/**
2110 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
20d49473 2111 * @sk: socket sending this packet
c14ac945 2112 * @tsflags: timestamping flags to use
140c55d4
ED
2113 * @tx_flags: completed with instructions for time stamping
2114 *
2115 * Note : callers should take care of initial *tx_flags value (usually 0)
20d49473 2116 */
c14ac945
SHY
2117static inline void sock_tx_timestamp(const struct sock *sk, __u16 tsflags,
2118 __u8 *tx_flags)
67cc0d40 2119{
c14ac945
SHY
2120 if (unlikely(tsflags))
2121 __sock_tx_timestamp(tsflags, tx_flags);
67cc0d40
WB
2122 if (unlikely(sock_flag(sk, SOCK_WIFI_STATUS)))
2123 *tx_flags |= SKBTX_WIFI_STATUS;
2124}
20d49473 2125
1da177e4
LT
2126/**
2127 * sk_eat_skb - Release a skb if it is no longer needed
4dc3b16b
PP
2128 * @sk: socket to eat this skb from
2129 * @skb: socket buffer to eat
1da177e4
LT
2130 *
2131 * This routine must be called with interrupts disabled or with the socket
2132 * locked so that the sk_buff queue operation is ok.
2133*/
7bced397 2134static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb)
1da177e4
LT
2135{
2136 __skb_unlink(skb, &sk->sk_receive_queue);
2137 __kfree_skb(skb);
2138}
2139
3b1e0a65
YH
2140static inline
2141struct net *sock_net(const struct sock *sk)
2142{
c2d9ba9b 2143 return read_pnet(&sk->sk_net);
3b1e0a65
YH
2144}
2145
2146static inline
f5aa23fd 2147void sock_net_set(struct sock *sk, struct net *net)
3b1e0a65 2148{
c2d9ba9b 2149 write_pnet(&sk->sk_net, net);
3b1e0a65
YH
2150}
2151
23542618
KK
2152static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2153{
efc27f8c 2154 if (skb->sk) {
23542618
KK
2155 struct sock *sk = skb->sk;
2156
2157 skb->destructor = NULL;
2158 skb->sk = NULL;
2159 return sk;
2160 }
2161 return NULL;
2162}
2163
1d0ab253
ED
2164/* This helper checks if a socket is a full socket,
2165 * ie _not_ a timewait or request socket.
2166 */
2167static inline bool sk_fullsock(const struct sock *sk)
2168{
2169 return (1 << sk->sk_state) & ~(TCPF_TIME_WAIT | TCPF_NEW_SYN_RECV);
2170}
2171
e446f9df
ED
2172/* This helper checks if a socket is a LISTEN or NEW_SYN_RECV
2173 * SYNACK messages can be attached to either ones (depending on SYNCOOKIE)
2174 */
2175static inline bool sk_listener(const struct sock *sk)
2176{
2177 return (1 << sk->sk_state) & (TCPF_LISTEN | TCPF_NEW_SYN_RECV);
2178}
2179
00fd38d9
ED
2180/**
2181 * sk_state_load - read sk->sk_state for lockless contexts
2182 * @sk: socket pointer
2183 *
2184 * Paired with sk_state_store(). Used in places we do not hold socket lock :
2185 * tcp_diag_get_info(), tcp_get_info(), tcp_poll(), get_tcp4_sock() ...
2186 */
2187static inline int sk_state_load(const struct sock *sk)
2188{
2189 return smp_load_acquire(&sk->sk_state);
2190}
2191
2192/**
2193 * sk_state_store - update sk->sk_state
2194 * @sk: socket pointer
2195 * @newstate: new state
2196 *
2197 * Paired with sk_state_load(). Should be used in contexts where
2198 * state change might impact lockless readers.
2199 */
2200static inline void sk_state_store(struct sock *sk, int newstate)
2201{
2202 smp_store_release(&sk->sk_state, newstate);
2203}
2204
69336bd2
JP
2205void sock_enable_timestamp(struct sock *sk, int flag);
2206int sock_get_timestamp(struct sock *, struct timeval __user *);
2207int sock_get_timestampns(struct sock *, struct timespec __user *);
2208int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len, int level,
2209 int type);
1da177e4 2210
a3b299da
EB
2211bool sk_ns_capable(const struct sock *sk,
2212 struct user_namespace *user_ns, int cap);
2213bool sk_capable(const struct sock *sk, int cap);
2214bool sk_net_capable(const struct sock *sk, int cap);
2215
1da177e4
LT
2216extern __u32 sysctl_wmem_max;
2217extern __u32 sysctl_rmem_max;
2218
b245be1f 2219extern int sysctl_tstamp_allow_data;
6baf1f41
DM
2220extern int sysctl_optmem_max;
2221
20380731
ACM
2222extern __u32 sysctl_wmem_default;
2223extern __u32 sysctl_rmem_default;
20380731 2224
1da177e4 2225#endif /* _SOCK_H */
This page took 1.53781 seconds and 4 git commands to generate.