blob: f5ea148853e2f76257aca80a4aef65a30d587538 [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;
181 struct hlist_nulls_node skc_portaddr_node;
182 };
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
Tom Herbert28448b82014-05-23 08:47:19 -0700255 * @sk_no_check_tx: %SO_NO_CHECK setting, set checksum in TX packets
256 * @sk_no_check_rx: allow zero checksum in RX packets
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700257 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
Eric Dumazeta4654192010-05-16 00:36:33 -0700258 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
Herbert Xubcd76112006-06-30 13:36:35 -0700259 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -0700260 * @sk_gso_max_size: Maximum GSO segment size to build
Ben Hutchings14853482012-07-30 16:11:42 +0000261 * @sk_gso_max_segs: Maximum number of GSO segments
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700262 * @sk_lingertime: %SO_LINGER l_linger setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700263 * @sk_backlog: always used with the per-socket spinlock held
264 * @sk_callback_lock: used with the callbacks in the end of this struct
265 * @sk_error_queue: rarely used
Wang Chen33c732c2007-11-13 20:30:01 -0800266 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
267 * IPV6_ADDRFORM for instance)
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700268 * @sk_err: last error
Wang Chen33c732c2007-11-13 20:30:01 -0800269 * @sk_err_soft: errors that don't cause failure but are the cause of a
270 * persistent failure not just 'timed out'
Eric Dumazetcb61cb92008-06-17 21:04:56 -0700271 * @sk_drops: raw/udp drops counter
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700272 * @sk_ack_backlog: current listen backlog
273 * @sk_max_ack_backlog: listen backlog set in listen()
274 * @sk_priority: %SO_PRIORITY setting
275 * @sk_type: socket type (%SOCK_STREAM, etc)
276 * @sk_protocol: which protocol this socket belongs in this network family
Randy Dunlap53c3fa22010-08-09 13:41:07 +0000277 * @sk_peer_pid: &struct pid for this socket's peer
278 * @sk_peer_cred: %SO_PEERCRED setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700279 * @sk_rcvlowat: %SO_RCVLOWAT setting
280 * @sk_rcvtimeo: %SO_RCVTIMEO setting
281 * @sk_sndtimeo: %SO_SNDTIMEO setting
Tom Herbertb73c3d02014-07-01 21:32:17 -0700282 * @sk_txhash: computed flow hash for use on transmit
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700283 * @sk_filter: socket filtering instructions
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700284 * @sk_timer: sock cleanup timer
285 * @sk_stamp: time stamp of last packet received
Willem de Bruijnb9f40e22014-08-04 22:11:46 -0400286 * @sk_tsflags: SO_TIMESTAMPING socket options
Willem de Bruijn09c2d252014-08-04 22:11:47 -0400287 * @sk_tskey: counter to disambiguate concurrent tstamp requests
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700288 * @sk_socket: Identd and reporting IO signals
289 * @sk_user_data: RPC layer private data
Eric Dumazet5640f762012-09-23 23:04:42 +0000290 * @sk_frag: cached page frag
Randy Dunlapd3d4f0a2012-04-17 14:03:53 +0000291 * @sk_peek_off: current peek_offset value
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700292 * @sk_send_head: front of stuff to transmit
Martin Waitz67be2dd2005-05-01 08:59:26 -0700293 * @sk_security: used by security modules
Randy Dunlap31729362008-02-18 20:52:13 -0800294 * @sk_mark: generic packet mark
Tejun Heo2a56a1f2015-12-07 17:38:52 -0500295 * @sk_cgrp_data: cgroup data for this cgroup
Johannes Weinerbaac50b2016-01-14 15:21:17 -0800296 * @sk_memcg: this socket's memory cgroup association
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700297 * @sk_write_pending: a write to stream socket waits to start
298 * @sk_state_change: callback to indicate change in the state of the sock
299 * @sk_data_ready: callback to indicate there is data to be processed
300 * @sk_write_space: callback to indicate there is bf sending space available
301 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
302 * @sk_backlog_rcv: callback to process the backlog
303 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
Craig Gallekef456142016-01-04 17:41:45 -0500304 * @sk_reuseport_cb: reuseport group container
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305 */
306struct sock {
307 /*
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700308 * Now struct inet_timewait_sock also uses sock_common, so please just
Linus Torvalds1da177e2005-04-16 15:20:36 -0700309 * don't add nothing before this first member (__sk_common) --acme
310 */
311 struct sock_common __sk_common;
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000312#define sk_node __sk_common.skc_node
313#define sk_nulls_node __sk_common.skc_nulls_node
314#define sk_refcnt __sk_common.skc_refcnt
Krishna Kumare022f0b2009-10-19 23:46:20 +0000315#define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000316
Eric Dumazet68835ab2010-11-30 19:04:07 +0000317#define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
318#define sk_dontcopy_end __sk_common.skc_dontcopy_end
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000319#define sk_hash __sk_common.skc_hash
Eric Dumazet50805462013-10-02 04:29:50 -0700320#define sk_portpair __sk_common.skc_portpair
Eric Dumazet05dbc7b2013-10-03 00:22:02 -0700321#define sk_num __sk_common.skc_num
322#define sk_dport __sk_common.skc_dport
Eric Dumazet50805462013-10-02 04:29:50 -0700323#define sk_addrpair __sk_common.skc_addrpair
324#define sk_daddr __sk_common.skc_daddr
325#define sk_rcv_saddr __sk_common.skc_rcv_saddr
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326#define sk_family __sk_common.skc_family
327#define sk_state __sk_common.skc_state
328#define sk_reuse __sk_common.skc_reuse
Tom Herbert055dc212013-01-22 09:49:50 +0000329#define sk_reuseport __sk_common.skc_reuseport
Eric Dumazet9fe516b2014-06-27 08:36:16 -0700330#define sk_ipv6only __sk_common.skc_ipv6only
Eric W. Biederman26abe142015-05-08 21:10:31 -0500331#define sk_net_refcnt __sk_common.skc_net_refcnt
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332#define sk_bound_dev_if __sk_common.skc_bound_dev_if
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333#define sk_bind_node __sk_common.skc_bind_node
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700334#define sk_prot __sk_common.skc_prot
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200335#define sk_net __sk_common.skc_net
Eric Dumazetefe42082013-10-03 15:42:29 -0700336#define sk_v6_daddr __sk_common.skc_v6_daddr
337#define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr
Eric Dumazet33cf7c92015-03-11 18:53:14 -0700338#define sk_cookie __sk_common.skc_cookie
Eric Dumazet70da2682015-10-08 19:33:21 -0700339#define sk_incoming_cpu __sk_common.skc_incoming_cpu
Eric Dumazet8e5eb542015-10-08 19:33:22 -0700340#define sk_flags __sk_common.skc_flags
Eric Dumazeted53d0a2015-10-08 19:33:23 -0700341#define sk_rxhash __sk_common.skc_rxhash
Eric Dumazetefe42082013-10-03 15:42:29 -0700342
Eric Dumazetb178bb32010-11-16 05:56:04 +0000343 socket_lock_t sk_lock;
344 struct sk_buff_head sk_receive_queue;
345 /*
346 * The backlog queue is special, it is always used with
347 * the per-socket spinlock held and requires low latency
348 * access. Therefore we special case it's implementation.
349 * Note : rmem_alloc is in this structure to fill a hole
350 * on 64bit arches, not because its logically part of
351 * backlog.
352 */
353 struct {
354 atomic_t rmem_alloc;
355 int len;
356 struct sk_buff *head;
357 struct sk_buff *tail;
358 } sk_backlog;
359#define sk_rmem_alloc sk_backlog.rmem_alloc
360 int sk_forward_alloc;
Eric Dumazet2c8c56e2014-11-11 05:54:28 -0800361
Tom Herbertb73c3d02014-07-01 21:32:17 -0700362 __u32 sk_txhash;
Cong Wange0d10952013-08-01 11:10:25 +0800363#ifdef CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir06021292013-06-10 11:39:50 +0300364 unsigned int sk_napi_id;
Eliezer Tamirdafcc432013-06-14 16:33:57 +0300365 unsigned int sk_ll_usec;
Eliezer Tamir06021292013-06-10 11:39:50 +0300366#endif
Eric Dumazetb178bb32010-11-16 05:56:04 +0000367 atomic_t sk_drops;
368 int sk_rcvbuf;
369
370 struct sk_filter __rcu *sk_filter;
Eric Dumazetceb5d582015-11-29 20:03:11 -0800371 union {
372 struct socket_wq __rcu *sk_wq;
373 struct socket_wq *sk_wq_raw;
374 };
Eric Dumazetb178bb32010-11-16 05:56:04 +0000375#ifdef CONFIG_XFRM
Eric Dumazetd188ba82015-12-08 07:22:02 -0800376 struct xfrm_policy __rcu *sk_policy[2];
Eric Dumazetb178bb32010-11-16 05:56:04 +0000377#endif
Eric Dumazetdeaa5852012-06-24 20:22:49 +0000378 struct dst_entry *sk_rx_dst;
Cong Wang0e36cbb2013-01-22 21:09:51 +0000379 struct dst_entry __rcu *sk_dst_cache;
Eric Dumazet6bd4f352015-12-02 21:53:57 -0800380 /* Note: 32bit hole on 64bit arches */
Eric Dumazetb178bb32010-11-16 05:56:04 +0000381 atomic_t sk_wmem_alloc;
382 atomic_t sk_omem_alloc;
383 int sk_sndbuf;
384 struct sk_buff_head sk_write_queue;
Vegard Nossuma98b65a2009-02-26 14:46:57 +0100385 kmemcheck_bitfield_begin(flags);
Eric Dumazet5fdb9972009-10-08 22:50:25 +0000386 unsigned int sk_shutdown : 2,
Tom Herbert28448b82014-05-23 08:47:19 -0700387 sk_no_check_tx : 1,
388 sk_no_check_rx : 1,
Eric Dumazet5fdb9972009-10-08 22:50:25 +0000389 sk_userlocks : 4,
390 sk_protocol : 8,
391 sk_type : 16;
Hannes Frederic Sowa7bbadd22015-12-14 23:30:43 +0100392#define SK_PROTOCOL_MAX U8_MAX
Vegard Nossuma98b65a2009-02-26 14:46:57 +0100393 kmemcheck_bitfield_end(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394 int sk_wmem_queued;
Al Viro7d877f32005-10-21 03:20:43 -0400395 gfp_t sk_allocation;
Eric Dumazet95bd09e2013-08-27 05:46:32 -0700396 u32 sk_pacing_rate; /* bytes per second */
Eric Dumazet62748f32013-09-24 08:20:52 -0700397 u32 sk_max_pacing_rate;
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000398 netdev_features_t sk_route_caps;
399 netdev_features_t sk_route_nocaps;
Herbert Xubcd76112006-06-30 13:36:35 -0700400 int sk_gso_type;
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -0700401 unsigned int sk_gso_max_size;
Ben Hutchings14853482012-07-30 16:11:42 +0000402 u16 sk_gso_max_segs;
David S. Miller9932cf92006-03-24 15:12:37 -0800403 int sk_rcvlowat;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404 unsigned long sk_lingertime;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 struct sk_buff_head sk_error_queue;
Arnaldo Carvalho de Melo476e19c2005-05-05 13:35:15 -0700406 struct proto *sk_prot_creator;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407 rwlock_t sk_callback_lock;
408 int sk_err,
409 sk_err_soft;
Eric Dumazetbecb74f2015-03-19 19:04:21 -0700410 u32 sk_ack_backlog;
411 u32 sk_max_ack_backlog;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412 __u32 sk_priority;
Tejun Heo297dbde2015-12-07 17:38:51 -0500413 __u32 sk_mark;
Eric W. Biederman109f6e32010-06-13 03:30:14 +0000414 struct pid *sk_peer_pid;
415 const struct cred *sk_peer_cred;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416 long sk_rcvtimeo;
417 long sk_sndtimeo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418 struct timer_list sk_timer;
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700419 ktime_t sk_stamp;
Willem de Bruijnb9f40e22014-08-04 22:11:46 -0400420 u16 sk_tsflags;
Willem de Bruijn09c2d252014-08-04 22:11:47 -0400421 u32 sk_tskey;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422 struct socket *sk_socket;
423 void *sk_user_data;
Eric Dumazet5640f762012-09-23 23:04:42 +0000424 struct page_frag sk_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425 struct sk_buff *sk_send_head;
Pavel Emelyanovef64a542012-02-21 07:31:34 +0000426 __s32 sk_peek_off;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427 int sk_write_pending;
Alexey Dobriyand5f64232008-11-04 14:45:58 -0800428#ifdef CONFIG_SECURITY
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429 void *sk_security;
Alexey Dobriyand5f64232008-11-04 14:45:58 -0800430#endif
Tejun Heo2a56a1f2015-12-07 17:38:52 -0500431 struct sock_cgroup_data sk_cgrp_data;
Johannes Weinerbaac50b2016-01-14 15:21:17 -0800432 struct mem_cgroup *sk_memcg;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433 void (*sk_state_change)(struct sock *sk);
David S. Miller676d2362014-04-11 16:15:36 -0400434 void (*sk_data_ready)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435 void (*sk_write_space)(struct sock *sk);
436 void (*sk_error_report)(struct sock *sk);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000437 int (*sk_backlog_rcv)(struct sock *sk,
438 struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439 void (*sk_destruct)(struct sock *sk);
Craig Gallekef456142016-01-04 17:41:45 -0500440 struct sock_reuseport __rcu *sk_reuseport_cb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441};
442
Pravin B Shelar559835ea2013-09-24 10:25:40 -0700443#define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
444
445#define rcu_dereference_sk_user_data(sk) rcu_dereference(__sk_user_data((sk)))
446#define rcu_assign_sk_user_data(sk, ptr) rcu_assign_pointer(__sk_user_data((sk)), ptr)
447
Pavel Emelyanov4a17fd52012-04-19 03:39:36 +0000448/*
449 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
450 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
451 * on a socket means that the socket will reuse everybody else's port
452 * without looking at the other's sk_reuse value.
453 */
454
455#define SK_NO_REUSE 0
456#define SK_CAN_REUSE 1
457#define SK_FORCE_REUSE 2
458
Pavel Emelyanovef64a542012-02-21 07:31:34 +0000459static inline int sk_peek_offset(struct sock *sk, int flags)
460{
461 if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0))
462 return sk->sk_peek_off;
463 else
464 return 0;
465}
466
467static inline void sk_peek_offset_bwd(struct sock *sk, int val)
468{
469 if (sk->sk_peek_off >= 0) {
470 if (sk->sk_peek_off >= val)
471 sk->sk_peek_off -= val;
472 else
473 sk->sk_peek_off = 0;
474 }
475}
476
477static inline void sk_peek_offset_fwd(struct sock *sk, int val)
478{
479 if (sk->sk_peek_off >= 0)
480 sk->sk_peek_off += val;
481}
482
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483/*
484 * Hashed lists helper routines
485 */
Li Zefanc4146642010-02-08 23:18:45 +0000486static inline struct sock *sk_entry(const struct hlist_node *node)
487{
488 return hlist_entry(node, struct sock, sk_node);
489}
490
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700491static inline struct sock *__sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492{
493 return hlist_entry(head->first, struct sock, sk_node);
494}
495
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700496static inline struct sock *sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497{
498 return hlist_empty(head) ? NULL : __sk_head(head);
499}
500
Eric Dumazet88ab1932008-11-16 19:39:21 -0800501static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
502{
503 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
504}
505
506static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
507{
508 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
509}
510
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700511static inline struct sock *sk_next(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512{
513 return sk->sk_node.next ?
514 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
515}
516
Eric Dumazet88ab1932008-11-16 19:39:21 -0800517static inline struct sock *sk_nulls_next(const struct sock *sk)
518{
519 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
520 hlist_nulls_entry(sk->sk_nulls_node.next,
521 struct sock, sk_nulls_node) :
522 NULL;
523}
524
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000525static inline bool sk_unhashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526{
527 return hlist_unhashed(&sk->sk_node);
528}
529
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000530static inline bool sk_hashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531{
Akinobu Mitada753be2006-04-28 15:21:23 -0700532 return !sk_unhashed(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533}
534
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000535static inline void sk_node_init(struct hlist_node *node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536{
537 node->pprev = NULL;
538}
539
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000540static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800541{
542 node->pprev = NULL;
543}
544
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000545static inline void __sk_del_node(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546{
547 __hlist_del(&sk->sk_node);
548}
549
stephen hemminger808f5112010-02-22 07:57:18 +0000550/* NB: equivalent to hlist_del_init_rcu */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000551static inline bool __sk_del_node_init(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552{
553 if (sk_hashed(sk)) {
554 __sk_del_node(sk);
555 sk_node_init(&sk->sk_node);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000556 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557 }
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000558 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559}
560
561/* Grab socket reference count. This operation is valid only
562 when sk is ALREADY grabbed f.e. it is found in hash table
563 or a list and the lookup is made under lock preventing hash table
564 modifications.
565 */
566
567static inline void sock_hold(struct sock *sk)
568{
569 atomic_inc(&sk->sk_refcnt);
570}
571
572/* Ungrab socket in the context, which assumes that socket refcnt
573 cannot hit zero, f.e. it is true in context of any socketcall.
574 */
575static inline void __sock_put(struct sock *sk)
576{
577 atomic_dec(&sk->sk_refcnt);
578}
579
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000580static inline bool sk_del_node_init(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581{
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000582 bool rc = __sk_del_node_init(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583
584 if (rc) {
585 /* paranoid for a while -acme */
586 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
587 __sock_put(sk);
588 }
589 return rc;
590}
stephen hemminger808f5112010-02-22 07:57:18 +0000591#define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000593static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700594{
595 if (sk_hashed(sk)) {
Eric Dumazet88ab1932008-11-16 19:39:21 -0800596 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000597 return true;
Eric Dumazet271b72c2008-10-29 02:11:14 -0700598 }
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000599 return false;
Eric Dumazet271b72c2008-10-29 02:11:14 -0700600}
601
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000602static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700603{
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000604 bool rc = __sk_nulls_del_node_init_rcu(sk);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700605
606 if (rc) {
607 /* paranoid for a while -acme */
608 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
609 __sock_put(sk);
610 }
611 return rc;
612}
613
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000614static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615{
616 hlist_add_head(&sk->sk_node, list);
617}
618
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000619static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700620{
621 sock_hold(sk);
622 __sk_add_node(sk, list);
623}
624
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000625static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
stephen hemminger808f5112010-02-22 07:57:18 +0000626{
627 sock_hold(sk);
628 hlist_add_head_rcu(&sk->sk_node, list);
629}
630
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000631static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700632{
Eric Dumazet88ab1932008-11-16 19:39:21 -0800633 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700634}
635
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000636static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700637{
638 sock_hold(sk);
Eric Dumazet88ab1932008-11-16 19:39:21 -0800639 __sk_nulls_add_node_rcu(sk, list);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700640}
641
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000642static inline void __sk_del_bind_node(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643{
644 __hlist_del(&sk->sk_bind_node);
645}
646
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000647static inline void sk_add_bind_node(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648 struct hlist_head *list)
649{
650 hlist_add_head(&sk->sk_bind_node, list);
651}
652
Sasha Levinb67bfe02013-02-27 17:06:00 -0800653#define sk_for_each(__sk, list) \
654 hlist_for_each_entry(__sk, list, sk_node)
655#define sk_for_each_rcu(__sk, list) \
656 hlist_for_each_entry_rcu(__sk, list, sk_node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800657#define sk_nulls_for_each(__sk, node, list) \
658 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
659#define sk_nulls_for_each_rcu(__sk, node, list) \
660 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
Sasha Levinb67bfe02013-02-27 17:06:00 -0800661#define sk_for_each_from(__sk) \
662 hlist_for_each_entry_from(__sk, sk_node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800663#define sk_nulls_for_each_from(__sk, node) \
664 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
665 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
Sasha Levinb67bfe02013-02-27 17:06:00 -0800666#define sk_for_each_safe(__sk, tmp, list) \
667 hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
668#define sk_for_each_bound(__sk, list) \
669 hlist_for_each_entry(__sk, list, sk_bind_node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670
David Held2dc41cf2014-07-15 23:28:32 -0400671/**
672 * sk_nulls_for_each_entry_offset - iterate over a list at a given struct offset
673 * @tpos: the type * to use as a loop cursor.
674 * @pos: the &struct hlist_node to use as a loop cursor.
675 * @head: the head for your list.
676 * @offset: offset of hlist_node within the struct.
677 *
678 */
679#define sk_nulls_for_each_entry_offset(tpos, pos, head, offset) \
680 for (pos = (head)->first; \
681 (!is_a_nulls(pos)) && \
682 ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;}); \
683 pos = pos->next)
684
Eric W. Biedermanc336d142012-05-24 17:56:43 -0600685static inline struct user_namespace *sk_user_ns(struct sock *sk)
686{
687 /* Careful only use this in a context where these parameters
688 * can not change and must all be valid, such as recvmsg from
689 * userspace.
690 */
691 return sk->sk_socket->file->f_cred->user_ns;
692}
693
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694/* Sock flags */
695enum sock_flags {
696 SOCK_DEAD,
697 SOCK_DONE,
698 SOCK_URGINLINE,
699 SOCK_KEEPOPEN,
700 SOCK_LINGER,
701 SOCK_DESTROY,
702 SOCK_BROADCAST,
703 SOCK_TIMESTAMP,
704 SOCK_ZAPPED,
705 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
706 SOCK_DBG, /* %SO_DEBUG setting */
707 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700708 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
710 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
Mel Gorman7cb02402012-07-31 16:44:16 -0700711 SOCK_MEMALLOC, /* VM depends on this socket for swapping */
Patrick Ohly20d49472009-02-12 05:03:38 +0000712 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
Eric Dumazetbcdce712009-10-06 17:28:29 -0700713 SOCK_FASYNC, /* fasync() active */
Neil Horman3b885782009-10-12 13:26:31 -0700714 SOCK_RXQ_OVFL,
Shirley Ma1cdebb42011-07-06 12:17:30 +0000715 SOCK_ZEROCOPY, /* buffers from userspace */
Johannes Berg6e3e9392011-11-09 10:15:42 +0100716 SOCK_WIFI_STATUS, /* push wifi status to userspace */
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000717 SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
718 * Will use last 4 bytes of packet sent from
719 * user-space instead.
720 */
Vincent Bernatd59577b2013-01-16 22:55:49 +0100721 SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */
Keller, Jacob E7d4c04f2013-03-28 11:19:25 +0000722 SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723};
724
Marcelo Ricardo Leitner01ce63c2015-12-04 15:14:04 -0200725#define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
726
Ralf Baechle53b924b2005-08-23 10:11:30 -0700727static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
728{
729 nsk->sk_flags = osk->sk_flags;
730}
731
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
733{
734 __set_bit(flag, &sk->sk_flags);
735}
736
737static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
738{
739 __clear_bit(flag, &sk->sk_flags);
740}
741
Eric Dumazet1b23a5d2012-05-16 05:57:07 +0000742static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743{
744 return test_bit(flag, &sk->sk_flags);
745}
746
Mel Gormanc93bdd02012-07-31 16:44:19 -0700747#ifdef CONFIG_NET
748extern struct static_key memalloc_socks;
749static inline int sk_memalloc_socks(void)
750{
751 return static_key_false(&memalloc_socks);
752}
753#else
754
755static inline int sk_memalloc_socks(void)
756{
757 return 0;
758}
759
760#endif
761
Eric Dumazet7450aaf2015-11-30 08:57:28 -0800762static inline gfp_t sk_gfp_mask(const struct sock *sk, gfp_t gfp_mask)
Mel Gorman99a1dec2012-07-31 16:44:14 -0700763{
Eric Dumazet7450aaf2015-11-30 08:57:28 -0800764 return gfp_mask | (sk->sk_allocation & __GFP_MEMALLOC);
Mel Gorman99a1dec2012-07-31 16:44:14 -0700765}
766
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767static inline void sk_acceptq_removed(struct sock *sk)
768{
769 sk->sk_ack_backlog--;
770}
771
772static inline void sk_acceptq_added(struct sock *sk)
773{
774 sk->sk_ack_backlog++;
775}
776
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000777static inline bool sk_acceptq_is_full(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778{
David S. Miller64a14652007-03-06 11:21:05 -0800779 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780}
781
782/*
783 * Compute minimal free write space needed to queue new packets.
784 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000785static inline int sk_stream_min_wspace(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786{
Eric Dumazet8df09ea2007-12-21 03:07:41 -0800787 return sk->sk_wmem_queued >> 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788}
789
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000790static inline int sk_stream_wspace(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791{
792 return sk->sk_sndbuf - sk->sk_wmem_queued;
793}
794
Joe Perches69336bd2013-09-22 10:32:26 -0700795void sk_stream_write_space(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796
Zhu Yi8eae9392010-03-04 18:01:40 +0000797/* OOB backlog add */
Zhu Yia3a858f2010-03-04 18:01:47 +0000798static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800799{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000800 /* dont let skb dst not refcounted, we are going to leave rcu lock */
Eric Dumazet5037e9e2015-12-14 14:08:53 -0800801 skb_dst_force_safe(skb);
Eric Dumazet7fee2262010-05-11 23:19:48 +0000802
803 if (!sk->sk_backlog.tail)
804 sk->sk_backlog.head = skb;
805 else
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800806 sk->sk_backlog.tail->next = skb;
Eric Dumazet7fee2262010-05-11 23:19:48 +0000807
808 sk->sk_backlog.tail = skb;
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800809 skb->next = NULL;
810}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811
Eric Dumazetc3774112010-04-27 15:13:20 -0700812/*
813 * Take into account size of receive queue and backlog queue
Eric Dumazet0fd7bac2011-12-21 07:11:44 +0000814 * Do not take into account this skb truesize,
815 * to allow even a single big packet to come.
Eric Dumazetc3774112010-04-27 15:13:20 -0700816 */
Sorin Dumitru274f4822014-07-22 21:16:51 +0300817static inline bool sk_rcvqueues_full(const struct sock *sk, unsigned int limit)
Eric Dumazetc3774112010-04-27 15:13:20 -0700818{
819 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
820
Eric Dumazetf545a382012-04-22 23:34:26 +0000821 return qsize > limit;
Eric Dumazetc3774112010-04-27 15:13:20 -0700822}
823
Zhu Yi8eae9392010-03-04 18:01:40 +0000824/* The per-socket spinlock must be held here. */
Eric Dumazetf545a382012-04-22 23:34:26 +0000825static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
826 unsigned int limit)
Zhu Yi8eae9392010-03-04 18:01:40 +0000827{
Sorin Dumitru274f4822014-07-22 21:16:51 +0300828 if (sk_rcvqueues_full(sk, limit))
Zhu Yi8eae9392010-03-04 18:01:40 +0000829 return -ENOBUFS;
830
Eric Dumazetc7c49b82015-09-29 18:52:25 -0700831 /*
832 * If the skb was allocated from pfmemalloc reserves, only
833 * allow SOCK_MEMALLOC sockets to use it as this socket is
834 * helping free memory
835 */
836 if (skb_pfmemalloc(skb) && !sock_flag(sk, SOCK_MEMALLOC))
837 return -ENOMEM;
838
Zhu Yia3a858f2010-03-04 18:01:47 +0000839 __sk_add_backlog(sk, skb);
Zhu Yi8eae9392010-03-04 18:01:40 +0000840 sk->sk_backlog.len += skb->truesize;
841 return 0;
842}
843
Joe Perches69336bd2013-09-22 10:32:26 -0700844int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
Mel Gormanb4b9e352012-07-31 16:44:26 -0700845
Peter Zijlstrac57943a2008-10-07 14:18:42 -0700846static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
847{
Mel Gormanb4b9e352012-07-31 16:44:26 -0700848 if (sk_memalloc_socks() && skb_pfmemalloc(skb))
849 return __sk_backlog_rcv(sk, skb);
850
Peter Zijlstrac57943a2008-10-07 14:18:42 -0700851 return sk->sk_backlog_rcv(sk, skb);
852}
853
Eric Dumazet2c8c56e2014-11-11 05:54:28 -0800854static inline void sk_incoming_cpu_update(struct sock *sk)
855{
856 sk->sk_incoming_cpu = raw_smp_processor_id();
857}
858
Tom Herbertfe477552013-12-22 18:54:31 +0800859static inline void sock_rps_record_flow_hash(__u32 hash)
David S. Millerc58dc012010-04-27 15:05:31 -0700860{
861#ifdef CONFIG_RPS
862 struct rps_sock_flow_table *sock_flow_table;
863
864 rcu_read_lock();
865 sock_flow_table = rcu_dereference(rps_sock_flow_table);
Tom Herbertfe477552013-12-22 18:54:31 +0800866 rps_record_sock_flow(sock_flow_table, hash);
David S. Millerc58dc012010-04-27 15:05:31 -0700867 rcu_read_unlock();
868#endif
869}
870
Tom Herbertfe477552013-12-22 18:54:31 +0800871static inline void sock_rps_record_flow(const struct sock *sk)
872{
Zhi Yong Wuc9d8ca02014-01-01 04:31:01 +0800873#ifdef CONFIG_RPS
Tom Herbertfe477552013-12-22 18:54:31 +0800874 sock_rps_record_flow_hash(sk->sk_rxhash);
Zhi Yong Wuc9d8ca02014-01-01 04:31:01 +0800875#endif
Tom Herbertfe477552013-12-22 18:54:31 +0800876}
877
Tom Herbertbdeab992011-08-14 19:45:55 +0000878static inline void sock_rps_save_rxhash(struct sock *sk,
879 const struct sk_buff *skb)
David S. Millerc58dc012010-04-27 15:05:31 -0700880{
881#ifdef CONFIG_RPS
Eric Dumazet567e4b72015-02-06 12:59:01 -0800882 if (unlikely(sk->sk_rxhash != skb->hash))
Tom Herbert61b905d2014-03-24 15:34:47 -0700883 sk->sk_rxhash = skb->hash;
David S. Millerc58dc012010-04-27 15:05:31 -0700884#endif
885}
886
Tom Herbertbdeab992011-08-14 19:45:55 +0000887static inline void sock_rps_reset_rxhash(struct sock *sk)
888{
889#ifdef CONFIG_RPS
Tom Herbertbdeab992011-08-14 19:45:55 +0000890 sk->sk_rxhash = 0;
891#endif
892}
893
Stephen Hemmingercfcabdc2007-10-09 01:59:42 -0700894#define sk_wait_event(__sk, __timeo, __condition) \
895 ({ int __rc; \
896 release_sock(__sk); \
897 __rc = __condition; \
898 if (!__rc) { \
899 *(__timeo) = schedule_timeout(*(__timeo)); \
900 } \
Peter Zijlstra26cabd32014-09-24 10:18:54 +0200901 sched_annotate_sleep(); \
Stephen Hemmingercfcabdc2007-10-09 01:59:42 -0700902 lock_sock(__sk); \
903 __rc = __condition; \
904 __rc; \
905 })
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906
Joe Perches69336bd2013-09-22 10:32:26 -0700907int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
908int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
909void sk_stream_wait_close(struct sock *sk, long timeo_p);
910int sk_stream_error(struct sock *sk, int flags, int err);
911void sk_stream_kill_queues(struct sock *sk);
912void sk_set_memalloc(struct sock *sk);
913void sk_clear_memalloc(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914
Sabrina Dubrocadfbafc92015-07-24 18:19:25 +0200915int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700917struct request_sock_ops;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800918struct timewait_sock_ops;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -0800919struct inet_hashinfo;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -0700920struct raw_hashinfo;
Paul Gortmakerde477252011-05-26 13:46:22 -0400921struct module;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700922
Eric Dumazetf77d6022013-05-09 10:28:16 +0000923/*
924 * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
925 * un-modified. Special care is taken when initializing object to zero.
926 */
927static inline void sk_prot_clear_nulls(struct sock *sk, int size)
928{
929 if (offsetof(struct sock, sk_node.next) != 0)
930 memset(sk, 0, offsetof(struct sock, sk_node.next));
931 memset(&sk->sk_node.pprev, 0,
932 size - offsetof(struct sock, sk_node.pprev));
933}
934
Linus Torvalds1da177e2005-04-16 15:20:36 -0700935/* Networking protocol blocks we attach to sockets.
936 * socket layer -> transport layer interface
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937 */
938struct proto {
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000939 void (*close)(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940 long timeout);
941 int (*connect)(struct sock *sk,
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000942 struct sockaddr *uaddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700943 int addr_len);
944 int (*disconnect)(struct sock *sk, int flags);
945
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000946 struct sock * (*accept)(struct sock *sk, int flags, int *err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700947
948 int (*ioctl)(struct sock *sk, int cmd,
949 unsigned long arg);
950 int (*init)(struct sock *sk);
Brian Haley7d06b2e2008-06-14 17:04:49 -0700951 void (*destroy)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700952 void (*shutdown)(struct sock *sk, int how);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000953 int (*setsockopt)(struct sock *sk, int level,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700954 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700955 unsigned int optlen);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000956 int (*getsockopt)(struct sock *sk, int level,
957 int optname, char __user *optval,
958 int __user *option);
Alexey Dobriyanaf01d532008-08-28 02:53:51 -0700959#ifdef CONFIG_COMPAT
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800960 int (*compat_setsockopt)(struct sock *sk,
961 int level,
962 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700963 unsigned int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800964 int (*compat_getsockopt)(struct sock *sk,
965 int level,
966 int optname, char __user *optval,
967 int __user *option);
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000968 int (*compat_ioctl)(struct sock *sk,
969 unsigned int cmd, unsigned long arg);
Alexey Dobriyanaf01d532008-08-28 02:53:51 -0700970#endif
Ying Xue1b784142015-03-02 15:37:48 +0800971 int (*sendmsg)(struct sock *sk, struct msghdr *msg,
972 size_t len);
973 int (*recvmsg)(struct sock *sk, struct msghdr *msg,
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000974 size_t len, int noblock, int flags,
975 int *addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700976 int (*sendpage)(struct sock *sk, struct page *page,
977 int offset, size_t size, int flags);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000978 int (*bind)(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700979 struct sockaddr *uaddr, int addr_len);
980
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000981 int (*backlog_rcv) (struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982 struct sk_buff *skb);
983
Eric Dumazet46d3cea2012-07-11 05:50:31 +0000984 void (*release_cb)(struct sock *sk);
985
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986 /* Keeping track of sk's, looking them up, and port selection methods. */
987 void (*hash)(struct sock *sk);
988 void (*unhash)(struct sock *sk);
Eric Dumazet719f8352010-09-08 05:08:44 +0000989 void (*rehash)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990 int (*get_port)(struct sock *sk, unsigned short snum);
Octavian Purdilafcbdf092010-12-16 14:26:56 -0800991 void (*clear_sk)(struct sock *sk, int size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992
Eric Dumazet286ab3d2007-11-05 23:38:39 -0800993 /* Keeping track of sockets in use */
Eric Dumazet65f76512008-01-03 20:46:48 -0800994#ifdef CONFIG_PROC_FS
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -0700995 unsigned int inuse_idx;
Eric Dumazet65f76512008-01-03 20:46:48 -0800996#endif
Arnaldo Carvalho de Meloebb53d72007-11-21 22:08:50 +0800997
Eric Dumazetc9bee3b72013-07-22 20:27:07 -0700998 bool (*stream_memory_free)(const struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999 /* Memory pressure */
Pavel Emelyanov5c52ba12008-07-16 20:28:10 -07001000 void (*enter_memory_pressure)(struct sock *sk);
Eric Dumazet8d987e52010-11-09 23:24:26 +00001001 atomic_long_t *memory_allocated; /* Current allocated memory. */
Eric Dumazet17483762008-11-25 21:16:35 -08001002 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003 /*
1004 * Pressure flag: try to collapse.
1005 * Technical note: it is used by multiple contexts non atomically.
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001006 * All the __sk_mem_schedule() is of this nature: accounting
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 * is strict, actions are advisory and have some latency.
1008 */
1009 int *memory_pressure;
Eric Dumazet8d987e52010-11-09 23:24:26 +00001010 long *sysctl_mem;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011 int *sysctl_wmem;
1012 int *sysctl_rmem;
1013 int max_header;
Changli Gao7ba42912010-07-10 20:41:55 +00001014 bool no_autobind;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015
Eric Dumazet271b72c2008-10-29 02:11:14 -07001016 struct kmem_cache *slab;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017 unsigned int obj_size;
Eric Dumazet271b72c2008-10-29 02:11:14 -07001018 int slab_flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019
Eric Dumazetdd24c002008-11-25 21:17:14 -08001020 struct percpu_counter *orphan_count;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -07001021
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -07001022 struct request_sock_ops *rsk_prot;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -08001023 struct timewait_sock_ops *twsk_prot;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07001024
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -07001025 union {
1026 struct inet_hashinfo *hashinfo;
Eric Dumazet645ca702008-10-29 01:41:45 -07001027 struct udp_table *udp_table;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -07001028 struct raw_hashinfo *raw_hash;
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -07001029 } h;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -08001030
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031 struct module *owner;
1032
1033 char name[32];
1034
1035 struct list_head node;
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -07001036#ifdef SOCK_REFCNT_DEBUG
1037 atomic_t socks;
1038#endif
Lorenzo Colitti64be0ae2015-12-16 12:30:03 +09001039 int (*diag_destroy)(struct sock *sk, int err);
Glauber Costae1aab162011-12-11 21:47:03 +00001040};
1041
Joe Perches69336bd2013-09-22 10:32:26 -07001042int proto_register(struct proto *prot, int alloc_slab);
1043void proto_unregister(struct proto *prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -07001045#ifdef SOCK_REFCNT_DEBUG
1046static inline void sk_refcnt_debug_inc(struct sock *sk)
1047{
1048 atomic_inc(&sk->sk_prot->socks);
1049}
1050
1051static inline void sk_refcnt_debug_dec(struct sock *sk)
1052{
1053 atomic_dec(&sk->sk_prot->socks);
1054 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
1055 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
1056}
1057
Ying Xuedec34fb2013-02-15 22:28:25 +00001058static inline void sk_refcnt_debug_release(const struct sock *sk)
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -07001059{
1060 if (atomic_read(&sk->sk_refcnt) != 1)
1061 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
1062 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
1063}
1064#else /* SOCK_REFCNT_DEBUG */
1065#define sk_refcnt_debug_inc(sk) do { } while (0)
1066#define sk_refcnt_debug_dec(sk) do { } while (0)
1067#define sk_refcnt_debug_release(sk) do { } while (0)
1068#endif /* SOCK_REFCNT_DEBUG */
1069
Eric Dumazetc9bee3b72013-07-22 20:27:07 -07001070static inline bool sk_stream_memory_free(const struct sock *sk)
1071{
1072 if (sk->sk_wmem_queued >= sk->sk_sndbuf)
1073 return false;
1074
1075 return sk->sk_prot->stream_memory_free ?
1076 sk->sk_prot->stream_memory_free(sk) : true;
1077}
1078
Eric Dumazet64dc6132013-07-22 20:26:31 -07001079static inline bool sk_stream_is_writeable(const struct sock *sk)
1080{
Eric Dumazetc9bee3b72013-07-22 20:27:07 -07001081 return sk_stream_wspace(sk) >= sk_stream_min_wspace(sk) &&
1082 sk_stream_memory_free(sk);
Eric Dumazet64dc6132013-07-22 20:26:31 -07001083}
Glauber Costae1aab162011-12-11 21:47:03 +00001084
Eric Dumazetc9bee3b72013-07-22 20:27:07 -07001085
Glauber Costa180d8cd2011-12-11 21:47:02 +00001086static inline bool sk_has_memory_pressure(const struct sock *sk)
1087{
1088 return sk->sk_prot->memory_pressure != NULL;
1089}
1090
1091static inline bool sk_under_memory_pressure(const struct sock *sk)
1092{
1093 if (!sk->sk_prot->memory_pressure)
1094 return false;
Glauber Costae1aab162011-12-11 21:47:03 +00001095
Johannes Weinerbaac50b2016-01-14 15:21:17 -08001096 if (mem_cgroup_sockets_enabled && sk->sk_memcg &&
1097 mem_cgroup_under_socket_pressure(sk->sk_memcg))
Johannes Weinere8056052016-01-14 15:21:14 -08001098 return true;
Glauber Costae1aab162011-12-11 21:47:03 +00001099
Christoph Paasch35b87f62013-10-23 12:49:21 -07001100 return !!*sk->sk_prot->memory_pressure;
Glauber Costa180d8cd2011-12-11 21:47:02 +00001101}
1102
1103static inline void sk_leave_memory_pressure(struct sock *sk)
1104{
1105 int *memory_pressure = sk->sk_prot->memory_pressure;
1106
Glauber Costae1aab162011-12-11 21:47:03 +00001107 if (!memory_pressure)
1108 return;
1109
1110 if (*memory_pressure)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001111 *memory_pressure = 0;
1112}
1113
1114static inline void sk_enter_memory_pressure(struct sock *sk)
1115{
Glauber Costae1aab162011-12-11 21:47:03 +00001116 if (!sk->sk_prot->enter_memory_pressure)
1117 return;
1118
Glauber Costae1aab162011-12-11 21:47:03 +00001119 sk->sk_prot->enter_memory_pressure(sk);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001120}
1121
1122static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1123{
Johannes Weinere8056052016-01-14 15:21:14 -08001124 return sk->sk_prot->sysctl_mem[index];
Glauber Costae1aab162011-12-11 21:47:03 +00001125}
1126
Glauber Costa180d8cd2011-12-11 21:47:02 +00001127static inline long
1128sk_memory_allocated(const struct sock *sk)
1129{
Johannes Weinere8056052016-01-14 15:21:14 -08001130 return atomic_long_read(sk->sk_prot->memory_allocated);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001131}
1132
1133static inline long
Johannes Weinere8056052016-01-14 15:21:14 -08001134sk_memory_allocated_add(struct sock *sk, int amt)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001135{
Johannes Weinere8056052016-01-14 15:21:14 -08001136 return atomic_long_add_return(amt, sk->sk_prot->memory_allocated);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001137}
1138
1139static inline void
Glauber Costa0e90b312012-01-20 04:57:16 +00001140sk_memory_allocated_sub(struct sock *sk, int amt)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001141{
Johannes Weinere8056052016-01-14 15:21:14 -08001142 atomic_long_sub(amt, sk->sk_prot->memory_allocated);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001143}
1144
1145static inline void sk_sockets_allocated_dec(struct sock *sk)
1146{
Johannes Weineraf95d7d2016-01-14 15:21:08 -08001147 percpu_counter_dec(sk->sk_prot->sockets_allocated);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001148}
1149
1150static inline void sk_sockets_allocated_inc(struct sock *sk)
1151{
Johannes Weineraf95d7d2016-01-14 15:21:08 -08001152 percpu_counter_inc(sk->sk_prot->sockets_allocated);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001153}
1154
1155static inline int
1156sk_sockets_allocated_read_positive(struct sock *sk)
1157{
Johannes Weineraf95d7d2016-01-14 15:21:08 -08001158 return percpu_counter_read_positive(sk->sk_prot->sockets_allocated);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001159}
1160
1161static inline int
1162proto_sockets_allocated_sum_positive(struct proto *prot)
1163{
1164 return percpu_counter_sum_positive(prot->sockets_allocated);
1165}
1166
1167static inline long
1168proto_memory_allocated(struct proto *prot)
1169{
1170 return atomic_long_read(prot->memory_allocated);
1171}
1172
1173static inline bool
1174proto_memory_pressure(struct proto *prot)
1175{
1176 if (!prot->memory_pressure)
1177 return false;
1178 return !!*prot->memory_pressure;
1179}
1180
Eric Dumazet65f76512008-01-03 20:46:48 -08001181
1182#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183/* Called with local bh disabled */
Joe Perches69336bd2013-09-22 10:32:26 -07001184void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1185int sock_prot_inuse_get(struct net *net, struct proto *proto);
Eric Dumazet65f76512008-01-03 20:46:48 -08001186#else
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001187static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -07001188 int inc)
Eric Dumazet65f76512008-01-03 20:46:48 -08001189{
1190}
Eric Dumazet65f76512008-01-03 20:46:48 -08001191#endif
1192
Linus Torvalds1da177e2005-04-16 15:20:36 -07001193
Arnaldo Carvalho de Melo614c6cb2005-08-09 19:47:37 -07001194/* With per-bucket locks this operation is not-atomic, so that
1195 * this version is not worse.
1196 */
1197static inline void __sk_prot_rehash(struct sock *sk)
1198{
1199 sk->sk_prot->unhash(sk);
1200 sk->sk_prot->hash(sk);
1201}
1202
Octavian Purdilafcbdf092010-12-16 14:26:56 -08001203void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
1204
Linus Torvalds1da177e2005-04-16 15:20:36 -07001205/* About 10 seconds */
1206#define SOCK_DESTROY_TIME (10*HZ)
1207
1208/* Sockets 0-1023 can't be bound to unless you are superuser */
1209#define PROT_SOCK 1024
1210
1211#define SHUTDOWN_MASK 3
1212#define RCV_SHUTDOWN 1
1213#define SEND_SHUTDOWN 2
1214
1215#define SOCK_SNDBUF_LOCK 1
1216#define SOCK_RCVBUF_LOCK 2
1217#define SOCK_BINDADDR_LOCK 4
1218#define SOCK_BINDPORT_LOCK 8
1219
Linus Torvalds1da177e2005-04-16 15:20:36 -07001220struct socket_alloc {
1221 struct socket socket;
1222 struct inode vfs_inode;
1223};
1224
1225static inline struct socket *SOCKET_I(struct inode *inode)
1226{
1227 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1228}
1229
1230static inline struct inode *SOCK_INODE(struct socket *socket)
1231{
1232 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1233}
1234
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001235/*
1236 * Functions for memory accounting
1237 */
Joe Perches69336bd2013-09-22 10:32:26 -07001238int __sk_mem_schedule(struct sock *sk, int size, int kind);
Eric Dumazet1a24e042015-05-15 12:39:25 -07001239void __sk_mem_reclaim(struct sock *sk, int amount);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001240
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001241#define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1242#define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1243#define SK_MEM_SEND 0
1244#define SK_MEM_RECV 1
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001246static inline int sk_mem_pages(int amt)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001247{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001248 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249}
1250
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001251static inline bool sk_has_account(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001252{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001253 /* return true if protocol supports memory accounting */
1254 return !!sk->sk_prot->memory_allocated;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255}
1256
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001257static inline bool sk_wmem_schedule(struct sock *sk, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001259 if (!sk_has_account(sk))
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001260 return true;
Herbert Xud80d99d2005-09-01 17:48:23 -07001261 return size <= sk->sk_forward_alloc ||
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001262 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1263}
1264
Mel Gormanc76562b2012-07-31 16:44:41 -07001265static inline bool
Chuck Lever35c448a2012-09-17 14:09:11 -07001266sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size)
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001267{
1268 if (!sk_has_account(sk))
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001269 return true;
Mel Gormanc76562b2012-07-31 16:44:41 -07001270 return size<= sk->sk_forward_alloc ||
1271 __sk_mem_schedule(sk, size, SK_MEM_RECV) ||
1272 skb_pfmemalloc(skb);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001273}
1274
1275static inline void sk_mem_reclaim(struct sock *sk)
1276{
1277 if (!sk_has_account(sk))
1278 return;
1279 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
Eric Dumazet1a24e042015-05-15 12:39:25 -07001280 __sk_mem_reclaim(sk, sk->sk_forward_alloc);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001281}
1282
David S. Miller9993e7d2008-01-10 21:56:38 -08001283static inline void sk_mem_reclaim_partial(struct sock *sk)
1284{
1285 if (!sk_has_account(sk))
1286 return;
1287 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
Eric Dumazet1a24e042015-05-15 12:39:25 -07001288 __sk_mem_reclaim(sk, sk->sk_forward_alloc - 1);
David S. Miller9993e7d2008-01-10 21:56:38 -08001289}
1290
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001291static inline void sk_mem_charge(struct sock *sk, int size)
1292{
1293 if (!sk_has_account(sk))
1294 return;
1295 sk->sk_forward_alloc -= size;
1296}
1297
1298static inline void sk_mem_uncharge(struct sock *sk, int size)
1299{
1300 if (!sk_has_account(sk))
1301 return;
1302 sk->sk_forward_alloc += size;
1303}
1304
1305static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1306{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001307 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1308 sk->sk_wmem_queued -= skb->truesize;
1309 sk_mem_uncharge(sk, skb->truesize);
1310 __kfree_skb(skb);
Herbert Xud80d99d2005-09-01 17:48:23 -07001311}
1312
Linus Torvalds1da177e2005-04-16 15:20:36 -07001313/* Used by processes to "lock" a socket state, so that
1314 * interrupts and bottom half handlers won't change it
1315 * from under us. It essentially blocks any incoming
1316 * packets, so that we won't get any new data or any
1317 * packets that change the state of the socket.
1318 *
1319 * While locked, BH processing will add new packets to
1320 * the backlog queue. This queue is processed by the
1321 * owner of the socket lock right before it is released.
1322 *
1323 * Since ~2.3.5 it is also exclusive sleep lock serializing
1324 * accesses from user process context.
1325 */
John Heffnerd2e91172007-09-12 10:44:19 +02001326#define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001327
Eric Dumazetc3f9b012014-03-10 09:50:11 -07001328static inline void sock_release_ownership(struct sock *sk)
1329{
1330 sk->sk_lock.owned = 0;
1331}
1332
Peter Zijlstraed075362006-12-06 20:35:24 -08001333/*
1334 * Macro so as to not evaluate some arguments when
1335 * lockdep is not enabled.
1336 *
1337 * Mark both the sk_lock and the sk_lock.slock as a
1338 * per-address-family lock class.
1339 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001340#define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
Peter Zijlstraed075362006-12-06 20:35:24 -08001341do { \
Ingo Molnare8f6fbf2008-11-12 01:38:36 +00001342 sk->sk_lock.owned = 0; \
Peter Zijlstraed075362006-12-06 20:35:24 -08001343 init_waitqueue_head(&sk->sk_lock.wq); \
1344 spin_lock_init(&(sk)->sk_lock.slock); \
1345 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1346 sizeof((sk)->sk_lock)); \
1347 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001348 (skey), (sname)); \
Peter Zijlstraed075362006-12-06 20:35:24 -08001349 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1350} while (0)
1351
Joe Perches69336bd2013-09-22 10:32:26 -07001352void lock_sock_nested(struct sock *sk, int subclass);
Peter Zijlstrafcc70d52006-11-08 22:44:35 -08001353
1354static inline void lock_sock(struct sock *sk)
1355{
1356 lock_sock_nested(sk, 0);
1357}
1358
Joe Perches69336bd2013-09-22 10:32:26 -07001359void release_sock(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360
1361/* BH context may only use the following locking interface. */
1362#define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
Ingo Molnarc6366182006-07-03 00:25:13 -07001363#define bh_lock_sock_nested(__sk) \
1364 spin_lock_nested(&((__sk)->sk_lock.slock), \
1365 SINGLE_DEPTH_NESTING)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001366#define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1367
Joe Perches69336bd2013-09-22 10:32:26 -07001368bool lock_sock_fast(struct sock *sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001369/**
1370 * unlock_sock_fast - complement of lock_sock_fast
1371 * @sk: socket
1372 * @slow: slow mode
1373 *
1374 * fast unlock socket for user context.
1375 * If slow mode is on, we call regular release_sock()
1376 */
1377static inline void unlock_sock_fast(struct sock *sk, bool slow)
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001378{
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001379 if (slow)
1380 release_sock(sk);
1381 else
1382 spin_unlock_bh(&sk->sk_lock.slock);
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001383}
1384
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001385
Joe Perches69336bd2013-09-22 10:32:26 -07001386struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
Eric W. Biederman11aa9c22015-05-08 21:09:13 -05001387 struct proto *prot, int kern);
Joe Perches69336bd2013-09-22 10:32:26 -07001388void sk_free(struct sock *sk);
Craig Gallekeb4cb002015-06-15 11:26:18 -04001389void sk_destruct(struct sock *sk);
Joe Perches69336bd2013-09-22 10:32:26 -07001390struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001391
Joe Perches69336bd2013-09-22 10:32:26 -07001392struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
1393 gfp_t priority);
Joe Perches69336bd2013-09-22 10:32:26 -07001394void sock_wfree(struct sk_buff *skb);
1395void skb_orphan_partial(struct sk_buff *skb);
1396void sock_rfree(struct sk_buff *skb);
Alexander Duyck62bccb82014-09-04 13:31:35 -04001397void sock_efree(struct sk_buff *skb);
Alexander Duyck82eabd92014-09-04 13:32:11 -04001398#ifdef CONFIG_INET
Joe Perches69336bd2013-09-22 10:32:26 -07001399void sock_edemux(struct sk_buff *skb);
Alexander Duyck82eabd92014-09-04 13:32:11 -04001400#else
1401#define sock_edemux(skb) sock_efree(skb)
1402#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001403
Joe Perches69336bd2013-09-22 10:32:26 -07001404int sock_setsockopt(struct socket *sock, int level, int op,
1405 char __user *optval, unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001406
Joe Perches69336bd2013-09-22 10:32:26 -07001407int sock_getsockopt(struct socket *sock, int level, int op,
1408 char __user *optval, int __user *optlen);
1409struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1410 int noblock, int *errcode);
1411struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1412 unsigned long data_len, int noblock,
1413 int *errcode, int max_page_order);
1414void *sock_kmalloc(struct sock *sk, int size, gfp_t priority);
1415void sock_kfree_s(struct sock *sk, void *mem, int size);
Daniel Borkmann79e88652014-11-19 17:13:11 +01001416void sock_kzfree_s(struct sock *sk, void *mem, int size);
Joe Perches69336bd2013-09-22 10:32:26 -07001417void sk_send_sigurg(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418
Edward Jeef28ea362015-10-08 14:56:48 -07001419struct sockcm_cookie {
1420 u32 mark;
1421};
1422
1423int sock_cmsg_send(struct sock *sk, struct msghdr *msg,
1424 struct sockcm_cookie *sockc);
1425
Linus Torvalds1da177e2005-04-16 15:20:36 -07001426/*
1427 * Functions to fill in entries in struct proto_ops when a protocol
1428 * does not implement a particular function.
1429 */
Joe Perches69336bd2013-09-22 10:32:26 -07001430int sock_no_bind(struct socket *, struct sockaddr *, int);
1431int sock_no_connect(struct socket *, struct sockaddr *, int, int);
1432int sock_no_socketpair(struct socket *, struct socket *);
1433int sock_no_accept(struct socket *, struct socket *, int);
1434int sock_no_getname(struct socket *, struct sockaddr *, int *, int);
1435unsigned int sock_no_poll(struct file *, struct socket *,
1436 struct poll_table_struct *);
1437int sock_no_ioctl(struct socket *, unsigned int, unsigned long);
1438int sock_no_listen(struct socket *, int);
1439int sock_no_shutdown(struct socket *, int);
1440int sock_no_getsockopt(struct socket *, int , int, char __user *, int __user *);
1441int sock_no_setsockopt(struct socket *, int, int, char __user *, unsigned int);
Ying Xue1b784142015-03-02 15:37:48 +08001442int sock_no_sendmsg(struct socket *, struct msghdr *, size_t);
1443int sock_no_recvmsg(struct socket *, struct msghdr *, size_t, int);
Joe Perches69336bd2013-09-22 10:32:26 -07001444int sock_no_mmap(struct file *file, struct socket *sock,
1445 struct vm_area_struct *vma);
1446ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset,
1447 size_t size, int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448
1449/*
1450 * Functions to fill in entries in struct proto_ops when a protocol
1451 * uses the inet style.
1452 */
Joe Perches69336bd2013-09-22 10:32:26 -07001453int sock_common_getsockopt(struct socket *sock, int level, int optname,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001454 char __user *optval, int __user *optlen);
Ying Xue1b784142015-03-02 15:37:48 +08001455int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
1456 int flags);
Joe Perches69336bd2013-09-22 10:32:26 -07001457int sock_common_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07001458 char __user *optval, unsigned int optlen);
Joe Perches69336bd2013-09-22 10:32:26 -07001459int compat_sock_common_getsockopt(struct socket *sock, int level,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001460 int optname, char __user *optval, int __user *optlen);
Joe Perches69336bd2013-09-22 10:32:26 -07001461int compat_sock_common_setsockopt(struct socket *sock, int level,
David S. Millerb7058842009-09-30 16:12:20 -07001462 int optname, char __user *optval, unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001463
Joe Perches69336bd2013-09-22 10:32:26 -07001464void sk_common_release(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001465
1466/*
1467 * Default socket callbacks and setup code
1468 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001469
Linus Torvalds1da177e2005-04-16 15:20:36 -07001470/* Initialise core socket variables */
Joe Perches69336bd2013-09-22 10:32:26 -07001471void sock_init_data(struct socket *sock, struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472
Linus Torvalds1da177e2005-04-16 15:20:36 -07001473/*
1474 * Socket reference counting postulates.
1475 *
1476 * * Each user of socket SHOULD hold a reference count.
1477 * * Each access point to socket (an hash table bucket, reference from a list,
1478 * running timer, skb in flight MUST hold a reference count.
1479 * * When reference count hits 0, it means it will never increase back.
1480 * * When reference count hits 0, it means that no references from
1481 * outside exist to this socket and current process on current CPU
1482 * is last user and may/should destroy this socket.
1483 * * sk_free is called from any context: process, BH, IRQ. When
1484 * it is called, socket has no references from outside -> sk_free
1485 * may release descendant resources allocated by the socket, but
1486 * to the time when it is called, socket is NOT referenced by any
1487 * hash tables, lists etc.
1488 * * Packets, delivered from outside (from network or from another process)
1489 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1490 * when they sit in queue. Otherwise, packets will leak to hole, when
1491 * socket is looked up by one cpu and unhasing is made by another CPU.
1492 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1493 * (leak to backlog). Packet socket does all the processing inside
1494 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1495 * use separate SMP lock, so that they are prone too.
1496 */
1497
1498/* Ungrab socket and destroy it, if it was the last reference. */
1499static inline void sock_put(struct sock *sk)
1500{
1501 if (atomic_dec_and_test(&sk->sk_refcnt))
1502 sk_free(sk);
1503}
Eric Dumazet05dbc7b2013-10-03 00:22:02 -07001504/* Generic version of sock_put(), dealing with all sockets
Eric Dumazet41b822c2015-03-12 16:44:08 -07001505 * (TCP_TIMEWAIT, TCP_NEW_SYN_RECV, ESTABLISHED...)
Eric Dumazet05dbc7b2013-10-03 00:22:02 -07001506 */
1507void sock_gen_put(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001508
Joe Perches69336bd2013-09-22 10:32:26 -07001509int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested);
Arnaldo Carvalho de Melo25995ff2005-12-27 02:42:22 -02001510
Krishna Kumare022f0b2009-10-19 23:46:20 +00001511static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1512{
1513 sk->sk_tx_queue_mapping = tx_queue;
1514}
1515
1516static inline void sk_tx_queue_clear(struct sock *sk)
1517{
1518 sk->sk_tx_queue_mapping = -1;
1519}
1520
1521static inline int sk_tx_queue_get(const struct sock *sk)
1522{
Tom Herbertb0f77d02010-07-14 20:50:29 -07001523 return sk ? sk->sk_tx_queue_mapping : -1;
Krishna Kumare022f0b2009-10-19 23:46:20 +00001524}
1525
David S. Miller972692e2008-06-17 22:41:38 -07001526static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1527{
Krishna Kumare022f0b2009-10-19 23:46:20 +00001528 sk_tx_queue_clear(sk);
David S. Miller972692e2008-06-17 22:41:38 -07001529 sk->sk_socket = sock;
1530}
1531
Eric Dumazetaa395142010-04-20 13:03:51 +00001532static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1533{
Eric Dumazeteaefd112011-02-18 03:26:36 +00001534 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1535 return &rcu_dereference_raw(sk->sk_wq)->wait;
Eric Dumazetaa395142010-04-20 13:03:51 +00001536}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001537/* Detach socket from process context.
1538 * Announce socket dead, detach it from wait queue and inode.
1539 * Note that parent inode held reference count on this struct sock,
1540 * we do not release it in this function, because protocol
1541 * probably wants some additional cleanups or even continuing
1542 * to work with this socket (TCP).
1543 */
1544static inline void sock_orphan(struct sock *sk)
1545{
1546 write_lock_bh(&sk->sk_callback_lock);
1547 sock_set_flag(sk, SOCK_DEAD);
David S. Miller972692e2008-06-17 22:41:38 -07001548 sk_set_socket(sk, NULL);
Eric Dumazet43815482010-04-29 11:01:49 +00001549 sk->sk_wq = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550 write_unlock_bh(&sk->sk_callback_lock);
1551}
1552
1553static inline void sock_graft(struct sock *sk, struct socket *parent)
1554{
1555 write_lock_bh(&sk->sk_callback_lock);
Eric Dumazeteaefd112011-02-18 03:26:36 +00001556 sk->sk_wq = parent->wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557 parent->sk = sk;
David S. Miller972692e2008-06-17 22:41:38 -07001558 sk_set_socket(sk, parent);
Venkat Yekkirala4237c752006-07-24 23:32:50 -07001559 security_sock_graft(sk, parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001560 write_unlock_bh(&sk->sk_callback_lock);
1561}
1562
Joe Perches69336bd2013-09-22 10:32:26 -07001563kuid_t sock_i_uid(struct sock *sk);
1564unsigned long sock_i_ino(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565
Eric Dumazet58d607d2015-09-15 15:24:20 -07001566static inline u32 net_tx_rndhash(void)
1567{
1568 u32 v = prandom_u32();
1569
1570 return v ?: 1;
1571}
1572
Tom Herbert877d1f62015-07-28 16:02:05 -07001573static inline void sk_set_txhash(struct sock *sk)
1574{
Eric Dumazet58d607d2015-09-15 15:24:20 -07001575 sk->sk_txhash = net_tx_rndhash();
Tom Herbert877d1f62015-07-28 16:02:05 -07001576}
1577
Tom Herbert265f94f2015-07-28 16:02:06 -07001578static inline void sk_rethink_txhash(struct sock *sk)
1579{
1580 if (sk->sk_txhash)
1581 sk_set_txhash(sk);
1582}
1583
Linus Torvalds1da177e2005-04-16 15:20:36 -07001584static inline struct dst_entry *
1585__sk_dst_get(struct sock *sk)
1586{
Michal Hockod8bf4ca2011-07-08 14:39:41 +02001587 return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
Eric Dumazetf68c2242010-04-22 16:06:59 -07001588 lockdep_is_held(&sk->sk_lock.slock));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001589}
1590
1591static inline struct dst_entry *
1592sk_dst_get(struct sock *sk)
1593{
1594 struct dst_entry *dst;
1595
Eric Dumazetb6c67122010-04-08 23:03:29 +00001596 rcu_read_lock();
1597 dst = rcu_dereference(sk->sk_dst_cache);
Eric Dumazetf8864972014-06-24 10:05:11 -07001598 if (dst && !atomic_inc_not_zero(&dst->__refcnt))
1599 dst = NULL;
Eric Dumazetb6c67122010-04-08 23:03:29 +00001600 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001601 return dst;
1602}
1603
Eric Dumazetb6c67122010-04-08 23:03:29 +00001604static inline void dst_negative_advice(struct sock *sk)
1605{
1606 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1607
Tom Herbert265f94f2015-07-28 16:02:06 -07001608 sk_rethink_txhash(sk);
1609
Eric Dumazetb6c67122010-04-08 23:03:29 +00001610 if (dst && dst->ops->negative_advice) {
1611 ndst = dst->ops->negative_advice(dst);
1612
1613 if (ndst != dst) {
1614 rcu_assign_pointer(sk->sk_dst_cache, ndst);
ZHAO Gang0a6957e2013-10-22 16:23:38 +08001615 sk_tx_queue_clear(sk);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001616 }
1617 }
1618}
1619
Linus Torvalds1da177e2005-04-16 15:20:36 -07001620static inline void
1621__sk_dst_set(struct sock *sk, struct dst_entry *dst)
1622{
1623 struct dst_entry *old_dst;
1624
Krishna Kumare022f0b2009-10-19 23:46:20 +00001625 sk_tx_queue_clear(sk);
Eric Dumazet0b53ff22010-04-26 20:40:43 +00001626 /*
1627 * This can be called while sk is owned by the caller only,
1628 * with no state that can be checked in a rcu_dereference_check() cond
1629 */
1630 old_dst = rcu_dereference_raw(sk->sk_dst_cache);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001631 rcu_assign_pointer(sk->sk_dst_cache, dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001632 dst_release(old_dst);
1633}
1634
1635static inline void
1636sk_dst_set(struct sock *sk, struct dst_entry *dst)
1637{
Eric Dumazet7f502362014-06-30 01:26:23 -07001638 struct dst_entry *old_dst;
1639
1640 sk_tx_queue_clear(sk);
Eric Dumazet5925a052014-07-02 02:39:38 -07001641 old_dst = xchg((__force struct dst_entry **)&sk->sk_dst_cache, dst);
Eric Dumazet7f502362014-06-30 01:26:23 -07001642 dst_release(old_dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001643}
1644
1645static inline void
1646__sk_dst_reset(struct sock *sk)
1647{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001648 __sk_dst_set(sk, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001649}
1650
1651static inline void
1652sk_dst_reset(struct sock *sk)
1653{
Eric Dumazet7f502362014-06-30 01:26:23 -07001654 sk_dst_set(sk, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001655}
1656
Joe Perches69336bd2013-09-22 10:32:26 -07001657struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001658
Joe Perches69336bd2013-09-22 10:32:26 -07001659struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001660
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +02001661bool sk_mc_loop(struct sock *sk);
1662
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001663static inline bool sk_can_gso(const struct sock *sk)
Herbert Xubcd76112006-06-30 13:36:35 -07001664{
1665 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1666}
1667
Joe Perches69336bd2013-09-22 10:32:26 -07001668void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
Arnaldo Carvalho de Melo6cbb0df2005-08-09 19:49:02 -07001669
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001670static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
Eric Dumazeta4654192010-05-16 00:36:33 -07001671{
1672 sk->sk_route_nocaps |= flags;
1673 sk->sk_route_caps &= ~flags;
1674}
1675
Tom Herbert9a498502015-12-14 11:19:45 -08001676static inline bool sk_check_csum_caps(struct sock *sk)
1677{
1678 return (sk->sk_route_caps & NETIF_F_HW_CSUM) ||
1679 (sk->sk_family == PF_INET &&
1680 (sk->sk_route_caps & NETIF_F_IP_CSUM)) ||
1681 (sk->sk_family == PF_INET6 &&
1682 (sk->sk_route_caps & NETIF_F_IPV6_CSUM));
1683}
1684
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001685static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
Al Viro57be5bd2014-11-28 13:40:20 -05001686 struct iov_iter *from, char *to,
Wei Yongjun912d3982011-04-06 18:40:12 +00001687 int copy, int offset)
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001688{
1689 if (skb->ip_summed == CHECKSUM_NONE) {
Al Viro57be5bd2014-11-28 13:40:20 -05001690 __wsum csum = 0;
1691 if (csum_and_copy_from_iter(to, copy, &csum, from) != copy)
1692 return -EFAULT;
Wei Yongjun912d3982011-04-06 18:40:12 +00001693 skb->csum = csum_block_add(skb->csum, csum, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001694 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
Al Viro57be5bd2014-11-28 13:40:20 -05001695 if (copy_from_iter_nocache(to, copy, from) != copy)
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001696 return -EFAULT;
Al Viro57be5bd2014-11-28 13:40:20 -05001697 } else if (copy_from_iter(to, copy, from) != copy)
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001698 return -EFAULT;
1699
1700 return 0;
1701}
1702
1703static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
Al Viro57be5bd2014-11-28 13:40:20 -05001704 struct iov_iter *from, int copy)
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001705{
Wei Yongjun912d3982011-04-06 18:40:12 +00001706 int err, offset = skb->len;
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001707
Wei Yongjun912d3982011-04-06 18:40:12 +00001708 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1709 copy, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001710 if (err)
Wei Yongjun912d3982011-04-06 18:40:12 +00001711 __skb_trim(skb, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001712
1713 return err;
1714}
1715
Al Viro57be5bd2014-11-28 13:40:20 -05001716static inline int skb_copy_to_page_nocache(struct sock *sk, struct iov_iter *from,
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001717 struct sk_buff *skb,
1718 struct page *page,
1719 int off, int copy)
1720{
1721 int err;
1722
Wei Yongjun912d3982011-04-06 18:40:12 +00001723 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1724 copy, skb->len);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001725 if (err)
1726 return err;
1727
1728 skb->len += copy;
1729 skb->data_len += copy;
1730 skb->truesize += copy;
1731 sk->sk_wmem_queued += copy;
1732 sk_mem_charge(sk, copy);
1733 return 0;
1734}
1735
Eric Dumazetc5640392009-06-16 10:12:03 +00001736/**
1737 * sk_wmem_alloc_get - returns write allocations
1738 * @sk: socket
1739 *
1740 * Returns sk_wmem_alloc minus initial offset of one
1741 */
1742static inline int sk_wmem_alloc_get(const struct sock *sk)
1743{
1744 return atomic_read(&sk->sk_wmem_alloc) - 1;
1745}
1746
1747/**
1748 * sk_rmem_alloc_get - returns read allocations
1749 * @sk: socket
1750 *
1751 * Returns sk_rmem_alloc
1752 */
1753static inline int sk_rmem_alloc_get(const struct sock *sk)
1754{
1755 return atomic_read(&sk->sk_rmem_alloc);
1756}
1757
1758/**
1759 * sk_has_allocations - check if allocations are outstanding
1760 * @sk: socket
1761 *
1762 * Returns true if socket has write or read allocations
1763 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001764static inline bool sk_has_allocations(const struct sock *sk)
Eric Dumazetc5640392009-06-16 10:12:03 +00001765{
1766 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1767}
1768
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001769/**
Herbert Xu1ce0bf52015-11-26 13:55:39 +08001770 * skwq_has_sleeper - check if there are any waiting processes
Randy Dunlapacfbe962010-05-24 23:54:18 -07001771 * @wq: struct socket_wq
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001772 *
Eric Dumazet43815482010-04-29 11:01:49 +00001773 * Returns true if socket_wq has waiting processes
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001774 *
Herbert Xu1ce0bf52015-11-26 13:55:39 +08001775 * The purpose of the skwq_has_sleeper and sock_poll_wait is to wrap the memory
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001776 * barrier call. They were added due to the race found within the tcp code.
1777 *
1778 * Consider following tcp code paths:
1779 *
1780 * CPU1 CPU2
1781 *
1782 * sys_select receive packet
1783 * ... ...
1784 * __add_wait_queue update tp->rcv_nxt
1785 * ... ...
1786 * tp->rcv_nxt check sock_def_readable
1787 * ... {
Eric Dumazet43815482010-04-29 11:01:49 +00001788 * schedule rcu_read_lock();
1789 * wq = rcu_dereference(sk->sk_wq);
1790 * if (wq && waitqueue_active(&wq->wait))
1791 * wake_up_interruptible(&wq->wait)
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001792 * ...
1793 * }
1794 *
1795 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1796 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1797 * could then endup calling schedule and sleep forever if there are no more
1798 * data on the socket.
Jiri Olsaad462762009-07-08 12:10:31 +00001799 *
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001800 */
Herbert Xu1ce0bf52015-11-26 13:55:39 +08001801static inline bool skwq_has_sleeper(struct socket_wq *wq)
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001802{
Herbert Xu1ce0bf52015-11-26 13:55:39 +08001803 return wq && wq_has_sleeper(&wq->wait);
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001804}
1805
1806/**
1807 * sock_poll_wait - place memory barrier behind the poll_wait call.
1808 * @filp: file
1809 * @wait_address: socket wait queue
1810 * @p: poll_table
1811 *
Eric Dumazet43815482010-04-29 11:01:49 +00001812 * See the comments in the wq_has_sleeper function.
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001813 */
1814static inline void sock_poll_wait(struct file *filp,
1815 wait_queue_head_t *wait_address, poll_table *p)
1816{
Hans Verkuil626cf232012-03-23 15:02:27 -07001817 if (!poll_does_not_wait(p) && wait_address) {
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001818 poll_wait(filp, wait_address, p);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001819 /* We need to be sure we are in sync with the
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001820 * socket flags modification.
1821 *
Eric Dumazet43815482010-04-29 11:01:49 +00001822 * This memory barrier is paired in the wq_has_sleeper.
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001823 */
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001824 smp_mb();
1825 }
1826}
1827
Tom Herbertb73c3d02014-07-01 21:32:17 -07001828static inline void skb_set_hash_from_sk(struct sk_buff *skb, struct sock *sk)
1829{
1830 if (sk->sk_txhash) {
1831 skb->l4_hash = 1;
1832 skb->hash = sk->sk_txhash;
1833 }
1834}
1835
Eric Dumazet9e17f8a2015-11-01 15:36:55 -08001836void skb_set_owner_w(struct sk_buff *skb, struct sock *sk);
1837
Linus Torvalds1da177e2005-04-16 15:20:36 -07001838/*
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001839 * Queue a received datagram if it will fit. Stream and sequenced
Linus Torvalds1da177e2005-04-16 15:20:36 -07001840 * protocols can't normally use this as they need to fit buffers in
1841 * and play with them.
1842 *
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001843 * Inlined as it's very short and called for pretty much every
Linus Torvalds1da177e2005-04-16 15:20:36 -07001844 * packet ever received.
1845 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001846static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1847{
Herbert Xud55d87f2009-06-22 02:25:25 +00001848 skb_orphan(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001849 skb->sk = sk;
1850 skb->destructor = sock_rfree;
1851 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001852 sk_mem_charge(sk, skb->truesize);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001853}
1854
Joe Perches69336bd2013-09-22 10:32:26 -07001855void sk_reset_timer(struct sock *sk, struct timer_list *timer,
1856 unsigned long expires);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001857
Joe Perches69336bd2013-09-22 10:32:26 -07001858void sk_stop_timer(struct sock *sk, struct timer_list *timer);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001859
Joe Perches69336bd2013-09-22 10:32:26 -07001860int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001861
Joe Perches69336bd2013-09-22 10:32:26 -07001862int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
Willem de Bruijn364a9e92014-08-31 21:30:27 -04001863struct sk_buff *sock_dequeue_err_skb(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001864
1865/*
1866 * Recover an error report and clear atomically
1867 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001868
Linus Torvalds1da177e2005-04-16 15:20:36 -07001869static inline int sock_error(struct sock *sk)
1870{
Benjamin LaHaisec1cbe4b2005-12-13 23:22:19 -08001871 int err;
1872 if (likely(!sk->sk_err))
1873 return 0;
1874 err = xchg(&sk->sk_err, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001875 return -err;
1876}
1877
1878static inline unsigned long sock_wspace(struct sock *sk)
1879{
1880 int amt = 0;
1881
1882 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1883 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001884 if (amt < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001885 amt = 0;
1886 }
1887 return amt;
1888}
1889
Eric Dumazetceb5d582015-11-29 20:03:11 -08001890/* Note:
1891 * We use sk->sk_wq_raw, from contexts knowing this
1892 * pointer is not NULL and cannot disappear/change.
1893 */
Eric Dumazet9cd3e072015-11-29 20:03:10 -08001894static inline void sk_set_bit(int nr, struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001895{
Eric Dumazetceb5d582015-11-29 20:03:11 -08001896 set_bit(nr, &sk->sk_wq_raw->flags);
Eric Dumazet9cd3e072015-11-29 20:03:10 -08001897}
1898
1899static inline void sk_clear_bit(int nr, struct sock *sk)
1900{
Eric Dumazetceb5d582015-11-29 20:03:11 -08001901 clear_bit(nr, &sk->sk_wq_raw->flags);
Eric Dumazet9cd3e072015-11-29 20:03:10 -08001902}
1903
Eric Dumazetceb5d582015-11-29 20:03:11 -08001904static inline void sk_wake_async(const struct sock *sk, int how, int band)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001905{
Eric Dumazetceb5d582015-11-29 20:03:11 -08001906 if (sock_flag(sk, SOCK_FASYNC)) {
1907 rcu_read_lock();
1908 sock_wake_async(rcu_dereference(sk->sk_wq), how, band);
1909 rcu_read_unlock();
1910 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001911}
1912
Daniel Borkmanneea86af2013-06-19 12:51:20 +02001913/* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
1914 * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
1915 * Note: for send buffers, TCP works better if we can build two skbs at
1916 * minimum.
Eric Dumazet7a91b4342010-09-26 18:53:07 -07001917 */
Eric Dumazet9eb5bf82013-07-03 05:02:22 -07001918#define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
Daniel Borkmanneea86af2013-06-19 12:51:20 +02001919
1920#define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2)
1921#define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE
Linus Torvalds1da177e2005-04-16 15:20:36 -07001922
1923static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1924{
1925 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
Eric Dumazet8df09ea2007-12-21 03:07:41 -08001926 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
Daniel Borkmanneea86af2013-06-19 12:51:20 +02001927 sk->sk_sndbuf = max_t(u32, sk->sk_sndbuf, SOCK_MIN_SNDBUF);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001928 }
1929}
1930
Eric Dumazeteb934472015-05-19 13:26:55 -07001931struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp,
1932 bool force_schedule);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001933
Eric Dumazet5640f762012-09-23 23:04:42 +00001934/**
1935 * sk_page_frag - return an appropriate page_frag
1936 * @sk: socket
1937 *
1938 * If socket allocation mode allows current thread to sleep, it means its
1939 * safe to use the per task page_frag instead of the per socket one.
1940 */
1941static inline struct page_frag *sk_page_frag(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001942{
Mel Gormand0164ad2015-11-06 16:28:21 -08001943 if (gfpflags_allow_blocking(sk->sk_allocation))
Eric Dumazet5640f762012-09-23 23:04:42 +00001944 return &current->task_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001945
Eric Dumazet5640f762012-09-23 23:04:42 +00001946 return &sk->sk_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001947}
1948
Joe Perches69336bd2013-09-22 10:32:26 -07001949bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag);
Eric Dumazet5640f762012-09-23 23:04:42 +00001950
Linus Torvalds1da177e2005-04-16 15:20:36 -07001951/*
1952 * Default write policy as shown to user space via poll/select/SIGIO
1953 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001954static inline bool sock_writeable(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001955{
Eric Dumazet8df09ea2007-12-21 03:07:41 -08001956 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957}
1958
Al Virodd0fc662005-10-07 07:46:04 +01001959static inline gfp_t gfp_any(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001960{
Andrew Morton99709372009-02-12 16:43:17 -08001961 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001962}
1963
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001964static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001965{
1966 return noblock ? 0 : sk->sk_rcvtimeo;
1967}
1968
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001969static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001970{
1971 return noblock ? 0 : sk->sk_sndtimeo;
1972}
1973
1974static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1975{
1976 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1977}
1978
1979/* Alas, with timeout socket operations are not restartable.
1980 * Compare this to poll().
1981 */
1982static inline int sock_intr_errno(long timeo)
1983{
1984 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1985}
1986
Eyal Birger744d5a32015-03-01 14:58:31 +02001987struct sock_skb_cb {
1988 u32 dropcount;
1989};
1990
1991/* Store sock_skb_cb at the end of skb->cb[] so protocol families
1992 * using skb->cb[] would keep using it directly and utilize its
1993 * alignement guarantee.
1994 */
1995#define SOCK_SKB_CB_OFFSET ((FIELD_SIZEOF(struct sk_buff, cb) - \
1996 sizeof(struct sock_skb_cb)))
1997
1998#define SOCK_SKB_CB(__skb) ((struct sock_skb_cb *)((__skb)->cb + \
1999 SOCK_SKB_CB_OFFSET))
2000
Eyal Birgerb4772ef2015-03-01 14:58:29 +02002001#define sock_skb_cb_check_size(size) \
Eyal Birger744d5a32015-03-01 14:58:31 +02002002 BUILD_BUG_ON((size) > SOCK_SKB_CB_OFFSET)
Eyal Birgerb4772ef2015-03-01 14:58:29 +02002003
Eyal Birger3bc3b962015-03-01 14:58:30 +02002004static inline void
2005sock_skb_set_dropcount(const struct sock *sk, struct sk_buff *skb)
2006{
Eyal Birger744d5a32015-03-01 14:58:31 +02002007 SOCK_SKB_CB(skb)->dropcount = atomic_read(&sk->sk_drops);
Eyal Birger3bc3b962015-03-01 14:58:30 +02002008}
2009
Joe Perches69336bd2013-09-22 10:32:26 -07002010void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2011 struct sk_buff *skb);
2012void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2013 struct sk_buff *skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -07002014
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002015static inline void
Linus Torvalds1da177e2005-04-16 15:20:36 -07002016sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2017{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002018 ktime_t kt = skb->tstamp;
Patrick Ohly20d49472009-02-12 05:03:38 +00002019 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002020
Patrick Ohly20d49472009-02-12 05:03:38 +00002021 /*
2022 * generate control messages if
Willem de Bruijnb9f40e22014-08-04 22:11:46 -04002023 * - receive time stamping in software requested
Patrick Ohly20d49472009-02-12 05:03:38 +00002024 * - software time stamp available and wanted
Patrick Ohly20d49472009-02-12 05:03:38 +00002025 * - hardware time stamps available and wanted
Patrick Ohly20d49472009-02-12 05:03:38 +00002026 */
2027 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
Willem de Bruijnb9f40e22014-08-04 22:11:46 -04002028 (sk->sk_tsflags & SOF_TIMESTAMPING_RX_SOFTWARE) ||
Willem de Bruijnc1991052014-09-03 12:01:18 -04002029 (kt.tv64 && sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) ||
Patrick Ohly20d49472009-02-12 05:03:38 +00002030 (hwtstamps->hwtstamp.tv64 &&
Willem de Bruijnb9f40e22014-08-04 22:11:46 -04002031 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE)))
Eric Dumazet92f37fd2007-03-25 22:14:49 -07002032 __sock_recv_timestamp(msg, sk, skb);
2033 else
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002034 sk->sk_stamp = kt;
Johannes Berg6e3e9392011-11-09 10:15:42 +01002035
2036 if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2037 __sock_recv_wifi_status(msg, sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002038}
2039
Joe Perches69336bd2013-09-22 10:32:26 -07002040void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2041 struct sk_buff *skb);
Eric Dumazet767dd032010-04-28 19:14:43 +00002042
2043static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2044 struct sk_buff *skb)
2045{
2046#define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
Willem de Bruijnb9f40e22014-08-04 22:11:46 -04002047 (1UL << SOCK_RCVTSTAMP))
2048#define TSFLAGS_ANY (SOF_TIMESTAMPING_SOFTWARE | \
2049 SOF_TIMESTAMPING_RAW_HARDWARE)
Eric Dumazet767dd032010-04-28 19:14:43 +00002050
Willem de Bruijnb9f40e22014-08-04 22:11:46 -04002051 if (sk->sk_flags & FLAGS_TS_OR_DROPS || sk->sk_tsflags & TSFLAGS_ANY)
Eric Dumazet767dd032010-04-28 19:14:43 +00002052 __sock_recv_ts_and_drops(msg, sk, skb);
2053 else
2054 sk->sk_stamp = skb->tstamp;
2055}
Neil Horman3b885782009-10-12 13:26:31 -07002056
Willem de Bruijn67cc0d42014-09-08 19:58:58 -04002057void __sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags);
2058
Linus Torvalds1da177e2005-04-16 15:20:36 -07002059/**
Patrick Ohly20d49472009-02-12 05:03:38 +00002060 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
Patrick Ohly20d49472009-02-12 05:03:38 +00002061 * @sk: socket sending this packet
Eric Dumazet140c55d2014-08-06 11:49:29 +02002062 * @tx_flags: completed with instructions for time stamping
2063 *
2064 * Note : callers should take care of initial *tx_flags value (usually 0)
Patrick Ohly20d49472009-02-12 05:03:38 +00002065 */
Willem de Bruijn67cc0d42014-09-08 19:58:58 -04002066static inline void sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags)
2067{
2068 if (unlikely(sk->sk_tsflags))
2069 __sock_tx_timestamp(sk, tx_flags);
2070 if (unlikely(sock_flag(sk, SOCK_WIFI_STATUS)))
2071 *tx_flags |= SKBTX_WIFI_STATUS;
2072}
Patrick Ohly20d49472009-02-12 05:03:38 +00002073
2074/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002075 * sk_eat_skb - Release a skb if it is no longer needed
Pavel Pisa4dc3b162005-05-01 08:59:25 -07002076 * @sk: socket to eat this skb from
2077 * @skb: socket buffer to eat
Linus Torvalds1da177e2005-04-16 15:20:36 -07002078 *
2079 * This routine must be called with interrupts disabled or with the socket
2080 * locked so that the sk_buff queue operation is ok.
2081*/
Dan Williams7bced392013-12-30 12:37:29 -08002082static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002083{
2084 __skb_unlink(skb, &sk->sk_receive_queue);
2085 __kfree_skb(skb);
2086}
2087
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002088static inline
2089struct net *sock_net(const struct sock *sk)
2090{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002091 return read_pnet(&sk->sk_net);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002092}
2093
2094static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07002095void sock_net_set(struct sock *sk, struct net *net)
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002096{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002097 write_pnet(&sk->sk_net, net);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002098}
2099
KOVACS Krisztian23542612008-10-07 12:41:01 -07002100static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2101{
Vijay Subramanianefc27f82012-06-24 13:03:07 +00002102 if (skb->sk) {
KOVACS Krisztian23542612008-10-07 12:41:01 -07002103 struct sock *sk = skb->sk;
2104
2105 skb->destructor = NULL;
2106 skb->sk = NULL;
2107 return sk;
2108 }
2109 return NULL;
2110}
2111
Eric Dumazet1d0ab252015-03-15 21:12:12 -07002112/* This helper checks if a socket is a full socket,
2113 * ie _not_ a timewait or request socket.
2114 */
2115static inline bool sk_fullsock(const struct sock *sk)
2116{
2117 return (1 << sk->sk_state) & ~(TCPF_TIME_WAIT | TCPF_NEW_SYN_RECV);
2118}
2119
Eric Dumazete446f9d2015-10-08 05:01:55 -07002120/* This helper checks if a socket is a LISTEN or NEW_SYN_RECV
2121 * SYNACK messages can be attached to either ones (depending on SYNCOOKIE)
2122 */
2123static inline bool sk_listener(const struct sock *sk)
2124{
2125 return (1 << sk->sk_state) & (TCPF_LISTEN | TCPF_NEW_SYN_RECV);
2126}
2127
Eric Dumazet00fd38d2015-11-12 08:43:18 -08002128/**
2129 * sk_state_load - read sk->sk_state for lockless contexts
2130 * @sk: socket pointer
2131 *
2132 * Paired with sk_state_store(). Used in places we do not hold socket lock :
2133 * tcp_diag_get_info(), tcp_get_info(), tcp_poll(), get_tcp4_sock() ...
2134 */
2135static inline int sk_state_load(const struct sock *sk)
2136{
2137 return smp_load_acquire(&sk->sk_state);
2138}
2139
2140/**
2141 * sk_state_store - update sk->sk_state
2142 * @sk: socket pointer
2143 * @newstate: new state
2144 *
2145 * Paired with sk_state_load(). Should be used in contexts where
2146 * state change might impact lockless readers.
2147 */
2148static inline void sk_state_store(struct sock *sk, int newstate)
2149{
2150 smp_store_release(&sk->sk_state, newstate);
2151}
2152
Joe Perches69336bd2013-09-22 10:32:26 -07002153void sock_enable_timestamp(struct sock *sk, int flag);
2154int sock_get_timestamp(struct sock *, struct timeval __user *);
2155int sock_get_timestampns(struct sock *, struct timespec __user *);
2156int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len, int level,
2157 int type);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002158
Eric W. Biedermana3b299d2014-04-23 14:26:56 -07002159bool sk_ns_capable(const struct sock *sk,
2160 struct user_namespace *user_ns, int cap);
2161bool sk_capable(const struct sock *sk, int cap);
2162bool sk_net_capable(const struct sock *sk, int cap);
2163
Linus Torvalds1da177e2005-04-16 15:20:36 -07002164extern __u32 sysctl_wmem_max;
2165extern __u32 sysctl_rmem_max;
2166
Willem de Bruijnb245be12015-01-30 13:29:32 -05002167extern int sysctl_tstamp_allow_data;
David S. Miller6baf1f42005-09-05 18:14:11 -07002168extern int sysctl_optmem_max;
2169
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002170extern __u32 sysctl_wmem_default;
2171extern __u32 sysctl_rmem_default;
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002172
Linus Torvalds1da177e2005-04-16 15:20:36 -07002173#endif /* _SOCK_H */