blob: 73c6b008f1b768e6ca51a00532723c3739bb02bf [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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 *
Jesper Juhl02c30a82005-05-05 16:16:16 -070010 * Authors: Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -070011 * 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
Alexey Dobriyana6b7a402011-06-06 10:43:46 +000043#include <linux/hardirq.h>
Ilpo Järvinen172589c2007-08-28 15:50:33 -070044#include <linux/kernel.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070045#include <linux/list.h>
Eric Dumazet88ab1932008-11-16 19:39:21 -080046#include <linux/list_nulls.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070047#include <linux/timer.h>
48#include <linux/cache.h>
Glauber Costa3f134612012-05-29 15:07:11 -070049#include <linux/bitops.h>
Ingo Molnara5b5bb92006-07-03 00:25:35 -070050#include <linux/lockdep.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070051#include <linux/netdevice.h>
52#include <linux/skbuff.h> /* struct sk_buff */
Al Virod7fe0f22006-12-03 23:15:30 -050053#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070054#include <linux/security.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090055#include <linux/slab.h>
Tom Herbertc6e1a0d2011-04-04 22:30:30 -070056#include <linux/uaccess.h>
Johannes Weiner3e32cb22014-12-10 15:42:31 -080057#include <linux/page_counter.h>
Glauber Costa180d8cd2011-12-11 21:47:02 +000058#include <linux/memcontrol.h>
Ingo Molnarc5905af2012-02-24 08:31:31 +010059#include <linux/static_key.h>
Al Viro40401532012-02-13 03:58:52 +000060#include <linux/sched.h>
Herbert Xu1ce0bf52015-11-26 13:55:39 +080061#include <linux/wait.h>
Tejun Heo2a56a1f2015-12-07 17:38:52 -050062#include <linux/cgroup-defs.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070063
64#include <linux/filter.h>
Eric Dumazet88ab1932008-11-16 19:39:21 -080065#include <linux/rculist_nulls.h>
Jiri Olsaa57de0b2009-07-08 12:09:13 +000066#include <linux/poll.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070067
Eric Dumazetc31504d2010-11-15 19:58:26 +000068#include <linux/atomic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070069#include <net/dst.h>
70#include <net/checksum.h>
Eric Dumazet1d0ab252015-03-15 21:12:12 -070071#include <net/tcp_states.h>
Willem de Bruijnb9f40e22014-08-04 22:11:46 -040072#include <linux/net_tstamp.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070073
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
Stephen Hemminger4cd90292008-03-21 15:54:53 -070086/* Validate arguments and do nothing */
Joe Perchesb9075fa2011-10-31 17:11:33 -070087static inline __printf(2, 3)
Eric Dumazetdc6b9b72012-05-16 22:48:15 +000088void SOCK_DEBUG(const struct sock *sk, const char *msg, ...)
Stephen Hemminger4cd90292008-03-21 15:54:53 -070089{
90}
Linus Torvalds1da177e2005-04-16 15:20:36 -070091#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 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070097typedef struct {
98 spinlock_t slock;
John Heffnerd2e91172007-09-12 10:44:19 +020099 int owned;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100 wait_queue_head_t wq;
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700101 /*
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
Linus Torvalds1da177e2005-04-16 15:20:36 -0700110} socket_lock_t;
111
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112struct sock;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700113struct proto;
Denis V. Lunev0eeb8ff2007-12-04 01:15:45 -0800114struct net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115
Eric Dumazet077b3932012-12-02 07:33:10 +0000116typedef __u32 __bitwise __portpair;
117typedef __u64 __bitwise __addrpair;
118
Linus Torvalds1da177e2005-04-16 15:20:36 -0700119/**
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700120 * struct sock_common - minimal network layer representation of sockets
Eric Dumazet68835ab2010-11-30 19:04:07 +0000121 * @skc_daddr: Foreign IPv4 addr
122 * @skc_rcv_saddr: Bound local IPv4 addr
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000123 * @skc_hash: hash value used with various protocol lookup tables
Eric Dumazetd4cada42009-11-08 10:17:30 +0000124 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
Eric Dumazetce43b032012-11-30 09:49:27 +0000125 * @skc_dport: placeholder for inet_dport/tw_dport
126 * @skc_num: placeholder for inet_num/tw_num
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700127 * @skc_family: network address family
128 * @skc_state: Connection state
129 * @skc_reuse: %SO_REUSEADDR setting
Tom Herbert055dc212013-01-22 09:49:50 +0000130 * @skc_reuseport: %SO_REUSEPORT setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700131 * @skc_bound_dev_if: bound device index if != 0
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700132 * @skc_bind_node: bind hash linkage for various protocol lookup tables
Eric Dumazet512615b2009-11-08 10:17:58 +0000133 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700134 * @skc_prot: protocol handlers inside a network family
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200135 * @skc_net: reference to the network namespace of this socket
Eric Dumazet68835ab2010-11-30 19:04:07 +0000136 * @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
Eric Dumazet8e5eb542015-10-08 19:33:22 -0700139 * @skc_flags: place holder for sk_flags
140 * %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
141 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
Eric Dumazet70da2682015-10-08 19:33:21 -0700142 * @skc_incoming_cpu: record/match cpu processing incoming packets
Eric Dumazet68835ab2010-11-30 19:04:07 +0000143 * @skc_refcnt: reference count
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700144 *
145 * This is the minimal network layer representation of sockets, the header
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700146 * for struct sock and struct inet_timewait_sock.
147 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148struct sock_common {
Eric Dumazetce43b032012-11-30 09:49:27 +0000149 /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned
Eric Dumazet05dbc7b2013-10-03 00:22:02 -0700150 * address on 64bit arches : cf INET_MATCH()
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000151 */
Eric Dumazetce43b032012-11-30 09:49:27 +0000152 union {
Eric Dumazet077b3932012-12-02 07:33:10 +0000153 __addrpair skc_addrpair;
Eric Dumazetce43b032012-11-30 09:49:27 +0000154 struct {
155 __be32 skc_daddr;
156 __be32 skc_rcv_saddr;
157 };
158 };
Eric Dumazetd4cada42009-11-08 10:17:30 +0000159 union {
160 unsigned int skc_hash;
161 __u16 skc_u16hashes[2];
162 };
Eric Dumazetce43b032012-11-30 09:49:27 +0000163 /* skc_dport && skc_num must be grouped as well */
164 union {
Eric Dumazet077b3932012-12-02 07:33:10 +0000165 __portpair skc_portpair;
Eric Dumazetce43b032012-11-30 09:49:27 +0000166 struct {
167 __be16 skc_dport;
168 __u16 skc_num;
169 };
170 };
171
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000172 unsigned short skc_family;
173 volatile unsigned char skc_state;
Tom Herbert055dc212013-01-22 09:49:50 +0000174 unsigned char skc_reuse:4;
Eric Dumazet9fe516b2014-06-27 08:36:16 -0700175 unsigned char skc_reuseport:1;
176 unsigned char skc_ipv6only:1;
Eric W. Biederman26abe142015-05-08 21:10:31 -0500177 unsigned char skc_net_refcnt:1;
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000178 int skc_bound_dev_if;
Eric Dumazet512615b2009-11-08 10:17:58 +0000179 union {
180 struct hlist_node skc_bind_node;
Eric Dumazetca065d02016-04-01 08:52:13 -0700181 struct hlist_node skc_portaddr_node;
Eric Dumazet512615b2009-11-08 10:17:58 +0000182 };
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700183 struct proto *skc_prot;
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -0500184 possible_net_t skc_net;
Eric Dumazetefe42082013-10-03 15:42:29 -0700185
186#if IS_ENABLED(CONFIG_IPV6)
187 struct in6_addr skc_v6_daddr;
188 struct in6_addr skc_v6_rcv_saddr;
189#endif
190
Eric Dumazet33cf7c92015-03-11 18:53:14 -0700191 atomic64_t skc_cookie;
192
Eric Dumazet8e5eb542015-10-08 19:33:22 -0700193 /* 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 };
Eric Dumazet68835ab2010-11-30 19:04:07 +0000203 /*
204 * fields between dontcopy_begin/dontcopy_end
205 * are not copied in sock_copy()
206 */
Randy Dunlap928c41e2011-01-08 17:39:21 +0000207 /* private: */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000208 int skc_dontcopy_begin[0];
Randy Dunlap928c41e2011-01-08 17:39:21 +0000209 /* public: */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000210 union {
211 struct hlist_node skc_node;
212 struct hlist_nulls_node skc_nulls_node;
213 };
214 int skc_tx_queue_mapping;
Eric Dumazeted53d0a2015-10-08 19:33:23 -0700215 union {
216 int skc_incoming_cpu;
217 u32 skc_rcv_wnd;
Eric Dumazetd475f092015-10-08 19:33:24 -0700218 u32 skc_tw_rcv_nxt; /* struct tcp_timewait_sock */
Eric Dumazeted53d0a2015-10-08 19:33:23 -0700219 };
Eric Dumazet70da2682015-10-08 19:33:21 -0700220
Eric Dumazet68835ab2010-11-30 19:04:07 +0000221 atomic_t skc_refcnt;
Randy Dunlap928c41e2011-01-08 17:39:21 +0000222 /* private: */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000223 int skc_dontcopy_end[0];
Eric Dumazeted53d0a2015-10-08 19:33:23 -0700224 union {
225 u32 skc_rxhash;
226 u32 skc_window_clamp;
Eric Dumazetd475f092015-10-08 19:33:24 -0700227 u32 skc_tw_snd_nxt; /* struct tcp_timewait_sock */
Eric Dumazeted53d0a2015-10-08 19:33:23 -0700228 };
Randy Dunlap928c41e2011-01-08 17:39:21 +0000229 /* public: */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230};
231
232/**
233 * struct sock - network layer representation of sockets
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700234 * @__sk_common: shared layout with inet_timewait_sock
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700235 * @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
Eric Dumazet43815482010-04-29 11:01:49 +0000239 * @sk_wq: sock wait queue and async head
Shawn Bohrer421b3882013-10-07 11:01:39 -0500240 * @sk_rx_dst: receive input route used by early demux
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700241 * @sk_dst_cache: destination cache
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700242 * @sk_policy: flow policy
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700243 * @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
Eliezer Tamir06021292013-06-10 11:39:50 +0300249 * @sk_napi_id: id of the last napi context to receive data for sk
Eliezer Tamirdafcc432013-06-14 16:33:57 +0300250 * @sk_ll_usec: usecs to busypoll when there is no data
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700251 * @sk_allocation: allocation mode
Eric Dumazet95bd09e2013-08-27 05:46:32 -0700252 * @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
Eric Dumazetc3f40d72013-09-29 01:12:40 -0700253 * @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE)
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700254 * @sk_sndbuf: size of send buffer in bytes
Stephen Hemminger293de7d2016-10-23 09:28:29 -0700255 * @sk_padding: unused element for alignment
Tom Herbert28448b82014-05-23 08:47:19 -0700256 * @sk_no_check_tx: %SO_NO_CHECK setting, set checksum in TX packets
257 * @sk_no_check_rx: allow zero checksum in RX packets
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700258 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
Eric Dumazeta4654192010-05-16 00:36:33 -0700259 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
Herbert Xubcd76112006-06-30 13:36:35 -0700260 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -0700261 * @sk_gso_max_size: Maximum GSO segment size to build
Ben Hutchings14853482012-07-30 16:11:42 +0000262 * @sk_gso_max_segs: Maximum number of GSO segments
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700263 * @sk_lingertime: %SO_LINGER l_linger setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700264 * @sk_backlog: always used with the per-socket spinlock held
265 * @sk_callback_lock: used with the callbacks in the end of this struct
266 * @sk_error_queue: rarely used
Wang Chen33c732c2007-11-13 20:30:01 -0800267 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
268 * IPV6_ADDRFORM for instance)
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700269 * @sk_err: last error
Wang Chen33c732c2007-11-13 20:30:01 -0800270 * @sk_err_soft: errors that don't cause failure but are the cause of a
271 * persistent failure not just 'timed out'
Eric Dumazetcb61cb92008-06-17 21:04:56 -0700272 * @sk_drops: raw/udp drops counter
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700273 * @sk_ack_backlog: current listen backlog
274 * @sk_max_ack_backlog: listen backlog set in listen()
275 * @sk_priority: %SO_PRIORITY setting
276 * @sk_type: socket type (%SOCK_STREAM, etc)
277 * @sk_protocol: which protocol this socket belongs in this network family
Randy Dunlap53c3fa22010-08-09 13:41:07 +0000278 * @sk_peer_pid: &struct pid for this socket's peer
279 * @sk_peer_cred: %SO_PEERCRED setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700280 * @sk_rcvlowat: %SO_RCVLOWAT setting
281 * @sk_rcvtimeo: %SO_RCVTIMEO setting
282 * @sk_sndtimeo: %SO_SNDTIMEO setting
Tom Herbertb73c3d02014-07-01 21:32:17 -0700283 * @sk_txhash: computed flow hash for use on transmit
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700284 * @sk_filter: socket filtering instructions
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700285 * @sk_timer: sock cleanup timer
286 * @sk_stamp: time stamp of last packet received
Willem de Bruijnb9f40e22014-08-04 22:11:46 -0400287 * @sk_tsflags: SO_TIMESTAMPING socket options
Willem de Bruijn09c2d252014-08-04 22:11:47 -0400288 * @sk_tskey: counter to disambiguate concurrent tstamp requests
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700289 * @sk_socket: Identd and reporting IO signals
290 * @sk_user_data: RPC layer private data
Eric Dumazet5640f762012-09-23 23:04:42 +0000291 * @sk_frag: cached page frag
Randy Dunlapd3d4f0a2012-04-17 14:03:53 +0000292 * @sk_peek_off: current peek_offset value
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700293 * @sk_send_head: front of stuff to transmit
Martin Waitz67be2dd2005-05-01 08:59:26 -0700294 * @sk_security: used by security modules
Randy Dunlap31729362008-02-18 20:52:13 -0800295 * @sk_mark: generic packet mark
Tejun Heo2a56a1f2015-12-07 17:38:52 -0500296 * @sk_cgrp_data: cgroup data for this cgroup
Johannes Weinerbaac50b2016-01-14 15:21:17 -0800297 * @sk_memcg: this socket's memory cgroup association
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700298 * @sk_write_pending: a write to stream socket waits to start
299 * @sk_state_change: callback to indicate change in the state of the sock
300 * @sk_data_ready: callback to indicate there is data to be processed
301 * @sk_write_space: callback to indicate there is bf sending space available
302 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
303 * @sk_backlog_rcv: callback to process the backlog
304 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
Craig Gallekef456142016-01-04 17:41:45 -0500305 * @sk_reuseport_cb: reuseport group container
Stephen Hemminger293de7d2016-10-23 09:28:29 -0700306 * @sk_rcu: used during RCU grace period
307 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308struct sock {
309 /*
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700310 * Now struct inet_timewait_sock also uses sock_common, so please just
Linus Torvalds1da177e2005-04-16 15:20:36 -0700311 * don't add nothing before this first member (__sk_common) --acme
312 */
313 struct sock_common __sk_common;
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000314#define sk_node __sk_common.skc_node
315#define sk_nulls_node __sk_common.skc_nulls_node
316#define sk_refcnt __sk_common.skc_refcnt
Krishna Kumare022f0b2009-10-19 23:46:20 +0000317#define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000318
Eric Dumazet68835ab2010-11-30 19:04:07 +0000319#define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
320#define sk_dontcopy_end __sk_common.skc_dontcopy_end
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000321#define sk_hash __sk_common.skc_hash
Eric Dumazet50805462013-10-02 04:29:50 -0700322#define sk_portpair __sk_common.skc_portpair
Eric Dumazet05dbc7b2013-10-03 00:22:02 -0700323#define sk_num __sk_common.skc_num
324#define sk_dport __sk_common.skc_dport
Eric Dumazet50805462013-10-02 04:29:50 -0700325#define sk_addrpair __sk_common.skc_addrpair
326#define sk_daddr __sk_common.skc_daddr
327#define sk_rcv_saddr __sk_common.skc_rcv_saddr
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328#define sk_family __sk_common.skc_family
329#define sk_state __sk_common.skc_state
330#define sk_reuse __sk_common.skc_reuse
Tom Herbert055dc212013-01-22 09:49:50 +0000331#define sk_reuseport __sk_common.skc_reuseport
Eric Dumazet9fe516b2014-06-27 08:36:16 -0700332#define sk_ipv6only __sk_common.skc_ipv6only
Eric W. Biederman26abe142015-05-08 21:10:31 -0500333#define sk_net_refcnt __sk_common.skc_net_refcnt
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334#define sk_bound_dev_if __sk_common.skc_bound_dev_if
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335#define sk_bind_node __sk_common.skc_bind_node
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700336#define sk_prot __sk_common.skc_prot
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200337#define sk_net __sk_common.skc_net
Eric Dumazetefe42082013-10-03 15:42:29 -0700338#define sk_v6_daddr __sk_common.skc_v6_daddr
339#define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr
Eric Dumazet33cf7c92015-03-11 18:53:14 -0700340#define sk_cookie __sk_common.skc_cookie
Eric Dumazet70da2682015-10-08 19:33:21 -0700341#define sk_incoming_cpu __sk_common.skc_incoming_cpu
Eric Dumazet8e5eb542015-10-08 19:33:22 -0700342#define sk_flags __sk_common.skc_flags
Eric Dumazeted53d0a2015-10-08 19:33:23 -0700343#define sk_rxhash __sk_common.skc_rxhash
Eric Dumazetefe42082013-10-03 15:42:29 -0700344
Eric Dumazetb178bb32010-11-16 05:56:04 +0000345 socket_lock_t sk_lock;
346 struct sk_buff_head sk_receive_queue;
347 /*
348 * The backlog queue is special, it is always used with
349 * the per-socket spinlock held and requires low latency
350 * access. Therefore we special case it's implementation.
351 * Note : rmem_alloc is in this structure to fill a hole
352 * on 64bit arches, not because its logically part of
353 * backlog.
354 */
355 struct {
356 atomic_t rmem_alloc;
357 int len;
358 struct sk_buff *head;
359 struct sk_buff *tail;
360 } sk_backlog;
361#define sk_rmem_alloc sk_backlog.rmem_alloc
362 int sk_forward_alloc;
Eric Dumazet2c8c56e2014-11-11 05:54:28 -0800363
Tom Herbertb73c3d02014-07-01 21:32:17 -0700364 __u32 sk_txhash;
Cong Wange0d10952013-08-01 11:10:25 +0800365#ifdef CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir06021292013-06-10 11:39:50 +0300366 unsigned int sk_napi_id;
Eliezer Tamirdafcc432013-06-14 16:33:57 +0300367 unsigned int sk_ll_usec;
Eliezer Tamir06021292013-06-10 11:39:50 +0300368#endif
Eric Dumazetb178bb32010-11-16 05:56:04 +0000369 atomic_t sk_drops;
370 int sk_rcvbuf;
371
372 struct sk_filter __rcu *sk_filter;
Eric Dumazetceb5d582015-11-29 20:03:11 -0800373 union {
374 struct socket_wq __rcu *sk_wq;
375 struct socket_wq *sk_wq_raw;
376 };
Eric Dumazetb178bb32010-11-16 05:56:04 +0000377#ifdef CONFIG_XFRM
Eric Dumazetd188ba82015-12-08 07:22:02 -0800378 struct xfrm_policy __rcu *sk_policy[2];
Eric Dumazetb178bb32010-11-16 05:56:04 +0000379#endif
Eric Dumazetdeaa5852012-06-24 20:22:49 +0000380 struct dst_entry *sk_rx_dst;
Cong Wang0e36cbb2013-01-22 21:09:51 +0000381 struct dst_entry __rcu *sk_dst_cache;
Eric Dumazet6bd4f352015-12-02 21:53:57 -0800382 /* Note: 32bit hole on 64bit arches */
Eric Dumazetb178bb32010-11-16 05:56:04 +0000383 atomic_t sk_wmem_alloc;
384 atomic_t sk_omem_alloc;
385 int sk_sndbuf;
386 struct sk_buff_head sk_write_queue;
Andrey Ryabininfc648692016-05-18 19:19:27 +0300387
388 /*
389 * Because of non atomicity rules, all
390 * changes are protected by socket lock.
391 */
Vegard Nossuma98b65a2009-02-26 14:46:57 +0100392 kmemcheck_bitfield_begin(flags);
Andrey Ryabininfc648692016-05-18 19:19:27 +0300393 unsigned int sk_padding : 2,
Tom Herbert28448b82014-05-23 08:47:19 -0700394 sk_no_check_tx : 1,
395 sk_no_check_rx : 1,
Eric Dumazet5fdb9972009-10-08 22:50:25 +0000396 sk_userlocks : 4,
397 sk_protocol : 8,
398 sk_type : 16;
Hannes Frederic Sowa7bbadd22015-12-14 23:30:43 +0100399#define SK_PROTOCOL_MAX U8_MAX
Vegard Nossuma98b65a2009-02-26 14:46:57 +0100400 kmemcheck_bitfield_end(flags);
Andrey Ryabininfc648692016-05-18 19:19:27 +0300401
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402 int sk_wmem_queued;
Al Viro7d877f32005-10-21 03:20:43 -0400403 gfp_t sk_allocation;
Eric Dumazet95bd09e2013-08-27 05:46:32 -0700404 u32 sk_pacing_rate; /* bytes per second */
Eric Dumazet62748f32013-09-24 08:20:52 -0700405 u32 sk_max_pacing_rate;
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000406 netdev_features_t sk_route_caps;
407 netdev_features_t sk_route_nocaps;
Herbert Xubcd76112006-06-30 13:36:35 -0700408 int sk_gso_type;
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -0700409 unsigned int sk_gso_max_size;
Ben Hutchings14853482012-07-30 16:11:42 +0000410 u16 sk_gso_max_segs;
David S. Miller9932cf92006-03-24 15:12:37 -0800411 int sk_rcvlowat;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412 unsigned long sk_lingertime;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700413 struct sk_buff_head sk_error_queue;
Arnaldo Carvalho de Melo476e19c2005-05-05 13:35:15 -0700414 struct proto *sk_prot_creator;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 rwlock_t sk_callback_lock;
416 int sk_err,
417 sk_err_soft;
Eric Dumazetbecb74f2015-03-19 19:04:21 -0700418 u32 sk_ack_backlog;
419 u32 sk_max_ack_backlog;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420 __u32 sk_priority;
Tejun Heo297dbde2015-12-07 17:38:51 -0500421 __u32 sk_mark;
Eric W. Biederman109f6e32010-06-13 03:30:14 +0000422 struct pid *sk_peer_pid;
423 const struct cred *sk_peer_cred;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424 long sk_rcvtimeo;
425 long sk_sndtimeo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426 struct timer_list sk_timer;
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700427 ktime_t sk_stamp;
Willem de Bruijnb9f40e22014-08-04 22:11:46 -0400428 u16 sk_tsflags;
Andrey Ryabininfc648692016-05-18 19:19:27 +0300429 u8 sk_shutdown;
Willem de Bruijn09c2d252014-08-04 22:11:47 -0400430 u32 sk_tskey;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431 struct socket *sk_socket;
432 void *sk_user_data;
Eric Dumazet5640f762012-09-23 23:04:42 +0000433 struct page_frag sk_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 struct sk_buff *sk_send_head;
Pavel Emelyanovef64a542012-02-21 07:31:34 +0000435 __s32 sk_peek_off;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436 int sk_write_pending;
Alexey Dobriyand5f64232008-11-04 14:45:58 -0800437#ifdef CONFIG_SECURITY
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438 void *sk_security;
Alexey Dobriyand5f64232008-11-04 14:45:58 -0800439#endif
Tejun Heo2a56a1f2015-12-07 17:38:52 -0500440 struct sock_cgroup_data sk_cgrp_data;
Johannes Weinerbaac50b2016-01-14 15:21:17 -0800441 struct mem_cgroup *sk_memcg;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442 void (*sk_state_change)(struct sock *sk);
David S. Miller676d2362014-04-11 16:15:36 -0400443 void (*sk_data_ready)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444 void (*sk_write_space)(struct sock *sk);
445 void (*sk_error_report)(struct sock *sk);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000446 int (*sk_backlog_rcv)(struct sock *sk,
447 struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448 void (*sk_destruct)(struct sock *sk);
Craig Gallekef456142016-01-04 17:41:45 -0500449 struct sock_reuseport __rcu *sk_reuseport_cb;
Eric Dumazeta4298e42016-04-01 08:52:12 -0700450 struct rcu_head sk_rcu;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451};
452
Pravin B Shelar559835ea2013-09-24 10:25:40 -0700453#define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
454
455#define rcu_dereference_sk_user_data(sk) rcu_dereference(__sk_user_data((sk)))
456#define rcu_assign_sk_user_data(sk, ptr) rcu_assign_pointer(__sk_user_data((sk)), ptr)
457
Pavel Emelyanov4a17fd52012-04-19 03:39:36 +0000458/*
459 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
460 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
461 * on a socket means that the socket will reuse everybody else's port
462 * without looking at the other's sk_reuse value.
463 */
464
465#define SK_NO_REUSE 0
466#define SK_CAN_REUSE 1
467#define SK_FORCE_REUSE 2
468
samanthakumar627d2d62016-04-05 12:41:16 -0400469int sk_set_peek_off(struct sock *sk, int val);
470
Pavel Emelyanovef64a542012-02-21 07:31:34 +0000471static inline int sk_peek_offset(struct sock *sk, int flags)
472{
Willem de Bruijnb9bb53f2016-04-05 12:41:14 -0400473 if (unlikely(flags & MSG_PEEK)) {
474 s32 off = READ_ONCE(sk->sk_peek_off);
475 if (off >= 0)
476 return off;
477 }
478
479 return 0;
Pavel Emelyanovef64a542012-02-21 07:31:34 +0000480}
481
482static inline void sk_peek_offset_bwd(struct sock *sk, int val)
483{
Willem de Bruijnb9bb53f2016-04-05 12:41:14 -0400484 s32 off = READ_ONCE(sk->sk_peek_off);
485
486 if (unlikely(off >= 0)) {
487 off = max_t(s32, off - val, 0);
488 WRITE_ONCE(sk->sk_peek_off, off);
Pavel Emelyanovef64a542012-02-21 07:31:34 +0000489 }
490}
491
492static inline void sk_peek_offset_fwd(struct sock *sk, int val)
493{
Willem de Bruijnb9bb53f2016-04-05 12:41:14 -0400494 sk_peek_offset_bwd(sk, -val);
Pavel Emelyanovef64a542012-02-21 07:31:34 +0000495}
496
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497/*
498 * Hashed lists helper routines
499 */
Li Zefanc4146642010-02-08 23:18:45 +0000500static inline struct sock *sk_entry(const struct hlist_node *node)
501{
502 return hlist_entry(node, struct sock, sk_node);
503}
504
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700505static inline struct sock *__sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506{
507 return hlist_entry(head->first, struct sock, sk_node);
508}
509
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700510static inline struct sock *sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511{
512 return hlist_empty(head) ? NULL : __sk_head(head);
513}
514
Eric Dumazet88ab1932008-11-16 19:39:21 -0800515static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
516{
517 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
518}
519
520static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
521{
522 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
523}
524
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700525static inline struct sock *sk_next(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526{
527 return sk->sk_node.next ?
528 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
529}
530
Eric Dumazet88ab1932008-11-16 19:39:21 -0800531static inline struct sock *sk_nulls_next(const struct sock *sk)
532{
533 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
534 hlist_nulls_entry(sk->sk_nulls_node.next,
535 struct sock, sk_nulls_node) :
536 NULL;
537}
538
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000539static inline bool sk_unhashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540{
541 return hlist_unhashed(&sk->sk_node);
542}
543
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000544static inline bool sk_hashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545{
Akinobu Mitada753be2006-04-28 15:21:23 -0700546 return !sk_unhashed(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547}
548
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000549static inline void sk_node_init(struct hlist_node *node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550{
551 node->pprev = NULL;
552}
553
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000554static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800555{
556 node->pprev = NULL;
557}
558
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000559static inline void __sk_del_node(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560{
561 __hlist_del(&sk->sk_node);
562}
563
stephen hemminger808f5112010-02-22 07:57:18 +0000564/* NB: equivalent to hlist_del_init_rcu */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000565static inline bool __sk_del_node_init(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566{
567 if (sk_hashed(sk)) {
568 __sk_del_node(sk);
569 sk_node_init(&sk->sk_node);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000570 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571 }
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000572 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573}
574
575/* Grab socket reference count. This operation is valid only
576 when sk is ALREADY grabbed f.e. it is found in hash table
577 or a list and the lookup is made under lock preventing hash table
578 modifications.
579 */
580
Denys Vlasenkof9a7cbb2016-04-08 17:51:54 +0200581static __always_inline void sock_hold(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582{
583 atomic_inc(&sk->sk_refcnt);
584}
585
586/* Ungrab socket in the context, which assumes that socket refcnt
587 cannot hit zero, f.e. it is true in context of any socketcall.
588 */
Denys Vlasenkof9a7cbb2016-04-08 17:51:54 +0200589static __always_inline void __sock_put(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590{
591 atomic_dec(&sk->sk_refcnt);
592}
593
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000594static inline bool sk_del_node_init(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595{
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000596 bool rc = __sk_del_node_init(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597
598 if (rc) {
599 /* paranoid for a while -acme */
600 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
601 __sock_put(sk);
602 }
603 return rc;
604}
stephen hemminger808f5112010-02-22 07:57:18 +0000605#define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000607static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700608{
609 if (sk_hashed(sk)) {
Eric Dumazet88ab1932008-11-16 19:39:21 -0800610 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000611 return true;
Eric Dumazet271b72c2008-10-29 02:11:14 -0700612 }
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000613 return false;
Eric Dumazet271b72c2008-10-29 02:11:14 -0700614}
615
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000616static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700617{
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000618 bool rc = __sk_nulls_del_node_init_rcu(sk);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700619
620 if (rc) {
621 /* paranoid for a while -acme */
622 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
623 __sock_put(sk);
624 }
625 return rc;
626}
627
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000628static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629{
630 hlist_add_head(&sk->sk_node, list);
631}
632
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000633static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634{
635 sock_hold(sk);
636 __sk_add_node(sk, list);
637}
638
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000639static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
stephen hemminger808f5112010-02-22 07:57:18 +0000640{
641 sock_hold(sk);
Craig Gallekd296ba62016-04-25 10:42:12 -0400642 if (IS_ENABLED(CONFIG_IPV6) && sk->sk_reuseport &&
643 sk->sk_family == AF_INET6)
644 hlist_add_tail_rcu(&sk->sk_node, list);
645 else
646 hlist_add_head_rcu(&sk->sk_node, list);
stephen hemminger808f5112010-02-22 07:57:18 +0000647}
648
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000649static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700650{
Craig Gallekd894ba12016-04-12 13:11:25 -0400651 if (IS_ENABLED(CONFIG_IPV6) && sk->sk_reuseport &&
652 sk->sk_family == AF_INET6)
653 hlist_nulls_add_tail_rcu(&sk->sk_nulls_node, list);
654 else
655 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700656}
657
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000658static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700659{
660 sock_hold(sk);
Eric Dumazet88ab1932008-11-16 19:39:21 -0800661 __sk_nulls_add_node_rcu(sk, list);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700662}
663
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000664static inline void __sk_del_bind_node(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665{
666 __hlist_del(&sk->sk_bind_node);
667}
668
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000669static inline void sk_add_bind_node(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670 struct hlist_head *list)
671{
672 hlist_add_head(&sk->sk_bind_node, list);
673}
674
Sasha Levinb67bfe02013-02-27 17:06:00 -0800675#define sk_for_each(__sk, list) \
676 hlist_for_each_entry(__sk, list, sk_node)
677#define sk_for_each_rcu(__sk, list) \
678 hlist_for_each_entry_rcu(__sk, list, sk_node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800679#define sk_nulls_for_each(__sk, node, list) \
680 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
681#define sk_nulls_for_each_rcu(__sk, node, list) \
682 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
Sasha Levinb67bfe02013-02-27 17:06:00 -0800683#define sk_for_each_from(__sk) \
684 hlist_for_each_entry_from(__sk, sk_node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800685#define sk_nulls_for_each_from(__sk, node) \
686 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
687 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
Sasha Levinb67bfe02013-02-27 17:06:00 -0800688#define sk_for_each_safe(__sk, tmp, list) \
689 hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
690#define sk_for_each_bound(__sk, list) \
691 hlist_for_each_entry(__sk, list, sk_bind_node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692
David Held2dc41cf2014-07-15 23:28:32 -0400693/**
Eric Dumazetca065d02016-04-01 08:52:13 -0700694 * sk_for_each_entry_offset_rcu - iterate over a list at a given struct offset
David Held2dc41cf2014-07-15 23:28:32 -0400695 * @tpos: the type * to use as a loop cursor.
696 * @pos: the &struct hlist_node to use as a loop cursor.
697 * @head: the head for your list.
698 * @offset: offset of hlist_node within the struct.
699 *
700 */
Eric Dumazetca065d02016-04-01 08:52:13 -0700701#define sk_for_each_entry_offset_rcu(tpos, pos, head, offset) \
702 for (pos = rcu_dereference((head)->first); \
703 pos != NULL && \
David Held2dc41cf2014-07-15 23:28:32 -0400704 ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;}); \
Eric Dumazetca065d02016-04-01 08:52:13 -0700705 pos = rcu_dereference(pos->next))
David Held2dc41cf2014-07-15 23:28:32 -0400706
Eric W. Biedermanc336d142012-05-24 17:56:43 -0600707static inline struct user_namespace *sk_user_ns(struct sock *sk)
708{
709 /* Careful only use this in a context where these parameters
710 * can not change and must all be valid, such as recvmsg from
711 * userspace.
712 */
713 return sk->sk_socket->file->f_cred->user_ns;
714}
715
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716/* Sock flags */
717enum sock_flags {
718 SOCK_DEAD,
719 SOCK_DONE,
720 SOCK_URGINLINE,
721 SOCK_KEEPOPEN,
722 SOCK_LINGER,
723 SOCK_DESTROY,
724 SOCK_BROADCAST,
725 SOCK_TIMESTAMP,
726 SOCK_ZAPPED,
727 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
728 SOCK_DBG, /* %SO_DEBUG setting */
729 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700730 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
732 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
Mel Gorman7cb02402012-07-31 16:44:16 -0700733 SOCK_MEMALLOC, /* VM depends on this socket for swapping */
Patrick Ohly20d49472009-02-12 05:03:38 +0000734 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
Eric Dumazetbcdce712009-10-06 17:28:29 -0700735 SOCK_FASYNC, /* fasync() active */
Neil Horman3b885782009-10-12 13:26:31 -0700736 SOCK_RXQ_OVFL,
Shirley Ma1cdebb42011-07-06 12:17:30 +0000737 SOCK_ZEROCOPY, /* buffers from userspace */
Johannes Berg6e3e9392011-11-09 10:15:42 +0100738 SOCK_WIFI_STATUS, /* push wifi status to userspace */
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000739 SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
740 * Will use last 4 bytes of packet sent from
741 * user-space instead.
742 */
Vincent Bernatd59577b2013-01-16 22:55:49 +0100743 SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */
Keller, Jacob E7d4c04f2013-03-28 11:19:25 +0000744 SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */
Eric Dumazeta4298e42016-04-01 08:52:12 -0700745 SOCK_RCU_FREE, /* wait rcu grace period in sk_destruct() */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746};
747
Marcelo Ricardo Leitner01ce63c2015-12-04 15:14:04 -0200748#define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
749
Ralf Baechle53b924b2005-08-23 10:11:30 -0700750static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
751{
752 nsk->sk_flags = osk->sk_flags;
753}
754
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
756{
757 __set_bit(flag, &sk->sk_flags);
758}
759
760static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
761{
762 __clear_bit(flag, &sk->sk_flags);
763}
764
Eric Dumazet1b23a5d2012-05-16 05:57:07 +0000765static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766{
767 return test_bit(flag, &sk->sk_flags);
768}
769
Mel Gormanc93bdd02012-07-31 16:44:19 -0700770#ifdef CONFIG_NET
771extern struct static_key memalloc_socks;
772static inline int sk_memalloc_socks(void)
773{
774 return static_key_false(&memalloc_socks);
775}
776#else
777
778static inline int sk_memalloc_socks(void)
779{
780 return 0;
781}
782
783#endif
784
Eric Dumazet7450aaf2015-11-30 08:57:28 -0800785static inline gfp_t sk_gfp_mask(const struct sock *sk, gfp_t gfp_mask)
Mel Gorman99a1dec2012-07-31 16:44:14 -0700786{
Eric Dumazet7450aaf2015-11-30 08:57:28 -0800787 return gfp_mask | (sk->sk_allocation & __GFP_MEMALLOC);
Mel Gorman99a1dec2012-07-31 16:44:14 -0700788}
789
Linus Torvalds1da177e2005-04-16 15:20:36 -0700790static inline void sk_acceptq_removed(struct sock *sk)
791{
792 sk->sk_ack_backlog--;
793}
794
795static inline void sk_acceptq_added(struct sock *sk)
796{
797 sk->sk_ack_backlog++;
798}
799
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000800static inline bool sk_acceptq_is_full(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801{
David S. Miller64a14652007-03-06 11:21:05 -0800802 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803}
804
805/*
806 * Compute minimal free write space needed to queue new packets.
807 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000808static inline int sk_stream_min_wspace(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809{
Eric Dumazet8df09ea2007-12-21 03:07:41 -0800810 return sk->sk_wmem_queued >> 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811}
812
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000813static inline int sk_stream_wspace(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814{
815 return sk->sk_sndbuf - sk->sk_wmem_queued;
816}
817
Joe Perches69336bd2013-09-22 10:32:26 -0700818void sk_stream_write_space(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819
Zhu Yi8eae9392010-03-04 18:01:40 +0000820/* OOB backlog add */
Zhu Yia3a858f2010-03-04 18:01:47 +0000821static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800822{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000823 /* dont let skb dst not refcounted, we are going to leave rcu lock */
Eric Dumazet5037e9e2015-12-14 14:08:53 -0800824 skb_dst_force_safe(skb);
Eric Dumazet7fee2262010-05-11 23:19:48 +0000825
826 if (!sk->sk_backlog.tail)
827 sk->sk_backlog.head = skb;
828 else
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800829 sk->sk_backlog.tail->next = skb;
Eric Dumazet7fee2262010-05-11 23:19:48 +0000830
831 sk->sk_backlog.tail = skb;
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800832 skb->next = NULL;
833}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834
Eric Dumazetc3774112010-04-27 15:13:20 -0700835/*
836 * Take into account size of receive queue and backlog queue
Eric Dumazet0fd7bac2011-12-21 07:11:44 +0000837 * Do not take into account this skb truesize,
838 * to allow even a single big packet to come.
Eric Dumazetc3774112010-04-27 15:13:20 -0700839 */
Sorin Dumitru274f4822014-07-22 21:16:51 +0300840static inline bool sk_rcvqueues_full(const struct sock *sk, unsigned int limit)
Eric Dumazetc3774112010-04-27 15:13:20 -0700841{
842 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
843
Eric Dumazetf545a382012-04-22 23:34:26 +0000844 return qsize > limit;
Eric Dumazetc3774112010-04-27 15:13:20 -0700845}
846
Zhu Yi8eae9392010-03-04 18:01:40 +0000847/* The per-socket spinlock must be held here. */
Eric Dumazetf545a382012-04-22 23:34:26 +0000848static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
849 unsigned int limit)
Zhu Yi8eae9392010-03-04 18:01:40 +0000850{
Sorin Dumitru274f4822014-07-22 21:16:51 +0300851 if (sk_rcvqueues_full(sk, limit))
Zhu Yi8eae9392010-03-04 18:01:40 +0000852 return -ENOBUFS;
853
Eric Dumazetc7c49b82015-09-29 18:52:25 -0700854 /*
855 * If the skb was allocated from pfmemalloc reserves, only
856 * allow SOCK_MEMALLOC sockets to use it as this socket is
857 * helping free memory
858 */
859 if (skb_pfmemalloc(skb) && !sock_flag(sk, SOCK_MEMALLOC))
860 return -ENOMEM;
861
Zhu Yia3a858f2010-03-04 18:01:47 +0000862 __sk_add_backlog(sk, skb);
Zhu Yi8eae9392010-03-04 18:01:40 +0000863 sk->sk_backlog.len += skb->truesize;
864 return 0;
865}
866
Joe Perches69336bd2013-09-22 10:32:26 -0700867int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
Mel Gormanb4b9e352012-07-31 16:44:26 -0700868
Peter Zijlstrac57943a2008-10-07 14:18:42 -0700869static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
870{
Mel Gormanb4b9e352012-07-31 16:44:26 -0700871 if (sk_memalloc_socks() && skb_pfmemalloc(skb))
872 return __sk_backlog_rcv(sk, skb);
873
Peter Zijlstrac57943a2008-10-07 14:18:42 -0700874 return sk->sk_backlog_rcv(sk, skb);
875}
876
Eric Dumazet2c8c56e2014-11-11 05:54:28 -0800877static inline void sk_incoming_cpu_update(struct sock *sk)
878{
879 sk->sk_incoming_cpu = raw_smp_processor_id();
880}
881
Tom Herbertfe477552013-12-22 18:54:31 +0800882static inline void sock_rps_record_flow_hash(__u32 hash)
David S. Millerc58dc012010-04-27 15:05:31 -0700883{
884#ifdef CONFIG_RPS
885 struct rps_sock_flow_table *sock_flow_table;
886
887 rcu_read_lock();
888 sock_flow_table = rcu_dereference(rps_sock_flow_table);
Tom Herbertfe477552013-12-22 18:54:31 +0800889 rps_record_sock_flow(sock_flow_table, hash);
David S. Millerc58dc012010-04-27 15:05:31 -0700890 rcu_read_unlock();
891#endif
892}
893
Tom Herbertfe477552013-12-22 18:54:31 +0800894static inline void sock_rps_record_flow(const struct sock *sk)
895{
Zhi Yong Wuc9d8ca02014-01-01 04:31:01 +0800896#ifdef CONFIG_RPS
Tom Herbertfe477552013-12-22 18:54:31 +0800897 sock_rps_record_flow_hash(sk->sk_rxhash);
Zhi Yong Wuc9d8ca02014-01-01 04:31:01 +0800898#endif
Tom Herbertfe477552013-12-22 18:54:31 +0800899}
900
Tom Herbertbdeab992011-08-14 19:45:55 +0000901static inline void sock_rps_save_rxhash(struct sock *sk,
902 const struct sk_buff *skb)
David S. Millerc58dc012010-04-27 15:05:31 -0700903{
904#ifdef CONFIG_RPS
Eric Dumazet567e4b72015-02-06 12:59:01 -0800905 if (unlikely(sk->sk_rxhash != skb->hash))
Tom Herbert61b905d2014-03-24 15:34:47 -0700906 sk->sk_rxhash = skb->hash;
David S. Millerc58dc012010-04-27 15:05:31 -0700907#endif
908}
909
Tom Herbertbdeab992011-08-14 19:45:55 +0000910static inline void sock_rps_reset_rxhash(struct sock *sk)
911{
912#ifdef CONFIG_RPS
Tom Herbertbdeab992011-08-14 19:45:55 +0000913 sk->sk_rxhash = 0;
914#endif
915}
916
Stephen Hemmingercfcabdc2007-10-09 01:59:42 -0700917#define sk_wait_event(__sk, __timeo, __condition) \
918 ({ int __rc; \
919 release_sock(__sk); \
920 __rc = __condition; \
921 if (!__rc) { \
922 *(__timeo) = schedule_timeout(*(__timeo)); \
923 } \
Peter Zijlstra26cabd32014-09-24 10:18:54 +0200924 sched_annotate_sleep(); \
Stephen Hemmingercfcabdc2007-10-09 01:59:42 -0700925 lock_sock(__sk); \
926 __rc = __condition; \
927 __rc; \
928 })
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929
Joe Perches69336bd2013-09-22 10:32:26 -0700930int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
931int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
932void sk_stream_wait_close(struct sock *sk, long timeo_p);
933int sk_stream_error(struct sock *sk, int flags, int err);
934void sk_stream_kill_queues(struct sock *sk);
935void sk_set_memalloc(struct sock *sk);
936void sk_clear_memalloc(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937
Eric Dumazetd41a69f2016-04-29 14:16:53 -0700938void __sk_flush_backlog(struct sock *sk);
939
940static inline bool sk_flush_backlog(struct sock *sk)
941{
942 if (unlikely(READ_ONCE(sk->sk_backlog.tail))) {
943 __sk_flush_backlog(sk);
944 return true;
945 }
946 return false;
947}
948
Sabrina Dubrocadfbafc92015-07-24 18:19:25 +0200949int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700950
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700951struct request_sock_ops;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800952struct timewait_sock_ops;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -0800953struct inet_hashinfo;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -0700954struct raw_hashinfo;
Paul Gortmakerde477252011-05-26 13:46:22 -0400955struct module;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700956
Eric Dumazetf77d6022013-05-09 10:28:16 +0000957/*
958 * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
959 * un-modified. Special care is taken when initializing object to zero.
960 */
961static inline void sk_prot_clear_nulls(struct sock *sk, int size)
962{
963 if (offsetof(struct sock, sk_node.next) != 0)
964 memset(sk, 0, offsetof(struct sock, sk_node.next));
965 memset(&sk->sk_node.pprev, 0,
966 size - offsetof(struct sock, sk_node.pprev));
967}
968
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969/* Networking protocol blocks we attach to sockets.
970 * socket layer -> transport layer interface
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971 */
972struct proto {
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000973 void (*close)(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974 long timeout);
975 int (*connect)(struct sock *sk,
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000976 struct sockaddr *uaddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977 int addr_len);
978 int (*disconnect)(struct sock *sk, int flags);
979
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000980 struct sock * (*accept)(struct sock *sk, int flags, int *err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981
982 int (*ioctl)(struct sock *sk, int cmd,
983 unsigned long arg);
984 int (*init)(struct sock *sk);
Brian Haley7d06b2e2008-06-14 17:04:49 -0700985 void (*destroy)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986 void (*shutdown)(struct sock *sk, int how);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000987 int (*setsockopt)(struct sock *sk, int level,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700988 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700989 unsigned int optlen);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000990 int (*getsockopt)(struct sock *sk, int level,
991 int optname, char __user *optval,
992 int __user *option);
Alexey Dobriyanaf01d532008-08-28 02:53:51 -0700993#ifdef CONFIG_COMPAT
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800994 int (*compat_setsockopt)(struct sock *sk,
995 int level,
996 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700997 unsigned int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800998 int (*compat_getsockopt)(struct sock *sk,
999 int level,
1000 int optname, char __user *optval,
1001 int __user *option);
Eric W. Biederman709b46e2011-01-29 16:15:56 +00001002 int (*compat_ioctl)(struct sock *sk,
1003 unsigned int cmd, unsigned long arg);
Alexey Dobriyanaf01d532008-08-28 02:53:51 -07001004#endif
Ying Xue1b784142015-03-02 15:37:48 +08001005 int (*sendmsg)(struct sock *sk, struct msghdr *msg,
1006 size_t len);
1007 int (*recvmsg)(struct sock *sk, struct msghdr *msg,
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001008 size_t len, int noblock, int flags,
1009 int *addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010 int (*sendpage)(struct sock *sk, struct page *page,
1011 int offset, size_t size, int flags);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001012 int (*bind)(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013 struct sockaddr *uaddr, int addr_len);
1014
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001015 int (*backlog_rcv) (struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016 struct sk_buff *skb);
1017
Eric Dumazet46d3cea2012-07-11 05:50:31 +00001018 void (*release_cb)(struct sock *sk);
1019
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020 /* Keeping track of sk's, looking them up, and port selection methods. */
Craig Gallek086c6532016-02-10 11:50:35 -05001021 int (*hash)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022 void (*unhash)(struct sock *sk);
Eric Dumazet719f8352010-09-08 05:08:44 +00001023 void (*rehash)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024 int (*get_port)(struct sock *sk, unsigned short snum);
1025
Eric Dumazet286ab3d2007-11-05 23:38:39 -08001026 /* Keeping track of sockets in use */
Eric Dumazet65f76512008-01-03 20:46:48 -08001027#ifdef CONFIG_PROC_FS
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -07001028 unsigned int inuse_idx;
Eric Dumazet65f76512008-01-03 20:46:48 -08001029#endif
Arnaldo Carvalho de Meloebb53d72007-11-21 22:08:50 +08001030
Eric Dumazetc9bee3b72013-07-22 20:27:07 -07001031 bool (*stream_memory_free)(const struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032 /* Memory pressure */
Pavel Emelyanov5c52ba12008-07-16 20:28:10 -07001033 void (*enter_memory_pressure)(struct sock *sk);
Eric Dumazet8d987e52010-11-09 23:24:26 +00001034 atomic_long_t *memory_allocated; /* Current allocated memory. */
Eric Dumazet17483762008-11-25 21:16:35 -08001035 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036 /*
1037 * Pressure flag: try to collapse.
1038 * Technical note: it is used by multiple contexts non atomically.
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001039 * All the __sk_mem_schedule() is of this nature: accounting
Linus Torvalds1da177e2005-04-16 15:20:36 -07001040 * is strict, actions are advisory and have some latency.
1041 */
1042 int *memory_pressure;
Eric Dumazet8d987e52010-11-09 23:24:26 +00001043 long *sysctl_mem;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044 int *sysctl_wmem;
1045 int *sysctl_rmem;
1046 int max_header;
Changli Gao7ba42912010-07-10 20:41:55 +00001047 bool no_autobind;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048
Eric Dumazet271b72c2008-10-29 02:11:14 -07001049 struct kmem_cache *slab;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050 unsigned int obj_size;
Eric Dumazet271b72c2008-10-29 02:11:14 -07001051 int slab_flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001052
Eric Dumazetdd24c002008-11-25 21:17:14 -08001053 struct percpu_counter *orphan_count;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -07001054
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -07001055 struct request_sock_ops *rsk_prot;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -08001056 struct timewait_sock_ops *twsk_prot;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07001057
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -07001058 union {
1059 struct inet_hashinfo *hashinfo;
Eric Dumazet645ca702008-10-29 01:41:45 -07001060 struct udp_table *udp_table;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -07001061 struct raw_hashinfo *raw_hash;
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -07001062 } h;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -08001063
Linus Torvalds1da177e2005-04-16 15:20:36 -07001064 struct module *owner;
1065
1066 char name[32];
1067
1068 struct list_head node;
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -07001069#ifdef SOCK_REFCNT_DEBUG
1070 atomic_t socks;
1071#endif
Lorenzo Colitti64be0ae2015-12-16 12:30:03 +09001072 int (*diag_destroy)(struct sock *sk, int err);
Glauber Costae1aab162011-12-11 21:47:03 +00001073};
1074
Joe Perches69336bd2013-09-22 10:32:26 -07001075int proto_register(struct proto *prot, int alloc_slab);
1076void proto_unregister(struct proto *prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001077
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -07001078#ifdef SOCK_REFCNT_DEBUG
1079static inline void sk_refcnt_debug_inc(struct sock *sk)
1080{
1081 atomic_inc(&sk->sk_prot->socks);
1082}
1083
1084static inline void sk_refcnt_debug_dec(struct sock *sk)
1085{
1086 atomic_dec(&sk->sk_prot->socks);
1087 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
1088 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
1089}
1090
Ying Xuedec34fb2013-02-15 22:28:25 +00001091static inline void sk_refcnt_debug_release(const struct sock *sk)
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -07001092{
1093 if (atomic_read(&sk->sk_refcnt) != 1)
1094 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
1095 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
1096}
1097#else /* SOCK_REFCNT_DEBUG */
1098#define sk_refcnt_debug_inc(sk) do { } while (0)
1099#define sk_refcnt_debug_dec(sk) do { } while (0)
1100#define sk_refcnt_debug_release(sk) do { } while (0)
1101#endif /* SOCK_REFCNT_DEBUG */
1102
Eric Dumazetc9bee3b72013-07-22 20:27:07 -07001103static inline bool sk_stream_memory_free(const struct sock *sk)
1104{
1105 if (sk->sk_wmem_queued >= sk->sk_sndbuf)
1106 return false;
1107
1108 return sk->sk_prot->stream_memory_free ?
1109 sk->sk_prot->stream_memory_free(sk) : true;
1110}
1111
Eric Dumazet64dc6132013-07-22 20:26:31 -07001112static inline bool sk_stream_is_writeable(const struct sock *sk)
1113{
Eric Dumazetc9bee3b72013-07-22 20:27:07 -07001114 return sk_stream_wspace(sk) >= sk_stream_min_wspace(sk) &&
1115 sk_stream_memory_free(sk);
Eric Dumazet64dc6132013-07-22 20:26:31 -07001116}
Glauber Costae1aab162011-12-11 21:47:03 +00001117
Daniel Borkmann54fd9c22016-08-18 01:00:41 +02001118static inline int sk_under_cgroup_hierarchy(struct sock *sk,
1119 struct cgroup *ancestor)
1120{
1121#ifdef CONFIG_SOCK_CGROUP_DATA
1122 return cgroup_is_descendant(sock_cgroup_ptr(&sk->sk_cgrp_data),
1123 ancestor);
1124#else
1125 return -ENOTSUPP;
1126#endif
1127}
Eric Dumazetc9bee3b72013-07-22 20:27:07 -07001128
Glauber Costa180d8cd2011-12-11 21:47:02 +00001129static inline bool sk_has_memory_pressure(const struct sock *sk)
1130{
1131 return sk->sk_prot->memory_pressure != NULL;
1132}
1133
1134static inline bool sk_under_memory_pressure(const struct sock *sk)
1135{
1136 if (!sk->sk_prot->memory_pressure)
1137 return false;
Glauber Costae1aab162011-12-11 21:47:03 +00001138
Johannes Weinerbaac50b2016-01-14 15:21:17 -08001139 if (mem_cgroup_sockets_enabled && sk->sk_memcg &&
1140 mem_cgroup_under_socket_pressure(sk->sk_memcg))
Johannes Weinere8056052016-01-14 15:21:14 -08001141 return true;
Glauber Costae1aab162011-12-11 21:47:03 +00001142
Christoph Paasch35b87f62013-10-23 12:49:21 -07001143 return !!*sk->sk_prot->memory_pressure;
Glauber Costa180d8cd2011-12-11 21:47:02 +00001144}
1145
1146static inline void sk_leave_memory_pressure(struct sock *sk)
1147{
1148 int *memory_pressure = sk->sk_prot->memory_pressure;
1149
Glauber Costae1aab162011-12-11 21:47:03 +00001150 if (!memory_pressure)
1151 return;
1152
1153 if (*memory_pressure)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001154 *memory_pressure = 0;
1155}
1156
1157static inline void sk_enter_memory_pressure(struct sock *sk)
1158{
Glauber Costae1aab162011-12-11 21:47:03 +00001159 if (!sk->sk_prot->enter_memory_pressure)
1160 return;
1161
Glauber Costae1aab162011-12-11 21:47:03 +00001162 sk->sk_prot->enter_memory_pressure(sk);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001163}
1164
1165static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1166{
Johannes Weinere8056052016-01-14 15:21:14 -08001167 return sk->sk_prot->sysctl_mem[index];
Glauber Costae1aab162011-12-11 21:47:03 +00001168}
1169
Glauber Costa180d8cd2011-12-11 21:47:02 +00001170static inline long
1171sk_memory_allocated(const struct sock *sk)
1172{
Johannes Weinere8056052016-01-14 15:21:14 -08001173 return atomic_long_read(sk->sk_prot->memory_allocated);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001174}
1175
1176static inline long
Johannes Weinere8056052016-01-14 15:21:14 -08001177sk_memory_allocated_add(struct sock *sk, int amt)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001178{
Johannes Weinere8056052016-01-14 15:21:14 -08001179 return atomic_long_add_return(amt, sk->sk_prot->memory_allocated);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001180}
1181
1182static inline void
Glauber Costa0e90b312012-01-20 04:57:16 +00001183sk_memory_allocated_sub(struct sock *sk, int amt)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001184{
Johannes Weinere8056052016-01-14 15:21:14 -08001185 atomic_long_sub(amt, sk->sk_prot->memory_allocated);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001186}
1187
1188static inline void sk_sockets_allocated_dec(struct sock *sk)
1189{
Johannes Weineraf95d7d2016-01-14 15:21:08 -08001190 percpu_counter_dec(sk->sk_prot->sockets_allocated);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001191}
1192
1193static inline void sk_sockets_allocated_inc(struct sock *sk)
1194{
Johannes Weineraf95d7d2016-01-14 15:21:08 -08001195 percpu_counter_inc(sk->sk_prot->sockets_allocated);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001196}
1197
1198static inline int
1199sk_sockets_allocated_read_positive(struct sock *sk)
1200{
Johannes Weineraf95d7d2016-01-14 15:21:08 -08001201 return percpu_counter_read_positive(sk->sk_prot->sockets_allocated);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001202}
1203
1204static inline int
1205proto_sockets_allocated_sum_positive(struct proto *prot)
1206{
1207 return percpu_counter_sum_positive(prot->sockets_allocated);
1208}
1209
1210static inline long
1211proto_memory_allocated(struct proto *prot)
1212{
1213 return atomic_long_read(prot->memory_allocated);
1214}
1215
1216static inline bool
1217proto_memory_pressure(struct proto *prot)
1218{
1219 if (!prot->memory_pressure)
1220 return false;
1221 return !!*prot->memory_pressure;
1222}
1223
Eric Dumazet65f76512008-01-03 20:46:48 -08001224
1225#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001226/* Called with local bh disabled */
Joe Perches69336bd2013-09-22 10:32:26 -07001227void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1228int sock_prot_inuse_get(struct net *net, struct proto *proto);
Eric Dumazet65f76512008-01-03 20:46:48 -08001229#else
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001230static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -07001231 int inc)
Eric Dumazet65f76512008-01-03 20:46:48 -08001232{
1233}
Eric Dumazet65f76512008-01-03 20:46:48 -08001234#endif
1235
Linus Torvalds1da177e2005-04-16 15:20:36 -07001236
Arnaldo Carvalho de Melo614c6cb2005-08-09 19:47:37 -07001237/* With per-bucket locks this operation is not-atomic, so that
1238 * this version is not worse.
1239 */
Craig Gallek086c6532016-02-10 11:50:35 -05001240static inline int __sk_prot_rehash(struct sock *sk)
Arnaldo Carvalho de Melo614c6cb2005-08-09 19:47:37 -07001241{
1242 sk->sk_prot->unhash(sk);
Craig Gallek086c6532016-02-10 11:50:35 -05001243 return sk->sk_prot->hash(sk);
Arnaldo Carvalho de Melo614c6cb2005-08-09 19:47:37 -07001244}
1245
Linus Torvalds1da177e2005-04-16 15:20:36 -07001246/* About 10 seconds */
1247#define SOCK_DESTROY_TIME (10*HZ)
1248
1249/* Sockets 0-1023 can't be bound to unless you are superuser */
1250#define PROT_SOCK 1024
1251
1252#define SHUTDOWN_MASK 3
1253#define RCV_SHUTDOWN 1
1254#define SEND_SHUTDOWN 2
1255
1256#define SOCK_SNDBUF_LOCK 1
1257#define SOCK_RCVBUF_LOCK 2
1258#define SOCK_BINDADDR_LOCK 4
1259#define SOCK_BINDPORT_LOCK 8
1260
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261struct socket_alloc {
1262 struct socket socket;
1263 struct inode vfs_inode;
1264};
1265
1266static inline struct socket *SOCKET_I(struct inode *inode)
1267{
1268 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1269}
1270
1271static inline struct inode *SOCK_INODE(struct socket *socket)
1272{
1273 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1274}
1275
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001276/*
1277 * Functions for memory accounting
1278 */
Joe Perches69336bd2013-09-22 10:32:26 -07001279int __sk_mem_schedule(struct sock *sk, int size, int kind);
Eric Dumazet1a24e042015-05-15 12:39:25 -07001280void __sk_mem_reclaim(struct sock *sk, int amount);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001281
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001282#define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1283#define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1284#define SK_MEM_SEND 0
1285#define SK_MEM_RECV 1
Linus Torvalds1da177e2005-04-16 15:20:36 -07001286
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001287static inline int sk_mem_pages(int amt)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001288{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001289 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001290}
1291
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001292static inline bool sk_has_account(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001293{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001294 /* return true if protocol supports memory accounting */
1295 return !!sk->sk_prot->memory_allocated;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001296}
1297
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001298static inline bool sk_wmem_schedule(struct sock *sk, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001299{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001300 if (!sk_has_account(sk))
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001301 return true;
Herbert Xud80d99d2005-09-01 17:48:23 -07001302 return size <= sk->sk_forward_alloc ||
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001303 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1304}
1305
Mel Gormanc76562b2012-07-31 16:44:41 -07001306static inline bool
Chuck Lever35c448a2012-09-17 14:09:11 -07001307sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size)
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001308{
1309 if (!sk_has_account(sk))
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001310 return true;
Mel Gormanc76562b2012-07-31 16:44:41 -07001311 return size<= sk->sk_forward_alloc ||
1312 __sk_mem_schedule(sk, size, SK_MEM_RECV) ||
1313 skb_pfmemalloc(skb);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001314}
1315
1316static inline void sk_mem_reclaim(struct sock *sk)
1317{
1318 if (!sk_has_account(sk))
1319 return;
1320 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
Eric Dumazet1a24e042015-05-15 12:39:25 -07001321 __sk_mem_reclaim(sk, sk->sk_forward_alloc);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001322}
1323
David S. Miller9993e7d2008-01-10 21:56:38 -08001324static inline void sk_mem_reclaim_partial(struct sock *sk)
1325{
1326 if (!sk_has_account(sk))
1327 return;
1328 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
Eric Dumazet1a24e042015-05-15 12:39:25 -07001329 __sk_mem_reclaim(sk, sk->sk_forward_alloc - 1);
David S. Miller9993e7d2008-01-10 21:56:38 -08001330}
1331
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001332static inline void sk_mem_charge(struct sock *sk, int size)
1333{
1334 if (!sk_has_account(sk))
1335 return;
1336 sk->sk_forward_alloc -= size;
1337}
1338
1339static inline void sk_mem_uncharge(struct sock *sk, int size)
1340{
1341 if (!sk_has_account(sk))
1342 return;
1343 sk->sk_forward_alloc += size;
Eric Dumazet20c64d52016-09-15 08:48:46 -07001344
1345 /* Avoid a possible overflow.
1346 * TCP send queues can make this happen, if sk_mem_reclaim()
1347 * is not called and more than 2 GBytes are released at once.
1348 *
1349 * If we reach 2 MBytes, reclaim 1 MBytes right now, there is
1350 * no need to hold that much forward allocation anyway.
1351 */
1352 if (unlikely(sk->sk_forward_alloc >= 1 << 21))
1353 __sk_mem_reclaim(sk, 1 << 20);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001354}
1355
1356static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1357{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001358 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1359 sk->sk_wmem_queued -= skb->truesize;
1360 sk_mem_uncharge(sk, skb->truesize);
1361 __kfree_skb(skb);
Herbert Xud80d99d2005-09-01 17:48:23 -07001362}
1363
Eric Dumazetc3f9b012014-03-10 09:50:11 -07001364static inline void sock_release_ownership(struct sock *sk)
1365{
Hannes Frederic Sowa61881cf2016-04-05 17:10:14 +02001366 if (sk->sk_lock.owned) {
1367 sk->sk_lock.owned = 0;
1368
1369 /* The sk_lock has mutex_unlock() semantics: */
1370 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
1371 }
Eric Dumazetc3f9b012014-03-10 09:50:11 -07001372}
1373
Peter Zijlstraed075362006-12-06 20:35:24 -08001374/*
1375 * Macro so as to not evaluate some arguments when
1376 * lockdep is not enabled.
1377 *
1378 * Mark both the sk_lock and the sk_lock.slock as a
1379 * per-address-family lock class.
1380 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001381#define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
Peter Zijlstraed075362006-12-06 20:35:24 -08001382do { \
Ingo Molnare8f6fbf2008-11-12 01:38:36 +00001383 sk->sk_lock.owned = 0; \
Peter Zijlstraed075362006-12-06 20:35:24 -08001384 init_waitqueue_head(&sk->sk_lock.wq); \
1385 spin_lock_init(&(sk)->sk_lock.slock); \
1386 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1387 sizeof((sk)->sk_lock)); \
1388 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001389 (skey), (sname)); \
Peter Zijlstraed075362006-12-06 20:35:24 -08001390 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1391} while (0)
1392
David S. Millerb33b0a12016-04-07 20:40:25 -04001393#ifdef CONFIG_LOCKDEP
Hannes Frederic Sowa03be9822016-04-07 23:53:35 +02001394static inline bool lockdep_sock_is_held(const struct sock *csk)
Hannes Frederic Sowa1e1d04e2016-04-05 17:10:15 +02001395{
1396 struct sock *sk = (struct sock *)csk;
1397
1398 return lockdep_is_held(&sk->sk_lock) ||
1399 lockdep_is_held(&sk->sk_lock.slock);
1400}
David S. Millerb33b0a12016-04-07 20:40:25 -04001401#endif
Hannes Frederic Sowa1e1d04e2016-04-05 17:10:15 +02001402
Joe Perches69336bd2013-09-22 10:32:26 -07001403void lock_sock_nested(struct sock *sk, int subclass);
Peter Zijlstrafcc70d52006-11-08 22:44:35 -08001404
1405static inline void lock_sock(struct sock *sk)
1406{
1407 lock_sock_nested(sk, 0);
1408}
1409
Joe Perches69336bd2013-09-22 10:32:26 -07001410void release_sock(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001411
1412/* BH context may only use the following locking interface. */
1413#define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
Ingo Molnarc6366182006-07-03 00:25:13 -07001414#define bh_lock_sock_nested(__sk) \
1415 spin_lock_nested(&((__sk)->sk_lock.slock), \
1416 SINGLE_DEPTH_NESTING)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001417#define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1418
Joe Perches69336bd2013-09-22 10:32:26 -07001419bool lock_sock_fast(struct sock *sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001420/**
1421 * unlock_sock_fast - complement of lock_sock_fast
1422 * @sk: socket
1423 * @slow: slow mode
1424 *
1425 * fast unlock socket for user context.
1426 * If slow mode is on, we call regular release_sock()
1427 */
1428static inline void unlock_sock_fast(struct sock *sk, bool slow)
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001429{
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001430 if (slow)
1431 release_sock(sk);
1432 else
1433 spin_unlock_bh(&sk->sk_lock.slock);
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001434}
1435
Hannes Frederic Sowafafc4e12016-04-08 15:11:27 +02001436/* Used by processes to "lock" a socket state, so that
1437 * interrupts and bottom half handlers won't change it
1438 * from under us. It essentially blocks any incoming
1439 * packets, so that we won't get any new data or any
1440 * packets that change the state of the socket.
1441 *
1442 * While locked, BH processing will add new packets to
1443 * the backlog queue. This queue is processed by the
1444 * owner of the socket lock right before it is released.
1445 *
1446 * Since ~2.3.5 it is also exclusive sleep lock serializing
1447 * accesses from user process context.
1448 */
1449
Eric Dumazet46cc6e42016-05-03 16:56:03 -07001450static inline void sock_owned_by_me(const struct sock *sk)
Hannes Frederic Sowafafc4e12016-04-08 15:11:27 +02001451{
1452#ifdef CONFIG_LOCKDEP
Eric Dumazet5e91f6c2016-04-25 06:34:09 -07001453 WARN_ON_ONCE(!lockdep_sock_is_held(sk) && debug_locks);
Hannes Frederic Sowafafc4e12016-04-08 15:11:27 +02001454#endif
Eric Dumazet46cc6e42016-05-03 16:56:03 -07001455}
1456
1457static inline bool sock_owned_by_user(const struct sock *sk)
1458{
1459 sock_owned_by_me(sk);
Hannes Frederic Sowafafc4e12016-04-08 15:11:27 +02001460 return sk->sk_lock.owned;
1461}
1462
1463/* no reclassification while locks are held */
1464static inline bool sock_allow_reclassification(const struct sock *csk)
1465{
1466 struct sock *sk = (struct sock *)csk;
1467
1468 return !sk->sk_lock.owned && !spin_is_locked(&sk->sk_lock.slock);
1469}
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001470
Joe Perches69336bd2013-09-22 10:32:26 -07001471struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
Eric W. Biederman11aa9c22015-05-08 21:09:13 -05001472 struct proto *prot, int kern);
Joe Perches69336bd2013-09-22 10:32:26 -07001473void sk_free(struct sock *sk);
Craig Gallekeb4cb002015-06-15 11:26:18 -04001474void sk_destruct(struct sock *sk);
Joe Perches69336bd2013-09-22 10:32:26 -07001475struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476
Joe Perches69336bd2013-09-22 10:32:26 -07001477struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
1478 gfp_t priority);
Eric Dumazet1d2077a2016-05-02 10:56:27 -07001479void __sock_wfree(struct sk_buff *skb);
Joe Perches69336bd2013-09-22 10:32:26 -07001480void sock_wfree(struct sk_buff *skb);
1481void skb_orphan_partial(struct sk_buff *skb);
1482void sock_rfree(struct sk_buff *skb);
Alexander Duyck62bccb82014-09-04 13:31:35 -04001483void sock_efree(struct sk_buff *skb);
Alexander Duyck82eabd92014-09-04 13:32:11 -04001484#ifdef CONFIG_INET
Joe Perches69336bd2013-09-22 10:32:26 -07001485void sock_edemux(struct sk_buff *skb);
Alexander Duyck82eabd92014-09-04 13:32:11 -04001486#else
1487#define sock_edemux(skb) sock_efree(skb)
1488#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001489
Joe Perches69336bd2013-09-22 10:32:26 -07001490int sock_setsockopt(struct socket *sock, int level, int op,
1491 char __user *optval, unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001492
Joe Perches69336bd2013-09-22 10:32:26 -07001493int sock_getsockopt(struct socket *sock, int level, int op,
1494 char __user *optval, int __user *optlen);
1495struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1496 int noblock, int *errcode);
1497struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1498 unsigned long data_len, int noblock,
1499 int *errcode, int max_page_order);
1500void *sock_kmalloc(struct sock *sk, int size, gfp_t priority);
1501void sock_kfree_s(struct sock *sk, void *mem, int size);
Daniel Borkmann79e88652014-11-19 17:13:11 +01001502void sock_kzfree_s(struct sock *sk, void *mem, int size);
Joe Perches69336bd2013-09-22 10:32:26 -07001503void sk_send_sigurg(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001504
Edward Jeef28ea362015-10-08 14:56:48 -07001505struct sockcm_cookie {
1506 u32 mark;
Soheil Hassas Yeganeh3dd17e62016-04-02 23:08:09 -04001507 u16 tsflags;
Edward Jeef28ea362015-10-08 14:56:48 -07001508};
1509
Willem de Bruijn39771b12016-04-02 23:08:06 -04001510int __sock_cmsg_send(struct sock *sk, struct msghdr *msg, struct cmsghdr *cmsg,
1511 struct sockcm_cookie *sockc);
Edward Jeef28ea362015-10-08 14:56:48 -07001512int sock_cmsg_send(struct sock *sk, struct msghdr *msg,
1513 struct sockcm_cookie *sockc);
1514
Linus Torvalds1da177e2005-04-16 15:20:36 -07001515/*
1516 * Functions to fill in entries in struct proto_ops when a protocol
1517 * does not implement a particular function.
1518 */
Joe Perches69336bd2013-09-22 10:32:26 -07001519int sock_no_bind(struct socket *, struct sockaddr *, int);
1520int sock_no_connect(struct socket *, struct sockaddr *, int, int);
1521int sock_no_socketpair(struct socket *, struct socket *);
1522int sock_no_accept(struct socket *, struct socket *, int);
1523int sock_no_getname(struct socket *, struct sockaddr *, int *, int);
1524unsigned int sock_no_poll(struct file *, struct socket *,
1525 struct poll_table_struct *);
1526int sock_no_ioctl(struct socket *, unsigned int, unsigned long);
1527int sock_no_listen(struct socket *, int);
1528int sock_no_shutdown(struct socket *, int);
1529int sock_no_getsockopt(struct socket *, int , int, char __user *, int __user *);
1530int sock_no_setsockopt(struct socket *, int, int, char __user *, unsigned int);
Ying Xue1b784142015-03-02 15:37:48 +08001531int sock_no_sendmsg(struct socket *, struct msghdr *, size_t);
1532int sock_no_recvmsg(struct socket *, struct msghdr *, size_t, int);
Joe Perches69336bd2013-09-22 10:32:26 -07001533int sock_no_mmap(struct file *file, struct socket *sock,
1534 struct vm_area_struct *vma);
1535ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset,
1536 size_t size, int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001537
1538/*
1539 * Functions to fill in entries in struct proto_ops when a protocol
1540 * uses the inet style.
1541 */
Joe Perches69336bd2013-09-22 10:32:26 -07001542int sock_common_getsockopt(struct socket *sock, int level, int optname,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001543 char __user *optval, int __user *optlen);
Ying Xue1b784142015-03-02 15:37:48 +08001544int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
1545 int flags);
Joe Perches69336bd2013-09-22 10:32:26 -07001546int sock_common_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07001547 char __user *optval, unsigned int optlen);
Joe Perches69336bd2013-09-22 10:32:26 -07001548int compat_sock_common_getsockopt(struct socket *sock, int level,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001549 int optname, char __user *optval, int __user *optlen);
Joe Perches69336bd2013-09-22 10:32:26 -07001550int compat_sock_common_setsockopt(struct socket *sock, int level,
David S. Millerb7058842009-09-30 16:12:20 -07001551 int optname, char __user *optval, unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552
Joe Perches69336bd2013-09-22 10:32:26 -07001553void sk_common_release(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001554
1555/*
1556 * Default socket callbacks and setup code
1557 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001558
Linus Torvalds1da177e2005-04-16 15:20:36 -07001559/* Initialise core socket variables */
Joe Perches69336bd2013-09-22 10:32:26 -07001560void sock_init_data(struct socket *sock, struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001561
Linus Torvalds1da177e2005-04-16 15:20:36 -07001562/*
1563 * Socket reference counting postulates.
1564 *
1565 * * Each user of socket SHOULD hold a reference count.
1566 * * Each access point to socket (an hash table bucket, reference from a list,
1567 * running timer, skb in flight MUST hold a reference count.
1568 * * When reference count hits 0, it means it will never increase back.
1569 * * When reference count hits 0, it means that no references from
1570 * outside exist to this socket and current process on current CPU
1571 * is last user and may/should destroy this socket.
1572 * * sk_free is called from any context: process, BH, IRQ. When
1573 * it is called, socket has no references from outside -> sk_free
1574 * may release descendant resources allocated by the socket, but
1575 * to the time when it is called, socket is NOT referenced by any
1576 * hash tables, lists etc.
1577 * * Packets, delivered from outside (from network or from another process)
1578 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1579 * when they sit in queue. Otherwise, packets will leak to hole, when
1580 * socket is looked up by one cpu and unhasing is made by another CPU.
1581 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1582 * (leak to backlog). Packet socket does all the processing inside
1583 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1584 * use separate SMP lock, so that they are prone too.
1585 */
1586
1587/* Ungrab socket and destroy it, if it was the last reference. */
1588static inline void sock_put(struct sock *sk)
1589{
1590 if (atomic_dec_and_test(&sk->sk_refcnt))
1591 sk_free(sk);
1592}
Eric Dumazet05dbc7b2013-10-03 00:22:02 -07001593/* Generic version of sock_put(), dealing with all sockets
Eric Dumazet41b822c2015-03-12 16:44:08 -07001594 * (TCP_TIMEWAIT, TCP_NEW_SYN_RECV, ESTABLISHED...)
Eric Dumazet05dbc7b2013-10-03 00:22:02 -07001595 */
1596void sock_gen_put(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597
Willem de Bruijn4f0c40d92016-07-12 18:18:57 -04001598int __sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested,
1599 unsigned int trim_cap);
1600static inline int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1601 const int nested)
1602{
1603 return __sk_receive_skb(sk, skb, nested, 1);
1604}
Arnaldo Carvalho de Melo25995ff2005-12-27 02:42:22 -02001605
Krishna Kumare022f0b2009-10-19 23:46:20 +00001606static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1607{
1608 sk->sk_tx_queue_mapping = tx_queue;
1609}
1610
1611static inline void sk_tx_queue_clear(struct sock *sk)
1612{
1613 sk->sk_tx_queue_mapping = -1;
1614}
1615
1616static inline int sk_tx_queue_get(const struct sock *sk)
1617{
Tom Herbertb0f77d02010-07-14 20:50:29 -07001618 return sk ? sk->sk_tx_queue_mapping : -1;
Krishna Kumare022f0b2009-10-19 23:46:20 +00001619}
1620
David S. Miller972692e2008-06-17 22:41:38 -07001621static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1622{
Krishna Kumare022f0b2009-10-19 23:46:20 +00001623 sk_tx_queue_clear(sk);
David S. Miller972692e2008-06-17 22:41:38 -07001624 sk->sk_socket = sock;
1625}
1626
Eric Dumazetaa395142010-04-20 13:03:51 +00001627static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1628{
Eric Dumazeteaefd112011-02-18 03:26:36 +00001629 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1630 return &rcu_dereference_raw(sk->sk_wq)->wait;
Eric Dumazetaa395142010-04-20 13:03:51 +00001631}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001632/* Detach socket from process context.
1633 * Announce socket dead, detach it from wait queue and inode.
1634 * Note that parent inode held reference count on this struct sock,
1635 * we do not release it in this function, because protocol
1636 * probably wants some additional cleanups or even continuing
1637 * to work with this socket (TCP).
1638 */
1639static inline void sock_orphan(struct sock *sk)
1640{
1641 write_lock_bh(&sk->sk_callback_lock);
1642 sock_set_flag(sk, SOCK_DEAD);
David S. Miller972692e2008-06-17 22:41:38 -07001643 sk_set_socket(sk, NULL);
Eric Dumazet43815482010-04-29 11:01:49 +00001644 sk->sk_wq = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001645 write_unlock_bh(&sk->sk_callback_lock);
1646}
1647
1648static inline void sock_graft(struct sock *sk, struct socket *parent)
1649{
1650 write_lock_bh(&sk->sk_callback_lock);
Eric Dumazeteaefd112011-02-18 03:26:36 +00001651 sk->sk_wq = parent->wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001652 parent->sk = sk;
David S. Miller972692e2008-06-17 22:41:38 -07001653 sk_set_socket(sk, parent);
Venkat Yekkirala4237c752006-07-24 23:32:50 -07001654 security_sock_graft(sk, parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001655 write_unlock_bh(&sk->sk_callback_lock);
1656}
1657
Joe Perches69336bd2013-09-22 10:32:26 -07001658kuid_t sock_i_uid(struct sock *sk);
1659unsigned long sock_i_ino(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001660
Eric Dumazet58d607d2015-09-15 15:24:20 -07001661static inline u32 net_tx_rndhash(void)
1662{
1663 u32 v = prandom_u32();
1664
1665 return v ?: 1;
1666}
1667
Tom Herbert877d1f62015-07-28 16:02:05 -07001668static inline void sk_set_txhash(struct sock *sk)
1669{
Eric Dumazet58d607d2015-09-15 15:24:20 -07001670 sk->sk_txhash = net_tx_rndhash();
Tom Herbert877d1f62015-07-28 16:02:05 -07001671}
1672
Tom Herbert265f94f2015-07-28 16:02:06 -07001673static inline void sk_rethink_txhash(struct sock *sk)
1674{
1675 if (sk->sk_txhash)
1676 sk_set_txhash(sk);
1677}
1678
Linus Torvalds1da177e2005-04-16 15:20:36 -07001679static inline struct dst_entry *
1680__sk_dst_get(struct sock *sk)
1681{
Hannes Frederic Sowa1e1d04e2016-04-05 17:10:15 +02001682 return rcu_dereference_check(sk->sk_dst_cache,
1683 lockdep_sock_is_held(sk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001684}
1685
1686static inline struct dst_entry *
1687sk_dst_get(struct sock *sk)
1688{
1689 struct dst_entry *dst;
1690
Eric Dumazetb6c67122010-04-08 23:03:29 +00001691 rcu_read_lock();
1692 dst = rcu_dereference(sk->sk_dst_cache);
Eric Dumazetf8864972014-06-24 10:05:11 -07001693 if (dst && !atomic_inc_not_zero(&dst->__refcnt))
1694 dst = NULL;
Eric Dumazetb6c67122010-04-08 23:03:29 +00001695 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001696 return dst;
1697}
1698
Eric Dumazetb6c67122010-04-08 23:03:29 +00001699static inline void dst_negative_advice(struct sock *sk)
1700{
1701 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1702
Tom Herbert265f94f2015-07-28 16:02:06 -07001703 sk_rethink_txhash(sk);
1704
Eric Dumazetb6c67122010-04-08 23:03:29 +00001705 if (dst && dst->ops->negative_advice) {
1706 ndst = dst->ops->negative_advice(dst);
1707
1708 if (ndst != dst) {
1709 rcu_assign_pointer(sk->sk_dst_cache, ndst);
ZHAO Gang0a6957e2013-10-22 16:23:38 +08001710 sk_tx_queue_clear(sk);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001711 }
1712 }
1713}
1714
Linus Torvalds1da177e2005-04-16 15:20:36 -07001715static inline void
1716__sk_dst_set(struct sock *sk, struct dst_entry *dst)
1717{
1718 struct dst_entry *old_dst;
1719
Krishna Kumare022f0b2009-10-19 23:46:20 +00001720 sk_tx_queue_clear(sk);
Eric Dumazet0b53ff22010-04-26 20:40:43 +00001721 /*
1722 * This can be called while sk is owned by the caller only,
1723 * with no state that can be checked in a rcu_dereference_check() cond
1724 */
1725 old_dst = rcu_dereference_raw(sk->sk_dst_cache);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001726 rcu_assign_pointer(sk->sk_dst_cache, dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001727 dst_release(old_dst);
1728}
1729
1730static inline void
1731sk_dst_set(struct sock *sk, struct dst_entry *dst)
1732{
Eric Dumazet7f502362014-06-30 01:26:23 -07001733 struct dst_entry *old_dst;
1734
1735 sk_tx_queue_clear(sk);
Eric Dumazet5925a052014-07-02 02:39:38 -07001736 old_dst = xchg((__force struct dst_entry **)&sk->sk_dst_cache, dst);
Eric Dumazet7f502362014-06-30 01:26:23 -07001737 dst_release(old_dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001738}
1739
1740static inline void
1741__sk_dst_reset(struct sock *sk)
1742{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001743 __sk_dst_set(sk, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744}
1745
1746static inline void
1747sk_dst_reset(struct sock *sk)
1748{
Eric Dumazet7f502362014-06-30 01:26:23 -07001749 sk_dst_set(sk, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750}
1751
Joe Perches69336bd2013-09-22 10:32:26 -07001752struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001753
Joe Perches69336bd2013-09-22 10:32:26 -07001754struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001755
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +02001756bool sk_mc_loop(struct sock *sk);
1757
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001758static inline bool sk_can_gso(const struct sock *sk)
Herbert Xubcd76112006-06-30 13:36:35 -07001759{
1760 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1761}
1762
Joe Perches69336bd2013-09-22 10:32:26 -07001763void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
Arnaldo Carvalho de Melo6cbb0df2005-08-09 19:49:02 -07001764
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001765static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
Eric Dumazeta4654192010-05-16 00:36:33 -07001766{
1767 sk->sk_route_nocaps |= flags;
1768 sk->sk_route_caps &= ~flags;
1769}
1770
Tom Herbert9a498502015-12-14 11:19:45 -08001771static inline bool sk_check_csum_caps(struct sock *sk)
1772{
1773 return (sk->sk_route_caps & NETIF_F_HW_CSUM) ||
1774 (sk->sk_family == PF_INET &&
1775 (sk->sk_route_caps & NETIF_F_IP_CSUM)) ||
1776 (sk->sk_family == PF_INET6 &&
1777 (sk->sk_route_caps & NETIF_F_IPV6_CSUM));
1778}
1779
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001780static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
Al Viro57be5bd2014-11-28 13:40:20 -05001781 struct iov_iter *from, char *to,
Wei Yongjun912d3982011-04-06 18:40:12 +00001782 int copy, int offset)
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001783{
1784 if (skb->ip_summed == CHECKSUM_NONE) {
Al Viro57be5bd2014-11-28 13:40:20 -05001785 __wsum csum = 0;
1786 if (csum_and_copy_from_iter(to, copy, &csum, from) != copy)
1787 return -EFAULT;
Wei Yongjun912d3982011-04-06 18:40:12 +00001788 skb->csum = csum_block_add(skb->csum, csum, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001789 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
Al Viro57be5bd2014-11-28 13:40:20 -05001790 if (copy_from_iter_nocache(to, copy, from) != copy)
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001791 return -EFAULT;
Al Viro57be5bd2014-11-28 13:40:20 -05001792 } else if (copy_from_iter(to, copy, from) != copy)
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001793 return -EFAULT;
1794
1795 return 0;
1796}
1797
1798static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
Al Viro57be5bd2014-11-28 13:40:20 -05001799 struct iov_iter *from, int copy)
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001800{
Wei Yongjun912d3982011-04-06 18:40:12 +00001801 int err, offset = skb->len;
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001802
Wei Yongjun912d3982011-04-06 18:40:12 +00001803 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1804 copy, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001805 if (err)
Wei Yongjun912d3982011-04-06 18:40:12 +00001806 __skb_trim(skb, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001807
1808 return err;
1809}
1810
Al Viro57be5bd2014-11-28 13:40:20 -05001811static inline int skb_copy_to_page_nocache(struct sock *sk, struct iov_iter *from,
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001812 struct sk_buff *skb,
1813 struct page *page,
1814 int off, int copy)
1815{
1816 int err;
1817
Wei Yongjun912d3982011-04-06 18:40:12 +00001818 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1819 copy, skb->len);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001820 if (err)
1821 return err;
1822
1823 skb->len += copy;
1824 skb->data_len += copy;
1825 skb->truesize += copy;
1826 sk->sk_wmem_queued += copy;
1827 sk_mem_charge(sk, copy);
1828 return 0;
1829}
1830
Eric Dumazetc5640392009-06-16 10:12:03 +00001831/**
1832 * sk_wmem_alloc_get - returns write allocations
1833 * @sk: socket
1834 *
1835 * Returns sk_wmem_alloc minus initial offset of one
1836 */
1837static inline int sk_wmem_alloc_get(const struct sock *sk)
1838{
1839 return atomic_read(&sk->sk_wmem_alloc) - 1;
1840}
1841
1842/**
1843 * sk_rmem_alloc_get - returns read allocations
1844 * @sk: socket
1845 *
1846 * Returns sk_rmem_alloc
1847 */
1848static inline int sk_rmem_alloc_get(const struct sock *sk)
1849{
1850 return atomic_read(&sk->sk_rmem_alloc);
1851}
1852
1853/**
1854 * sk_has_allocations - check if allocations are outstanding
1855 * @sk: socket
1856 *
1857 * Returns true if socket has write or read allocations
1858 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001859static inline bool sk_has_allocations(const struct sock *sk)
Eric Dumazetc5640392009-06-16 10:12:03 +00001860{
1861 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1862}
1863
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001864/**
Herbert Xu1ce0bf52015-11-26 13:55:39 +08001865 * skwq_has_sleeper - check if there are any waiting processes
Randy Dunlapacfbe962010-05-24 23:54:18 -07001866 * @wq: struct socket_wq
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001867 *
Eric Dumazet43815482010-04-29 11:01:49 +00001868 * Returns true if socket_wq has waiting processes
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001869 *
Herbert Xu1ce0bf52015-11-26 13:55:39 +08001870 * The purpose of the skwq_has_sleeper and sock_poll_wait is to wrap the memory
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001871 * barrier call. They were added due to the race found within the tcp code.
1872 *
1873 * Consider following tcp code paths:
1874 *
1875 * CPU1 CPU2
1876 *
1877 * sys_select receive packet
1878 * ... ...
1879 * __add_wait_queue update tp->rcv_nxt
1880 * ... ...
1881 * tp->rcv_nxt check sock_def_readable
1882 * ... {
Eric Dumazet43815482010-04-29 11:01:49 +00001883 * schedule rcu_read_lock();
1884 * wq = rcu_dereference(sk->sk_wq);
1885 * if (wq && waitqueue_active(&wq->wait))
1886 * wake_up_interruptible(&wq->wait)
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001887 * ...
1888 * }
1889 *
1890 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1891 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1892 * could then endup calling schedule and sleep forever if there are no more
1893 * data on the socket.
Jiri Olsaad462762009-07-08 12:10:31 +00001894 *
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001895 */
Herbert Xu1ce0bf52015-11-26 13:55:39 +08001896static inline bool skwq_has_sleeper(struct socket_wq *wq)
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001897{
Herbert Xu1ce0bf52015-11-26 13:55:39 +08001898 return wq && wq_has_sleeper(&wq->wait);
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001899}
1900
1901/**
1902 * sock_poll_wait - place memory barrier behind the poll_wait call.
1903 * @filp: file
1904 * @wait_address: socket wait queue
1905 * @p: poll_table
1906 *
Eric Dumazet43815482010-04-29 11:01:49 +00001907 * See the comments in the wq_has_sleeper function.
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001908 */
1909static inline void sock_poll_wait(struct file *filp,
1910 wait_queue_head_t *wait_address, poll_table *p)
1911{
Hans Verkuil626cf232012-03-23 15:02:27 -07001912 if (!poll_does_not_wait(p) && wait_address) {
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001913 poll_wait(filp, wait_address, p);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001914 /* We need to be sure we are in sync with the
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001915 * socket flags modification.
1916 *
Eric Dumazet43815482010-04-29 11:01:49 +00001917 * This memory barrier is paired in the wq_has_sleeper.
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001918 */
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001919 smp_mb();
1920 }
1921}
1922
Tom Herbertb73c3d02014-07-01 21:32:17 -07001923static inline void skb_set_hash_from_sk(struct sk_buff *skb, struct sock *sk)
1924{
1925 if (sk->sk_txhash) {
1926 skb->l4_hash = 1;
1927 skb->hash = sk->sk_txhash;
1928 }
1929}
1930
Eric Dumazet9e17f8a2015-11-01 15:36:55 -08001931void skb_set_owner_w(struct sk_buff *skb, struct sock *sk);
1932
Linus Torvalds1da177e2005-04-16 15:20:36 -07001933/*
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001934 * Queue a received datagram if it will fit. Stream and sequenced
Linus Torvalds1da177e2005-04-16 15:20:36 -07001935 * protocols can't normally use this as they need to fit buffers in
1936 * and play with them.
1937 *
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001938 * Inlined as it's very short and called for pretty much every
Linus Torvalds1da177e2005-04-16 15:20:36 -07001939 * packet ever received.
1940 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001941static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1942{
Herbert Xud55d87f2009-06-22 02:25:25 +00001943 skb_orphan(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001944 skb->sk = sk;
1945 skb->destructor = sock_rfree;
1946 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001947 sk_mem_charge(sk, skb->truesize);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001948}
1949
Joe Perches69336bd2013-09-22 10:32:26 -07001950void sk_reset_timer(struct sock *sk, struct timer_list *timer,
1951 unsigned long expires);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001952
Joe Perches69336bd2013-09-22 10:32:26 -07001953void sk_stop_timer(struct sock *sk, struct timer_list *timer);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001954
samanthakumare6afc8a2016-04-05 12:41:15 -04001955int __sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
Joe Perches69336bd2013-09-22 10:32:26 -07001956int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957
Joe Perches69336bd2013-09-22 10:32:26 -07001958int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
Willem de Bruijn364a9e92014-08-31 21:30:27 -04001959struct sk_buff *sock_dequeue_err_skb(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001960
1961/*
1962 * Recover an error report and clear atomically
1963 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001964
Linus Torvalds1da177e2005-04-16 15:20:36 -07001965static inline int sock_error(struct sock *sk)
1966{
Benjamin LaHaisec1cbe4b2005-12-13 23:22:19 -08001967 int err;
1968 if (likely(!sk->sk_err))
1969 return 0;
1970 err = xchg(&sk->sk_err, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001971 return -err;
1972}
1973
1974static inline unsigned long sock_wspace(struct sock *sk)
1975{
1976 int amt = 0;
1977
1978 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1979 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001980 if (amt < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001981 amt = 0;
1982 }
1983 return amt;
1984}
1985
Eric Dumazetceb5d582015-11-29 20:03:11 -08001986/* Note:
1987 * We use sk->sk_wq_raw, from contexts knowing this
1988 * pointer is not NULL and cannot disappear/change.
1989 */
Eric Dumazet9cd3e072015-11-29 20:03:10 -08001990static inline void sk_set_bit(int nr, struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001991{
Eric Dumazet4be73522016-04-25 10:39:34 -07001992 if ((nr == SOCKWQ_ASYNC_NOSPACE || nr == SOCKWQ_ASYNC_WAITDATA) &&
1993 !sock_flag(sk, SOCK_FASYNC))
Eric Dumazet9317bb62016-04-25 10:39:32 -07001994 return;
1995
Eric Dumazetceb5d582015-11-29 20:03:11 -08001996 set_bit(nr, &sk->sk_wq_raw->flags);
Eric Dumazet9cd3e072015-11-29 20:03:10 -08001997}
1998
1999static inline void sk_clear_bit(int nr, struct sock *sk)
2000{
Eric Dumazet4be73522016-04-25 10:39:34 -07002001 if ((nr == SOCKWQ_ASYNC_NOSPACE || nr == SOCKWQ_ASYNC_WAITDATA) &&
2002 !sock_flag(sk, SOCK_FASYNC))
Eric Dumazet9317bb62016-04-25 10:39:32 -07002003 return;
2004
Eric Dumazetceb5d582015-11-29 20:03:11 -08002005 clear_bit(nr, &sk->sk_wq_raw->flags);
Eric Dumazet9cd3e072015-11-29 20:03:10 -08002006}
2007
Eric Dumazetceb5d582015-11-29 20:03:11 -08002008static inline void sk_wake_async(const struct sock *sk, int how, int band)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009{
Eric Dumazetceb5d582015-11-29 20:03:11 -08002010 if (sock_flag(sk, SOCK_FASYNC)) {
2011 rcu_read_lock();
2012 sock_wake_async(rcu_dereference(sk->sk_wq), how, band);
2013 rcu_read_unlock();
2014 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002015}
2016
Daniel Borkmanneea86af2013-06-19 12:51:20 +02002017/* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
2018 * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
2019 * Note: for send buffers, TCP works better if we can build two skbs at
2020 * minimum.
Eric Dumazet7a91b4342010-09-26 18:53:07 -07002021 */
Eric Dumazet9eb5bf82013-07-03 05:02:22 -07002022#define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
Daniel Borkmanneea86af2013-06-19 12:51:20 +02002023
2024#define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2)
2025#define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE
Linus Torvalds1da177e2005-04-16 15:20:36 -07002026
2027static inline void sk_stream_moderate_sndbuf(struct sock *sk)
2028{
2029 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
Eric Dumazet8df09ea2007-12-21 03:07:41 -08002030 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
Daniel Borkmanneea86af2013-06-19 12:51:20 +02002031 sk->sk_sndbuf = max_t(u32, sk->sk_sndbuf, SOCK_MIN_SNDBUF);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002032 }
2033}
2034
Eric Dumazeteb934472015-05-19 13:26:55 -07002035struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp,
2036 bool force_schedule);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002037
Eric Dumazet5640f762012-09-23 23:04:42 +00002038/**
2039 * sk_page_frag - return an appropriate page_frag
2040 * @sk: socket
2041 *
2042 * If socket allocation mode allows current thread to sleep, it means its
2043 * safe to use the per task page_frag instead of the per socket one.
2044 */
2045static inline struct page_frag *sk_page_frag(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002046{
Mel Gormand0164ad2015-11-06 16:28:21 -08002047 if (gfpflags_allow_blocking(sk->sk_allocation))
Eric Dumazet5640f762012-09-23 23:04:42 +00002048 return &current->task_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002049
Eric Dumazet5640f762012-09-23 23:04:42 +00002050 return &sk->sk_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002051}
2052
Joe Perches69336bd2013-09-22 10:32:26 -07002053bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag);
Eric Dumazet5640f762012-09-23 23:04:42 +00002054
Linus Torvalds1da177e2005-04-16 15:20:36 -07002055/*
2056 * Default write policy as shown to user space via poll/select/SIGIO
2057 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002058static inline bool sock_writeable(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002059{
Eric Dumazet8df09ea2007-12-21 03:07:41 -08002060 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002061}
2062
Al Virodd0fc662005-10-07 07:46:04 +01002063static inline gfp_t gfp_any(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002064{
Andrew Morton99709372009-02-12 16:43:17 -08002065 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002066}
2067
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002068static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002069{
2070 return noblock ? 0 : sk->sk_rcvtimeo;
2071}
2072
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002073static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002074{
2075 return noblock ? 0 : sk->sk_sndtimeo;
2076}
2077
2078static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2079{
2080 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
2081}
2082
2083/* Alas, with timeout socket operations are not restartable.
2084 * Compare this to poll().
2085 */
2086static inline int sock_intr_errno(long timeo)
2087{
2088 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2089}
2090
Eyal Birger744d5a32015-03-01 14:58:31 +02002091struct sock_skb_cb {
2092 u32 dropcount;
2093};
2094
2095/* Store sock_skb_cb at the end of skb->cb[] so protocol families
2096 * using skb->cb[] would keep using it directly and utilize its
2097 * alignement guarantee.
2098 */
2099#define SOCK_SKB_CB_OFFSET ((FIELD_SIZEOF(struct sk_buff, cb) - \
2100 sizeof(struct sock_skb_cb)))
2101
2102#define SOCK_SKB_CB(__skb) ((struct sock_skb_cb *)((__skb)->cb + \
2103 SOCK_SKB_CB_OFFSET))
2104
Eyal Birgerb4772ef2015-03-01 14:58:29 +02002105#define sock_skb_cb_check_size(size) \
Eyal Birger744d5a32015-03-01 14:58:31 +02002106 BUILD_BUG_ON((size) > SOCK_SKB_CB_OFFSET)
Eyal Birgerb4772ef2015-03-01 14:58:29 +02002107
Eyal Birger3bc3b962015-03-01 14:58:30 +02002108static inline void
2109sock_skb_set_dropcount(const struct sock *sk, struct sk_buff *skb)
2110{
Eyal Birger744d5a32015-03-01 14:58:31 +02002111 SOCK_SKB_CB(skb)->dropcount = atomic_read(&sk->sk_drops);
Eyal Birger3bc3b962015-03-01 14:58:30 +02002112}
2113
Eric Dumazet532182c2016-04-01 08:52:19 -07002114static inline void sk_drops_add(struct sock *sk, const struct sk_buff *skb)
2115{
2116 int segs = max_t(u16, 1, skb_shinfo(skb)->gso_segs);
2117
2118 atomic_add(segs, &sk->sk_drops);
2119}
2120
Joe Perches69336bd2013-09-22 10:32:26 -07002121void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2122 struct sk_buff *skb);
2123void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2124 struct sk_buff *skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -07002125
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002126static inline void
Linus Torvalds1da177e2005-04-16 15:20:36 -07002127sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2128{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002129 ktime_t kt = skb->tstamp;
Patrick Ohly20d49472009-02-12 05:03:38 +00002130 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002131
Patrick Ohly20d49472009-02-12 05:03:38 +00002132 /*
2133 * generate control messages if
Willem de Bruijnb9f40e22014-08-04 22:11:46 -04002134 * - receive time stamping in software requested
Patrick Ohly20d49472009-02-12 05:03:38 +00002135 * - software time stamp available and wanted
Patrick Ohly20d49472009-02-12 05:03:38 +00002136 * - hardware time stamps available and wanted
Patrick Ohly20d49472009-02-12 05:03:38 +00002137 */
2138 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
Willem de Bruijnb9f40e22014-08-04 22:11:46 -04002139 (sk->sk_tsflags & SOF_TIMESTAMPING_RX_SOFTWARE) ||
Willem de Bruijnc1991052014-09-03 12:01:18 -04002140 (kt.tv64 && sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) ||
Patrick Ohly20d49472009-02-12 05:03:38 +00002141 (hwtstamps->hwtstamp.tv64 &&
Willem de Bruijnb9f40e22014-08-04 22:11:46 -04002142 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE)))
Eric Dumazet92f37fd2007-03-25 22:14:49 -07002143 __sock_recv_timestamp(msg, sk, skb);
2144 else
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002145 sk->sk_stamp = kt;
Johannes Berg6e3e9392011-11-09 10:15:42 +01002146
2147 if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2148 __sock_recv_wifi_status(msg, sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002149}
2150
Joe Perches69336bd2013-09-22 10:32:26 -07002151void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2152 struct sk_buff *skb);
Eric Dumazet767dd032010-04-28 19:14:43 +00002153
2154static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2155 struct sk_buff *skb)
2156{
2157#define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
Willem de Bruijnb9f40e22014-08-04 22:11:46 -04002158 (1UL << SOCK_RCVTSTAMP))
2159#define TSFLAGS_ANY (SOF_TIMESTAMPING_SOFTWARE | \
2160 SOF_TIMESTAMPING_RAW_HARDWARE)
Eric Dumazet767dd032010-04-28 19:14:43 +00002161
Willem de Bruijnb9f40e22014-08-04 22:11:46 -04002162 if (sk->sk_flags & FLAGS_TS_OR_DROPS || sk->sk_tsflags & TSFLAGS_ANY)
Eric Dumazet767dd032010-04-28 19:14:43 +00002163 __sock_recv_ts_and_drops(msg, sk, skb);
2164 else
2165 sk->sk_stamp = skb->tstamp;
2166}
Neil Horman3b885782009-10-12 13:26:31 -07002167
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -04002168void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags);
Willem de Bruijn67cc0d42014-09-08 19:58:58 -04002169
Linus Torvalds1da177e2005-04-16 15:20:36 -07002170/**
Patrick Ohly20d49472009-02-12 05:03:38 +00002171 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
Patrick Ohly20d49472009-02-12 05:03:38 +00002172 * @sk: socket sending this packet
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -04002173 * @tsflags: timestamping flags to use
Eric Dumazet140c55d2014-08-06 11:49:29 +02002174 * @tx_flags: completed with instructions for time stamping
2175 *
2176 * Note : callers should take care of initial *tx_flags value (usually 0)
Patrick Ohly20d49472009-02-12 05:03:38 +00002177 */
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -04002178static inline void sock_tx_timestamp(const struct sock *sk, __u16 tsflags,
2179 __u8 *tx_flags)
Willem de Bruijn67cc0d42014-09-08 19:58:58 -04002180{
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -04002181 if (unlikely(tsflags))
2182 __sock_tx_timestamp(tsflags, tx_flags);
Willem de Bruijn67cc0d42014-09-08 19:58:58 -04002183 if (unlikely(sock_flag(sk, SOCK_WIFI_STATUS)))
2184 *tx_flags |= SKBTX_WIFI_STATUS;
2185}
Patrick Ohly20d49472009-02-12 05:03:38 +00002186
2187/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002188 * sk_eat_skb - Release a skb if it is no longer needed
Pavel Pisa4dc3b162005-05-01 08:59:25 -07002189 * @sk: socket to eat this skb from
2190 * @skb: socket buffer to eat
Linus Torvalds1da177e2005-04-16 15:20:36 -07002191 *
2192 * This routine must be called with interrupts disabled or with the socket
2193 * locked so that the sk_buff queue operation is ok.
2194*/
Dan Williams7bced392013-12-30 12:37:29 -08002195static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002196{
2197 __skb_unlink(skb, &sk->sk_receive_queue);
2198 __kfree_skb(skb);
2199}
2200
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002201static inline
2202struct net *sock_net(const struct sock *sk)
2203{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002204 return read_pnet(&sk->sk_net);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002205}
2206
2207static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07002208void sock_net_set(struct sock *sk, struct net *net)
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002209{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002210 write_pnet(&sk->sk_net, net);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002211}
2212
KOVACS Krisztian23542612008-10-07 12:41:01 -07002213static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2214{
Vijay Subramanianefc27f82012-06-24 13:03:07 +00002215 if (skb->sk) {
KOVACS Krisztian23542612008-10-07 12:41:01 -07002216 struct sock *sk = skb->sk;
2217
2218 skb->destructor = NULL;
2219 skb->sk = NULL;
2220 return sk;
2221 }
2222 return NULL;
2223}
2224
Eric Dumazet1d0ab252015-03-15 21:12:12 -07002225/* This helper checks if a socket is a full socket,
2226 * ie _not_ a timewait or request socket.
2227 */
2228static inline bool sk_fullsock(const struct sock *sk)
2229{
2230 return (1 << sk->sk_state) & ~(TCPF_TIME_WAIT | TCPF_NEW_SYN_RECV);
2231}
2232
Eric Dumazete446f9d2015-10-08 05:01:55 -07002233/* This helper checks if a socket is a LISTEN or NEW_SYN_RECV
2234 * SYNACK messages can be attached to either ones (depending on SYNCOOKIE)
2235 */
2236static inline bool sk_listener(const struct sock *sk)
2237{
2238 return (1 << sk->sk_state) & (TCPF_LISTEN | TCPF_NEW_SYN_RECV);
2239}
2240
Eric Dumazet00fd38d2015-11-12 08:43:18 -08002241/**
2242 * sk_state_load - read sk->sk_state for lockless contexts
2243 * @sk: socket pointer
2244 *
2245 * Paired with sk_state_store(). Used in places we do not hold socket lock :
2246 * tcp_diag_get_info(), tcp_get_info(), tcp_poll(), get_tcp4_sock() ...
2247 */
2248static inline int sk_state_load(const struct sock *sk)
2249{
2250 return smp_load_acquire(&sk->sk_state);
2251}
2252
2253/**
2254 * sk_state_store - update sk->sk_state
2255 * @sk: socket pointer
2256 * @newstate: new state
2257 *
2258 * Paired with sk_state_load(). Should be used in contexts where
2259 * state change might impact lockless readers.
2260 */
2261static inline void sk_state_store(struct sock *sk, int newstate)
2262{
2263 smp_store_release(&sk->sk_state, newstate);
2264}
2265
Joe Perches69336bd2013-09-22 10:32:26 -07002266void sock_enable_timestamp(struct sock *sk, int flag);
2267int sock_get_timestamp(struct sock *, struct timeval __user *);
2268int sock_get_timestampns(struct sock *, struct timespec __user *);
2269int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len, int level,
2270 int type);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002271
Eric W. Biedermana3b299d2014-04-23 14:26:56 -07002272bool sk_ns_capable(const struct sock *sk,
2273 struct user_namespace *user_ns, int cap);
2274bool sk_capable(const struct sock *sk, int cap);
2275bool sk_net_capable(const struct sock *sk, int cap);
2276
Linus Torvalds1da177e2005-04-16 15:20:36 -07002277extern __u32 sysctl_wmem_max;
2278extern __u32 sysctl_rmem_max;
2279
Willem de Bruijnb245be12015-01-30 13:29:32 -05002280extern int sysctl_tstamp_allow_data;
David S. Miller6baf1f42005-09-05 18:14:11 -07002281extern int sysctl_optmem_max;
2282
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002283extern __u32 sysctl_wmem_default;
2284extern __u32 sysctl_rmem_default;
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002285
Linus Torvalds1da177e2005-04-16 15:20:36 -07002286#endif /* _SOCK_H */