blob: 02f5b35e65f18675d758322f4d0be9b91cf7fcd1 [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>
Glauber Costa180d8cd2011-12-11 21:47:02 +000057#include <linux/memcontrol.h>
Glauber Costae1aab162011-12-11 21:47:03 +000058#include <linux/res_counter.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>
70
Eric Dumazet9f048bf2011-12-13 03:59:08 +000071struct cgroup;
72struct cgroup_subsys;
Glauber Costac607b2e2011-12-16 00:52:00 +000073#ifdef CONFIG_NET
Glauber Costa1d62e432012-04-09 19:36:33 -030074int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss);
75void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg);
Glauber Costac607b2e2011-12-16 00:52:00 +000076#else
77static inline
Glauber Costa1d62e432012-04-09 19:36:33 -030078int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
Glauber Costac607b2e2011-12-16 00:52:00 +000079{
80 return 0;
81}
82static inline
Glauber Costa1d62e432012-04-09 19:36:33 -030083void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg)
Glauber Costac607b2e2011-12-16 00:52:00 +000084{
85}
86#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -070087/*
88 * This structure really needs to be cleaned up.
89 * Most of it is for TCP, and not used by any of
90 * the other protocols.
91 */
92
93/* Define this to get the SOCK_DBG debugging facility. */
94#define SOCK_DEBUGGING
95#ifdef SOCK_DEBUGGING
96#define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
97 printk(KERN_DEBUG msg); } while (0)
98#else
Stephen Hemminger4cd90292008-03-21 15:54:53 -070099/* Validate arguments and do nothing */
Joe Perchesb9075fa2011-10-31 17:11:33 -0700100static inline __printf(2, 3)
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000101void SOCK_DEBUG(const struct sock *sk, const char *msg, ...)
Stephen Hemminger4cd90292008-03-21 15:54:53 -0700102{
103}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104#endif
105
106/* This is the per-socket lock. The spinlock provides a synchronization
107 * between user contexts and software interrupt processing, whereas the
108 * mini-semaphore synchronizes multiple users amongst themselves.
109 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700110typedef struct {
111 spinlock_t slock;
John Heffnerd2e91172007-09-12 10:44:19 +0200112 int owned;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113 wait_queue_head_t wq;
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700114 /*
115 * We express the mutex-alike socket_lock semantics
116 * to the lock validator by explicitly managing
117 * the slock as a lock variant (in addition to
118 * the slock itself):
119 */
120#ifdef CONFIG_DEBUG_LOCK_ALLOC
121 struct lockdep_map dep_map;
122#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123} socket_lock_t;
124
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125struct sock;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700126struct proto;
Denis V. Lunev0eeb8ff2007-12-04 01:15:45 -0800127struct net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128
Eric Dumazet077b3932012-12-02 07:33:10 +0000129typedef __u32 __bitwise __portpair;
130typedef __u64 __bitwise __addrpair;
131
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132/**
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700133 * struct sock_common - minimal network layer representation of sockets
Eric Dumazet68835ab2010-11-30 19:04:07 +0000134 * @skc_daddr: Foreign IPv4 addr
135 * @skc_rcv_saddr: Bound local IPv4 addr
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000136 * @skc_hash: hash value used with various protocol lookup tables
Eric Dumazetd4cada42009-11-08 10:17:30 +0000137 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
Eric Dumazetce43b032012-11-30 09:49:27 +0000138 * @skc_dport: placeholder for inet_dport/tw_dport
139 * @skc_num: placeholder for inet_num/tw_num
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700140 * @skc_family: network address family
141 * @skc_state: Connection state
142 * @skc_reuse: %SO_REUSEADDR setting
Tom Herbert055dc212013-01-22 09:49:50 +0000143 * @skc_reuseport: %SO_REUSEPORT setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700144 * @skc_bound_dev_if: bound device index if != 0
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700145 * @skc_bind_node: bind hash linkage for various protocol lookup tables
Eric Dumazet512615b2009-11-08 10:17:58 +0000146 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700147 * @skc_prot: protocol handlers inside a network family
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200148 * @skc_net: reference to the network namespace of this socket
Eric Dumazet68835ab2010-11-30 19:04:07 +0000149 * @skc_node: main hash linkage for various protocol lookup tables
150 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
151 * @skc_tx_queue_mapping: tx queue number for this connection
152 * @skc_refcnt: reference count
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700153 *
154 * This is the minimal network layer representation of sockets, the header
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700155 * for struct sock and struct inet_timewait_sock.
156 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157struct sock_common {
Eric Dumazetce43b032012-11-30 09:49:27 +0000158 /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned
Eric Dumazet05dbc7b2013-10-03 00:22:02 -0700159 * address on 64bit arches : cf INET_MATCH()
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000160 */
Eric Dumazetce43b032012-11-30 09:49:27 +0000161 union {
Eric Dumazet077b3932012-12-02 07:33:10 +0000162 __addrpair skc_addrpair;
Eric Dumazetce43b032012-11-30 09:49:27 +0000163 struct {
164 __be32 skc_daddr;
165 __be32 skc_rcv_saddr;
166 };
167 };
Eric Dumazetd4cada42009-11-08 10:17:30 +0000168 union {
169 unsigned int skc_hash;
170 __u16 skc_u16hashes[2];
171 };
Eric Dumazetce43b032012-11-30 09:49:27 +0000172 /* skc_dport && skc_num must be grouped as well */
173 union {
Eric Dumazet077b3932012-12-02 07:33:10 +0000174 __portpair skc_portpair;
Eric Dumazetce43b032012-11-30 09:49:27 +0000175 struct {
176 __be16 skc_dport;
177 __u16 skc_num;
178 };
179 };
180
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000181 unsigned short skc_family;
182 volatile unsigned char skc_state;
Tom Herbert055dc212013-01-22 09:49:50 +0000183 unsigned char skc_reuse:4;
Eric Dumazet9fe516b2014-06-27 08:36:16 -0700184 unsigned char skc_reuseport:1;
185 unsigned char skc_ipv6only:1;
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000186 int skc_bound_dev_if;
Eric Dumazet512615b2009-11-08 10:17:58 +0000187 union {
188 struct hlist_node skc_bind_node;
189 struct hlist_nulls_node skc_portaddr_node;
190 };
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700191 struct proto *skc_prot;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900192#ifdef CONFIG_NET_NS
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200193 struct net *skc_net;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900194#endif
Eric Dumazetefe42082013-10-03 15:42:29 -0700195
196#if IS_ENABLED(CONFIG_IPV6)
197 struct in6_addr skc_v6_daddr;
198 struct in6_addr skc_v6_rcv_saddr;
199#endif
200
Eric Dumazet68835ab2010-11-30 19:04:07 +0000201 /*
202 * fields between dontcopy_begin/dontcopy_end
203 * are not copied in sock_copy()
204 */
Randy Dunlap928c41e2011-01-08 17:39:21 +0000205 /* private: */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000206 int skc_dontcopy_begin[0];
Randy Dunlap928c41e2011-01-08 17:39:21 +0000207 /* public: */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000208 union {
209 struct hlist_node skc_node;
210 struct hlist_nulls_node skc_nulls_node;
211 };
212 int skc_tx_queue_mapping;
213 atomic_t skc_refcnt;
Randy Dunlap928c41e2011-01-08 17:39:21 +0000214 /* private: */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000215 int skc_dontcopy_end[0];
Randy Dunlap928c41e2011-01-08 17:39:21 +0000216 /* public: */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217};
218
Glauber Costae1aab162011-12-11 21:47:03 +0000219struct cg_proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220/**
221 * struct sock - network layer representation of sockets
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700222 * @__sk_common: shared layout with inet_timewait_sock
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700223 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
224 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
225 * @sk_lock: synchronizer
226 * @sk_rcvbuf: size of receive buffer in bytes
Eric Dumazet43815482010-04-29 11:01:49 +0000227 * @sk_wq: sock wait queue and async head
Shawn Bohrer421b3882013-10-07 11:01:39 -0500228 * @sk_rx_dst: receive input route used by early demux
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700229 * @sk_dst_cache: destination cache
230 * @sk_dst_lock: destination cache lock
231 * @sk_policy: flow policy
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700232 * @sk_receive_queue: incoming packets
233 * @sk_wmem_alloc: transmit queue bytes committed
234 * @sk_write_queue: Packet sending queue
Chris Leech97fc2f02006-05-23 17:55:33 -0700235 * @sk_async_wait_queue: DMA copied packets
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700236 * @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
Tom Herbertb73c3d02014-07-01 21:32:17 -0700276 * @sk_txhash: computed flow hash for use on transmit
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700277 * @sk_filter: socket filtering instructions
278 * @sk_protinfo: private area, net family specific, when not using slab
279 * @sk_timer: sock cleanup timer
280 * @sk_stamp: time stamp of last packet received
281 * @sk_socket: Identd and reporting IO signals
282 * @sk_user_data: RPC layer private data
Eric Dumazet5640f762012-09-23 23:04:42 +0000283 * @sk_frag: cached page frag
Randy Dunlapd3d4f0a2012-04-17 14:03:53 +0000284 * @sk_peek_off: current peek_offset value
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700285 * @sk_send_head: front of stuff to transmit
Martin Waitz67be2dd2005-05-01 08:59:26 -0700286 * @sk_security: used by security modules
Randy Dunlap31729362008-02-18 20:52:13 -0800287 * @sk_mark: generic packet mark
Randy Dunlap53c3fa22010-08-09 13:41:07 +0000288 * @sk_classid: this socket's cgroup classid
Glauber Costae1aab162011-12-11 21:47:03 +0000289 * @sk_cgrp: this socket's cgroup-specific proto data
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700290 * @sk_write_pending: a write to stream socket waits to start
291 * @sk_state_change: callback to indicate change in the state of the sock
292 * @sk_data_ready: callback to indicate there is data to be processed
293 * @sk_write_space: callback to indicate there is bf sending space available
294 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
295 * @sk_backlog_rcv: callback to process the backlog
296 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297 */
298struct sock {
299 /*
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700300 * Now struct inet_timewait_sock also uses sock_common, so please just
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301 * don't add nothing before this first member (__sk_common) --acme
302 */
303 struct sock_common __sk_common;
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000304#define sk_node __sk_common.skc_node
305#define sk_nulls_node __sk_common.skc_nulls_node
306#define sk_refcnt __sk_common.skc_refcnt
Krishna Kumare022f0b2009-10-19 23:46:20 +0000307#define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000308
Eric Dumazet68835ab2010-11-30 19:04:07 +0000309#define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
310#define sk_dontcopy_end __sk_common.skc_dontcopy_end
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000311#define sk_hash __sk_common.skc_hash
Eric Dumazet50805462013-10-02 04:29:50 -0700312#define sk_portpair __sk_common.skc_portpair
Eric Dumazet05dbc7b2013-10-03 00:22:02 -0700313#define sk_num __sk_common.skc_num
314#define sk_dport __sk_common.skc_dport
Eric Dumazet50805462013-10-02 04:29:50 -0700315#define sk_addrpair __sk_common.skc_addrpair
316#define sk_daddr __sk_common.skc_daddr
317#define sk_rcv_saddr __sk_common.skc_rcv_saddr
Linus Torvalds1da177e2005-04-16 15:20:36 -0700318#define sk_family __sk_common.skc_family
319#define sk_state __sk_common.skc_state
320#define sk_reuse __sk_common.skc_reuse
Tom Herbert055dc212013-01-22 09:49:50 +0000321#define sk_reuseport __sk_common.skc_reuseport
Eric Dumazet9fe516b2014-06-27 08:36:16 -0700322#define sk_ipv6only __sk_common.skc_ipv6only
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323#define sk_bound_dev_if __sk_common.skc_bound_dev_if
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324#define sk_bind_node __sk_common.skc_bind_node
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700325#define sk_prot __sk_common.skc_prot
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200326#define sk_net __sk_common.skc_net
Eric Dumazetefe42082013-10-03 15:42:29 -0700327#define sk_v6_daddr __sk_common.skc_v6_daddr
328#define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr
329
Eric Dumazetb178bb32010-11-16 05:56:04 +0000330 socket_lock_t sk_lock;
331 struct sk_buff_head sk_receive_queue;
332 /*
333 * The backlog queue is special, it is always used with
334 * the per-socket spinlock held and requires low latency
335 * access. Therefore we special case it's implementation.
336 * Note : rmem_alloc is in this structure to fill a hole
337 * on 64bit arches, not because its logically part of
338 * backlog.
339 */
340 struct {
341 atomic_t rmem_alloc;
342 int len;
343 struct sk_buff *head;
344 struct sk_buff *tail;
345 } sk_backlog;
346#define sk_rmem_alloc sk_backlog.rmem_alloc
347 int sk_forward_alloc;
348#ifdef CONFIG_RPS
349 __u32 sk_rxhash;
350#endif
Tom Herbertb73c3d02014-07-01 21:32:17 -0700351 __u32 sk_txhash;
Cong Wange0d10952013-08-01 11:10:25 +0800352#ifdef CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir06021292013-06-10 11:39:50 +0300353 unsigned int sk_napi_id;
Eliezer Tamirdafcc432013-06-14 16:33:57 +0300354 unsigned int sk_ll_usec;
Eliezer Tamir06021292013-06-10 11:39:50 +0300355#endif
Eric Dumazetb178bb32010-11-16 05:56:04 +0000356 atomic_t sk_drops;
357 int sk_rcvbuf;
358
359 struct sk_filter __rcu *sk_filter;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000360 struct socket_wq __rcu *sk_wq;
Eric Dumazetb178bb32010-11-16 05:56:04 +0000361
362#ifdef CONFIG_NET_DMA
363 struct sk_buff_head sk_async_wait_queue;
364#endif
365
366#ifdef CONFIG_XFRM
367 struct xfrm_policy *sk_policy[2];
368#endif
369 unsigned long sk_flags;
Eric Dumazetdeaa5852012-06-24 20:22:49 +0000370 struct dst_entry *sk_rx_dst;
Cong Wang0e36cbb2013-01-22 21:09:51 +0000371 struct dst_entry __rcu *sk_dst_cache;
Eric Dumazetb178bb32010-11-16 05:56:04 +0000372 spinlock_t sk_dst_lock;
373 atomic_t sk_wmem_alloc;
374 atomic_t sk_omem_alloc;
375 int sk_sndbuf;
376 struct sk_buff_head sk_write_queue;
Vegard Nossuma98b65a2009-02-26 14:46:57 +0100377 kmemcheck_bitfield_begin(flags);
Eric Dumazet5fdb9972009-10-08 22:50:25 +0000378 unsigned int sk_shutdown : 2,
Tom Herbert28448b82014-05-23 08:47:19 -0700379 sk_no_check_tx : 1,
380 sk_no_check_rx : 1,
Eric Dumazet5fdb9972009-10-08 22:50:25 +0000381 sk_userlocks : 4,
382 sk_protocol : 8,
383 sk_type : 16;
Vegard Nossuma98b65a2009-02-26 14:46:57 +0100384 kmemcheck_bitfield_end(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385 int sk_wmem_queued;
Al Viro7d877f32005-10-21 03:20:43 -0400386 gfp_t sk_allocation;
Eric Dumazet95bd09e2013-08-27 05:46:32 -0700387 u32 sk_pacing_rate; /* bytes per second */
Eric Dumazet62748f32013-09-24 08:20:52 -0700388 u32 sk_max_pacing_rate;
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000389 netdev_features_t sk_route_caps;
390 netdev_features_t sk_route_nocaps;
Herbert Xubcd76112006-06-30 13:36:35 -0700391 int sk_gso_type;
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -0700392 unsigned int sk_gso_max_size;
Ben Hutchings14853482012-07-30 16:11:42 +0000393 u16 sk_gso_max_segs;
David S. Miller9932cf92006-03-24 15:12:37 -0800394 int sk_rcvlowat;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395 unsigned long sk_lingertime;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396 struct sk_buff_head sk_error_queue;
Arnaldo Carvalho de Melo476e19c2005-05-05 13:35:15 -0700397 struct proto *sk_prot_creator;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398 rwlock_t sk_callback_lock;
399 int sk_err,
400 sk_err_soft;
401 unsigned short sk_ack_backlog;
402 unsigned short sk_max_ack_backlog;
403 __u32 sk_priority;
Daniel Borkmann86f85152013-12-29 17:27:11 +0100404#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
Neil Horman5bc14212011-11-22 05:10:51 +0000405 __u32 sk_cgrp_prioidx;
406#endif
Eric W. Biederman109f6e32010-06-13 03:30:14 +0000407 struct pid *sk_peer_pid;
408 const struct cred *sk_peer_cred;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409 long sk_rcvtimeo;
410 long sk_sndtimeo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411 void *sk_protinfo;
412 struct timer_list sk_timer;
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700413 ktime_t sk_stamp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414 struct socket *sk_socket;
415 void *sk_user_data;
Eric Dumazet5640f762012-09-23 23:04:42 +0000416 struct page_frag sk_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417 struct sk_buff *sk_send_head;
Pavel Emelyanovef64a542012-02-21 07:31:34 +0000418 __s32 sk_peek_off;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419 int sk_write_pending;
Alexey Dobriyand5f64232008-11-04 14:45:58 -0800420#ifdef CONFIG_SECURITY
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 void *sk_security;
Alexey Dobriyand5f64232008-11-04 14:45:58 -0800422#endif
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800423 __u32 sk_mark;
Herbert Xuf8451722010-05-24 00:12:34 -0700424 u32 sk_classid;
Glauber Costae1aab162011-12-11 21:47:03 +0000425 struct cg_proto *sk_cgrp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426 void (*sk_state_change)(struct sock *sk);
David S. Miller676d2362014-04-11 16:15:36 -0400427 void (*sk_data_ready)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428 void (*sk_write_space)(struct sock *sk);
429 void (*sk_error_report)(struct sock *sk);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000430 int (*sk_backlog_rcv)(struct sock *sk,
431 struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432 void (*sk_destruct)(struct sock *sk);
433};
434
Pravin B Shelar559835ea2013-09-24 10:25:40 -0700435#define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
436
437#define rcu_dereference_sk_user_data(sk) rcu_dereference(__sk_user_data((sk)))
438#define rcu_assign_sk_user_data(sk, ptr) rcu_assign_pointer(__sk_user_data((sk)), ptr)
439
Pavel Emelyanov4a17fd52012-04-19 03:39:36 +0000440/*
441 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
442 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
443 * on a socket means that the socket will reuse everybody else's port
444 * without looking at the other's sk_reuse value.
445 */
446
447#define SK_NO_REUSE 0
448#define SK_CAN_REUSE 1
449#define SK_FORCE_REUSE 2
450
Pavel Emelyanovef64a542012-02-21 07:31:34 +0000451static inline int sk_peek_offset(struct sock *sk, int flags)
452{
453 if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0))
454 return sk->sk_peek_off;
455 else
456 return 0;
457}
458
459static inline void sk_peek_offset_bwd(struct sock *sk, int val)
460{
461 if (sk->sk_peek_off >= 0) {
462 if (sk->sk_peek_off >= val)
463 sk->sk_peek_off -= val;
464 else
465 sk->sk_peek_off = 0;
466 }
467}
468
469static inline void sk_peek_offset_fwd(struct sock *sk, int val)
470{
471 if (sk->sk_peek_off >= 0)
472 sk->sk_peek_off += val;
473}
474
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475/*
476 * Hashed lists helper routines
477 */
Li Zefanc4146642010-02-08 23:18:45 +0000478static inline struct sock *sk_entry(const struct hlist_node *node)
479{
480 return hlist_entry(node, struct sock, sk_node);
481}
482
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700483static inline struct sock *__sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484{
485 return hlist_entry(head->first, struct sock, sk_node);
486}
487
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700488static inline struct sock *sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489{
490 return hlist_empty(head) ? NULL : __sk_head(head);
491}
492
Eric Dumazet88ab1932008-11-16 19:39:21 -0800493static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
494{
495 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
496}
497
498static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
499{
500 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
501}
502
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700503static inline struct sock *sk_next(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504{
505 return sk->sk_node.next ?
506 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
507}
508
Eric Dumazet88ab1932008-11-16 19:39:21 -0800509static inline struct sock *sk_nulls_next(const struct sock *sk)
510{
511 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
512 hlist_nulls_entry(sk->sk_nulls_node.next,
513 struct sock, sk_nulls_node) :
514 NULL;
515}
516
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000517static inline bool sk_unhashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518{
519 return hlist_unhashed(&sk->sk_node);
520}
521
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000522static inline bool sk_hashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523{
Akinobu Mitada753be2006-04-28 15:21:23 -0700524 return !sk_unhashed(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525}
526
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000527static inline void sk_node_init(struct hlist_node *node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528{
529 node->pprev = NULL;
530}
531
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000532static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800533{
534 node->pprev = NULL;
535}
536
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000537static inline void __sk_del_node(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538{
539 __hlist_del(&sk->sk_node);
540}
541
stephen hemminger808f5112010-02-22 07:57:18 +0000542/* NB: equivalent to hlist_del_init_rcu */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000543static inline bool __sk_del_node_init(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544{
545 if (sk_hashed(sk)) {
546 __sk_del_node(sk);
547 sk_node_init(&sk->sk_node);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000548 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549 }
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000550 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551}
552
553/* Grab socket reference count. This operation is valid only
554 when sk is ALREADY grabbed f.e. it is found in hash table
555 or a list and the lookup is made under lock preventing hash table
556 modifications.
557 */
558
559static inline void sock_hold(struct sock *sk)
560{
561 atomic_inc(&sk->sk_refcnt);
562}
563
564/* Ungrab socket in the context, which assumes that socket refcnt
565 cannot hit zero, f.e. it is true in context of any socketcall.
566 */
567static inline void __sock_put(struct sock *sk)
568{
569 atomic_dec(&sk->sk_refcnt);
570}
571
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000572static inline bool sk_del_node_init(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573{
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000574 bool rc = __sk_del_node_init(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575
576 if (rc) {
577 /* paranoid for a while -acme */
578 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
579 __sock_put(sk);
580 }
581 return rc;
582}
stephen hemminger808f5112010-02-22 07:57:18 +0000583#define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000585static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700586{
587 if (sk_hashed(sk)) {
Eric Dumazet88ab1932008-11-16 19:39:21 -0800588 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000589 return true;
Eric Dumazet271b72c2008-10-29 02:11:14 -0700590 }
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000591 return false;
Eric Dumazet271b72c2008-10-29 02:11:14 -0700592}
593
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000594static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700595{
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000596 bool rc = __sk_nulls_del_node_init_rcu(sk);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700597
598 if (rc) {
599 /* paranoid for a while -acme */
600 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
601 __sock_put(sk);
602 }
603 return rc;
604}
605
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000606static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607{
608 hlist_add_head(&sk->sk_node, list);
609}
610
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000611static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612{
613 sock_hold(sk);
614 __sk_add_node(sk, list);
615}
616
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000617static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
stephen hemminger808f5112010-02-22 07:57:18 +0000618{
619 sock_hold(sk);
620 hlist_add_head_rcu(&sk->sk_node, list);
621}
622
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000623static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700624{
Eric Dumazet88ab1932008-11-16 19:39:21 -0800625 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700626}
627
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000628static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700629{
630 sock_hold(sk);
Eric Dumazet88ab1932008-11-16 19:39:21 -0800631 __sk_nulls_add_node_rcu(sk, list);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700632}
633
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000634static inline void __sk_del_bind_node(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635{
636 __hlist_del(&sk->sk_bind_node);
637}
638
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000639static inline void sk_add_bind_node(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640 struct hlist_head *list)
641{
642 hlist_add_head(&sk->sk_bind_node, list);
643}
644
Sasha Levinb67bfe02013-02-27 17:06:00 -0800645#define sk_for_each(__sk, list) \
646 hlist_for_each_entry(__sk, list, sk_node)
647#define sk_for_each_rcu(__sk, list) \
648 hlist_for_each_entry_rcu(__sk, list, sk_node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800649#define sk_nulls_for_each(__sk, node, list) \
650 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
651#define sk_nulls_for_each_rcu(__sk, node, list) \
652 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
Sasha Levinb67bfe02013-02-27 17:06:00 -0800653#define sk_for_each_from(__sk) \
654 hlist_for_each_entry_from(__sk, sk_node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800655#define sk_nulls_for_each_from(__sk, node) \
656 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
657 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
Sasha Levinb67bfe02013-02-27 17:06:00 -0800658#define sk_for_each_safe(__sk, tmp, list) \
659 hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
660#define sk_for_each_bound(__sk, list) \
661 hlist_for_each_entry(__sk, list, sk_bind_node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662
David Held2dc41cf2014-07-15 23:28:32 -0400663/**
664 * sk_nulls_for_each_entry_offset - iterate over a list at a given struct offset
665 * @tpos: the type * to use as a loop cursor.
666 * @pos: the &struct hlist_node to use as a loop cursor.
667 * @head: the head for your list.
668 * @offset: offset of hlist_node within the struct.
669 *
670 */
671#define sk_nulls_for_each_entry_offset(tpos, pos, head, offset) \
672 for (pos = (head)->first; \
673 (!is_a_nulls(pos)) && \
674 ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;}); \
675 pos = pos->next)
676
Eric W. Biedermanc336d142012-05-24 17:56:43 -0600677static inline struct user_namespace *sk_user_ns(struct sock *sk)
678{
679 /* Careful only use this in a context where these parameters
680 * can not change and must all be valid, such as recvmsg from
681 * userspace.
682 */
683 return sk->sk_socket->file->f_cred->user_ns;
684}
685
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686/* Sock flags */
687enum sock_flags {
688 SOCK_DEAD,
689 SOCK_DONE,
690 SOCK_URGINLINE,
691 SOCK_KEEPOPEN,
692 SOCK_LINGER,
693 SOCK_DESTROY,
694 SOCK_BROADCAST,
695 SOCK_TIMESTAMP,
696 SOCK_ZAPPED,
697 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
698 SOCK_DBG, /* %SO_DEBUG setting */
699 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700700 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
702 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
Mel Gorman7cb02402012-07-31 16:44:16 -0700703 SOCK_MEMALLOC, /* VM depends on this socket for swapping */
Patrick Ohly20d49472009-02-12 05:03:38 +0000704 SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */
705 SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
706 SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */
707 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
708 SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */
709 SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
Eric Dumazetbcdce712009-10-06 17:28:29 -0700710 SOCK_FASYNC, /* fasync() active */
Neil Horman3b885782009-10-12 13:26:31 -0700711 SOCK_RXQ_OVFL,
Shirley Ma1cdebb42011-07-06 12:17:30 +0000712 SOCK_ZEROCOPY, /* buffers from userspace */
Johannes Berg6e3e9392011-11-09 10:15:42 +0100713 SOCK_WIFI_STATUS, /* push wifi status to userspace */
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000714 SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
715 * Will use last 4 bytes of packet sent from
716 * user-space instead.
717 */
Vincent Bernatd59577b2013-01-16 22:55:49 +0100718 SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */
Keller, Jacob E7d4c04f2013-03-28 11:19:25 +0000719 SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720};
721
Ralf Baechle53b924b2005-08-23 10:11:30 -0700722static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
723{
724 nsk->sk_flags = osk->sk_flags;
725}
726
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
728{
729 __set_bit(flag, &sk->sk_flags);
730}
731
732static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
733{
734 __clear_bit(flag, &sk->sk_flags);
735}
736
Eric Dumazet1b23a5d2012-05-16 05:57:07 +0000737static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738{
739 return test_bit(flag, &sk->sk_flags);
740}
741
Mel Gormanc93bdd02012-07-31 16:44:19 -0700742#ifdef CONFIG_NET
743extern struct static_key memalloc_socks;
744static inline int sk_memalloc_socks(void)
745{
746 return static_key_false(&memalloc_socks);
747}
748#else
749
750static inline int sk_memalloc_socks(void)
751{
752 return 0;
753}
754
755#endif
756
Mel Gorman99a1dec2012-07-31 16:44:14 -0700757static inline gfp_t sk_gfp_atomic(struct sock *sk, gfp_t gfp_mask)
758{
Mel Gorman7cb02402012-07-31 16:44:16 -0700759 return GFP_ATOMIC | (sk->sk_allocation & __GFP_MEMALLOC);
Mel Gorman99a1dec2012-07-31 16:44:14 -0700760}
761
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762static inline void sk_acceptq_removed(struct sock *sk)
763{
764 sk->sk_ack_backlog--;
765}
766
767static inline void sk_acceptq_added(struct sock *sk)
768{
769 sk->sk_ack_backlog++;
770}
771
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000772static inline bool sk_acceptq_is_full(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773{
David S. Miller64a14652007-03-06 11:21:05 -0800774 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775}
776
777/*
778 * Compute minimal free write space needed to queue new packets.
779 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000780static inline int sk_stream_min_wspace(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781{
Eric Dumazet8df09ea2007-12-21 03:07:41 -0800782 return sk->sk_wmem_queued >> 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783}
784
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000785static inline int sk_stream_wspace(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786{
787 return sk->sk_sndbuf - sk->sk_wmem_queued;
788}
789
Joe Perches69336bd2013-09-22 10:32:26 -0700790void sk_stream_write_space(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791
Zhu Yi8eae9392010-03-04 18:01:40 +0000792/* OOB backlog add */
Zhu Yia3a858f2010-03-04 18:01:47 +0000793static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800794{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000795 /* dont let skb dst not refcounted, we are going to leave rcu lock */
796 skb_dst_force(skb);
797
798 if (!sk->sk_backlog.tail)
799 sk->sk_backlog.head = skb;
800 else
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800801 sk->sk_backlog.tail->next = skb;
Eric Dumazet7fee2262010-05-11 23:19:48 +0000802
803 sk->sk_backlog.tail = skb;
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800804 skb->next = NULL;
805}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806
Eric Dumazetc3774112010-04-27 15:13:20 -0700807/*
808 * Take into account size of receive queue and backlog queue
Eric Dumazet0fd7bac2011-12-21 07:11:44 +0000809 * Do not take into account this skb truesize,
810 * to allow even a single big packet to come.
Eric Dumazetc3774112010-04-27 15:13:20 -0700811 */
Sorin Dumitru274f4822014-07-22 21:16:51 +0300812static inline bool sk_rcvqueues_full(const struct sock *sk, unsigned int limit)
Eric Dumazetc3774112010-04-27 15:13:20 -0700813{
814 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
815
Eric Dumazetf545a382012-04-22 23:34:26 +0000816 return qsize > limit;
Eric Dumazetc3774112010-04-27 15:13:20 -0700817}
818
Zhu Yi8eae9392010-03-04 18:01:40 +0000819/* The per-socket spinlock must be held here. */
Eric Dumazetf545a382012-04-22 23:34:26 +0000820static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
821 unsigned int limit)
Zhu Yi8eae9392010-03-04 18:01:40 +0000822{
Sorin Dumitru274f4822014-07-22 21:16:51 +0300823 if (sk_rcvqueues_full(sk, limit))
Zhu Yi8eae9392010-03-04 18:01:40 +0000824 return -ENOBUFS;
825
Zhu Yia3a858f2010-03-04 18:01:47 +0000826 __sk_add_backlog(sk, skb);
Zhu Yi8eae9392010-03-04 18:01:40 +0000827 sk->sk_backlog.len += skb->truesize;
828 return 0;
829}
830
Joe Perches69336bd2013-09-22 10:32:26 -0700831int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
Mel Gormanb4b9e352012-07-31 16:44:26 -0700832
Peter Zijlstrac57943a2008-10-07 14:18:42 -0700833static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
834{
Mel Gormanb4b9e352012-07-31 16:44:26 -0700835 if (sk_memalloc_socks() && skb_pfmemalloc(skb))
836 return __sk_backlog_rcv(sk, skb);
837
Peter Zijlstrac57943a2008-10-07 14:18:42 -0700838 return sk->sk_backlog_rcv(sk, skb);
839}
840
Tom Herbertfe477552013-12-22 18:54:31 +0800841static inline void sock_rps_record_flow_hash(__u32 hash)
David S. Millerc58dc012010-04-27 15:05:31 -0700842{
843#ifdef CONFIG_RPS
844 struct rps_sock_flow_table *sock_flow_table;
845
846 rcu_read_lock();
847 sock_flow_table = rcu_dereference(rps_sock_flow_table);
Tom Herbertfe477552013-12-22 18:54:31 +0800848 rps_record_sock_flow(sock_flow_table, hash);
David S. Millerc58dc012010-04-27 15:05:31 -0700849 rcu_read_unlock();
850#endif
851}
852
Tom Herbertfe477552013-12-22 18:54:31 +0800853static inline void sock_rps_reset_flow_hash(__u32 hash)
David S. Millerc58dc012010-04-27 15:05:31 -0700854{
855#ifdef CONFIG_RPS
856 struct rps_sock_flow_table *sock_flow_table;
857
858 rcu_read_lock();
859 sock_flow_table = rcu_dereference(rps_sock_flow_table);
Tom Herbertfe477552013-12-22 18:54:31 +0800860 rps_reset_sock_flow(sock_flow_table, hash);
David S. Millerc58dc012010-04-27 15:05:31 -0700861 rcu_read_unlock();
862#endif
863}
864
Tom Herbertfe477552013-12-22 18:54:31 +0800865static inline void sock_rps_record_flow(const struct sock *sk)
866{
Zhi Yong Wuc9d8ca02014-01-01 04:31:01 +0800867#ifdef CONFIG_RPS
Tom Herbertfe477552013-12-22 18:54:31 +0800868 sock_rps_record_flow_hash(sk->sk_rxhash);
Zhi Yong Wuc9d8ca02014-01-01 04:31:01 +0800869#endif
Tom Herbertfe477552013-12-22 18:54:31 +0800870}
871
872static inline void sock_rps_reset_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_reset_flow_hash(sk->sk_rxhash);
Zhi Yong Wuc9d8ca02014-01-01 04:31:01 +0800876#endif
Tom Herbertfe477552013-12-22 18:54:31 +0800877}
878
Tom Herbertbdeab992011-08-14 19:45:55 +0000879static inline void sock_rps_save_rxhash(struct sock *sk,
880 const struct sk_buff *skb)
David S. Millerc58dc012010-04-27 15:05:31 -0700881{
882#ifdef CONFIG_RPS
Tom Herbert61b905d2014-03-24 15:34:47 -0700883 if (unlikely(sk->sk_rxhash != skb->hash)) {
David S. Millerc58dc012010-04-27 15:05:31 -0700884 sock_rps_reset_flow(sk);
Tom Herbert61b905d2014-03-24 15:34:47 -0700885 sk->sk_rxhash = skb->hash;
David S. Millerc58dc012010-04-27 15:05:31 -0700886 }
887#endif
888}
889
Tom Herbertbdeab992011-08-14 19:45:55 +0000890static inline void sock_rps_reset_rxhash(struct sock *sk)
891{
892#ifdef CONFIG_RPS
893 sock_rps_reset_flow(sk);
894 sk->sk_rxhash = 0;
895#endif
896}
897
Stephen Hemmingercfcabdc2007-10-09 01:59:42 -0700898#define sk_wait_event(__sk, __timeo, __condition) \
899 ({ int __rc; \
900 release_sock(__sk); \
901 __rc = __condition; \
902 if (!__rc) { \
903 *(__timeo) = schedule_timeout(*(__timeo)); \
904 } \
905 lock_sock(__sk); \
906 __rc = __condition; \
907 __rc; \
908 })
Linus Torvalds1da177e2005-04-16 15:20:36 -0700909
Joe Perches69336bd2013-09-22 10:32:26 -0700910int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
911int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
912void sk_stream_wait_close(struct sock *sk, long timeo_p);
913int sk_stream_error(struct sock *sk, int flags, int err);
914void sk_stream_kill_queues(struct sock *sk);
915void sk_set_memalloc(struct sock *sk);
916void sk_clear_memalloc(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700917
Joe Perches69336bd2013-09-22 10:32:26 -0700918int sk_wait_data(struct sock *sk, long *timeo);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700919
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700920struct request_sock_ops;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800921struct timewait_sock_ops;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -0800922struct inet_hashinfo;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -0700923struct raw_hashinfo;
Paul Gortmakerde477252011-05-26 13:46:22 -0400924struct module;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700925
Eric Dumazetf77d6022013-05-09 10:28:16 +0000926/*
927 * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
928 * un-modified. Special care is taken when initializing object to zero.
929 */
930static inline void sk_prot_clear_nulls(struct sock *sk, int size)
931{
932 if (offsetof(struct sock, sk_node.next) != 0)
933 memset(sk, 0, offsetof(struct sock, sk_node.next));
934 memset(&sk->sk_node.pprev, 0,
935 size - offsetof(struct sock, sk_node.pprev));
936}
937
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938/* Networking protocol blocks we attach to sockets.
939 * socket layer -> transport layer interface
940 * transport -> network interface is defined by struct inet_proto
941 */
942struct proto {
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000943 void (*close)(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944 long timeout);
945 int (*connect)(struct sock *sk,
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000946 struct sockaddr *uaddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700947 int addr_len);
948 int (*disconnect)(struct sock *sk, int flags);
949
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000950 struct sock * (*accept)(struct sock *sk, int flags, int *err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700951
952 int (*ioctl)(struct sock *sk, int cmd,
953 unsigned long arg);
954 int (*init)(struct sock *sk);
Brian Haley7d06b2e2008-06-14 17:04:49 -0700955 void (*destroy)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700956 void (*shutdown)(struct sock *sk, int how);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000957 int (*setsockopt)(struct sock *sk, int level,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700959 unsigned int optlen);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000960 int (*getsockopt)(struct sock *sk, int level,
961 int optname, char __user *optval,
962 int __user *option);
Alexey Dobriyanaf01d532008-08-28 02:53:51 -0700963#ifdef CONFIG_COMPAT
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800964 int (*compat_setsockopt)(struct sock *sk,
965 int level,
966 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700967 unsigned int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800968 int (*compat_getsockopt)(struct sock *sk,
969 int level,
970 int optname, char __user *optval,
971 int __user *option);
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000972 int (*compat_ioctl)(struct sock *sk,
973 unsigned int cmd, unsigned long arg);
Alexey Dobriyanaf01d532008-08-28 02:53:51 -0700974#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
976 struct msghdr *msg, size_t len);
977 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
978 struct msghdr *msg,
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000979 size_t len, int noblock, int flags,
980 int *addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981 int (*sendpage)(struct sock *sk, struct page *page,
982 int offset, size_t size, int flags);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000983 int (*bind)(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984 struct sockaddr *uaddr, int addr_len);
985
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000986 int (*backlog_rcv) (struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987 struct sk_buff *skb);
988
Eric Dumazet46d3cea2012-07-11 05:50:31 +0000989 void (*release_cb)(struct sock *sk);
Eric Dumazet563d34d2012-07-23 09:48:52 +0200990 void (*mtu_reduced)(struct sock *sk);
Eric Dumazet46d3cea2012-07-11 05:50:31 +0000991
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992 /* Keeping track of sk's, looking them up, and port selection methods. */
993 void (*hash)(struct sock *sk);
994 void (*unhash)(struct sock *sk);
Eric Dumazet719f8352010-09-08 05:08:44 +0000995 void (*rehash)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996 int (*get_port)(struct sock *sk, unsigned short snum);
Octavian Purdilafcbdf092010-12-16 14:26:56 -0800997 void (*clear_sk)(struct sock *sk, int size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998
Eric Dumazet286ab3d2007-11-05 23:38:39 -0800999 /* Keeping track of sockets in use */
Eric Dumazet65f76512008-01-03 20:46:48 -08001000#ifdef CONFIG_PROC_FS
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -07001001 unsigned int inuse_idx;
Eric Dumazet65f76512008-01-03 20:46:48 -08001002#endif
Arnaldo Carvalho de Meloebb53d72007-11-21 22:08:50 +08001003
Eric Dumazetc9bee3b72013-07-22 20:27:07 -07001004 bool (*stream_memory_free)(const struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005 /* Memory pressure */
Pavel Emelyanov5c52ba12008-07-16 20:28:10 -07001006 void (*enter_memory_pressure)(struct sock *sk);
Eric Dumazet8d987e52010-11-09 23:24:26 +00001007 atomic_long_t *memory_allocated; /* Current allocated memory. */
Eric Dumazet17483762008-11-25 21:16:35 -08001008 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009 /*
1010 * Pressure flag: try to collapse.
1011 * Technical note: it is used by multiple contexts non atomically.
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001012 * All the __sk_mem_schedule() is of this nature: accounting
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013 * is strict, actions are advisory and have some latency.
1014 */
1015 int *memory_pressure;
Eric Dumazet8d987e52010-11-09 23:24:26 +00001016 long *sysctl_mem;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017 int *sysctl_wmem;
1018 int *sysctl_rmem;
1019 int max_header;
Changli Gao7ba42912010-07-10 20:41:55 +00001020 bool no_autobind;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021
Eric Dumazet271b72c2008-10-29 02:11:14 -07001022 struct kmem_cache *slab;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023 unsigned int obj_size;
Eric Dumazet271b72c2008-10-29 02:11:14 -07001024 int slab_flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025
Eric Dumazetdd24c002008-11-25 21:17:14 -08001026 struct percpu_counter *orphan_count;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -07001027
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -07001028 struct request_sock_ops *rsk_prot;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -08001029 struct timewait_sock_ops *twsk_prot;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07001030
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -07001031 union {
1032 struct inet_hashinfo *hashinfo;
Eric Dumazet645ca702008-10-29 01:41:45 -07001033 struct udp_table *udp_table;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -07001034 struct raw_hashinfo *raw_hash;
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -07001035 } h;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -08001036
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037 struct module *owner;
1038
1039 char name[32];
1040
1041 struct list_head node;
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -07001042#ifdef SOCK_REFCNT_DEBUG
1043 atomic_t socks;
1044#endif
Andrew Mortonc255a452012-07-31 16:43:02 -07001045#ifdef CONFIG_MEMCG_KMEM
Glauber Costae1aab162011-12-11 21:47:03 +00001046 /*
1047 * cgroup specific init/deinit functions. Called once for all
1048 * protocols that implement it, from cgroups populate function.
1049 * This function has to setup any files the protocol want to
1050 * appear in the kmem cgroup filesystem.
1051 */
Glauber Costa1d62e432012-04-09 19:36:33 -03001052 int (*init_cgroup)(struct mem_cgroup *memcg,
Glauber Costae1aab162011-12-11 21:47:03 +00001053 struct cgroup_subsys *ss);
Glauber Costa1d62e432012-04-09 19:36:33 -03001054 void (*destroy_cgroup)(struct mem_cgroup *memcg);
Glauber Costae1aab162011-12-11 21:47:03 +00001055 struct cg_proto *(*proto_cgroup)(struct mem_cgroup *memcg);
1056#endif
1057};
1058
Glauber Costa3f134612012-05-29 15:07:11 -07001059/*
1060 * Bits in struct cg_proto.flags
1061 */
1062enum cg_proto_flags {
1063 /* Currently active and new sockets should be assigned to cgroups */
1064 MEMCG_SOCK_ACTIVE,
1065 /* It was ever activated; we must disarm static keys on destruction */
1066 MEMCG_SOCK_ACTIVATED,
1067};
1068
Glauber Costae1aab162011-12-11 21:47:03 +00001069struct cg_proto {
Eric W. Biederman2e685cad2013-10-19 16:26:19 -07001070 struct res_counter memory_allocated; /* Current allocated memory. */
1071 struct percpu_counter sockets_allocated; /* Current number of sockets. */
1072 int memory_pressure;
1073 long sysctl_mem[3];
Glauber Costa3f134612012-05-29 15:07:11 -07001074 unsigned long flags;
Glauber Costae1aab162011-12-11 21:47:03 +00001075 /*
1076 * memcg field is used to find which memcg we belong directly
1077 * Each memcg struct can hold more than one cg_proto, so container_of
1078 * won't really cut.
1079 *
1080 * The elegant solution would be having an inverse function to
1081 * proto_cgroup in struct proto, but that means polluting the structure
1082 * for everybody, instead of just for memcg users.
1083 */
1084 struct mem_cgroup *memcg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001085};
1086
Joe Perches69336bd2013-09-22 10:32:26 -07001087int proto_register(struct proto *prot, int alloc_slab);
1088void proto_unregister(struct proto *prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001089
Glauber Costa3f134612012-05-29 15:07:11 -07001090static inline bool memcg_proto_active(struct cg_proto *cg_proto)
1091{
1092 return test_bit(MEMCG_SOCK_ACTIVE, &cg_proto->flags);
1093}
1094
1095static inline bool memcg_proto_activated(struct cg_proto *cg_proto)
1096{
1097 return test_bit(MEMCG_SOCK_ACTIVATED, &cg_proto->flags);
1098}
1099
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -07001100#ifdef SOCK_REFCNT_DEBUG
1101static inline void sk_refcnt_debug_inc(struct sock *sk)
1102{
1103 atomic_inc(&sk->sk_prot->socks);
1104}
1105
1106static inline void sk_refcnt_debug_dec(struct sock *sk)
1107{
1108 atomic_dec(&sk->sk_prot->socks);
1109 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
1110 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
1111}
1112
Ying Xuedec34fb2013-02-15 22:28:25 +00001113static inline void sk_refcnt_debug_release(const struct sock *sk)
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -07001114{
1115 if (atomic_read(&sk->sk_refcnt) != 1)
1116 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
1117 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
1118}
1119#else /* SOCK_REFCNT_DEBUG */
1120#define sk_refcnt_debug_inc(sk) do { } while (0)
1121#define sk_refcnt_debug_dec(sk) do { } while (0)
1122#define sk_refcnt_debug_release(sk) do { } while (0)
1123#endif /* SOCK_REFCNT_DEBUG */
1124
Andrew Mortonc255a452012-07-31 16:43:02 -07001125#if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_NET)
Ingo Molnarc5905af2012-02-24 08:31:31 +01001126extern struct static_key memcg_socket_limit_enabled;
Glauber Costae1aab162011-12-11 21:47:03 +00001127static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1128 struct cg_proto *cg_proto)
1129{
1130 return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg));
1131}
Ingo Molnarc5905af2012-02-24 08:31:31 +01001132#define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
Glauber Costae1aab162011-12-11 21:47:03 +00001133#else
1134#define mem_cgroup_sockets_enabled 0
1135static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1136 struct cg_proto *cg_proto)
1137{
1138 return NULL;
1139}
1140#endif
1141
Eric Dumazetc9bee3b72013-07-22 20:27:07 -07001142static inline bool sk_stream_memory_free(const struct sock *sk)
1143{
1144 if (sk->sk_wmem_queued >= sk->sk_sndbuf)
1145 return false;
1146
1147 return sk->sk_prot->stream_memory_free ?
1148 sk->sk_prot->stream_memory_free(sk) : true;
1149}
1150
Eric Dumazet64dc6132013-07-22 20:26:31 -07001151static inline bool sk_stream_is_writeable(const struct sock *sk)
1152{
Eric Dumazetc9bee3b72013-07-22 20:27:07 -07001153 return sk_stream_wspace(sk) >= sk_stream_min_wspace(sk) &&
1154 sk_stream_memory_free(sk);
Eric Dumazet64dc6132013-07-22 20:26:31 -07001155}
Glauber Costae1aab162011-12-11 21:47:03 +00001156
Eric Dumazetc9bee3b72013-07-22 20:27:07 -07001157
Glauber Costa180d8cd2011-12-11 21:47:02 +00001158static inline bool sk_has_memory_pressure(const struct sock *sk)
1159{
1160 return sk->sk_prot->memory_pressure != NULL;
1161}
1162
1163static inline bool sk_under_memory_pressure(const struct sock *sk)
1164{
1165 if (!sk->sk_prot->memory_pressure)
1166 return false;
Glauber Costae1aab162011-12-11 21:47:03 +00001167
1168 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
Eric W. Biederman2e685cad2013-10-19 16:26:19 -07001169 return !!sk->sk_cgrp->memory_pressure;
Glauber Costae1aab162011-12-11 21:47:03 +00001170
Christoph Paasch35b87f62013-10-23 12:49:21 -07001171 return !!*sk->sk_prot->memory_pressure;
Glauber Costa180d8cd2011-12-11 21:47:02 +00001172}
1173
1174static inline void sk_leave_memory_pressure(struct sock *sk)
1175{
1176 int *memory_pressure = sk->sk_prot->memory_pressure;
1177
Glauber Costae1aab162011-12-11 21:47:03 +00001178 if (!memory_pressure)
1179 return;
1180
1181 if (*memory_pressure)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001182 *memory_pressure = 0;
Glauber Costae1aab162011-12-11 21:47:03 +00001183
1184 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1185 struct cg_proto *cg_proto = sk->sk_cgrp;
1186 struct proto *prot = sk->sk_prot;
1187
1188 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
Eric W. Biederman7f2cbdc2013-12-04 20:12:04 -08001189 cg_proto->memory_pressure = 0;
Glauber Costae1aab162011-12-11 21:47:03 +00001190 }
1191
Glauber Costa180d8cd2011-12-11 21:47:02 +00001192}
1193
1194static inline void sk_enter_memory_pressure(struct sock *sk)
1195{
Glauber Costae1aab162011-12-11 21:47:03 +00001196 if (!sk->sk_prot->enter_memory_pressure)
1197 return;
1198
1199 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1200 struct cg_proto *cg_proto = sk->sk_cgrp;
1201 struct proto *prot = sk->sk_prot;
1202
1203 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
Eric W. Biederman7f2cbdc2013-12-04 20:12:04 -08001204 cg_proto->memory_pressure = 1;
Glauber Costae1aab162011-12-11 21:47:03 +00001205 }
1206
1207 sk->sk_prot->enter_memory_pressure(sk);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001208}
1209
1210static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1211{
1212 long *prot = sk->sk_prot->sysctl_mem;
Glauber Costae1aab162011-12-11 21:47:03 +00001213 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1214 prot = sk->sk_cgrp->sysctl_mem;
Glauber Costa180d8cd2011-12-11 21:47:02 +00001215 return prot[index];
1216}
1217
Glauber Costae1aab162011-12-11 21:47:03 +00001218static inline void memcg_memory_allocated_add(struct cg_proto *prot,
1219 unsigned long amt,
1220 int *parent_status)
1221{
1222 struct res_counter *fail;
1223 int ret;
1224
Eric W. Biederman2e685cad2013-10-19 16:26:19 -07001225 ret = res_counter_charge_nofail(&prot->memory_allocated,
Glauber Costa0e90b312012-01-20 04:57:16 +00001226 amt << PAGE_SHIFT, &fail);
Glauber Costae1aab162011-12-11 21:47:03 +00001227 if (ret < 0)
1228 *parent_status = OVER_LIMIT;
1229}
1230
1231static inline void memcg_memory_allocated_sub(struct cg_proto *prot,
1232 unsigned long amt)
1233{
Eric W. Biederman2e685cad2013-10-19 16:26:19 -07001234 res_counter_uncharge(&prot->memory_allocated, amt << PAGE_SHIFT);
Glauber Costae1aab162011-12-11 21:47:03 +00001235}
1236
1237static inline u64 memcg_memory_allocated_read(struct cg_proto *prot)
1238{
1239 u64 ret;
Eric W. Biederman2e685cad2013-10-19 16:26:19 -07001240 ret = res_counter_read_u64(&prot->memory_allocated, RES_USAGE);
Glauber Costae1aab162011-12-11 21:47:03 +00001241 return ret >> PAGE_SHIFT;
1242}
1243
Glauber Costa180d8cd2011-12-11 21:47:02 +00001244static inline long
1245sk_memory_allocated(const struct sock *sk)
1246{
1247 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001248 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1249 return memcg_memory_allocated_read(sk->sk_cgrp);
1250
Glauber Costa180d8cd2011-12-11 21:47:02 +00001251 return atomic_long_read(prot->memory_allocated);
1252}
1253
1254static inline long
Glauber Costae1aab162011-12-11 21:47:03 +00001255sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001256{
1257 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001258
1259 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1260 memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status);
1261 /* update the root cgroup regardless */
1262 atomic_long_add_return(amt, prot->memory_allocated);
1263 return memcg_memory_allocated_read(sk->sk_cgrp);
1264 }
1265
Glauber Costa180d8cd2011-12-11 21:47:02 +00001266 return atomic_long_add_return(amt, prot->memory_allocated);
1267}
1268
1269static inline void
Glauber Costa0e90b312012-01-20 04:57:16 +00001270sk_memory_allocated_sub(struct sock *sk, int amt)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001271{
1272 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001273
Glauber Costa0e90b312012-01-20 04:57:16 +00001274 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
Glauber Costae1aab162011-12-11 21:47:03 +00001275 memcg_memory_allocated_sub(sk->sk_cgrp, amt);
1276
Glauber Costa180d8cd2011-12-11 21:47:02 +00001277 atomic_long_sub(amt, prot->memory_allocated);
1278}
1279
1280static inline void sk_sockets_allocated_dec(struct sock *sk)
1281{
1282 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001283
1284 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1285 struct cg_proto *cg_proto = sk->sk_cgrp;
1286
1287 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
Eric W. Biederman2e685cad2013-10-19 16:26:19 -07001288 percpu_counter_dec(&cg_proto->sockets_allocated);
Glauber Costae1aab162011-12-11 21:47:03 +00001289 }
1290
Glauber Costa180d8cd2011-12-11 21:47:02 +00001291 percpu_counter_dec(prot->sockets_allocated);
1292}
1293
1294static inline void sk_sockets_allocated_inc(struct sock *sk)
1295{
1296 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001297
1298 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1299 struct cg_proto *cg_proto = sk->sk_cgrp;
1300
1301 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
Eric W. Biederman2e685cad2013-10-19 16:26:19 -07001302 percpu_counter_inc(&cg_proto->sockets_allocated);
Glauber Costae1aab162011-12-11 21:47:03 +00001303 }
1304
Glauber Costa180d8cd2011-12-11 21:47:02 +00001305 percpu_counter_inc(prot->sockets_allocated);
1306}
1307
1308static inline int
1309sk_sockets_allocated_read_positive(struct sock *sk)
1310{
1311 struct proto *prot = sk->sk_prot;
1312
Glauber Costae1aab162011-12-11 21:47:03 +00001313 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
Eric W. Biederman2e685cad2013-10-19 16:26:19 -07001314 return percpu_counter_read_positive(&sk->sk_cgrp->sockets_allocated);
Glauber Costae1aab162011-12-11 21:47:03 +00001315
Eric Dumazet518fbf92012-04-28 23:21:56 +00001316 return percpu_counter_read_positive(prot->sockets_allocated);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001317}
1318
1319static inline int
1320proto_sockets_allocated_sum_positive(struct proto *prot)
1321{
1322 return percpu_counter_sum_positive(prot->sockets_allocated);
1323}
1324
1325static inline long
1326proto_memory_allocated(struct proto *prot)
1327{
1328 return atomic_long_read(prot->memory_allocated);
1329}
1330
1331static inline bool
1332proto_memory_pressure(struct proto *prot)
1333{
1334 if (!prot->memory_pressure)
1335 return false;
1336 return !!*prot->memory_pressure;
1337}
1338
Eric Dumazet65f76512008-01-03 20:46:48 -08001339
1340#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001341/* Called with local bh disabled */
Joe Perches69336bd2013-09-22 10:32:26 -07001342void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1343int sock_prot_inuse_get(struct net *net, struct proto *proto);
Eric Dumazet65f76512008-01-03 20:46:48 -08001344#else
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001345static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -07001346 int inc)
Eric Dumazet65f76512008-01-03 20:46:48 -08001347{
1348}
Eric Dumazet65f76512008-01-03 20:46:48 -08001349#endif
1350
Linus Torvalds1da177e2005-04-16 15:20:36 -07001351
Arnaldo Carvalho de Melo614c6cb2005-08-09 19:47:37 -07001352/* With per-bucket locks this operation is not-atomic, so that
1353 * this version is not worse.
1354 */
1355static inline void __sk_prot_rehash(struct sock *sk)
1356{
1357 sk->sk_prot->unhash(sk);
1358 sk->sk_prot->hash(sk);
1359}
1360
Octavian Purdilafcbdf092010-12-16 14:26:56 -08001361void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
1362
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363/* About 10 seconds */
1364#define SOCK_DESTROY_TIME (10*HZ)
1365
1366/* Sockets 0-1023 can't be bound to unless you are superuser */
1367#define PROT_SOCK 1024
1368
1369#define SHUTDOWN_MASK 3
1370#define RCV_SHUTDOWN 1
1371#define SEND_SHUTDOWN 2
1372
1373#define SOCK_SNDBUF_LOCK 1
1374#define SOCK_RCVBUF_LOCK 2
1375#define SOCK_BINDADDR_LOCK 4
1376#define SOCK_BINDPORT_LOCK 8
1377
1378/* sock_iocb: used to kick off async processing of socket ios */
1379struct sock_iocb {
1380 struct list_head list;
1381
1382 int flags;
1383 int size;
1384 struct socket *sock;
1385 struct sock *sk;
1386 struct scm_cookie *scm;
1387 struct msghdr *msg, async_msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001388 struct kiocb *kiocb;
1389};
1390
1391static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
1392{
1393 return (struct sock_iocb *)iocb->private;
1394}
1395
1396static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
1397{
1398 return si->kiocb;
1399}
1400
1401struct socket_alloc {
1402 struct socket socket;
1403 struct inode vfs_inode;
1404};
1405
1406static inline struct socket *SOCKET_I(struct inode *inode)
1407{
1408 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1409}
1410
1411static inline struct inode *SOCK_INODE(struct socket *socket)
1412{
1413 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1414}
1415
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001416/*
1417 * Functions for memory accounting
1418 */
Joe Perches69336bd2013-09-22 10:32:26 -07001419int __sk_mem_schedule(struct sock *sk, int size, int kind);
1420void __sk_mem_reclaim(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001422#define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1423#define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1424#define SK_MEM_SEND 0
1425#define SK_MEM_RECV 1
Linus Torvalds1da177e2005-04-16 15:20:36 -07001426
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001427static inline int sk_mem_pages(int amt)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001429 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430}
1431
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001432static inline bool sk_has_account(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001433{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001434 /* return true if protocol supports memory accounting */
1435 return !!sk->sk_prot->memory_allocated;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001436}
1437
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001438static inline bool sk_wmem_schedule(struct sock *sk, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001439{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001440 if (!sk_has_account(sk))
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001441 return true;
Herbert Xud80d99d2005-09-01 17:48:23 -07001442 return size <= sk->sk_forward_alloc ||
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001443 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1444}
1445
Mel Gormanc76562b2012-07-31 16:44:41 -07001446static inline bool
Chuck Lever35c448a2012-09-17 14:09:11 -07001447sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size)
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001448{
1449 if (!sk_has_account(sk))
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001450 return true;
Mel Gormanc76562b2012-07-31 16:44:41 -07001451 return size<= sk->sk_forward_alloc ||
1452 __sk_mem_schedule(sk, size, SK_MEM_RECV) ||
1453 skb_pfmemalloc(skb);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001454}
1455
1456static inline void sk_mem_reclaim(struct sock *sk)
1457{
1458 if (!sk_has_account(sk))
1459 return;
1460 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1461 __sk_mem_reclaim(sk);
1462}
1463
David S. Miller9993e7d2008-01-10 21:56:38 -08001464static inline void sk_mem_reclaim_partial(struct sock *sk)
1465{
1466 if (!sk_has_account(sk))
1467 return;
1468 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1469 __sk_mem_reclaim(sk);
1470}
1471
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001472static inline void sk_mem_charge(struct sock *sk, int size)
1473{
1474 if (!sk_has_account(sk))
1475 return;
1476 sk->sk_forward_alloc -= size;
1477}
1478
1479static inline void sk_mem_uncharge(struct sock *sk, int size)
1480{
1481 if (!sk_has_account(sk))
1482 return;
1483 sk->sk_forward_alloc += size;
1484}
1485
1486static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1487{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001488 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1489 sk->sk_wmem_queued -= skb->truesize;
1490 sk_mem_uncharge(sk, skb->truesize);
1491 __kfree_skb(skb);
Herbert Xud80d99d2005-09-01 17:48:23 -07001492}
1493
Linus Torvalds1da177e2005-04-16 15:20:36 -07001494/* Used by processes to "lock" a socket state, so that
1495 * interrupts and bottom half handlers won't change it
1496 * from under us. It essentially blocks any incoming
1497 * packets, so that we won't get any new data or any
1498 * packets that change the state of the socket.
1499 *
1500 * While locked, BH processing will add new packets to
1501 * the backlog queue. This queue is processed by the
1502 * owner of the socket lock right before it is released.
1503 *
1504 * Since ~2.3.5 it is also exclusive sleep lock serializing
1505 * accesses from user process context.
1506 */
John Heffnerd2e91172007-09-12 10:44:19 +02001507#define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001508
Eric Dumazetc3f9b012014-03-10 09:50:11 -07001509static inline void sock_release_ownership(struct sock *sk)
1510{
1511 sk->sk_lock.owned = 0;
1512}
1513
Peter Zijlstraed075362006-12-06 20:35:24 -08001514/*
1515 * Macro so as to not evaluate some arguments when
1516 * lockdep is not enabled.
1517 *
1518 * Mark both the sk_lock and the sk_lock.slock as a
1519 * per-address-family lock class.
1520 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001521#define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
Peter Zijlstraed075362006-12-06 20:35:24 -08001522do { \
Ingo Molnare8f6fbf2008-11-12 01:38:36 +00001523 sk->sk_lock.owned = 0; \
Peter Zijlstraed075362006-12-06 20:35:24 -08001524 init_waitqueue_head(&sk->sk_lock.wq); \
1525 spin_lock_init(&(sk)->sk_lock.slock); \
1526 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1527 sizeof((sk)->sk_lock)); \
1528 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001529 (skey), (sname)); \
Peter Zijlstraed075362006-12-06 20:35:24 -08001530 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1531} while (0)
1532
Joe Perches69336bd2013-09-22 10:32:26 -07001533void lock_sock_nested(struct sock *sk, int subclass);
Peter Zijlstrafcc70d52006-11-08 22:44:35 -08001534
1535static inline void lock_sock(struct sock *sk)
1536{
1537 lock_sock_nested(sk, 0);
1538}
1539
Joe Perches69336bd2013-09-22 10:32:26 -07001540void release_sock(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001541
1542/* BH context may only use the following locking interface. */
1543#define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
Ingo Molnarc6366182006-07-03 00:25:13 -07001544#define bh_lock_sock_nested(__sk) \
1545 spin_lock_nested(&((__sk)->sk_lock.slock), \
1546 SINGLE_DEPTH_NESTING)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001547#define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1548
Joe Perches69336bd2013-09-22 10:32:26 -07001549bool lock_sock_fast(struct sock *sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001550/**
1551 * unlock_sock_fast - complement of lock_sock_fast
1552 * @sk: socket
1553 * @slow: slow mode
1554 *
1555 * fast unlock socket for user context.
1556 * If slow mode is on, we call regular release_sock()
1557 */
1558static inline void unlock_sock_fast(struct sock *sk, bool slow)
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001559{
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001560 if (slow)
1561 release_sock(sk);
1562 else
1563 spin_unlock_bh(&sk->sk_lock.slock);
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001564}
1565
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001566
Joe Perches69336bd2013-09-22 10:32:26 -07001567struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
1568 struct proto *prot);
1569void sk_free(struct sock *sk);
1570void sk_release_kernel(struct sock *sk);
1571struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001572
Joe Perches69336bd2013-09-22 10:32:26 -07001573struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
1574 gfp_t priority);
Joe Perches69336bd2013-09-22 10:32:26 -07001575void sock_wfree(struct sk_buff *skb);
1576void skb_orphan_partial(struct sk_buff *skb);
1577void sock_rfree(struct sk_buff *skb);
1578void sock_edemux(struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001579
Joe Perches69336bd2013-09-22 10:32:26 -07001580int sock_setsockopt(struct socket *sock, int level, int op,
1581 char __user *optval, unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582
Joe Perches69336bd2013-09-22 10:32:26 -07001583int sock_getsockopt(struct socket *sock, int level, int op,
1584 char __user *optval, int __user *optlen);
1585struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1586 int noblock, int *errcode);
1587struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1588 unsigned long data_len, int noblock,
1589 int *errcode, int max_page_order);
1590void *sock_kmalloc(struct sock *sk, int size, gfp_t priority);
1591void sock_kfree_s(struct sock *sk, void *mem, int size);
1592void sk_send_sigurg(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001593
1594/*
1595 * Functions to fill in entries in struct proto_ops when a protocol
1596 * does not implement a particular function.
1597 */
Joe Perches69336bd2013-09-22 10:32:26 -07001598int sock_no_bind(struct socket *, struct sockaddr *, int);
1599int sock_no_connect(struct socket *, struct sockaddr *, int, int);
1600int sock_no_socketpair(struct socket *, struct socket *);
1601int sock_no_accept(struct socket *, struct socket *, int);
1602int sock_no_getname(struct socket *, struct sockaddr *, int *, int);
1603unsigned int sock_no_poll(struct file *, struct socket *,
1604 struct poll_table_struct *);
1605int sock_no_ioctl(struct socket *, unsigned int, unsigned long);
1606int sock_no_listen(struct socket *, int);
1607int sock_no_shutdown(struct socket *, int);
1608int sock_no_getsockopt(struct socket *, int , int, char __user *, int __user *);
1609int sock_no_setsockopt(struct socket *, int, int, char __user *, unsigned int);
1610int sock_no_sendmsg(struct kiocb *, struct socket *, struct msghdr *, size_t);
1611int sock_no_recvmsg(struct kiocb *, struct socket *, struct msghdr *, size_t,
1612 int);
1613int sock_no_mmap(struct file *file, struct socket *sock,
1614 struct vm_area_struct *vma);
1615ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset,
1616 size_t size, int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001617
1618/*
1619 * Functions to fill in entries in struct proto_ops when a protocol
1620 * uses the inet style.
1621 */
Joe Perches69336bd2013-09-22 10:32:26 -07001622int sock_common_getsockopt(struct socket *sock, int level, int optname,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001623 char __user *optval, int __user *optlen);
Joe Perches69336bd2013-09-22 10:32:26 -07001624int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001625 struct msghdr *msg, size_t size, int flags);
Joe Perches69336bd2013-09-22 10:32:26 -07001626int sock_common_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07001627 char __user *optval, unsigned int optlen);
Joe Perches69336bd2013-09-22 10:32:26 -07001628int compat_sock_common_getsockopt(struct socket *sock, int level,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001629 int optname, char __user *optval, int __user *optlen);
Joe Perches69336bd2013-09-22 10:32:26 -07001630int compat_sock_common_setsockopt(struct socket *sock, int level,
David S. Millerb7058842009-09-30 16:12:20 -07001631 int optname, char __user *optval, unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001632
Joe Perches69336bd2013-09-22 10:32:26 -07001633void sk_common_release(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001634
1635/*
1636 * Default socket callbacks and setup code
1637 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001638
Linus Torvalds1da177e2005-04-16 15:20:36 -07001639/* Initialise core socket variables */
Joe Perches69336bd2013-09-22 10:32:26 -07001640void sock_init_data(struct socket *sock, struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001641
Linus Torvalds1da177e2005-04-16 15:20:36 -07001642/*
1643 * Socket reference counting postulates.
1644 *
1645 * * Each user of socket SHOULD hold a reference count.
1646 * * Each access point to socket (an hash table bucket, reference from a list,
1647 * running timer, skb in flight MUST hold a reference count.
1648 * * When reference count hits 0, it means it will never increase back.
1649 * * When reference count hits 0, it means that no references from
1650 * outside exist to this socket and current process on current CPU
1651 * is last user and may/should destroy this socket.
1652 * * sk_free is called from any context: process, BH, IRQ. When
1653 * it is called, socket has no references from outside -> sk_free
1654 * may release descendant resources allocated by the socket, but
1655 * to the time when it is called, socket is NOT referenced by any
1656 * hash tables, lists etc.
1657 * * Packets, delivered from outside (from network or from another process)
1658 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1659 * when they sit in queue. Otherwise, packets will leak to hole, when
1660 * socket is looked up by one cpu and unhasing is made by another CPU.
1661 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1662 * (leak to backlog). Packet socket does all the processing inside
1663 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1664 * use separate SMP lock, so that they are prone too.
1665 */
1666
1667/* Ungrab socket and destroy it, if it was the last reference. */
1668static inline void sock_put(struct sock *sk)
1669{
1670 if (atomic_dec_and_test(&sk->sk_refcnt))
1671 sk_free(sk);
1672}
Eric Dumazet05dbc7b2013-10-03 00:22:02 -07001673/* Generic version of sock_put(), dealing with all sockets
1674 * (TCP_TIMEWAIT, ESTABLISHED...)
1675 */
1676void sock_gen_put(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677
Joe Perches69336bd2013-09-22 10:32:26 -07001678int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested);
Arnaldo Carvalho de Melo25995ff2005-12-27 02:42:22 -02001679
Krishna Kumare022f0b2009-10-19 23:46:20 +00001680static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1681{
1682 sk->sk_tx_queue_mapping = tx_queue;
1683}
1684
1685static inline void sk_tx_queue_clear(struct sock *sk)
1686{
1687 sk->sk_tx_queue_mapping = -1;
1688}
1689
1690static inline int sk_tx_queue_get(const struct sock *sk)
1691{
Tom Herbertb0f77d02010-07-14 20:50:29 -07001692 return sk ? sk->sk_tx_queue_mapping : -1;
Krishna Kumare022f0b2009-10-19 23:46:20 +00001693}
1694
David S. Miller972692e2008-06-17 22:41:38 -07001695static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1696{
Krishna Kumare022f0b2009-10-19 23:46:20 +00001697 sk_tx_queue_clear(sk);
David S. Miller972692e2008-06-17 22:41:38 -07001698 sk->sk_socket = sock;
1699}
1700
Eric Dumazetaa395142010-04-20 13:03:51 +00001701static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1702{
Eric Dumazeteaefd112011-02-18 03:26:36 +00001703 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1704 return &rcu_dereference_raw(sk->sk_wq)->wait;
Eric Dumazetaa395142010-04-20 13:03:51 +00001705}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001706/* Detach socket from process context.
1707 * Announce socket dead, detach it from wait queue and inode.
1708 * Note that parent inode held reference count on this struct sock,
1709 * we do not release it in this function, because protocol
1710 * probably wants some additional cleanups or even continuing
1711 * to work with this socket (TCP).
1712 */
1713static inline void sock_orphan(struct sock *sk)
1714{
1715 write_lock_bh(&sk->sk_callback_lock);
1716 sock_set_flag(sk, SOCK_DEAD);
David S. Miller972692e2008-06-17 22:41:38 -07001717 sk_set_socket(sk, NULL);
Eric Dumazet43815482010-04-29 11:01:49 +00001718 sk->sk_wq = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001719 write_unlock_bh(&sk->sk_callback_lock);
1720}
1721
1722static inline void sock_graft(struct sock *sk, struct socket *parent)
1723{
1724 write_lock_bh(&sk->sk_callback_lock);
Eric Dumazeteaefd112011-02-18 03:26:36 +00001725 sk->sk_wq = parent->wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001726 parent->sk = sk;
David S. Miller972692e2008-06-17 22:41:38 -07001727 sk_set_socket(sk, parent);
Venkat Yekkirala4237c752006-07-24 23:32:50 -07001728 security_sock_graft(sk, parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001729 write_unlock_bh(&sk->sk_callback_lock);
1730}
1731
Joe Perches69336bd2013-09-22 10:32:26 -07001732kuid_t sock_i_uid(struct sock *sk);
1733unsigned long sock_i_ino(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001734
1735static inline struct dst_entry *
1736__sk_dst_get(struct sock *sk)
1737{
Michal Hockod8bf4ca2011-07-08 14:39:41 +02001738 return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
Eric Dumazetf68c2242010-04-22 16:06:59 -07001739 lockdep_is_held(&sk->sk_lock.slock));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001740}
1741
1742static inline struct dst_entry *
1743sk_dst_get(struct sock *sk)
1744{
1745 struct dst_entry *dst;
1746
Eric Dumazetb6c67122010-04-08 23:03:29 +00001747 rcu_read_lock();
1748 dst = rcu_dereference(sk->sk_dst_cache);
Eric Dumazetf8864972014-06-24 10:05:11 -07001749 if (dst && !atomic_inc_not_zero(&dst->__refcnt))
1750 dst = NULL;
Eric Dumazetb6c67122010-04-08 23:03:29 +00001751 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001752 return dst;
1753}
1754
Eric Dumazetb6c67122010-04-08 23:03:29 +00001755static inline void dst_negative_advice(struct sock *sk)
1756{
1757 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1758
1759 if (dst && dst->ops->negative_advice) {
1760 ndst = dst->ops->negative_advice(dst);
1761
1762 if (ndst != dst) {
1763 rcu_assign_pointer(sk->sk_dst_cache, ndst);
ZHAO Gang0a6957e2013-10-22 16:23:38 +08001764 sk_tx_queue_clear(sk);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001765 }
1766 }
1767}
1768
Linus Torvalds1da177e2005-04-16 15:20:36 -07001769static inline void
1770__sk_dst_set(struct sock *sk, struct dst_entry *dst)
1771{
1772 struct dst_entry *old_dst;
1773
Krishna Kumare022f0b2009-10-19 23:46:20 +00001774 sk_tx_queue_clear(sk);
Eric Dumazet0b53ff22010-04-26 20:40:43 +00001775 /*
1776 * This can be called while sk is owned by the caller only,
1777 * with no state that can be checked in a rcu_dereference_check() cond
1778 */
1779 old_dst = rcu_dereference_raw(sk->sk_dst_cache);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001780 rcu_assign_pointer(sk->sk_dst_cache, dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001781 dst_release(old_dst);
1782}
1783
1784static inline void
1785sk_dst_set(struct sock *sk, struct dst_entry *dst)
1786{
Eric Dumazet7f502362014-06-30 01:26:23 -07001787 struct dst_entry *old_dst;
1788
1789 sk_tx_queue_clear(sk);
Eric Dumazet5925a052014-07-02 02:39:38 -07001790 old_dst = xchg((__force struct dst_entry **)&sk->sk_dst_cache, dst);
Eric Dumazet7f502362014-06-30 01:26:23 -07001791 dst_release(old_dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001792}
1793
1794static inline void
1795__sk_dst_reset(struct sock *sk)
1796{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001797 __sk_dst_set(sk, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001798}
1799
1800static inline void
1801sk_dst_reset(struct sock *sk)
1802{
Eric Dumazet7f502362014-06-30 01:26:23 -07001803 sk_dst_set(sk, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001804}
1805
Joe Perches69336bd2013-09-22 10:32:26 -07001806struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001807
Joe Perches69336bd2013-09-22 10:32:26 -07001808struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001809
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001810static inline bool sk_can_gso(const struct sock *sk)
Herbert Xubcd76112006-06-30 13:36:35 -07001811{
1812 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1813}
1814
Joe Perches69336bd2013-09-22 10:32:26 -07001815void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
Arnaldo Carvalho de Melo6cbb0df2005-08-09 19:49:02 -07001816
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001817static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
Eric Dumazeta4654192010-05-16 00:36:33 -07001818{
1819 sk->sk_route_nocaps |= flags;
1820 sk->sk_route_caps &= ~flags;
1821}
1822
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001823static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1824 char __user *from, char *to,
Wei Yongjun912d3982011-04-06 18:40:12 +00001825 int copy, int offset)
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001826{
1827 if (skb->ip_summed == CHECKSUM_NONE) {
1828 int err = 0;
1829 __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err);
1830 if (err)
1831 return err;
Wei Yongjun912d3982011-04-06 18:40:12 +00001832 skb->csum = csum_block_add(skb->csum, csum, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001833 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1834 if (!access_ok(VERIFY_READ, from, copy) ||
1835 __copy_from_user_nocache(to, from, copy))
1836 return -EFAULT;
1837 } else if (copy_from_user(to, from, copy))
1838 return -EFAULT;
1839
1840 return 0;
1841}
1842
1843static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
1844 char __user *from, int copy)
1845{
Wei Yongjun912d3982011-04-06 18:40:12 +00001846 int err, offset = skb->len;
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001847
Wei Yongjun912d3982011-04-06 18:40:12 +00001848 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1849 copy, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001850 if (err)
Wei Yongjun912d3982011-04-06 18:40:12 +00001851 __skb_trim(skb, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001852
1853 return err;
1854}
1855
1856static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from,
1857 struct sk_buff *skb,
1858 struct page *page,
1859 int off, int copy)
1860{
1861 int err;
1862
Wei Yongjun912d3982011-04-06 18:40:12 +00001863 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1864 copy, skb->len);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001865 if (err)
1866 return err;
1867
1868 skb->len += copy;
1869 skb->data_len += copy;
1870 skb->truesize += copy;
1871 sk->sk_wmem_queued += copy;
1872 sk_mem_charge(sk, copy);
1873 return 0;
1874}
1875
Linus Torvalds1da177e2005-04-16 15:20:36 -07001876static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1877 struct sk_buff *skb, struct page *page,
1878 int off, int copy)
1879{
1880 if (skb->ip_summed == CHECKSUM_NONE) {
1881 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08001882 __wsum csum = csum_and_copy_from_user(from,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001883 page_address(page) + off,
1884 copy, 0, &err);
1885 if (err)
1886 return err;
1887 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1888 } else if (copy_from_user(page_address(page) + off, from, copy))
1889 return -EFAULT;
1890
1891 skb->len += copy;
1892 skb->data_len += copy;
1893 skb->truesize += copy;
1894 sk->sk_wmem_queued += copy;
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001895 sk_mem_charge(sk, copy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001896 return 0;
1897}
1898
Eric Dumazetc5640392009-06-16 10:12:03 +00001899/**
1900 * sk_wmem_alloc_get - returns write allocations
1901 * @sk: socket
1902 *
1903 * Returns sk_wmem_alloc minus initial offset of one
1904 */
1905static inline int sk_wmem_alloc_get(const struct sock *sk)
1906{
1907 return atomic_read(&sk->sk_wmem_alloc) - 1;
1908}
1909
1910/**
1911 * sk_rmem_alloc_get - returns read allocations
1912 * @sk: socket
1913 *
1914 * Returns sk_rmem_alloc
1915 */
1916static inline int sk_rmem_alloc_get(const struct sock *sk)
1917{
1918 return atomic_read(&sk->sk_rmem_alloc);
1919}
1920
1921/**
1922 * sk_has_allocations - check if allocations are outstanding
1923 * @sk: socket
1924 *
1925 * Returns true if socket has write or read allocations
1926 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001927static inline bool sk_has_allocations(const struct sock *sk)
Eric Dumazetc5640392009-06-16 10:12:03 +00001928{
1929 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1930}
1931
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001932/**
Eric Dumazet43815482010-04-29 11:01:49 +00001933 * wq_has_sleeper - check if there are any waiting processes
Randy Dunlapacfbe962010-05-24 23:54:18 -07001934 * @wq: struct socket_wq
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001935 *
Eric Dumazet43815482010-04-29 11:01:49 +00001936 * Returns true if socket_wq has waiting processes
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001937 *
Eric Dumazet43815482010-04-29 11:01:49 +00001938 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001939 * barrier call. They were added due to the race found within the tcp code.
1940 *
1941 * Consider following tcp code paths:
1942 *
1943 * CPU1 CPU2
1944 *
1945 * sys_select receive packet
1946 * ... ...
1947 * __add_wait_queue update tp->rcv_nxt
1948 * ... ...
1949 * tp->rcv_nxt check sock_def_readable
1950 * ... {
Eric Dumazet43815482010-04-29 11:01:49 +00001951 * schedule rcu_read_lock();
1952 * wq = rcu_dereference(sk->sk_wq);
1953 * if (wq && waitqueue_active(&wq->wait))
1954 * wake_up_interruptible(&wq->wait)
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001955 * ...
1956 * }
1957 *
1958 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1959 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1960 * could then endup calling schedule and sleep forever if there are no more
1961 * data on the socket.
Jiri Olsaad462762009-07-08 12:10:31 +00001962 *
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001963 */
Eric Dumazet43815482010-04-29 11:01:49 +00001964static inline bool wq_has_sleeper(struct socket_wq *wq)
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001965{
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001966 /* We need to be sure we are in sync with the
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001967 * add_wait_queue modifications to the wait queue.
1968 *
1969 * This memory barrier is paired in the sock_poll_wait.
1970 */
Eric Dumazet43815482010-04-29 11:01:49 +00001971 smp_mb();
1972 return wq && waitqueue_active(&wq->wait);
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001973}
1974
1975/**
1976 * sock_poll_wait - place memory barrier behind the poll_wait call.
1977 * @filp: file
1978 * @wait_address: socket wait queue
1979 * @p: poll_table
1980 *
Eric Dumazet43815482010-04-29 11:01:49 +00001981 * See the comments in the wq_has_sleeper function.
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001982 */
1983static inline void sock_poll_wait(struct file *filp,
1984 wait_queue_head_t *wait_address, poll_table *p)
1985{
Hans Verkuil626cf232012-03-23 15:02:27 -07001986 if (!poll_does_not_wait(p) && wait_address) {
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001987 poll_wait(filp, wait_address, p);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001988 /* We need to be sure we are in sync with the
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001989 * socket flags modification.
1990 *
Eric Dumazet43815482010-04-29 11:01:49 +00001991 * This memory barrier is paired in the wq_has_sleeper.
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001992 */
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001993 smp_mb();
1994 }
1995}
1996
Tom Herbertb73c3d02014-07-01 21:32:17 -07001997static inline void skb_set_hash_from_sk(struct sk_buff *skb, struct sock *sk)
1998{
1999 if (sk->sk_txhash) {
2000 skb->l4_hash = 1;
2001 skb->hash = sk->sk_txhash;
2002 }
2003}
2004
Linus Torvalds1da177e2005-04-16 15:20:36 -07002005/*
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002006 * Queue a received datagram if it will fit. Stream and sequenced
Linus Torvalds1da177e2005-04-16 15:20:36 -07002007 * protocols can't normally use this as they need to fit buffers in
2008 * and play with them.
2009 *
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002010 * Inlined as it's very short and called for pretty much every
Linus Torvalds1da177e2005-04-16 15:20:36 -07002011 * packet ever received.
2012 */
2013
2014static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
2015{
Herbert Xud55d87f2009-06-22 02:25:25 +00002016 skb_orphan(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002017 skb->sk = sk;
2018 skb->destructor = sock_wfree;
Tom Herbertb73c3d02014-07-01 21:32:17 -07002019 skb_set_hash_from_sk(skb, sk);
Eric Dumazet2b85a342009-06-11 02:55:43 -07002020 /*
2021 * We used to take a refcount on sk, but following operation
2022 * is enough to guarantee sk_free() wont free this sock until
2023 * all in-flight packets are completed
2024 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002025 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
2026}
2027
2028static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
2029{
Herbert Xud55d87f2009-06-22 02:25:25 +00002030 skb_orphan(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002031 skb->sk = sk;
2032 skb->destructor = sock_rfree;
2033 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08002034 sk_mem_charge(sk, skb->truesize);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002035}
2036
Joe Perches69336bd2013-09-22 10:32:26 -07002037void sk_reset_timer(struct sock *sk, struct timer_list *timer,
2038 unsigned long expires);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002039
Joe Perches69336bd2013-09-22 10:32:26 -07002040void sk_stop_timer(struct sock *sk, struct timer_list *timer);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002041
Joe Perches69336bd2013-09-22 10:32:26 -07002042int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002043
Joe Perches69336bd2013-09-22 10:32:26 -07002044int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002045
2046/*
2047 * Recover an error report and clear atomically
2048 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002049
Linus Torvalds1da177e2005-04-16 15:20:36 -07002050static inline int sock_error(struct sock *sk)
2051{
Benjamin LaHaisec1cbe4b2005-12-13 23:22:19 -08002052 int err;
2053 if (likely(!sk->sk_err))
2054 return 0;
2055 err = xchg(&sk->sk_err, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002056 return -err;
2057}
2058
2059static inline unsigned long sock_wspace(struct sock *sk)
2060{
2061 int amt = 0;
2062
2063 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
2064 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002065 if (amt < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002066 amt = 0;
2067 }
2068 return amt;
2069}
2070
2071static inline void sk_wake_async(struct sock *sk, int how, int band)
2072{
Eric Dumazetbcdce712009-10-06 17:28:29 -07002073 if (sock_flag(sk, SOCK_FASYNC))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002074 sock_wake_async(sk->sk_socket, how, band);
2075}
2076
Daniel Borkmanneea86af2013-06-19 12:51:20 +02002077/* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
2078 * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
2079 * Note: for send buffers, TCP works better if we can build two skbs at
2080 * minimum.
Eric Dumazet7a91b4342010-09-26 18:53:07 -07002081 */
Eric Dumazet9eb5bf82013-07-03 05:02:22 -07002082#define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
Daniel Borkmanneea86af2013-06-19 12:51:20 +02002083
2084#define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2)
2085#define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE
Linus Torvalds1da177e2005-04-16 15:20:36 -07002086
2087static inline void sk_stream_moderate_sndbuf(struct sock *sk)
2088{
2089 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
Eric Dumazet8df09ea2007-12-21 03:07:41 -08002090 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
Daniel Borkmanneea86af2013-06-19 12:51:20 +02002091 sk->sk_sndbuf = max_t(u32, sk->sk_sndbuf, SOCK_MIN_SNDBUF);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002092 }
2093}
2094
Pavel Emelyanovdf97c702007-11-29 21:22:33 +11002095struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002096
Eric Dumazet5640f762012-09-23 23:04:42 +00002097/**
2098 * sk_page_frag - return an appropriate page_frag
2099 * @sk: socket
2100 *
2101 * If socket allocation mode allows current thread to sleep, it means its
2102 * safe to use the per task page_frag instead of the per socket one.
2103 */
2104static inline struct page_frag *sk_page_frag(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002105{
Eric Dumazet5640f762012-09-23 23:04:42 +00002106 if (sk->sk_allocation & __GFP_WAIT)
2107 return &current->task_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002108
Eric Dumazet5640f762012-09-23 23:04:42 +00002109 return &sk->sk_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002110}
2111
Joe Perches69336bd2013-09-22 10:32:26 -07002112bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag);
Eric Dumazet5640f762012-09-23 23:04:42 +00002113
Linus Torvalds1da177e2005-04-16 15:20:36 -07002114/*
2115 * Default write policy as shown to user space via poll/select/SIGIO
2116 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002117static inline bool sock_writeable(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002118{
Eric Dumazet8df09ea2007-12-21 03:07:41 -08002119 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002120}
2121
Al Virodd0fc662005-10-07 07:46:04 +01002122static inline gfp_t gfp_any(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002123{
Andrew Morton99709372009-02-12 16:43:17 -08002124 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002125}
2126
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002127static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002128{
2129 return noblock ? 0 : sk->sk_rcvtimeo;
2130}
2131
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002132static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002133{
2134 return noblock ? 0 : sk->sk_sndtimeo;
2135}
2136
2137static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2138{
2139 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
2140}
2141
2142/* Alas, with timeout socket operations are not restartable.
2143 * Compare this to poll().
2144 */
2145static inline int sock_intr_errno(long timeo)
2146{
2147 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2148}
2149
Joe Perches69336bd2013-09-22 10:32:26 -07002150void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2151 struct sk_buff *skb);
2152void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2153 struct sk_buff *skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -07002154
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002155static inline void
Linus Torvalds1da177e2005-04-16 15:20:36 -07002156sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2157{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002158 ktime_t kt = skb->tstamp;
Patrick Ohly20d49472009-02-12 05:03:38 +00002159 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002160
Patrick Ohly20d49472009-02-12 05:03:38 +00002161 /*
2162 * generate control messages if
2163 * - receive time stamping in software requested (SOCK_RCVTSTAMP
2164 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
2165 * - software time stamp available and wanted
2166 * (SOCK_TIMESTAMPING_SOFTWARE)
2167 * - hardware time stamps available and wanted
Willem de Bruijn4d276eb2014-07-25 18:01:32 -04002168 * SOCK_TIMESTAMPING_RAW_HARDWARE
Patrick Ohly20d49472009-02-12 05:03:38 +00002169 */
2170 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
2171 sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
Willem de Bruijnf24b9be2014-08-04 22:11:45 -04002172 (kt.tv64 &&
2173 (sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE) ||
2174 skb_shinfo(skb)->tx_flags & SKBTX_ANY_SW_TSTAMP)) ||
Patrick Ohly20d49472009-02-12 05:03:38 +00002175 (hwtstamps->hwtstamp.tv64 &&
Willem de Bruijn4d276eb2014-07-25 18:01:32 -04002176 sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)))
Eric Dumazet92f37fd2007-03-25 22:14:49 -07002177 __sock_recv_timestamp(msg, sk, skb);
2178 else
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002179 sk->sk_stamp = kt;
Johannes Berg6e3e9392011-11-09 10:15:42 +01002180
2181 if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2182 __sock_recv_wifi_status(msg, sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002183}
2184
Joe Perches69336bd2013-09-22 10:32:26 -07002185void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2186 struct sk_buff *skb);
Eric Dumazet767dd032010-04-28 19:14:43 +00002187
2188static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2189 struct sk_buff *skb)
2190{
2191#define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2192 (1UL << SOCK_RCVTSTAMP) | \
Eric Dumazet767dd032010-04-28 19:14:43 +00002193 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
Willem de Bruijn4d276eb2014-07-25 18:01:32 -04002194 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE))
Eric Dumazet767dd032010-04-28 19:14:43 +00002195
2196 if (sk->sk_flags & FLAGS_TS_OR_DROPS)
2197 __sock_recv_ts_and_drops(msg, sk, skb);
2198 else
2199 sk->sk_stamp = skb->tstamp;
2200}
Neil Horman3b885782009-10-12 13:26:31 -07002201
Linus Torvalds1da177e2005-04-16 15:20:36 -07002202/**
Patrick Ohly20d49472009-02-12 05:03:38 +00002203 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
Patrick Ohly20d49472009-02-12 05:03:38 +00002204 * @sk: socket sending this packet
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002205 * @tx_flags: filled with instructions for time stamping
Patrick Ohly20d49472009-02-12 05:03:38 +00002206 *
Daniel Borkmannbf84a012013-04-14 08:08:13 +00002207 * Currently only depends on SOCK_TIMESTAMPING* flags.
Patrick Ohly20d49472009-02-12 05:03:38 +00002208 */
Joe Perches69336bd2013-09-22 10:32:26 -07002209void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags);
Patrick Ohly20d49472009-02-12 05:03:38 +00002210
2211/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002212 * sk_eat_skb - Release a skb if it is no longer needed
Pavel Pisa4dc3b162005-05-01 08:59:25 -07002213 * @sk: socket to eat this skb from
2214 * @skb: socket buffer to eat
Randy Dunlapf4b8ea72006-06-22 16:00:11 -07002215 * @copied_early: flag indicating whether DMA operations copied this data early
Linus Torvalds1da177e2005-04-16 15:20:36 -07002216 *
2217 * This routine must be called with interrupts disabled or with the socket
2218 * locked so that the sk_buff queue operation is ok.
2219*/
Chris Leech624d1162006-05-23 18:01:28 -07002220#ifdef CONFIG_NET_DMA
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002221static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
Chris Leech624d1162006-05-23 18:01:28 -07002222{
2223 __skb_unlink(skb, &sk->sk_receive_queue);
2224 if (!copied_early)
2225 __kfree_skb(skb);
2226 else
2227 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
2228}
2229#else
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002230static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002231{
2232 __skb_unlink(skb, &sk->sk_receive_queue);
2233 __kfree_skb(skb);
2234}
Chris Leech624d1162006-05-23 18:01:28 -07002235#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002236
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002237static inline
2238struct net *sock_net(const struct sock *sk)
2239{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002240 return read_pnet(&sk->sk_net);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002241}
2242
2243static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07002244void sock_net_set(struct sock *sk, struct net *net)
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002245{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002246 write_pnet(&sk->sk_net, net);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002247}
2248
Denis V. Lunevedf02082008-02-29 11:18:32 -08002249/*
2250 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
Lucas De Marchi25985ed2011-03-30 22:57:33 -03002251 * They should not hold a reference to a namespace in order to allow
Denis V. Lunevedf02082008-02-29 11:18:32 -08002252 * to stop it.
2253 * Sockets after sk_change_net should be released using sk_release_kernel
2254 */
2255static inline void sk_change_net(struct sock *sk, struct net *net)
2256{
Gu Zheng5812521b2014-03-10 09:57:34 +08002257 struct net *current_net = sock_net(sk);
2258
2259 if (!net_eq(current_net, net)) {
2260 put_net(current_net);
2261 sock_net_set(sk, hold_net(net));
2262 }
Denis V. Lunevedf02082008-02-29 11:18:32 -08002263}
2264
KOVACS Krisztian23542612008-10-07 12:41:01 -07002265static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2266{
Vijay Subramanianefc27f82012-06-24 13:03:07 +00002267 if (skb->sk) {
KOVACS Krisztian23542612008-10-07 12:41:01 -07002268 struct sock *sk = skb->sk;
2269
2270 skb->destructor = NULL;
2271 skb->sk = NULL;
2272 return sk;
2273 }
2274 return NULL;
2275}
2276
Joe Perches69336bd2013-09-22 10:32:26 -07002277void sock_enable_timestamp(struct sock *sk, int flag);
2278int sock_get_timestamp(struct sock *, struct timeval __user *);
2279int sock_get_timestampns(struct sock *, struct timespec __user *);
2280int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len, int level,
2281 int type);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002282
Eric W. Biedermana3b299d2014-04-23 14:26:56 -07002283bool sk_ns_capable(const struct sock *sk,
2284 struct user_namespace *user_ns, int cap);
2285bool sk_capable(const struct sock *sk, int cap);
2286bool sk_net_capable(const struct sock *sk, int cap);
2287
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002288/*
2289 * Enable debug/info messages
Linus Torvalds1da177e2005-04-16 15:20:36 -07002290 */
Stephen Hemmingera2a316f2007-03-08 20:41:08 -08002291extern int net_msg_warn;
2292#define NETDEBUG(fmt, args...) \
2293 do { if (net_msg_warn) printk(fmt,##args); } while (0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002294
Stephen Hemmingera2a316f2007-03-08 20:41:08 -08002295#define LIMIT_NETDEBUG(fmt, args...) \
2296 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002297
Linus Torvalds1da177e2005-04-16 15:20:36 -07002298extern __u32 sysctl_wmem_max;
2299extern __u32 sysctl_rmem_max;
2300
David S. Miller6baf1f42005-09-05 18:14:11 -07002301extern int sysctl_optmem_max;
2302
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002303extern __u32 sysctl_wmem_default;
2304extern __u32 sysctl_rmem_default;
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002305
Linus Torvalds1da177e2005-04-16 15:20:36 -07002306#endif /* _SOCK_H */