blob: 07b7fcd60d808a33f9e6fff208c07fe412da8c7e [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;
184 unsigned char skc_reuseport:4;
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000185 int skc_bound_dev_if;
Eric Dumazet512615b2009-11-08 10:17:58 +0000186 union {
187 struct hlist_node skc_bind_node;
188 struct hlist_nulls_node skc_portaddr_node;
189 };
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700190 struct proto *skc_prot;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900191#ifdef CONFIG_NET_NS
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200192 struct net *skc_net;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900193#endif
Eric Dumazetefe42082013-10-03 15:42:29 -0700194
195#if IS_ENABLED(CONFIG_IPV6)
196 struct in6_addr skc_v6_daddr;
197 struct in6_addr skc_v6_rcv_saddr;
198#endif
199
Eric Dumazet68835ab2010-11-30 19:04:07 +0000200 /*
201 * fields between dontcopy_begin/dontcopy_end
202 * are not copied in sock_copy()
203 */
Randy Dunlap928c41e2011-01-08 17:39:21 +0000204 /* private: */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000205 int skc_dontcopy_begin[0];
Randy Dunlap928c41e2011-01-08 17:39:21 +0000206 /* public: */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000207 union {
208 struct hlist_node skc_node;
209 struct hlist_nulls_node skc_nulls_node;
210 };
211 int skc_tx_queue_mapping;
212 atomic_t skc_refcnt;
Randy Dunlap928c41e2011-01-08 17:39:21 +0000213 /* private: */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000214 int skc_dontcopy_end[0];
Randy Dunlap928c41e2011-01-08 17:39:21 +0000215 /* public: */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216};
217
Glauber Costae1aab162011-12-11 21:47:03 +0000218struct cg_proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219/**
220 * struct sock - network layer representation of sockets
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700221 * @__sk_common: shared layout with inet_timewait_sock
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700222 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
223 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
224 * @sk_lock: synchronizer
225 * @sk_rcvbuf: size of receive buffer in bytes
Eric Dumazet43815482010-04-29 11:01:49 +0000226 * @sk_wq: sock wait queue and async head
Shawn Bohrer421b3882013-10-07 11:01:39 -0500227 * @sk_rx_dst: receive input route used by early demux
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700228 * @sk_dst_cache: destination cache
229 * @sk_dst_lock: destination cache lock
230 * @sk_policy: flow policy
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700231 * @sk_receive_queue: incoming packets
232 * @sk_wmem_alloc: transmit queue bytes committed
233 * @sk_write_queue: Packet sending queue
Chris Leech97fc2f02006-05-23 17:55:33 -0700234 * @sk_async_wait_queue: DMA copied packets
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700235 * @sk_omem_alloc: "o" is "option" or "other"
236 * @sk_wmem_queued: persistent queue size
237 * @sk_forward_alloc: space allocated forward
Eliezer Tamir06021292013-06-10 11:39:50 +0300238 * @sk_napi_id: id of the last napi context to receive data for sk
Eliezer Tamirdafcc432013-06-14 16:33:57 +0300239 * @sk_ll_usec: usecs to busypoll when there is no data
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700240 * @sk_allocation: allocation mode
Eric Dumazet95bd09e2013-08-27 05:46:32 -0700241 * @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
Eric Dumazetc3f40d72013-09-29 01:12:40 -0700242 * @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE)
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700243 * @sk_sndbuf: size of send buffer in bytes
Wang Chen33c732c2007-11-13 20:30:01 -0800244 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
Patrick Ohly20d49472009-02-12 05:03:38 +0000245 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
Tom Herbert28448b82014-05-23 08:47:19 -0700246 * @sk_no_check_tx: %SO_NO_CHECK setting, set checksum in TX packets
247 * @sk_no_check_rx: allow zero checksum in RX packets
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700248 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
Eric Dumazeta4654192010-05-16 00:36:33 -0700249 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
Herbert Xubcd76112006-06-30 13:36:35 -0700250 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -0700251 * @sk_gso_max_size: Maximum GSO segment size to build
Ben Hutchings14853482012-07-30 16:11:42 +0000252 * @sk_gso_max_segs: Maximum number of GSO segments
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700253 * @sk_lingertime: %SO_LINGER l_linger setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700254 * @sk_backlog: always used with the per-socket spinlock held
255 * @sk_callback_lock: used with the callbacks in the end of this struct
256 * @sk_error_queue: rarely used
Wang Chen33c732c2007-11-13 20:30:01 -0800257 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
258 * IPV6_ADDRFORM for instance)
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700259 * @sk_err: last error
Wang Chen33c732c2007-11-13 20:30:01 -0800260 * @sk_err_soft: errors that don't cause failure but are the cause of a
261 * persistent failure not just 'timed out'
Eric Dumazetcb61cb92008-06-17 21:04:56 -0700262 * @sk_drops: raw/udp drops counter
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700263 * @sk_ack_backlog: current listen backlog
264 * @sk_max_ack_backlog: listen backlog set in listen()
265 * @sk_priority: %SO_PRIORITY setting
Randy Dunlap1a3bc362012-01-21 09:03:10 +0000266 * @sk_cgrp_prioidx: socket group's priority map index
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700267 * @sk_type: socket type (%SOCK_STREAM, etc)
268 * @sk_protocol: which protocol this socket belongs in this network family
Randy Dunlap53c3fa22010-08-09 13:41:07 +0000269 * @sk_peer_pid: &struct pid for this socket's peer
270 * @sk_peer_cred: %SO_PEERCRED setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700271 * @sk_rcvlowat: %SO_RCVLOWAT setting
272 * @sk_rcvtimeo: %SO_RCVTIMEO setting
273 * @sk_sndtimeo: %SO_SNDTIMEO setting
David S. Millerc58dc012010-04-27 15:05:31 -0700274 * @sk_rxhash: flow hash received from netif layer
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700275 * @sk_filter: socket filtering instructions
276 * @sk_protinfo: private area, net family specific, when not using slab
277 * @sk_timer: sock cleanup timer
278 * @sk_stamp: time stamp of last packet received
279 * @sk_socket: Identd and reporting IO signals
280 * @sk_user_data: RPC layer private data
Eric Dumazet5640f762012-09-23 23:04:42 +0000281 * @sk_frag: cached page frag
Randy Dunlapd3d4f0a2012-04-17 14:03:53 +0000282 * @sk_peek_off: current peek_offset value
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700283 * @sk_send_head: front of stuff to transmit
Martin Waitz67be2dd2005-05-01 08:59:26 -0700284 * @sk_security: used by security modules
Randy Dunlap31729362008-02-18 20:52:13 -0800285 * @sk_mark: generic packet mark
Randy Dunlap53c3fa22010-08-09 13:41:07 +0000286 * @sk_classid: this socket's cgroup classid
Glauber Costae1aab162011-12-11 21:47:03 +0000287 * @sk_cgrp: this socket's cgroup-specific proto data
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700288 * @sk_write_pending: a write to stream socket waits to start
289 * @sk_state_change: callback to indicate change in the state of the sock
290 * @sk_data_ready: callback to indicate there is data to be processed
291 * @sk_write_space: callback to indicate there is bf sending space available
292 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
293 * @sk_backlog_rcv: callback to process the backlog
294 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
Linus Torvalds1da177e2005-04-16 15:20:36 -0700295 */
296struct sock {
297 /*
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700298 * Now struct inet_timewait_sock also uses sock_common, so please just
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299 * don't add nothing before this first member (__sk_common) --acme
300 */
301 struct sock_common __sk_common;
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000302#define sk_node __sk_common.skc_node
303#define sk_nulls_node __sk_common.skc_nulls_node
304#define sk_refcnt __sk_common.skc_refcnt
Krishna Kumare022f0b2009-10-19 23:46:20 +0000305#define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000306
Eric Dumazet68835ab2010-11-30 19:04:07 +0000307#define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
308#define sk_dontcopy_end __sk_common.skc_dontcopy_end
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000309#define sk_hash __sk_common.skc_hash
Eric Dumazet50805462013-10-02 04:29:50 -0700310#define sk_portpair __sk_common.skc_portpair
Eric Dumazet05dbc7b2013-10-03 00:22:02 -0700311#define sk_num __sk_common.skc_num
312#define sk_dport __sk_common.skc_dport
Eric Dumazet50805462013-10-02 04:29:50 -0700313#define sk_addrpair __sk_common.skc_addrpair
314#define sk_daddr __sk_common.skc_daddr
315#define sk_rcv_saddr __sk_common.skc_rcv_saddr
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316#define sk_family __sk_common.skc_family
317#define sk_state __sk_common.skc_state
318#define sk_reuse __sk_common.skc_reuse
Tom Herbert055dc212013-01-22 09:49:50 +0000319#define sk_reuseport __sk_common.skc_reuseport
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320#define sk_bound_dev_if __sk_common.skc_bound_dev_if
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321#define sk_bind_node __sk_common.skc_bind_node
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700322#define sk_prot __sk_common.skc_prot
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200323#define sk_net __sk_common.skc_net
Eric Dumazetefe42082013-10-03 15:42:29 -0700324#define sk_v6_daddr __sk_common.skc_v6_daddr
325#define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr
326
Eric Dumazetb178bb32010-11-16 05:56:04 +0000327 socket_lock_t sk_lock;
328 struct sk_buff_head sk_receive_queue;
329 /*
330 * The backlog queue is special, it is always used with
331 * the per-socket spinlock held and requires low latency
332 * access. Therefore we special case it's implementation.
333 * Note : rmem_alloc is in this structure to fill a hole
334 * on 64bit arches, not because its logically part of
335 * backlog.
336 */
337 struct {
338 atomic_t rmem_alloc;
339 int len;
340 struct sk_buff *head;
341 struct sk_buff *tail;
342 } sk_backlog;
343#define sk_rmem_alloc sk_backlog.rmem_alloc
344 int sk_forward_alloc;
345#ifdef CONFIG_RPS
346 __u32 sk_rxhash;
347#endif
Cong Wange0d10952013-08-01 11:10:25 +0800348#ifdef CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir06021292013-06-10 11:39:50 +0300349 unsigned int sk_napi_id;
Eliezer Tamirdafcc432013-06-14 16:33:57 +0300350 unsigned int sk_ll_usec;
Eliezer Tamir06021292013-06-10 11:39:50 +0300351#endif
Eric Dumazetb178bb32010-11-16 05:56:04 +0000352 atomic_t sk_drops;
353 int sk_rcvbuf;
354
355 struct sk_filter __rcu *sk_filter;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000356 struct socket_wq __rcu *sk_wq;
Eric Dumazetb178bb32010-11-16 05:56:04 +0000357
358#ifdef CONFIG_NET_DMA
359 struct sk_buff_head sk_async_wait_queue;
360#endif
361
362#ifdef CONFIG_XFRM
363 struct xfrm_policy *sk_policy[2];
364#endif
365 unsigned long sk_flags;
Eric Dumazetdeaa5852012-06-24 20:22:49 +0000366 struct dst_entry *sk_rx_dst;
Cong Wang0e36cbb2013-01-22 21:09:51 +0000367 struct dst_entry __rcu *sk_dst_cache;
Eric Dumazetb178bb32010-11-16 05:56:04 +0000368 spinlock_t sk_dst_lock;
369 atomic_t sk_wmem_alloc;
370 atomic_t sk_omem_alloc;
371 int sk_sndbuf;
372 struct sk_buff_head sk_write_queue;
Vegard Nossuma98b65a2009-02-26 14:46:57 +0100373 kmemcheck_bitfield_begin(flags);
Eric Dumazet5fdb9972009-10-08 22:50:25 +0000374 unsigned int sk_shutdown : 2,
Tom Herbert28448b82014-05-23 08:47:19 -0700375 sk_no_check_tx : 1,
376 sk_no_check_rx : 1,
Eric Dumazet5fdb9972009-10-08 22:50:25 +0000377 sk_userlocks : 4,
378 sk_protocol : 8,
379 sk_type : 16;
Vegard Nossuma98b65a2009-02-26 14:46:57 +0100380 kmemcheck_bitfield_end(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381 int sk_wmem_queued;
Al Viro7d877f32005-10-21 03:20:43 -0400382 gfp_t sk_allocation;
Eric Dumazet95bd09e2013-08-27 05:46:32 -0700383 u32 sk_pacing_rate; /* bytes per second */
Eric Dumazet62748f32013-09-24 08:20:52 -0700384 u32 sk_max_pacing_rate;
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000385 netdev_features_t sk_route_caps;
386 netdev_features_t sk_route_nocaps;
Herbert Xubcd76112006-06-30 13:36:35 -0700387 int sk_gso_type;
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -0700388 unsigned int sk_gso_max_size;
Ben Hutchings14853482012-07-30 16:11:42 +0000389 u16 sk_gso_max_segs;
David S. Miller9932cf92006-03-24 15:12:37 -0800390 int sk_rcvlowat;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391 unsigned long sk_lingertime;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392 struct sk_buff_head sk_error_queue;
Arnaldo Carvalho de Melo476e19c2005-05-05 13:35:15 -0700393 struct proto *sk_prot_creator;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394 rwlock_t sk_callback_lock;
395 int sk_err,
396 sk_err_soft;
397 unsigned short sk_ack_backlog;
398 unsigned short sk_max_ack_backlog;
399 __u32 sk_priority;
Daniel Borkmann86f85152013-12-29 17:27:11 +0100400#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
Neil Horman5bc14212011-11-22 05:10:51 +0000401 __u32 sk_cgrp_prioidx;
402#endif
Eric W. Biederman109f6e32010-06-13 03:30:14 +0000403 struct pid *sk_peer_pid;
404 const struct cred *sk_peer_cred;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 long sk_rcvtimeo;
406 long sk_sndtimeo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407 void *sk_protinfo;
408 struct timer_list sk_timer;
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700409 ktime_t sk_stamp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410 struct socket *sk_socket;
411 void *sk_user_data;
Eric Dumazet5640f762012-09-23 23:04:42 +0000412 struct page_frag sk_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700413 struct sk_buff *sk_send_head;
Pavel Emelyanovef64a542012-02-21 07:31:34 +0000414 __s32 sk_peek_off;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 int sk_write_pending;
Alexey Dobriyand5f64232008-11-04 14:45:58 -0800416#ifdef CONFIG_SECURITY
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417 void *sk_security;
Alexey Dobriyand5f64232008-11-04 14:45:58 -0800418#endif
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800419 __u32 sk_mark;
Herbert Xuf8451722010-05-24 00:12:34 -0700420 u32 sk_classid;
Glauber Costae1aab162011-12-11 21:47:03 +0000421 struct cg_proto *sk_cgrp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422 void (*sk_state_change)(struct sock *sk);
David S. Miller676d2362014-04-11 16:15:36 -0400423 void (*sk_data_ready)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424 void (*sk_write_space)(struct sock *sk);
425 void (*sk_error_report)(struct sock *sk);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000426 int (*sk_backlog_rcv)(struct sock *sk,
427 struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428 void (*sk_destruct)(struct sock *sk);
429};
430
Pravin B Shelar559835ea2013-09-24 10:25:40 -0700431#define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
432
433#define rcu_dereference_sk_user_data(sk) rcu_dereference(__sk_user_data((sk)))
434#define rcu_assign_sk_user_data(sk, ptr) rcu_assign_pointer(__sk_user_data((sk)), ptr)
435
Pavel Emelyanov4a17fd52012-04-19 03:39:36 +0000436/*
437 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
438 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
439 * on a socket means that the socket will reuse everybody else's port
440 * without looking at the other's sk_reuse value.
441 */
442
443#define SK_NO_REUSE 0
444#define SK_CAN_REUSE 1
445#define SK_FORCE_REUSE 2
446
Pavel Emelyanovef64a542012-02-21 07:31:34 +0000447static inline int sk_peek_offset(struct sock *sk, int flags)
448{
449 if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0))
450 return sk->sk_peek_off;
451 else
452 return 0;
453}
454
455static inline void sk_peek_offset_bwd(struct sock *sk, int val)
456{
457 if (sk->sk_peek_off >= 0) {
458 if (sk->sk_peek_off >= val)
459 sk->sk_peek_off -= val;
460 else
461 sk->sk_peek_off = 0;
462 }
463}
464
465static inline void sk_peek_offset_fwd(struct sock *sk, int val)
466{
467 if (sk->sk_peek_off >= 0)
468 sk->sk_peek_off += val;
469}
470
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471/*
472 * Hashed lists helper routines
473 */
Li Zefanc4146642010-02-08 23:18:45 +0000474static inline struct sock *sk_entry(const struct hlist_node *node)
475{
476 return hlist_entry(node, struct sock, sk_node);
477}
478
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700479static inline struct sock *__sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480{
481 return hlist_entry(head->first, struct sock, sk_node);
482}
483
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700484static inline struct sock *sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485{
486 return hlist_empty(head) ? NULL : __sk_head(head);
487}
488
Eric Dumazet88ab1932008-11-16 19:39:21 -0800489static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
490{
491 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
492}
493
494static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
495{
496 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
497}
498
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700499static inline struct sock *sk_next(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500{
501 return sk->sk_node.next ?
502 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
503}
504
Eric Dumazet88ab1932008-11-16 19:39:21 -0800505static inline struct sock *sk_nulls_next(const struct sock *sk)
506{
507 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
508 hlist_nulls_entry(sk->sk_nulls_node.next,
509 struct sock, sk_nulls_node) :
510 NULL;
511}
512
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000513static inline bool sk_unhashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514{
515 return hlist_unhashed(&sk->sk_node);
516}
517
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000518static inline bool sk_hashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519{
Akinobu Mitada753be2006-04-28 15:21:23 -0700520 return !sk_unhashed(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521}
522
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000523static inline void sk_node_init(struct hlist_node *node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524{
525 node->pprev = NULL;
526}
527
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000528static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800529{
530 node->pprev = NULL;
531}
532
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000533static inline void __sk_del_node(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534{
535 __hlist_del(&sk->sk_node);
536}
537
stephen hemminger808f5112010-02-22 07:57:18 +0000538/* NB: equivalent to hlist_del_init_rcu */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000539static inline bool __sk_del_node_init(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540{
541 if (sk_hashed(sk)) {
542 __sk_del_node(sk);
543 sk_node_init(&sk->sk_node);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000544 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545 }
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000546 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547}
548
549/* Grab socket reference count. This operation is valid only
550 when sk is ALREADY grabbed f.e. it is found in hash table
551 or a list and the lookup is made under lock preventing hash table
552 modifications.
553 */
554
555static inline void sock_hold(struct sock *sk)
556{
557 atomic_inc(&sk->sk_refcnt);
558}
559
560/* Ungrab socket in the context, which assumes that socket refcnt
561 cannot hit zero, f.e. it is true in context of any socketcall.
562 */
563static inline void __sock_put(struct sock *sk)
564{
565 atomic_dec(&sk->sk_refcnt);
566}
567
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000568static inline bool sk_del_node_init(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569{
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000570 bool rc = __sk_del_node_init(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571
572 if (rc) {
573 /* paranoid for a while -acme */
574 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
575 __sock_put(sk);
576 }
577 return rc;
578}
stephen hemminger808f5112010-02-22 07:57:18 +0000579#define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000581static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700582{
583 if (sk_hashed(sk)) {
Eric Dumazet88ab1932008-11-16 19:39:21 -0800584 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000585 return true;
Eric Dumazet271b72c2008-10-29 02:11:14 -0700586 }
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000587 return false;
Eric Dumazet271b72c2008-10-29 02:11:14 -0700588}
589
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000590static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700591{
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000592 bool rc = __sk_nulls_del_node_init_rcu(sk);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700593
594 if (rc) {
595 /* paranoid for a while -acme */
596 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
597 __sock_put(sk);
598 }
599 return rc;
600}
601
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000602static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603{
604 hlist_add_head(&sk->sk_node, list);
605}
606
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000607static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608{
609 sock_hold(sk);
610 __sk_add_node(sk, list);
611}
612
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000613static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
stephen hemminger808f5112010-02-22 07:57:18 +0000614{
615 sock_hold(sk);
616 hlist_add_head_rcu(&sk->sk_node, list);
617}
618
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000619static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700620{
Eric Dumazet88ab1932008-11-16 19:39:21 -0800621 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700622}
623
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000624static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700625{
626 sock_hold(sk);
Eric Dumazet88ab1932008-11-16 19:39:21 -0800627 __sk_nulls_add_node_rcu(sk, list);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700628}
629
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000630static inline void __sk_del_bind_node(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631{
632 __hlist_del(&sk->sk_bind_node);
633}
634
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000635static inline void sk_add_bind_node(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636 struct hlist_head *list)
637{
638 hlist_add_head(&sk->sk_bind_node, list);
639}
640
Sasha Levinb67bfe02013-02-27 17:06:00 -0800641#define sk_for_each(__sk, list) \
642 hlist_for_each_entry(__sk, list, sk_node)
643#define sk_for_each_rcu(__sk, list) \
644 hlist_for_each_entry_rcu(__sk, list, sk_node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800645#define sk_nulls_for_each(__sk, node, list) \
646 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
647#define sk_nulls_for_each_rcu(__sk, node, list) \
648 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
Sasha Levinb67bfe02013-02-27 17:06:00 -0800649#define sk_for_each_from(__sk) \
650 hlist_for_each_entry_from(__sk, sk_node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800651#define sk_nulls_for_each_from(__sk, node) \
652 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
653 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
Sasha Levinb67bfe02013-02-27 17:06:00 -0800654#define sk_for_each_safe(__sk, tmp, list) \
655 hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
656#define sk_for_each_bound(__sk, list) \
657 hlist_for_each_entry(__sk, list, sk_bind_node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658
Eric W. Biedermanc336d142012-05-24 17:56:43 -0600659static inline struct user_namespace *sk_user_ns(struct sock *sk)
660{
661 /* Careful only use this in a context where these parameters
662 * can not change and must all be valid, such as recvmsg from
663 * userspace.
664 */
665 return sk->sk_socket->file->f_cred->user_ns;
666}
667
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668/* Sock flags */
669enum sock_flags {
670 SOCK_DEAD,
671 SOCK_DONE,
672 SOCK_URGINLINE,
673 SOCK_KEEPOPEN,
674 SOCK_LINGER,
675 SOCK_DESTROY,
676 SOCK_BROADCAST,
677 SOCK_TIMESTAMP,
678 SOCK_ZAPPED,
679 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
680 SOCK_DBG, /* %SO_DEBUG setting */
681 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700682 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
684 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
Mel Gorman7cb02402012-07-31 16:44:16 -0700685 SOCK_MEMALLOC, /* VM depends on this socket for swapping */
Patrick Ohly20d49472009-02-12 05:03:38 +0000686 SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */
687 SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
688 SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */
689 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
690 SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */
691 SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
692 SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
Eric Dumazetbcdce712009-10-06 17:28:29 -0700693 SOCK_FASYNC, /* fasync() active */
Neil Horman3b885782009-10-12 13:26:31 -0700694 SOCK_RXQ_OVFL,
Shirley Ma1cdebb42011-07-06 12:17:30 +0000695 SOCK_ZEROCOPY, /* buffers from userspace */
Johannes Berg6e3e9392011-11-09 10:15:42 +0100696 SOCK_WIFI_STATUS, /* push wifi status to userspace */
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000697 SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
698 * Will use last 4 bytes of packet sent from
699 * user-space instead.
700 */
Vincent Bernatd59577b2013-01-16 22:55:49 +0100701 SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */
Keller, Jacob E7d4c04f2013-03-28 11:19:25 +0000702 SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703};
704
Ralf Baechle53b924b2005-08-23 10:11:30 -0700705static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
706{
707 nsk->sk_flags = osk->sk_flags;
708}
709
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
711{
712 __set_bit(flag, &sk->sk_flags);
713}
714
715static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
716{
717 __clear_bit(flag, &sk->sk_flags);
718}
719
Eric Dumazet1b23a5d2012-05-16 05:57:07 +0000720static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721{
722 return test_bit(flag, &sk->sk_flags);
723}
724
Mel Gormanc93bdd02012-07-31 16:44:19 -0700725#ifdef CONFIG_NET
726extern struct static_key memalloc_socks;
727static inline int sk_memalloc_socks(void)
728{
729 return static_key_false(&memalloc_socks);
730}
731#else
732
733static inline int sk_memalloc_socks(void)
734{
735 return 0;
736}
737
738#endif
739
Mel Gorman99a1dec2012-07-31 16:44:14 -0700740static inline gfp_t sk_gfp_atomic(struct sock *sk, gfp_t gfp_mask)
741{
Mel Gorman7cb02402012-07-31 16:44:16 -0700742 return GFP_ATOMIC | (sk->sk_allocation & __GFP_MEMALLOC);
Mel Gorman99a1dec2012-07-31 16:44:14 -0700743}
744
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745static inline void sk_acceptq_removed(struct sock *sk)
746{
747 sk->sk_ack_backlog--;
748}
749
750static inline void sk_acceptq_added(struct sock *sk)
751{
752 sk->sk_ack_backlog++;
753}
754
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000755static inline bool sk_acceptq_is_full(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756{
David S. Miller64a14652007-03-06 11:21:05 -0800757 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758}
759
760/*
761 * Compute minimal free write space needed to queue new packets.
762 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000763static inline int sk_stream_min_wspace(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764{
Eric Dumazet8df09ea2007-12-21 03:07:41 -0800765 return sk->sk_wmem_queued >> 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766}
767
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000768static inline int sk_stream_wspace(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769{
770 return sk->sk_sndbuf - sk->sk_wmem_queued;
771}
772
Joe Perches69336bd2013-09-22 10:32:26 -0700773void sk_stream_write_space(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774
Zhu Yi8eae9392010-03-04 18:01:40 +0000775/* OOB backlog add */
Zhu Yia3a858f2010-03-04 18:01:47 +0000776static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800777{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000778 /* dont let skb dst not refcounted, we are going to leave rcu lock */
779 skb_dst_force(skb);
780
781 if (!sk->sk_backlog.tail)
782 sk->sk_backlog.head = skb;
783 else
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800784 sk->sk_backlog.tail->next = skb;
Eric Dumazet7fee2262010-05-11 23:19:48 +0000785
786 sk->sk_backlog.tail = skb;
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800787 skb->next = NULL;
788}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789
Eric Dumazetc3774112010-04-27 15:13:20 -0700790/*
791 * Take into account size of receive queue and backlog queue
Eric Dumazet0fd7bac2011-12-21 07:11:44 +0000792 * Do not take into account this skb truesize,
793 * to allow even a single big packet to come.
Eric Dumazetc3774112010-04-27 15:13:20 -0700794 */
Eric Dumazetf545a382012-04-22 23:34:26 +0000795static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb,
796 unsigned int limit)
Eric Dumazetc3774112010-04-27 15:13:20 -0700797{
798 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
799
Eric Dumazetf545a382012-04-22 23:34:26 +0000800 return qsize > limit;
Eric Dumazetc3774112010-04-27 15:13:20 -0700801}
802
Zhu Yi8eae9392010-03-04 18:01:40 +0000803/* The per-socket spinlock must be held here. */
Eric Dumazetf545a382012-04-22 23:34:26 +0000804static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
805 unsigned int limit)
Zhu Yi8eae9392010-03-04 18:01:40 +0000806{
Eric Dumazetf545a382012-04-22 23:34:26 +0000807 if (sk_rcvqueues_full(sk, skb, limit))
Zhu Yi8eae9392010-03-04 18:01:40 +0000808 return -ENOBUFS;
809
Zhu Yia3a858f2010-03-04 18:01:47 +0000810 __sk_add_backlog(sk, skb);
Zhu Yi8eae9392010-03-04 18:01:40 +0000811 sk->sk_backlog.len += skb->truesize;
812 return 0;
813}
814
Joe Perches69336bd2013-09-22 10:32:26 -0700815int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
Mel Gormanb4b9e352012-07-31 16:44:26 -0700816
Peter Zijlstrac57943a2008-10-07 14:18:42 -0700817static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
818{
Mel Gormanb4b9e352012-07-31 16:44:26 -0700819 if (sk_memalloc_socks() && skb_pfmemalloc(skb))
820 return __sk_backlog_rcv(sk, skb);
821
Peter Zijlstrac57943a2008-10-07 14:18:42 -0700822 return sk->sk_backlog_rcv(sk, skb);
823}
824
Tom Herbertfe477552013-12-22 18:54:31 +0800825static inline void sock_rps_record_flow_hash(__u32 hash)
David S. Millerc58dc012010-04-27 15:05:31 -0700826{
827#ifdef CONFIG_RPS
828 struct rps_sock_flow_table *sock_flow_table;
829
830 rcu_read_lock();
831 sock_flow_table = rcu_dereference(rps_sock_flow_table);
Tom Herbertfe477552013-12-22 18:54:31 +0800832 rps_record_sock_flow(sock_flow_table, hash);
David S. Millerc58dc012010-04-27 15:05:31 -0700833 rcu_read_unlock();
834#endif
835}
836
Tom Herbertfe477552013-12-22 18:54:31 +0800837static inline void sock_rps_reset_flow_hash(__u32 hash)
David S. Millerc58dc012010-04-27 15:05:31 -0700838{
839#ifdef CONFIG_RPS
840 struct rps_sock_flow_table *sock_flow_table;
841
842 rcu_read_lock();
843 sock_flow_table = rcu_dereference(rps_sock_flow_table);
Tom Herbertfe477552013-12-22 18:54:31 +0800844 rps_reset_sock_flow(sock_flow_table, hash);
David S. Millerc58dc012010-04-27 15:05:31 -0700845 rcu_read_unlock();
846#endif
847}
848
Tom Herbertfe477552013-12-22 18:54:31 +0800849static inline void sock_rps_record_flow(const struct sock *sk)
850{
Zhi Yong Wuc9d8ca02014-01-01 04:31:01 +0800851#ifdef CONFIG_RPS
Tom Herbertfe477552013-12-22 18:54:31 +0800852 sock_rps_record_flow_hash(sk->sk_rxhash);
Zhi Yong Wuc9d8ca02014-01-01 04:31:01 +0800853#endif
Tom Herbertfe477552013-12-22 18:54:31 +0800854}
855
856static inline void sock_rps_reset_flow(const struct sock *sk)
857{
Zhi Yong Wuc9d8ca02014-01-01 04:31:01 +0800858#ifdef CONFIG_RPS
Tom Herbertfe477552013-12-22 18:54:31 +0800859 sock_rps_reset_flow_hash(sk->sk_rxhash);
Zhi Yong Wuc9d8ca02014-01-01 04:31:01 +0800860#endif
Tom Herbertfe477552013-12-22 18:54:31 +0800861}
862
Tom Herbertbdeab992011-08-14 19:45:55 +0000863static inline void sock_rps_save_rxhash(struct sock *sk,
864 const struct sk_buff *skb)
David S. Millerc58dc012010-04-27 15:05:31 -0700865{
866#ifdef CONFIG_RPS
Tom Herbert61b905d2014-03-24 15:34:47 -0700867 if (unlikely(sk->sk_rxhash != skb->hash)) {
David S. Millerc58dc012010-04-27 15:05:31 -0700868 sock_rps_reset_flow(sk);
Tom Herbert61b905d2014-03-24 15:34:47 -0700869 sk->sk_rxhash = skb->hash;
David S. Millerc58dc012010-04-27 15:05:31 -0700870 }
871#endif
872}
873
Tom Herbertbdeab992011-08-14 19:45:55 +0000874static inline void sock_rps_reset_rxhash(struct sock *sk)
875{
876#ifdef CONFIG_RPS
877 sock_rps_reset_flow(sk);
878 sk->sk_rxhash = 0;
879#endif
880}
881
Stephen Hemmingercfcabdc2007-10-09 01:59:42 -0700882#define sk_wait_event(__sk, __timeo, __condition) \
883 ({ int __rc; \
884 release_sock(__sk); \
885 __rc = __condition; \
886 if (!__rc) { \
887 *(__timeo) = schedule_timeout(*(__timeo)); \
888 } \
889 lock_sock(__sk); \
890 __rc = __condition; \
891 __rc; \
892 })
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893
Joe Perches69336bd2013-09-22 10:32:26 -0700894int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
895int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
896void sk_stream_wait_close(struct sock *sk, long timeo_p);
897int sk_stream_error(struct sock *sk, int flags, int err);
898void sk_stream_kill_queues(struct sock *sk);
899void sk_set_memalloc(struct sock *sk);
900void sk_clear_memalloc(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700901
Joe Perches69336bd2013-09-22 10:32:26 -0700902int sk_wait_data(struct sock *sk, long *timeo);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700904struct request_sock_ops;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800905struct timewait_sock_ops;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -0800906struct inet_hashinfo;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -0700907struct raw_hashinfo;
Paul Gortmakerde477252011-05-26 13:46:22 -0400908struct module;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700909
Eric Dumazetf77d6022013-05-09 10:28:16 +0000910/*
911 * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
912 * un-modified. Special care is taken when initializing object to zero.
913 */
914static inline void sk_prot_clear_nulls(struct sock *sk, int size)
915{
916 if (offsetof(struct sock, sk_node.next) != 0)
917 memset(sk, 0, offsetof(struct sock, sk_node.next));
918 memset(&sk->sk_node.pprev, 0,
919 size - offsetof(struct sock, sk_node.pprev));
920}
921
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922/* Networking protocol blocks we attach to sockets.
923 * socket layer -> transport layer interface
924 * transport -> network interface is defined by struct inet_proto
925 */
926struct proto {
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000927 void (*close)(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700928 long timeout);
929 int (*connect)(struct sock *sk,
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000930 struct sockaddr *uaddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931 int addr_len);
932 int (*disconnect)(struct sock *sk, int flags);
933
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000934 struct sock * (*accept)(struct sock *sk, int flags, int *err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700935
936 int (*ioctl)(struct sock *sk, int cmd,
937 unsigned long arg);
938 int (*init)(struct sock *sk);
Brian Haley7d06b2e2008-06-14 17:04:49 -0700939 void (*destroy)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940 void (*shutdown)(struct sock *sk, int how);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000941 int (*setsockopt)(struct sock *sk, int level,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700943 unsigned int optlen);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000944 int (*getsockopt)(struct sock *sk, int level,
945 int optname, char __user *optval,
946 int __user *option);
Alexey Dobriyanaf01d532008-08-28 02:53:51 -0700947#ifdef CONFIG_COMPAT
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800948 int (*compat_setsockopt)(struct sock *sk,
949 int level,
950 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700951 unsigned int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800952 int (*compat_getsockopt)(struct sock *sk,
953 int level,
954 int optname, char __user *optval,
955 int __user *option);
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000956 int (*compat_ioctl)(struct sock *sk,
957 unsigned int cmd, unsigned long arg);
Alexey Dobriyanaf01d532008-08-28 02:53:51 -0700958#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
960 struct msghdr *msg, size_t len);
961 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
962 struct msghdr *msg,
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000963 size_t len, int noblock, int flags,
964 int *addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965 int (*sendpage)(struct sock *sk, struct page *page,
966 int offset, size_t size, int flags);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000967 int (*bind)(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968 struct sockaddr *uaddr, int addr_len);
969
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000970 int (*backlog_rcv) (struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971 struct sk_buff *skb);
972
Eric Dumazet46d3cea2012-07-11 05:50:31 +0000973 void (*release_cb)(struct sock *sk);
Eric Dumazet563d34d2012-07-23 09:48:52 +0200974 void (*mtu_reduced)(struct sock *sk);
Eric Dumazet46d3cea2012-07-11 05:50:31 +0000975
Linus Torvalds1da177e2005-04-16 15:20:36 -0700976 /* Keeping track of sk's, looking them up, and port selection methods. */
977 void (*hash)(struct sock *sk);
978 void (*unhash)(struct sock *sk);
Eric Dumazet719f8352010-09-08 05:08:44 +0000979 void (*rehash)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980 int (*get_port)(struct sock *sk, unsigned short snum);
Octavian Purdilafcbdf092010-12-16 14:26:56 -0800981 void (*clear_sk)(struct sock *sk, int size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982
Eric Dumazet286ab3d2007-11-05 23:38:39 -0800983 /* Keeping track of sockets in use */
Eric Dumazet65f76512008-01-03 20:46:48 -0800984#ifdef CONFIG_PROC_FS
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -0700985 unsigned int inuse_idx;
Eric Dumazet65f76512008-01-03 20:46:48 -0800986#endif
Arnaldo Carvalho de Meloebb53d72007-11-21 22:08:50 +0800987
Eric Dumazetc9bee3b72013-07-22 20:27:07 -0700988 bool (*stream_memory_free)(const struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989 /* Memory pressure */
Pavel Emelyanov5c52ba12008-07-16 20:28:10 -0700990 void (*enter_memory_pressure)(struct sock *sk);
Eric Dumazet8d987e52010-11-09 23:24:26 +0000991 atomic_long_t *memory_allocated; /* Current allocated memory. */
Eric Dumazet17483762008-11-25 21:16:35 -0800992 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700993 /*
994 * Pressure flag: try to collapse.
995 * Technical note: it is used by multiple contexts non atomically.
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800996 * All the __sk_mem_schedule() is of this nature: accounting
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997 * is strict, actions are advisory and have some latency.
998 */
999 int *memory_pressure;
Eric Dumazet8d987e52010-11-09 23:24:26 +00001000 long *sysctl_mem;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001 int *sysctl_wmem;
1002 int *sysctl_rmem;
1003 int max_header;
Changli Gao7ba42912010-07-10 20:41:55 +00001004 bool no_autobind;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005
Eric Dumazet271b72c2008-10-29 02:11:14 -07001006 struct kmem_cache *slab;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 unsigned int obj_size;
Eric Dumazet271b72c2008-10-29 02:11:14 -07001008 int slab_flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009
Eric Dumazetdd24c002008-11-25 21:17:14 -08001010 struct percpu_counter *orphan_count;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -07001011
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -07001012 struct request_sock_ops *rsk_prot;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -08001013 struct timewait_sock_ops *twsk_prot;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07001014
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -07001015 union {
1016 struct inet_hashinfo *hashinfo;
Eric Dumazet645ca702008-10-29 01:41:45 -07001017 struct udp_table *udp_table;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -07001018 struct raw_hashinfo *raw_hash;
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -07001019 } h;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -08001020
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021 struct module *owner;
1022
1023 char name[32];
1024
1025 struct list_head node;
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -07001026#ifdef SOCK_REFCNT_DEBUG
1027 atomic_t socks;
1028#endif
Andrew Mortonc255a452012-07-31 16:43:02 -07001029#ifdef CONFIG_MEMCG_KMEM
Glauber Costae1aab162011-12-11 21:47:03 +00001030 /*
1031 * cgroup specific init/deinit functions. Called once for all
1032 * protocols that implement it, from cgroups populate function.
1033 * This function has to setup any files the protocol want to
1034 * appear in the kmem cgroup filesystem.
1035 */
Glauber Costa1d62e432012-04-09 19:36:33 -03001036 int (*init_cgroup)(struct mem_cgroup *memcg,
Glauber Costae1aab162011-12-11 21:47:03 +00001037 struct cgroup_subsys *ss);
Glauber Costa1d62e432012-04-09 19:36:33 -03001038 void (*destroy_cgroup)(struct mem_cgroup *memcg);
Glauber Costae1aab162011-12-11 21:47:03 +00001039 struct cg_proto *(*proto_cgroup)(struct mem_cgroup *memcg);
1040#endif
1041};
1042
Glauber Costa3f134612012-05-29 15:07:11 -07001043/*
1044 * Bits in struct cg_proto.flags
1045 */
1046enum cg_proto_flags {
1047 /* Currently active and new sockets should be assigned to cgroups */
1048 MEMCG_SOCK_ACTIVE,
1049 /* It was ever activated; we must disarm static keys on destruction */
1050 MEMCG_SOCK_ACTIVATED,
1051};
1052
Glauber Costae1aab162011-12-11 21:47:03 +00001053struct cg_proto {
Eric W. Biederman2e685cad2013-10-19 16:26:19 -07001054 struct res_counter memory_allocated; /* Current allocated memory. */
1055 struct percpu_counter sockets_allocated; /* Current number of sockets. */
1056 int memory_pressure;
1057 long sysctl_mem[3];
Glauber Costa3f134612012-05-29 15:07:11 -07001058 unsigned long flags;
Glauber Costae1aab162011-12-11 21:47:03 +00001059 /*
1060 * memcg field is used to find which memcg we belong directly
1061 * Each memcg struct can hold more than one cg_proto, so container_of
1062 * won't really cut.
1063 *
1064 * The elegant solution would be having an inverse function to
1065 * proto_cgroup in struct proto, but that means polluting the structure
1066 * for everybody, instead of just for memcg users.
1067 */
1068 struct mem_cgroup *memcg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069};
1070
Joe Perches69336bd2013-09-22 10:32:26 -07001071int proto_register(struct proto *prot, int alloc_slab);
1072void proto_unregister(struct proto *prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001073
Glauber Costa3f134612012-05-29 15:07:11 -07001074static inline bool memcg_proto_active(struct cg_proto *cg_proto)
1075{
1076 return test_bit(MEMCG_SOCK_ACTIVE, &cg_proto->flags);
1077}
1078
1079static inline bool memcg_proto_activated(struct cg_proto *cg_proto)
1080{
1081 return test_bit(MEMCG_SOCK_ACTIVATED, &cg_proto->flags);
1082}
1083
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -07001084#ifdef SOCK_REFCNT_DEBUG
1085static inline void sk_refcnt_debug_inc(struct sock *sk)
1086{
1087 atomic_inc(&sk->sk_prot->socks);
1088}
1089
1090static inline void sk_refcnt_debug_dec(struct sock *sk)
1091{
1092 atomic_dec(&sk->sk_prot->socks);
1093 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
1094 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
1095}
1096
Ying Xuedec34fb2013-02-15 22:28:25 +00001097static inline void sk_refcnt_debug_release(const struct sock *sk)
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -07001098{
1099 if (atomic_read(&sk->sk_refcnt) != 1)
1100 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
1101 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
1102}
1103#else /* SOCK_REFCNT_DEBUG */
1104#define sk_refcnt_debug_inc(sk) do { } while (0)
1105#define sk_refcnt_debug_dec(sk) do { } while (0)
1106#define sk_refcnt_debug_release(sk) do { } while (0)
1107#endif /* SOCK_REFCNT_DEBUG */
1108
Andrew Mortonc255a452012-07-31 16:43:02 -07001109#if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_NET)
Ingo Molnarc5905af2012-02-24 08:31:31 +01001110extern struct static_key memcg_socket_limit_enabled;
Glauber Costae1aab162011-12-11 21:47:03 +00001111static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1112 struct cg_proto *cg_proto)
1113{
1114 return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg));
1115}
Ingo Molnarc5905af2012-02-24 08:31:31 +01001116#define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
Glauber Costae1aab162011-12-11 21:47:03 +00001117#else
1118#define mem_cgroup_sockets_enabled 0
1119static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1120 struct cg_proto *cg_proto)
1121{
1122 return NULL;
1123}
1124#endif
1125
Eric Dumazetc9bee3b72013-07-22 20:27:07 -07001126static inline bool sk_stream_memory_free(const struct sock *sk)
1127{
1128 if (sk->sk_wmem_queued >= sk->sk_sndbuf)
1129 return false;
1130
1131 return sk->sk_prot->stream_memory_free ?
1132 sk->sk_prot->stream_memory_free(sk) : true;
1133}
1134
Eric Dumazet64dc6132013-07-22 20:26:31 -07001135static inline bool sk_stream_is_writeable(const struct sock *sk)
1136{
Eric Dumazetc9bee3b72013-07-22 20:27:07 -07001137 return sk_stream_wspace(sk) >= sk_stream_min_wspace(sk) &&
1138 sk_stream_memory_free(sk);
Eric Dumazet64dc6132013-07-22 20:26:31 -07001139}
Glauber Costae1aab162011-12-11 21:47:03 +00001140
Eric Dumazetc9bee3b72013-07-22 20:27:07 -07001141
Glauber Costa180d8cd2011-12-11 21:47:02 +00001142static inline bool sk_has_memory_pressure(const struct sock *sk)
1143{
1144 return sk->sk_prot->memory_pressure != NULL;
1145}
1146
1147static inline bool sk_under_memory_pressure(const struct sock *sk)
1148{
1149 if (!sk->sk_prot->memory_pressure)
1150 return false;
Glauber Costae1aab162011-12-11 21:47:03 +00001151
1152 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
Eric W. Biederman2e685cad2013-10-19 16:26:19 -07001153 return !!sk->sk_cgrp->memory_pressure;
Glauber Costae1aab162011-12-11 21:47:03 +00001154
Christoph Paasch35b87f62013-10-23 12:49:21 -07001155 return !!*sk->sk_prot->memory_pressure;
Glauber Costa180d8cd2011-12-11 21:47:02 +00001156}
1157
1158static inline void sk_leave_memory_pressure(struct sock *sk)
1159{
1160 int *memory_pressure = sk->sk_prot->memory_pressure;
1161
Glauber Costae1aab162011-12-11 21:47:03 +00001162 if (!memory_pressure)
1163 return;
1164
1165 if (*memory_pressure)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001166 *memory_pressure = 0;
Glauber Costae1aab162011-12-11 21:47:03 +00001167
1168 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1169 struct cg_proto *cg_proto = sk->sk_cgrp;
1170 struct proto *prot = sk->sk_prot;
1171
1172 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
Eric W. Biederman7f2cbdc2013-12-04 20:12:04 -08001173 cg_proto->memory_pressure = 0;
Glauber Costae1aab162011-12-11 21:47:03 +00001174 }
1175
Glauber Costa180d8cd2011-12-11 21:47:02 +00001176}
1177
1178static inline void sk_enter_memory_pressure(struct sock *sk)
1179{
Glauber Costae1aab162011-12-11 21:47:03 +00001180 if (!sk->sk_prot->enter_memory_pressure)
1181 return;
1182
1183 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1184 struct cg_proto *cg_proto = sk->sk_cgrp;
1185 struct proto *prot = sk->sk_prot;
1186
1187 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
Eric W. Biederman7f2cbdc2013-12-04 20:12:04 -08001188 cg_proto->memory_pressure = 1;
Glauber Costae1aab162011-12-11 21:47:03 +00001189 }
1190
1191 sk->sk_prot->enter_memory_pressure(sk);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001192}
1193
1194static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1195{
1196 long *prot = sk->sk_prot->sysctl_mem;
Glauber Costae1aab162011-12-11 21:47:03 +00001197 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1198 prot = sk->sk_cgrp->sysctl_mem;
Glauber Costa180d8cd2011-12-11 21:47:02 +00001199 return prot[index];
1200}
1201
Glauber Costae1aab162011-12-11 21:47:03 +00001202static inline void memcg_memory_allocated_add(struct cg_proto *prot,
1203 unsigned long amt,
1204 int *parent_status)
1205{
1206 struct res_counter *fail;
1207 int ret;
1208
Eric W. Biederman2e685cad2013-10-19 16:26:19 -07001209 ret = res_counter_charge_nofail(&prot->memory_allocated,
Glauber Costa0e90b312012-01-20 04:57:16 +00001210 amt << PAGE_SHIFT, &fail);
Glauber Costae1aab162011-12-11 21:47:03 +00001211 if (ret < 0)
1212 *parent_status = OVER_LIMIT;
1213}
1214
1215static inline void memcg_memory_allocated_sub(struct cg_proto *prot,
1216 unsigned long amt)
1217{
Eric W. Biederman2e685cad2013-10-19 16:26:19 -07001218 res_counter_uncharge(&prot->memory_allocated, amt << PAGE_SHIFT);
Glauber Costae1aab162011-12-11 21:47:03 +00001219}
1220
1221static inline u64 memcg_memory_allocated_read(struct cg_proto *prot)
1222{
1223 u64 ret;
Eric W. Biederman2e685cad2013-10-19 16:26:19 -07001224 ret = res_counter_read_u64(&prot->memory_allocated, RES_USAGE);
Glauber Costae1aab162011-12-11 21:47:03 +00001225 return ret >> PAGE_SHIFT;
1226}
1227
Glauber Costa180d8cd2011-12-11 21:47:02 +00001228static inline long
1229sk_memory_allocated(const struct sock *sk)
1230{
1231 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001232 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1233 return memcg_memory_allocated_read(sk->sk_cgrp);
1234
Glauber Costa180d8cd2011-12-11 21:47:02 +00001235 return atomic_long_read(prot->memory_allocated);
1236}
1237
1238static inline long
Glauber Costae1aab162011-12-11 21:47:03 +00001239sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001240{
1241 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001242
1243 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1244 memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status);
1245 /* update the root cgroup regardless */
1246 atomic_long_add_return(amt, prot->memory_allocated);
1247 return memcg_memory_allocated_read(sk->sk_cgrp);
1248 }
1249
Glauber Costa180d8cd2011-12-11 21:47:02 +00001250 return atomic_long_add_return(amt, prot->memory_allocated);
1251}
1252
1253static inline void
Glauber Costa0e90b312012-01-20 04:57:16 +00001254sk_memory_allocated_sub(struct sock *sk, int amt)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001255{
1256 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001257
Glauber Costa0e90b312012-01-20 04:57:16 +00001258 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
Glauber Costae1aab162011-12-11 21:47:03 +00001259 memcg_memory_allocated_sub(sk->sk_cgrp, amt);
1260
Glauber Costa180d8cd2011-12-11 21:47:02 +00001261 atomic_long_sub(amt, prot->memory_allocated);
1262}
1263
1264static inline void sk_sockets_allocated_dec(struct sock *sk)
1265{
1266 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001267
1268 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1269 struct cg_proto *cg_proto = sk->sk_cgrp;
1270
1271 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
Eric W. Biederman2e685cad2013-10-19 16:26:19 -07001272 percpu_counter_dec(&cg_proto->sockets_allocated);
Glauber Costae1aab162011-12-11 21:47:03 +00001273 }
1274
Glauber Costa180d8cd2011-12-11 21:47:02 +00001275 percpu_counter_dec(prot->sockets_allocated);
1276}
1277
1278static inline void sk_sockets_allocated_inc(struct sock *sk)
1279{
1280 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001281
1282 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1283 struct cg_proto *cg_proto = sk->sk_cgrp;
1284
1285 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
Eric W. Biederman2e685cad2013-10-19 16:26:19 -07001286 percpu_counter_inc(&cg_proto->sockets_allocated);
Glauber Costae1aab162011-12-11 21:47:03 +00001287 }
1288
Glauber Costa180d8cd2011-12-11 21:47:02 +00001289 percpu_counter_inc(prot->sockets_allocated);
1290}
1291
1292static inline int
1293sk_sockets_allocated_read_positive(struct sock *sk)
1294{
1295 struct proto *prot = sk->sk_prot;
1296
Glauber Costae1aab162011-12-11 21:47:03 +00001297 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
Eric W. Biederman2e685cad2013-10-19 16:26:19 -07001298 return percpu_counter_read_positive(&sk->sk_cgrp->sockets_allocated);
Glauber Costae1aab162011-12-11 21:47:03 +00001299
Eric Dumazet518fbf92012-04-28 23:21:56 +00001300 return percpu_counter_read_positive(prot->sockets_allocated);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001301}
1302
1303static inline int
1304proto_sockets_allocated_sum_positive(struct proto *prot)
1305{
1306 return percpu_counter_sum_positive(prot->sockets_allocated);
1307}
1308
1309static inline long
1310proto_memory_allocated(struct proto *prot)
1311{
1312 return atomic_long_read(prot->memory_allocated);
1313}
1314
1315static inline bool
1316proto_memory_pressure(struct proto *prot)
1317{
1318 if (!prot->memory_pressure)
1319 return false;
1320 return !!*prot->memory_pressure;
1321}
1322
Eric Dumazet65f76512008-01-03 20:46:48 -08001323
1324#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325/* Called with local bh disabled */
Joe Perches69336bd2013-09-22 10:32:26 -07001326void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1327int sock_prot_inuse_get(struct net *net, struct proto *proto);
Eric Dumazet65f76512008-01-03 20:46:48 -08001328#else
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001329static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -07001330 int inc)
Eric Dumazet65f76512008-01-03 20:46:48 -08001331{
1332}
Eric Dumazet65f76512008-01-03 20:46:48 -08001333#endif
1334
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335
Arnaldo Carvalho de Melo614c6cb2005-08-09 19:47:37 -07001336/* With per-bucket locks this operation is not-atomic, so that
1337 * this version is not worse.
1338 */
1339static inline void __sk_prot_rehash(struct sock *sk)
1340{
1341 sk->sk_prot->unhash(sk);
1342 sk->sk_prot->hash(sk);
1343}
1344
Octavian Purdilafcbdf092010-12-16 14:26:56 -08001345void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
1346
Linus Torvalds1da177e2005-04-16 15:20:36 -07001347/* About 10 seconds */
1348#define SOCK_DESTROY_TIME (10*HZ)
1349
1350/* Sockets 0-1023 can't be bound to unless you are superuser */
1351#define PROT_SOCK 1024
1352
1353#define SHUTDOWN_MASK 3
1354#define RCV_SHUTDOWN 1
1355#define SEND_SHUTDOWN 2
1356
1357#define SOCK_SNDBUF_LOCK 1
1358#define SOCK_RCVBUF_LOCK 2
1359#define SOCK_BINDADDR_LOCK 4
1360#define SOCK_BINDPORT_LOCK 8
1361
1362/* sock_iocb: used to kick off async processing of socket ios */
1363struct sock_iocb {
1364 struct list_head list;
1365
1366 int flags;
1367 int size;
1368 struct socket *sock;
1369 struct sock *sk;
1370 struct scm_cookie *scm;
1371 struct msghdr *msg, async_msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372 struct kiocb *kiocb;
1373};
1374
1375static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
1376{
1377 return (struct sock_iocb *)iocb->private;
1378}
1379
1380static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
1381{
1382 return si->kiocb;
1383}
1384
1385struct socket_alloc {
1386 struct socket socket;
1387 struct inode vfs_inode;
1388};
1389
1390static inline struct socket *SOCKET_I(struct inode *inode)
1391{
1392 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1393}
1394
1395static inline struct inode *SOCK_INODE(struct socket *socket)
1396{
1397 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1398}
1399
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001400/*
1401 * Functions for memory accounting
1402 */
Joe Perches69336bd2013-09-22 10:32:26 -07001403int __sk_mem_schedule(struct sock *sk, int size, int kind);
1404void __sk_mem_reclaim(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001405
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001406#define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1407#define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1408#define SK_MEM_SEND 0
1409#define SK_MEM_RECV 1
Linus Torvalds1da177e2005-04-16 15:20:36 -07001410
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001411static inline int sk_mem_pages(int amt)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001413 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001414}
1415
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001416static inline bool sk_has_account(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001417{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001418 /* return true if protocol supports memory accounting */
1419 return !!sk->sk_prot->memory_allocated;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420}
1421
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001422static inline bool sk_wmem_schedule(struct sock *sk, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001423{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001424 if (!sk_has_account(sk))
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001425 return true;
Herbert Xud80d99d2005-09-01 17:48:23 -07001426 return size <= sk->sk_forward_alloc ||
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001427 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1428}
1429
Mel Gormanc76562b2012-07-31 16:44:41 -07001430static inline bool
Chuck Lever35c448a2012-09-17 14:09:11 -07001431sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size)
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001432{
1433 if (!sk_has_account(sk))
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001434 return true;
Mel Gormanc76562b2012-07-31 16:44:41 -07001435 return size<= sk->sk_forward_alloc ||
1436 __sk_mem_schedule(sk, size, SK_MEM_RECV) ||
1437 skb_pfmemalloc(skb);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001438}
1439
1440static inline void sk_mem_reclaim(struct sock *sk)
1441{
1442 if (!sk_has_account(sk))
1443 return;
1444 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1445 __sk_mem_reclaim(sk);
1446}
1447
David S. Miller9993e7d2008-01-10 21:56:38 -08001448static inline void sk_mem_reclaim_partial(struct sock *sk)
1449{
1450 if (!sk_has_account(sk))
1451 return;
1452 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1453 __sk_mem_reclaim(sk);
1454}
1455
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001456static inline void sk_mem_charge(struct sock *sk, int size)
1457{
1458 if (!sk_has_account(sk))
1459 return;
1460 sk->sk_forward_alloc -= size;
1461}
1462
1463static inline void sk_mem_uncharge(struct sock *sk, int size)
1464{
1465 if (!sk_has_account(sk))
1466 return;
1467 sk->sk_forward_alloc += size;
1468}
1469
1470static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1471{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001472 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1473 sk->sk_wmem_queued -= skb->truesize;
1474 sk_mem_uncharge(sk, skb->truesize);
1475 __kfree_skb(skb);
Herbert Xud80d99d2005-09-01 17:48:23 -07001476}
1477
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478/* Used by processes to "lock" a socket state, so that
1479 * interrupts and bottom half handlers won't change it
1480 * from under us. It essentially blocks any incoming
1481 * packets, so that we won't get any new data or any
1482 * packets that change the state of the socket.
1483 *
1484 * While locked, BH processing will add new packets to
1485 * the backlog queue. This queue is processed by the
1486 * owner of the socket lock right before it is released.
1487 *
1488 * Since ~2.3.5 it is also exclusive sleep lock serializing
1489 * accesses from user process context.
1490 */
John Heffnerd2e91172007-09-12 10:44:19 +02001491#define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001492
Eric Dumazetc3f9b012014-03-10 09:50:11 -07001493static inline void sock_release_ownership(struct sock *sk)
1494{
1495 sk->sk_lock.owned = 0;
1496}
1497
Peter Zijlstraed075362006-12-06 20:35:24 -08001498/*
1499 * Macro so as to not evaluate some arguments when
1500 * lockdep is not enabled.
1501 *
1502 * Mark both the sk_lock and the sk_lock.slock as a
1503 * per-address-family lock class.
1504 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001505#define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
Peter Zijlstraed075362006-12-06 20:35:24 -08001506do { \
Ingo Molnare8f6fbf2008-11-12 01:38:36 +00001507 sk->sk_lock.owned = 0; \
Peter Zijlstraed075362006-12-06 20:35:24 -08001508 init_waitqueue_head(&sk->sk_lock.wq); \
1509 spin_lock_init(&(sk)->sk_lock.slock); \
1510 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1511 sizeof((sk)->sk_lock)); \
1512 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001513 (skey), (sname)); \
Peter Zijlstraed075362006-12-06 20:35:24 -08001514 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1515} while (0)
1516
Joe Perches69336bd2013-09-22 10:32:26 -07001517void lock_sock_nested(struct sock *sk, int subclass);
Peter Zijlstrafcc70d52006-11-08 22:44:35 -08001518
1519static inline void lock_sock(struct sock *sk)
1520{
1521 lock_sock_nested(sk, 0);
1522}
1523
Joe Perches69336bd2013-09-22 10:32:26 -07001524void release_sock(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001525
1526/* BH context may only use the following locking interface. */
1527#define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
Ingo Molnarc6366182006-07-03 00:25:13 -07001528#define bh_lock_sock_nested(__sk) \
1529 spin_lock_nested(&((__sk)->sk_lock.slock), \
1530 SINGLE_DEPTH_NESTING)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531#define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1532
Joe Perches69336bd2013-09-22 10:32:26 -07001533bool lock_sock_fast(struct sock *sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001534/**
1535 * unlock_sock_fast - complement of lock_sock_fast
1536 * @sk: socket
1537 * @slow: slow mode
1538 *
1539 * fast unlock socket for user context.
1540 * If slow mode is on, we call regular release_sock()
1541 */
1542static inline void unlock_sock_fast(struct sock *sk, bool slow)
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001543{
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001544 if (slow)
1545 release_sock(sk);
1546 else
1547 spin_unlock_bh(&sk->sk_lock.slock);
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001548}
1549
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001550
Joe Perches69336bd2013-09-22 10:32:26 -07001551struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
1552 struct proto *prot);
1553void sk_free(struct sock *sk);
1554void sk_release_kernel(struct sock *sk);
1555struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001556
Joe Perches69336bd2013-09-22 10:32:26 -07001557struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
1558 gfp_t priority);
Joe Perches69336bd2013-09-22 10:32:26 -07001559void sock_wfree(struct sk_buff *skb);
1560void skb_orphan_partial(struct sk_buff *skb);
1561void sock_rfree(struct sk_buff *skb);
1562void sock_edemux(struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001563
Joe Perches69336bd2013-09-22 10:32:26 -07001564int sock_setsockopt(struct socket *sock, int level, int op,
1565 char __user *optval, unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001566
Joe Perches69336bd2013-09-22 10:32:26 -07001567int sock_getsockopt(struct socket *sock, int level, int op,
1568 char __user *optval, int __user *optlen);
1569struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1570 int noblock, int *errcode);
1571struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1572 unsigned long data_len, int noblock,
1573 int *errcode, int max_page_order);
1574void *sock_kmalloc(struct sock *sk, int size, gfp_t priority);
1575void sock_kfree_s(struct sock *sk, void *mem, int size);
1576void sk_send_sigurg(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001577
1578/*
1579 * Functions to fill in entries in struct proto_ops when a protocol
1580 * does not implement a particular function.
1581 */
Joe Perches69336bd2013-09-22 10:32:26 -07001582int sock_no_bind(struct socket *, struct sockaddr *, int);
1583int sock_no_connect(struct socket *, struct sockaddr *, int, int);
1584int sock_no_socketpair(struct socket *, struct socket *);
1585int sock_no_accept(struct socket *, struct socket *, int);
1586int sock_no_getname(struct socket *, struct sockaddr *, int *, int);
1587unsigned int sock_no_poll(struct file *, struct socket *,
1588 struct poll_table_struct *);
1589int sock_no_ioctl(struct socket *, unsigned int, unsigned long);
1590int sock_no_listen(struct socket *, int);
1591int sock_no_shutdown(struct socket *, int);
1592int sock_no_getsockopt(struct socket *, int , int, char __user *, int __user *);
1593int sock_no_setsockopt(struct socket *, int, int, char __user *, unsigned int);
1594int sock_no_sendmsg(struct kiocb *, struct socket *, struct msghdr *, size_t);
1595int sock_no_recvmsg(struct kiocb *, struct socket *, struct msghdr *, size_t,
1596 int);
1597int sock_no_mmap(struct file *file, struct socket *sock,
1598 struct vm_area_struct *vma);
1599ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset,
1600 size_t size, int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001601
1602/*
1603 * Functions to fill in entries in struct proto_ops when a protocol
1604 * uses the inet style.
1605 */
Joe Perches69336bd2013-09-22 10:32:26 -07001606int sock_common_getsockopt(struct socket *sock, int level, int optname,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001607 char __user *optval, int __user *optlen);
Joe Perches69336bd2013-09-22 10:32:26 -07001608int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001609 struct msghdr *msg, size_t size, int flags);
Joe Perches69336bd2013-09-22 10:32:26 -07001610int sock_common_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07001611 char __user *optval, unsigned int optlen);
Joe Perches69336bd2013-09-22 10:32:26 -07001612int compat_sock_common_getsockopt(struct socket *sock, int level,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001613 int optname, char __user *optval, int __user *optlen);
Joe Perches69336bd2013-09-22 10:32:26 -07001614int compat_sock_common_setsockopt(struct socket *sock, int level,
David S. Millerb7058842009-09-30 16:12:20 -07001615 int optname, char __user *optval, unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001616
Joe Perches69336bd2013-09-22 10:32:26 -07001617void sk_common_release(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001618
1619/*
1620 * Default socket callbacks and setup code
1621 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001622
Linus Torvalds1da177e2005-04-16 15:20:36 -07001623/* Initialise core socket variables */
Joe Perches69336bd2013-09-22 10:32:26 -07001624void sock_init_data(struct socket *sock, struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001625
Linus Torvalds1da177e2005-04-16 15:20:36 -07001626/*
1627 * Socket reference counting postulates.
1628 *
1629 * * Each user of socket SHOULD hold a reference count.
1630 * * Each access point to socket (an hash table bucket, reference from a list,
1631 * running timer, skb in flight MUST hold a reference count.
1632 * * When reference count hits 0, it means it will never increase back.
1633 * * When reference count hits 0, it means that no references from
1634 * outside exist to this socket and current process on current CPU
1635 * is last user and may/should destroy this socket.
1636 * * sk_free is called from any context: process, BH, IRQ. When
1637 * it is called, socket has no references from outside -> sk_free
1638 * may release descendant resources allocated by the socket, but
1639 * to the time when it is called, socket is NOT referenced by any
1640 * hash tables, lists etc.
1641 * * Packets, delivered from outside (from network or from another process)
1642 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1643 * when they sit in queue. Otherwise, packets will leak to hole, when
1644 * socket is looked up by one cpu and unhasing is made by another CPU.
1645 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1646 * (leak to backlog). Packet socket does all the processing inside
1647 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1648 * use separate SMP lock, so that they are prone too.
1649 */
1650
1651/* Ungrab socket and destroy it, if it was the last reference. */
1652static inline void sock_put(struct sock *sk)
1653{
1654 if (atomic_dec_and_test(&sk->sk_refcnt))
1655 sk_free(sk);
1656}
Eric Dumazet05dbc7b2013-10-03 00:22:02 -07001657/* Generic version of sock_put(), dealing with all sockets
1658 * (TCP_TIMEWAIT, ESTABLISHED...)
1659 */
1660void sock_gen_put(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001661
Joe Perches69336bd2013-09-22 10:32:26 -07001662int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested);
Arnaldo Carvalho de Melo25995ff2005-12-27 02:42:22 -02001663
Krishna Kumare022f0b2009-10-19 23:46:20 +00001664static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1665{
1666 sk->sk_tx_queue_mapping = tx_queue;
1667}
1668
1669static inline void sk_tx_queue_clear(struct sock *sk)
1670{
1671 sk->sk_tx_queue_mapping = -1;
1672}
1673
1674static inline int sk_tx_queue_get(const struct sock *sk)
1675{
Tom Herbertb0f77d02010-07-14 20:50:29 -07001676 return sk ? sk->sk_tx_queue_mapping : -1;
Krishna Kumare022f0b2009-10-19 23:46:20 +00001677}
1678
David S. Miller972692e2008-06-17 22:41:38 -07001679static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1680{
Krishna Kumare022f0b2009-10-19 23:46:20 +00001681 sk_tx_queue_clear(sk);
David S. Miller972692e2008-06-17 22:41:38 -07001682 sk->sk_socket = sock;
1683}
1684
Eric Dumazetaa395142010-04-20 13:03:51 +00001685static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1686{
Eric Dumazeteaefd112011-02-18 03:26:36 +00001687 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1688 return &rcu_dereference_raw(sk->sk_wq)->wait;
Eric Dumazetaa395142010-04-20 13:03:51 +00001689}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001690/* Detach socket from process context.
1691 * Announce socket dead, detach it from wait queue and inode.
1692 * Note that parent inode held reference count on this struct sock,
1693 * we do not release it in this function, because protocol
1694 * probably wants some additional cleanups or even continuing
1695 * to work with this socket (TCP).
1696 */
1697static inline void sock_orphan(struct sock *sk)
1698{
1699 write_lock_bh(&sk->sk_callback_lock);
1700 sock_set_flag(sk, SOCK_DEAD);
David S. Miller972692e2008-06-17 22:41:38 -07001701 sk_set_socket(sk, NULL);
Eric Dumazet43815482010-04-29 11:01:49 +00001702 sk->sk_wq = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001703 write_unlock_bh(&sk->sk_callback_lock);
1704}
1705
1706static inline void sock_graft(struct sock *sk, struct socket *parent)
1707{
1708 write_lock_bh(&sk->sk_callback_lock);
Eric Dumazeteaefd112011-02-18 03:26:36 +00001709 sk->sk_wq = parent->wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710 parent->sk = sk;
David S. Miller972692e2008-06-17 22:41:38 -07001711 sk_set_socket(sk, parent);
Venkat Yekkirala4237c752006-07-24 23:32:50 -07001712 security_sock_graft(sk, parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001713 write_unlock_bh(&sk->sk_callback_lock);
1714}
1715
Joe Perches69336bd2013-09-22 10:32:26 -07001716kuid_t sock_i_uid(struct sock *sk);
1717unsigned long sock_i_ino(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001718
1719static inline struct dst_entry *
1720__sk_dst_get(struct sock *sk)
1721{
Michal Hockod8bf4ca2011-07-08 14:39:41 +02001722 return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
Eric Dumazetf68c2242010-04-22 16:06:59 -07001723 lockdep_is_held(&sk->sk_lock.slock));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724}
1725
1726static inline struct dst_entry *
1727sk_dst_get(struct sock *sk)
1728{
1729 struct dst_entry *dst;
1730
Eric Dumazetb6c67122010-04-08 23:03:29 +00001731 rcu_read_lock();
1732 dst = rcu_dereference(sk->sk_dst_cache);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733 if (dst)
1734 dst_hold(dst);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001735 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001736 return dst;
1737}
1738
Eric Dumazetb6c67122010-04-08 23:03:29 +00001739static inline void dst_negative_advice(struct sock *sk)
1740{
1741 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1742
1743 if (dst && dst->ops->negative_advice) {
1744 ndst = dst->ops->negative_advice(dst);
1745
1746 if (ndst != dst) {
1747 rcu_assign_pointer(sk->sk_dst_cache, ndst);
ZHAO Gang0a6957e2013-10-22 16:23:38 +08001748 sk_tx_queue_clear(sk);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001749 }
1750 }
1751}
1752
Linus Torvalds1da177e2005-04-16 15:20:36 -07001753static inline void
1754__sk_dst_set(struct sock *sk, struct dst_entry *dst)
1755{
1756 struct dst_entry *old_dst;
1757
Krishna Kumare022f0b2009-10-19 23:46:20 +00001758 sk_tx_queue_clear(sk);
Eric Dumazet0b53ff22010-04-26 20:40:43 +00001759 /*
1760 * This can be called while sk is owned by the caller only,
1761 * with no state that can be checked in a rcu_dereference_check() cond
1762 */
1763 old_dst = rcu_dereference_raw(sk->sk_dst_cache);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001764 rcu_assign_pointer(sk->sk_dst_cache, dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001765 dst_release(old_dst);
1766}
1767
1768static inline void
1769sk_dst_set(struct sock *sk, struct dst_entry *dst)
1770{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001771 spin_lock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001772 __sk_dst_set(sk, dst);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001773 spin_unlock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001774}
1775
1776static inline void
1777__sk_dst_reset(struct sock *sk)
1778{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001779 __sk_dst_set(sk, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001780}
1781
1782static inline void
1783sk_dst_reset(struct sock *sk)
1784{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001785 spin_lock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001786 __sk_dst_reset(sk);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001787 spin_unlock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001788}
1789
Joe Perches69336bd2013-09-22 10:32:26 -07001790struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001791
Joe Perches69336bd2013-09-22 10:32:26 -07001792struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001793
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001794static inline bool sk_can_gso(const struct sock *sk)
Herbert Xubcd76112006-06-30 13:36:35 -07001795{
1796 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1797}
1798
Joe Perches69336bd2013-09-22 10:32:26 -07001799void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
Arnaldo Carvalho de Melo6cbb0df2005-08-09 19:49:02 -07001800
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001801static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
Eric Dumazeta4654192010-05-16 00:36:33 -07001802{
1803 sk->sk_route_nocaps |= flags;
1804 sk->sk_route_caps &= ~flags;
1805}
1806
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001807static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1808 char __user *from, char *to,
Wei Yongjun912d3982011-04-06 18:40:12 +00001809 int copy, int offset)
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001810{
1811 if (skb->ip_summed == CHECKSUM_NONE) {
1812 int err = 0;
1813 __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err);
1814 if (err)
1815 return err;
Wei Yongjun912d3982011-04-06 18:40:12 +00001816 skb->csum = csum_block_add(skb->csum, csum, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001817 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1818 if (!access_ok(VERIFY_READ, from, copy) ||
1819 __copy_from_user_nocache(to, from, copy))
1820 return -EFAULT;
1821 } else if (copy_from_user(to, from, copy))
1822 return -EFAULT;
1823
1824 return 0;
1825}
1826
1827static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
1828 char __user *from, int copy)
1829{
Wei Yongjun912d3982011-04-06 18:40:12 +00001830 int err, offset = skb->len;
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001831
Wei Yongjun912d3982011-04-06 18:40:12 +00001832 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1833 copy, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001834 if (err)
Wei Yongjun912d3982011-04-06 18:40:12 +00001835 __skb_trim(skb, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001836
1837 return err;
1838}
1839
1840static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from,
1841 struct sk_buff *skb,
1842 struct page *page,
1843 int off, int copy)
1844{
1845 int err;
1846
Wei Yongjun912d3982011-04-06 18:40:12 +00001847 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1848 copy, skb->len);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001849 if (err)
1850 return err;
1851
1852 skb->len += copy;
1853 skb->data_len += copy;
1854 skb->truesize += copy;
1855 sk->sk_wmem_queued += copy;
1856 sk_mem_charge(sk, copy);
1857 return 0;
1858}
1859
Linus Torvalds1da177e2005-04-16 15:20:36 -07001860static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1861 struct sk_buff *skb, struct page *page,
1862 int off, int copy)
1863{
1864 if (skb->ip_summed == CHECKSUM_NONE) {
1865 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08001866 __wsum csum = csum_and_copy_from_user(from,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001867 page_address(page) + off,
1868 copy, 0, &err);
1869 if (err)
1870 return err;
1871 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1872 } else if (copy_from_user(page_address(page) + off, from, copy))
1873 return -EFAULT;
1874
1875 skb->len += copy;
1876 skb->data_len += copy;
1877 skb->truesize += copy;
1878 sk->sk_wmem_queued += copy;
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001879 sk_mem_charge(sk, copy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001880 return 0;
1881}
1882
Eric Dumazetc5640392009-06-16 10:12:03 +00001883/**
1884 * sk_wmem_alloc_get - returns write allocations
1885 * @sk: socket
1886 *
1887 * Returns sk_wmem_alloc minus initial offset of one
1888 */
1889static inline int sk_wmem_alloc_get(const struct sock *sk)
1890{
1891 return atomic_read(&sk->sk_wmem_alloc) - 1;
1892}
1893
1894/**
1895 * sk_rmem_alloc_get - returns read allocations
1896 * @sk: socket
1897 *
1898 * Returns sk_rmem_alloc
1899 */
1900static inline int sk_rmem_alloc_get(const struct sock *sk)
1901{
1902 return atomic_read(&sk->sk_rmem_alloc);
1903}
1904
1905/**
1906 * sk_has_allocations - check if allocations are outstanding
1907 * @sk: socket
1908 *
1909 * Returns true if socket has write or read allocations
1910 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001911static inline bool sk_has_allocations(const struct sock *sk)
Eric Dumazetc5640392009-06-16 10:12:03 +00001912{
1913 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1914}
1915
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001916/**
Eric Dumazet43815482010-04-29 11:01:49 +00001917 * wq_has_sleeper - check if there are any waiting processes
Randy Dunlapacfbe962010-05-24 23:54:18 -07001918 * @wq: struct socket_wq
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001919 *
Eric Dumazet43815482010-04-29 11:01:49 +00001920 * Returns true if socket_wq has waiting processes
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001921 *
Eric Dumazet43815482010-04-29 11:01:49 +00001922 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001923 * barrier call. They were added due to the race found within the tcp code.
1924 *
1925 * Consider following tcp code paths:
1926 *
1927 * CPU1 CPU2
1928 *
1929 * sys_select receive packet
1930 * ... ...
1931 * __add_wait_queue update tp->rcv_nxt
1932 * ... ...
1933 * tp->rcv_nxt check sock_def_readable
1934 * ... {
Eric Dumazet43815482010-04-29 11:01:49 +00001935 * schedule rcu_read_lock();
1936 * wq = rcu_dereference(sk->sk_wq);
1937 * if (wq && waitqueue_active(&wq->wait))
1938 * wake_up_interruptible(&wq->wait)
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001939 * ...
1940 * }
1941 *
1942 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1943 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1944 * could then endup calling schedule and sleep forever if there are no more
1945 * data on the socket.
Jiri Olsaad462762009-07-08 12:10:31 +00001946 *
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001947 */
Eric Dumazet43815482010-04-29 11:01:49 +00001948static inline bool wq_has_sleeper(struct socket_wq *wq)
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001949{
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001950 /* We need to be sure we are in sync with the
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001951 * add_wait_queue modifications to the wait queue.
1952 *
1953 * This memory barrier is paired in the sock_poll_wait.
1954 */
Eric Dumazet43815482010-04-29 11:01:49 +00001955 smp_mb();
1956 return wq && waitqueue_active(&wq->wait);
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001957}
1958
1959/**
1960 * sock_poll_wait - place memory barrier behind the poll_wait call.
1961 * @filp: file
1962 * @wait_address: socket wait queue
1963 * @p: poll_table
1964 *
Eric Dumazet43815482010-04-29 11:01:49 +00001965 * See the comments in the wq_has_sleeper function.
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001966 */
1967static inline void sock_poll_wait(struct file *filp,
1968 wait_queue_head_t *wait_address, poll_table *p)
1969{
Hans Verkuil626cf232012-03-23 15:02:27 -07001970 if (!poll_does_not_wait(p) && wait_address) {
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001971 poll_wait(filp, wait_address, p);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001972 /* We need to be sure we are in sync with the
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001973 * socket flags modification.
1974 *
Eric Dumazet43815482010-04-29 11:01:49 +00001975 * This memory barrier is paired in the wq_has_sleeper.
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001976 */
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001977 smp_mb();
1978 }
1979}
1980
Linus Torvalds1da177e2005-04-16 15:20:36 -07001981/*
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001982 * Queue a received datagram if it will fit. Stream and sequenced
Linus Torvalds1da177e2005-04-16 15:20:36 -07001983 * protocols can't normally use this as they need to fit buffers in
1984 * and play with them.
1985 *
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001986 * Inlined as it's very short and called for pretty much every
Linus Torvalds1da177e2005-04-16 15:20:36 -07001987 * packet ever received.
1988 */
1989
1990static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1991{
Herbert Xud55d87f2009-06-22 02:25:25 +00001992 skb_orphan(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001993 skb->sk = sk;
1994 skb->destructor = sock_wfree;
Eric Dumazet2b85a342009-06-11 02:55:43 -07001995 /*
1996 * We used to take a refcount on sk, but following operation
1997 * is enough to guarantee sk_free() wont free this sock until
1998 * all in-flight packets are completed
1999 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002000 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
2001}
2002
2003static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
2004{
Herbert Xud55d87f2009-06-22 02:25:25 +00002005 skb_orphan(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002006 skb->sk = sk;
2007 skb->destructor = sock_rfree;
2008 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08002009 sk_mem_charge(sk, skb->truesize);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002010}
2011
Joe Perches69336bd2013-09-22 10:32:26 -07002012void sk_reset_timer(struct sock *sk, struct timer_list *timer,
2013 unsigned long expires);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002014
Joe Perches69336bd2013-09-22 10:32:26 -07002015void sk_stop_timer(struct sock *sk, struct timer_list *timer);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002016
Joe Perches69336bd2013-09-22 10:32:26 -07002017int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002018
Joe Perches69336bd2013-09-22 10:32:26 -07002019int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002020
2021/*
2022 * Recover an error report and clear atomically
2023 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002024
Linus Torvalds1da177e2005-04-16 15:20:36 -07002025static inline int sock_error(struct sock *sk)
2026{
Benjamin LaHaisec1cbe4b2005-12-13 23:22:19 -08002027 int err;
2028 if (likely(!sk->sk_err))
2029 return 0;
2030 err = xchg(&sk->sk_err, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002031 return -err;
2032}
2033
2034static inline unsigned long sock_wspace(struct sock *sk)
2035{
2036 int amt = 0;
2037
2038 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
2039 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002040 if (amt < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002041 amt = 0;
2042 }
2043 return amt;
2044}
2045
2046static inline void sk_wake_async(struct sock *sk, int how, int band)
2047{
Eric Dumazetbcdce712009-10-06 17:28:29 -07002048 if (sock_flag(sk, SOCK_FASYNC))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002049 sock_wake_async(sk->sk_socket, how, band);
2050}
2051
Daniel Borkmanneea86af2013-06-19 12:51:20 +02002052/* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
2053 * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
2054 * Note: for send buffers, TCP works better if we can build two skbs at
2055 * minimum.
Eric Dumazet7a91b4342010-09-26 18:53:07 -07002056 */
Eric Dumazet9eb5bf82013-07-03 05:02:22 -07002057#define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
Daniel Borkmanneea86af2013-06-19 12:51:20 +02002058
2059#define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2)
2060#define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE
Linus Torvalds1da177e2005-04-16 15:20:36 -07002061
2062static inline void sk_stream_moderate_sndbuf(struct sock *sk)
2063{
2064 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
Eric Dumazet8df09ea2007-12-21 03:07:41 -08002065 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
Daniel Borkmanneea86af2013-06-19 12:51:20 +02002066 sk->sk_sndbuf = max_t(u32, sk->sk_sndbuf, SOCK_MIN_SNDBUF);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002067 }
2068}
2069
Pavel Emelyanovdf97c702007-11-29 21:22:33 +11002070struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002071
Eric Dumazet5640f762012-09-23 23:04:42 +00002072/**
2073 * sk_page_frag - return an appropriate page_frag
2074 * @sk: socket
2075 *
2076 * If socket allocation mode allows current thread to sleep, it means its
2077 * safe to use the per task page_frag instead of the per socket one.
2078 */
2079static inline struct page_frag *sk_page_frag(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002080{
Eric Dumazet5640f762012-09-23 23:04:42 +00002081 if (sk->sk_allocation & __GFP_WAIT)
2082 return &current->task_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002083
Eric Dumazet5640f762012-09-23 23:04:42 +00002084 return &sk->sk_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002085}
2086
Joe Perches69336bd2013-09-22 10:32:26 -07002087bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag);
Eric Dumazet5640f762012-09-23 23:04:42 +00002088
Linus Torvalds1da177e2005-04-16 15:20:36 -07002089/*
2090 * Default write policy as shown to user space via poll/select/SIGIO
2091 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002092static inline bool sock_writeable(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002093{
Eric Dumazet8df09ea2007-12-21 03:07:41 -08002094 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002095}
2096
Al Virodd0fc662005-10-07 07:46:04 +01002097static inline gfp_t gfp_any(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002098{
Andrew Morton99709372009-02-12 16:43:17 -08002099 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002100}
2101
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002102static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002103{
2104 return noblock ? 0 : sk->sk_rcvtimeo;
2105}
2106
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002107static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002108{
2109 return noblock ? 0 : sk->sk_sndtimeo;
2110}
2111
2112static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2113{
2114 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
2115}
2116
2117/* Alas, with timeout socket operations are not restartable.
2118 * Compare this to poll().
2119 */
2120static inline int sock_intr_errno(long timeo)
2121{
2122 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2123}
2124
Joe Perches69336bd2013-09-22 10:32:26 -07002125void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2126 struct sk_buff *skb);
2127void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2128 struct sk_buff *skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -07002129
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002130static inline void
Linus Torvalds1da177e2005-04-16 15:20:36 -07002131sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2132{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002133 ktime_t kt = skb->tstamp;
Patrick Ohly20d49472009-02-12 05:03:38 +00002134 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002135
Patrick Ohly20d49472009-02-12 05:03:38 +00002136 /*
2137 * generate control messages if
2138 * - receive time stamping in software requested (SOCK_RCVTSTAMP
2139 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
2140 * - software time stamp available and wanted
2141 * (SOCK_TIMESTAMPING_SOFTWARE)
2142 * - hardware time stamps available and wanted
2143 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
2144 * SOCK_TIMESTAMPING_RAW_HARDWARE)
2145 */
2146 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
2147 sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
2148 (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
2149 (hwtstamps->hwtstamp.tv64 &&
2150 sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
2151 (hwtstamps->syststamp.tv64 &&
2152 sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
Eric Dumazet92f37fd2007-03-25 22:14:49 -07002153 __sock_recv_timestamp(msg, sk, skb);
2154 else
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002155 sk->sk_stamp = kt;
Johannes Berg6e3e9392011-11-09 10:15:42 +01002156
2157 if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2158 __sock_recv_wifi_status(msg, sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002159}
2160
Joe Perches69336bd2013-09-22 10:32:26 -07002161void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2162 struct sk_buff *skb);
Eric Dumazet767dd032010-04-28 19:14:43 +00002163
2164static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2165 struct sk_buff *skb)
2166{
2167#define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2168 (1UL << SOCK_RCVTSTAMP) | \
Eric Dumazet767dd032010-04-28 19:14:43 +00002169 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002170 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
Eric Dumazet767dd032010-04-28 19:14:43 +00002171 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
2172
2173 if (sk->sk_flags & FLAGS_TS_OR_DROPS)
2174 __sock_recv_ts_and_drops(msg, sk, skb);
2175 else
2176 sk->sk_stamp = skb->tstamp;
2177}
Neil Horman3b885782009-10-12 13:26:31 -07002178
Linus Torvalds1da177e2005-04-16 15:20:36 -07002179/**
Patrick Ohly20d49472009-02-12 05:03:38 +00002180 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
Patrick Ohly20d49472009-02-12 05:03:38 +00002181 * @sk: socket sending this packet
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002182 * @tx_flags: filled with instructions for time stamping
Patrick Ohly20d49472009-02-12 05:03:38 +00002183 *
Daniel Borkmannbf84a012013-04-14 08:08:13 +00002184 * Currently only depends on SOCK_TIMESTAMPING* flags.
Patrick Ohly20d49472009-02-12 05:03:38 +00002185 */
Joe Perches69336bd2013-09-22 10:32:26 -07002186void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags);
Patrick Ohly20d49472009-02-12 05:03:38 +00002187
2188/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002189 * sk_eat_skb - Release a skb if it is no longer needed
Pavel Pisa4dc3b162005-05-01 08:59:25 -07002190 * @sk: socket to eat this skb from
2191 * @skb: socket buffer to eat
Randy Dunlapf4b8ea72006-06-22 16:00:11 -07002192 * @copied_early: flag indicating whether DMA operations copied this data early
Linus Torvalds1da177e2005-04-16 15:20:36 -07002193 *
2194 * This routine must be called with interrupts disabled or with the socket
2195 * locked so that the sk_buff queue operation is ok.
2196*/
Chris Leech624d1162006-05-23 18:01:28 -07002197#ifdef CONFIG_NET_DMA
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002198static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
Chris Leech624d1162006-05-23 18:01:28 -07002199{
2200 __skb_unlink(skb, &sk->sk_receive_queue);
2201 if (!copied_early)
2202 __kfree_skb(skb);
2203 else
2204 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
2205}
2206#else
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002207static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002208{
2209 __skb_unlink(skb, &sk->sk_receive_queue);
2210 __kfree_skb(skb);
2211}
Chris Leech624d1162006-05-23 18:01:28 -07002212#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002213
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002214static inline
2215struct net *sock_net(const struct sock *sk)
2216{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002217 return read_pnet(&sk->sk_net);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002218}
2219
2220static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07002221void sock_net_set(struct sock *sk, struct net *net)
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002222{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002223 write_pnet(&sk->sk_net, net);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002224}
2225
Denis V. Lunevedf02082008-02-29 11:18:32 -08002226/*
2227 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
Lucas De Marchi25985ed2011-03-30 22:57:33 -03002228 * They should not hold a reference to a namespace in order to allow
Denis V. Lunevedf02082008-02-29 11:18:32 -08002229 * to stop it.
2230 * Sockets after sk_change_net should be released using sk_release_kernel
2231 */
2232static inline void sk_change_net(struct sock *sk, struct net *net)
2233{
Gu Zheng5812521b2014-03-10 09:57:34 +08002234 struct net *current_net = sock_net(sk);
2235
2236 if (!net_eq(current_net, net)) {
2237 put_net(current_net);
2238 sock_net_set(sk, hold_net(net));
2239 }
Denis V. Lunevedf02082008-02-29 11:18:32 -08002240}
2241
KOVACS Krisztian23542612008-10-07 12:41:01 -07002242static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2243{
Vijay Subramanianefc27f82012-06-24 13:03:07 +00002244 if (skb->sk) {
KOVACS Krisztian23542612008-10-07 12:41:01 -07002245 struct sock *sk = skb->sk;
2246
2247 skb->destructor = NULL;
2248 skb->sk = NULL;
2249 return sk;
2250 }
2251 return NULL;
2252}
2253
Joe Perches69336bd2013-09-22 10:32:26 -07002254void sock_enable_timestamp(struct sock *sk, int flag);
2255int sock_get_timestamp(struct sock *, struct timeval __user *);
2256int sock_get_timestampns(struct sock *, struct timespec __user *);
2257int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len, int level,
2258 int type);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002259
Eric W. Biedermana3b299d2014-04-23 14:26:56 -07002260bool sk_ns_capable(const struct sock *sk,
2261 struct user_namespace *user_ns, int cap);
2262bool sk_capable(const struct sock *sk, int cap);
2263bool sk_net_capable(const struct sock *sk, int cap);
2264
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002265/*
2266 * Enable debug/info messages
Linus Torvalds1da177e2005-04-16 15:20:36 -07002267 */
Stephen Hemmingera2a316f2007-03-08 20:41:08 -08002268extern int net_msg_warn;
2269#define NETDEBUG(fmt, args...) \
2270 do { if (net_msg_warn) printk(fmt,##args); } while (0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002271
Stephen Hemmingera2a316f2007-03-08 20:41:08 -08002272#define LIMIT_NETDEBUG(fmt, args...) \
2273 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002274
Linus Torvalds1da177e2005-04-16 15:20:36 -07002275extern __u32 sysctl_wmem_max;
2276extern __u32 sysctl_rmem_max;
2277
David S. Miller6baf1f42005-09-05 18:14:11 -07002278extern int sysctl_optmem_max;
2279
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002280extern __u32 sysctl_wmem_default;
2281extern __u32 sysctl_rmem_default;
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002282
Linus Torvalds1da177e2005-04-16 15:20:36 -07002283#endif /* _SOCK_H */