blob: 19032437de56521f3183cacd23c88be988e0fa1b [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070061
62#include <linux/filter.h>
Eric Dumazet88ab1932008-11-16 19:39:21 -080063#include <linux/rculist_nulls.h>
Jiri Olsaa57de0b2009-07-08 12:09:13 +000064#include <linux/poll.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070065
Eric Dumazetc31504d2010-11-15 19:58:26 +000066#include <linux/atomic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070067#include <net/dst.h>
68#include <net/checksum.h>
Eric Dumazet1d0ab252015-03-15 21:12:12 -070069#include <net/tcp_states.h>
Willem de Bruijnb9f40e22014-08-04 22:11:46 -040070#include <linux/net_tstamp.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070071
Eric Dumazet9f048bf2011-12-13 03:59:08 +000072struct cgroup;
73struct cgroup_subsys;
Glauber Costac607b2ed2011-12-16 00:52:00 +000074#ifdef CONFIG_NET
Glauber Costa1d62e432012-04-09 19:36:33 -030075int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss);
76void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg);
Glauber Costac607b2ed2011-12-16 00:52:00 +000077#else
78static inline
Glauber Costa1d62e432012-04-09 19:36:33 -030079int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
Glauber Costac607b2ed2011-12-16 00:52:00 +000080{
81 return 0;
82}
83static inline
Glauber Costa1d62e432012-04-09 19:36:33 -030084void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg)
Glauber Costac607b2ed2011-12-16 00:52:00 +000085{
86}
87#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -070088/*
89 * This structure really needs to be cleaned up.
90 * Most of it is for TCP, and not used by any of
91 * the other protocols.
92 */
93
94/* Define this to get the SOCK_DBG debugging facility. */
95#define SOCK_DEBUGGING
96#ifdef SOCK_DEBUGGING
97#define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
98 printk(KERN_DEBUG msg); } while (0)
99#else
Stephen Hemminger4cd90292008-03-21 15:54:53 -0700100/* Validate arguments and do nothing */
Joe Perchesb9075fa2011-10-31 17:11:33 -0700101static inline __printf(2, 3)
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000102void SOCK_DEBUG(const struct sock *sk, const char *msg, ...)
Stephen Hemminger4cd90292008-03-21 15:54:53 -0700103{
104}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105#endif
106
107/* This is the per-socket lock. The spinlock provides a synchronization
108 * between user contexts and software interrupt processing, whereas the
109 * mini-semaphore synchronizes multiple users amongst themselves.
110 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111typedef struct {
112 spinlock_t slock;
John Heffnerd2e91172007-09-12 10:44:19 +0200113 int owned;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114 wait_queue_head_t wq;
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700115 /*
116 * We express the mutex-alike socket_lock semantics
117 * to the lock validator by explicitly managing
118 * the slock as a lock variant (in addition to
119 * the slock itself):
120 */
121#ifdef CONFIG_DEBUG_LOCK_ALLOC
122 struct lockdep_map dep_map;
123#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124} socket_lock_t;
125
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126struct sock;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700127struct proto;
Denis V. Lunev0eeb8ff2007-12-04 01:15:45 -0800128struct net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129
Eric Dumazet077b3932012-12-02 07:33:10 +0000130typedef __u32 __bitwise __portpair;
131typedef __u64 __bitwise __addrpair;
132
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133/**
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700134 * struct sock_common - minimal network layer representation of sockets
Eric Dumazet68835ab2010-11-30 19:04:07 +0000135 * @skc_daddr: Foreign IPv4 addr
136 * @skc_rcv_saddr: Bound local IPv4 addr
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000137 * @skc_hash: hash value used with various protocol lookup tables
Eric Dumazetd4cada42009-11-08 10:17:30 +0000138 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
Eric Dumazetce43b032012-11-30 09:49:27 +0000139 * @skc_dport: placeholder for inet_dport/tw_dport
140 * @skc_num: placeholder for inet_num/tw_num
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700141 * @skc_family: network address family
142 * @skc_state: Connection state
143 * @skc_reuse: %SO_REUSEADDR setting
Tom Herbert055dc212013-01-22 09:49:50 +0000144 * @skc_reuseport: %SO_REUSEPORT setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700145 * @skc_bound_dev_if: bound device index if != 0
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700146 * @skc_bind_node: bind hash linkage for various protocol lookup tables
Eric Dumazet512615b2009-11-08 10:17:58 +0000147 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700148 * @skc_prot: protocol handlers inside a network family
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200149 * @skc_net: reference to the network namespace of this socket
Eric Dumazet68835ab2010-11-30 19:04:07 +0000150 * @skc_node: main hash linkage for various protocol lookup tables
151 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
152 * @skc_tx_queue_mapping: tx queue number for this connection
153 * @skc_refcnt: reference count
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700154 *
155 * This is the minimal network layer representation of sockets, the header
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700156 * for struct sock and struct inet_timewait_sock.
157 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158struct sock_common {
Eric Dumazetce43b032012-11-30 09:49:27 +0000159 /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned
Eric Dumazet05dbc7b2013-10-03 00:22:02 -0700160 * address on 64bit arches : cf INET_MATCH()
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000161 */
Eric Dumazetce43b032012-11-30 09:49:27 +0000162 union {
Eric Dumazet077b3932012-12-02 07:33:10 +0000163 __addrpair skc_addrpair;
Eric Dumazetce43b032012-11-30 09:49:27 +0000164 struct {
165 __be32 skc_daddr;
166 __be32 skc_rcv_saddr;
167 };
168 };
Eric Dumazetd4cada42009-11-08 10:17:30 +0000169 union {
170 unsigned int skc_hash;
171 __u16 skc_u16hashes[2];
172 };
Eric Dumazetce43b032012-11-30 09:49:27 +0000173 /* skc_dport && skc_num must be grouped as well */
174 union {
Eric Dumazet077b3932012-12-02 07:33:10 +0000175 __portpair skc_portpair;
Eric Dumazetce43b032012-11-30 09:49:27 +0000176 struct {
177 __be16 skc_dport;
178 __u16 skc_num;
179 };
180 };
181
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000182 unsigned short skc_family;
183 volatile unsigned char skc_state;
Tom Herbert055dc212013-01-22 09:49:50 +0000184 unsigned char skc_reuse:4;
Eric Dumazet9fe516b2014-06-27 08:36:16 -0700185 unsigned char skc_reuseport:1;
186 unsigned char skc_ipv6only:1;
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000187 int skc_bound_dev_if;
Eric Dumazet512615b2009-11-08 10:17:58 +0000188 union {
189 struct hlist_node skc_bind_node;
190 struct hlist_nulls_node skc_portaddr_node;
191 };
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700192 struct proto *skc_prot;
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -0500193 possible_net_t skc_net;
Eric Dumazetefe42082013-10-03 15:42:29 -0700194
195#if IS_ENABLED(CONFIG_IPV6)
196 struct in6_addr skc_v6_daddr;
197 struct in6_addr skc_v6_rcv_saddr;
198#endif
199
Eric Dumazet33cf7c92015-03-11 18:53:14 -0700200 atomic64_t skc_cookie;
201
Eric Dumazet68835ab2010-11-30 19:04:07 +0000202 /*
203 * fields between dontcopy_begin/dontcopy_end
204 * are not copied in sock_copy()
205 */
Randy Dunlap928c41e2011-01-08 17:39:21 +0000206 /* private: */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000207 int skc_dontcopy_begin[0];
Randy Dunlap928c41e2011-01-08 17:39:21 +0000208 /* public: */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000209 union {
210 struct hlist_node skc_node;
211 struct hlist_nulls_node skc_nulls_node;
212 };
213 int skc_tx_queue_mapping;
214 atomic_t skc_refcnt;
Randy Dunlap928c41e2011-01-08 17:39:21 +0000215 /* private: */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000216 int skc_dontcopy_end[0];
Randy Dunlap928c41e2011-01-08 17:39:21 +0000217 /* public: */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700218};
219
Glauber Costae1aab162011-12-11 21:47:03 +0000220struct cg_proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221/**
222 * struct sock - network layer representation of sockets
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700223 * @__sk_common: shared layout with inet_timewait_sock
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700224 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
225 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
226 * @sk_lock: synchronizer
227 * @sk_rcvbuf: size of receive buffer in bytes
Eric Dumazet43815482010-04-29 11:01:49 +0000228 * @sk_wq: sock wait queue and async head
Shawn Bohrer421b3882013-10-07 11:01:39 -0500229 * @sk_rx_dst: receive input route used by early demux
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700230 * @sk_dst_cache: destination cache
231 * @sk_dst_lock: destination cache lock
232 * @sk_policy: flow policy
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700233 * @sk_receive_queue: incoming packets
234 * @sk_wmem_alloc: transmit queue bytes committed
235 * @sk_write_queue: Packet sending queue
236 * @sk_omem_alloc: "o" is "option" or "other"
237 * @sk_wmem_queued: persistent queue size
238 * @sk_forward_alloc: space allocated forward
Eliezer Tamir06021292013-06-10 11:39:50 +0300239 * @sk_napi_id: id of the last napi context to receive data for sk
Eliezer Tamirdafcc432013-06-14 16:33:57 +0300240 * @sk_ll_usec: usecs to busypoll when there is no data
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700241 * @sk_allocation: allocation mode
Eric Dumazet95bd09e2013-08-27 05:46:32 -0700242 * @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
Eric Dumazetc3f40d72013-09-29 01:12:40 -0700243 * @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE)
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700244 * @sk_sndbuf: size of send buffer in bytes
Wang Chen33c732c2007-11-13 20:30:01 -0800245 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
Patrick Ohly20d49472009-02-12 05:03:38 +0000246 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
Tom Herbert28448b82014-05-23 08:47:19 -0700247 * @sk_no_check_tx: %SO_NO_CHECK setting, set checksum in TX packets
248 * @sk_no_check_rx: allow zero checksum in RX packets
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700249 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
Eric Dumazeta4654192010-05-16 00:36:33 -0700250 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
Herbert Xubcd76112006-06-30 13:36:35 -0700251 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -0700252 * @sk_gso_max_size: Maximum GSO segment size to build
Ben Hutchings14853482012-07-30 16:11:42 +0000253 * @sk_gso_max_segs: Maximum number of GSO segments
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700254 * @sk_lingertime: %SO_LINGER l_linger setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700255 * @sk_backlog: always used with the per-socket spinlock held
256 * @sk_callback_lock: used with the callbacks in the end of this struct
257 * @sk_error_queue: rarely used
Wang Chen33c732c2007-11-13 20:30:01 -0800258 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
259 * IPV6_ADDRFORM for instance)
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700260 * @sk_err: last error
Wang Chen33c732c2007-11-13 20:30:01 -0800261 * @sk_err_soft: errors that don't cause failure but are the cause of a
262 * persistent failure not just 'timed out'
Eric Dumazetcb61cb92008-06-17 21:04:56 -0700263 * @sk_drops: raw/udp drops counter
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700264 * @sk_ack_backlog: current listen backlog
265 * @sk_max_ack_backlog: listen backlog set in listen()
266 * @sk_priority: %SO_PRIORITY setting
Randy Dunlap1a3bc362012-01-21 09:03:10 +0000267 * @sk_cgrp_prioidx: socket group's priority map index
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700268 * @sk_type: socket type (%SOCK_STREAM, etc)
269 * @sk_protocol: which protocol this socket belongs in this network family
Randy Dunlap53c3fa22010-08-09 13:41:07 +0000270 * @sk_peer_pid: &struct pid for this socket's peer
271 * @sk_peer_cred: %SO_PEERCRED setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700272 * @sk_rcvlowat: %SO_RCVLOWAT setting
273 * @sk_rcvtimeo: %SO_RCVTIMEO setting
274 * @sk_sndtimeo: %SO_SNDTIMEO setting
David S. Millerc58dc012010-04-27 15:05:31 -0700275 * @sk_rxhash: flow hash received from netif layer
Eric Dumazet2c8c56e2014-11-11 05:54:28 -0800276 * @sk_incoming_cpu: record cpu processing incoming packets
Tom Herbertb73c3d02014-07-01 21:32:17 -0700277 * @sk_txhash: computed flow hash for use on transmit
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700278 * @sk_filter: socket filtering instructions
279 * @sk_protinfo: private area, net family specific, when not using slab
280 * @sk_timer: sock cleanup timer
281 * @sk_stamp: time stamp of last packet received
Willem de Bruijnb9f40e22014-08-04 22:11:46 -0400282 * @sk_tsflags: SO_TIMESTAMPING socket options
Willem de Bruijn09c2d252014-08-04 22:11:47 -0400283 * @sk_tskey: counter to disambiguate concurrent tstamp requests
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700284 * @sk_socket: Identd and reporting IO signals
285 * @sk_user_data: RPC layer private data
Eric Dumazet5640f762012-09-23 23:04:42 +0000286 * @sk_frag: cached page frag
Randy Dunlapd3d4f0a2012-04-17 14:03:53 +0000287 * @sk_peek_off: current peek_offset value
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700288 * @sk_send_head: front of stuff to transmit
Martin Waitz67be2dd2005-05-01 08:59:26 -0700289 * @sk_security: used by security modules
Randy Dunlap31729362008-02-18 20:52:13 -0800290 * @sk_mark: generic packet mark
Randy Dunlap53c3fa22010-08-09 13:41:07 +0000291 * @sk_classid: this socket's cgroup classid
Glauber Costae1aab162011-12-11 21:47:03 +0000292 * @sk_cgrp: this socket's cgroup-specific proto data
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700293 * @sk_write_pending: a write to stream socket waits to start
294 * @sk_state_change: callback to indicate change in the state of the sock
295 * @sk_data_ready: callback to indicate there is data to be processed
296 * @sk_write_space: callback to indicate there is bf sending space available
297 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
298 * @sk_backlog_rcv: callback to process the backlog
299 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300 */
301struct sock {
302 /*
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700303 * Now struct inet_timewait_sock also uses sock_common, so please just
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304 * don't add nothing before this first member (__sk_common) --acme
305 */
306 struct sock_common __sk_common;
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000307#define sk_node __sk_common.skc_node
308#define sk_nulls_node __sk_common.skc_nulls_node
309#define sk_refcnt __sk_common.skc_refcnt
Krishna Kumare022f0b2009-10-19 23:46:20 +0000310#define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000311
Eric Dumazet68835ab2010-11-30 19:04:07 +0000312#define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
313#define sk_dontcopy_end __sk_common.skc_dontcopy_end
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000314#define sk_hash __sk_common.skc_hash
Eric Dumazet50805462013-10-02 04:29:50 -0700315#define sk_portpair __sk_common.skc_portpair
Eric Dumazet05dbc7b2013-10-03 00:22:02 -0700316#define sk_num __sk_common.skc_num
317#define sk_dport __sk_common.skc_dport
Eric Dumazet50805462013-10-02 04:29:50 -0700318#define sk_addrpair __sk_common.skc_addrpair
319#define sk_daddr __sk_common.skc_daddr
320#define sk_rcv_saddr __sk_common.skc_rcv_saddr
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321#define sk_family __sk_common.skc_family
322#define sk_state __sk_common.skc_state
323#define sk_reuse __sk_common.skc_reuse
Tom Herbert055dc212013-01-22 09:49:50 +0000324#define sk_reuseport __sk_common.skc_reuseport
Eric Dumazet9fe516b2014-06-27 08:36:16 -0700325#define sk_ipv6only __sk_common.skc_ipv6only
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326#define sk_bound_dev_if __sk_common.skc_bound_dev_if
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327#define sk_bind_node __sk_common.skc_bind_node
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700328#define sk_prot __sk_common.skc_prot
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200329#define sk_net __sk_common.skc_net
Eric Dumazetefe42082013-10-03 15:42:29 -0700330#define sk_v6_daddr __sk_common.skc_v6_daddr
331#define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr
Eric Dumazet33cf7c92015-03-11 18:53:14 -0700332#define sk_cookie __sk_common.skc_cookie
Eric Dumazetefe42082013-10-03 15:42:29 -0700333
Eric Dumazetb178bb32010-11-16 05:56:04 +0000334 socket_lock_t sk_lock;
335 struct sk_buff_head sk_receive_queue;
336 /*
337 * The backlog queue is special, it is always used with
338 * the per-socket spinlock held and requires low latency
339 * access. Therefore we special case it's implementation.
340 * Note : rmem_alloc is in this structure to fill a hole
341 * on 64bit arches, not because its logically part of
342 * backlog.
343 */
344 struct {
345 atomic_t rmem_alloc;
346 int len;
347 struct sk_buff *head;
348 struct sk_buff *tail;
349 } sk_backlog;
350#define sk_rmem_alloc sk_backlog.rmem_alloc
351 int sk_forward_alloc;
352#ifdef CONFIG_RPS
353 __u32 sk_rxhash;
354#endif
Eric Dumazet2c8c56e2014-11-11 05:54:28 -0800355 u16 sk_incoming_cpu;
356 /* 16bit hole
357 * Warned : sk_incoming_cpu can be set from softirq,
358 * Do not use this hole without fully understanding possible issues.
359 */
360
Tom Herbertb73c3d02014-07-01 21:32:17 -0700361 __u32 sk_txhash;
Cong Wange0d10952013-08-01 11:10:25 +0800362#ifdef CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir06021292013-06-10 11:39:50 +0300363 unsigned int sk_napi_id;
Eliezer Tamirdafcc432013-06-14 16:33:57 +0300364 unsigned int sk_ll_usec;
Eliezer Tamir06021292013-06-10 11:39:50 +0300365#endif
Eric Dumazetb178bb32010-11-16 05:56:04 +0000366 atomic_t sk_drops;
367 int sk_rcvbuf;
368
369 struct sk_filter __rcu *sk_filter;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000370 struct socket_wq __rcu *sk_wq;
Eric Dumazetb178bb32010-11-16 05:56:04 +0000371
Eric Dumazetb178bb32010-11-16 05:56:04 +0000372#ifdef CONFIG_XFRM
373 struct xfrm_policy *sk_policy[2];
374#endif
375 unsigned long sk_flags;
Eric Dumazetdeaa5852012-06-24 20:22:49 +0000376 struct dst_entry *sk_rx_dst;
Cong Wang0e36cbb2013-01-22 21:09:51 +0000377 struct dst_entry __rcu *sk_dst_cache;
Eric Dumazetb178bb32010-11-16 05:56:04 +0000378 spinlock_t sk_dst_lock;
379 atomic_t sk_wmem_alloc;
380 atomic_t sk_omem_alloc;
381 int sk_sndbuf;
382 struct sk_buff_head sk_write_queue;
Vegard Nossuma98b65a2009-02-26 14:46:57 +0100383 kmemcheck_bitfield_begin(flags);
Eric Dumazet5fdb9972009-10-08 22:50:25 +0000384 unsigned int sk_shutdown : 2,
Tom Herbert28448b82014-05-23 08:47:19 -0700385 sk_no_check_tx : 1,
386 sk_no_check_rx : 1,
Eric Dumazet5fdb9972009-10-08 22:50:25 +0000387 sk_userlocks : 4,
388 sk_protocol : 8,
Hannes Frederic Sowabc8f79b2015-12-14 22:03:39 +0100389#define SK_PROTOCOL_MAX U8_MAX
Eric Dumazet5fdb9972009-10-08 22:50:25 +0000390 sk_type : 16;
Vegard Nossuma98b65a2009-02-26 14:46:57 +0100391 kmemcheck_bitfield_end(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392 int sk_wmem_queued;
Al Viro7d877f32005-10-21 03:20:43 -0400393 gfp_t sk_allocation;
Eric Dumazet95bd09e2013-08-27 05:46:32 -0700394 u32 sk_pacing_rate; /* bytes per second */
Eric Dumazet62748f32013-09-24 08:20:52 -0700395 u32 sk_max_pacing_rate;
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000396 netdev_features_t sk_route_caps;
397 netdev_features_t sk_route_nocaps;
Herbert Xubcd76112006-06-30 13:36:35 -0700398 int sk_gso_type;
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -0700399 unsigned int sk_gso_max_size;
Ben Hutchings14853482012-07-30 16:11:42 +0000400 u16 sk_gso_max_segs;
David S. Miller9932cf92006-03-24 15:12:37 -0800401 int sk_rcvlowat;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402 unsigned long sk_lingertime;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403 struct sk_buff_head sk_error_queue;
Arnaldo Carvalho de Melo476e19c2005-05-05 13:35:15 -0700404 struct proto *sk_prot_creator;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 rwlock_t sk_callback_lock;
406 int sk_err,
407 sk_err_soft;
Eric Dumazetbecb74f2015-03-19 19:04:21 -0700408 u32 sk_ack_backlog;
409 u32 sk_max_ack_backlog;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410 __u32 sk_priority;
Daniel Borkmann86f85152013-12-29 17:27:11 +0100411#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
Neil Horman5bc14212011-11-22 05:10:51 +0000412 __u32 sk_cgrp_prioidx;
413#endif
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 void *sk_protinfo;
419 struct timer_list sk_timer;
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700420 ktime_t sk_stamp;
Willem de Bruijnb9f40e22014-08-04 22:11:46 -0400421 u16 sk_tsflags;
Willem de Bruijn09c2d252014-08-04 22:11:47 -0400422 u32 sk_tskey;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423 struct socket *sk_socket;
424 void *sk_user_data;
Eric Dumazet5640f762012-09-23 23:04:42 +0000425 struct page_frag sk_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426 struct sk_buff *sk_send_head;
Pavel Emelyanovef64a542012-02-21 07:31:34 +0000427 __s32 sk_peek_off;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428 int sk_write_pending;
Alexey Dobriyand5f64232008-11-04 14:45:58 -0800429#ifdef CONFIG_SECURITY
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430 void *sk_security;
Alexey Dobriyand5f64232008-11-04 14:45:58 -0800431#endif
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800432 __u32 sk_mark;
Herbert Xuf8451722010-05-24 00:12:34 -0700433 u32 sk_classid;
Glauber Costae1aab162011-12-11 21:47:03 +0000434 struct cg_proto *sk_cgrp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435 void (*sk_state_change)(struct sock *sk);
David S. Miller676d2362014-04-11 16:15:36 -0400436 void (*sk_data_ready)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437 void (*sk_write_space)(struct sock *sk);
438 void (*sk_error_report)(struct sock *sk);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000439 int (*sk_backlog_rcv)(struct sock *sk,
440 struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441 void (*sk_destruct)(struct sock *sk);
442};
443
Pravin B Shelar559835ea2013-09-24 10:25:40 -0700444#define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
445
446#define rcu_dereference_sk_user_data(sk) rcu_dereference(__sk_user_data((sk)))
447#define rcu_assign_sk_user_data(sk, ptr) rcu_assign_pointer(__sk_user_data((sk)), ptr)
448
Pavel Emelyanov4a17fd52012-04-19 03:39:36 +0000449/*
450 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
451 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
452 * on a socket means that the socket will reuse everybody else's port
453 * without looking at the other's sk_reuse value.
454 */
455
456#define SK_NO_REUSE 0
457#define SK_CAN_REUSE 1
458#define SK_FORCE_REUSE 2
459
Pavel Emelyanovef64a542012-02-21 07:31:34 +0000460static inline int sk_peek_offset(struct sock *sk, int flags)
461{
462 if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0))
463 return sk->sk_peek_off;
464 else
465 return 0;
466}
467
468static inline void sk_peek_offset_bwd(struct sock *sk, int val)
469{
470 if (sk->sk_peek_off >= 0) {
471 if (sk->sk_peek_off >= val)
472 sk->sk_peek_off -= val;
473 else
474 sk->sk_peek_off = 0;
475 }
476}
477
478static inline void sk_peek_offset_fwd(struct sock *sk, int val)
479{
480 if (sk->sk_peek_off >= 0)
481 sk->sk_peek_off += val;
482}
483
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484/*
485 * Hashed lists helper routines
486 */
Li Zefanc4146642010-02-08 23:18:45 +0000487static inline struct sock *sk_entry(const struct hlist_node *node)
488{
489 return hlist_entry(node, struct sock, sk_node);
490}
491
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700492static inline struct sock *__sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493{
494 return hlist_entry(head->first, struct sock, sk_node);
495}
496
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700497static inline struct sock *sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498{
499 return hlist_empty(head) ? NULL : __sk_head(head);
500}
501
Eric Dumazet88ab1932008-11-16 19:39:21 -0800502static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
503{
504 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
505}
506
507static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
508{
509 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
510}
511
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700512static inline struct sock *sk_next(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513{
514 return sk->sk_node.next ?
515 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
516}
517
Eric Dumazet88ab1932008-11-16 19:39:21 -0800518static inline struct sock *sk_nulls_next(const struct sock *sk)
519{
520 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
521 hlist_nulls_entry(sk->sk_nulls_node.next,
522 struct sock, sk_nulls_node) :
523 NULL;
524}
525
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000526static inline bool sk_unhashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527{
528 return hlist_unhashed(&sk->sk_node);
529}
530
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000531static inline bool sk_hashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532{
Akinobu Mitada753be2006-04-28 15:21:23 -0700533 return !sk_unhashed(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534}
535
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000536static inline void sk_node_init(struct hlist_node *node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537{
538 node->pprev = NULL;
539}
540
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000541static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800542{
543 node->pprev = NULL;
544}
545
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000546static inline void __sk_del_node(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547{
548 __hlist_del(&sk->sk_node);
549}
550
stephen hemminger808f5112010-02-22 07:57:18 +0000551/* NB: equivalent to hlist_del_init_rcu */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000552static inline bool __sk_del_node_init(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553{
554 if (sk_hashed(sk)) {
555 __sk_del_node(sk);
556 sk_node_init(&sk->sk_node);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000557 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558 }
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000559 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560}
561
562/* Grab socket reference count. This operation is valid only
563 when sk is ALREADY grabbed f.e. it is found in hash table
564 or a list and the lookup is made under lock preventing hash table
565 modifications.
566 */
567
568static inline void sock_hold(struct sock *sk)
569{
570 atomic_inc(&sk->sk_refcnt);
571}
572
573/* Ungrab socket in the context, which assumes that socket refcnt
574 cannot hit zero, f.e. it is true in context of any socketcall.
575 */
576static inline void __sock_put(struct sock *sk)
577{
578 atomic_dec(&sk->sk_refcnt);
579}
580
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000581static inline bool sk_del_node_init(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582{
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000583 bool rc = __sk_del_node_init(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584
585 if (rc) {
586 /* paranoid for a while -acme */
587 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
588 __sock_put(sk);
589 }
590 return rc;
591}
stephen hemminger808f5112010-02-22 07:57:18 +0000592#define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000594static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700595{
596 if (sk_hashed(sk)) {
Eric Dumazet88ab1932008-11-16 19:39:21 -0800597 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000598 return true;
Eric Dumazet271b72c2008-10-29 02:11:14 -0700599 }
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000600 return false;
Eric Dumazet271b72c2008-10-29 02:11:14 -0700601}
602
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000603static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700604{
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000605 bool rc = __sk_nulls_del_node_init_rcu(sk);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700606
607 if (rc) {
608 /* paranoid for a while -acme */
609 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
610 __sock_put(sk);
611 }
612 return rc;
613}
614
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000615static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616{
617 hlist_add_head(&sk->sk_node, list);
618}
619
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000620static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621{
622 sock_hold(sk);
623 __sk_add_node(sk, list);
624}
625
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000626static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
stephen hemminger808f5112010-02-22 07:57:18 +0000627{
628 sock_hold(sk);
629 hlist_add_head_rcu(&sk->sk_node, list);
630}
631
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000632static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700633{
Eric Dumazet88ab1932008-11-16 19:39:21 -0800634 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700635}
636
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000637static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700638{
639 sock_hold(sk);
Eric Dumazet88ab1932008-11-16 19:39:21 -0800640 __sk_nulls_add_node_rcu(sk, list);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700641}
642
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000643static inline void __sk_del_bind_node(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644{
645 __hlist_del(&sk->sk_bind_node);
646}
647
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000648static inline void sk_add_bind_node(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649 struct hlist_head *list)
650{
651 hlist_add_head(&sk->sk_bind_node, list);
652}
653
Sasha Levinb67bfe02013-02-27 17:06:00 -0800654#define sk_for_each(__sk, list) \
655 hlist_for_each_entry(__sk, list, sk_node)
656#define sk_for_each_rcu(__sk, list) \
657 hlist_for_each_entry_rcu(__sk, list, sk_node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800658#define sk_nulls_for_each(__sk, node, list) \
659 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
660#define sk_nulls_for_each_rcu(__sk, node, list) \
661 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
Sasha Levinb67bfe02013-02-27 17:06:00 -0800662#define sk_for_each_from(__sk) \
663 hlist_for_each_entry_from(__sk, sk_node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800664#define sk_nulls_for_each_from(__sk, node) \
665 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
666 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
Sasha Levinb67bfe02013-02-27 17:06:00 -0800667#define sk_for_each_safe(__sk, tmp, list) \
668 hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
669#define sk_for_each_bound(__sk, list) \
670 hlist_for_each_entry(__sk, list, sk_bind_node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671
David Held2dc41cf2014-07-15 23:28:32 -0400672/**
673 * sk_nulls_for_each_entry_offset - iterate over a list at a given struct offset
674 * @tpos: the type * to use as a loop cursor.
675 * @pos: the &struct hlist_node to use as a loop cursor.
676 * @head: the head for your list.
677 * @offset: offset of hlist_node within the struct.
678 *
679 */
680#define sk_nulls_for_each_entry_offset(tpos, pos, head, offset) \
681 for (pos = (head)->first; \
682 (!is_a_nulls(pos)) && \
683 ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;}); \
684 pos = pos->next)
685
Eric W. Biedermanc336d142012-05-24 17:56:43 -0600686static inline struct user_namespace *sk_user_ns(struct sock *sk)
687{
688 /* Careful only use this in a context where these parameters
689 * can not change and must all be valid, such as recvmsg from
690 * userspace.
691 */
692 return sk->sk_socket->file->f_cred->user_ns;
693}
694
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695/* Sock flags */
696enum sock_flags {
697 SOCK_DEAD,
698 SOCK_DONE,
699 SOCK_URGINLINE,
700 SOCK_KEEPOPEN,
701 SOCK_LINGER,
702 SOCK_DESTROY,
703 SOCK_BROADCAST,
704 SOCK_TIMESTAMP,
705 SOCK_ZAPPED,
706 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
707 SOCK_DBG, /* %SO_DEBUG setting */
708 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700709 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
711 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
Mel Gorman7cb02402012-07-31 16:44:16 -0700712 SOCK_MEMALLOC, /* VM depends on this socket for swapping */
Patrick Ohly20d49472009-02-12 05:03:38 +0000713 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
Eric Dumazetbcdce712009-10-06 17:28:29 -0700714 SOCK_FASYNC, /* fasync() active */
Neil Horman3b885782009-10-12 13:26:31 -0700715 SOCK_RXQ_OVFL,
Shirley Ma1cdebb42011-07-06 12:17:30 +0000716 SOCK_ZEROCOPY, /* buffers from userspace */
Johannes Berg6e3e9392011-11-09 10:15:42 +0100717 SOCK_WIFI_STATUS, /* push wifi status to userspace */
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000718 SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
719 * Will use last 4 bytes of packet sent from
720 * user-space instead.
721 */
Vincent Bernatd59577b2013-01-16 22:55:49 +0100722 SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */
Keller, Jacob E7d4c04f2013-03-28 11:19:25 +0000723 SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724};
725
Marcelo Ricardo Leitner2e2b9372015-12-04 15:14:04 -0200726#define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
727
Ralf Baechle53b924b2005-08-23 10:11:30 -0700728static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
729{
730 nsk->sk_flags = osk->sk_flags;
731}
732
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
734{
735 __set_bit(flag, &sk->sk_flags);
736}
737
738static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
739{
740 __clear_bit(flag, &sk->sk_flags);
741}
742
Eric Dumazet1b23a5d2012-05-16 05:57:07 +0000743static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744{
745 return test_bit(flag, &sk->sk_flags);
746}
747
Mel Gormanc93bdd02012-07-31 16:44:19 -0700748#ifdef CONFIG_NET
749extern struct static_key memalloc_socks;
750static inline int sk_memalloc_socks(void)
751{
752 return static_key_false(&memalloc_socks);
753}
754#else
755
756static inline int sk_memalloc_socks(void)
757{
758 return 0;
759}
760
761#endif
762
Mel Gorman99a1dec2012-07-31 16:44:14 -0700763static inline gfp_t sk_gfp_atomic(struct sock *sk, gfp_t gfp_mask)
764{
Mel Gorman7cb02402012-07-31 16:44:16 -0700765 return GFP_ATOMIC | (sk->sk_allocation & __GFP_MEMALLOC);
Mel Gorman99a1dec2012-07-31 16:44:14 -0700766}
767
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768static inline void sk_acceptq_removed(struct sock *sk)
769{
770 sk->sk_ack_backlog--;
771}
772
773static inline void sk_acceptq_added(struct sock *sk)
774{
775 sk->sk_ack_backlog++;
776}
777
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000778static inline bool sk_acceptq_is_full(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779{
David S. Miller64a14652007-03-06 11:21:05 -0800780 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781}
782
783/*
784 * Compute minimal free write space needed to queue new packets.
785 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000786static inline int sk_stream_min_wspace(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787{
Eric Dumazet8df09ea2007-12-21 03:07:41 -0800788 return sk->sk_wmem_queued >> 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789}
790
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000791static inline int sk_stream_wspace(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792{
793 return sk->sk_sndbuf - sk->sk_wmem_queued;
794}
795
Joe Perches69336bd2013-09-22 10:32:26 -0700796void sk_stream_write_space(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797
Zhu Yi8eae9392010-03-04 18:01:40 +0000798/* OOB backlog add */
Zhu Yia3a858f2010-03-04 18:01:47 +0000799static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800800{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000801 /* dont let skb dst not refcounted, we are going to leave rcu lock */
Eric Dumazet2dea6352015-12-14 14:08:53 -0800802 skb_dst_force_safe(skb);
Eric Dumazet7fee2262010-05-11 23:19:48 +0000803
804 if (!sk->sk_backlog.tail)
805 sk->sk_backlog.head = skb;
806 else
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800807 sk->sk_backlog.tail->next = skb;
Eric Dumazet7fee2262010-05-11 23:19:48 +0000808
809 sk->sk_backlog.tail = skb;
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800810 skb->next = NULL;
811}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812
Eric Dumazetc3774112010-04-27 15:13:20 -0700813/*
814 * Take into account size of receive queue and backlog queue
Eric Dumazet0fd7bac2011-12-21 07:11:44 +0000815 * Do not take into account this skb truesize,
816 * to allow even a single big packet to come.
Eric Dumazetc3774112010-04-27 15:13:20 -0700817 */
Sorin Dumitru274f4822014-07-22 21:16:51 +0300818static inline bool sk_rcvqueues_full(const struct sock *sk, unsigned int limit)
Eric Dumazetc3774112010-04-27 15:13:20 -0700819{
820 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
821
Eric Dumazetf545a382012-04-22 23:34:26 +0000822 return qsize > limit;
Eric Dumazetc3774112010-04-27 15:13:20 -0700823}
824
Zhu Yi8eae9392010-03-04 18:01:40 +0000825/* The per-socket spinlock must be held here. */
Eric Dumazetf545a382012-04-22 23:34:26 +0000826static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
827 unsigned int limit)
Zhu Yi8eae9392010-03-04 18:01:40 +0000828{
Sorin Dumitru274f4822014-07-22 21:16:51 +0300829 if (sk_rcvqueues_full(sk, limit))
Zhu Yi8eae9392010-03-04 18:01:40 +0000830 return -ENOBUFS;
831
Eric Dumazetfac948a2015-09-29 18:52:25 -0700832 /*
833 * If the skb was allocated from pfmemalloc reserves, only
834 * allow SOCK_MEMALLOC sockets to use it as this socket is
835 * helping free memory
836 */
837 if (skb_pfmemalloc(skb) && !sock_flag(sk, SOCK_MEMALLOC))
838 return -ENOMEM;
839
Zhu Yia3a858f2010-03-04 18:01:47 +0000840 __sk_add_backlog(sk, skb);
Zhu Yi8eae9392010-03-04 18:01:40 +0000841 sk->sk_backlog.len += skb->truesize;
842 return 0;
843}
844
Joe Perches69336bd2013-09-22 10:32:26 -0700845int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
Mel Gormanb4b9e352012-07-31 16:44:26 -0700846
Peter Zijlstrac57943a2008-10-07 14:18:42 -0700847static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
848{
Mel Gormanb4b9e352012-07-31 16:44:26 -0700849 if (sk_memalloc_socks() && skb_pfmemalloc(skb))
850 return __sk_backlog_rcv(sk, skb);
851
Peter Zijlstrac57943a2008-10-07 14:18:42 -0700852 return sk->sk_backlog_rcv(sk, skb);
853}
854
Eric Dumazet2c8c56e2014-11-11 05:54:28 -0800855static inline void sk_incoming_cpu_update(struct sock *sk)
856{
857 sk->sk_incoming_cpu = raw_smp_processor_id();
858}
859
Tom Herbertfe477552013-12-22 18:54:31 +0800860static inline void sock_rps_record_flow_hash(__u32 hash)
David S. Millerc58dc012010-04-27 15:05:31 -0700861{
862#ifdef CONFIG_RPS
863 struct rps_sock_flow_table *sock_flow_table;
864
865 rcu_read_lock();
866 sock_flow_table = rcu_dereference(rps_sock_flow_table);
Tom Herbertfe477552013-12-22 18:54:31 +0800867 rps_record_sock_flow(sock_flow_table, hash);
David S. Millerc58dc012010-04-27 15:05:31 -0700868 rcu_read_unlock();
869#endif
870}
871
Tom Herbertfe477552013-12-22 18:54:31 +0800872static inline void sock_rps_record_flow(const struct sock *sk)
873{
Zhi Yong Wuc9d8ca02014-01-01 04:31:01 +0800874#ifdef CONFIG_RPS
Tom Herbertfe477552013-12-22 18:54:31 +0800875 sock_rps_record_flow_hash(sk->sk_rxhash);
Zhi Yong Wuc9d8ca02014-01-01 04:31:01 +0800876#endif
Tom Herbertfe477552013-12-22 18:54:31 +0800877}
878
Tom Herbertbdeab992011-08-14 19:45:55 +0000879static inline void sock_rps_save_rxhash(struct sock *sk,
880 const struct sk_buff *skb)
David S. Millerc58dc012010-04-27 15:05:31 -0700881{
882#ifdef CONFIG_RPS
Eric Dumazet567e4b72015-02-06 12:59:01 -0800883 if (unlikely(sk->sk_rxhash != skb->hash))
Tom Herbert61b905d2014-03-24 15:34:47 -0700884 sk->sk_rxhash = skb->hash;
David S. Millerc58dc012010-04-27 15:05:31 -0700885#endif
886}
887
Tom Herbertbdeab992011-08-14 19:45:55 +0000888static inline void sock_rps_reset_rxhash(struct sock *sk)
889{
890#ifdef CONFIG_RPS
Tom Herbertbdeab992011-08-14 19:45:55 +0000891 sk->sk_rxhash = 0;
892#endif
893}
894
Stephen Hemmingercfcabdc2007-10-09 01:59:42 -0700895#define sk_wait_event(__sk, __timeo, __condition) \
896 ({ int __rc; \
897 release_sock(__sk); \
898 __rc = __condition; \
899 if (!__rc) { \
900 *(__timeo) = schedule_timeout(*(__timeo)); \
901 } \
Peter Zijlstra26cabd32014-09-24 10:18:54 +0200902 sched_annotate_sleep(); \
Stephen Hemmingercfcabdc2007-10-09 01:59:42 -0700903 lock_sock(__sk); \
904 __rc = __condition; \
905 __rc; \
906 })
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907
Joe Perches69336bd2013-09-22 10:32:26 -0700908int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
909int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
910void sk_stream_wait_close(struct sock *sk, long timeo_p);
911int sk_stream_error(struct sock *sk, int flags, int err);
912void sk_stream_kill_queues(struct sock *sk);
913void sk_set_memalloc(struct sock *sk);
914void sk_clear_memalloc(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915
Joe Perches69336bd2013-09-22 10:32:26 -0700916int sk_wait_data(struct sock *sk, long *timeo);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700917
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700918struct request_sock_ops;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800919struct timewait_sock_ops;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -0800920struct inet_hashinfo;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -0700921struct raw_hashinfo;
Paul Gortmakerde477252011-05-26 13:46:22 -0400922struct module;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700923
Eric Dumazetf77d6022013-05-09 10:28:16 +0000924/*
925 * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
926 * un-modified. Special care is taken when initializing object to zero.
927 */
928static inline void sk_prot_clear_nulls(struct sock *sk, int size)
929{
930 if (offsetof(struct sock, sk_node.next) != 0)
931 memset(sk, 0, offsetof(struct sock, sk_node.next));
932 memset(&sk->sk_node.pprev, 0,
933 size - offsetof(struct sock, sk_node.pprev));
934}
935
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936/* Networking protocol blocks we attach to sockets.
937 * socket layer -> transport layer interface
938 * transport -> network interface is defined by struct inet_proto
939 */
940struct proto {
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000941 void (*close)(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942 long timeout);
943 int (*connect)(struct sock *sk,
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000944 struct sockaddr *uaddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700945 int addr_len);
946 int (*disconnect)(struct sock *sk, int flags);
947
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000948 struct sock * (*accept)(struct sock *sk, int flags, int *err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700949
950 int (*ioctl)(struct sock *sk, int cmd,
951 unsigned long arg);
952 int (*init)(struct sock *sk);
Brian Haley7d06b2e2008-06-14 17:04:49 -0700953 void (*destroy)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700954 void (*shutdown)(struct sock *sk, int how);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000955 int (*setsockopt)(struct sock *sk, int level,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700956 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700957 unsigned int optlen);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000958 int (*getsockopt)(struct sock *sk, int level,
959 int optname, char __user *optval,
960 int __user *option);
Alexey Dobriyanaf01d532008-08-28 02:53:51 -0700961#ifdef CONFIG_COMPAT
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800962 int (*compat_setsockopt)(struct sock *sk,
963 int level,
964 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700965 unsigned int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800966 int (*compat_getsockopt)(struct sock *sk,
967 int level,
968 int optname, char __user *optval,
969 int __user *option);
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000970 int (*compat_ioctl)(struct sock *sk,
971 unsigned int cmd, unsigned long arg);
Alexey Dobriyanaf01d532008-08-28 02:53:51 -0700972#endif
Ying Xue1b784142015-03-02 15:37:48 +0800973 int (*sendmsg)(struct sock *sk, struct msghdr *msg,
974 size_t len);
975 int (*recvmsg)(struct sock *sk, struct msghdr *msg,
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000976 size_t len, int noblock, int flags,
977 int *addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700978 int (*sendpage)(struct sock *sk, struct page *page,
979 int offset, size_t size, int flags);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000980 int (*bind)(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981 struct sockaddr *uaddr, int addr_len);
982
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000983 int (*backlog_rcv) (struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984 struct sk_buff *skb);
985
Eric Dumazet46d3cea2012-07-11 05:50:31 +0000986 void (*release_cb)(struct sock *sk);
987
Linus Torvalds1da177e2005-04-16 15:20:36 -0700988 /* Keeping track of sk's, looking them up, and port selection methods. */
989 void (*hash)(struct sock *sk);
990 void (*unhash)(struct sock *sk);
Eric Dumazet719f8352010-09-08 05:08:44 +0000991 void (*rehash)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992 int (*get_port)(struct sock *sk, unsigned short snum);
Octavian Purdilafcbdf092010-12-16 14:26:56 -0800993 void (*clear_sk)(struct sock *sk, int size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994
Eric Dumazet286ab3d2007-11-05 23:38:39 -0800995 /* Keeping track of sockets in use */
Eric Dumazet65f76512008-01-03 20:46:48 -0800996#ifdef CONFIG_PROC_FS
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -0700997 unsigned int inuse_idx;
Eric Dumazet65f76512008-01-03 20:46:48 -0800998#endif
Arnaldo Carvalho de Meloebb53d72007-11-21 22:08:50 +0800999
Eric Dumazetc9bee3b72013-07-22 20:27:07 -07001000 bool (*stream_memory_free)(const struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001 /* Memory pressure */
Pavel Emelyanov5c52ba12008-07-16 20:28:10 -07001002 void (*enter_memory_pressure)(struct sock *sk);
Eric Dumazet8d987e52010-11-09 23:24:26 +00001003 atomic_long_t *memory_allocated; /* Current allocated memory. */
Eric Dumazet17483762008-11-25 21:16:35 -08001004 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005 /*
1006 * Pressure flag: try to collapse.
1007 * Technical note: it is used by multiple contexts non atomically.
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001008 * All the __sk_mem_schedule() is of this nature: accounting
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009 * is strict, actions are advisory and have some latency.
1010 */
1011 int *memory_pressure;
Eric Dumazet8d987e52010-11-09 23:24:26 +00001012 long *sysctl_mem;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013 int *sysctl_wmem;
1014 int *sysctl_rmem;
1015 int max_header;
Changli Gao7ba42912010-07-10 20:41:55 +00001016 bool no_autobind;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017
Eric Dumazet271b72c2008-10-29 02:11:14 -07001018 struct kmem_cache *slab;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019 unsigned int obj_size;
Eric Dumazet271b72c2008-10-29 02:11:14 -07001020 int slab_flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021
Eric Dumazetdd24c002008-11-25 21:17:14 -08001022 struct percpu_counter *orphan_count;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -07001023
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -07001024 struct request_sock_ops *rsk_prot;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -08001025 struct timewait_sock_ops *twsk_prot;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07001026
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -07001027 union {
1028 struct inet_hashinfo *hashinfo;
Eric Dumazet645ca702008-10-29 01:41:45 -07001029 struct udp_table *udp_table;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -07001030 struct raw_hashinfo *raw_hash;
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -07001031 } h;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -08001032
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033 struct module *owner;
1034
1035 char name[32];
1036
1037 struct list_head node;
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -07001038#ifdef SOCK_REFCNT_DEBUG
1039 atomic_t socks;
1040#endif
Andrew Mortonc255a452012-07-31 16:43:02 -07001041#ifdef CONFIG_MEMCG_KMEM
Glauber Costae1aab162011-12-11 21:47:03 +00001042 /*
1043 * cgroup specific init/deinit functions. Called once for all
1044 * protocols that implement it, from cgroups populate function.
1045 * This function has to setup any files the protocol want to
1046 * appear in the kmem cgroup filesystem.
1047 */
Glauber Costa1d62e432012-04-09 19:36:33 -03001048 int (*init_cgroup)(struct mem_cgroup *memcg,
Glauber Costae1aab162011-12-11 21:47:03 +00001049 struct cgroup_subsys *ss);
Glauber Costa1d62e432012-04-09 19:36:33 -03001050 void (*destroy_cgroup)(struct mem_cgroup *memcg);
Glauber Costae1aab162011-12-11 21:47:03 +00001051 struct cg_proto *(*proto_cgroup)(struct mem_cgroup *memcg);
1052#endif
Lorenzo Colitti034d6bc2015-12-16 12:30:03 +09001053 int (*diag_destroy)(struct sock *sk, int err);
Glauber Costae1aab162011-12-11 21:47:03 +00001054};
1055
Joe Perches69336bd2013-09-22 10:32:26 -07001056int proto_register(struct proto *prot, int alloc_slab);
1057void proto_unregister(struct proto *prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -07001059#ifdef SOCK_REFCNT_DEBUG
1060static inline void sk_refcnt_debug_inc(struct sock *sk)
1061{
1062 atomic_inc(&sk->sk_prot->socks);
1063}
1064
1065static inline void sk_refcnt_debug_dec(struct sock *sk)
1066{
1067 atomic_dec(&sk->sk_prot->socks);
1068 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
1069 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
1070}
1071
Ying Xuedec34fb2013-02-15 22:28:25 +00001072static inline void sk_refcnt_debug_release(const struct sock *sk)
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -07001073{
1074 if (atomic_read(&sk->sk_refcnt) != 1)
1075 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
1076 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
1077}
1078#else /* SOCK_REFCNT_DEBUG */
1079#define sk_refcnt_debug_inc(sk) do { } while (0)
1080#define sk_refcnt_debug_dec(sk) do { } while (0)
1081#define sk_refcnt_debug_release(sk) do { } while (0)
1082#endif /* SOCK_REFCNT_DEBUG */
1083
Andrew Mortonc255a452012-07-31 16:43:02 -07001084#if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_NET)
Ingo Molnarc5905af2012-02-24 08:31:31 +01001085extern struct static_key memcg_socket_limit_enabled;
Glauber Costae1aab162011-12-11 21:47:03 +00001086static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1087 struct cg_proto *cg_proto)
1088{
1089 return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg));
1090}
Ingo Molnarc5905af2012-02-24 08:31:31 +01001091#define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
Glauber Costae1aab162011-12-11 21:47:03 +00001092#else
1093#define mem_cgroup_sockets_enabled 0
1094static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1095 struct cg_proto *cg_proto)
1096{
1097 return NULL;
1098}
1099#endif
1100
Eric Dumazetc9bee3b72013-07-22 20:27:07 -07001101static inline bool sk_stream_memory_free(const struct sock *sk)
1102{
1103 if (sk->sk_wmem_queued >= sk->sk_sndbuf)
1104 return false;
1105
1106 return sk->sk_prot->stream_memory_free ?
1107 sk->sk_prot->stream_memory_free(sk) : true;
1108}
1109
Eric Dumazet64dc6132013-07-22 20:26:31 -07001110static inline bool sk_stream_is_writeable(const struct sock *sk)
1111{
Eric Dumazetc9bee3b72013-07-22 20:27:07 -07001112 return sk_stream_wspace(sk) >= sk_stream_min_wspace(sk) &&
1113 sk_stream_memory_free(sk);
Eric Dumazet64dc6132013-07-22 20:26:31 -07001114}
Glauber Costae1aab162011-12-11 21:47:03 +00001115
Eric Dumazetc9bee3b72013-07-22 20:27:07 -07001116
Glauber Costa180d8cd2011-12-11 21:47:02 +00001117static inline bool sk_has_memory_pressure(const struct sock *sk)
1118{
1119 return sk->sk_prot->memory_pressure != NULL;
1120}
1121
1122static inline bool sk_under_memory_pressure(const struct sock *sk)
1123{
1124 if (!sk->sk_prot->memory_pressure)
1125 return false;
Glauber Costae1aab162011-12-11 21:47:03 +00001126
1127 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
Eric W. Biederman2e685ca2013-10-19 16:26:19 -07001128 return !!sk->sk_cgrp->memory_pressure;
Glauber Costae1aab162011-12-11 21:47:03 +00001129
Christoph Paasch35b87f62013-10-23 12:49:21 -07001130 return !!*sk->sk_prot->memory_pressure;
Glauber Costa180d8cd2011-12-11 21:47:02 +00001131}
1132
1133static inline void sk_leave_memory_pressure(struct sock *sk)
1134{
1135 int *memory_pressure = sk->sk_prot->memory_pressure;
1136
Glauber Costae1aab162011-12-11 21:47:03 +00001137 if (!memory_pressure)
1138 return;
1139
1140 if (*memory_pressure)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001141 *memory_pressure = 0;
Glauber Costae1aab162011-12-11 21:47:03 +00001142
1143 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1144 struct cg_proto *cg_proto = sk->sk_cgrp;
1145 struct proto *prot = sk->sk_prot;
1146
1147 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
Eric W. Biederman7f2cbdc2013-12-04 20:12:04 -08001148 cg_proto->memory_pressure = 0;
Glauber Costae1aab162011-12-11 21:47:03 +00001149 }
1150
Glauber Costa180d8cd2011-12-11 21:47:02 +00001151}
1152
1153static inline void sk_enter_memory_pressure(struct sock *sk)
1154{
Glauber Costae1aab162011-12-11 21:47:03 +00001155 if (!sk->sk_prot->enter_memory_pressure)
1156 return;
1157
1158 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1159 struct cg_proto *cg_proto = sk->sk_cgrp;
1160 struct proto *prot = sk->sk_prot;
1161
1162 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
Eric W. Biederman7f2cbdc2013-12-04 20:12:04 -08001163 cg_proto->memory_pressure = 1;
Glauber Costae1aab162011-12-11 21:47:03 +00001164 }
1165
1166 sk->sk_prot->enter_memory_pressure(sk);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001167}
1168
1169static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1170{
1171 long *prot = sk->sk_prot->sysctl_mem;
Glauber Costae1aab162011-12-11 21:47:03 +00001172 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1173 prot = sk->sk_cgrp->sysctl_mem;
Glauber Costa180d8cd2011-12-11 21:47:02 +00001174 return prot[index];
1175}
1176
Glauber Costae1aab162011-12-11 21:47:03 +00001177static inline void memcg_memory_allocated_add(struct cg_proto *prot,
1178 unsigned long amt,
1179 int *parent_status)
1180{
Johannes Weiner3e32cb22014-12-10 15:42:31 -08001181 page_counter_charge(&prot->memory_allocated, amt);
Glauber Costae1aab162011-12-11 21:47:03 +00001182
Johannes Weiner3e32cb22014-12-10 15:42:31 -08001183 if (page_counter_read(&prot->memory_allocated) >
1184 prot->memory_allocated.limit)
Glauber Costae1aab162011-12-11 21:47:03 +00001185 *parent_status = OVER_LIMIT;
1186}
1187
1188static inline void memcg_memory_allocated_sub(struct cg_proto *prot,
1189 unsigned long amt)
1190{
Johannes Weiner3e32cb22014-12-10 15:42:31 -08001191 page_counter_uncharge(&prot->memory_allocated, amt);
Glauber Costae1aab162011-12-11 21:47:03 +00001192}
1193
Glauber Costa180d8cd2011-12-11 21:47:02 +00001194static inline long
1195sk_memory_allocated(const struct sock *sk)
1196{
1197 struct proto *prot = sk->sk_prot;
Johannes Weiner3e32cb22014-12-10 15:42:31 -08001198
Glauber Costae1aab162011-12-11 21:47:03 +00001199 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
Johannes Weiner3e32cb22014-12-10 15:42:31 -08001200 return page_counter_read(&sk->sk_cgrp->memory_allocated);
Glauber Costae1aab162011-12-11 21:47:03 +00001201
Glauber Costa180d8cd2011-12-11 21:47:02 +00001202 return atomic_long_read(prot->memory_allocated);
1203}
1204
1205static inline long
Glauber Costae1aab162011-12-11 21:47:03 +00001206sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001207{
1208 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001209
1210 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1211 memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status);
1212 /* update the root cgroup regardless */
1213 atomic_long_add_return(amt, prot->memory_allocated);
Johannes Weiner3e32cb22014-12-10 15:42:31 -08001214 return page_counter_read(&sk->sk_cgrp->memory_allocated);
Glauber Costae1aab162011-12-11 21:47:03 +00001215 }
1216
Glauber Costa180d8cd2011-12-11 21:47:02 +00001217 return atomic_long_add_return(amt, prot->memory_allocated);
1218}
1219
1220static inline void
Glauber Costa0e90b312012-01-20 04:57:16 +00001221sk_memory_allocated_sub(struct sock *sk, int amt)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001222{
1223 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001224
Glauber Costa0e90b312012-01-20 04:57:16 +00001225 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
Glauber Costae1aab162011-12-11 21:47:03 +00001226 memcg_memory_allocated_sub(sk->sk_cgrp, amt);
1227
Glauber Costa180d8cd2011-12-11 21:47:02 +00001228 atomic_long_sub(amt, prot->memory_allocated);
1229}
1230
1231static inline void sk_sockets_allocated_dec(struct sock *sk)
1232{
1233 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001234
1235 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1236 struct cg_proto *cg_proto = sk->sk_cgrp;
1237
1238 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
Eric W. Biederman2e685ca2013-10-19 16:26:19 -07001239 percpu_counter_dec(&cg_proto->sockets_allocated);
Glauber Costae1aab162011-12-11 21:47:03 +00001240 }
1241
Glauber Costa180d8cd2011-12-11 21:47:02 +00001242 percpu_counter_dec(prot->sockets_allocated);
1243}
1244
1245static inline void sk_sockets_allocated_inc(struct sock *sk)
1246{
1247 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001248
1249 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1250 struct cg_proto *cg_proto = sk->sk_cgrp;
1251
1252 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
Eric W. Biederman2e685ca2013-10-19 16:26:19 -07001253 percpu_counter_inc(&cg_proto->sockets_allocated);
Glauber Costae1aab162011-12-11 21:47:03 +00001254 }
1255
Glauber Costa180d8cd2011-12-11 21:47:02 +00001256 percpu_counter_inc(prot->sockets_allocated);
1257}
1258
1259static inline int
1260sk_sockets_allocated_read_positive(struct sock *sk)
1261{
1262 struct proto *prot = sk->sk_prot;
1263
Glauber Costae1aab162011-12-11 21:47:03 +00001264 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
Eric W. Biederman2e685ca2013-10-19 16:26:19 -07001265 return percpu_counter_read_positive(&sk->sk_cgrp->sockets_allocated);
Glauber Costae1aab162011-12-11 21:47:03 +00001266
Eric Dumazet518fbf92012-04-28 23:21:56 +00001267 return percpu_counter_read_positive(prot->sockets_allocated);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001268}
1269
1270static inline int
1271proto_sockets_allocated_sum_positive(struct proto *prot)
1272{
1273 return percpu_counter_sum_positive(prot->sockets_allocated);
1274}
1275
1276static inline long
1277proto_memory_allocated(struct proto *prot)
1278{
1279 return atomic_long_read(prot->memory_allocated);
1280}
1281
1282static inline bool
1283proto_memory_pressure(struct proto *prot)
1284{
1285 if (!prot->memory_pressure)
1286 return false;
1287 return !!*prot->memory_pressure;
1288}
1289
Eric Dumazet65f76512008-01-03 20:46:48 -08001290
1291#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292/* Called with local bh disabled */
Joe Perches69336bd2013-09-22 10:32:26 -07001293void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1294int sock_prot_inuse_get(struct net *net, struct proto *proto);
Eric Dumazet65f76512008-01-03 20:46:48 -08001295#else
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001296static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -07001297 int inc)
Eric Dumazet65f76512008-01-03 20:46:48 -08001298{
1299}
Eric Dumazet65f76512008-01-03 20:46:48 -08001300#endif
1301
Linus Torvalds1da177e2005-04-16 15:20:36 -07001302
Arnaldo Carvalho de Melo614c6cb2005-08-09 19:47:37 -07001303/* With per-bucket locks this operation is not-atomic, so that
1304 * this version is not worse.
1305 */
1306static inline void __sk_prot_rehash(struct sock *sk)
1307{
1308 sk->sk_prot->unhash(sk);
1309 sk->sk_prot->hash(sk);
1310}
1311
Octavian Purdilafcbdf092010-12-16 14:26:56 -08001312void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
1313
Linus Torvalds1da177e2005-04-16 15:20:36 -07001314/* About 10 seconds */
1315#define SOCK_DESTROY_TIME (10*HZ)
1316
1317/* Sockets 0-1023 can't be bound to unless you are superuser */
1318#define PROT_SOCK 1024
1319
1320#define SHUTDOWN_MASK 3
1321#define RCV_SHUTDOWN 1
1322#define SEND_SHUTDOWN 2
1323
1324#define SOCK_SNDBUF_LOCK 1
1325#define SOCK_RCVBUF_LOCK 2
1326#define SOCK_BINDADDR_LOCK 4
1327#define SOCK_BINDPORT_LOCK 8
1328
Linus Torvalds1da177e2005-04-16 15:20:36 -07001329struct socket_alloc {
1330 struct socket socket;
1331 struct inode vfs_inode;
1332};
1333
1334static inline struct socket *SOCKET_I(struct inode *inode)
1335{
1336 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1337}
1338
1339static inline struct inode *SOCK_INODE(struct socket *socket)
1340{
1341 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1342}
1343
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001344/*
1345 * Functions for memory accounting
1346 */
Joe Perches69336bd2013-09-22 10:32:26 -07001347int __sk_mem_schedule(struct sock *sk, int size, int kind);
1348void __sk_mem_reclaim(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001349
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001350#define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1351#define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1352#define SK_MEM_SEND 0
1353#define SK_MEM_RECV 1
Linus Torvalds1da177e2005-04-16 15:20:36 -07001354
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001355static inline int sk_mem_pages(int amt)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001357 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358}
1359
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001360static inline bool sk_has_account(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001361{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001362 /* return true if protocol supports memory accounting */
1363 return !!sk->sk_prot->memory_allocated;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364}
1365
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001366static inline bool sk_wmem_schedule(struct sock *sk, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001368 if (!sk_has_account(sk))
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001369 return true;
Herbert Xud80d99d2005-09-01 17:48:23 -07001370 return size <= sk->sk_forward_alloc ||
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001371 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1372}
1373
Mel Gormanc76562b2012-07-31 16:44:41 -07001374static inline bool
Chuck Lever35c448a2012-09-17 14:09:11 -07001375sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size)
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001376{
1377 if (!sk_has_account(sk))
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001378 return true;
Mel Gormanc76562b2012-07-31 16:44:41 -07001379 return size<= sk->sk_forward_alloc ||
1380 __sk_mem_schedule(sk, size, SK_MEM_RECV) ||
1381 skb_pfmemalloc(skb);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001382}
1383
1384static inline void sk_mem_reclaim(struct sock *sk)
1385{
1386 if (!sk_has_account(sk))
1387 return;
1388 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1389 __sk_mem_reclaim(sk);
1390}
1391
David S. Miller9993e7d2008-01-10 21:56:38 -08001392static inline void sk_mem_reclaim_partial(struct sock *sk)
1393{
1394 if (!sk_has_account(sk))
1395 return;
1396 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1397 __sk_mem_reclaim(sk);
1398}
1399
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001400static inline void sk_mem_charge(struct sock *sk, int size)
1401{
1402 if (!sk_has_account(sk))
1403 return;
1404 sk->sk_forward_alloc -= size;
1405}
1406
1407static inline void sk_mem_uncharge(struct sock *sk, int size)
1408{
1409 if (!sk_has_account(sk))
1410 return;
1411 sk->sk_forward_alloc += size;
1412}
1413
1414static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1415{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001416 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1417 sk->sk_wmem_queued -= skb->truesize;
1418 sk_mem_uncharge(sk, skb->truesize);
1419 __kfree_skb(skb);
Herbert Xud80d99d2005-09-01 17:48:23 -07001420}
1421
Linus Torvalds1da177e2005-04-16 15:20:36 -07001422/* Used by processes to "lock" a socket state, so that
1423 * interrupts and bottom half handlers won't change it
1424 * from under us. It essentially blocks any incoming
1425 * packets, so that we won't get any new data or any
1426 * packets that change the state of the socket.
1427 *
1428 * While locked, BH processing will add new packets to
1429 * the backlog queue. This queue is processed by the
1430 * owner of the socket lock right before it is released.
1431 *
1432 * Since ~2.3.5 it is also exclusive sleep lock serializing
1433 * accesses from user process context.
1434 */
John Heffnerd2e91172007-09-12 10:44:19 +02001435#define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001436
Eric Dumazetc3f9b012014-03-10 09:50:11 -07001437static inline void sock_release_ownership(struct sock *sk)
1438{
1439 sk->sk_lock.owned = 0;
1440}
1441
Peter Zijlstraed075362006-12-06 20:35:24 -08001442/*
1443 * Macro so as to not evaluate some arguments when
1444 * lockdep is not enabled.
1445 *
1446 * Mark both the sk_lock and the sk_lock.slock as a
1447 * per-address-family lock class.
1448 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001449#define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
Peter Zijlstraed075362006-12-06 20:35:24 -08001450do { \
Ingo Molnare8f6fbf2008-11-12 01:38:36 +00001451 sk->sk_lock.owned = 0; \
Peter Zijlstraed075362006-12-06 20:35:24 -08001452 init_waitqueue_head(&sk->sk_lock.wq); \
1453 spin_lock_init(&(sk)->sk_lock.slock); \
1454 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1455 sizeof((sk)->sk_lock)); \
1456 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001457 (skey), (sname)); \
Peter Zijlstraed075362006-12-06 20:35:24 -08001458 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1459} while (0)
1460
Joe Perches69336bd2013-09-22 10:32:26 -07001461void lock_sock_nested(struct sock *sk, int subclass);
Peter Zijlstrafcc70d52006-11-08 22:44:35 -08001462
1463static inline void lock_sock(struct sock *sk)
1464{
1465 lock_sock_nested(sk, 0);
1466}
1467
Joe Perches69336bd2013-09-22 10:32:26 -07001468void release_sock(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001469
1470/* BH context may only use the following locking interface. */
1471#define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
Ingo Molnarc6366182006-07-03 00:25:13 -07001472#define bh_lock_sock_nested(__sk) \
1473 spin_lock_nested(&((__sk)->sk_lock.slock), \
1474 SINGLE_DEPTH_NESTING)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001475#define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1476
Joe Perches69336bd2013-09-22 10:32:26 -07001477bool lock_sock_fast(struct sock *sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001478/**
1479 * unlock_sock_fast - complement of lock_sock_fast
1480 * @sk: socket
1481 * @slow: slow mode
1482 *
1483 * fast unlock socket for user context.
1484 * If slow mode is on, we call regular release_sock()
1485 */
1486static inline void unlock_sock_fast(struct sock *sk, bool slow)
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001487{
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001488 if (slow)
1489 release_sock(sk);
1490 else
1491 spin_unlock_bh(&sk->sk_lock.slock);
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001492}
1493
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001494
Joe Perches69336bd2013-09-22 10:32:26 -07001495struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
1496 struct proto *prot);
1497void sk_free(struct sock *sk);
1498void sk_release_kernel(struct sock *sk);
1499struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001500
Joe Perches69336bd2013-09-22 10:32:26 -07001501struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
1502 gfp_t priority);
Joe Perches69336bd2013-09-22 10:32:26 -07001503void sock_wfree(struct sk_buff *skb);
1504void skb_orphan_partial(struct sk_buff *skb);
1505void sock_rfree(struct sk_buff *skb);
Alexander Duyck62bccb82014-09-04 13:31:35 -04001506void sock_efree(struct sk_buff *skb);
Alexander Duyck82eabd92014-09-04 13:32:11 -04001507#ifdef CONFIG_INET
Joe Perches69336bd2013-09-22 10:32:26 -07001508void sock_edemux(struct sk_buff *skb);
Alexander Duyck82eabd92014-09-04 13:32:11 -04001509#else
1510#define sock_edemux(skb) sock_efree(skb)
1511#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001512
Joe Perches69336bd2013-09-22 10:32:26 -07001513int sock_setsockopt(struct socket *sock, int level, int op,
1514 char __user *optval, unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001515
Joe Perches69336bd2013-09-22 10:32:26 -07001516int sock_getsockopt(struct socket *sock, int level, int op,
1517 char __user *optval, int __user *optlen);
1518struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1519 int noblock, int *errcode);
1520struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1521 unsigned long data_len, int noblock,
1522 int *errcode, int max_page_order);
1523void *sock_kmalloc(struct sock *sk, int size, gfp_t priority);
1524void sock_kfree_s(struct sock *sk, void *mem, int size);
Daniel Borkmann79e88652014-11-19 17:13:11 +01001525void sock_kzfree_s(struct sock *sk, void *mem, int size);
Joe Perches69336bd2013-09-22 10:32:26 -07001526void sk_send_sigurg(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001527
1528/*
1529 * Functions to fill in entries in struct proto_ops when a protocol
1530 * does not implement a particular function.
1531 */
Joe Perches69336bd2013-09-22 10:32:26 -07001532int sock_no_bind(struct socket *, struct sockaddr *, int);
1533int sock_no_connect(struct socket *, struct sockaddr *, int, int);
1534int sock_no_socketpair(struct socket *, struct socket *);
1535int sock_no_accept(struct socket *, struct socket *, int);
1536int sock_no_getname(struct socket *, struct sockaddr *, int *, int);
1537unsigned int sock_no_poll(struct file *, struct socket *,
1538 struct poll_table_struct *);
1539int sock_no_ioctl(struct socket *, unsigned int, unsigned long);
1540int sock_no_listen(struct socket *, int);
1541int sock_no_shutdown(struct socket *, int);
1542int sock_no_getsockopt(struct socket *, int , int, char __user *, int __user *);
1543int sock_no_setsockopt(struct socket *, int, int, char __user *, unsigned int);
Ying Xue1b784142015-03-02 15:37:48 +08001544int sock_no_sendmsg(struct socket *, struct msghdr *, size_t);
1545int sock_no_recvmsg(struct socket *, struct msghdr *, size_t, int);
Joe Perches69336bd2013-09-22 10:32:26 -07001546int sock_no_mmap(struct file *file, struct socket *sock,
1547 struct vm_area_struct *vma);
1548ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset,
1549 size_t size, int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550
1551/*
1552 * Functions to fill in entries in struct proto_ops when a protocol
1553 * uses the inet style.
1554 */
Joe Perches69336bd2013-09-22 10:32:26 -07001555int sock_common_getsockopt(struct socket *sock, int level, int optname,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001556 char __user *optval, int __user *optlen);
Ying Xue1b784142015-03-02 15:37:48 +08001557int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
1558 int flags);
Joe Perches69336bd2013-09-22 10:32:26 -07001559int sock_common_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07001560 char __user *optval, unsigned int optlen);
Joe Perches69336bd2013-09-22 10:32:26 -07001561int compat_sock_common_getsockopt(struct socket *sock, int level,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001562 int optname, char __user *optval, int __user *optlen);
Joe Perches69336bd2013-09-22 10:32:26 -07001563int compat_sock_common_setsockopt(struct socket *sock, int level,
David S. Millerb7058842009-09-30 16:12:20 -07001564 int optname, char __user *optval, unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565
Joe Perches69336bd2013-09-22 10:32:26 -07001566void sk_common_release(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567
1568/*
1569 * Default socket callbacks and setup code
1570 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001571
Linus Torvalds1da177e2005-04-16 15:20:36 -07001572/* Initialise core socket variables */
Joe Perches69336bd2013-09-22 10:32:26 -07001573void sock_init_data(struct socket *sock, struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574
Linus Torvalds1da177e2005-04-16 15:20:36 -07001575/*
1576 * Socket reference counting postulates.
1577 *
1578 * * Each user of socket SHOULD hold a reference count.
1579 * * Each access point to socket (an hash table bucket, reference from a list,
1580 * running timer, skb in flight MUST hold a reference count.
1581 * * When reference count hits 0, it means it will never increase back.
1582 * * When reference count hits 0, it means that no references from
1583 * outside exist to this socket and current process on current CPU
1584 * is last user and may/should destroy this socket.
1585 * * sk_free is called from any context: process, BH, IRQ. When
1586 * it is called, socket has no references from outside -> sk_free
1587 * may release descendant resources allocated by the socket, but
1588 * to the time when it is called, socket is NOT referenced by any
1589 * hash tables, lists etc.
1590 * * Packets, delivered from outside (from network or from another process)
1591 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1592 * when they sit in queue. Otherwise, packets will leak to hole, when
1593 * socket is looked up by one cpu and unhasing is made by another CPU.
1594 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1595 * (leak to backlog). Packet socket does all the processing inside
1596 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1597 * use separate SMP lock, so that they are prone too.
1598 */
1599
1600/* Ungrab socket and destroy it, if it was the last reference. */
1601static inline void sock_put(struct sock *sk)
1602{
1603 if (atomic_dec_and_test(&sk->sk_refcnt))
1604 sk_free(sk);
1605}
Eric Dumazet05dbc7b2013-10-03 00:22:02 -07001606/* Generic version of sock_put(), dealing with all sockets
Eric Dumazet41b822c2015-03-12 16:44:08 -07001607 * (TCP_TIMEWAIT, TCP_NEW_SYN_RECV, ESTABLISHED...)
Eric Dumazet05dbc7b2013-10-03 00:22:02 -07001608 */
1609void sock_gen_put(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001610
Joe Perches69336bd2013-09-22 10:32:26 -07001611int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested);
Arnaldo Carvalho de Melo25995ff2005-12-27 02:42:22 -02001612
Krishna Kumare022f0b2009-10-19 23:46:20 +00001613static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1614{
1615 sk->sk_tx_queue_mapping = tx_queue;
1616}
1617
1618static inline void sk_tx_queue_clear(struct sock *sk)
1619{
1620 sk->sk_tx_queue_mapping = -1;
1621}
1622
1623static inline int sk_tx_queue_get(const struct sock *sk)
1624{
Tom Herbertb0f77d02010-07-14 20:50:29 -07001625 return sk ? sk->sk_tx_queue_mapping : -1;
Krishna Kumare022f0b2009-10-19 23:46:20 +00001626}
1627
David S. Miller972692e2008-06-17 22:41:38 -07001628static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1629{
Krishna Kumare022f0b2009-10-19 23:46:20 +00001630 sk_tx_queue_clear(sk);
David S. Miller972692e2008-06-17 22:41:38 -07001631 sk->sk_socket = sock;
1632}
1633
Eric Dumazetaa395142010-04-20 13:03:51 +00001634static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1635{
Eric Dumazeteaefd112011-02-18 03:26:36 +00001636 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1637 return &rcu_dereference_raw(sk->sk_wq)->wait;
Eric Dumazetaa395142010-04-20 13:03:51 +00001638}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001639/* Detach socket from process context.
1640 * Announce socket dead, detach it from wait queue and inode.
1641 * Note that parent inode held reference count on this struct sock,
1642 * we do not release it in this function, because protocol
1643 * probably wants some additional cleanups or even continuing
1644 * to work with this socket (TCP).
1645 */
1646static inline void sock_orphan(struct sock *sk)
1647{
1648 write_lock_bh(&sk->sk_callback_lock);
1649 sock_set_flag(sk, SOCK_DEAD);
David S. Miller972692e2008-06-17 22:41:38 -07001650 sk_set_socket(sk, NULL);
Eric Dumazet43815482010-04-29 11:01:49 +00001651 sk->sk_wq = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001652 write_unlock_bh(&sk->sk_callback_lock);
1653}
1654
1655static inline void sock_graft(struct sock *sk, struct socket *parent)
1656{
1657 write_lock_bh(&sk->sk_callback_lock);
Eric Dumazeteaefd112011-02-18 03:26:36 +00001658 sk->sk_wq = parent->wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001659 parent->sk = sk;
David S. Miller972692e2008-06-17 22:41:38 -07001660 sk_set_socket(sk, parent);
Venkat Yekkirala4237c752006-07-24 23:32:50 -07001661 security_sock_graft(sk, parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662 write_unlock_bh(&sk->sk_callback_lock);
1663}
1664
Joe Perches69336bd2013-09-22 10:32:26 -07001665kuid_t sock_i_uid(struct sock *sk);
1666unsigned long sock_i_ino(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001667
1668static inline struct dst_entry *
1669__sk_dst_get(struct sock *sk)
1670{
Michal Hockod8bf4ca2011-07-08 14:39:41 +02001671 return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
Eric Dumazetf68c2242010-04-22 16:06:59 -07001672 lockdep_is_held(&sk->sk_lock.slock));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001673}
1674
1675static inline struct dst_entry *
1676sk_dst_get(struct sock *sk)
1677{
1678 struct dst_entry *dst;
1679
Eric Dumazetb6c67122010-04-08 23:03:29 +00001680 rcu_read_lock();
1681 dst = rcu_dereference(sk->sk_dst_cache);
Eric Dumazetf8864972014-06-24 10:05:11 -07001682 if (dst && !atomic_inc_not_zero(&dst->__refcnt))
1683 dst = NULL;
Eric Dumazetb6c67122010-04-08 23:03:29 +00001684 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001685 return dst;
1686}
1687
Eric Dumazetb6c67122010-04-08 23:03:29 +00001688static inline void dst_negative_advice(struct sock *sk)
1689{
1690 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1691
1692 if (dst && dst->ops->negative_advice) {
1693 ndst = dst->ops->negative_advice(dst);
1694
1695 if (ndst != dst) {
1696 rcu_assign_pointer(sk->sk_dst_cache, ndst);
ZHAO Gang0a6957e2013-10-22 16:23:38 +08001697 sk_tx_queue_clear(sk);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001698 }
1699 }
1700}
1701
Linus Torvalds1da177e2005-04-16 15:20:36 -07001702static inline void
1703__sk_dst_set(struct sock *sk, struct dst_entry *dst)
1704{
1705 struct dst_entry *old_dst;
1706
Krishna Kumare022f0b2009-10-19 23:46:20 +00001707 sk_tx_queue_clear(sk);
Eric Dumazet0b53ff22010-04-26 20:40:43 +00001708 /*
1709 * This can be called while sk is owned by the caller only,
1710 * with no state that can be checked in a rcu_dereference_check() cond
1711 */
1712 old_dst = rcu_dereference_raw(sk->sk_dst_cache);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001713 rcu_assign_pointer(sk->sk_dst_cache, dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001714 dst_release(old_dst);
1715}
1716
1717static inline void
1718sk_dst_set(struct sock *sk, struct dst_entry *dst)
1719{
Eric Dumazet7f502362014-06-30 01:26:23 -07001720 struct dst_entry *old_dst;
1721
1722 sk_tx_queue_clear(sk);
Eric Dumazet5925a052014-07-02 02:39:38 -07001723 old_dst = xchg((__force struct dst_entry **)&sk->sk_dst_cache, dst);
Eric Dumazet7f502362014-06-30 01:26:23 -07001724 dst_release(old_dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001725}
1726
1727static inline void
1728__sk_dst_reset(struct sock *sk)
1729{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001730 __sk_dst_set(sk, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731}
1732
1733static inline void
1734sk_dst_reset(struct sock *sk)
1735{
Eric Dumazet7f502362014-06-30 01:26:23 -07001736 sk_dst_set(sk, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001737}
1738
Joe Perches69336bd2013-09-22 10:32:26 -07001739struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001740
Joe Perches69336bd2013-09-22 10:32:26 -07001741struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001742
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +02001743bool sk_mc_loop(struct sock *sk);
1744
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001745static inline bool sk_can_gso(const struct sock *sk)
Herbert Xubcd76112006-06-30 13:36:35 -07001746{
1747 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1748}
1749
Joe Perches69336bd2013-09-22 10:32:26 -07001750void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
Arnaldo Carvalho de Melo6cbb0df2005-08-09 19:49:02 -07001751
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001752static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
Eric Dumazeta4654192010-05-16 00:36:33 -07001753{
1754 sk->sk_route_nocaps |= flags;
1755 sk->sk_route_caps &= ~flags;
1756}
1757
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001758static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
Al Viro57be5bd2014-11-28 13:40:20 -05001759 struct iov_iter *from, char *to,
Wei Yongjun912d3982011-04-06 18:40:12 +00001760 int copy, int offset)
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001761{
1762 if (skb->ip_summed == CHECKSUM_NONE) {
Al Viro57be5bd2014-11-28 13:40:20 -05001763 __wsum csum = 0;
1764 if (csum_and_copy_from_iter(to, copy, &csum, from) != copy)
1765 return -EFAULT;
Wei Yongjun912d3982011-04-06 18:40:12 +00001766 skb->csum = csum_block_add(skb->csum, csum, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001767 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
Al Viro57be5bd2014-11-28 13:40:20 -05001768 if (copy_from_iter_nocache(to, copy, from) != copy)
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001769 return -EFAULT;
Al Viro57be5bd2014-11-28 13:40:20 -05001770 } else if (copy_from_iter(to, copy, from) != copy)
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001771 return -EFAULT;
1772
1773 return 0;
1774}
1775
1776static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
Al Viro57be5bd2014-11-28 13:40:20 -05001777 struct iov_iter *from, int copy)
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001778{
Wei Yongjun912d3982011-04-06 18:40:12 +00001779 int err, offset = skb->len;
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001780
Wei Yongjun912d3982011-04-06 18:40:12 +00001781 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1782 copy, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001783 if (err)
Wei Yongjun912d3982011-04-06 18:40:12 +00001784 __skb_trim(skb, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001785
1786 return err;
1787}
1788
Al Viro57be5bd2014-11-28 13:40:20 -05001789static inline int skb_copy_to_page_nocache(struct sock *sk, struct iov_iter *from,
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001790 struct sk_buff *skb,
1791 struct page *page,
1792 int off, int copy)
1793{
1794 int err;
1795
Wei Yongjun912d3982011-04-06 18:40:12 +00001796 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1797 copy, skb->len);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001798 if (err)
1799 return err;
1800
1801 skb->len += copy;
1802 skb->data_len += copy;
1803 skb->truesize += copy;
1804 sk->sk_wmem_queued += copy;
1805 sk_mem_charge(sk, copy);
1806 return 0;
1807}
1808
Eric Dumazetc5640392009-06-16 10:12:03 +00001809/**
1810 * sk_wmem_alloc_get - returns write allocations
1811 * @sk: socket
1812 *
1813 * Returns sk_wmem_alloc minus initial offset of one
1814 */
1815static inline int sk_wmem_alloc_get(const struct sock *sk)
1816{
1817 return atomic_read(&sk->sk_wmem_alloc) - 1;
1818}
1819
1820/**
1821 * sk_rmem_alloc_get - returns read allocations
1822 * @sk: socket
1823 *
1824 * Returns sk_rmem_alloc
1825 */
1826static inline int sk_rmem_alloc_get(const struct sock *sk)
1827{
1828 return atomic_read(&sk->sk_rmem_alloc);
1829}
1830
1831/**
1832 * sk_has_allocations - check if allocations are outstanding
1833 * @sk: socket
1834 *
1835 * Returns true if socket has write or read allocations
1836 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001837static inline bool sk_has_allocations(const struct sock *sk)
Eric Dumazetc5640392009-06-16 10:12:03 +00001838{
1839 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1840}
1841
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001842/**
Eric Dumazet43815482010-04-29 11:01:49 +00001843 * wq_has_sleeper - check if there are any waiting processes
Randy Dunlapacfbe962010-05-24 23:54:18 -07001844 * @wq: struct socket_wq
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001845 *
Eric Dumazet43815482010-04-29 11:01:49 +00001846 * Returns true if socket_wq has waiting processes
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001847 *
Eric Dumazet43815482010-04-29 11:01:49 +00001848 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001849 * barrier call. They were added due to the race found within the tcp code.
1850 *
1851 * Consider following tcp code paths:
1852 *
1853 * CPU1 CPU2
1854 *
1855 * sys_select receive packet
1856 * ... ...
1857 * __add_wait_queue update tp->rcv_nxt
1858 * ... ...
1859 * tp->rcv_nxt check sock_def_readable
1860 * ... {
Eric Dumazet43815482010-04-29 11:01:49 +00001861 * schedule rcu_read_lock();
1862 * wq = rcu_dereference(sk->sk_wq);
1863 * if (wq && waitqueue_active(&wq->wait))
1864 * wake_up_interruptible(&wq->wait)
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001865 * ...
1866 * }
1867 *
1868 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1869 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1870 * could then endup calling schedule and sleep forever if there are no more
1871 * data on the socket.
Jiri Olsaad462762009-07-08 12:10:31 +00001872 *
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001873 */
Eric Dumazet43815482010-04-29 11:01:49 +00001874static inline bool wq_has_sleeper(struct socket_wq *wq)
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001875{
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001876 /* We need to be sure we are in sync with the
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001877 * add_wait_queue modifications to the wait queue.
1878 *
1879 * This memory barrier is paired in the sock_poll_wait.
1880 */
Eric Dumazet43815482010-04-29 11:01:49 +00001881 smp_mb();
1882 return wq && waitqueue_active(&wq->wait);
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001883}
1884
1885/**
1886 * sock_poll_wait - place memory barrier behind the poll_wait call.
1887 * @filp: file
1888 * @wait_address: socket wait queue
1889 * @p: poll_table
1890 *
Eric Dumazet43815482010-04-29 11:01:49 +00001891 * See the comments in the wq_has_sleeper function.
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001892 */
1893static inline void sock_poll_wait(struct file *filp,
1894 wait_queue_head_t *wait_address, poll_table *p)
1895{
Hans Verkuil626cf232012-03-23 15:02:27 -07001896 if (!poll_does_not_wait(p) && wait_address) {
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001897 poll_wait(filp, wait_address, p);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001898 /* We need to be sure we are in sync with the
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001899 * socket flags modification.
1900 *
Eric Dumazet43815482010-04-29 11:01:49 +00001901 * This memory barrier is paired in the wq_has_sleeper.
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001902 */
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001903 smp_mb();
1904 }
1905}
1906
Tom Herbertb73c3d02014-07-01 21:32:17 -07001907static inline void skb_set_hash_from_sk(struct sk_buff *skb, struct sock *sk)
1908{
1909 if (sk->sk_txhash) {
1910 skb->l4_hash = 1;
1911 skb->hash = sk->sk_txhash;
1912 }
1913}
1914
Linus Torvalds1da177e2005-04-16 15:20:36 -07001915/*
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001916 * Queue a received datagram if it will fit. Stream and sequenced
Linus Torvalds1da177e2005-04-16 15:20:36 -07001917 * protocols can't normally use this as they need to fit buffers in
1918 * and play with them.
1919 *
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001920 * Inlined as it's very short and called for pretty much every
Linus Torvalds1da177e2005-04-16 15:20:36 -07001921 * packet ever received.
1922 */
1923
1924static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1925{
Herbert Xud55d87f2009-06-22 02:25:25 +00001926 skb_orphan(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001927 skb->sk = sk;
1928 skb->destructor = sock_wfree;
Tom Herbertb73c3d02014-07-01 21:32:17 -07001929 skb_set_hash_from_sk(skb, sk);
Eric Dumazet2b85a342009-06-11 02:55:43 -07001930 /*
1931 * We used to take a refcount on sk, but following operation
1932 * is enough to guarantee sk_free() wont free this sock until
1933 * all in-flight packets are completed
1934 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001935 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1936}
1937
1938static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1939{
Herbert Xud55d87f2009-06-22 02:25:25 +00001940 skb_orphan(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001941 skb->sk = sk;
1942 skb->destructor = sock_rfree;
1943 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001944 sk_mem_charge(sk, skb->truesize);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001945}
1946
Joe Perches69336bd2013-09-22 10:32:26 -07001947void sk_reset_timer(struct sock *sk, struct timer_list *timer,
1948 unsigned long expires);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001949
Joe Perches69336bd2013-09-22 10:32:26 -07001950void sk_stop_timer(struct sock *sk, struct timer_list *timer);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001951
Joe Perches69336bd2013-09-22 10:32:26 -07001952int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001953
Joe Perches69336bd2013-09-22 10:32:26 -07001954int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
Willem de Bruijn364a9e92014-08-31 21:30:27 -04001955struct sk_buff *sock_dequeue_err_skb(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001956
1957/*
1958 * Recover an error report and clear atomically
1959 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001960
Linus Torvalds1da177e2005-04-16 15:20:36 -07001961static inline int sock_error(struct sock *sk)
1962{
Benjamin LaHaisec1cbe4b2005-12-13 23:22:19 -08001963 int err;
1964 if (likely(!sk->sk_err))
1965 return 0;
1966 err = xchg(&sk->sk_err, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001967 return -err;
1968}
1969
1970static inline unsigned long sock_wspace(struct sock *sk)
1971{
1972 int amt = 0;
1973
1974 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1975 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001976 if (amt < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001977 amt = 0;
1978 }
1979 return amt;
1980}
1981
1982static inline void sk_wake_async(struct sock *sk, int how, int band)
1983{
Eric Dumazetbcdce712009-10-06 17:28:29 -07001984 if (sock_flag(sk, SOCK_FASYNC))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001985 sock_wake_async(sk->sk_socket, how, band);
1986}
1987
Daniel Borkmanneea86af2013-06-19 12:51:20 +02001988/* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
1989 * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
1990 * Note: for send buffers, TCP works better if we can build two skbs at
1991 * minimum.
Eric Dumazet7a91b4342010-09-26 18:53:07 -07001992 */
Eric Dumazet9eb5bf82013-07-03 05:02:22 -07001993#define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
Daniel Borkmanneea86af2013-06-19 12:51:20 +02001994
1995#define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2)
1996#define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE
Linus Torvalds1da177e2005-04-16 15:20:36 -07001997
1998static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1999{
2000 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
Eric Dumazet8df09ea2007-12-21 03:07:41 -08002001 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
Daniel Borkmanneea86af2013-06-19 12:51:20 +02002002 sk->sk_sndbuf = max_t(u32, sk->sk_sndbuf, SOCK_MIN_SNDBUF);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002003 }
2004}
2005
Pavel Emelyanovdf97c702007-11-29 21:22:33 +11002006struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002007
Eric Dumazet5640f762012-09-23 23:04:42 +00002008/**
2009 * sk_page_frag - return an appropriate page_frag
2010 * @sk: socket
2011 *
2012 * If socket allocation mode allows current thread to sleep, it means its
2013 * safe to use the per task page_frag instead of the per socket one.
2014 */
2015static inline struct page_frag *sk_page_frag(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002016{
Eric Dumazet5640f762012-09-23 23:04:42 +00002017 if (sk->sk_allocation & __GFP_WAIT)
2018 return &current->task_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002019
Eric Dumazet5640f762012-09-23 23:04:42 +00002020 return &sk->sk_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021}
2022
Joe Perches69336bd2013-09-22 10:32:26 -07002023bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag);
Eric Dumazet5640f762012-09-23 23:04:42 +00002024
Linus Torvalds1da177e2005-04-16 15:20:36 -07002025/*
2026 * Default write policy as shown to user space via poll/select/SIGIO
2027 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002028static inline bool sock_writeable(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002029{
Eric Dumazet8df09ea2007-12-21 03:07:41 -08002030 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002031}
2032
Al Virodd0fc662005-10-07 07:46:04 +01002033static inline gfp_t gfp_any(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002034{
Andrew Morton99709372009-02-12 16:43:17 -08002035 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002036}
2037
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002038static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002039{
2040 return noblock ? 0 : sk->sk_rcvtimeo;
2041}
2042
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002043static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002044{
2045 return noblock ? 0 : sk->sk_sndtimeo;
2046}
2047
2048static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2049{
2050 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
2051}
2052
2053/* Alas, with timeout socket operations are not restartable.
2054 * Compare this to poll().
2055 */
2056static inline int sock_intr_errno(long timeo)
2057{
2058 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2059}
2060
Eyal Birger744d5a32015-03-01 14:58:31 +02002061struct sock_skb_cb {
2062 u32 dropcount;
2063};
2064
2065/* Store sock_skb_cb at the end of skb->cb[] so protocol families
2066 * using skb->cb[] would keep using it directly and utilize its
2067 * alignement guarantee.
2068 */
2069#define SOCK_SKB_CB_OFFSET ((FIELD_SIZEOF(struct sk_buff, cb) - \
2070 sizeof(struct sock_skb_cb)))
2071
2072#define SOCK_SKB_CB(__skb) ((struct sock_skb_cb *)((__skb)->cb + \
2073 SOCK_SKB_CB_OFFSET))
2074
Eyal Birgerb4772ef2015-03-01 14:58:29 +02002075#define sock_skb_cb_check_size(size) \
Eyal Birger744d5a32015-03-01 14:58:31 +02002076 BUILD_BUG_ON((size) > SOCK_SKB_CB_OFFSET)
Eyal Birgerb4772ef2015-03-01 14:58:29 +02002077
Eyal Birger3bc3b962015-03-01 14:58:30 +02002078static inline void
2079sock_skb_set_dropcount(const struct sock *sk, struct sk_buff *skb)
2080{
Eyal Birger744d5a32015-03-01 14:58:31 +02002081 SOCK_SKB_CB(skb)->dropcount = atomic_read(&sk->sk_drops);
Eyal Birger3bc3b962015-03-01 14:58:30 +02002082}
2083
Joe Perches69336bd2013-09-22 10:32:26 -07002084void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2085 struct sk_buff *skb);
2086void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2087 struct sk_buff *skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -07002088
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002089static inline void
Linus Torvalds1da177e2005-04-16 15:20:36 -07002090sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2091{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002092 ktime_t kt = skb->tstamp;
Patrick Ohly20d49472009-02-12 05:03:38 +00002093 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002094
Patrick Ohly20d49472009-02-12 05:03:38 +00002095 /*
2096 * generate control messages if
Willem de Bruijnb9f40e22014-08-04 22:11:46 -04002097 * - receive time stamping in software requested
Patrick Ohly20d49472009-02-12 05:03:38 +00002098 * - software time stamp available and wanted
Patrick Ohly20d49472009-02-12 05:03:38 +00002099 * - hardware time stamps available and wanted
Patrick Ohly20d49472009-02-12 05:03:38 +00002100 */
2101 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
Willem de Bruijnb9f40e22014-08-04 22:11:46 -04002102 (sk->sk_tsflags & SOF_TIMESTAMPING_RX_SOFTWARE) ||
Willem de Bruijnc1991052014-09-03 12:01:18 -04002103 (kt.tv64 && sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) ||
Patrick Ohly20d49472009-02-12 05:03:38 +00002104 (hwtstamps->hwtstamp.tv64 &&
Willem de Bruijnb9f40e22014-08-04 22:11:46 -04002105 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE)))
Eric Dumazet92f37fd2007-03-25 22:14:49 -07002106 __sock_recv_timestamp(msg, sk, skb);
2107 else
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002108 sk->sk_stamp = kt;
Johannes Berg6e3e9392011-11-09 10:15:42 +01002109
2110 if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2111 __sock_recv_wifi_status(msg, sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002112}
2113
Joe Perches69336bd2013-09-22 10:32:26 -07002114void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2115 struct sk_buff *skb);
Eric Dumazet767dd032010-04-28 19:14:43 +00002116
2117static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2118 struct sk_buff *skb)
2119{
2120#define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
Willem de Bruijnb9f40e22014-08-04 22:11:46 -04002121 (1UL << SOCK_RCVTSTAMP))
2122#define TSFLAGS_ANY (SOF_TIMESTAMPING_SOFTWARE | \
2123 SOF_TIMESTAMPING_RAW_HARDWARE)
Eric Dumazet767dd032010-04-28 19:14:43 +00002124
Willem de Bruijnb9f40e22014-08-04 22:11:46 -04002125 if (sk->sk_flags & FLAGS_TS_OR_DROPS || sk->sk_tsflags & TSFLAGS_ANY)
Eric Dumazet767dd032010-04-28 19:14:43 +00002126 __sock_recv_ts_and_drops(msg, sk, skb);
2127 else
2128 sk->sk_stamp = skb->tstamp;
2129}
Neil Horman3b885782009-10-12 13:26:31 -07002130
Willem de Bruijn67cc0d42014-09-08 19:58:58 -04002131void __sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags);
2132
Linus Torvalds1da177e2005-04-16 15:20:36 -07002133/**
Patrick Ohly20d49472009-02-12 05:03:38 +00002134 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
Patrick Ohly20d49472009-02-12 05:03:38 +00002135 * @sk: socket sending this packet
Eric Dumazet140c55d2014-08-06 11:49:29 +02002136 * @tx_flags: completed with instructions for time stamping
2137 *
2138 * Note : callers should take care of initial *tx_flags value (usually 0)
Patrick Ohly20d49472009-02-12 05:03:38 +00002139 */
Willem de Bruijn67cc0d42014-09-08 19:58:58 -04002140static inline void sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags)
2141{
2142 if (unlikely(sk->sk_tsflags))
2143 __sock_tx_timestamp(sk, tx_flags);
2144 if (unlikely(sock_flag(sk, SOCK_WIFI_STATUS)))
2145 *tx_flags |= SKBTX_WIFI_STATUS;
2146}
Patrick Ohly20d49472009-02-12 05:03:38 +00002147
2148/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002149 * sk_eat_skb - Release a skb if it is no longer needed
Pavel Pisa4dc3b162005-05-01 08:59:25 -07002150 * @sk: socket to eat this skb from
2151 * @skb: socket buffer to eat
Linus Torvalds1da177e2005-04-16 15:20:36 -07002152 *
2153 * This routine must be called with interrupts disabled or with the socket
2154 * locked so that the sk_buff queue operation is ok.
2155*/
Dan Williams7bced392013-12-30 12:37:29 -08002156static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002157{
2158 __skb_unlink(skb, &sk->sk_receive_queue);
2159 __kfree_skb(skb);
2160}
2161
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002162static inline
2163struct net *sock_net(const struct sock *sk)
2164{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002165 return read_pnet(&sk->sk_net);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002166}
2167
2168static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07002169void sock_net_set(struct sock *sk, struct net *net)
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002170{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002171 write_pnet(&sk->sk_net, net);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002172}
2173
Denis V. Lunevedf02082008-02-29 11:18:32 -08002174/*
2175 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
Lucas De Marchi25985ed2011-03-30 22:57:33 -03002176 * They should not hold a reference to a namespace in order to allow
Denis V. Lunevedf02082008-02-29 11:18:32 -08002177 * to stop it.
2178 * Sockets after sk_change_net should be released using sk_release_kernel
2179 */
2180static inline void sk_change_net(struct sock *sk, struct net *net)
2181{
Gu Zheng5812521b2014-03-10 09:57:34 +08002182 struct net *current_net = sock_net(sk);
2183
2184 if (!net_eq(current_net, net)) {
2185 put_net(current_net);
Eric W. Biedermanefd7ef12015-03-11 23:04:08 -05002186 sock_net_set(sk, net);
Gu Zheng5812521b2014-03-10 09:57:34 +08002187 }
Denis V. Lunevedf02082008-02-29 11:18:32 -08002188}
2189
KOVACS Krisztian23542612008-10-07 12:41:01 -07002190static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2191{
Vijay Subramanianefc27f82012-06-24 13:03:07 +00002192 if (skb->sk) {
KOVACS Krisztian23542612008-10-07 12:41:01 -07002193 struct sock *sk = skb->sk;
2194
2195 skb->destructor = NULL;
2196 skb->sk = NULL;
2197 return sk;
2198 }
2199 return NULL;
2200}
2201
Eric Dumazet1d0ab252015-03-15 21:12:12 -07002202/* This helper checks if a socket is a full socket,
2203 * ie _not_ a timewait or request socket.
2204 */
2205static inline bool sk_fullsock(const struct sock *sk)
2206{
2207 return (1 << sk->sk_state) & ~(TCPF_TIME_WAIT | TCPF_NEW_SYN_RECV);
2208}
2209
Joe Perches69336bd2013-09-22 10:32:26 -07002210void sock_enable_timestamp(struct sock *sk, int flag);
2211int sock_get_timestamp(struct sock *, struct timeval __user *);
2212int sock_get_timestampns(struct sock *, struct timespec __user *);
2213int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len, int level,
2214 int type);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002215
Eric W. Biedermana3b299d2014-04-23 14:26:56 -07002216bool sk_ns_capable(const struct sock *sk,
2217 struct user_namespace *user_ns, int cap);
2218bool sk_capable(const struct sock *sk, int cap);
2219bool sk_net_capable(const struct sock *sk, int cap);
2220
Linus Torvalds1da177e2005-04-16 15:20:36 -07002221extern __u32 sysctl_wmem_max;
2222extern __u32 sysctl_rmem_max;
2223
Willem de Bruijnb245be12015-01-30 13:29:32 -05002224extern int sysctl_tstamp_allow_data;
David S. Miller6baf1f42005-09-05 18:14:11 -07002225extern int sysctl_optmem_max;
2226
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002227extern __u32 sysctl_wmem_default;
2228extern __u32 sysctl_rmem_default;
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002229
Linus Torvalds1da177e2005-04-16 15:20:36 -07002230#endif /* _SOCK_H */