blob: 5c3f7c3624aa00214572c47cd3b6dbb3777d93ee [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
Adam Buchbinder48fc7f72012-09-19 21:48:00 -0400246 * @sk_no_check: %SO_NO_CHECK setting, whether or not checkup packets
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700247 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
Eric Dumazeta4654192010-05-16 00:36:33 -0700248 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
Herbert Xubcd76112006-06-30 13:36:35 -0700249 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -0700250 * @sk_gso_max_size: Maximum GSO segment size to build
Ben Hutchings14853482012-07-30 16:11:42 +0000251 * @sk_gso_max_segs: Maximum number of GSO segments
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700252 * @sk_lingertime: %SO_LINGER l_linger setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700253 * @sk_backlog: always used with the per-socket spinlock held
254 * @sk_callback_lock: used with the callbacks in the end of this struct
255 * @sk_error_queue: rarely used
Wang Chen33c732c2007-11-13 20:30:01 -0800256 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
257 * IPV6_ADDRFORM for instance)
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700258 * @sk_err: last error
Wang Chen33c732c2007-11-13 20:30:01 -0800259 * @sk_err_soft: errors that don't cause failure but are the cause of a
260 * persistent failure not just 'timed out'
Eric Dumazetcb61cb92008-06-17 21:04:56 -0700261 * @sk_drops: raw/udp drops counter
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700262 * @sk_ack_backlog: current listen backlog
263 * @sk_max_ack_backlog: listen backlog set in listen()
264 * @sk_priority: %SO_PRIORITY setting
Randy Dunlap1a3bc362012-01-21 09:03:10 +0000265 * @sk_cgrp_prioidx: socket group's priority map index
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700266 * @sk_type: socket type (%SOCK_STREAM, etc)
267 * @sk_protocol: which protocol this socket belongs in this network family
Randy Dunlap53c3fa22010-08-09 13:41:07 +0000268 * @sk_peer_pid: &struct pid for this socket's peer
269 * @sk_peer_cred: %SO_PEERCRED setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700270 * @sk_rcvlowat: %SO_RCVLOWAT setting
271 * @sk_rcvtimeo: %SO_RCVTIMEO setting
272 * @sk_sndtimeo: %SO_SNDTIMEO setting
David S. Millerc58dc012010-04-27 15:05:31 -0700273 * @sk_rxhash: flow hash received from netif layer
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700274 * @sk_filter: socket filtering instructions
275 * @sk_protinfo: private area, net family specific, when not using slab
276 * @sk_timer: sock cleanup timer
277 * @sk_stamp: time stamp of last packet received
278 * @sk_socket: Identd and reporting IO signals
279 * @sk_user_data: RPC layer private data
Eric Dumazet5640f762012-09-23 23:04:42 +0000280 * @sk_frag: cached page frag
Randy Dunlapd3d4f0a2012-04-17 14:03:53 +0000281 * @sk_peek_off: current peek_offset value
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700282 * @sk_send_head: front of stuff to transmit
Martin Waitz67be2dd2005-05-01 08:59:26 -0700283 * @sk_security: used by security modules
Randy Dunlap31729362008-02-18 20:52:13 -0800284 * @sk_mark: generic packet mark
Randy Dunlap53c3fa22010-08-09 13:41:07 +0000285 * @sk_classid: this socket's cgroup classid
Glauber Costae1aab162011-12-11 21:47:03 +0000286 * @sk_cgrp: this socket's cgroup-specific proto data
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700287 * @sk_write_pending: a write to stream socket waits to start
288 * @sk_state_change: callback to indicate change in the state of the sock
289 * @sk_data_ready: callback to indicate there is data to be processed
290 * @sk_write_space: callback to indicate there is bf sending space available
291 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
292 * @sk_backlog_rcv: callback to process the backlog
293 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294 */
295struct sock {
296 /*
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700297 * Now struct inet_timewait_sock also uses sock_common, so please just
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298 * don't add nothing before this first member (__sk_common) --acme
299 */
300 struct sock_common __sk_common;
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000301#define sk_node __sk_common.skc_node
302#define sk_nulls_node __sk_common.skc_nulls_node
303#define sk_refcnt __sk_common.skc_refcnt
Krishna Kumare022f0b2009-10-19 23:46:20 +0000304#define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000305
Eric Dumazet68835ab2010-11-30 19:04:07 +0000306#define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
307#define sk_dontcopy_end __sk_common.skc_dontcopy_end
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000308#define sk_hash __sk_common.skc_hash
Eric Dumazet50805462013-10-02 04:29:50 -0700309#define sk_portpair __sk_common.skc_portpair
Eric Dumazet05dbc7b2013-10-03 00:22:02 -0700310#define sk_num __sk_common.skc_num
311#define sk_dport __sk_common.skc_dport
Eric Dumazet50805462013-10-02 04:29:50 -0700312#define sk_addrpair __sk_common.skc_addrpair
313#define sk_daddr __sk_common.skc_daddr
314#define sk_rcv_saddr __sk_common.skc_rcv_saddr
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315#define sk_family __sk_common.skc_family
316#define sk_state __sk_common.skc_state
317#define sk_reuse __sk_common.skc_reuse
Tom Herbert055dc212013-01-22 09:49:50 +0000318#define sk_reuseport __sk_common.skc_reuseport
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319#define sk_bound_dev_if __sk_common.skc_bound_dev_if
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320#define sk_bind_node __sk_common.skc_bind_node
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700321#define sk_prot __sk_common.skc_prot
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200322#define sk_net __sk_common.skc_net
Eric Dumazetefe42082013-10-03 15:42:29 -0700323#define sk_v6_daddr __sk_common.skc_v6_daddr
324#define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr
325
Eric Dumazetb178bb32010-11-16 05:56:04 +0000326 socket_lock_t sk_lock;
327 struct sk_buff_head sk_receive_queue;
328 /*
329 * The backlog queue is special, it is always used with
330 * the per-socket spinlock held and requires low latency
331 * access. Therefore we special case it's implementation.
332 * Note : rmem_alloc is in this structure to fill a hole
333 * on 64bit arches, not because its logically part of
334 * backlog.
335 */
336 struct {
337 atomic_t rmem_alloc;
338 int len;
339 struct sk_buff *head;
340 struct sk_buff *tail;
341 } sk_backlog;
342#define sk_rmem_alloc sk_backlog.rmem_alloc
343 int sk_forward_alloc;
344#ifdef CONFIG_RPS
345 __u32 sk_rxhash;
346#endif
Cong Wange0d10952013-08-01 11:10:25 +0800347#ifdef CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir06021292013-06-10 11:39:50 +0300348 unsigned int sk_napi_id;
Eliezer Tamirdafcc432013-06-14 16:33:57 +0300349 unsigned int sk_ll_usec;
Eliezer Tamir06021292013-06-10 11:39:50 +0300350#endif
Eric Dumazetb178bb32010-11-16 05:56:04 +0000351 atomic_t sk_drops;
352 int sk_rcvbuf;
353
354 struct sk_filter __rcu *sk_filter;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000355 struct socket_wq __rcu *sk_wq;
Eric Dumazetb178bb32010-11-16 05:56:04 +0000356
357#ifdef CONFIG_NET_DMA
358 struct sk_buff_head sk_async_wait_queue;
359#endif
360
361#ifdef CONFIG_XFRM
362 struct xfrm_policy *sk_policy[2];
363#endif
364 unsigned long sk_flags;
Eric Dumazetdeaa5852012-06-24 20:22:49 +0000365 struct dst_entry *sk_rx_dst;
Cong Wang0e36cbb2013-01-22 21:09:51 +0000366 struct dst_entry __rcu *sk_dst_cache;
Eric Dumazetb178bb32010-11-16 05:56:04 +0000367 spinlock_t sk_dst_lock;
368 atomic_t sk_wmem_alloc;
369 atomic_t sk_omem_alloc;
370 int sk_sndbuf;
371 struct sk_buff_head sk_write_queue;
Vegard Nossuma98b65a2009-02-26 14:46:57 +0100372 kmemcheck_bitfield_begin(flags);
Eric Dumazet5fdb9972009-10-08 22:50:25 +0000373 unsigned int sk_shutdown : 2,
374 sk_no_check : 2,
375 sk_userlocks : 4,
376 sk_protocol : 8,
377 sk_type : 16;
Vegard Nossuma98b65a2009-02-26 14:46:57 +0100378 kmemcheck_bitfield_end(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379 int sk_wmem_queued;
Al Viro7d877f32005-10-21 03:20:43 -0400380 gfp_t sk_allocation;
Eric Dumazet95bd09e2013-08-27 05:46:32 -0700381 u32 sk_pacing_rate; /* bytes per second */
Eric Dumazet62748f32013-09-24 08:20:52 -0700382 u32 sk_max_pacing_rate;
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000383 netdev_features_t sk_route_caps;
384 netdev_features_t sk_route_nocaps;
Herbert Xubcd76112006-06-30 13:36:35 -0700385 int sk_gso_type;
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -0700386 unsigned int sk_gso_max_size;
Ben Hutchings14853482012-07-30 16:11:42 +0000387 u16 sk_gso_max_segs;
David S. Miller9932cf92006-03-24 15:12:37 -0800388 int sk_rcvlowat;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389 unsigned long sk_lingertime;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390 struct sk_buff_head sk_error_queue;
Arnaldo Carvalho de Melo476e19c2005-05-05 13:35:15 -0700391 struct proto *sk_prot_creator;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392 rwlock_t sk_callback_lock;
393 int sk_err,
394 sk_err_soft;
395 unsigned short sk_ack_backlog;
396 unsigned short sk_max_ack_backlog;
397 __u32 sk_priority;
Daniel Borkmann86f85152013-12-29 17:27:11 +0100398#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
Neil Horman5bc14212011-11-22 05:10:51 +0000399 __u32 sk_cgrp_prioidx;
400#endif
Eric W. Biederman109f6e32010-06-13 03:30:14 +0000401 struct pid *sk_peer_pid;
402 const struct cred *sk_peer_cred;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403 long sk_rcvtimeo;
404 long sk_sndtimeo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 void *sk_protinfo;
406 struct timer_list sk_timer;
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700407 ktime_t sk_stamp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408 struct socket *sk_socket;
409 void *sk_user_data;
Eric Dumazet5640f762012-09-23 23:04:42 +0000410 struct page_frag sk_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411 struct sk_buff *sk_send_head;
Pavel Emelyanovef64a542012-02-21 07:31:34 +0000412 __s32 sk_peek_off;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700413 int sk_write_pending;
Alexey Dobriyand5f64232008-11-04 14:45:58 -0800414#ifdef CONFIG_SECURITY
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 void *sk_security;
Alexey Dobriyand5f64232008-11-04 14:45:58 -0800416#endif
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800417 __u32 sk_mark;
Herbert Xuf8451722010-05-24 00:12:34 -0700418 u32 sk_classid;
Glauber Costae1aab162011-12-11 21:47:03 +0000419 struct cg_proto *sk_cgrp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420 void (*sk_state_change)(struct sock *sk);
421 void (*sk_data_ready)(struct sock *sk, int bytes);
422 void (*sk_write_space)(struct sock *sk);
423 void (*sk_error_report)(struct sock *sk);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000424 int (*sk_backlog_rcv)(struct sock *sk,
425 struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426 void (*sk_destruct)(struct sock *sk);
427};
428
Pravin B Shelar559835ea2013-09-24 10:25:40 -0700429#define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
430
431#define rcu_dereference_sk_user_data(sk) rcu_dereference(__sk_user_data((sk)))
432#define rcu_assign_sk_user_data(sk, ptr) rcu_assign_pointer(__sk_user_data((sk)), ptr)
433
Pavel Emelyanov4a17fd52012-04-19 03:39:36 +0000434/*
435 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
436 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
437 * on a socket means that the socket will reuse everybody else's port
438 * without looking at the other's sk_reuse value.
439 */
440
441#define SK_NO_REUSE 0
442#define SK_CAN_REUSE 1
443#define SK_FORCE_REUSE 2
444
Pavel Emelyanovef64a542012-02-21 07:31:34 +0000445static inline int sk_peek_offset(struct sock *sk, int flags)
446{
447 if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0))
448 return sk->sk_peek_off;
449 else
450 return 0;
451}
452
453static inline void sk_peek_offset_bwd(struct sock *sk, int val)
454{
455 if (sk->sk_peek_off >= 0) {
456 if (sk->sk_peek_off >= val)
457 sk->sk_peek_off -= val;
458 else
459 sk->sk_peek_off = 0;
460 }
461}
462
463static inline void sk_peek_offset_fwd(struct sock *sk, int val)
464{
465 if (sk->sk_peek_off >= 0)
466 sk->sk_peek_off += val;
467}
468
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469/*
470 * Hashed lists helper routines
471 */
Li Zefanc4146642010-02-08 23:18:45 +0000472static inline struct sock *sk_entry(const struct hlist_node *node)
473{
474 return hlist_entry(node, struct sock, sk_node);
475}
476
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700477static inline struct sock *__sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478{
479 return hlist_entry(head->first, struct sock, sk_node);
480}
481
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700482static inline struct sock *sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483{
484 return hlist_empty(head) ? NULL : __sk_head(head);
485}
486
Eric Dumazet88ab1932008-11-16 19:39:21 -0800487static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
488{
489 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
490}
491
492static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
493{
494 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
495}
496
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700497static inline struct sock *sk_next(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498{
499 return sk->sk_node.next ?
500 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
501}
502
Eric Dumazet88ab1932008-11-16 19:39:21 -0800503static inline struct sock *sk_nulls_next(const struct sock *sk)
504{
505 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
506 hlist_nulls_entry(sk->sk_nulls_node.next,
507 struct sock, sk_nulls_node) :
508 NULL;
509}
510
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000511static inline bool sk_unhashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512{
513 return hlist_unhashed(&sk->sk_node);
514}
515
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000516static inline bool sk_hashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517{
Akinobu Mitada753be2006-04-28 15:21:23 -0700518 return !sk_unhashed(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519}
520
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000521static inline void sk_node_init(struct hlist_node *node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522{
523 node->pprev = NULL;
524}
525
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000526static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800527{
528 node->pprev = NULL;
529}
530
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000531static inline void __sk_del_node(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532{
533 __hlist_del(&sk->sk_node);
534}
535
stephen hemminger808f5112010-02-22 07:57:18 +0000536/* NB: equivalent to hlist_del_init_rcu */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000537static inline bool __sk_del_node_init(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538{
539 if (sk_hashed(sk)) {
540 __sk_del_node(sk);
541 sk_node_init(&sk->sk_node);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000542 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 }
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000544 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545}
546
547/* Grab socket reference count. This operation is valid only
548 when sk is ALREADY grabbed f.e. it is found in hash table
549 or a list and the lookup is made under lock preventing hash table
550 modifications.
551 */
552
553static inline void sock_hold(struct sock *sk)
554{
555 atomic_inc(&sk->sk_refcnt);
556}
557
558/* Ungrab socket in the context, which assumes that socket refcnt
559 cannot hit zero, f.e. it is true in context of any socketcall.
560 */
561static inline void __sock_put(struct sock *sk)
562{
563 atomic_dec(&sk->sk_refcnt);
564}
565
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000566static inline bool sk_del_node_init(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567{
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000568 bool rc = __sk_del_node_init(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569
570 if (rc) {
571 /* paranoid for a while -acme */
572 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
573 __sock_put(sk);
574 }
575 return rc;
576}
stephen hemminger808f5112010-02-22 07:57:18 +0000577#define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000579static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700580{
581 if (sk_hashed(sk)) {
Eric Dumazet88ab1932008-11-16 19:39:21 -0800582 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000583 return true;
Eric Dumazet271b72c2008-10-29 02:11:14 -0700584 }
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000585 return false;
Eric Dumazet271b72c2008-10-29 02:11:14 -0700586}
587
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000588static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700589{
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000590 bool rc = __sk_nulls_del_node_init_rcu(sk);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700591
592 if (rc) {
593 /* paranoid for a while -acme */
594 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
595 __sock_put(sk);
596 }
597 return rc;
598}
599
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000600static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601{
602 hlist_add_head(&sk->sk_node, list);
603}
604
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000605static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606{
607 sock_hold(sk);
608 __sk_add_node(sk, list);
609}
610
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000611static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
stephen hemminger808f5112010-02-22 07:57:18 +0000612{
613 sock_hold(sk);
614 hlist_add_head_rcu(&sk->sk_node, list);
615}
616
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000617static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700618{
Eric Dumazet88ab1932008-11-16 19:39:21 -0800619 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700620}
621
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000622static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700623{
624 sock_hold(sk);
Eric Dumazet88ab1932008-11-16 19:39:21 -0800625 __sk_nulls_add_node_rcu(sk, list);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700626}
627
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000628static inline void __sk_del_bind_node(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629{
630 __hlist_del(&sk->sk_bind_node);
631}
632
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000633static inline void sk_add_bind_node(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634 struct hlist_head *list)
635{
636 hlist_add_head(&sk->sk_bind_node, list);
637}
638
Sasha Levinb67bfe02013-02-27 17:06:00 -0800639#define sk_for_each(__sk, list) \
640 hlist_for_each_entry(__sk, list, sk_node)
641#define sk_for_each_rcu(__sk, list) \
642 hlist_for_each_entry_rcu(__sk, list, sk_node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800643#define sk_nulls_for_each(__sk, node, list) \
644 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
645#define sk_nulls_for_each_rcu(__sk, node, list) \
646 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
Sasha Levinb67bfe02013-02-27 17:06:00 -0800647#define sk_for_each_from(__sk) \
648 hlist_for_each_entry_from(__sk, sk_node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800649#define sk_nulls_for_each_from(__sk, node) \
650 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
651 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
Sasha Levinb67bfe02013-02-27 17:06:00 -0800652#define sk_for_each_safe(__sk, tmp, list) \
653 hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
654#define sk_for_each_bound(__sk, list) \
655 hlist_for_each_entry(__sk, list, sk_bind_node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700656
Eric W. Biedermanc336d142012-05-24 17:56:43 -0600657static inline struct user_namespace *sk_user_ns(struct sock *sk)
658{
659 /* Careful only use this in a context where these parameters
660 * can not change and must all be valid, such as recvmsg from
661 * userspace.
662 */
663 return sk->sk_socket->file->f_cred->user_ns;
664}
665
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666/* Sock flags */
667enum sock_flags {
668 SOCK_DEAD,
669 SOCK_DONE,
670 SOCK_URGINLINE,
671 SOCK_KEEPOPEN,
672 SOCK_LINGER,
673 SOCK_DESTROY,
674 SOCK_BROADCAST,
675 SOCK_TIMESTAMP,
676 SOCK_ZAPPED,
677 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
678 SOCK_DBG, /* %SO_DEBUG setting */
679 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700680 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
682 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
Mel Gorman7cb02402012-07-31 16:44:16 -0700683 SOCK_MEMALLOC, /* VM depends on this socket for swapping */
Patrick Ohly20d49472009-02-12 05:03:38 +0000684 SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */
685 SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
686 SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */
687 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
688 SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */
689 SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
690 SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
Eric Dumazetbcdce712009-10-06 17:28:29 -0700691 SOCK_FASYNC, /* fasync() active */
Neil Horman3b885782009-10-12 13:26:31 -0700692 SOCK_RXQ_OVFL,
Shirley Ma1cdebb42011-07-06 12:17:30 +0000693 SOCK_ZEROCOPY, /* buffers from userspace */
Johannes Berg6e3e9392011-11-09 10:15:42 +0100694 SOCK_WIFI_STATUS, /* push wifi status to userspace */
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000695 SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
696 * Will use last 4 bytes of packet sent from
697 * user-space instead.
698 */
Vincent Bernatd59577b2013-01-16 22:55:49 +0100699 SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */
Keller, Jacob E7d4c04f2013-03-28 11:19:25 +0000700 SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701};
702
Ralf Baechle53b924b2005-08-23 10:11:30 -0700703static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
704{
705 nsk->sk_flags = osk->sk_flags;
706}
707
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
709{
710 __set_bit(flag, &sk->sk_flags);
711}
712
713static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
714{
715 __clear_bit(flag, &sk->sk_flags);
716}
717
Eric Dumazet1b23a5d2012-05-16 05:57:07 +0000718static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700719{
720 return test_bit(flag, &sk->sk_flags);
721}
722
Mel Gormanc93bdd02012-07-31 16:44:19 -0700723#ifdef CONFIG_NET
724extern struct static_key memalloc_socks;
725static inline int sk_memalloc_socks(void)
726{
727 return static_key_false(&memalloc_socks);
728}
729#else
730
731static inline int sk_memalloc_socks(void)
732{
733 return 0;
734}
735
736#endif
737
Mel Gorman99a1dec2012-07-31 16:44:14 -0700738static inline gfp_t sk_gfp_atomic(struct sock *sk, gfp_t gfp_mask)
739{
Mel Gorman7cb02402012-07-31 16:44:16 -0700740 return GFP_ATOMIC | (sk->sk_allocation & __GFP_MEMALLOC);
Mel Gorman99a1dec2012-07-31 16:44:14 -0700741}
742
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743static inline void sk_acceptq_removed(struct sock *sk)
744{
745 sk->sk_ack_backlog--;
746}
747
748static inline void sk_acceptq_added(struct sock *sk)
749{
750 sk->sk_ack_backlog++;
751}
752
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000753static inline bool sk_acceptq_is_full(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754{
David S. Miller64a14652007-03-06 11:21:05 -0800755 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756}
757
758/*
759 * Compute minimal free write space needed to queue new packets.
760 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000761static inline int sk_stream_min_wspace(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762{
Eric Dumazet8df09ea2007-12-21 03:07:41 -0800763 return sk->sk_wmem_queued >> 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764}
765
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000766static inline int sk_stream_wspace(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767{
768 return sk->sk_sndbuf - sk->sk_wmem_queued;
769}
770
Joe Perches69336bd2013-09-22 10:32:26 -0700771void sk_stream_write_space(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700772
Zhu Yi8eae9392010-03-04 18:01:40 +0000773/* OOB backlog add */
Zhu Yia3a858f2010-03-04 18:01:47 +0000774static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800775{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000776 /* dont let skb dst not refcounted, we are going to leave rcu lock */
777 skb_dst_force(skb);
778
779 if (!sk->sk_backlog.tail)
780 sk->sk_backlog.head = skb;
781 else
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800782 sk->sk_backlog.tail->next = skb;
Eric Dumazet7fee2262010-05-11 23:19:48 +0000783
784 sk->sk_backlog.tail = skb;
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800785 skb->next = NULL;
786}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787
Eric Dumazetc3774112010-04-27 15:13:20 -0700788/*
789 * Take into account size of receive queue and backlog queue
Eric Dumazet0fd7bac2011-12-21 07:11:44 +0000790 * Do not take into account this skb truesize,
791 * to allow even a single big packet to come.
Eric Dumazetc3774112010-04-27 15:13:20 -0700792 */
Eric Dumazetf545a382012-04-22 23:34:26 +0000793static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb,
794 unsigned int limit)
Eric Dumazetc3774112010-04-27 15:13:20 -0700795{
796 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
797
Eric Dumazetf545a382012-04-22 23:34:26 +0000798 return qsize > limit;
Eric Dumazetc3774112010-04-27 15:13:20 -0700799}
800
Zhu Yi8eae9392010-03-04 18:01:40 +0000801/* The per-socket spinlock must be held here. */
Eric Dumazetf545a382012-04-22 23:34:26 +0000802static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
803 unsigned int limit)
Zhu Yi8eae9392010-03-04 18:01:40 +0000804{
Eric Dumazetf545a382012-04-22 23:34:26 +0000805 if (sk_rcvqueues_full(sk, skb, limit))
Zhu Yi8eae9392010-03-04 18:01:40 +0000806 return -ENOBUFS;
807
Zhu Yia3a858f2010-03-04 18:01:47 +0000808 __sk_add_backlog(sk, skb);
Zhu Yi8eae9392010-03-04 18:01:40 +0000809 sk->sk_backlog.len += skb->truesize;
810 return 0;
811}
812
Joe Perches69336bd2013-09-22 10:32:26 -0700813int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
Mel Gormanb4b9e352012-07-31 16:44:26 -0700814
Peter Zijlstrac57943a2008-10-07 14:18:42 -0700815static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
816{
Mel Gormanb4b9e352012-07-31 16:44:26 -0700817 if (sk_memalloc_socks() && skb_pfmemalloc(skb))
818 return __sk_backlog_rcv(sk, skb);
819
Peter Zijlstrac57943a2008-10-07 14:18:42 -0700820 return sk->sk_backlog_rcv(sk, skb);
821}
822
Tom Herbertfe477552013-12-22 18:54:31 +0800823static inline void sock_rps_record_flow_hash(__u32 hash)
David S. Millerc58dc012010-04-27 15:05:31 -0700824{
825#ifdef CONFIG_RPS
826 struct rps_sock_flow_table *sock_flow_table;
827
828 rcu_read_lock();
829 sock_flow_table = rcu_dereference(rps_sock_flow_table);
Tom Herbertfe477552013-12-22 18:54:31 +0800830 rps_record_sock_flow(sock_flow_table, hash);
David S. Millerc58dc012010-04-27 15:05:31 -0700831 rcu_read_unlock();
832#endif
833}
834
Tom Herbertfe477552013-12-22 18:54:31 +0800835static inline void sock_rps_reset_flow_hash(__u32 hash)
David S. Millerc58dc012010-04-27 15:05:31 -0700836{
837#ifdef CONFIG_RPS
838 struct rps_sock_flow_table *sock_flow_table;
839
840 rcu_read_lock();
841 sock_flow_table = rcu_dereference(rps_sock_flow_table);
Tom Herbertfe477552013-12-22 18:54:31 +0800842 rps_reset_sock_flow(sock_flow_table, hash);
David S. Millerc58dc012010-04-27 15:05:31 -0700843 rcu_read_unlock();
844#endif
845}
846
Tom Herbertfe477552013-12-22 18:54:31 +0800847static inline void sock_rps_record_flow(const struct sock *sk)
848{
Zhi Yong Wuc9d8ca02014-01-01 04:31:01 +0800849#ifdef CONFIG_RPS
Tom Herbertfe477552013-12-22 18:54:31 +0800850 sock_rps_record_flow_hash(sk->sk_rxhash);
Zhi Yong Wuc9d8ca02014-01-01 04:31:01 +0800851#endif
Tom Herbertfe477552013-12-22 18:54:31 +0800852}
853
854static inline void sock_rps_reset_flow(const struct sock *sk)
855{
Zhi Yong Wuc9d8ca02014-01-01 04:31:01 +0800856#ifdef CONFIG_RPS
Tom Herbertfe477552013-12-22 18:54:31 +0800857 sock_rps_reset_flow_hash(sk->sk_rxhash);
Zhi Yong Wuc9d8ca02014-01-01 04:31:01 +0800858#endif
Tom Herbertfe477552013-12-22 18:54:31 +0800859}
860
Tom Herbertbdeab992011-08-14 19:45:55 +0000861static inline void sock_rps_save_rxhash(struct sock *sk,
862 const struct sk_buff *skb)
David S. Millerc58dc012010-04-27 15:05:31 -0700863{
864#ifdef CONFIG_RPS
Tom Herbertbdeab992011-08-14 19:45:55 +0000865 if (unlikely(sk->sk_rxhash != skb->rxhash)) {
David S. Millerc58dc012010-04-27 15:05:31 -0700866 sock_rps_reset_flow(sk);
Tom Herbertbdeab992011-08-14 19:45:55 +0000867 sk->sk_rxhash = skb->rxhash;
David S. Millerc58dc012010-04-27 15:05:31 -0700868 }
869#endif
870}
871
Tom Herbertbdeab992011-08-14 19:45:55 +0000872static inline void sock_rps_reset_rxhash(struct sock *sk)
873{
874#ifdef CONFIG_RPS
875 sock_rps_reset_flow(sk);
876 sk->sk_rxhash = 0;
877#endif
878}
879
Stephen Hemmingercfcabdc2007-10-09 01:59:42 -0700880#define sk_wait_event(__sk, __timeo, __condition) \
881 ({ int __rc; \
882 release_sock(__sk); \
883 __rc = __condition; \
884 if (!__rc) { \
885 *(__timeo) = schedule_timeout(*(__timeo)); \
886 } \
887 lock_sock(__sk); \
888 __rc = __condition; \
889 __rc; \
890 })
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891
Joe Perches69336bd2013-09-22 10:32:26 -0700892int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
893int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
894void sk_stream_wait_close(struct sock *sk, long timeo_p);
895int sk_stream_error(struct sock *sk, int flags, int err);
896void sk_stream_kill_queues(struct sock *sk);
897void sk_set_memalloc(struct sock *sk);
898void sk_clear_memalloc(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899
Joe Perches69336bd2013-09-22 10:32:26 -0700900int sk_wait_data(struct sock *sk, long *timeo);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700901
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700902struct request_sock_ops;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800903struct timewait_sock_ops;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -0800904struct inet_hashinfo;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -0700905struct raw_hashinfo;
Paul Gortmakerde477252011-05-26 13:46:22 -0400906struct module;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700907
Eric Dumazetf77d6022013-05-09 10:28:16 +0000908/*
909 * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
910 * un-modified. Special care is taken when initializing object to zero.
911 */
912static inline void sk_prot_clear_nulls(struct sock *sk, int size)
913{
914 if (offsetof(struct sock, sk_node.next) != 0)
915 memset(sk, 0, offsetof(struct sock, sk_node.next));
916 memset(&sk->sk_node.pprev, 0,
917 size - offsetof(struct sock, sk_node.pprev));
918}
919
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920/* Networking protocol blocks we attach to sockets.
921 * socket layer -> transport layer interface
922 * transport -> network interface is defined by struct inet_proto
923 */
924struct proto {
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000925 void (*close)(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926 long timeout);
927 int (*connect)(struct sock *sk,
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000928 struct sockaddr *uaddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929 int addr_len);
930 int (*disconnect)(struct sock *sk, int flags);
931
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000932 struct sock * (*accept)(struct sock *sk, int flags, int *err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700933
934 int (*ioctl)(struct sock *sk, int cmd,
935 unsigned long arg);
936 int (*init)(struct sock *sk);
Brian Haley7d06b2e2008-06-14 17:04:49 -0700937 void (*destroy)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938 void (*shutdown)(struct sock *sk, int how);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000939 int (*setsockopt)(struct sock *sk, int level,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700941 unsigned int optlen);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000942 int (*getsockopt)(struct sock *sk, int level,
943 int optname, char __user *optval,
944 int __user *option);
Alexey Dobriyanaf01d532008-08-28 02:53:51 -0700945#ifdef CONFIG_COMPAT
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800946 int (*compat_setsockopt)(struct sock *sk,
947 int level,
948 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700949 unsigned int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800950 int (*compat_getsockopt)(struct sock *sk,
951 int level,
952 int optname, char __user *optval,
953 int __user *option);
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000954 int (*compat_ioctl)(struct sock *sk,
955 unsigned int cmd, unsigned long arg);
Alexey Dobriyanaf01d532008-08-28 02:53:51 -0700956#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
958 struct msghdr *msg, size_t len);
959 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
960 struct msghdr *msg,
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000961 size_t len, int noblock, int flags,
962 int *addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963 int (*sendpage)(struct sock *sk, struct page *page,
964 int offset, size_t size, int flags);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000965 int (*bind)(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966 struct sockaddr *uaddr, int addr_len);
967
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000968 int (*backlog_rcv) (struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969 struct sk_buff *skb);
970
Eric Dumazet46d3cea2012-07-11 05:50:31 +0000971 void (*release_cb)(struct sock *sk);
Eric Dumazet563d34d2012-07-23 09:48:52 +0200972 void (*mtu_reduced)(struct sock *sk);
Eric Dumazet46d3cea2012-07-11 05:50:31 +0000973
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974 /* Keeping track of sk's, looking them up, and port selection methods. */
975 void (*hash)(struct sock *sk);
976 void (*unhash)(struct sock *sk);
Eric Dumazet719f8352010-09-08 05:08:44 +0000977 void (*rehash)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700978 int (*get_port)(struct sock *sk, unsigned short snum);
Octavian Purdilafcbdf092010-12-16 14:26:56 -0800979 void (*clear_sk)(struct sock *sk, int size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980
Eric Dumazet286ab3d2007-11-05 23:38:39 -0800981 /* Keeping track of sockets in use */
Eric Dumazet65f76512008-01-03 20:46:48 -0800982#ifdef CONFIG_PROC_FS
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -0700983 unsigned int inuse_idx;
Eric Dumazet65f76512008-01-03 20:46:48 -0800984#endif
Arnaldo Carvalho de Meloebb53d72007-11-21 22:08:50 +0800985
Eric Dumazetc9bee3b72013-07-22 20:27:07 -0700986 bool (*stream_memory_free)(const struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987 /* Memory pressure */
Pavel Emelyanov5c52ba12008-07-16 20:28:10 -0700988 void (*enter_memory_pressure)(struct sock *sk);
Eric Dumazet8d987e52010-11-09 23:24:26 +0000989 atomic_long_t *memory_allocated; /* Current allocated memory. */
Eric Dumazet17483762008-11-25 21:16:35 -0800990 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700991 /*
992 * Pressure flag: try to collapse.
993 * Technical note: it is used by multiple contexts non atomically.
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800994 * All the __sk_mem_schedule() is of this nature: accounting
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995 * is strict, actions are advisory and have some latency.
996 */
997 int *memory_pressure;
Eric Dumazet8d987e52010-11-09 23:24:26 +0000998 long *sysctl_mem;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999 int *sysctl_wmem;
1000 int *sysctl_rmem;
1001 int max_header;
Changli Gao7ba42912010-07-10 20:41:55 +00001002 bool no_autobind;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003
Eric Dumazet271b72c2008-10-29 02:11:14 -07001004 struct kmem_cache *slab;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005 unsigned int obj_size;
Eric Dumazet271b72c2008-10-29 02:11:14 -07001006 int slab_flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007
Eric Dumazetdd24c002008-11-25 21:17:14 -08001008 struct percpu_counter *orphan_count;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -07001009
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -07001010 struct request_sock_ops *rsk_prot;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -08001011 struct timewait_sock_ops *twsk_prot;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07001012
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -07001013 union {
1014 struct inet_hashinfo *hashinfo;
Eric Dumazet645ca702008-10-29 01:41:45 -07001015 struct udp_table *udp_table;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -07001016 struct raw_hashinfo *raw_hash;
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -07001017 } h;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -08001018
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019 struct module *owner;
1020
1021 char name[32];
1022
1023 struct list_head node;
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -07001024#ifdef SOCK_REFCNT_DEBUG
1025 atomic_t socks;
1026#endif
Andrew Mortonc255a452012-07-31 16:43:02 -07001027#ifdef CONFIG_MEMCG_KMEM
Glauber Costae1aab162011-12-11 21:47:03 +00001028 /*
1029 * cgroup specific init/deinit functions. Called once for all
1030 * protocols that implement it, from cgroups populate function.
1031 * This function has to setup any files the protocol want to
1032 * appear in the kmem cgroup filesystem.
1033 */
Glauber Costa1d62e432012-04-09 19:36:33 -03001034 int (*init_cgroup)(struct mem_cgroup *memcg,
Glauber Costae1aab162011-12-11 21:47:03 +00001035 struct cgroup_subsys *ss);
Glauber Costa1d62e432012-04-09 19:36:33 -03001036 void (*destroy_cgroup)(struct mem_cgroup *memcg);
Glauber Costae1aab162011-12-11 21:47:03 +00001037 struct cg_proto *(*proto_cgroup)(struct mem_cgroup *memcg);
1038#endif
1039};
1040
Glauber Costa3f134612012-05-29 15:07:11 -07001041/*
1042 * Bits in struct cg_proto.flags
1043 */
1044enum cg_proto_flags {
1045 /* Currently active and new sockets should be assigned to cgroups */
1046 MEMCG_SOCK_ACTIVE,
1047 /* It was ever activated; we must disarm static keys on destruction */
1048 MEMCG_SOCK_ACTIVATED,
1049};
1050
Glauber Costae1aab162011-12-11 21:47:03 +00001051struct cg_proto {
Eric W. Biederman2e685ca2013-10-19 16:26:19 -07001052 struct res_counter memory_allocated; /* Current allocated memory. */
1053 struct percpu_counter sockets_allocated; /* Current number of sockets. */
1054 int memory_pressure;
1055 long sysctl_mem[3];
Glauber Costa3f134612012-05-29 15:07:11 -07001056 unsigned long flags;
Glauber Costae1aab162011-12-11 21:47:03 +00001057 /*
1058 * memcg field is used to find which memcg we belong directly
1059 * Each memcg struct can hold more than one cg_proto, so container_of
1060 * won't really cut.
1061 *
1062 * The elegant solution would be having an inverse function to
1063 * proto_cgroup in struct proto, but that means polluting the structure
1064 * for everybody, instead of just for memcg users.
1065 */
1066 struct mem_cgroup *memcg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067};
1068
Joe Perches69336bd2013-09-22 10:32:26 -07001069int proto_register(struct proto *prot, int alloc_slab);
1070void proto_unregister(struct proto *prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071
Glauber Costa3f134612012-05-29 15:07:11 -07001072static inline bool memcg_proto_active(struct cg_proto *cg_proto)
1073{
1074 return test_bit(MEMCG_SOCK_ACTIVE, &cg_proto->flags);
1075}
1076
1077static inline bool memcg_proto_activated(struct cg_proto *cg_proto)
1078{
1079 return test_bit(MEMCG_SOCK_ACTIVATED, &cg_proto->flags);
1080}
1081
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -07001082#ifdef SOCK_REFCNT_DEBUG
1083static inline void sk_refcnt_debug_inc(struct sock *sk)
1084{
1085 atomic_inc(&sk->sk_prot->socks);
1086}
1087
1088static inline void sk_refcnt_debug_dec(struct sock *sk)
1089{
1090 atomic_dec(&sk->sk_prot->socks);
1091 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
1092 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
1093}
1094
Ying Xuedec34fb2013-02-15 22:28:25 +00001095static inline void sk_refcnt_debug_release(const struct sock *sk)
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -07001096{
1097 if (atomic_read(&sk->sk_refcnt) != 1)
1098 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
1099 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
1100}
1101#else /* SOCK_REFCNT_DEBUG */
1102#define sk_refcnt_debug_inc(sk) do { } while (0)
1103#define sk_refcnt_debug_dec(sk) do { } while (0)
1104#define sk_refcnt_debug_release(sk) do { } while (0)
1105#endif /* SOCK_REFCNT_DEBUG */
1106
Andrew Mortonc255a452012-07-31 16:43:02 -07001107#if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_NET)
Ingo Molnarc5905af2012-02-24 08:31:31 +01001108extern struct static_key memcg_socket_limit_enabled;
Glauber Costae1aab162011-12-11 21:47:03 +00001109static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1110 struct cg_proto *cg_proto)
1111{
1112 return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg));
1113}
Ingo Molnarc5905af2012-02-24 08:31:31 +01001114#define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
Glauber Costae1aab162011-12-11 21:47:03 +00001115#else
1116#define mem_cgroup_sockets_enabled 0
1117static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1118 struct cg_proto *cg_proto)
1119{
1120 return NULL;
1121}
1122#endif
1123
Eric Dumazetc9bee3b72013-07-22 20:27:07 -07001124static inline bool sk_stream_memory_free(const struct sock *sk)
1125{
1126 if (sk->sk_wmem_queued >= sk->sk_sndbuf)
1127 return false;
1128
1129 return sk->sk_prot->stream_memory_free ?
1130 sk->sk_prot->stream_memory_free(sk) : true;
1131}
1132
Eric Dumazet64dc6132013-07-22 20:26:31 -07001133static inline bool sk_stream_is_writeable(const struct sock *sk)
1134{
Eric Dumazetc9bee3b72013-07-22 20:27:07 -07001135 return sk_stream_wspace(sk) >= sk_stream_min_wspace(sk) &&
1136 sk_stream_memory_free(sk);
Eric Dumazet64dc6132013-07-22 20:26:31 -07001137}
Glauber Costae1aab162011-12-11 21:47:03 +00001138
Eric Dumazetc9bee3b72013-07-22 20:27:07 -07001139
Glauber Costa180d8cd2011-12-11 21:47:02 +00001140static inline bool sk_has_memory_pressure(const struct sock *sk)
1141{
1142 return sk->sk_prot->memory_pressure != NULL;
1143}
1144
1145static inline bool sk_under_memory_pressure(const struct sock *sk)
1146{
1147 if (!sk->sk_prot->memory_pressure)
1148 return false;
Glauber Costae1aab162011-12-11 21:47:03 +00001149
1150 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
Eric W. Biederman2e685ca2013-10-19 16:26:19 -07001151 return !!sk->sk_cgrp->memory_pressure;
Glauber Costae1aab162011-12-11 21:47:03 +00001152
Christoph Paasch35b87f62013-10-23 12:49:21 -07001153 return !!*sk->sk_prot->memory_pressure;
Glauber Costa180d8cd2011-12-11 21:47:02 +00001154}
1155
1156static inline void sk_leave_memory_pressure(struct sock *sk)
1157{
1158 int *memory_pressure = sk->sk_prot->memory_pressure;
1159
Glauber Costae1aab162011-12-11 21:47:03 +00001160 if (!memory_pressure)
1161 return;
1162
1163 if (*memory_pressure)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001164 *memory_pressure = 0;
Glauber Costae1aab162011-12-11 21:47:03 +00001165
1166 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1167 struct cg_proto *cg_proto = sk->sk_cgrp;
1168 struct proto *prot = sk->sk_prot;
1169
1170 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
Eric W. Biederman7f2cbdc2013-12-04 20:12:04 -08001171 cg_proto->memory_pressure = 0;
Glauber Costae1aab162011-12-11 21:47:03 +00001172 }
1173
Glauber Costa180d8cd2011-12-11 21:47:02 +00001174}
1175
1176static inline void sk_enter_memory_pressure(struct sock *sk)
1177{
Glauber Costae1aab162011-12-11 21:47:03 +00001178 if (!sk->sk_prot->enter_memory_pressure)
1179 return;
1180
1181 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1182 struct cg_proto *cg_proto = sk->sk_cgrp;
1183 struct proto *prot = sk->sk_prot;
1184
1185 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
Eric W. Biederman7f2cbdc2013-12-04 20:12:04 -08001186 cg_proto->memory_pressure = 1;
Glauber Costae1aab162011-12-11 21:47:03 +00001187 }
1188
1189 sk->sk_prot->enter_memory_pressure(sk);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001190}
1191
1192static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1193{
1194 long *prot = sk->sk_prot->sysctl_mem;
Glauber Costae1aab162011-12-11 21:47:03 +00001195 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1196 prot = sk->sk_cgrp->sysctl_mem;
Glauber Costa180d8cd2011-12-11 21:47:02 +00001197 return prot[index];
1198}
1199
Glauber Costae1aab162011-12-11 21:47:03 +00001200static inline void memcg_memory_allocated_add(struct cg_proto *prot,
1201 unsigned long amt,
1202 int *parent_status)
1203{
1204 struct res_counter *fail;
1205 int ret;
1206
Eric W. Biederman2e685ca2013-10-19 16:26:19 -07001207 ret = res_counter_charge_nofail(&prot->memory_allocated,
Glauber Costa0e90b312012-01-20 04:57:16 +00001208 amt << PAGE_SHIFT, &fail);
Glauber Costae1aab162011-12-11 21:47:03 +00001209 if (ret < 0)
1210 *parent_status = OVER_LIMIT;
1211}
1212
1213static inline void memcg_memory_allocated_sub(struct cg_proto *prot,
1214 unsigned long amt)
1215{
Eric W. Biederman2e685ca2013-10-19 16:26:19 -07001216 res_counter_uncharge(&prot->memory_allocated, amt << PAGE_SHIFT);
Glauber Costae1aab162011-12-11 21:47:03 +00001217}
1218
1219static inline u64 memcg_memory_allocated_read(struct cg_proto *prot)
1220{
1221 u64 ret;
Eric W. Biederman2e685ca2013-10-19 16:26:19 -07001222 ret = res_counter_read_u64(&prot->memory_allocated, RES_USAGE);
Glauber Costae1aab162011-12-11 21:47:03 +00001223 return ret >> PAGE_SHIFT;
1224}
1225
Glauber Costa180d8cd2011-12-11 21:47:02 +00001226static inline long
1227sk_memory_allocated(const struct sock *sk)
1228{
1229 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001230 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1231 return memcg_memory_allocated_read(sk->sk_cgrp);
1232
Glauber Costa180d8cd2011-12-11 21:47:02 +00001233 return atomic_long_read(prot->memory_allocated);
1234}
1235
1236static inline long
Glauber Costae1aab162011-12-11 21:47:03 +00001237sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001238{
1239 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001240
1241 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1242 memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status);
1243 /* update the root cgroup regardless */
1244 atomic_long_add_return(amt, prot->memory_allocated);
1245 return memcg_memory_allocated_read(sk->sk_cgrp);
1246 }
1247
Glauber Costa180d8cd2011-12-11 21:47:02 +00001248 return atomic_long_add_return(amt, prot->memory_allocated);
1249}
1250
1251static inline void
Glauber Costa0e90b312012-01-20 04:57:16 +00001252sk_memory_allocated_sub(struct sock *sk, int amt)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001253{
1254 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001255
Glauber Costa0e90b312012-01-20 04:57:16 +00001256 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
Glauber Costae1aab162011-12-11 21:47:03 +00001257 memcg_memory_allocated_sub(sk->sk_cgrp, amt);
1258
Glauber Costa180d8cd2011-12-11 21:47:02 +00001259 atomic_long_sub(amt, prot->memory_allocated);
1260}
1261
1262static inline void sk_sockets_allocated_dec(struct sock *sk)
1263{
1264 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001265
1266 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1267 struct cg_proto *cg_proto = sk->sk_cgrp;
1268
1269 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
Eric W. Biederman2e685ca2013-10-19 16:26:19 -07001270 percpu_counter_dec(&cg_proto->sockets_allocated);
Glauber Costae1aab162011-12-11 21:47:03 +00001271 }
1272
Glauber Costa180d8cd2011-12-11 21:47:02 +00001273 percpu_counter_dec(prot->sockets_allocated);
1274}
1275
1276static inline void sk_sockets_allocated_inc(struct sock *sk)
1277{
1278 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001279
1280 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1281 struct cg_proto *cg_proto = sk->sk_cgrp;
1282
1283 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
Eric W. Biederman2e685ca2013-10-19 16:26:19 -07001284 percpu_counter_inc(&cg_proto->sockets_allocated);
Glauber Costae1aab162011-12-11 21:47:03 +00001285 }
1286
Glauber Costa180d8cd2011-12-11 21:47:02 +00001287 percpu_counter_inc(prot->sockets_allocated);
1288}
1289
1290static inline int
1291sk_sockets_allocated_read_positive(struct sock *sk)
1292{
1293 struct proto *prot = sk->sk_prot;
1294
Glauber Costae1aab162011-12-11 21:47:03 +00001295 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
Eric W. Biederman2e685ca2013-10-19 16:26:19 -07001296 return percpu_counter_read_positive(&sk->sk_cgrp->sockets_allocated);
Glauber Costae1aab162011-12-11 21:47:03 +00001297
Eric Dumazet518fbf92012-04-28 23:21:56 +00001298 return percpu_counter_read_positive(prot->sockets_allocated);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001299}
1300
1301static inline int
1302proto_sockets_allocated_sum_positive(struct proto *prot)
1303{
1304 return percpu_counter_sum_positive(prot->sockets_allocated);
1305}
1306
1307static inline long
1308proto_memory_allocated(struct proto *prot)
1309{
1310 return atomic_long_read(prot->memory_allocated);
1311}
1312
1313static inline bool
1314proto_memory_pressure(struct proto *prot)
1315{
1316 if (!prot->memory_pressure)
1317 return false;
1318 return !!*prot->memory_pressure;
1319}
1320
Eric Dumazet65f76512008-01-03 20:46:48 -08001321
1322#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323/* Called with local bh disabled */
Joe Perches69336bd2013-09-22 10:32:26 -07001324void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1325int sock_prot_inuse_get(struct net *net, struct proto *proto);
Eric Dumazet65f76512008-01-03 20:46:48 -08001326#else
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001327static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -07001328 int inc)
Eric Dumazet65f76512008-01-03 20:46:48 -08001329{
1330}
Eric Dumazet65f76512008-01-03 20:46:48 -08001331#endif
1332
Linus Torvalds1da177e2005-04-16 15:20:36 -07001333
Arnaldo Carvalho de Melo614c6cb2005-08-09 19:47:37 -07001334/* With per-bucket locks this operation is not-atomic, so that
1335 * this version is not worse.
1336 */
1337static inline void __sk_prot_rehash(struct sock *sk)
1338{
1339 sk->sk_prot->unhash(sk);
1340 sk->sk_prot->hash(sk);
1341}
1342
Octavian Purdilafcbdf092010-12-16 14:26:56 -08001343void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
1344
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345/* About 10 seconds */
1346#define SOCK_DESTROY_TIME (10*HZ)
1347
1348/* Sockets 0-1023 can't be bound to unless you are superuser */
1349#define PROT_SOCK 1024
1350
1351#define SHUTDOWN_MASK 3
1352#define RCV_SHUTDOWN 1
1353#define SEND_SHUTDOWN 2
1354
1355#define SOCK_SNDBUF_LOCK 1
1356#define SOCK_RCVBUF_LOCK 2
1357#define SOCK_BINDADDR_LOCK 4
1358#define SOCK_BINDPORT_LOCK 8
1359
1360/* sock_iocb: used to kick off async processing of socket ios */
1361struct sock_iocb {
1362 struct list_head list;
1363
1364 int flags;
1365 int size;
1366 struct socket *sock;
1367 struct sock *sk;
1368 struct scm_cookie *scm;
1369 struct msghdr *msg, async_msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370 struct kiocb *kiocb;
1371};
1372
1373static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
1374{
1375 return (struct sock_iocb *)iocb->private;
1376}
1377
1378static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
1379{
1380 return si->kiocb;
1381}
1382
1383struct socket_alloc {
1384 struct socket socket;
1385 struct inode vfs_inode;
1386};
1387
1388static inline struct socket *SOCKET_I(struct inode *inode)
1389{
1390 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1391}
1392
1393static inline struct inode *SOCK_INODE(struct socket *socket)
1394{
1395 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1396}
1397
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001398/*
1399 * Functions for memory accounting
1400 */
Joe Perches69336bd2013-09-22 10:32:26 -07001401int __sk_mem_schedule(struct sock *sk, int size, int kind);
1402void __sk_mem_reclaim(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001403
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001404#define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1405#define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1406#define SK_MEM_SEND 0
1407#define SK_MEM_RECV 1
Linus Torvalds1da177e2005-04-16 15:20:36 -07001408
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001409static inline int sk_mem_pages(int amt)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001410{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001411 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412}
1413
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001414static inline bool sk_has_account(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001415{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001416 /* return true if protocol supports memory accounting */
1417 return !!sk->sk_prot->memory_allocated;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418}
1419
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001420static inline bool sk_wmem_schedule(struct sock *sk, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001422 if (!sk_has_account(sk))
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001423 return true;
Herbert Xud80d99d2005-09-01 17:48:23 -07001424 return size <= sk->sk_forward_alloc ||
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001425 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1426}
1427
Mel Gormanc76562b2012-07-31 16:44:41 -07001428static inline bool
Chuck Lever35c448a2012-09-17 14:09:11 -07001429sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size)
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001430{
1431 if (!sk_has_account(sk))
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001432 return true;
Mel Gormanc76562b2012-07-31 16:44:41 -07001433 return size<= sk->sk_forward_alloc ||
1434 __sk_mem_schedule(sk, size, SK_MEM_RECV) ||
1435 skb_pfmemalloc(skb);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001436}
1437
1438static inline void sk_mem_reclaim(struct sock *sk)
1439{
1440 if (!sk_has_account(sk))
1441 return;
1442 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1443 __sk_mem_reclaim(sk);
1444}
1445
David S. Miller9993e7d2008-01-10 21:56:38 -08001446static inline void sk_mem_reclaim_partial(struct sock *sk)
1447{
1448 if (!sk_has_account(sk))
1449 return;
1450 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1451 __sk_mem_reclaim(sk);
1452}
1453
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001454static inline void sk_mem_charge(struct sock *sk, int size)
1455{
1456 if (!sk_has_account(sk))
1457 return;
1458 sk->sk_forward_alloc -= size;
1459}
1460
1461static inline void sk_mem_uncharge(struct sock *sk, int size)
1462{
1463 if (!sk_has_account(sk))
1464 return;
1465 sk->sk_forward_alloc += size;
1466}
1467
1468static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1469{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001470 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1471 sk->sk_wmem_queued -= skb->truesize;
1472 sk_mem_uncharge(sk, skb->truesize);
1473 __kfree_skb(skb);
Herbert Xud80d99d2005-09-01 17:48:23 -07001474}
1475
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476/* Used by processes to "lock" a socket state, so that
1477 * interrupts and bottom half handlers won't change it
1478 * from under us. It essentially blocks any incoming
1479 * packets, so that we won't get any new data or any
1480 * packets that change the state of the socket.
1481 *
1482 * While locked, BH processing will add new packets to
1483 * the backlog queue. This queue is processed by the
1484 * owner of the socket lock right before it is released.
1485 *
1486 * Since ~2.3.5 it is also exclusive sleep lock serializing
1487 * accesses from user process context.
1488 */
John Heffnerd2e91172007-09-12 10:44:19 +02001489#define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001490
Peter Zijlstraed075362006-12-06 20:35:24 -08001491/*
1492 * Macro so as to not evaluate some arguments when
1493 * lockdep is not enabled.
1494 *
1495 * Mark both the sk_lock and the sk_lock.slock as a
1496 * per-address-family lock class.
1497 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001498#define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
Peter Zijlstraed075362006-12-06 20:35:24 -08001499do { \
Ingo Molnare8f6fbf2008-11-12 01:38:36 +00001500 sk->sk_lock.owned = 0; \
Peter Zijlstraed075362006-12-06 20:35:24 -08001501 init_waitqueue_head(&sk->sk_lock.wq); \
1502 spin_lock_init(&(sk)->sk_lock.slock); \
1503 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1504 sizeof((sk)->sk_lock)); \
1505 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001506 (skey), (sname)); \
Peter Zijlstraed075362006-12-06 20:35:24 -08001507 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1508} while (0)
1509
Joe Perches69336bd2013-09-22 10:32:26 -07001510void lock_sock_nested(struct sock *sk, int subclass);
Peter Zijlstrafcc70d52006-11-08 22:44:35 -08001511
1512static inline void lock_sock(struct sock *sk)
1513{
1514 lock_sock_nested(sk, 0);
1515}
1516
Joe Perches69336bd2013-09-22 10:32:26 -07001517void release_sock(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001518
1519/* BH context may only use the following locking interface. */
1520#define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
Ingo Molnarc6366182006-07-03 00:25:13 -07001521#define bh_lock_sock_nested(__sk) \
1522 spin_lock_nested(&((__sk)->sk_lock.slock), \
1523 SINGLE_DEPTH_NESTING)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524#define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1525
Joe Perches69336bd2013-09-22 10:32:26 -07001526bool lock_sock_fast(struct sock *sk);
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001527/**
1528 * unlock_sock_fast - complement of lock_sock_fast
1529 * @sk: socket
1530 * @slow: slow mode
1531 *
1532 * fast unlock socket for user context.
1533 * If slow mode is on, we call regular release_sock()
1534 */
1535static inline void unlock_sock_fast(struct sock *sk, bool slow)
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001536{
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001537 if (slow)
1538 release_sock(sk);
1539 else
1540 spin_unlock_bh(&sk->sk_lock.slock);
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001541}
1542
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001543
Joe Perches69336bd2013-09-22 10:32:26 -07001544struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
1545 struct proto *prot);
1546void sk_free(struct sock *sk);
1547void sk_release_kernel(struct sock *sk);
1548struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549
Joe Perches69336bd2013-09-22 10:32:26 -07001550struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
1551 gfp_t priority);
Joe Perches69336bd2013-09-22 10:32:26 -07001552void sock_wfree(struct sk_buff *skb);
1553void skb_orphan_partial(struct sk_buff *skb);
1554void sock_rfree(struct sk_buff *skb);
1555void sock_edemux(struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001556
Joe Perches69336bd2013-09-22 10:32:26 -07001557int sock_setsockopt(struct socket *sock, int level, int op,
1558 char __user *optval, unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001559
Joe Perches69336bd2013-09-22 10:32:26 -07001560int sock_getsockopt(struct socket *sock, int level, int op,
1561 char __user *optval, int __user *optlen);
1562struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1563 int noblock, int *errcode);
1564struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1565 unsigned long data_len, int noblock,
1566 int *errcode, int max_page_order);
1567void *sock_kmalloc(struct sock *sk, int size, gfp_t priority);
1568void sock_kfree_s(struct sock *sk, void *mem, int size);
1569void sk_send_sigurg(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001570
1571/*
1572 * Functions to fill in entries in struct proto_ops when a protocol
1573 * does not implement a particular function.
1574 */
Joe Perches69336bd2013-09-22 10:32:26 -07001575int sock_no_bind(struct socket *, struct sockaddr *, int);
1576int sock_no_connect(struct socket *, struct sockaddr *, int, int);
1577int sock_no_socketpair(struct socket *, struct socket *);
1578int sock_no_accept(struct socket *, struct socket *, int);
1579int sock_no_getname(struct socket *, struct sockaddr *, int *, int);
1580unsigned int sock_no_poll(struct file *, struct socket *,
1581 struct poll_table_struct *);
1582int sock_no_ioctl(struct socket *, unsigned int, unsigned long);
1583int sock_no_listen(struct socket *, int);
1584int sock_no_shutdown(struct socket *, int);
1585int sock_no_getsockopt(struct socket *, int , int, char __user *, int __user *);
1586int sock_no_setsockopt(struct socket *, int, int, char __user *, unsigned int);
1587int sock_no_sendmsg(struct kiocb *, struct socket *, struct msghdr *, size_t);
1588int sock_no_recvmsg(struct kiocb *, struct socket *, struct msghdr *, size_t,
1589 int);
1590int sock_no_mmap(struct file *file, struct socket *sock,
1591 struct vm_area_struct *vma);
1592ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset,
1593 size_t size, int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001594
1595/*
1596 * Functions to fill in entries in struct proto_ops when a protocol
1597 * uses the inet style.
1598 */
Joe Perches69336bd2013-09-22 10:32:26 -07001599int sock_common_getsockopt(struct socket *sock, int level, int optname,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001600 char __user *optval, int __user *optlen);
Joe Perches69336bd2013-09-22 10:32:26 -07001601int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001602 struct msghdr *msg, size_t size, int flags);
Joe Perches69336bd2013-09-22 10:32:26 -07001603int sock_common_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07001604 char __user *optval, unsigned int optlen);
Joe Perches69336bd2013-09-22 10:32:26 -07001605int compat_sock_common_getsockopt(struct socket *sock, int level,
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001606 int optname, char __user *optval, int __user *optlen);
Joe Perches69336bd2013-09-22 10:32:26 -07001607int compat_sock_common_setsockopt(struct socket *sock, int level,
David S. Millerb7058842009-09-30 16:12:20 -07001608 int optname, char __user *optval, unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001609
Joe Perches69336bd2013-09-22 10:32:26 -07001610void sk_common_release(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001611
1612/*
1613 * Default socket callbacks and setup code
1614 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001615
Linus Torvalds1da177e2005-04-16 15:20:36 -07001616/* Initialise core socket variables */
Joe Perches69336bd2013-09-22 10:32:26 -07001617void sock_init_data(struct socket *sock, struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001618
Joe Perches69336bd2013-09-22 10:32:26 -07001619void sk_filter_release_rcu(struct rcu_head *rcu);
Eric Dumazet46bcf142010-12-06 09:29:43 -08001620
Linus Torvalds1da177e2005-04-16 15:20:36 -07001621/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001622 * sk_filter_release - release a socket filter
Paul Bonserdc9b3342006-11-23 17:56:13 -08001623 * @fp: filter to remove
1624 *
1625 * Remove a filter from a socket and release its resources.
1626 */
1627
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -07001628static inline void sk_filter_release(struct sk_filter *fp)
1629{
1630 if (atomic_dec_and_test(&fp->refcnt))
Eric Dumazet80f8f102011-01-18 07:46:52 +00001631 call_rcu(&fp->rcu, sk_filter_release_rcu);
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -07001632}
1633
1634static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001635{
Alexei Starovoitovd45ed4a2013-10-04 00:14:06 -07001636 atomic_sub(sk_filter_size(fp->len), &sk->sk_omem_alloc);
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -07001637 sk_filter_release(fp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001638}
1639
1640static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
1641{
1642 atomic_inc(&fp->refcnt);
Alexei Starovoitovd45ed4a2013-10-04 00:14:06 -07001643 atomic_add(sk_filter_size(fp->len), &sk->sk_omem_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001644}
1645
1646/*
1647 * Socket reference counting postulates.
1648 *
1649 * * Each user of socket SHOULD hold a reference count.
1650 * * Each access point to socket (an hash table bucket, reference from a list,
1651 * running timer, skb in flight MUST hold a reference count.
1652 * * When reference count hits 0, it means it will never increase back.
1653 * * When reference count hits 0, it means that no references from
1654 * outside exist to this socket and current process on current CPU
1655 * is last user and may/should destroy this socket.
1656 * * sk_free is called from any context: process, BH, IRQ. When
1657 * it is called, socket has no references from outside -> sk_free
1658 * may release descendant resources allocated by the socket, but
1659 * to the time when it is called, socket is NOT referenced by any
1660 * hash tables, lists etc.
1661 * * Packets, delivered from outside (from network or from another process)
1662 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1663 * when they sit in queue. Otherwise, packets will leak to hole, when
1664 * socket is looked up by one cpu and unhasing is made by another CPU.
1665 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1666 * (leak to backlog). Packet socket does all the processing inside
1667 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1668 * use separate SMP lock, so that they are prone too.
1669 */
1670
1671/* Ungrab socket and destroy it, if it was the last reference. */
1672static inline void sock_put(struct sock *sk)
1673{
1674 if (atomic_dec_and_test(&sk->sk_refcnt))
1675 sk_free(sk);
1676}
Eric Dumazet05dbc7b2013-10-03 00:22:02 -07001677/* Generic version of sock_put(), dealing with all sockets
1678 * (TCP_TIMEWAIT, ESTABLISHED...)
1679 */
1680void sock_gen_put(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001681
Joe Perches69336bd2013-09-22 10:32:26 -07001682int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested);
Arnaldo Carvalho de Melo25995ff2005-12-27 02:42:22 -02001683
Krishna Kumare022f0b2009-10-19 23:46:20 +00001684static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1685{
1686 sk->sk_tx_queue_mapping = tx_queue;
1687}
1688
1689static inline void sk_tx_queue_clear(struct sock *sk)
1690{
1691 sk->sk_tx_queue_mapping = -1;
1692}
1693
1694static inline int sk_tx_queue_get(const struct sock *sk)
1695{
Tom Herbertb0f77d02010-07-14 20:50:29 -07001696 return sk ? sk->sk_tx_queue_mapping : -1;
Krishna Kumare022f0b2009-10-19 23:46:20 +00001697}
1698
David S. Miller972692e2008-06-17 22:41:38 -07001699static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1700{
Krishna Kumare022f0b2009-10-19 23:46:20 +00001701 sk_tx_queue_clear(sk);
David S. Miller972692e2008-06-17 22:41:38 -07001702 sk->sk_socket = sock;
1703}
1704
Eric Dumazetaa395142010-04-20 13:03:51 +00001705static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1706{
Eric Dumazeteaefd112011-02-18 03:26:36 +00001707 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1708 return &rcu_dereference_raw(sk->sk_wq)->wait;
Eric Dumazetaa395142010-04-20 13:03:51 +00001709}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710/* Detach socket from process context.
1711 * Announce socket dead, detach it from wait queue and inode.
1712 * Note that parent inode held reference count on this struct sock,
1713 * we do not release it in this function, because protocol
1714 * probably wants some additional cleanups or even continuing
1715 * to work with this socket (TCP).
1716 */
1717static inline void sock_orphan(struct sock *sk)
1718{
1719 write_lock_bh(&sk->sk_callback_lock);
1720 sock_set_flag(sk, SOCK_DEAD);
David S. Miller972692e2008-06-17 22:41:38 -07001721 sk_set_socket(sk, NULL);
Eric Dumazet43815482010-04-29 11:01:49 +00001722 sk->sk_wq = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001723 write_unlock_bh(&sk->sk_callback_lock);
1724}
1725
1726static inline void sock_graft(struct sock *sk, struct socket *parent)
1727{
1728 write_lock_bh(&sk->sk_callback_lock);
Eric Dumazeteaefd112011-02-18 03:26:36 +00001729 sk->sk_wq = parent->wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001730 parent->sk = sk;
David S. Miller972692e2008-06-17 22:41:38 -07001731 sk_set_socket(sk, parent);
Venkat Yekkirala4237c752006-07-24 23:32:50 -07001732 security_sock_graft(sk, parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733 write_unlock_bh(&sk->sk_callback_lock);
1734}
1735
Joe Perches69336bd2013-09-22 10:32:26 -07001736kuid_t sock_i_uid(struct sock *sk);
1737unsigned long sock_i_ino(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001738
1739static inline struct dst_entry *
1740__sk_dst_get(struct sock *sk)
1741{
Michal Hockod8bf4ca2011-07-08 14:39:41 +02001742 return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
Eric Dumazetf68c2242010-04-22 16:06:59 -07001743 lockdep_is_held(&sk->sk_lock.slock));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744}
1745
1746static inline struct dst_entry *
1747sk_dst_get(struct sock *sk)
1748{
1749 struct dst_entry *dst;
1750
Eric Dumazetb6c67122010-04-08 23:03:29 +00001751 rcu_read_lock();
1752 dst = rcu_dereference(sk->sk_dst_cache);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001753 if (dst)
1754 dst_hold(dst);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001755 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001756 return dst;
1757}
1758
Eric Dumazetb6c67122010-04-08 23:03:29 +00001759static inline void dst_negative_advice(struct sock *sk)
1760{
1761 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1762
1763 if (dst && dst->ops->negative_advice) {
1764 ndst = dst->ops->negative_advice(dst);
1765
1766 if (ndst != dst) {
1767 rcu_assign_pointer(sk->sk_dst_cache, ndst);
ZHAO Gang0a6957e2013-10-22 16:23:38 +08001768 sk_tx_queue_clear(sk);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001769 }
1770 }
1771}
1772
Linus Torvalds1da177e2005-04-16 15:20:36 -07001773static inline void
1774__sk_dst_set(struct sock *sk, struct dst_entry *dst)
1775{
1776 struct dst_entry *old_dst;
1777
Krishna Kumare022f0b2009-10-19 23:46:20 +00001778 sk_tx_queue_clear(sk);
Eric Dumazet0b53ff22010-04-26 20:40:43 +00001779 /*
1780 * This can be called while sk is owned by the caller only,
1781 * with no state that can be checked in a rcu_dereference_check() cond
1782 */
1783 old_dst = rcu_dereference_raw(sk->sk_dst_cache);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001784 rcu_assign_pointer(sk->sk_dst_cache, dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001785 dst_release(old_dst);
1786}
1787
1788static inline void
1789sk_dst_set(struct sock *sk, struct dst_entry *dst)
1790{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001791 spin_lock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001792 __sk_dst_set(sk, dst);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001793 spin_unlock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001794}
1795
1796static inline void
1797__sk_dst_reset(struct sock *sk)
1798{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001799 __sk_dst_set(sk, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001800}
1801
1802static inline void
1803sk_dst_reset(struct sock *sk)
1804{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001805 spin_lock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001806 __sk_dst_reset(sk);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001807 spin_unlock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001808}
1809
Joe Perches69336bd2013-09-22 10:32:26 -07001810struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001811
Joe Perches69336bd2013-09-22 10:32:26 -07001812struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001813
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001814static inline bool sk_can_gso(const struct sock *sk)
Herbert Xubcd76112006-06-30 13:36:35 -07001815{
1816 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1817}
1818
Joe Perches69336bd2013-09-22 10:32:26 -07001819void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
Arnaldo Carvalho de Melo6cbb0df2005-08-09 19:49:02 -07001820
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001821static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
Eric Dumazeta4654192010-05-16 00:36:33 -07001822{
1823 sk->sk_route_nocaps |= flags;
1824 sk->sk_route_caps &= ~flags;
1825}
1826
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001827static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1828 char __user *from, char *to,
Wei Yongjun912d3982011-04-06 18:40:12 +00001829 int copy, int offset)
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001830{
1831 if (skb->ip_summed == CHECKSUM_NONE) {
1832 int err = 0;
1833 __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err);
1834 if (err)
1835 return err;
Wei Yongjun912d3982011-04-06 18:40:12 +00001836 skb->csum = csum_block_add(skb->csum, csum, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001837 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1838 if (!access_ok(VERIFY_READ, from, copy) ||
1839 __copy_from_user_nocache(to, from, copy))
1840 return -EFAULT;
1841 } else if (copy_from_user(to, from, copy))
1842 return -EFAULT;
1843
1844 return 0;
1845}
1846
1847static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
1848 char __user *from, int copy)
1849{
Wei Yongjun912d3982011-04-06 18:40:12 +00001850 int err, offset = skb->len;
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001851
Wei Yongjun912d3982011-04-06 18:40:12 +00001852 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1853 copy, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001854 if (err)
Wei Yongjun912d3982011-04-06 18:40:12 +00001855 __skb_trim(skb, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001856
1857 return err;
1858}
1859
1860static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from,
1861 struct sk_buff *skb,
1862 struct page *page,
1863 int off, int copy)
1864{
1865 int err;
1866
Wei Yongjun912d3982011-04-06 18:40:12 +00001867 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1868 copy, skb->len);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001869 if (err)
1870 return err;
1871
1872 skb->len += copy;
1873 skb->data_len += copy;
1874 skb->truesize += copy;
1875 sk->sk_wmem_queued += copy;
1876 sk_mem_charge(sk, copy);
1877 return 0;
1878}
1879
Linus Torvalds1da177e2005-04-16 15:20:36 -07001880static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1881 struct sk_buff *skb, struct page *page,
1882 int off, int copy)
1883{
1884 if (skb->ip_summed == CHECKSUM_NONE) {
1885 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08001886 __wsum csum = csum_and_copy_from_user(from,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001887 page_address(page) + off,
1888 copy, 0, &err);
1889 if (err)
1890 return err;
1891 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1892 } else if (copy_from_user(page_address(page) + off, from, copy))
1893 return -EFAULT;
1894
1895 skb->len += copy;
1896 skb->data_len += copy;
1897 skb->truesize += copy;
1898 sk->sk_wmem_queued += copy;
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001899 sk_mem_charge(sk, copy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001900 return 0;
1901}
1902
Eric Dumazetc5640392009-06-16 10:12:03 +00001903/**
1904 * sk_wmem_alloc_get - returns write allocations
1905 * @sk: socket
1906 *
1907 * Returns sk_wmem_alloc minus initial offset of one
1908 */
1909static inline int sk_wmem_alloc_get(const struct sock *sk)
1910{
1911 return atomic_read(&sk->sk_wmem_alloc) - 1;
1912}
1913
1914/**
1915 * sk_rmem_alloc_get - returns read allocations
1916 * @sk: socket
1917 *
1918 * Returns sk_rmem_alloc
1919 */
1920static inline int sk_rmem_alloc_get(const struct sock *sk)
1921{
1922 return atomic_read(&sk->sk_rmem_alloc);
1923}
1924
1925/**
1926 * sk_has_allocations - check if allocations are outstanding
1927 * @sk: socket
1928 *
1929 * Returns true if socket has write or read allocations
1930 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001931static inline bool sk_has_allocations(const struct sock *sk)
Eric Dumazetc5640392009-06-16 10:12:03 +00001932{
1933 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1934}
1935
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001936/**
Eric Dumazet43815482010-04-29 11:01:49 +00001937 * wq_has_sleeper - check if there are any waiting processes
Randy Dunlapacfbe962010-05-24 23:54:18 -07001938 * @wq: struct socket_wq
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001939 *
Eric Dumazet43815482010-04-29 11:01:49 +00001940 * Returns true if socket_wq has waiting processes
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001941 *
Eric Dumazet43815482010-04-29 11:01:49 +00001942 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001943 * barrier call. They were added due to the race found within the tcp code.
1944 *
1945 * Consider following tcp code paths:
1946 *
1947 * CPU1 CPU2
1948 *
1949 * sys_select receive packet
1950 * ... ...
1951 * __add_wait_queue update tp->rcv_nxt
1952 * ... ...
1953 * tp->rcv_nxt check sock_def_readable
1954 * ... {
Eric Dumazet43815482010-04-29 11:01:49 +00001955 * schedule rcu_read_lock();
1956 * wq = rcu_dereference(sk->sk_wq);
1957 * if (wq && waitqueue_active(&wq->wait))
1958 * wake_up_interruptible(&wq->wait)
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001959 * ...
1960 * }
1961 *
1962 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1963 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1964 * could then endup calling schedule and sleep forever if there are no more
1965 * data on the socket.
Jiri Olsaad462762009-07-08 12:10:31 +00001966 *
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001967 */
Eric Dumazet43815482010-04-29 11:01:49 +00001968static inline bool wq_has_sleeper(struct socket_wq *wq)
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001969{
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001970 /* We need to be sure we are in sync with the
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001971 * add_wait_queue modifications to the wait queue.
1972 *
1973 * This memory barrier is paired in the sock_poll_wait.
1974 */
Eric Dumazet43815482010-04-29 11:01:49 +00001975 smp_mb();
1976 return wq && waitqueue_active(&wq->wait);
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001977}
1978
1979/**
1980 * sock_poll_wait - place memory barrier behind the poll_wait call.
1981 * @filp: file
1982 * @wait_address: socket wait queue
1983 * @p: poll_table
1984 *
Eric Dumazet43815482010-04-29 11:01:49 +00001985 * See the comments in the wq_has_sleeper function.
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001986 */
1987static inline void sock_poll_wait(struct file *filp,
1988 wait_queue_head_t *wait_address, poll_table *p)
1989{
Hans Verkuil626cf232012-03-23 15:02:27 -07001990 if (!poll_does_not_wait(p) && wait_address) {
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001991 poll_wait(filp, wait_address, p);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001992 /* We need to be sure we are in sync with the
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001993 * socket flags modification.
1994 *
Eric Dumazet43815482010-04-29 11:01:49 +00001995 * This memory barrier is paired in the wq_has_sleeper.
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001996 */
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001997 smp_mb();
1998 }
1999}
2000
Linus Torvalds1da177e2005-04-16 15:20:36 -07002001/*
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002002 * Queue a received datagram if it will fit. Stream and sequenced
Linus Torvalds1da177e2005-04-16 15:20:36 -07002003 * protocols can't normally use this as they need to fit buffers in
2004 * and play with them.
2005 *
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002006 * Inlined as it's very short and called for pretty much every
Linus Torvalds1da177e2005-04-16 15:20:36 -07002007 * packet ever received.
2008 */
2009
2010static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
2011{
Herbert Xud55d87f2009-06-22 02:25:25 +00002012 skb_orphan(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002013 skb->sk = sk;
2014 skb->destructor = sock_wfree;
Eric Dumazet2b85a342009-06-11 02:55:43 -07002015 /*
2016 * We used to take a refcount on sk, but following operation
2017 * is enough to guarantee sk_free() wont free this sock until
2018 * all in-flight packets are completed
2019 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002020 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
2021}
2022
2023static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
2024{
Herbert Xud55d87f2009-06-22 02:25:25 +00002025 skb_orphan(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002026 skb->sk = sk;
2027 skb->destructor = sock_rfree;
2028 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08002029 sk_mem_charge(sk, skb->truesize);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002030}
2031
Joe Perches69336bd2013-09-22 10:32:26 -07002032void sk_reset_timer(struct sock *sk, struct timer_list *timer,
2033 unsigned long expires);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002034
Joe Perches69336bd2013-09-22 10:32:26 -07002035void sk_stop_timer(struct sock *sk, struct timer_list *timer);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002036
Joe Perches69336bd2013-09-22 10:32:26 -07002037int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002038
Joe Perches69336bd2013-09-22 10:32:26 -07002039int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002040
2041/*
2042 * Recover an error report and clear atomically
2043 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002044
Linus Torvalds1da177e2005-04-16 15:20:36 -07002045static inline int sock_error(struct sock *sk)
2046{
Benjamin LaHaisec1cbe4b2005-12-13 23:22:19 -08002047 int err;
2048 if (likely(!sk->sk_err))
2049 return 0;
2050 err = xchg(&sk->sk_err, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002051 return -err;
2052}
2053
2054static inline unsigned long sock_wspace(struct sock *sk)
2055{
2056 int amt = 0;
2057
2058 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
2059 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002060 if (amt < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002061 amt = 0;
2062 }
2063 return amt;
2064}
2065
2066static inline void sk_wake_async(struct sock *sk, int how, int band)
2067{
Eric Dumazetbcdce712009-10-06 17:28:29 -07002068 if (sock_flag(sk, SOCK_FASYNC))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002069 sock_wake_async(sk->sk_socket, how, band);
2070}
2071
Daniel Borkmanneea86af2013-06-19 12:51:20 +02002072/* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
2073 * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
2074 * Note: for send buffers, TCP works better if we can build two skbs at
2075 * minimum.
Eric Dumazet7a91b4342010-09-26 18:53:07 -07002076 */
Eric Dumazet9eb5bf82013-07-03 05:02:22 -07002077#define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
Daniel Borkmanneea86af2013-06-19 12:51:20 +02002078
2079#define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2)
2080#define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE
Linus Torvalds1da177e2005-04-16 15:20:36 -07002081
2082static inline void sk_stream_moderate_sndbuf(struct sock *sk)
2083{
2084 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
Eric Dumazet8df09ea2007-12-21 03:07:41 -08002085 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
Daniel Borkmanneea86af2013-06-19 12:51:20 +02002086 sk->sk_sndbuf = max_t(u32, sk->sk_sndbuf, SOCK_MIN_SNDBUF);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002087 }
2088}
2089
Pavel Emelyanovdf97c702007-11-29 21:22:33 +11002090struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002091
Eric Dumazet5640f762012-09-23 23:04:42 +00002092/**
2093 * sk_page_frag - return an appropriate page_frag
2094 * @sk: socket
2095 *
2096 * If socket allocation mode allows current thread to sleep, it means its
2097 * safe to use the per task page_frag instead of the per socket one.
2098 */
2099static inline struct page_frag *sk_page_frag(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002100{
Eric Dumazet5640f762012-09-23 23:04:42 +00002101 if (sk->sk_allocation & __GFP_WAIT)
2102 return &current->task_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002103
Eric Dumazet5640f762012-09-23 23:04:42 +00002104 return &sk->sk_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002105}
2106
Joe Perches69336bd2013-09-22 10:32:26 -07002107bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag);
Eric Dumazet5640f762012-09-23 23:04:42 +00002108
Linus Torvalds1da177e2005-04-16 15:20:36 -07002109/*
2110 * Default write policy as shown to user space via poll/select/SIGIO
2111 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002112static inline bool sock_writeable(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002113{
Eric Dumazet8df09ea2007-12-21 03:07:41 -08002114 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002115}
2116
Al Virodd0fc662005-10-07 07:46:04 +01002117static inline gfp_t gfp_any(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002118{
Andrew Morton99709372009-02-12 16:43:17 -08002119 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002120}
2121
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002122static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002123{
2124 return noblock ? 0 : sk->sk_rcvtimeo;
2125}
2126
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002127static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002128{
2129 return noblock ? 0 : sk->sk_sndtimeo;
2130}
2131
2132static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2133{
2134 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
2135}
2136
2137/* Alas, with timeout socket operations are not restartable.
2138 * Compare this to poll().
2139 */
2140static inline int sock_intr_errno(long timeo)
2141{
2142 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2143}
2144
Joe Perches69336bd2013-09-22 10:32:26 -07002145void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2146 struct sk_buff *skb);
2147void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2148 struct sk_buff *skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -07002149
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002150static inline void
Linus Torvalds1da177e2005-04-16 15:20:36 -07002151sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2152{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002153 ktime_t kt = skb->tstamp;
Patrick Ohly20d49472009-02-12 05:03:38 +00002154 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002155
Patrick Ohly20d49472009-02-12 05:03:38 +00002156 /*
2157 * generate control messages if
2158 * - receive time stamping in software requested (SOCK_RCVTSTAMP
2159 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
2160 * - software time stamp available and wanted
2161 * (SOCK_TIMESTAMPING_SOFTWARE)
2162 * - hardware time stamps available and wanted
2163 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
2164 * SOCK_TIMESTAMPING_RAW_HARDWARE)
2165 */
2166 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
2167 sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
2168 (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
2169 (hwtstamps->hwtstamp.tv64 &&
2170 sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
2171 (hwtstamps->syststamp.tv64 &&
2172 sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
Eric Dumazet92f37fd2007-03-25 22:14:49 -07002173 __sock_recv_timestamp(msg, sk, skb);
2174 else
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002175 sk->sk_stamp = kt;
Johannes Berg6e3e9392011-11-09 10:15:42 +01002176
2177 if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2178 __sock_recv_wifi_status(msg, sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002179}
2180
Joe Perches69336bd2013-09-22 10:32:26 -07002181void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2182 struct sk_buff *skb);
Eric Dumazet767dd032010-04-28 19:14:43 +00002183
2184static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2185 struct sk_buff *skb)
2186{
2187#define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2188 (1UL << SOCK_RCVTSTAMP) | \
2189 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
2190 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002191 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
Eric Dumazet767dd032010-04-28 19:14:43 +00002192 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
2193
2194 if (sk->sk_flags & FLAGS_TS_OR_DROPS)
2195 __sock_recv_ts_and_drops(msg, sk, skb);
2196 else
2197 sk->sk_stamp = skb->tstamp;
2198}
Neil Horman3b885782009-10-12 13:26:31 -07002199
Linus Torvalds1da177e2005-04-16 15:20:36 -07002200/**
Patrick Ohly20d49472009-02-12 05:03:38 +00002201 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
Patrick Ohly20d49472009-02-12 05:03:38 +00002202 * @sk: socket sending this packet
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002203 * @tx_flags: filled with instructions for time stamping
Patrick Ohly20d49472009-02-12 05:03:38 +00002204 *
Daniel Borkmannbf84a012013-04-14 08:08:13 +00002205 * Currently only depends on SOCK_TIMESTAMPING* flags.
Patrick Ohly20d49472009-02-12 05:03:38 +00002206 */
Joe Perches69336bd2013-09-22 10:32:26 -07002207void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags);
Patrick Ohly20d49472009-02-12 05:03:38 +00002208
2209/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002210 * sk_eat_skb - Release a skb if it is no longer needed
Pavel Pisa4dc3b162005-05-01 08:59:25 -07002211 * @sk: socket to eat this skb from
2212 * @skb: socket buffer to eat
Randy Dunlapf4b8ea72006-06-22 16:00:11 -07002213 * @copied_early: flag indicating whether DMA operations copied this data early
Linus Torvalds1da177e2005-04-16 15:20:36 -07002214 *
2215 * This routine must be called with interrupts disabled or with the socket
2216 * locked so that the sk_buff queue operation is ok.
2217*/
Chris Leech624d1162006-05-23 18:01:28 -07002218#ifdef CONFIG_NET_DMA
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002219static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
Chris Leech624d1162006-05-23 18:01:28 -07002220{
2221 __skb_unlink(skb, &sk->sk_receive_queue);
2222 if (!copied_early)
2223 __kfree_skb(skb);
2224 else
2225 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
2226}
2227#else
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002228static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002229{
2230 __skb_unlink(skb, &sk->sk_receive_queue);
2231 __kfree_skb(skb);
2232}
Chris Leech624d1162006-05-23 18:01:28 -07002233#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002234
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002235static inline
2236struct net *sock_net(const struct sock *sk)
2237{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002238 return read_pnet(&sk->sk_net);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002239}
2240
2241static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07002242void sock_net_set(struct sock *sk, struct net *net)
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002243{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002244 write_pnet(&sk->sk_net, net);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002245}
2246
Denis V. Lunevedf02082008-02-29 11:18:32 -08002247/*
2248 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
Lucas De Marchi25985ed2011-03-30 22:57:33 -03002249 * They should not hold a reference to a namespace in order to allow
Denis V. Lunevedf02082008-02-29 11:18:32 -08002250 * to stop it.
2251 * Sockets after sk_change_net should be released using sk_release_kernel
2252 */
2253static inline void sk_change_net(struct sock *sk, struct net *net)
2254{
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002255 put_net(sock_net(sk));
Denis V. Lunev65a18ec2008-04-16 01:59:46 -07002256 sock_net_set(sk, hold_net(net));
Denis V. Lunevedf02082008-02-29 11:18:32 -08002257}
2258
KOVACS Krisztian23542612008-10-07 12:41:01 -07002259static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2260{
Vijay Subramanianefc27f82012-06-24 13:03:07 +00002261 if (skb->sk) {
KOVACS Krisztian23542612008-10-07 12:41:01 -07002262 struct sock *sk = skb->sk;
2263
2264 skb->destructor = NULL;
2265 skb->sk = NULL;
2266 return sk;
2267 }
2268 return NULL;
2269}
2270
Joe Perches69336bd2013-09-22 10:32:26 -07002271void sock_enable_timestamp(struct sock *sk, int flag);
2272int sock_get_timestamp(struct sock *, struct timeval __user *);
2273int sock_get_timestampns(struct sock *, struct timespec __user *);
2274int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len, int level,
2275 int type);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002276
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002277/*
2278 * Enable debug/info messages
Linus Torvalds1da177e2005-04-16 15:20:36 -07002279 */
Stephen Hemmingera2a316f2007-03-08 20:41:08 -08002280extern int net_msg_warn;
2281#define NETDEBUG(fmt, args...) \
2282 do { if (net_msg_warn) printk(fmt,##args); } while (0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002283
Stephen Hemmingera2a316f2007-03-08 20:41:08 -08002284#define LIMIT_NETDEBUG(fmt, args...) \
2285 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002286
Linus Torvalds1da177e2005-04-16 15:20:36 -07002287extern __u32 sysctl_wmem_max;
2288extern __u32 sysctl_rmem_max;
2289
David S. Miller6baf1f42005-09-05 18:14:11 -07002290extern int sysctl_optmem_max;
2291
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002292extern __u32 sysctl_wmem_default;
2293extern __u32 sysctl_rmem_default;
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002294
Linus Torvalds1da177e2005-04-16 15:20:36 -07002295#endif /* _SOCK_H */