blob: 2210fec65669c870384fe87f1d536eca34ab3586 [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/aio.h>
61#include <linux/sched.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070062
63#include <linux/filter.h>
Eric Dumazet88ab1932008-11-16 19:39:21 -080064#include <linux/rculist_nulls.h>
Jiri Olsaa57de0b2009-07-08 12:09:13 +000065#include <linux/poll.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070066
Eric Dumazetc31504d2010-11-15 19:58:26 +000067#include <linux/atomic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070068#include <net/dst.h>
69#include <net/checksum.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 Costac607b2e2011-12-16 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 Costac607b2e2011-12-16 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 Costac607b2e2011-12-16 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 Costac607b2e2011-12-16 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;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900193#ifdef CONFIG_NET_NS
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200194 struct net *skc_net;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900195#endif
Eric Dumazetefe42082013-10-03 15:42:29 -0700196
197#if IS_ENABLED(CONFIG_IPV6)
198 struct in6_addr skc_v6_daddr;
199 struct in6_addr skc_v6_rcv_saddr;
200#endif
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
332
Eric Dumazetb178bb32010-11-16 05:56:04 +0000333 socket_lock_t sk_lock;
334 struct sk_buff_head sk_receive_queue;
335 /*
336 * The backlog queue is special, it is always used with
337 * the per-socket spinlock held and requires low latency
338 * access. Therefore we special case it's implementation.
339 * Note : rmem_alloc is in this structure to fill a hole
340 * on 64bit arches, not because its logically part of
341 * backlog.
342 */
343 struct {
344 atomic_t rmem_alloc;
345 int len;
346 struct sk_buff *head;
347 struct sk_buff *tail;
348 } sk_backlog;
349#define sk_rmem_alloc sk_backlog.rmem_alloc
350 int sk_forward_alloc;
351#ifdef CONFIG_RPS
352 __u32 sk_rxhash;
353#endif
Eric Dumazet2c8c56e2014-11-11 05:54:28 -0800354 u16 sk_incoming_cpu;
355 /* 16bit hole
356 * Warned : sk_incoming_cpu can be set from softirq,
357 * Do not use this hole without fully understanding possible issues.
358 */
359
Tom Herbertb73c3d02014-07-01 21:32:17 -0700360 __u32 sk_txhash;
Cong Wange0d10952013-08-01 11:10:25 +0800361#ifdef CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir06021292013-06-10 11:39:50 +0300362 unsigned int sk_napi_id;
Eliezer Tamirdafcc432013-06-14 16:33:57 +0300363 unsigned int sk_ll_usec;
Eliezer Tamir06021292013-06-10 11:39:50 +0300364#endif
Eric Dumazetb178bb32010-11-16 05:56:04 +0000365 atomic_t sk_drops;
366 int sk_rcvbuf;
367
368 struct sk_filter __rcu *sk_filter;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000369 struct socket_wq __rcu *sk_wq;
Eric Dumazetb178bb32010-11-16 05:56:04 +0000370
Eric Dumazetb178bb32010-11-16 05:56:04 +0000371#ifdef CONFIG_XFRM
372 struct xfrm_policy *sk_policy[2];
373#endif
374 unsigned long sk_flags;
Eric Dumazetdeaa5852012-06-24 20:22:49 +0000375 struct dst_entry *sk_rx_dst;
Cong Wang0e36cbb2013-01-22 21:09:51 +0000376 struct dst_entry __rcu *sk_dst_cache;
Eric Dumazetb178bb32010-11-16 05:56:04 +0000377 spinlock_t sk_dst_lock;
378 atomic_t sk_wmem_alloc;
379 atomic_t sk_omem_alloc;
380 int sk_sndbuf;
381 struct sk_buff_head sk_write_queue;
Vegard Nossuma98b65a2009-02-26 14:46:57 +0100382 kmemcheck_bitfield_begin(flags);
Eric Dumazet5fdb9972009-10-08 22:50:25 +0000383 unsigned int sk_shutdown : 2,
Tom Herbert28448b82014-05-23 08:47:19 -0700384 sk_no_check_tx : 1,
385 sk_no_check_rx : 1,
Eric Dumazet5fdb9972009-10-08 22:50:25 +0000386 sk_userlocks : 4,
387 sk_protocol : 8,
388 sk_type : 16;
Vegard Nossuma98b65a2009-02-26 14:46:57 +0100389 kmemcheck_bitfield_end(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390 int sk_wmem_queued;
Al Viro7d877f32005-10-21 03:20:43 -0400391 gfp_t sk_allocation;
Eric Dumazet95bd09e2013-08-27 05:46:32 -0700392 u32 sk_pacing_rate; /* bytes per second */
Eric Dumazet62748f32013-09-24 08:20:52 -0700393 u32 sk_max_pacing_rate;
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000394 netdev_features_t sk_route_caps;
395 netdev_features_t sk_route_nocaps;
Herbert Xubcd76112006-06-30 13:36:35 -0700396 int sk_gso_type;
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -0700397 unsigned int sk_gso_max_size;
Ben Hutchings14853482012-07-30 16:11:42 +0000398 u16 sk_gso_max_segs;
David S. Miller9932cf92006-03-24 15:12:37 -0800399 int sk_rcvlowat;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400 unsigned long sk_lingertime;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401 struct sk_buff_head sk_error_queue;
Arnaldo Carvalho de Melo476e19c2005-05-05 13:35:15 -0700402 struct proto *sk_prot_creator;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403 rwlock_t sk_callback_lock;
404 int sk_err,
405 sk_err_soft;
406 unsigned short sk_ack_backlog;
407 unsigned short sk_max_ack_backlog;
408 __u32 sk_priority;
Daniel Borkmann86f85152013-12-29 17:27:11 +0100409#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
Neil Horman5bc14212011-11-22 05:10:51 +0000410 __u32 sk_cgrp_prioidx;
411#endif
Eric W. Biederman109f6e32010-06-13 03:30:14 +0000412 struct pid *sk_peer_pid;
413 const struct cred *sk_peer_cred;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414 long sk_rcvtimeo;
415 long sk_sndtimeo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416 void *sk_protinfo;
417 struct timer_list sk_timer;
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700418 ktime_t sk_stamp;
Willem de Bruijnb9f40e22014-08-04 22:11:46 -0400419 u16 sk_tsflags;
Willem de Bruijn09c2d252014-08-04 22:11:47 -0400420 u32 sk_tskey;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 struct socket *sk_socket;
422 void *sk_user_data;
Eric Dumazet5640f762012-09-23 23:04:42 +0000423 struct page_frag sk_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424 struct sk_buff *sk_send_head;
Pavel Emelyanovef64a542012-02-21 07:31:34 +0000425 __s32 sk_peek_off;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426 int sk_write_pending;
Alexey Dobriyand5f64232008-11-04 14:45:58 -0800427#ifdef CONFIG_SECURITY
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428 void *sk_security;
Alexey Dobriyand5f64232008-11-04 14:45:58 -0800429#endif
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800430 __u32 sk_mark;
Herbert Xuf8451722010-05-24 00:12:34 -0700431 u32 sk_classid;
Glauber Costae1aab162011-12-11 21:47:03 +0000432 struct cg_proto *sk_cgrp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433 void (*sk_state_change)(struct sock *sk);
David S. Miller676d2362014-04-11 16:15:36 -0400434 void (*sk_data_ready)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435 void (*sk_write_space)(struct sock *sk);
436 void (*sk_error_report)(struct sock *sk);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000437 int (*sk_backlog_rcv)(struct sock *sk,
438 struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439 void (*sk_destruct)(struct sock *sk);
440};
441
Pravin B Shelar559835ea2013-09-24 10:25:40 -0700442#define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
443
444#define rcu_dereference_sk_user_data(sk) rcu_dereference(__sk_user_data((sk)))
445#define rcu_assign_sk_user_data(sk, ptr) rcu_assign_pointer(__sk_user_data((sk)), ptr)
446
Pavel Emelyanov4a17fd52012-04-19 03:39:36 +0000447/*
448 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
449 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
450 * on a socket means that the socket will reuse everybody else's port
451 * without looking at the other's sk_reuse value.
452 */
453
454#define SK_NO_REUSE 0
455#define SK_CAN_REUSE 1
456#define SK_FORCE_REUSE 2
457
Pavel Emelyanovef64a542012-02-21 07:31:34 +0000458static inline int sk_peek_offset(struct sock *sk, int flags)
459{
460 if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0))
461 return sk->sk_peek_off;
462 else
463 return 0;
464}
465
466static inline void sk_peek_offset_bwd(struct sock *sk, int val)
467{
468 if (sk->sk_peek_off >= 0) {
469 if (sk->sk_peek_off >= val)
470 sk->sk_peek_off -= val;
471 else
472 sk->sk_peek_off = 0;
473 }
474}
475
476static inline void sk_peek_offset_fwd(struct sock *sk, int val)
477{
478 if (sk->sk_peek_off >= 0)
479 sk->sk_peek_off += val;
480}
481
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482/*
483 * Hashed lists helper routines
484 */
Li Zefanc4146642010-02-08 23:18:45 +0000485static inline struct sock *sk_entry(const struct hlist_node *node)
486{
487 return hlist_entry(node, struct sock, sk_node);
488}
489
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700490static inline struct sock *__sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491{
492 return hlist_entry(head->first, struct sock, sk_node);
493}
494
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700495static inline struct sock *sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496{
497 return hlist_empty(head) ? NULL : __sk_head(head);
498}
499
Eric Dumazet88ab1932008-11-16 19:39:21 -0800500static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
501{
502 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
503}
504
505static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
506{
507 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
508}
509
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700510static inline struct sock *sk_next(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511{
512 return sk->sk_node.next ?
513 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
514}
515
Eric Dumazet88ab1932008-11-16 19:39:21 -0800516static inline struct sock *sk_nulls_next(const struct sock *sk)
517{
518 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
519 hlist_nulls_entry(sk->sk_nulls_node.next,
520 struct sock, sk_nulls_node) :
521 NULL;
522}
523
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000524static inline bool sk_unhashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525{
526 return hlist_unhashed(&sk->sk_node);
527}
528
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000529static inline bool sk_hashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530{
Akinobu Mitada753be2006-04-28 15:21:23 -0700531 return !sk_unhashed(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532}
533
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000534static inline void sk_node_init(struct hlist_node *node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535{
536 node->pprev = NULL;
537}
538
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000539static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800540{
541 node->pprev = NULL;
542}
543
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000544static inline void __sk_del_node(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545{
546 __hlist_del(&sk->sk_node);
547}
548
stephen hemminger808f5112010-02-22 07:57:18 +0000549/* NB: equivalent to hlist_del_init_rcu */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000550static inline bool __sk_del_node_init(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551{
552 if (sk_hashed(sk)) {
553 __sk_del_node(sk);
554 sk_node_init(&sk->sk_node);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000555 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556 }
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000557 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558}
559
560/* Grab socket reference count. This operation is valid only
561 when sk is ALREADY grabbed f.e. it is found in hash table
562 or a list and the lookup is made under lock preventing hash table
563 modifications.
564 */
565
566static inline void sock_hold(struct sock *sk)
567{
568 atomic_inc(&sk->sk_refcnt);
569}
570
571/* Ungrab socket in the context, which assumes that socket refcnt
572 cannot hit zero, f.e. it is true in context of any socketcall.
573 */
574static inline void __sock_put(struct sock *sk)
575{
576 atomic_dec(&sk->sk_refcnt);
577}
578
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000579static inline bool sk_del_node_init(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580{
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000581 bool rc = __sk_del_node_init(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582
583 if (rc) {
584 /* paranoid for a while -acme */
585 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
586 __sock_put(sk);
587 }
588 return rc;
589}
stephen hemminger808f5112010-02-22 07:57:18 +0000590#define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000592static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700593{
594 if (sk_hashed(sk)) {
Eric Dumazet88ab1932008-11-16 19:39:21 -0800595 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000596 return true;
Eric Dumazet271b72c2008-10-29 02:11:14 -0700597 }
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000598 return false;
Eric Dumazet271b72c2008-10-29 02:11:14 -0700599}
600
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000601static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700602{
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000603 bool rc = __sk_nulls_del_node_init_rcu(sk);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700604
605 if (rc) {
606 /* paranoid for a while -acme */
607 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
608 __sock_put(sk);
609 }
610 return rc;
611}
612
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000613static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700614{
615 hlist_add_head(&sk->sk_node, list);
616}
617
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000618static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619{
620 sock_hold(sk);
621 __sk_add_node(sk, list);
622}
623
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000624static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
stephen hemminger808f5112010-02-22 07:57:18 +0000625{
626 sock_hold(sk);
627 hlist_add_head_rcu(&sk->sk_node, list);
628}
629
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000630static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700631{
Eric Dumazet88ab1932008-11-16 19:39:21 -0800632 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700633}
634
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000635static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700636{
637 sock_hold(sk);
Eric Dumazet88ab1932008-11-16 19:39:21 -0800638 __sk_nulls_add_node_rcu(sk, list);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700639}
640
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000641static inline void __sk_del_bind_node(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642{
643 __hlist_del(&sk->sk_bind_node);
644}
645
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000646static inline void sk_add_bind_node(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647 struct hlist_head *list)
648{
649 hlist_add_head(&sk->sk_bind_node, list);
650}
651
Sasha Levinb67bfe02013-02-27 17:06:00 -0800652#define sk_for_each(__sk, list) \
653 hlist_for_each_entry(__sk, list, sk_node)
654#define sk_for_each_rcu(__sk, list) \
655 hlist_for_each_entry_rcu(__sk, list, sk_node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800656#define sk_nulls_for_each(__sk, node, list) \
657 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
658#define sk_nulls_for_each_rcu(__sk, node, list) \
659 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
Sasha Levinb67bfe02013-02-27 17:06:00 -0800660#define sk_for_each_from(__sk) \
661 hlist_for_each_entry_from(__sk, sk_node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800662#define sk_nulls_for_each_from(__sk, node) \
663 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
664 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
Sasha Levinb67bfe02013-02-27 17:06:00 -0800665#define sk_for_each_safe(__sk, tmp, list) \
666 hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
667#define sk_for_each_bound(__sk, list) \
668 hlist_for_each_entry(__sk, list, sk_bind_node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669
David Held2dc41cf2014-07-15 23:28:32 -0400670/**
671 * sk_nulls_for_each_entry_offset - iterate over a list at a given struct offset
672 * @tpos: the type * to use as a loop cursor.
673 * @pos: the &struct hlist_node to use as a loop cursor.
674 * @head: the head for your list.
675 * @offset: offset of hlist_node within the struct.
676 *
677 */
678#define sk_nulls_for_each_entry_offset(tpos, pos, head, offset) \
679 for (pos = (head)->first; \
680 (!is_a_nulls(pos)) && \
681 ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;}); \
682 pos = pos->next)
683
Eric W. Biedermanc336d142012-05-24 17:56:43 -0600684static inline struct user_namespace *sk_user_ns(struct sock *sk)
685{
686 /* Careful only use this in a context where these parameters
687 * can not change and must all be valid, such as recvmsg from
688 * userspace.
689 */
690 return sk->sk_socket->file->f_cred->user_ns;
691}
692
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693/* Sock flags */
694enum sock_flags {
695 SOCK_DEAD,
696 SOCK_DONE,
697 SOCK_URGINLINE,
698 SOCK_KEEPOPEN,
699 SOCK_LINGER,
700 SOCK_DESTROY,
701 SOCK_BROADCAST,
702 SOCK_TIMESTAMP,
703 SOCK_ZAPPED,
704 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
705 SOCK_DBG, /* %SO_DEBUG setting */
706 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700707 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
709 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
Mel Gorman7cb02402012-07-31 16:44:16 -0700710 SOCK_MEMALLOC, /* VM depends on this socket for swapping */
Patrick Ohly20d49472009-02-12 05:03:38 +0000711 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
Eric Dumazetbcdce712009-10-06 17:28:29 -0700712 SOCK_FASYNC, /* fasync() active */
Neil Horman3b885782009-10-12 13:26:31 -0700713 SOCK_RXQ_OVFL,
Shirley Ma1cdebb42011-07-06 12:17:30 +0000714 SOCK_ZEROCOPY, /* buffers from userspace */
Johannes Berg6e3e9392011-11-09 10:15:42 +0100715 SOCK_WIFI_STATUS, /* push wifi status to userspace */
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000716 SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
717 * Will use last 4 bytes of packet sent from
718 * user-space instead.
719 */
Vincent Bernatd59577b2013-01-16 22:55:49 +0100720 SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */
Keller, Jacob E7d4c04f2013-03-28 11:19:25 +0000721 SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722};
723
Ralf Baechle53b924b2005-08-23 10:11:30 -0700724static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
725{
726 nsk->sk_flags = osk->sk_flags;
727}
728
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
730{
731 __set_bit(flag, &sk->sk_flags);
732}
733
734static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
735{
736 __clear_bit(flag, &sk->sk_flags);
737}
738
Eric Dumazet1b23a5d2012-05-16 05:57:07 +0000739static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740{
741 return test_bit(flag, &sk->sk_flags);
742}
743
Mel Gormanc93bdd02012-07-31 16:44:19 -0700744#ifdef CONFIG_NET
745extern struct static_key memalloc_socks;
746static inline int sk_memalloc_socks(void)
747{
748 return static_key_false(&memalloc_socks);
749}
750#else
751
752static inline int sk_memalloc_socks(void)
753{
754 return 0;
755}
756
757#endif
758
Mel Gorman99a1dec2012-07-31 16:44:14 -0700759static inline gfp_t sk_gfp_atomic(struct sock *sk, gfp_t gfp_mask)
760{
Mel Gorman7cb02402012-07-31 16:44:16 -0700761 return GFP_ATOMIC | (sk->sk_allocation & __GFP_MEMALLOC);
Mel Gorman99a1dec2012-07-31 16:44:14 -0700762}
763
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764static inline void sk_acceptq_removed(struct sock *sk)
765{
766 sk->sk_ack_backlog--;
767}
768
769static inline void sk_acceptq_added(struct sock *sk)
770{
771 sk->sk_ack_backlog++;
772}
773
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000774static inline bool sk_acceptq_is_full(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775{
David S. Miller64a14652007-03-06 11:21:05 -0800776 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777}
778
779/*
780 * Compute minimal free write space needed to queue new packets.
781 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000782static inline int sk_stream_min_wspace(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783{
Eric Dumazet8df09ea2007-12-21 03:07:41 -0800784 return sk->sk_wmem_queued >> 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785}
786
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000787static inline int sk_stream_wspace(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788{
789 return sk->sk_sndbuf - sk->sk_wmem_queued;
790}
791
Joe Perches69336bd2013-09-22 10:32:26 -0700792void sk_stream_write_space(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793
Zhu Yi8eae9392010-03-04 18:01:40 +0000794/* OOB backlog add */
Zhu Yia3a858f2010-03-04 18:01:47 +0000795static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800796{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000797 /* dont let skb dst not refcounted, we are going to leave rcu lock */
798 skb_dst_force(skb);
799
800 if (!sk->sk_backlog.tail)
801 sk->sk_backlog.head = skb;
802 else
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800803 sk->sk_backlog.tail->next = skb;
Eric Dumazet7fee2262010-05-11 23:19:48 +0000804
805 sk->sk_backlog.tail = skb;
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800806 skb->next = NULL;
807}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808
Eric Dumazetc3774112010-04-27 15:13:20 -0700809/*
810 * Take into account size of receive queue and backlog queue
Eric Dumazet0fd7bac2011-12-21 07:11:44 +0000811 * Do not take into account this skb truesize,
812 * to allow even a single big packet to come.
Eric Dumazetc3774112010-04-27 15:13:20 -0700813 */
Sorin Dumitru274f4822014-07-22 21:16:51 +0300814static inline bool sk_rcvqueues_full(const struct sock *sk, unsigned int limit)
Eric Dumazetc3774112010-04-27 15:13:20 -0700815{
816 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
817
Eric Dumazetf545a382012-04-22 23:34:26 +0000818 return qsize > limit;
Eric Dumazetc3774112010-04-27 15:13:20 -0700819}
820
Zhu Yi8eae9392010-03-04 18:01:40 +0000821/* The per-socket spinlock must be held here. */
Eric Dumazetf545a382012-04-22 23:34:26 +0000822static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
823 unsigned int limit)
Zhu Yi8eae9392010-03-04 18:01:40 +0000824{
Sorin Dumitru274f4822014-07-22 21:16:51 +0300825 if (sk_rcvqueues_full(sk, limit))
Zhu Yi8eae9392010-03-04 18:01:40 +0000826 return -ENOBUFS;
827
Zhu Yia3a858f2010-03-04 18:01:47 +0000828 __sk_add_backlog(sk, skb);
Zhu Yi8eae9392010-03-04 18:01:40 +0000829 sk->sk_backlog.len += skb->truesize;
830 return 0;
831}
832
Joe Perches69336bd2013-09-22 10:32:26 -0700833int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
Mel Gormanb4b9e352012-07-31 16:44:26 -0700834
Peter Zijlstrac57943a2008-10-07 14:18:42 -0700835static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
836{
Mel Gormanb4b9e352012-07-31 16:44:26 -0700837 if (sk_memalloc_socks() && skb_pfmemalloc(skb))
838 return __sk_backlog_rcv(sk, skb);
839
Peter Zijlstrac57943a2008-10-07 14:18:42 -0700840 return sk->sk_backlog_rcv(sk, skb);
841}
842
Eric Dumazet2c8c56e2014-11-11 05:54:28 -0800843static inline void sk_incoming_cpu_update(struct sock *sk)
844{
845 sk->sk_incoming_cpu = raw_smp_processor_id();
846}
847
Tom Herbertfe477552013-12-22 18:54:31 +0800848static inline void sock_rps_record_flow_hash(__u32 hash)
David S. Millerc58dc012010-04-27 15:05:31 -0700849{
850#ifdef CONFIG_RPS
851 struct rps_sock_flow_table *sock_flow_table;
852
853 rcu_read_lock();
854 sock_flow_table = rcu_dereference(rps_sock_flow_table);
Tom Herbertfe477552013-12-22 18:54:31 +0800855 rps_record_sock_flow(sock_flow_table, hash);
David S. Millerc58dc012010-04-27 15:05:31 -0700856 rcu_read_unlock();
857#endif
858}
859
Tom Herbertfe477552013-12-22 18:54:31 +0800860static inline void sock_rps_reset_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_reset_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
879static inline void sock_rps_reset_flow(const struct sock *sk)
880{
Zhi Yong Wuc9d8ca02014-01-01 04:31:01 +0800881#ifdef CONFIG_RPS
Tom Herbertfe477552013-12-22 18:54:31 +0800882 sock_rps_reset_flow_hash(sk->sk_rxhash);
Zhi Yong Wuc9d8ca02014-01-01 04:31:01 +0800883#endif
Tom Herbertfe477552013-12-22 18:54:31 +0800884}
885
Tom Herbertbdeab992011-08-14 19:45:55 +0000886static inline void sock_rps_save_rxhash(struct sock *sk,
887 const struct sk_buff *skb)
David S. Millerc58dc012010-04-27 15:05:31 -0700888{
889#ifdef CONFIG_RPS
Tom Herbert61b905d2014-03-24 15:34:47 -0700890 if (unlikely(sk->sk_rxhash != skb->hash)) {
David S. Millerc58dc012010-04-27 15:05:31 -0700891 sock_rps_reset_flow(sk);
Tom Herbert61b905d2014-03-24 15:34:47 -0700892 sk->sk_rxhash = skb->hash;
David S. Millerc58dc012010-04-27 15:05:31 -0700893 }
894#endif
895}
896
Tom Herbertbdeab992011-08-14 19:45:55 +0000897static inline void sock_rps_reset_rxhash(struct sock *sk)
898{
899#ifdef CONFIG_RPS
900 sock_rps_reset_flow(sk);
901 sk->sk_rxhash = 0;
902#endif
903}
904
Stephen Hemmingercfcabdc2007-10-09 01:59:42 -0700905#define sk_wait_event(__sk, __timeo, __condition) \
906 ({ int __rc; \
907 release_sock(__sk); \
908 __rc = __condition; \
909 if (!__rc) { \
910 *(__timeo) = schedule_timeout(*(__timeo)); \
911 } \
Peter Zijlstra26cabd32014-09-24 10:18:54 +0200912 sched_annotate_sleep(); \
Stephen Hemmingercfcabdc2007-10-09 01:59:42 -0700913 lock_sock(__sk); \
914 __rc = __condition; \
915 __rc; \
916 })
Linus Torvalds1da177e2005-04-16 15:20:36 -0700917
Joe Perches69336bd2013-09-22 10:32:26 -0700918int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
919int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
920void sk_stream_wait_close(struct sock *sk, long timeo_p);
921int sk_stream_error(struct sock *sk, int flags, int err);
922void sk_stream_kill_queues(struct sock *sk);
923void sk_set_memalloc(struct sock *sk);
924void sk_clear_memalloc(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700925
Joe Perches69336bd2013-09-22 10:32:26 -0700926int sk_wait_data(struct sock *sk, long *timeo);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700928struct request_sock_ops;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800929struct timewait_sock_ops;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -0800930struct inet_hashinfo;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -0700931struct raw_hashinfo;
Paul Gortmakerde477252011-05-26 13:46:22 -0400932struct module;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700933
Eric Dumazetf77d6022013-05-09 10:28:16 +0000934/*
935 * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
936 * un-modified. Special care is taken when initializing object to zero.
937 */
938static inline void sk_prot_clear_nulls(struct sock *sk, int size)
939{
940 if (offsetof(struct sock, sk_node.next) != 0)
941 memset(sk, 0, offsetof(struct sock, sk_node.next));
942 memset(&sk->sk_node.pprev, 0,
943 size - offsetof(struct sock, sk_node.pprev));
944}
945
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946/* Networking protocol blocks we attach to sockets.
947 * socket layer -> transport layer interface
948 * transport -> network interface is defined by struct inet_proto
949 */
950struct proto {
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000951 void (*close)(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700952 long timeout);
953 int (*connect)(struct sock *sk,
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000954 struct sockaddr *uaddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955 int addr_len);
956 int (*disconnect)(struct sock *sk, int flags);
957
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000958 struct sock * (*accept)(struct sock *sk, int flags, int *err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959
960 int (*ioctl)(struct sock *sk, int cmd,
961 unsigned long arg);
962 int (*init)(struct sock *sk);
Brian Haley7d06b2e2008-06-14 17:04:49 -0700963 void (*destroy)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964 void (*shutdown)(struct sock *sk, int how);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000965 int (*setsockopt)(struct sock *sk, int level,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700967 unsigned int optlen);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000968 int (*getsockopt)(struct sock *sk, int level,
969 int optname, char __user *optval,
970 int __user *option);
Alexey Dobriyanaf01d532008-08-28 02:53:51 -0700971#ifdef CONFIG_COMPAT
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800972 int (*compat_setsockopt)(struct sock *sk,
973 int level,
974 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700975 unsigned int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800976 int (*compat_getsockopt)(struct sock *sk,
977 int level,
978 int optname, char __user *optval,
979 int __user *option);
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000980 int (*compat_ioctl)(struct sock *sk,
981 unsigned int cmd, unsigned long arg);
Alexey Dobriyanaf01d532008-08-28 02:53:51 -0700982#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700983 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
984 struct msghdr *msg, size_t len);
985 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
986 struct msghdr *msg,
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000987 size_t len, int noblock, int flags,
988 int *addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989 int (*sendpage)(struct sock *sk, struct page *page,
990 int offset, size_t size, int flags);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000991 int (*bind)(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992 struct sockaddr *uaddr, int addr_len);
993
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000994 int (*backlog_rcv) (struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995 struct sk_buff *skb);
996
Eric Dumazet46d3cea2012-07-11 05:50:31 +0000997 void (*release_cb)(struct sock *sk);
998
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999 /* Keeping track of sk's, looking them up, and port selection methods. */
1000 void (*hash)(struct sock *sk);
1001 void (*unhash)(struct sock *sk);
Eric Dumazet719f8352010-09-08 05:08:44 +00001002 void (*rehash)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003 int (*get_port)(struct sock *sk, unsigned short snum);
Octavian Purdilafcbdf092010-12-16 14:26:56 -08001004 void (*clear_sk)(struct sock *sk, int size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005
Eric Dumazet286ab3d2007-11-05 23:38:39 -08001006 /* Keeping track of sockets in use */
Eric Dumazet65f76512008-01-03 20:46:48 -08001007#ifdef CONFIG_PROC_FS
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -07001008 unsigned int inuse_idx;
Eric Dumazet65f76512008-01-03 20:46:48 -08001009#endif
Arnaldo Carvalho de Meloebb53d72007-11-21 22:08:50 +08001010
Eric Dumazetc9bee3b72013-07-22 20:27:07 -07001011 bool (*stream_memory_free)(const struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012 /* Memory pressure */
Pavel Emelyanov5c52ba12008-07-16 20:28:10 -07001013 void (*enter_memory_pressure)(struct sock *sk);
Eric Dumazet8d987e52010-11-09 23:24:26 +00001014 atomic_long_t *memory_allocated; /* Current allocated memory. */
Eric Dumazet17483762008-11-25 21:16:35 -08001015 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016 /*
1017 * Pressure flag: try to collapse.
1018 * Technical note: it is used by multiple contexts non atomically.
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001019 * All the __sk_mem_schedule() is of this nature: accounting
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020 * is strict, actions are advisory and have some latency.
1021 */
1022 int *memory_pressure;
Eric Dumazet8d987e52010-11-09 23:24:26 +00001023 long *sysctl_mem;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024 int *sysctl_wmem;
1025 int *sysctl_rmem;
1026 int max_header;
Changli Gao7ba42912010-07-10 20:41:55 +00001027 bool no_autobind;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028
Eric Dumazet271b72c2008-10-29 02:11:14 -07001029 struct kmem_cache *slab;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001030 unsigned int obj_size;
Eric Dumazet271b72c2008-10-29 02:11:14 -07001031 int slab_flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032
Eric Dumazetdd24c002008-11-25 21:17:14 -08001033 struct percpu_counter *orphan_count;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -07001034
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -07001035 struct request_sock_ops *rsk_prot;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -08001036 struct timewait_sock_ops *twsk_prot;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07001037
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -07001038 union {
1039 struct inet_hashinfo *hashinfo;
Eric Dumazet645ca702008-10-29 01:41:45 -07001040 struct udp_table *udp_table;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -07001041 struct raw_hashinfo *raw_hash;
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -07001042 } h;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -08001043
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044 struct module *owner;
1045
1046 char name[32];
1047
1048 struct list_head node;
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -07001049#ifdef SOCK_REFCNT_DEBUG
1050 atomic_t socks;
1051#endif
Andrew Mortonc255a452012-07-31 16:43:02 -07001052#ifdef CONFIG_MEMCG_KMEM
Glauber Costae1aab162011-12-11 21:47:03 +00001053 /*
1054 * cgroup specific init/deinit functions. Called once for all
1055 * protocols that implement it, from cgroups populate function.
1056 * This function has to setup any files the protocol want to
1057 * appear in the kmem cgroup filesystem.
1058 */
Glauber Costa1d62e432012-04-09 19:36:33 -03001059 int (*init_cgroup)(struct mem_cgroup *memcg,
Glauber Costae1aab162011-12-11 21:47:03 +00001060 struct cgroup_subsys *ss);
Glauber Costa1d62e432012-04-09 19:36:33 -03001061 void (*destroy_cgroup)(struct mem_cgroup *memcg);
Glauber Costae1aab162011-12-11 21:47:03 +00001062 struct cg_proto *(*proto_cgroup)(struct mem_cgroup *memcg);
1063#endif
1064};
1065
Glauber Costa3f134612012-05-29 15:07:11 -07001066/*
1067 * Bits in struct cg_proto.flags
1068 */
1069enum cg_proto_flags {
1070 /* Currently active and new sockets should be assigned to cgroups */
1071 MEMCG_SOCK_ACTIVE,
1072 /* It was ever activated; we must disarm static keys on destruction */
1073 MEMCG_SOCK_ACTIVATED,
1074};
1075
Glauber Costae1aab162011-12-11 21:47:03 +00001076struct cg_proto {
Johannes Weiner3e32cb22014-12-10 15:42:31 -08001077 struct page_counter memory_allocated; /* Current allocated memory. */
Eric W. Biederman2e685ca2013-10-19 16:26:19 -07001078 struct percpu_counter sockets_allocated; /* Current number of sockets. */
1079 int memory_pressure;
1080 long sysctl_mem[3];
Glauber Costa3f134612012-05-29 15:07:11 -07001081 unsigned long flags;
Glauber Costae1aab162011-12-11 21:47:03 +00001082 /*
1083 * memcg field is used to find which memcg we belong directly
1084 * Each memcg struct can hold more than one cg_proto, so container_of
1085 * won't really cut.
1086 *
1087 * The elegant solution would be having an inverse function to
1088 * proto_cgroup in struct proto, but that means polluting the structure
1089 * for everybody, instead of just for memcg users.
1090 */
1091 struct mem_cgroup *memcg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001092};
1093
Joe Perches69336bd2013-09-22 10:32:26 -07001094int proto_register(struct proto *prot, int alloc_slab);
1095void proto_unregister(struct proto *prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096
Glauber Costa3f134612012-05-29 15:07:11 -07001097static inline bool memcg_proto_active(struct cg_proto *cg_proto)
1098{
1099 return test_bit(MEMCG_SOCK_ACTIVE, &cg_proto->flags);
1100}
1101
1102static inline bool memcg_proto_activated(struct cg_proto *cg_proto)
1103{
1104 return test_bit(MEMCG_SOCK_ACTIVATED, &cg_proto->flags);
1105}
1106
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -07001107#ifdef SOCK_REFCNT_DEBUG
1108static inline void sk_refcnt_debug_inc(struct sock *sk)
1109{
1110 atomic_inc(&sk->sk_prot->socks);
1111}
1112
1113static inline void sk_refcnt_debug_dec(struct sock *sk)
1114{
1115 atomic_dec(&sk->sk_prot->socks);
1116 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
1117 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
1118}
1119
Ying Xuedec34fb2013-02-15 22:28:25 +00001120static inline void sk_refcnt_debug_release(const struct sock *sk)
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -07001121{
1122 if (atomic_read(&sk->sk_refcnt) != 1)
1123 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
1124 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
1125}
1126#else /* SOCK_REFCNT_DEBUG */
1127#define sk_refcnt_debug_inc(sk) do { } while (0)
1128#define sk_refcnt_debug_dec(sk) do { } while (0)
1129#define sk_refcnt_debug_release(sk) do { } while (0)
1130#endif /* SOCK_REFCNT_DEBUG */
1131
Andrew Mortonc255a452012-07-31 16:43:02 -07001132#if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_NET)
Ingo Molnarc5905af2012-02-24 08:31:31 +01001133extern struct static_key memcg_socket_limit_enabled;
Glauber Costae1aab162011-12-11 21:47:03 +00001134static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1135 struct cg_proto *cg_proto)
1136{
1137 return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg));
1138}
Ingo Molnarc5905af2012-02-24 08:31:31 +01001139#define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
Glauber Costae1aab162011-12-11 21:47:03 +00001140#else
1141#define mem_cgroup_sockets_enabled 0
1142static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1143 struct cg_proto *cg_proto)
1144{
1145 return NULL;
1146}
1147#endif
1148
Eric Dumazetc9bee3b72013-07-22 20:27:07 -07001149static inline bool sk_stream_memory_free(const struct sock *sk)
1150{
1151 if (sk->sk_wmem_queued >= sk->sk_sndbuf)
1152 return false;
1153
1154 return sk->sk_prot->stream_memory_free ?
1155 sk->sk_prot->stream_memory_free(sk) : true;
1156}
1157
Eric Dumazet64dc6132013-07-22 20:26:31 -07001158static inline bool sk_stream_is_writeable(const struct sock *sk)
1159{
Eric Dumazetc9bee3b72013-07-22 20:27:07 -07001160 return sk_stream_wspace(sk) >= sk_stream_min_wspace(sk) &&
1161 sk_stream_memory_free(sk);
Eric Dumazet64dc6132013-07-22 20:26:31 -07001162}
Glauber Costae1aab162011-12-11 21:47:03 +00001163
Eric Dumazetc9bee3b72013-07-22 20:27:07 -07001164
Glauber Costa180d8cd2011-12-11 21:47:02 +00001165static inline bool sk_has_memory_pressure(const struct sock *sk)
1166{
1167 return sk->sk_prot->memory_pressure != NULL;
1168}
1169
1170static inline bool sk_under_memory_pressure(const struct sock *sk)
1171{
1172 if (!sk->sk_prot->memory_pressure)
1173 return false;
Glauber Costae1aab162011-12-11 21:47:03 +00001174
1175 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
Eric W. Biederman2e685ca2013-10-19 16:26:19 -07001176 return !!sk->sk_cgrp->memory_pressure;
Glauber Costae1aab162011-12-11 21:47:03 +00001177
Christoph Paasch35b87f62013-10-23 12:49:21 -07001178 return !!*sk->sk_prot->memory_pressure;
Glauber Costa180d8cd2011-12-11 21:47:02 +00001179}
1180
1181static inline void sk_leave_memory_pressure(struct sock *sk)
1182{
1183 int *memory_pressure = sk->sk_prot->memory_pressure;
1184
Glauber Costae1aab162011-12-11 21:47:03 +00001185 if (!memory_pressure)
1186 return;
1187
1188 if (*memory_pressure)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001189 *memory_pressure = 0;
Glauber Costae1aab162011-12-11 21:47:03 +00001190
1191 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1192 struct cg_proto *cg_proto = sk->sk_cgrp;
1193 struct proto *prot = sk->sk_prot;
1194
1195 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
Eric W. Biederman7f2cbdc2013-12-04 20:12:04 -08001196 cg_proto->memory_pressure = 0;
Glauber Costae1aab162011-12-11 21:47:03 +00001197 }
1198
Glauber Costa180d8cd2011-12-11 21:47:02 +00001199}
1200
1201static inline void sk_enter_memory_pressure(struct sock *sk)
1202{
Glauber Costae1aab162011-12-11 21:47:03 +00001203 if (!sk->sk_prot->enter_memory_pressure)
1204 return;
1205
1206 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1207 struct cg_proto *cg_proto = sk->sk_cgrp;
1208 struct proto *prot = sk->sk_prot;
1209
1210 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
Eric W. Biederman7f2cbdc2013-12-04 20:12:04 -08001211 cg_proto->memory_pressure = 1;
Glauber Costae1aab162011-12-11 21:47:03 +00001212 }
1213
1214 sk->sk_prot->enter_memory_pressure(sk);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001215}
1216
1217static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1218{
1219 long *prot = sk->sk_prot->sysctl_mem;
Glauber Costae1aab162011-12-11 21:47:03 +00001220 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1221 prot = sk->sk_cgrp->sysctl_mem;
Glauber Costa180d8cd2011-12-11 21:47:02 +00001222 return prot[index];
1223}
1224
Glauber Costae1aab162011-12-11 21:47:03 +00001225static inline void memcg_memory_allocated_add(struct cg_proto *prot,
1226 unsigned long amt,
1227 int *parent_status)
1228{
Johannes Weiner3e32cb22014-12-10 15:42:31 -08001229 page_counter_charge(&prot->memory_allocated, amt);
Glauber Costae1aab162011-12-11 21:47:03 +00001230
Johannes Weiner3e32cb22014-12-10 15:42:31 -08001231 if (page_counter_read(&prot->memory_allocated) >
1232 prot->memory_allocated.limit)
Glauber Costae1aab162011-12-11 21:47:03 +00001233 *parent_status = OVER_LIMIT;
1234}
1235
1236static inline void memcg_memory_allocated_sub(struct cg_proto *prot,
1237 unsigned long amt)
1238{
Johannes Weiner3e32cb22014-12-10 15:42:31 -08001239 page_counter_uncharge(&prot->memory_allocated, amt);
Glauber Costae1aab162011-12-11 21:47:03 +00001240}
1241
Glauber Costa180d8cd2011-12-11 21:47:02 +00001242static inline long
1243sk_memory_allocated(const struct sock *sk)
1244{
1245 struct proto *prot = sk->sk_prot;
Johannes Weiner3e32cb22014-12-10 15:42:31 -08001246
Glauber Costae1aab162011-12-11 21:47:03 +00001247 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
Johannes Weiner3e32cb22014-12-10 15:42:31 -08001248 return page_counter_read(&sk->sk_cgrp->memory_allocated);
Glauber Costae1aab162011-12-11 21:47:03 +00001249
Glauber Costa180d8cd2011-12-11 21:47:02 +00001250 return atomic_long_read(prot->memory_allocated);
1251}
1252
1253static inline long
Glauber Costae1aab162011-12-11 21:47:03 +00001254sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001255{
1256 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001257
1258 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1259 memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status);
1260 /* update the root cgroup regardless */
1261 atomic_long_add_return(amt, prot->memory_allocated);
Johannes Weiner3e32cb22014-12-10 15:42:31 -08001262 return page_counter_read(&sk->sk_cgrp->memory_allocated);
Glauber Costae1aab162011-12-11 21:47:03 +00001263 }
1264
Glauber Costa180d8cd2011-12-11 21:47:02 +00001265 return atomic_long_add_return(amt, prot->memory_allocated);
1266}
1267
1268static inline void
Glauber Costa0e90b312012-01-20 04:57:16 +00001269sk_memory_allocated_sub(struct sock *sk, int amt)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001270{
1271 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001272
Glauber Costa0e90b312012-01-20 04:57:16 +00001273 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
Glauber Costae1aab162011-12-11 21:47:03 +00001274 memcg_memory_allocated_sub(sk->sk_cgrp, amt);
1275
Glauber Costa180d8cd2011-12-11 21:47:02 +00001276 atomic_long_sub(amt, prot->memory_allocated);
1277}
1278
1279static inline void sk_sockets_allocated_dec(struct sock *sk)
1280{
1281 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001282
1283 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1284 struct cg_proto *cg_proto = sk->sk_cgrp;
1285
1286 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
Eric W. Biederman2e685ca2013-10-19 16:26:19 -07001287 percpu_counter_dec(&cg_proto->sockets_allocated);
Glauber Costae1aab162011-12-11 21:47:03 +00001288 }
1289
Glauber Costa180d8cd2011-12-11 21:47:02 +00001290 percpu_counter_dec(prot->sockets_allocated);
1291}
1292
1293static inline void sk_sockets_allocated_inc(struct sock *sk)
1294{
1295 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001296
1297 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1298 struct cg_proto *cg_proto = sk->sk_cgrp;
1299
1300 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
Eric W. Biederman2e685ca2013-10-19 16:26:19 -07001301 percpu_counter_inc(&cg_proto->sockets_allocated);
Glauber Costae1aab162011-12-11 21:47:03 +00001302 }
1303
Glauber Costa180d8cd2011-12-11 21:47:02 +00001304 percpu_counter_inc(prot->sockets_allocated);
1305}
1306
1307static inline int
1308sk_sockets_allocated_read_positive(struct sock *sk)
1309{
1310 struct proto *prot = sk->sk_prot;
1311
Glauber Costae1aab162011-12-11 21:47:03 +00001312 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
Eric W. Biederman2e685ca2013-10-19 16:26:19 -07001313 return percpu_counter_read_positive(&sk->sk_cgrp->sockets_allocated);
Glauber Costae1aab162011-12-11 21:47:03 +00001314
Eric Dumazet518fbf92012-04-28 23:21:56 +00001315 return percpu_counter_read_positive(prot->sockets_allocated);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001316}
1317
1318static inline int
1319proto_sockets_allocated_sum_positive(struct proto *prot)
1320{
1321 return percpu_counter_sum_positive(prot->sockets_allocated);
1322}
1323
1324static inline long
1325proto_memory_allocated(struct proto *prot)
1326{
1327 return atomic_long_read(prot->memory_allocated);
1328}
1329
1330static inline bool
1331proto_memory_pressure(struct proto *prot)
1332{
1333 if (!prot->memory_pressure)
1334 return false;
1335 return !!*prot->memory_pressure;
1336}
1337
Eric Dumazet65f76512008-01-03 20:46:48 -08001338
1339#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001340/* Called with local bh disabled */
Joe Perches69336bd2013-09-22 10:32:26 -07001341void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1342int sock_prot_inuse_get(struct net *net, struct proto *proto);
Eric Dumazet65f76512008-01-03 20:46:48 -08001343#else
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001344static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -07001345 int inc)
Eric Dumazet65f76512008-01-03 20:46:48 -08001346{
1347}
Eric Dumazet65f76512008-01-03 20:46:48 -08001348#endif
1349
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350
Arnaldo Carvalho de Melo614c6cb2005-08-09 19:47:37 -07001351/* With per-bucket locks this operation is not-atomic, so that
1352 * this version is not worse.
1353 */
1354static inline void __sk_prot_rehash(struct sock *sk)
1355{
1356 sk->sk_prot->unhash(sk);
1357 sk->sk_prot->hash(sk);
1358}
1359
Octavian Purdilafcbdf092010-12-16 14:26:56 -08001360void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
1361
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362/* About 10 seconds */
1363#define SOCK_DESTROY_TIME (10*HZ)
1364
1365/* Sockets 0-1023 can't be bound to unless you are superuser */
1366#define PROT_SOCK 1024
1367
1368#define SHUTDOWN_MASK 3
1369#define RCV_SHUTDOWN 1
1370#define SEND_SHUTDOWN 2
1371
1372#define SOCK_SNDBUF_LOCK 1
1373#define SOCK_RCVBUF_LOCK 2
1374#define SOCK_BINDADDR_LOCK 4
1375#define SOCK_BINDPORT_LOCK 8
1376
1377/* sock_iocb: used to kick off async processing of socket ios */
1378struct sock_iocb {
1379 struct list_head list;
1380
1381 int flags;
1382 int size;
1383 struct socket *sock;
1384 struct sock *sk;
1385 struct scm_cookie *scm;
1386 struct msghdr *msg, async_msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387 struct kiocb *kiocb;
1388};
1389
1390static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
1391{
1392 return (struct sock_iocb *)iocb->private;
1393}
1394
1395static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
1396{
1397 return si->kiocb;
1398}
1399
1400struct socket_alloc {
1401 struct socket socket;
1402 struct inode vfs_inode;
1403};
1404
1405static inline struct socket *SOCKET_I(struct inode *inode)
1406{
1407 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1408}
1409
1410static inline struct inode *SOCK_INODE(struct socket *socket)
1411{
1412 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1413}
1414
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001415/*
1416 * Functions for memory accounting
1417 */
Joe Perches69336bd2013-09-22 10:32:26 -07001418int __sk_mem_schedule(struct sock *sk, int size, int kind);
1419void __sk_mem_reclaim(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001421#define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1422#define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1423#define SK_MEM_SEND 0
1424#define SK_MEM_RECV 1
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001426static inline int sk_mem_pages(int amt)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001427{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001428 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429}
1430
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001431static inline bool sk_has_account(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001432{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001433 /* return true if protocol supports memory accounting */
1434 return !!sk->sk_prot->memory_allocated;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001435}
1436
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001437static inline bool sk_wmem_schedule(struct sock *sk, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001438{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001439 if (!sk_has_account(sk))
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001440 return true;
Herbert Xud80d99d2005-09-01 17:48:23 -07001441 return size <= sk->sk_forward_alloc ||
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001442 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1443}
1444
Mel Gormanc76562b2012-07-31 16:44:41 -07001445static inline bool
Chuck Lever35c448a2012-09-17 14:09:11 -07001446sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size)
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001447{
1448 if (!sk_has_account(sk))
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001449 return true;
Mel Gormanc76562b2012-07-31 16:44:41 -07001450 return size<= sk->sk_forward_alloc ||
1451 __sk_mem_schedule(sk, size, SK_MEM_RECV) ||
1452 skb_pfmemalloc(skb);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001453}
1454
1455static inline void sk_mem_reclaim(struct sock *sk)
1456{
1457 if (!sk_has_account(sk))
1458 return;
1459 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1460 __sk_mem_reclaim(sk);
1461}
1462
David S. Miller9993e7d2008-01-10 21:56:38 -08001463static inline void sk_mem_reclaim_partial(struct sock *sk)
1464{
1465 if (!sk_has_account(sk))
1466 return;
1467 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1468 __sk_mem_reclaim(sk);
1469}
1470
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001471static inline void sk_mem_charge(struct sock *sk, int size)
1472{
1473 if (!sk_has_account(sk))
1474 return;
1475 sk->sk_forward_alloc -= size;
1476}
1477
1478static inline void sk_mem_uncharge(struct sock *sk, int size)
1479{
1480 if (!sk_has_account(sk))
1481 return;
1482 sk->sk_forward_alloc += size;
1483}
1484
1485static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1486{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001487 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1488 sk->sk_wmem_queued -= skb->truesize;
1489 sk_mem_uncharge(sk, skb->truesize);
1490 __kfree_skb(skb);
Herbert Xud80d99d2005-09-01 17:48:23 -07001491}
1492
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493/* Used by processes to "lock" a socket state, so that
1494 * interrupts and bottom half handlers won't change it
1495 * from under us. It essentially blocks any incoming
1496 * packets, so that we won't get any new data or any
1497 * packets that change the state of the socket.
1498 *
1499 * While locked, BH processing will add new packets to
1500 * the backlog queue. This queue is processed by the
1501 * owner of the socket lock right before it is released.
1502 *
1503 * Since ~2.3.5 it is also exclusive sleep lock serializing
1504 * accesses from user process context.
1505 */
John Heffnerd2e91172007-09-12 10:44:19 +02001506#define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507
Eric Dumazetc3f9b012014-03-10 09:50:11 -07001508static inline void sock_release_ownership(struct sock *sk)
1509{
1510 sk->sk_lock.owned = 0;
1511}
1512
Peter Zijlstraed075362006-12-06 20:35:24 -08001513/*
1514 * Macro so as to not evaluate some arguments when
1515 * lockdep is not enabled.
1516 *
1517 * Mark both the sk_lock and the sk_lock.slock as a
1518 * per-address-family lock class.
1519 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001520#define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
Peter Zijlstraed075362006-12-06 20:35:24 -08001521do { \
Ingo Molnare8f6fbf2008-11-12 01:38:36 +00001522 sk->sk_lock.owned = 0; \
Peter Zijlstraed075362006-12-06 20:35:24 -08001523 init_waitqueue_head(&sk->sk_lock.wq); \
1524 spin_lock_init(&(sk)->sk_lock.slock); \
1525 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1526 sizeof((sk)->sk_lock)); \
1527 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001528 (skey), (sname)); \
Peter Zijlstraed075362006-12-06 20:35:24 -08001529 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1530} while (0)
1531
Joe Perches69336bd2013-09-22 10:32:26 -07001532void lock_sock_nested(struct sock *sk, int subclass);
Peter Zijlstrafcc70d52006-11-08 22:44:35 -08001533
1534static inline void lock_sock(struct sock *sk)
1535{
1536 lock_sock_nested(sk, 0);
1537}
1538
Joe Perches69336bd2013-09-22 10:32:26 -07001539void release_sock(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001540
1541/* BH context may only use the following locking interface. */
1542#define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
Ingo Molnarc6366182006-07-03 00:25:13 -07001543#define bh_lock_sock_nested(__sk) \
1544 spin_lock_nested(&((__sk)->sk_lock.slock), \
1545 SINGLE_DEPTH_NESTING)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001546#define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1547
Joe Perches69336bd2013-09-22 10:32:26 -07001548bool lock_sock_fast(struct sock *sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001549/**
1550 * unlock_sock_fast - complement of lock_sock_fast
1551 * @sk: socket
1552 * @slow: slow mode
1553 *
1554 * fast unlock socket for user context.
1555 * If slow mode is on, we call regular release_sock()
1556 */
1557static inline void unlock_sock_fast(struct sock *sk, bool slow)
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001558{
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001559 if (slow)
1560 release_sock(sk);
1561 else
1562 spin_unlock_bh(&sk->sk_lock.slock);
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001563}
1564
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001565
Joe Perches69336bd2013-09-22 10:32:26 -07001566struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
1567 struct proto *prot);
1568void sk_free(struct sock *sk);
1569void sk_release_kernel(struct sock *sk);
1570struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001571
Joe Perches69336bd2013-09-22 10:32:26 -07001572struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
1573 gfp_t priority);
Joe Perches69336bd2013-09-22 10:32:26 -07001574void sock_wfree(struct sk_buff *skb);
1575void skb_orphan_partial(struct sk_buff *skb);
1576void sock_rfree(struct sk_buff *skb);
Alexander Duyck62bccb82014-09-04 13:31:35 -04001577void sock_efree(struct sk_buff *skb);
Alexander Duyck82eabd92014-09-04 13:32:11 -04001578#ifdef CONFIG_INET
Joe Perches69336bd2013-09-22 10:32:26 -07001579void sock_edemux(struct sk_buff *skb);
Alexander Duyck82eabd92014-09-04 13:32:11 -04001580#else
1581#define sock_edemux(skb) sock_efree(skb)
1582#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583
Joe Perches69336bd2013-09-22 10:32:26 -07001584int sock_setsockopt(struct socket *sock, int level, int op,
1585 char __user *optval, unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001586
Joe Perches69336bd2013-09-22 10:32:26 -07001587int sock_getsockopt(struct socket *sock, int level, int op,
1588 char __user *optval, int __user *optlen);
1589struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1590 int noblock, int *errcode);
1591struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1592 unsigned long data_len, int noblock,
1593 int *errcode, int max_page_order);
1594void *sock_kmalloc(struct sock *sk, int size, gfp_t priority);
1595void sock_kfree_s(struct sock *sk, void *mem, int size);
Daniel Borkmann79e88652014-11-19 17:13:11 +01001596void sock_kzfree_s(struct sock *sk, void *mem, int size);
Joe Perches69336bd2013-09-22 10:32:26 -07001597void sk_send_sigurg(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001598
1599/*
1600 * Functions to fill in entries in struct proto_ops when a protocol
1601 * does not implement a particular function.
1602 */
Joe Perches69336bd2013-09-22 10:32:26 -07001603int sock_no_bind(struct socket *, struct sockaddr *, int);
1604int sock_no_connect(struct socket *, struct sockaddr *, int, int);
1605int sock_no_socketpair(struct socket *, struct socket *);
1606int sock_no_accept(struct socket *, struct socket *, int);
1607int sock_no_getname(struct socket *, struct sockaddr *, int *, int);
1608unsigned int sock_no_poll(struct file *, struct socket *,
1609 struct poll_table_struct *);
1610int sock_no_ioctl(struct socket *, unsigned int, unsigned long);
1611int sock_no_listen(struct socket *, int);
1612int sock_no_shutdown(struct socket *, int);
1613int sock_no_getsockopt(struct socket *, int , int, char __user *, int __user *);
1614int sock_no_setsockopt(struct socket *, int, int, char __user *, unsigned int);
1615int sock_no_sendmsg(struct kiocb *, struct socket *, struct msghdr *, size_t);
1616int sock_no_recvmsg(struct kiocb *, struct socket *, struct msghdr *, size_t,
1617 int);
1618int sock_no_mmap(struct file *file, struct socket *sock,
1619 struct vm_area_struct *vma);
1620ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset,
1621 size_t size, int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622
1623/*
1624 * Functions to fill in entries in struct proto_ops when a protocol
1625 * uses the inet style.
1626 */
Joe Perches69336bd2013-09-22 10:32:26 -07001627int sock_common_getsockopt(struct socket *sock, int level, int optname,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001628 char __user *optval, int __user *optlen);
Joe Perches69336bd2013-09-22 10:32:26 -07001629int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630 struct msghdr *msg, size_t size, int flags);
Joe Perches69336bd2013-09-22 10:32:26 -07001631int sock_common_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07001632 char __user *optval, unsigned int optlen);
Joe Perches69336bd2013-09-22 10:32:26 -07001633int compat_sock_common_getsockopt(struct socket *sock, int level,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001634 int optname, char __user *optval, int __user *optlen);
Joe Perches69336bd2013-09-22 10:32:26 -07001635int compat_sock_common_setsockopt(struct socket *sock, int level,
David S. Millerb7058842009-09-30 16:12:20 -07001636 int optname, char __user *optval, unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001637
Joe Perches69336bd2013-09-22 10:32:26 -07001638void sk_common_release(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001639
1640/*
1641 * Default socket callbacks and setup code
1642 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001643
Linus Torvalds1da177e2005-04-16 15:20:36 -07001644/* Initialise core socket variables */
Joe Perches69336bd2013-09-22 10:32:26 -07001645void sock_init_data(struct socket *sock, struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001646
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647/*
1648 * Socket reference counting postulates.
1649 *
1650 * * Each user of socket SHOULD hold a reference count.
1651 * * Each access point to socket (an hash table bucket, reference from a list,
1652 * running timer, skb in flight MUST hold a reference count.
1653 * * When reference count hits 0, it means it will never increase back.
1654 * * When reference count hits 0, it means that no references from
1655 * outside exist to this socket and current process on current CPU
1656 * is last user and may/should destroy this socket.
1657 * * sk_free is called from any context: process, BH, IRQ. When
1658 * it is called, socket has no references from outside -> sk_free
1659 * may release descendant resources allocated by the socket, but
1660 * to the time when it is called, socket is NOT referenced by any
1661 * hash tables, lists etc.
1662 * * Packets, delivered from outside (from network or from another process)
1663 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1664 * when they sit in queue. Otherwise, packets will leak to hole, when
1665 * socket is looked up by one cpu and unhasing is made by another CPU.
1666 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1667 * (leak to backlog). Packet socket does all the processing inside
1668 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1669 * use separate SMP lock, so that they are prone too.
1670 */
1671
1672/* Ungrab socket and destroy it, if it was the last reference. */
1673static inline void sock_put(struct sock *sk)
1674{
1675 if (atomic_dec_and_test(&sk->sk_refcnt))
1676 sk_free(sk);
1677}
Eric Dumazet05dbc7b2013-10-03 00:22:02 -07001678/* Generic version of sock_put(), dealing with all sockets
1679 * (TCP_TIMEWAIT, ESTABLISHED...)
1680 */
1681void sock_gen_put(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001682
Joe Perches69336bd2013-09-22 10:32:26 -07001683int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested);
Arnaldo Carvalho de Melo25995ff2005-12-27 02:42:22 -02001684
Krishna Kumare022f0b2009-10-19 23:46:20 +00001685static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1686{
1687 sk->sk_tx_queue_mapping = tx_queue;
1688}
1689
1690static inline void sk_tx_queue_clear(struct sock *sk)
1691{
1692 sk->sk_tx_queue_mapping = -1;
1693}
1694
1695static inline int sk_tx_queue_get(const struct sock *sk)
1696{
Tom Herbertb0f77d02010-07-14 20:50:29 -07001697 return sk ? sk->sk_tx_queue_mapping : -1;
Krishna Kumare022f0b2009-10-19 23:46:20 +00001698}
1699
David S. Miller972692e2008-06-17 22:41:38 -07001700static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1701{
Krishna Kumare022f0b2009-10-19 23:46:20 +00001702 sk_tx_queue_clear(sk);
David S. Miller972692e2008-06-17 22:41:38 -07001703 sk->sk_socket = sock;
1704}
1705
Eric Dumazetaa395142010-04-20 13:03:51 +00001706static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1707{
Eric Dumazeteaefd112011-02-18 03:26:36 +00001708 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1709 return &rcu_dereference_raw(sk->sk_wq)->wait;
Eric Dumazetaa395142010-04-20 13:03:51 +00001710}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001711/* Detach socket from process context.
1712 * Announce socket dead, detach it from wait queue and inode.
1713 * Note that parent inode held reference count on this struct sock,
1714 * we do not release it in this function, because protocol
1715 * probably wants some additional cleanups or even continuing
1716 * to work with this socket (TCP).
1717 */
1718static inline void sock_orphan(struct sock *sk)
1719{
1720 write_lock_bh(&sk->sk_callback_lock);
1721 sock_set_flag(sk, SOCK_DEAD);
David S. Miller972692e2008-06-17 22:41:38 -07001722 sk_set_socket(sk, NULL);
Eric Dumazet43815482010-04-29 11:01:49 +00001723 sk->sk_wq = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724 write_unlock_bh(&sk->sk_callback_lock);
1725}
1726
1727static inline void sock_graft(struct sock *sk, struct socket *parent)
1728{
1729 write_lock_bh(&sk->sk_callback_lock);
Eric Dumazeteaefd112011-02-18 03:26:36 +00001730 sk->sk_wq = parent->wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731 parent->sk = sk;
David S. Miller972692e2008-06-17 22:41:38 -07001732 sk_set_socket(sk, parent);
Venkat Yekkirala4237c752006-07-24 23:32:50 -07001733 security_sock_graft(sk, parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001734 write_unlock_bh(&sk->sk_callback_lock);
1735}
1736
Joe Perches69336bd2013-09-22 10:32:26 -07001737kuid_t sock_i_uid(struct sock *sk);
1738unsigned long sock_i_ino(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001739
1740static inline struct dst_entry *
1741__sk_dst_get(struct sock *sk)
1742{
Michal Hockod8bf4ca2011-07-08 14:39:41 +02001743 return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
Eric Dumazetf68c2242010-04-22 16:06:59 -07001744 lockdep_is_held(&sk->sk_lock.slock));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001745}
1746
1747static inline struct dst_entry *
1748sk_dst_get(struct sock *sk)
1749{
1750 struct dst_entry *dst;
1751
Eric Dumazetb6c67122010-04-08 23:03:29 +00001752 rcu_read_lock();
1753 dst = rcu_dereference(sk->sk_dst_cache);
Eric Dumazetf8864972014-06-24 10:05:11 -07001754 if (dst && !atomic_inc_not_zero(&dst->__refcnt))
1755 dst = NULL;
Eric Dumazetb6c67122010-04-08 23:03:29 +00001756 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001757 return dst;
1758}
1759
Eric Dumazetb6c67122010-04-08 23:03:29 +00001760static inline void dst_negative_advice(struct sock *sk)
1761{
1762 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1763
1764 if (dst && dst->ops->negative_advice) {
1765 ndst = dst->ops->negative_advice(dst);
1766
1767 if (ndst != dst) {
1768 rcu_assign_pointer(sk->sk_dst_cache, ndst);
ZHAO Gang0a6957e2013-10-22 16:23:38 +08001769 sk_tx_queue_clear(sk);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001770 }
1771 }
1772}
1773
Linus Torvalds1da177e2005-04-16 15:20:36 -07001774static inline void
1775__sk_dst_set(struct sock *sk, struct dst_entry *dst)
1776{
1777 struct dst_entry *old_dst;
1778
Krishna Kumare022f0b2009-10-19 23:46:20 +00001779 sk_tx_queue_clear(sk);
Eric Dumazet0b53ff22010-04-26 20:40:43 +00001780 /*
1781 * This can be called while sk is owned by the caller only,
1782 * with no state that can be checked in a rcu_dereference_check() cond
1783 */
1784 old_dst = rcu_dereference_raw(sk->sk_dst_cache);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001785 rcu_assign_pointer(sk->sk_dst_cache, dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001786 dst_release(old_dst);
1787}
1788
1789static inline void
1790sk_dst_set(struct sock *sk, struct dst_entry *dst)
1791{
Eric Dumazet7f502362014-06-30 01:26:23 -07001792 struct dst_entry *old_dst;
1793
1794 sk_tx_queue_clear(sk);
Eric Dumazet5925a052014-07-02 02:39:38 -07001795 old_dst = xchg((__force struct dst_entry **)&sk->sk_dst_cache, dst);
Eric Dumazet7f502362014-06-30 01:26:23 -07001796 dst_release(old_dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001797}
1798
1799static inline void
1800__sk_dst_reset(struct sock *sk)
1801{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001802 __sk_dst_set(sk, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001803}
1804
1805static inline void
1806sk_dst_reset(struct sock *sk)
1807{
Eric Dumazet7f502362014-06-30 01:26:23 -07001808 sk_dst_set(sk, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001809}
1810
Joe Perches69336bd2013-09-22 10:32:26 -07001811struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001812
Joe Perches69336bd2013-09-22 10:32:26 -07001813struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001814
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001815static inline bool sk_can_gso(const struct sock *sk)
Herbert Xubcd76112006-06-30 13:36:35 -07001816{
1817 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1818}
1819
Joe Perches69336bd2013-09-22 10:32:26 -07001820void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
Arnaldo Carvalho de Melo6cbb0df2005-08-09 19:49:02 -07001821
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001822static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
Eric Dumazeta4654192010-05-16 00:36:33 -07001823{
1824 sk->sk_route_nocaps |= flags;
1825 sk->sk_route_caps &= ~flags;
1826}
1827
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001828static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1829 char __user *from, char *to,
Wei Yongjun912d3982011-04-06 18:40:12 +00001830 int copy, int offset)
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001831{
1832 if (skb->ip_summed == CHECKSUM_NONE) {
1833 int err = 0;
1834 __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err);
1835 if (err)
1836 return err;
Wei Yongjun912d3982011-04-06 18:40:12 +00001837 skb->csum = csum_block_add(skb->csum, csum, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001838 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1839 if (!access_ok(VERIFY_READ, from, copy) ||
1840 __copy_from_user_nocache(to, from, copy))
1841 return -EFAULT;
1842 } else if (copy_from_user(to, from, copy))
1843 return -EFAULT;
1844
1845 return 0;
1846}
1847
1848static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
1849 char __user *from, int copy)
1850{
Wei Yongjun912d3982011-04-06 18:40:12 +00001851 int err, offset = skb->len;
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001852
Wei Yongjun912d3982011-04-06 18:40:12 +00001853 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1854 copy, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001855 if (err)
Wei Yongjun912d3982011-04-06 18:40:12 +00001856 __skb_trim(skb, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001857
1858 return err;
1859}
1860
1861static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from,
1862 struct sk_buff *skb,
1863 struct page *page,
1864 int off, int copy)
1865{
1866 int err;
1867
Wei Yongjun912d3982011-04-06 18:40:12 +00001868 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1869 copy, skb->len);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001870 if (err)
1871 return err;
1872
1873 skb->len += copy;
1874 skb->data_len += copy;
1875 skb->truesize += copy;
1876 sk->sk_wmem_queued += copy;
1877 sk_mem_charge(sk, copy);
1878 return 0;
1879}
1880
Eric Dumazetc5640392009-06-16 10:12:03 +00001881/**
1882 * sk_wmem_alloc_get - returns write allocations
1883 * @sk: socket
1884 *
1885 * Returns sk_wmem_alloc minus initial offset of one
1886 */
1887static inline int sk_wmem_alloc_get(const struct sock *sk)
1888{
1889 return atomic_read(&sk->sk_wmem_alloc) - 1;
1890}
1891
1892/**
1893 * sk_rmem_alloc_get - returns read allocations
1894 * @sk: socket
1895 *
1896 * Returns sk_rmem_alloc
1897 */
1898static inline int sk_rmem_alloc_get(const struct sock *sk)
1899{
1900 return atomic_read(&sk->sk_rmem_alloc);
1901}
1902
1903/**
1904 * sk_has_allocations - check if allocations are outstanding
1905 * @sk: socket
1906 *
1907 * Returns true if socket has write or read allocations
1908 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001909static inline bool sk_has_allocations(const struct sock *sk)
Eric Dumazetc5640392009-06-16 10:12:03 +00001910{
1911 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1912}
1913
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001914/**
Eric Dumazet43815482010-04-29 11:01:49 +00001915 * wq_has_sleeper - check if there are any waiting processes
Randy Dunlapacfbe962010-05-24 23:54:18 -07001916 * @wq: struct socket_wq
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001917 *
Eric Dumazet43815482010-04-29 11:01:49 +00001918 * Returns true if socket_wq has waiting processes
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001919 *
Eric Dumazet43815482010-04-29 11:01:49 +00001920 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001921 * barrier call. They were added due to the race found within the tcp code.
1922 *
1923 * Consider following tcp code paths:
1924 *
1925 * CPU1 CPU2
1926 *
1927 * sys_select receive packet
1928 * ... ...
1929 * __add_wait_queue update tp->rcv_nxt
1930 * ... ...
1931 * tp->rcv_nxt check sock_def_readable
1932 * ... {
Eric Dumazet43815482010-04-29 11:01:49 +00001933 * schedule rcu_read_lock();
1934 * wq = rcu_dereference(sk->sk_wq);
1935 * if (wq && waitqueue_active(&wq->wait))
1936 * wake_up_interruptible(&wq->wait)
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001937 * ...
1938 * }
1939 *
1940 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1941 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1942 * could then endup calling schedule and sleep forever if there are no more
1943 * data on the socket.
Jiri Olsaad462762009-07-08 12:10:31 +00001944 *
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001945 */
Eric Dumazet43815482010-04-29 11:01:49 +00001946static inline bool wq_has_sleeper(struct socket_wq *wq)
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001947{
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001948 /* We need to be sure we are in sync with the
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001949 * add_wait_queue modifications to the wait queue.
1950 *
1951 * This memory barrier is paired in the sock_poll_wait.
1952 */
Eric Dumazet43815482010-04-29 11:01:49 +00001953 smp_mb();
1954 return wq && waitqueue_active(&wq->wait);
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001955}
1956
1957/**
1958 * sock_poll_wait - place memory barrier behind the poll_wait call.
1959 * @filp: file
1960 * @wait_address: socket wait queue
1961 * @p: poll_table
1962 *
Eric Dumazet43815482010-04-29 11:01:49 +00001963 * See the comments in the wq_has_sleeper function.
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001964 */
1965static inline void sock_poll_wait(struct file *filp,
1966 wait_queue_head_t *wait_address, poll_table *p)
1967{
Hans Verkuil626cf232012-03-23 15:02:27 -07001968 if (!poll_does_not_wait(p) && wait_address) {
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001969 poll_wait(filp, wait_address, p);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001970 /* We need to be sure we are in sync with the
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001971 * socket flags modification.
1972 *
Eric Dumazet43815482010-04-29 11:01:49 +00001973 * This memory barrier is paired in the wq_has_sleeper.
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001974 */
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001975 smp_mb();
1976 }
1977}
1978
Tom Herbertb73c3d02014-07-01 21:32:17 -07001979static inline void skb_set_hash_from_sk(struct sk_buff *skb, struct sock *sk)
1980{
1981 if (sk->sk_txhash) {
1982 skb->l4_hash = 1;
1983 skb->hash = sk->sk_txhash;
1984 }
1985}
1986
Linus Torvalds1da177e2005-04-16 15:20:36 -07001987/*
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001988 * Queue a received datagram if it will fit. Stream and sequenced
Linus Torvalds1da177e2005-04-16 15:20:36 -07001989 * protocols can't normally use this as they need to fit buffers in
1990 * and play with them.
1991 *
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001992 * Inlined as it's very short and called for pretty much every
Linus Torvalds1da177e2005-04-16 15:20:36 -07001993 * packet ever received.
1994 */
1995
1996static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1997{
Herbert Xud55d87f2009-06-22 02:25:25 +00001998 skb_orphan(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001999 skb->sk = sk;
2000 skb->destructor = sock_wfree;
Tom Herbertb73c3d02014-07-01 21:32:17 -07002001 skb_set_hash_from_sk(skb, sk);
Eric Dumazet2b85a342009-06-11 02:55:43 -07002002 /*
2003 * We used to take a refcount on sk, but following operation
2004 * is enough to guarantee sk_free() wont free this sock until
2005 * all in-flight packets are completed
2006 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002007 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
2008}
2009
2010static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
2011{
Herbert Xud55d87f2009-06-22 02:25:25 +00002012 skb_orphan(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002013 skb->sk = sk;
2014 skb->destructor = sock_rfree;
2015 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08002016 sk_mem_charge(sk, skb->truesize);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002017}
2018
Joe Perches69336bd2013-09-22 10:32:26 -07002019void sk_reset_timer(struct sock *sk, struct timer_list *timer,
2020 unsigned long expires);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021
Joe Perches69336bd2013-09-22 10:32:26 -07002022void sk_stop_timer(struct sock *sk, struct timer_list *timer);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002023
Joe Perches69336bd2013-09-22 10:32:26 -07002024int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002025
Joe Perches69336bd2013-09-22 10:32:26 -07002026int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
Willem de Bruijn364a9e92014-08-31 21:30:27 -04002027struct sk_buff *sock_dequeue_err_skb(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002028
2029/*
2030 * Recover an error report and clear atomically
2031 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002032
Linus Torvalds1da177e2005-04-16 15:20:36 -07002033static inline int sock_error(struct sock *sk)
2034{
Benjamin LaHaisec1cbe4b2005-12-13 23:22:19 -08002035 int err;
2036 if (likely(!sk->sk_err))
2037 return 0;
2038 err = xchg(&sk->sk_err, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002039 return -err;
2040}
2041
2042static inline unsigned long sock_wspace(struct sock *sk)
2043{
2044 int amt = 0;
2045
2046 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
2047 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002048 if (amt < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002049 amt = 0;
2050 }
2051 return amt;
2052}
2053
2054static inline void sk_wake_async(struct sock *sk, int how, int band)
2055{
Eric Dumazetbcdce712009-10-06 17:28:29 -07002056 if (sock_flag(sk, SOCK_FASYNC))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002057 sock_wake_async(sk->sk_socket, how, band);
2058}
2059
Daniel Borkmanneea86af2013-06-19 12:51:20 +02002060/* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
2061 * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
2062 * Note: for send buffers, TCP works better if we can build two skbs at
2063 * minimum.
Eric Dumazet7a91b4342010-09-26 18:53:07 -07002064 */
Eric Dumazet9eb5bf82013-07-03 05:02:22 -07002065#define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
Daniel Borkmanneea86af2013-06-19 12:51:20 +02002066
2067#define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2)
2068#define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE
Linus Torvalds1da177e2005-04-16 15:20:36 -07002069
2070static inline void sk_stream_moderate_sndbuf(struct sock *sk)
2071{
2072 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
Eric Dumazet8df09ea2007-12-21 03:07:41 -08002073 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
Daniel Borkmanneea86af2013-06-19 12:51:20 +02002074 sk->sk_sndbuf = max_t(u32, sk->sk_sndbuf, SOCK_MIN_SNDBUF);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002075 }
2076}
2077
Pavel Emelyanovdf97c702007-11-29 21:22:33 +11002078struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002079
Eric Dumazet5640f762012-09-23 23:04:42 +00002080/**
2081 * sk_page_frag - return an appropriate page_frag
2082 * @sk: socket
2083 *
2084 * If socket allocation mode allows current thread to sleep, it means its
2085 * safe to use the per task page_frag instead of the per socket one.
2086 */
2087static inline struct page_frag *sk_page_frag(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002088{
Eric Dumazet5640f762012-09-23 23:04:42 +00002089 if (sk->sk_allocation & __GFP_WAIT)
2090 return &current->task_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002091
Eric Dumazet5640f762012-09-23 23:04:42 +00002092 return &sk->sk_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002093}
2094
Joe Perches69336bd2013-09-22 10:32:26 -07002095bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag);
Eric Dumazet5640f762012-09-23 23:04:42 +00002096
Linus Torvalds1da177e2005-04-16 15:20:36 -07002097/*
2098 * Default write policy as shown to user space via poll/select/SIGIO
2099 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002100static inline bool sock_writeable(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002101{
Eric Dumazet8df09ea2007-12-21 03:07:41 -08002102 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002103}
2104
Al Virodd0fc662005-10-07 07:46:04 +01002105static inline gfp_t gfp_any(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002106{
Andrew Morton99709372009-02-12 16:43:17 -08002107 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002108}
2109
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002110static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002111{
2112 return noblock ? 0 : sk->sk_rcvtimeo;
2113}
2114
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002115static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002116{
2117 return noblock ? 0 : sk->sk_sndtimeo;
2118}
2119
2120static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2121{
2122 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
2123}
2124
2125/* Alas, with timeout socket operations are not restartable.
2126 * Compare this to poll().
2127 */
2128static inline int sock_intr_errno(long timeo)
2129{
2130 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2131}
2132
Joe Perches69336bd2013-09-22 10:32:26 -07002133void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2134 struct sk_buff *skb);
2135void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2136 struct sk_buff *skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -07002137
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002138static inline void
Linus Torvalds1da177e2005-04-16 15:20:36 -07002139sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2140{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002141 ktime_t kt = skb->tstamp;
Patrick Ohly20d49472009-02-12 05:03:38 +00002142 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002143
Patrick Ohly20d49472009-02-12 05:03:38 +00002144 /*
2145 * generate control messages if
Willem de Bruijnb9f40e22014-08-04 22:11:46 -04002146 * - receive time stamping in software requested
Patrick Ohly20d49472009-02-12 05:03:38 +00002147 * - software time stamp available and wanted
Patrick Ohly20d49472009-02-12 05:03:38 +00002148 * - hardware time stamps available and wanted
Patrick Ohly20d49472009-02-12 05:03:38 +00002149 */
2150 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
Willem de Bruijnb9f40e22014-08-04 22:11:46 -04002151 (sk->sk_tsflags & SOF_TIMESTAMPING_RX_SOFTWARE) ||
Willem de Bruijnc1991052014-09-03 12:01:18 -04002152 (kt.tv64 && sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) ||
Patrick Ohly20d49472009-02-12 05:03:38 +00002153 (hwtstamps->hwtstamp.tv64 &&
Willem de Bruijnb9f40e22014-08-04 22:11:46 -04002154 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE)))
Eric Dumazet92f37fd2007-03-25 22:14:49 -07002155 __sock_recv_timestamp(msg, sk, skb);
2156 else
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002157 sk->sk_stamp = kt;
Johannes Berg6e3e9392011-11-09 10:15:42 +01002158
2159 if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2160 __sock_recv_wifi_status(msg, sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002161}
2162
Joe Perches69336bd2013-09-22 10:32:26 -07002163void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2164 struct sk_buff *skb);
Eric Dumazet767dd032010-04-28 19:14:43 +00002165
2166static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2167 struct sk_buff *skb)
2168{
2169#define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
Willem de Bruijnb9f40e22014-08-04 22:11:46 -04002170 (1UL << SOCK_RCVTSTAMP))
2171#define TSFLAGS_ANY (SOF_TIMESTAMPING_SOFTWARE | \
2172 SOF_TIMESTAMPING_RAW_HARDWARE)
Eric Dumazet767dd032010-04-28 19:14:43 +00002173
Willem de Bruijnb9f40e22014-08-04 22:11:46 -04002174 if (sk->sk_flags & FLAGS_TS_OR_DROPS || sk->sk_tsflags & TSFLAGS_ANY)
Eric Dumazet767dd032010-04-28 19:14:43 +00002175 __sock_recv_ts_and_drops(msg, sk, skb);
2176 else
2177 sk->sk_stamp = skb->tstamp;
2178}
Neil Horman3b885782009-10-12 13:26:31 -07002179
Willem de Bruijn67cc0d42014-09-08 19:58:58 -04002180void __sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags);
2181
Linus Torvalds1da177e2005-04-16 15:20:36 -07002182/**
Patrick Ohly20d49472009-02-12 05:03:38 +00002183 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
Patrick Ohly20d49472009-02-12 05:03:38 +00002184 * @sk: socket sending this packet
Eric Dumazet140c55d2014-08-06 11:49:29 +02002185 * @tx_flags: completed with instructions for time stamping
2186 *
2187 * Note : callers should take care of initial *tx_flags value (usually 0)
Patrick Ohly20d49472009-02-12 05:03:38 +00002188 */
Willem de Bruijn67cc0d42014-09-08 19:58:58 -04002189static inline void sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags)
2190{
2191 if (unlikely(sk->sk_tsflags))
2192 __sock_tx_timestamp(sk, tx_flags);
2193 if (unlikely(sock_flag(sk, SOCK_WIFI_STATUS)))
2194 *tx_flags |= SKBTX_WIFI_STATUS;
2195}
Patrick Ohly20d49472009-02-12 05:03:38 +00002196
2197/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002198 * sk_eat_skb - Release a skb if it is no longer needed
Pavel Pisa4dc3b162005-05-01 08:59:25 -07002199 * @sk: socket to eat this skb from
2200 * @skb: socket buffer to eat
Linus Torvalds1da177e2005-04-16 15:20:36 -07002201 *
2202 * This routine must be called with interrupts disabled or with the socket
2203 * locked so that the sk_buff queue operation is ok.
2204*/
Dan Williams7bced392013-12-30 12:37:29 -08002205static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002206{
2207 __skb_unlink(skb, &sk->sk_receive_queue);
2208 __kfree_skb(skb);
2209}
2210
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002211static inline
2212struct net *sock_net(const struct sock *sk)
2213{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002214 return read_pnet(&sk->sk_net);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002215}
2216
2217static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07002218void sock_net_set(struct sock *sk, struct net *net)
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002219{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002220 write_pnet(&sk->sk_net, net);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002221}
2222
Denis V. Lunevedf02082008-02-29 11:18:32 -08002223/*
2224 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
Lucas De Marchi25985ed2011-03-30 22:57:33 -03002225 * They should not hold a reference to a namespace in order to allow
Denis V. Lunevedf02082008-02-29 11:18:32 -08002226 * to stop it.
2227 * Sockets after sk_change_net should be released using sk_release_kernel
2228 */
2229static inline void sk_change_net(struct sock *sk, struct net *net)
2230{
Gu Zheng5812521b2014-03-10 09:57:34 +08002231 struct net *current_net = sock_net(sk);
2232
2233 if (!net_eq(current_net, net)) {
2234 put_net(current_net);
2235 sock_net_set(sk, hold_net(net));
2236 }
Denis V. Lunevedf02082008-02-29 11:18:32 -08002237}
2238
KOVACS Krisztian23542612008-10-07 12:41:01 -07002239static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2240{
Vijay Subramanianefc27f82012-06-24 13:03:07 +00002241 if (skb->sk) {
KOVACS Krisztian23542612008-10-07 12:41:01 -07002242 struct sock *sk = skb->sk;
2243
2244 skb->destructor = NULL;
2245 skb->sk = NULL;
2246 return sk;
2247 }
2248 return NULL;
2249}
2250
Joe Perches69336bd2013-09-22 10:32:26 -07002251void sock_enable_timestamp(struct sock *sk, int flag);
2252int sock_get_timestamp(struct sock *, struct timeval __user *);
2253int sock_get_timestampns(struct sock *, struct timespec __user *);
2254int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len, int level,
2255 int type);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002256
Eric W. Biedermana3b299d2014-04-23 14:26:56 -07002257bool sk_ns_capable(const struct sock *sk,
2258 struct user_namespace *user_ns, int cap);
2259bool sk_capable(const struct sock *sk, int cap);
2260bool sk_net_capable(const struct sock *sk, int cap);
2261
Linus Torvalds1da177e2005-04-16 15:20:36 -07002262extern __u32 sysctl_wmem_max;
2263extern __u32 sysctl_rmem_max;
2264
David S. Miller6baf1f42005-09-05 18:14:11 -07002265extern int sysctl_optmem_max;
2266
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002267extern __u32 sysctl_wmem_default;
2268extern __u32 sysctl_rmem_default;
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002269
Linus Torvalds1da177e2005-04-16 15:20:36 -07002270#endif /* _SOCK_H */