blob: 14f6e9d19dc735135c263022fa394173e89aea6e [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
159 * address on 64bit arches : cf INET_MATCH() and INET_TW_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 Dumazet68835ab2010-11-30 19:04:07 +0000194 /*
195 * fields between dontcopy_begin/dontcopy_end
196 * are not copied in sock_copy()
197 */
Randy Dunlap928c41e2011-01-08 17:39:21 +0000198 /* private: */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000199 int skc_dontcopy_begin[0];
Randy Dunlap928c41e2011-01-08 17:39:21 +0000200 /* public: */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000201 union {
202 struct hlist_node skc_node;
203 struct hlist_nulls_node skc_nulls_node;
204 };
205 int skc_tx_queue_mapping;
206 atomic_t skc_refcnt;
Randy Dunlap928c41e2011-01-08 17:39:21 +0000207 /* private: */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000208 int skc_dontcopy_end[0];
Randy Dunlap928c41e2011-01-08 17:39:21 +0000209 /* public: */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210};
211
Glauber Costae1aab162011-12-11 21:47:03 +0000212struct cg_proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213/**
214 * struct sock - network layer representation of sockets
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700215 * @__sk_common: shared layout with inet_timewait_sock
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700216 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
217 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
218 * @sk_lock: synchronizer
219 * @sk_rcvbuf: size of receive buffer in bytes
Eric Dumazet43815482010-04-29 11:01:49 +0000220 * @sk_wq: sock wait queue and async head
Eric Dumazetdeaa5852012-06-24 20:22:49 +0000221 * @sk_rx_dst: receive input route used by early tcp demux
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700222 * @sk_dst_cache: destination cache
223 * @sk_dst_lock: destination cache lock
224 * @sk_policy: flow policy
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700225 * @sk_receive_queue: incoming packets
226 * @sk_wmem_alloc: transmit queue bytes committed
227 * @sk_write_queue: Packet sending queue
Chris Leech97fc2f02006-05-23 17:55:33 -0700228 * @sk_async_wait_queue: DMA copied packets
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700229 * @sk_omem_alloc: "o" is "option" or "other"
230 * @sk_wmem_queued: persistent queue size
231 * @sk_forward_alloc: space allocated forward
232 * @sk_allocation: allocation mode
233 * @sk_sndbuf: size of send buffer in bytes
Wang Chen33c732c2007-11-13 20:30:01 -0800234 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
Patrick Ohly20d49472009-02-12 05:03:38 +0000235 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
Adam Buchbinder48fc7f72012-09-19 21:48:00 -0400236 * @sk_no_check: %SO_NO_CHECK setting, whether or not checkup packets
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700237 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
Eric Dumazeta4654192010-05-16 00:36:33 -0700238 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
Herbert Xubcd76112006-06-30 13:36:35 -0700239 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -0700240 * @sk_gso_max_size: Maximum GSO segment size to build
Ben Hutchings14853482012-07-30 16:11:42 +0000241 * @sk_gso_max_segs: Maximum number of GSO segments
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700242 * @sk_lingertime: %SO_LINGER l_linger setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700243 * @sk_backlog: always used with the per-socket spinlock held
244 * @sk_callback_lock: used with the callbacks in the end of this struct
245 * @sk_error_queue: rarely used
Wang Chen33c732c2007-11-13 20:30:01 -0800246 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
247 * IPV6_ADDRFORM for instance)
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700248 * @sk_err: last error
Wang Chen33c732c2007-11-13 20:30:01 -0800249 * @sk_err_soft: errors that don't cause failure but are the cause of a
250 * persistent failure not just 'timed out'
Eric Dumazetcb61cb92008-06-17 21:04:56 -0700251 * @sk_drops: raw/udp drops counter
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700252 * @sk_ack_backlog: current listen backlog
253 * @sk_max_ack_backlog: listen backlog set in listen()
254 * @sk_priority: %SO_PRIORITY setting
Randy Dunlap1a3bc362012-01-21 09:03:10 +0000255 * @sk_cgrp_prioidx: socket group's priority map index
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700256 * @sk_type: socket type (%SOCK_STREAM, etc)
257 * @sk_protocol: which protocol this socket belongs in this network family
Randy Dunlap53c3fa22010-08-09 13:41:07 +0000258 * @sk_peer_pid: &struct pid for this socket's peer
259 * @sk_peer_cred: %SO_PEERCRED setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700260 * @sk_rcvlowat: %SO_RCVLOWAT setting
261 * @sk_rcvtimeo: %SO_RCVTIMEO setting
262 * @sk_sndtimeo: %SO_SNDTIMEO setting
David S. Millerc58dc012010-04-27 15:05:31 -0700263 * @sk_rxhash: flow hash received from netif layer
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700264 * @sk_filter: socket filtering instructions
265 * @sk_protinfo: private area, net family specific, when not using slab
266 * @sk_timer: sock cleanup timer
267 * @sk_stamp: time stamp of last packet received
268 * @sk_socket: Identd and reporting IO signals
269 * @sk_user_data: RPC layer private data
Eric Dumazet5640f762012-09-23 23:04:42 +0000270 * @sk_frag: cached page frag
Randy Dunlapd3d4f0a2012-04-17 14:03:53 +0000271 * @sk_peek_off: current peek_offset value
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700272 * @sk_send_head: front of stuff to transmit
Martin Waitz67be2dd2005-05-01 08:59:26 -0700273 * @sk_security: used by security modules
Randy Dunlap31729362008-02-18 20:52:13 -0800274 * @sk_mark: generic packet mark
Randy Dunlap53c3fa22010-08-09 13:41:07 +0000275 * @sk_classid: this socket's cgroup classid
Glauber Costae1aab162011-12-11 21:47:03 +0000276 * @sk_cgrp: this socket's cgroup-specific proto data
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700277 * @sk_write_pending: a write to stream socket waits to start
278 * @sk_state_change: callback to indicate change in the state of the sock
279 * @sk_data_ready: callback to indicate there is data to be processed
280 * @sk_write_space: callback to indicate there is bf sending space available
281 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
282 * @sk_backlog_rcv: callback to process the backlog
283 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284 */
285struct sock {
286 /*
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700287 * Now struct inet_timewait_sock also uses sock_common, so please just
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288 * don't add nothing before this first member (__sk_common) --acme
289 */
290 struct sock_common __sk_common;
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000291#define sk_node __sk_common.skc_node
292#define sk_nulls_node __sk_common.skc_nulls_node
293#define sk_refcnt __sk_common.skc_refcnt
Krishna Kumare022f0b2009-10-19 23:46:20 +0000294#define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000295
Eric Dumazet68835ab2010-11-30 19:04:07 +0000296#define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
297#define sk_dontcopy_end __sk_common.skc_dontcopy_end
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000298#define sk_hash __sk_common.skc_hash
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299#define sk_family __sk_common.skc_family
300#define sk_state __sk_common.skc_state
301#define sk_reuse __sk_common.skc_reuse
Tom Herbert055dc212013-01-22 09:49:50 +0000302#define sk_reuseport __sk_common.skc_reuseport
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303#define sk_bound_dev_if __sk_common.skc_bound_dev_if
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304#define sk_bind_node __sk_common.skc_bind_node
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700305#define sk_prot __sk_common.skc_prot
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200306#define sk_net __sk_common.skc_net
Eric Dumazetb178bb32010-11-16 05:56:04 +0000307 socket_lock_t sk_lock;
308 struct sk_buff_head sk_receive_queue;
309 /*
310 * The backlog queue is special, it is always used with
311 * the per-socket spinlock held and requires low latency
312 * access. Therefore we special case it's implementation.
313 * Note : rmem_alloc is in this structure to fill a hole
314 * on 64bit arches, not because its logically part of
315 * backlog.
316 */
317 struct {
318 atomic_t rmem_alloc;
319 int len;
320 struct sk_buff *head;
321 struct sk_buff *tail;
322 } sk_backlog;
323#define sk_rmem_alloc sk_backlog.rmem_alloc
324 int sk_forward_alloc;
325#ifdef CONFIG_RPS
326 __u32 sk_rxhash;
327#endif
328 atomic_t sk_drops;
329 int sk_rcvbuf;
330
331 struct sk_filter __rcu *sk_filter;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000332 struct socket_wq __rcu *sk_wq;
Eric Dumazetb178bb32010-11-16 05:56:04 +0000333
334#ifdef CONFIG_NET_DMA
335 struct sk_buff_head sk_async_wait_queue;
336#endif
337
338#ifdef CONFIG_XFRM
339 struct xfrm_policy *sk_policy[2];
340#endif
341 unsigned long sk_flags;
Eric Dumazetdeaa5852012-06-24 20:22:49 +0000342 struct dst_entry *sk_rx_dst;
Cong Wang0e36cbb2013-01-22 21:09:51 +0000343 struct dst_entry __rcu *sk_dst_cache;
Eric Dumazetb178bb32010-11-16 05:56:04 +0000344 spinlock_t sk_dst_lock;
345 atomic_t sk_wmem_alloc;
346 atomic_t sk_omem_alloc;
347 int sk_sndbuf;
348 struct sk_buff_head sk_write_queue;
Vegard Nossuma98b65a2009-02-26 14:46:57 +0100349 kmemcheck_bitfield_begin(flags);
Eric Dumazet5fdb9972009-10-08 22:50:25 +0000350 unsigned int sk_shutdown : 2,
351 sk_no_check : 2,
352 sk_userlocks : 4,
353 sk_protocol : 8,
354 sk_type : 16;
Vegard Nossuma98b65a2009-02-26 14:46:57 +0100355 kmemcheck_bitfield_end(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356 int sk_wmem_queued;
Al Viro7d877f32005-10-21 03:20:43 -0400357 gfp_t sk_allocation;
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000358 netdev_features_t sk_route_caps;
359 netdev_features_t sk_route_nocaps;
Herbert Xubcd76112006-06-30 13:36:35 -0700360 int sk_gso_type;
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -0700361 unsigned int sk_gso_max_size;
Ben Hutchings14853482012-07-30 16:11:42 +0000362 u16 sk_gso_max_segs;
David S. Miller9932cf92006-03-24 15:12:37 -0800363 int sk_rcvlowat;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364 unsigned long sk_lingertime;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365 struct sk_buff_head sk_error_queue;
Arnaldo Carvalho de Melo476e19c2005-05-05 13:35:15 -0700366 struct proto *sk_prot_creator;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367 rwlock_t sk_callback_lock;
368 int sk_err,
369 sk_err_soft;
370 unsigned short sk_ack_backlog;
371 unsigned short sk_max_ack_backlog;
372 __u32 sk_priority;
Li Zefan3d0dcfb2012-12-25 20:48:24 +0000373#if IS_ENABLED(CONFIG_NETPRIO_CGROUP)
Neil Horman5bc14212011-11-22 05:10:51 +0000374 __u32 sk_cgrp_prioidx;
375#endif
Eric W. Biederman109f6e32010-06-13 03:30:14 +0000376 struct pid *sk_peer_pid;
377 const struct cred *sk_peer_cred;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378 long sk_rcvtimeo;
379 long sk_sndtimeo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380 void *sk_protinfo;
381 struct timer_list sk_timer;
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700382 ktime_t sk_stamp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383 struct socket *sk_socket;
384 void *sk_user_data;
Eric Dumazet5640f762012-09-23 23:04:42 +0000385 struct page_frag sk_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386 struct sk_buff *sk_send_head;
Pavel Emelyanovef64a542012-02-21 07:31:34 +0000387 __s32 sk_peek_off;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388 int sk_write_pending;
Alexey Dobriyand5f64232008-11-04 14:45:58 -0800389#ifdef CONFIG_SECURITY
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390 void *sk_security;
Alexey Dobriyand5f64232008-11-04 14:45:58 -0800391#endif
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800392 __u32 sk_mark;
Herbert Xuf8451722010-05-24 00:12:34 -0700393 u32 sk_classid;
Glauber Costae1aab162011-12-11 21:47:03 +0000394 struct cg_proto *sk_cgrp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395 void (*sk_state_change)(struct sock *sk);
396 void (*sk_data_ready)(struct sock *sk, int bytes);
397 void (*sk_write_space)(struct sock *sk);
398 void (*sk_error_report)(struct sock *sk);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000399 int (*sk_backlog_rcv)(struct sock *sk,
400 struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401 void (*sk_destruct)(struct sock *sk);
402};
403
Pavel Emelyanov4a17fd52012-04-19 03:39:36 +0000404/*
405 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
406 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
407 * on a socket means that the socket will reuse everybody else's port
408 * without looking at the other's sk_reuse value.
409 */
410
411#define SK_NO_REUSE 0
412#define SK_CAN_REUSE 1
413#define SK_FORCE_REUSE 2
414
Pavel Emelyanovef64a542012-02-21 07:31:34 +0000415static inline int sk_peek_offset(struct sock *sk, int flags)
416{
417 if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0))
418 return sk->sk_peek_off;
419 else
420 return 0;
421}
422
423static inline void sk_peek_offset_bwd(struct sock *sk, int val)
424{
425 if (sk->sk_peek_off >= 0) {
426 if (sk->sk_peek_off >= val)
427 sk->sk_peek_off -= val;
428 else
429 sk->sk_peek_off = 0;
430 }
431}
432
433static inline void sk_peek_offset_fwd(struct sock *sk, int val)
434{
435 if (sk->sk_peek_off >= 0)
436 sk->sk_peek_off += val;
437}
438
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439/*
440 * Hashed lists helper routines
441 */
Li Zefanc4146642010-02-08 23:18:45 +0000442static inline struct sock *sk_entry(const struct hlist_node *node)
443{
444 return hlist_entry(node, struct sock, sk_node);
445}
446
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700447static inline struct sock *__sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448{
449 return hlist_entry(head->first, struct sock, sk_node);
450}
451
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700452static inline struct sock *sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453{
454 return hlist_empty(head) ? NULL : __sk_head(head);
455}
456
Eric Dumazet88ab1932008-11-16 19:39:21 -0800457static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
458{
459 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
460}
461
462static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
463{
464 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
465}
466
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700467static inline struct sock *sk_next(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468{
469 return sk->sk_node.next ?
470 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
471}
472
Eric Dumazet88ab1932008-11-16 19:39:21 -0800473static inline struct sock *sk_nulls_next(const struct sock *sk)
474{
475 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
476 hlist_nulls_entry(sk->sk_nulls_node.next,
477 struct sock, sk_nulls_node) :
478 NULL;
479}
480
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000481static inline bool sk_unhashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482{
483 return hlist_unhashed(&sk->sk_node);
484}
485
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000486static inline bool sk_hashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487{
Akinobu Mitada753be2006-04-28 15:21:23 -0700488 return !sk_unhashed(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489}
490
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000491static inline void sk_node_init(struct hlist_node *node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492{
493 node->pprev = NULL;
494}
495
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000496static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800497{
498 node->pprev = NULL;
499}
500
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000501static inline void __sk_del_node(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502{
503 __hlist_del(&sk->sk_node);
504}
505
stephen hemminger808f5112010-02-22 07:57:18 +0000506/* NB: equivalent to hlist_del_init_rcu */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000507static inline bool __sk_del_node_init(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508{
509 if (sk_hashed(sk)) {
510 __sk_del_node(sk);
511 sk_node_init(&sk->sk_node);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000512 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513 }
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000514 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515}
516
517/* Grab socket reference count. This operation is valid only
518 when sk is ALREADY grabbed f.e. it is found in hash table
519 or a list and the lookup is made under lock preventing hash table
520 modifications.
521 */
522
523static inline void sock_hold(struct sock *sk)
524{
525 atomic_inc(&sk->sk_refcnt);
526}
527
528/* Ungrab socket in the context, which assumes that socket refcnt
529 cannot hit zero, f.e. it is true in context of any socketcall.
530 */
531static inline void __sock_put(struct sock *sk)
532{
533 atomic_dec(&sk->sk_refcnt);
534}
535
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000536static inline bool sk_del_node_init(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537{
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000538 bool rc = __sk_del_node_init(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539
540 if (rc) {
541 /* paranoid for a while -acme */
542 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
543 __sock_put(sk);
544 }
545 return rc;
546}
stephen hemminger808f5112010-02-22 07:57:18 +0000547#define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700548
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000549static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700550{
551 if (sk_hashed(sk)) {
Eric Dumazet88ab1932008-11-16 19:39:21 -0800552 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000553 return true;
Eric Dumazet271b72c2008-10-29 02:11:14 -0700554 }
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000555 return false;
Eric Dumazet271b72c2008-10-29 02:11:14 -0700556}
557
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000558static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700559{
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000560 bool rc = __sk_nulls_del_node_init_rcu(sk);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700561
562 if (rc) {
563 /* paranoid for a while -acme */
564 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
565 __sock_put(sk);
566 }
567 return rc;
568}
569
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000570static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571{
572 hlist_add_head(&sk->sk_node, list);
573}
574
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000575static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576{
577 sock_hold(sk);
578 __sk_add_node(sk, list);
579}
580
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000581static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
stephen hemminger808f5112010-02-22 07:57:18 +0000582{
583 sock_hold(sk);
584 hlist_add_head_rcu(&sk->sk_node, list);
585}
586
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000587static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700588{
Eric Dumazet88ab1932008-11-16 19:39:21 -0800589 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700590}
591
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000592static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700593{
594 sock_hold(sk);
Eric Dumazet88ab1932008-11-16 19:39:21 -0800595 __sk_nulls_add_node_rcu(sk, list);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700596}
597
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000598static inline void __sk_del_bind_node(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700599{
600 __hlist_del(&sk->sk_bind_node);
601}
602
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000603static inline void sk_add_bind_node(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604 struct hlist_head *list)
605{
606 hlist_add_head(&sk->sk_bind_node, list);
607}
608
Sasha Levinb67bfe02013-02-27 17:06:00 -0800609#define sk_for_each(__sk, list) \
610 hlist_for_each_entry(__sk, list, sk_node)
611#define sk_for_each_rcu(__sk, list) \
612 hlist_for_each_entry_rcu(__sk, list, sk_node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800613#define sk_nulls_for_each(__sk, node, list) \
614 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
615#define sk_nulls_for_each_rcu(__sk, node, list) \
616 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
Sasha Levinb67bfe02013-02-27 17:06:00 -0800617#define sk_for_each_from(__sk) \
618 hlist_for_each_entry_from(__sk, sk_node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800619#define sk_nulls_for_each_from(__sk, node) \
620 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
621 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
Sasha Levinb67bfe02013-02-27 17:06:00 -0800622#define sk_for_each_safe(__sk, tmp, list) \
623 hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
624#define sk_for_each_bound(__sk, list) \
625 hlist_for_each_entry(__sk, list, sk_bind_node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626
Eric W. Biedermanc336d142012-05-24 17:56:43 -0600627static inline struct user_namespace *sk_user_ns(struct sock *sk)
628{
629 /* Careful only use this in a context where these parameters
630 * can not change and must all be valid, such as recvmsg from
631 * userspace.
632 */
633 return sk->sk_socket->file->f_cred->user_ns;
634}
635
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636/* Sock flags */
637enum sock_flags {
638 SOCK_DEAD,
639 SOCK_DONE,
640 SOCK_URGINLINE,
641 SOCK_KEEPOPEN,
642 SOCK_LINGER,
643 SOCK_DESTROY,
644 SOCK_BROADCAST,
645 SOCK_TIMESTAMP,
646 SOCK_ZAPPED,
647 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
648 SOCK_DBG, /* %SO_DEBUG setting */
649 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700650 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
652 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
Mel Gorman7cb02402012-07-31 16:44:16 -0700653 SOCK_MEMALLOC, /* VM depends on this socket for swapping */
Patrick Ohly20d49472009-02-12 05:03:38 +0000654 SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */
655 SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
656 SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */
657 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
658 SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */
659 SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
660 SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
Eric Dumazetbcdce712009-10-06 17:28:29 -0700661 SOCK_FASYNC, /* fasync() active */
Neil Horman3b885782009-10-12 13:26:31 -0700662 SOCK_RXQ_OVFL,
Shirley Ma1cdebb42011-07-06 12:17:30 +0000663 SOCK_ZEROCOPY, /* buffers from userspace */
Johannes Berg6e3e9392011-11-09 10:15:42 +0100664 SOCK_WIFI_STATUS, /* push wifi status to userspace */
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000665 SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
666 * Will use last 4 bytes of packet sent from
667 * user-space instead.
668 */
Vincent Bernatd59577b2013-01-16 22:55:49 +0100669 SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670};
671
Ralf Baechle53b924b2005-08-23 10:11:30 -0700672static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
673{
674 nsk->sk_flags = osk->sk_flags;
675}
676
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
678{
679 __set_bit(flag, &sk->sk_flags);
680}
681
682static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
683{
684 __clear_bit(flag, &sk->sk_flags);
685}
686
Eric Dumazet1b23a5d2012-05-16 05:57:07 +0000687static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688{
689 return test_bit(flag, &sk->sk_flags);
690}
691
Mel Gormanc93bdd02012-07-31 16:44:19 -0700692#ifdef CONFIG_NET
693extern struct static_key memalloc_socks;
694static inline int sk_memalloc_socks(void)
695{
696 return static_key_false(&memalloc_socks);
697}
698#else
699
700static inline int sk_memalloc_socks(void)
701{
702 return 0;
703}
704
705#endif
706
Mel Gorman99a1dec2012-07-31 16:44:14 -0700707static inline gfp_t sk_gfp_atomic(struct sock *sk, gfp_t gfp_mask)
708{
Mel Gorman7cb02402012-07-31 16:44:16 -0700709 return GFP_ATOMIC | (sk->sk_allocation & __GFP_MEMALLOC);
Mel Gorman99a1dec2012-07-31 16:44:14 -0700710}
711
Linus Torvalds1da177e2005-04-16 15:20:36 -0700712static inline void sk_acceptq_removed(struct sock *sk)
713{
714 sk->sk_ack_backlog--;
715}
716
717static inline void sk_acceptq_added(struct sock *sk)
718{
719 sk->sk_ack_backlog++;
720}
721
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000722static inline bool sk_acceptq_is_full(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723{
David S. Miller64a14652007-03-06 11:21:05 -0800724 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700725}
726
727/*
728 * Compute minimal free write space needed to queue new packets.
729 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000730static inline int sk_stream_min_wspace(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731{
Eric Dumazet8df09ea2007-12-21 03:07:41 -0800732 return sk->sk_wmem_queued >> 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733}
734
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000735static inline int sk_stream_wspace(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736{
737 return sk->sk_sndbuf - sk->sk_wmem_queued;
738}
739
740extern void sk_stream_write_space(struct sock *sk);
741
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000742static inline bool sk_stream_memory_free(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743{
744 return sk->sk_wmem_queued < sk->sk_sndbuf;
745}
746
Zhu Yi8eae9392010-03-04 18:01:40 +0000747/* OOB backlog add */
Zhu Yia3a858f2010-03-04 18:01:47 +0000748static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800749{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000750 /* dont let skb dst not refcounted, we are going to leave rcu lock */
751 skb_dst_force(skb);
752
753 if (!sk->sk_backlog.tail)
754 sk->sk_backlog.head = skb;
755 else
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800756 sk->sk_backlog.tail->next = skb;
Eric Dumazet7fee2262010-05-11 23:19:48 +0000757
758 sk->sk_backlog.tail = skb;
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800759 skb->next = NULL;
760}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761
Eric Dumazetc3774112010-04-27 15:13:20 -0700762/*
763 * Take into account size of receive queue and backlog queue
Eric Dumazet0fd7bac2011-12-21 07:11:44 +0000764 * Do not take into account this skb truesize,
765 * to allow even a single big packet to come.
Eric Dumazetc3774112010-04-27 15:13:20 -0700766 */
Eric Dumazetf545a382012-04-22 23:34:26 +0000767static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb,
768 unsigned int limit)
Eric Dumazetc3774112010-04-27 15:13:20 -0700769{
770 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
771
Eric Dumazetf545a382012-04-22 23:34:26 +0000772 return qsize > limit;
Eric Dumazetc3774112010-04-27 15:13:20 -0700773}
774
Zhu Yi8eae9392010-03-04 18:01:40 +0000775/* The per-socket spinlock must be held here. */
Eric Dumazetf545a382012-04-22 23:34:26 +0000776static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
777 unsigned int limit)
Zhu Yi8eae9392010-03-04 18:01:40 +0000778{
Eric Dumazetf545a382012-04-22 23:34:26 +0000779 if (sk_rcvqueues_full(sk, skb, limit))
Zhu Yi8eae9392010-03-04 18:01:40 +0000780 return -ENOBUFS;
781
Zhu Yia3a858f2010-03-04 18:01:47 +0000782 __sk_add_backlog(sk, skb);
Zhu Yi8eae9392010-03-04 18:01:40 +0000783 sk->sk_backlog.len += skb->truesize;
784 return 0;
785}
786
Mel Gormanb4b9e352012-07-31 16:44:26 -0700787extern int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
788
Peter Zijlstrac57943a2008-10-07 14:18:42 -0700789static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
790{
Mel Gormanb4b9e352012-07-31 16:44:26 -0700791 if (sk_memalloc_socks() && skb_pfmemalloc(skb))
792 return __sk_backlog_rcv(sk, skb);
793
Peter Zijlstrac57943a2008-10-07 14:18:42 -0700794 return sk->sk_backlog_rcv(sk, skb);
795}
796
David S. Millerc58dc012010-04-27 15:05:31 -0700797static inline void sock_rps_record_flow(const struct sock *sk)
798{
799#ifdef CONFIG_RPS
800 struct rps_sock_flow_table *sock_flow_table;
801
802 rcu_read_lock();
803 sock_flow_table = rcu_dereference(rps_sock_flow_table);
804 rps_record_sock_flow(sock_flow_table, sk->sk_rxhash);
805 rcu_read_unlock();
806#endif
807}
808
809static inline void sock_rps_reset_flow(const struct sock *sk)
810{
811#ifdef CONFIG_RPS
812 struct rps_sock_flow_table *sock_flow_table;
813
814 rcu_read_lock();
815 sock_flow_table = rcu_dereference(rps_sock_flow_table);
816 rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash);
817 rcu_read_unlock();
818#endif
819}
820
Tom Herbertbdeab992011-08-14 19:45:55 +0000821static inline void sock_rps_save_rxhash(struct sock *sk,
822 const struct sk_buff *skb)
David S. Millerc58dc012010-04-27 15:05:31 -0700823{
824#ifdef CONFIG_RPS
Tom Herbertbdeab992011-08-14 19:45:55 +0000825 if (unlikely(sk->sk_rxhash != skb->rxhash)) {
David S. Millerc58dc012010-04-27 15:05:31 -0700826 sock_rps_reset_flow(sk);
Tom Herbertbdeab992011-08-14 19:45:55 +0000827 sk->sk_rxhash = skb->rxhash;
David S. Millerc58dc012010-04-27 15:05:31 -0700828 }
829#endif
830}
831
Tom Herbertbdeab992011-08-14 19:45:55 +0000832static inline void sock_rps_reset_rxhash(struct sock *sk)
833{
834#ifdef CONFIG_RPS
835 sock_rps_reset_flow(sk);
836 sk->sk_rxhash = 0;
837#endif
838}
839
Stephen Hemmingercfcabdc2007-10-09 01:59:42 -0700840#define sk_wait_event(__sk, __timeo, __condition) \
841 ({ int __rc; \
842 release_sock(__sk); \
843 __rc = __condition; \
844 if (!__rc) { \
845 *(__timeo) = schedule_timeout(*(__timeo)); \
846 } \
847 lock_sock(__sk); \
848 __rc = __condition; \
849 __rc; \
850 })
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851
852extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
853extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
854extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
855extern int sk_stream_error(struct sock *sk, int flags, int err);
856extern void sk_stream_kill_queues(struct sock *sk);
Mel Gorman7cb02402012-07-31 16:44:16 -0700857extern void sk_set_memalloc(struct sock *sk);
858extern void sk_clear_memalloc(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859
860extern int sk_wait_data(struct sock *sk, long *timeo);
861
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700862struct request_sock_ops;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800863struct timewait_sock_ops;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -0800864struct inet_hashinfo;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -0700865struct raw_hashinfo;
Paul Gortmakerde477252011-05-26 13:46:22 -0400866struct module;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700867
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868/* Networking protocol blocks we attach to sockets.
869 * socket layer -> transport layer interface
870 * transport -> network interface is defined by struct inet_proto
871 */
872struct proto {
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000873 void (*close)(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874 long timeout);
875 int (*connect)(struct sock *sk,
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000876 struct sockaddr *uaddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877 int addr_len);
878 int (*disconnect)(struct sock *sk, int flags);
879
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000880 struct sock * (*accept)(struct sock *sk, int flags, int *err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881
882 int (*ioctl)(struct sock *sk, int cmd,
883 unsigned long arg);
884 int (*init)(struct sock *sk);
Brian Haley7d06b2e2008-06-14 17:04:49 -0700885 void (*destroy)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886 void (*shutdown)(struct sock *sk, int how);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000887 int (*setsockopt)(struct sock *sk, int level,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700889 unsigned int optlen);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000890 int (*getsockopt)(struct sock *sk, int level,
891 int optname, char __user *optval,
892 int __user *option);
Alexey Dobriyanaf01d532008-08-28 02:53:51 -0700893#ifdef CONFIG_COMPAT
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800894 int (*compat_setsockopt)(struct sock *sk,
895 int level,
896 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700897 unsigned int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800898 int (*compat_getsockopt)(struct sock *sk,
899 int level,
900 int optname, char __user *optval,
901 int __user *option);
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000902 int (*compat_ioctl)(struct sock *sk,
903 unsigned int cmd, unsigned long arg);
Alexey Dobriyanaf01d532008-08-28 02:53:51 -0700904#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
906 struct msghdr *msg, size_t len);
907 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
908 struct msghdr *msg,
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000909 size_t len, int noblock, int flags,
910 int *addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911 int (*sendpage)(struct sock *sk, struct page *page,
912 int offset, size_t size, int flags);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000913 int (*bind)(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914 struct sockaddr *uaddr, int addr_len);
915
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000916 int (*backlog_rcv) (struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700917 struct sk_buff *skb);
918
Eric Dumazet46d3cea2012-07-11 05:50:31 +0000919 void (*release_cb)(struct sock *sk);
Eric Dumazet563d34d2012-07-23 09:48:52 +0200920 void (*mtu_reduced)(struct sock *sk);
Eric Dumazet46d3cea2012-07-11 05:50:31 +0000921
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922 /* Keeping track of sk's, looking them up, and port selection methods. */
923 void (*hash)(struct sock *sk);
924 void (*unhash)(struct sock *sk);
Eric Dumazet719f8352010-09-08 05:08:44 +0000925 void (*rehash)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926 int (*get_port)(struct sock *sk, unsigned short snum);
Octavian Purdilafcbdf092010-12-16 14:26:56 -0800927 void (*clear_sk)(struct sock *sk, int size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700928
Eric Dumazet286ab3d2007-11-05 23:38:39 -0800929 /* Keeping track of sockets in use */
Eric Dumazet65f76512008-01-03 20:46:48 -0800930#ifdef CONFIG_PROC_FS
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -0700931 unsigned int inuse_idx;
Eric Dumazet65f76512008-01-03 20:46:48 -0800932#endif
Arnaldo Carvalho de Meloebb53d72007-11-21 22:08:50 +0800933
Linus Torvalds1da177e2005-04-16 15:20:36 -0700934 /* Memory pressure */
Pavel Emelyanov5c52ba12008-07-16 20:28:10 -0700935 void (*enter_memory_pressure)(struct sock *sk);
Eric Dumazet8d987e52010-11-09 23:24:26 +0000936 atomic_long_t *memory_allocated; /* Current allocated memory. */
Eric Dumazet17483762008-11-25 21:16:35 -0800937 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938 /*
939 * Pressure flag: try to collapse.
940 * Technical note: it is used by multiple contexts non atomically.
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800941 * All the __sk_mem_schedule() is of this nature: accounting
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942 * is strict, actions are advisory and have some latency.
943 */
944 int *memory_pressure;
Eric Dumazet8d987e52010-11-09 23:24:26 +0000945 long *sysctl_mem;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946 int *sysctl_wmem;
947 int *sysctl_rmem;
948 int max_header;
Changli Gao7ba42912010-07-10 20:41:55 +0000949 bool no_autobind;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700950
Eric Dumazet271b72c2008-10-29 02:11:14 -0700951 struct kmem_cache *slab;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700952 unsigned int obj_size;
Eric Dumazet271b72c2008-10-29 02:11:14 -0700953 int slab_flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700954
Eric Dumazetdd24c002008-11-25 21:17:14 -0800955 struct percpu_counter *orphan_count;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700956
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700957 struct request_sock_ops *rsk_prot;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800958 struct timewait_sock_ops *twsk_prot;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700959
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -0700960 union {
961 struct inet_hashinfo *hashinfo;
Eric Dumazet645ca702008-10-29 01:41:45 -0700962 struct udp_table *udp_table;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -0700963 struct raw_hashinfo *raw_hash;
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -0700964 } h;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -0800965
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966 struct module *owner;
967
968 char name[32];
969
970 struct list_head node;
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700971#ifdef SOCK_REFCNT_DEBUG
972 atomic_t socks;
973#endif
Andrew Mortonc255a452012-07-31 16:43:02 -0700974#ifdef CONFIG_MEMCG_KMEM
Glauber Costae1aab162011-12-11 21:47:03 +0000975 /*
976 * cgroup specific init/deinit functions. Called once for all
977 * protocols that implement it, from cgroups populate function.
978 * This function has to setup any files the protocol want to
979 * appear in the kmem cgroup filesystem.
980 */
Glauber Costa1d62e432012-04-09 19:36:33 -0300981 int (*init_cgroup)(struct mem_cgroup *memcg,
Glauber Costae1aab162011-12-11 21:47:03 +0000982 struct cgroup_subsys *ss);
Glauber Costa1d62e432012-04-09 19:36:33 -0300983 void (*destroy_cgroup)(struct mem_cgroup *memcg);
Glauber Costae1aab162011-12-11 21:47:03 +0000984 struct cg_proto *(*proto_cgroup)(struct mem_cgroup *memcg);
985#endif
986};
987
Glauber Costa3f134612012-05-29 15:07:11 -0700988/*
989 * Bits in struct cg_proto.flags
990 */
991enum cg_proto_flags {
992 /* Currently active and new sockets should be assigned to cgroups */
993 MEMCG_SOCK_ACTIVE,
994 /* It was ever activated; we must disarm static keys on destruction */
995 MEMCG_SOCK_ACTIVATED,
996};
997
Glauber Costae1aab162011-12-11 21:47:03 +0000998struct cg_proto {
999 void (*enter_memory_pressure)(struct sock *sk);
1000 struct res_counter *memory_allocated; /* Current allocated memory. */
1001 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
1002 int *memory_pressure;
1003 long *sysctl_mem;
Glauber Costa3f134612012-05-29 15:07:11 -07001004 unsigned long flags;
Glauber Costae1aab162011-12-11 21:47:03 +00001005 /*
1006 * memcg field is used to find which memcg we belong directly
1007 * Each memcg struct can hold more than one cg_proto, so container_of
1008 * won't really cut.
1009 *
1010 * The elegant solution would be having an inverse function to
1011 * proto_cgroup in struct proto, but that means polluting the structure
1012 * for everybody, instead of just for memcg users.
1013 */
1014 struct mem_cgroup *memcg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015};
1016
1017extern int proto_register(struct proto *prot, int alloc_slab);
1018extern void proto_unregister(struct proto *prot);
1019
Glauber Costa3f134612012-05-29 15:07:11 -07001020static inline bool memcg_proto_active(struct cg_proto *cg_proto)
1021{
1022 return test_bit(MEMCG_SOCK_ACTIVE, &cg_proto->flags);
1023}
1024
1025static inline bool memcg_proto_activated(struct cg_proto *cg_proto)
1026{
1027 return test_bit(MEMCG_SOCK_ACTIVATED, &cg_proto->flags);
1028}
1029
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -07001030#ifdef SOCK_REFCNT_DEBUG
1031static inline void sk_refcnt_debug_inc(struct sock *sk)
1032{
1033 atomic_inc(&sk->sk_prot->socks);
1034}
1035
1036static inline void sk_refcnt_debug_dec(struct sock *sk)
1037{
1038 atomic_dec(&sk->sk_prot->socks);
1039 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
1040 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
1041}
1042
Ying Xuedec34fb2013-02-15 22:28:25 +00001043static inline void sk_refcnt_debug_release(const struct sock *sk)
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -07001044{
1045 if (atomic_read(&sk->sk_refcnt) != 1)
1046 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
1047 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
1048}
1049#else /* SOCK_REFCNT_DEBUG */
1050#define sk_refcnt_debug_inc(sk) do { } while (0)
1051#define sk_refcnt_debug_dec(sk) do { } while (0)
1052#define sk_refcnt_debug_release(sk) do { } while (0)
1053#endif /* SOCK_REFCNT_DEBUG */
1054
Andrew Mortonc255a452012-07-31 16:43:02 -07001055#if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_NET)
Ingo Molnarc5905af2012-02-24 08:31:31 +01001056extern struct static_key memcg_socket_limit_enabled;
Glauber Costae1aab162011-12-11 21:47:03 +00001057static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1058 struct cg_proto *cg_proto)
1059{
1060 return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg));
1061}
Ingo Molnarc5905af2012-02-24 08:31:31 +01001062#define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
Glauber Costae1aab162011-12-11 21:47:03 +00001063#else
1064#define mem_cgroup_sockets_enabled 0
1065static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1066 struct cg_proto *cg_proto)
1067{
1068 return NULL;
1069}
1070#endif
1071
1072
Glauber Costa180d8cd2011-12-11 21:47:02 +00001073static inline bool sk_has_memory_pressure(const struct sock *sk)
1074{
1075 return sk->sk_prot->memory_pressure != NULL;
1076}
1077
1078static inline bool sk_under_memory_pressure(const struct sock *sk)
1079{
1080 if (!sk->sk_prot->memory_pressure)
1081 return false;
Glauber Costae1aab162011-12-11 21:47:03 +00001082
1083 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1084 return !!*sk->sk_cgrp->memory_pressure;
1085
Glauber Costa180d8cd2011-12-11 21:47:02 +00001086 return !!*sk->sk_prot->memory_pressure;
1087}
1088
1089static inline void sk_leave_memory_pressure(struct sock *sk)
1090{
1091 int *memory_pressure = sk->sk_prot->memory_pressure;
1092
Glauber Costae1aab162011-12-11 21:47:03 +00001093 if (!memory_pressure)
1094 return;
1095
1096 if (*memory_pressure)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001097 *memory_pressure = 0;
Glauber Costae1aab162011-12-11 21:47:03 +00001098
1099 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1100 struct cg_proto *cg_proto = sk->sk_cgrp;
1101 struct proto *prot = sk->sk_prot;
1102
1103 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1104 if (*cg_proto->memory_pressure)
1105 *cg_proto->memory_pressure = 0;
1106 }
1107
Glauber Costa180d8cd2011-12-11 21:47:02 +00001108}
1109
1110static inline void sk_enter_memory_pressure(struct sock *sk)
1111{
Glauber Costae1aab162011-12-11 21:47:03 +00001112 if (!sk->sk_prot->enter_memory_pressure)
1113 return;
1114
1115 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1116 struct cg_proto *cg_proto = sk->sk_cgrp;
1117 struct proto *prot = sk->sk_prot;
1118
1119 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1120 cg_proto->enter_memory_pressure(sk);
1121 }
1122
1123 sk->sk_prot->enter_memory_pressure(sk);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001124}
1125
1126static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1127{
1128 long *prot = sk->sk_prot->sysctl_mem;
Glauber Costae1aab162011-12-11 21:47:03 +00001129 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1130 prot = sk->sk_cgrp->sysctl_mem;
Glauber Costa180d8cd2011-12-11 21:47:02 +00001131 return prot[index];
1132}
1133
Glauber Costae1aab162011-12-11 21:47:03 +00001134static inline void memcg_memory_allocated_add(struct cg_proto *prot,
1135 unsigned long amt,
1136 int *parent_status)
1137{
1138 struct res_counter *fail;
1139 int ret;
1140
Glauber Costa0e90b312012-01-20 04:57:16 +00001141 ret = res_counter_charge_nofail(prot->memory_allocated,
1142 amt << PAGE_SHIFT, &fail);
Glauber Costae1aab162011-12-11 21:47:03 +00001143 if (ret < 0)
1144 *parent_status = OVER_LIMIT;
1145}
1146
1147static inline void memcg_memory_allocated_sub(struct cg_proto *prot,
1148 unsigned long amt)
1149{
1150 res_counter_uncharge(prot->memory_allocated, amt << PAGE_SHIFT);
1151}
1152
1153static inline u64 memcg_memory_allocated_read(struct cg_proto *prot)
1154{
1155 u64 ret;
1156 ret = res_counter_read_u64(prot->memory_allocated, RES_USAGE);
1157 return ret >> PAGE_SHIFT;
1158}
1159
Glauber Costa180d8cd2011-12-11 21:47:02 +00001160static inline long
1161sk_memory_allocated(const struct sock *sk)
1162{
1163 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001164 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1165 return memcg_memory_allocated_read(sk->sk_cgrp);
1166
Glauber Costa180d8cd2011-12-11 21:47:02 +00001167 return atomic_long_read(prot->memory_allocated);
1168}
1169
1170static inline long
Glauber Costae1aab162011-12-11 21:47:03 +00001171sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001172{
1173 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001174
1175 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1176 memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status);
1177 /* update the root cgroup regardless */
1178 atomic_long_add_return(amt, prot->memory_allocated);
1179 return memcg_memory_allocated_read(sk->sk_cgrp);
1180 }
1181
Glauber Costa180d8cd2011-12-11 21:47:02 +00001182 return atomic_long_add_return(amt, prot->memory_allocated);
1183}
1184
1185static inline void
Glauber Costa0e90b312012-01-20 04:57:16 +00001186sk_memory_allocated_sub(struct sock *sk, int amt)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001187{
1188 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001189
Glauber Costa0e90b312012-01-20 04:57:16 +00001190 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
Glauber Costae1aab162011-12-11 21:47:03 +00001191 memcg_memory_allocated_sub(sk->sk_cgrp, amt);
1192
Glauber Costa180d8cd2011-12-11 21:47:02 +00001193 atomic_long_sub(amt, prot->memory_allocated);
1194}
1195
1196static inline void sk_sockets_allocated_dec(struct sock *sk)
1197{
1198 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001199
1200 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1201 struct cg_proto *cg_proto = sk->sk_cgrp;
1202
1203 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1204 percpu_counter_dec(cg_proto->sockets_allocated);
1205 }
1206
Glauber Costa180d8cd2011-12-11 21:47:02 +00001207 percpu_counter_dec(prot->sockets_allocated);
1208}
1209
1210static inline void sk_sockets_allocated_inc(struct sock *sk)
1211{
1212 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001213
1214 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1215 struct cg_proto *cg_proto = sk->sk_cgrp;
1216
1217 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1218 percpu_counter_inc(cg_proto->sockets_allocated);
1219 }
1220
Glauber Costa180d8cd2011-12-11 21:47:02 +00001221 percpu_counter_inc(prot->sockets_allocated);
1222}
1223
1224static inline int
1225sk_sockets_allocated_read_positive(struct sock *sk)
1226{
1227 struct proto *prot = sk->sk_prot;
1228
Glauber Costae1aab162011-12-11 21:47:03 +00001229 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
Eric Dumazet518fbf92012-04-28 23:21:56 +00001230 return percpu_counter_read_positive(sk->sk_cgrp->sockets_allocated);
Glauber Costae1aab162011-12-11 21:47:03 +00001231
Eric Dumazet518fbf92012-04-28 23:21:56 +00001232 return percpu_counter_read_positive(prot->sockets_allocated);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001233}
1234
1235static inline int
1236proto_sockets_allocated_sum_positive(struct proto *prot)
1237{
1238 return percpu_counter_sum_positive(prot->sockets_allocated);
1239}
1240
1241static inline long
1242proto_memory_allocated(struct proto *prot)
1243{
1244 return atomic_long_read(prot->memory_allocated);
1245}
1246
1247static inline bool
1248proto_memory_pressure(struct proto *prot)
1249{
1250 if (!prot->memory_pressure)
1251 return false;
1252 return !!*prot->memory_pressure;
1253}
1254
Eric Dumazet65f76512008-01-03 20:46:48 -08001255
1256#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001257/* Called with local bh disabled */
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -07001258extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1259extern int sock_prot_inuse_get(struct net *net, struct proto *proto);
Eric Dumazet65f76512008-01-03 20:46:48 -08001260#else
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001261static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -07001262 int inc)
Eric Dumazet65f76512008-01-03 20:46:48 -08001263{
1264}
Eric Dumazet65f76512008-01-03 20:46:48 -08001265#endif
1266
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267
Arnaldo Carvalho de Melo614c6cb2005-08-09 19:47:37 -07001268/* With per-bucket locks this operation is not-atomic, so that
1269 * this version is not worse.
1270 */
1271static inline void __sk_prot_rehash(struct sock *sk)
1272{
1273 sk->sk_prot->unhash(sk);
1274 sk->sk_prot->hash(sk);
1275}
1276
Octavian Purdilafcbdf092010-12-16 14:26:56 -08001277void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
1278
Linus Torvalds1da177e2005-04-16 15:20:36 -07001279/* About 10 seconds */
1280#define SOCK_DESTROY_TIME (10*HZ)
1281
1282/* Sockets 0-1023 can't be bound to unless you are superuser */
1283#define PROT_SOCK 1024
1284
1285#define SHUTDOWN_MASK 3
1286#define RCV_SHUTDOWN 1
1287#define SEND_SHUTDOWN 2
1288
1289#define SOCK_SNDBUF_LOCK 1
1290#define SOCK_RCVBUF_LOCK 2
1291#define SOCK_BINDADDR_LOCK 4
1292#define SOCK_BINDPORT_LOCK 8
1293
1294/* sock_iocb: used to kick off async processing of socket ios */
1295struct sock_iocb {
1296 struct list_head list;
1297
1298 int flags;
1299 int size;
1300 struct socket *sock;
1301 struct sock *sk;
1302 struct scm_cookie *scm;
1303 struct msghdr *msg, async_msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001304 struct kiocb *kiocb;
1305};
1306
1307static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
1308{
1309 return (struct sock_iocb *)iocb->private;
1310}
1311
1312static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
1313{
1314 return si->kiocb;
1315}
1316
1317struct socket_alloc {
1318 struct socket socket;
1319 struct inode vfs_inode;
1320};
1321
1322static inline struct socket *SOCKET_I(struct inode *inode)
1323{
1324 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1325}
1326
1327static inline struct inode *SOCK_INODE(struct socket *socket)
1328{
1329 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1330}
1331
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001332/*
1333 * Functions for memory accounting
1334 */
1335extern int __sk_mem_schedule(struct sock *sk, int size, int kind);
1336extern void __sk_mem_reclaim(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001338#define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1339#define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1340#define SK_MEM_SEND 0
1341#define SK_MEM_RECV 1
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001343static inline int sk_mem_pages(int amt)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001345 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001346}
1347
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001348static inline bool sk_has_account(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001349{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001350 /* return true if protocol supports memory accounting */
1351 return !!sk->sk_prot->memory_allocated;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352}
1353
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001354static inline bool sk_wmem_schedule(struct sock *sk, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001355{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001356 if (!sk_has_account(sk))
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001357 return true;
Herbert Xud80d99d2005-09-01 17:48:23 -07001358 return size <= sk->sk_forward_alloc ||
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001359 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1360}
1361
Mel Gormanc76562b2012-07-31 16:44:41 -07001362static inline bool
Chuck Lever35c448a2012-09-17 14:09:11 -07001363sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size)
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001364{
1365 if (!sk_has_account(sk))
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001366 return true;
Mel Gormanc76562b2012-07-31 16:44:41 -07001367 return size<= sk->sk_forward_alloc ||
1368 __sk_mem_schedule(sk, size, SK_MEM_RECV) ||
1369 skb_pfmemalloc(skb);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001370}
1371
1372static inline void sk_mem_reclaim(struct sock *sk)
1373{
1374 if (!sk_has_account(sk))
1375 return;
1376 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1377 __sk_mem_reclaim(sk);
1378}
1379
David S. Miller9993e7d2008-01-10 21:56:38 -08001380static inline void sk_mem_reclaim_partial(struct sock *sk)
1381{
1382 if (!sk_has_account(sk))
1383 return;
1384 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1385 __sk_mem_reclaim(sk);
1386}
1387
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001388static inline void sk_mem_charge(struct sock *sk, int size)
1389{
1390 if (!sk_has_account(sk))
1391 return;
1392 sk->sk_forward_alloc -= size;
1393}
1394
1395static inline void sk_mem_uncharge(struct sock *sk, int size)
1396{
1397 if (!sk_has_account(sk))
1398 return;
1399 sk->sk_forward_alloc += size;
1400}
1401
1402static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1403{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001404 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1405 sk->sk_wmem_queued -= skb->truesize;
1406 sk_mem_uncharge(sk, skb->truesize);
1407 __kfree_skb(skb);
Herbert Xud80d99d2005-09-01 17:48:23 -07001408}
1409
Linus Torvalds1da177e2005-04-16 15:20:36 -07001410/* Used by processes to "lock" a socket state, so that
1411 * interrupts and bottom half handlers won't change it
1412 * from under us. It essentially blocks any incoming
1413 * packets, so that we won't get any new data or any
1414 * packets that change the state of the socket.
1415 *
1416 * While locked, BH processing will add new packets to
1417 * the backlog queue. This queue is processed by the
1418 * owner of the socket lock right before it is released.
1419 *
1420 * Since ~2.3.5 it is also exclusive sleep lock serializing
1421 * accesses from user process context.
1422 */
John Heffnerd2e91172007-09-12 10:44:19 +02001423#define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001424
Peter Zijlstraed075362006-12-06 20:35:24 -08001425/*
1426 * Macro so as to not evaluate some arguments when
1427 * lockdep is not enabled.
1428 *
1429 * Mark both the sk_lock and the sk_lock.slock as a
1430 * per-address-family lock class.
1431 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001432#define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
Peter Zijlstraed075362006-12-06 20:35:24 -08001433do { \
Ingo Molnare8f6fbf2008-11-12 01:38:36 +00001434 sk->sk_lock.owned = 0; \
Peter Zijlstraed075362006-12-06 20:35:24 -08001435 init_waitqueue_head(&sk->sk_lock.wq); \
1436 spin_lock_init(&(sk)->sk_lock.slock); \
1437 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1438 sizeof((sk)->sk_lock)); \
1439 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001440 (skey), (sname)); \
Peter Zijlstraed075362006-12-06 20:35:24 -08001441 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1442} while (0)
1443
Harvey Harrison41380932007-12-12 10:46:51 -08001444extern void lock_sock_nested(struct sock *sk, int subclass);
Peter Zijlstrafcc70d52006-11-08 22:44:35 -08001445
1446static inline void lock_sock(struct sock *sk)
1447{
1448 lock_sock_nested(sk, 0);
1449}
1450
Harvey Harrison41380932007-12-12 10:46:51 -08001451extern void release_sock(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001452
1453/* BH context may only use the following locking interface. */
1454#define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
Ingo Molnarc6366182006-07-03 00:25:13 -07001455#define bh_lock_sock_nested(__sk) \
1456 spin_lock_nested(&((__sk)->sk_lock.slock), \
1457 SINGLE_DEPTH_NESTING)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001458#define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1459
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001460extern bool lock_sock_fast(struct sock *sk);
1461/**
1462 * unlock_sock_fast - complement of lock_sock_fast
1463 * @sk: socket
1464 * @slow: slow mode
1465 *
1466 * fast unlock socket for user context.
1467 * If slow mode is on, we call regular release_sock()
1468 */
1469static inline void unlock_sock_fast(struct sock *sk, bool slow)
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001470{
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001471 if (slow)
1472 release_sock(sk);
1473 else
1474 spin_unlock_bh(&sk->sk_lock.slock);
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001475}
1476
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001477
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001478extern struct sock *sk_alloc(struct net *net, int family,
Al Virodd0fc662005-10-07 07:46:04 +01001479 gfp_t priority,
Pavel Emelyanov6257ff22007-11-01 00:39:31 -07001480 struct proto *prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481extern void sk_free(struct sock *sk);
Denis V. Lunevedf02082008-02-29 11:18:32 -08001482extern void sk_release_kernel(struct sock *sk);
Eric Dumazete56c57d2011-11-08 17:07:07 -05001483extern struct sock *sk_clone_lock(const struct sock *sk,
1484 const gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001485
1486extern struct sk_buff *sock_wmalloc(struct sock *sk,
1487 unsigned long size, int force,
Al Virodd0fc662005-10-07 07:46:04 +01001488 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001489extern struct sk_buff *sock_rmalloc(struct sock *sk,
1490 unsigned long size, int force,
Al Virodd0fc662005-10-07 07:46:04 +01001491 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001492extern void sock_wfree(struct sk_buff *skb);
1493extern void sock_rfree(struct sk_buff *skb);
David S. Miller41063e92012-06-19 21:22:05 -07001494extern void sock_edemux(struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001495
1496extern int sock_setsockopt(struct socket *sock, int level,
1497 int op, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07001498 unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499
1500extern int sock_getsockopt(struct socket *sock, int level,
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001501 int op, char __user *optval,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001502 int __user *optlen);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001503extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001504 unsigned long size,
1505 int noblock,
1506 int *errcode);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001507extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
Herbert Xu4cc7f682009-02-04 16:55:54 -08001508 unsigned long header_len,
1509 unsigned long data_len,
1510 int noblock,
1511 int *errcode);
Victor Fusco86a76ca2005-07-08 14:57:47 -07001512extern void *sock_kmalloc(struct sock *sk, int size,
Al Virodd0fc662005-10-07 07:46:04 +01001513 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001514extern void sock_kfree_s(struct sock *sk, void *mem, int size);
1515extern void sk_send_sigurg(struct sock *sk);
1516
1517/*
1518 * Functions to fill in entries in struct proto_ops when a protocol
1519 * does not implement a particular function.
1520 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001521extern int sock_no_bind(struct socket *,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001522 struct sockaddr *, int);
1523extern int sock_no_connect(struct socket *,
1524 struct sockaddr *, int, int);
1525extern int sock_no_socketpair(struct socket *,
1526 struct socket *);
1527extern int sock_no_accept(struct socket *,
1528 struct socket *, int);
1529extern int sock_no_getname(struct socket *,
1530 struct sockaddr *, int *, int);
1531extern unsigned int sock_no_poll(struct file *, struct socket *,
1532 struct poll_table_struct *);
1533extern int sock_no_ioctl(struct socket *, unsigned int,
1534 unsigned long);
1535extern int sock_no_listen(struct socket *, int);
1536extern int sock_no_shutdown(struct socket *, int);
1537extern int sock_no_getsockopt(struct socket *, int , int,
1538 char __user *, int __user *);
1539extern int sock_no_setsockopt(struct socket *, int, int,
David S. Millerb7058842009-09-30 16:12:20 -07001540 char __user *, unsigned int);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001541extern int sock_no_sendmsg(struct kiocb *, struct socket *,
1542 struct msghdr *, size_t);
1543extern int sock_no_recvmsg(struct kiocb *, struct socket *,
1544 struct msghdr *, size_t, int);
1545extern int sock_no_mmap(struct file *file,
1546 struct socket *sock,
1547 struct vm_area_struct *vma);
1548extern ssize_t sock_no_sendpage(struct socket *sock,
1549 struct page *page,
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001550 int offset, size_t size,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001551 int flags);
1552
1553/*
1554 * Functions to fill in entries in struct proto_ops when a protocol
1555 * uses the inet style.
1556 */
1557extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
1558 char __user *optval, int __user *optlen);
1559extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1560 struct msghdr *msg, size_t size, int flags);
1561extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07001562 char __user *optval, unsigned int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001563extern int compat_sock_common_getsockopt(struct socket *sock, int level,
1564 int optname, char __user *optval, int __user *optlen);
1565extern int compat_sock_common_setsockopt(struct socket *sock, int level,
David S. Millerb7058842009-09-30 16:12:20 -07001566 int optname, char __user *optval, unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567
1568extern void sk_common_release(struct sock *sk);
1569
1570/*
1571 * Default socket callbacks and setup code
1572 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001573
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574/* Initialise core socket variables */
1575extern void sock_init_data(struct socket *sock, struct sock *sk);
1576
Eric Dumazet46bcf142010-12-06 09:29:43 -08001577extern void sk_filter_release_rcu(struct rcu_head *rcu);
1578
Linus Torvalds1da177e2005-04-16 15:20:36 -07001579/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001580 * sk_filter_release - release a socket filter
Paul Bonserdc9b3342006-11-23 17:56:13 -08001581 * @fp: filter to remove
1582 *
1583 * Remove a filter from a socket and release its resources.
1584 */
1585
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -07001586static inline void sk_filter_release(struct sk_filter *fp)
1587{
1588 if (atomic_dec_and_test(&fp->refcnt))
Eric Dumazet80f8f102011-01-18 07:46:52 +00001589 call_rcu(&fp->rcu, sk_filter_release_rcu);
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -07001590}
1591
1592static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001593{
1594 unsigned int size = sk_filter_len(fp);
1595
1596 atomic_sub(size, &sk->sk_omem_alloc);
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -07001597 sk_filter_release(fp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001598}
1599
1600static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
1601{
1602 atomic_inc(&fp->refcnt);
1603 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
1604}
1605
1606/*
1607 * Socket reference counting postulates.
1608 *
1609 * * Each user of socket SHOULD hold a reference count.
1610 * * Each access point to socket (an hash table bucket, reference from a list,
1611 * running timer, skb in flight MUST hold a reference count.
1612 * * When reference count hits 0, it means it will never increase back.
1613 * * When reference count hits 0, it means that no references from
1614 * outside exist to this socket and current process on current CPU
1615 * is last user and may/should destroy this socket.
1616 * * sk_free is called from any context: process, BH, IRQ. When
1617 * it is called, socket has no references from outside -> sk_free
1618 * may release descendant resources allocated by the socket, but
1619 * to the time when it is called, socket is NOT referenced by any
1620 * hash tables, lists etc.
1621 * * Packets, delivered from outside (from network or from another process)
1622 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1623 * when they sit in queue. Otherwise, packets will leak to hole, when
1624 * socket is looked up by one cpu and unhasing is made by another CPU.
1625 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1626 * (leak to backlog). Packet socket does all the processing inside
1627 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1628 * use separate SMP lock, so that they are prone too.
1629 */
1630
1631/* Ungrab socket and destroy it, if it was the last reference. */
1632static inline void sock_put(struct sock *sk)
1633{
1634 if (atomic_dec_and_test(&sk->sk_refcnt))
1635 sk_free(sk);
1636}
1637
Arnaldo Carvalho de Melo58a5a7b2006-11-16 14:06:06 -02001638extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1639 const int nested);
Arnaldo Carvalho de Melo25995ff2005-12-27 02:42:22 -02001640
Krishna Kumare022f0b2009-10-19 23:46:20 +00001641static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1642{
1643 sk->sk_tx_queue_mapping = tx_queue;
1644}
1645
1646static inline void sk_tx_queue_clear(struct sock *sk)
1647{
1648 sk->sk_tx_queue_mapping = -1;
1649}
1650
1651static inline int sk_tx_queue_get(const struct sock *sk)
1652{
Tom Herbertb0f77d02010-07-14 20:50:29 -07001653 return sk ? sk->sk_tx_queue_mapping : -1;
Krishna Kumare022f0b2009-10-19 23:46:20 +00001654}
1655
David S. Miller972692e2008-06-17 22:41:38 -07001656static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1657{
Krishna Kumare022f0b2009-10-19 23:46:20 +00001658 sk_tx_queue_clear(sk);
David S. Miller972692e2008-06-17 22:41:38 -07001659 sk->sk_socket = sock;
1660}
1661
Eric Dumazetaa395142010-04-20 13:03:51 +00001662static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1663{
Eric Dumazeteaefd112011-02-18 03:26:36 +00001664 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1665 return &rcu_dereference_raw(sk->sk_wq)->wait;
Eric Dumazetaa395142010-04-20 13:03:51 +00001666}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001667/* Detach socket from process context.
1668 * Announce socket dead, detach it from wait queue and inode.
1669 * Note that parent inode held reference count on this struct sock,
1670 * we do not release it in this function, because protocol
1671 * probably wants some additional cleanups or even continuing
1672 * to work with this socket (TCP).
1673 */
1674static inline void sock_orphan(struct sock *sk)
1675{
1676 write_lock_bh(&sk->sk_callback_lock);
1677 sock_set_flag(sk, SOCK_DEAD);
David S. Miller972692e2008-06-17 22:41:38 -07001678 sk_set_socket(sk, NULL);
Eric Dumazet43815482010-04-29 11:01:49 +00001679 sk->sk_wq = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001680 write_unlock_bh(&sk->sk_callback_lock);
1681}
1682
1683static inline void sock_graft(struct sock *sk, struct socket *parent)
1684{
1685 write_lock_bh(&sk->sk_callback_lock);
Eric Dumazeteaefd112011-02-18 03:26:36 +00001686 sk->sk_wq = parent->wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687 parent->sk = sk;
David S. Miller972692e2008-06-17 22:41:38 -07001688 sk_set_socket(sk, parent);
Venkat Yekkirala4237c752006-07-24 23:32:50 -07001689 security_sock_graft(sk, parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001690 write_unlock_bh(&sk->sk_callback_lock);
1691}
1692
Eric W. Biederman976d02012012-05-23 17:16:53 -06001693extern kuid_t sock_i_uid(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001694extern unsigned long sock_i_ino(struct sock *sk);
1695
1696static inline struct dst_entry *
1697__sk_dst_get(struct sock *sk)
1698{
Michal Hockod8bf4ca2011-07-08 14:39:41 +02001699 return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
Eric Dumazetf68c2242010-04-22 16:06:59 -07001700 lockdep_is_held(&sk->sk_lock.slock));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001701}
1702
1703static inline struct dst_entry *
1704sk_dst_get(struct sock *sk)
1705{
1706 struct dst_entry *dst;
1707
Eric Dumazetb6c67122010-04-08 23:03:29 +00001708 rcu_read_lock();
1709 dst = rcu_dereference(sk->sk_dst_cache);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710 if (dst)
1711 dst_hold(dst);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001712 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001713 return dst;
1714}
1715
Eric Dumazetb6c67122010-04-08 23:03:29 +00001716extern void sk_reset_txq(struct sock *sk);
1717
1718static inline void dst_negative_advice(struct sock *sk)
1719{
1720 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1721
1722 if (dst && dst->ops->negative_advice) {
1723 ndst = dst->ops->negative_advice(dst);
1724
1725 if (ndst != dst) {
1726 rcu_assign_pointer(sk->sk_dst_cache, ndst);
1727 sk_reset_txq(sk);
1728 }
1729 }
1730}
1731
Linus Torvalds1da177e2005-04-16 15:20:36 -07001732static inline void
1733__sk_dst_set(struct sock *sk, struct dst_entry *dst)
1734{
1735 struct dst_entry *old_dst;
1736
Krishna Kumare022f0b2009-10-19 23:46:20 +00001737 sk_tx_queue_clear(sk);
Eric Dumazet0b53ff22010-04-26 20:40:43 +00001738 /*
1739 * This can be called while sk is owned by the caller only,
1740 * with no state that can be checked in a rcu_dereference_check() cond
1741 */
1742 old_dst = rcu_dereference_raw(sk->sk_dst_cache);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001743 rcu_assign_pointer(sk->sk_dst_cache, dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744 dst_release(old_dst);
1745}
1746
1747static inline void
1748sk_dst_set(struct sock *sk, struct dst_entry *dst)
1749{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001750 spin_lock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751 __sk_dst_set(sk, dst);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001752 spin_unlock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001753}
1754
1755static inline void
1756__sk_dst_reset(struct sock *sk)
1757{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001758 __sk_dst_set(sk, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001759}
1760
1761static inline void
1762sk_dst_reset(struct sock *sk)
1763{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001764 spin_lock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001765 __sk_dst_reset(sk);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001766 spin_unlock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001767}
1768
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001769extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001770
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001771extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001772
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001773static inline bool sk_can_gso(const struct sock *sk)
Herbert Xubcd76112006-06-30 13:36:35 -07001774{
1775 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1776}
1777
Andi Kleen99580892007-04-20 17:12:43 -07001778extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
Arnaldo Carvalho de Melo6cbb0df2005-08-09 19:49:02 -07001779
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001780static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
Eric Dumazeta4654192010-05-16 00:36:33 -07001781{
1782 sk->sk_route_nocaps |= flags;
1783 sk->sk_route_caps &= ~flags;
1784}
1785
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001786static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1787 char __user *from, char *to,
Wei Yongjun912d3982011-04-06 18:40:12 +00001788 int copy, int offset)
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001789{
1790 if (skb->ip_summed == CHECKSUM_NONE) {
1791 int err = 0;
1792 __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err);
1793 if (err)
1794 return err;
Wei Yongjun912d3982011-04-06 18:40:12 +00001795 skb->csum = csum_block_add(skb->csum, csum, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001796 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1797 if (!access_ok(VERIFY_READ, from, copy) ||
1798 __copy_from_user_nocache(to, from, copy))
1799 return -EFAULT;
1800 } else if (copy_from_user(to, from, copy))
1801 return -EFAULT;
1802
1803 return 0;
1804}
1805
1806static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
1807 char __user *from, int copy)
1808{
Wei Yongjun912d3982011-04-06 18:40:12 +00001809 int err, offset = skb->len;
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001810
Wei Yongjun912d3982011-04-06 18:40:12 +00001811 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1812 copy, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001813 if (err)
Wei Yongjun912d3982011-04-06 18:40:12 +00001814 __skb_trim(skb, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001815
1816 return err;
1817}
1818
1819static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from,
1820 struct sk_buff *skb,
1821 struct page *page,
1822 int off, int copy)
1823{
1824 int err;
1825
Wei Yongjun912d3982011-04-06 18:40:12 +00001826 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1827 copy, skb->len);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001828 if (err)
1829 return err;
1830
1831 skb->len += copy;
1832 skb->data_len += copy;
1833 skb->truesize += copy;
1834 sk->sk_wmem_queued += copy;
1835 sk_mem_charge(sk, copy);
1836 return 0;
1837}
1838
Linus Torvalds1da177e2005-04-16 15:20:36 -07001839static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1840 struct sk_buff *skb, struct page *page,
1841 int off, int copy)
1842{
1843 if (skb->ip_summed == CHECKSUM_NONE) {
1844 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08001845 __wsum csum = csum_and_copy_from_user(from,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001846 page_address(page) + off,
1847 copy, 0, &err);
1848 if (err)
1849 return err;
1850 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1851 } else if (copy_from_user(page_address(page) + off, from, copy))
1852 return -EFAULT;
1853
1854 skb->len += copy;
1855 skb->data_len += copy;
1856 skb->truesize += copy;
1857 sk->sk_wmem_queued += copy;
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001858 sk_mem_charge(sk, copy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001859 return 0;
1860}
1861
Eric Dumazetc5640392009-06-16 10:12:03 +00001862/**
1863 * sk_wmem_alloc_get - returns write allocations
1864 * @sk: socket
1865 *
1866 * Returns sk_wmem_alloc minus initial offset of one
1867 */
1868static inline int sk_wmem_alloc_get(const struct sock *sk)
1869{
1870 return atomic_read(&sk->sk_wmem_alloc) - 1;
1871}
1872
1873/**
1874 * sk_rmem_alloc_get - returns read allocations
1875 * @sk: socket
1876 *
1877 * Returns sk_rmem_alloc
1878 */
1879static inline int sk_rmem_alloc_get(const struct sock *sk)
1880{
1881 return atomic_read(&sk->sk_rmem_alloc);
1882}
1883
1884/**
1885 * sk_has_allocations - check if allocations are outstanding
1886 * @sk: socket
1887 *
1888 * Returns true if socket has write or read allocations
1889 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001890static inline bool sk_has_allocations(const struct sock *sk)
Eric Dumazetc5640392009-06-16 10:12:03 +00001891{
1892 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1893}
1894
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001895/**
Eric Dumazet43815482010-04-29 11:01:49 +00001896 * wq_has_sleeper - check if there are any waiting processes
Randy Dunlapacfbe962010-05-24 23:54:18 -07001897 * @wq: struct socket_wq
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001898 *
Eric Dumazet43815482010-04-29 11:01:49 +00001899 * Returns true if socket_wq has waiting processes
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001900 *
Eric Dumazet43815482010-04-29 11:01:49 +00001901 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001902 * barrier call. They were added due to the race found within the tcp code.
1903 *
1904 * Consider following tcp code paths:
1905 *
1906 * CPU1 CPU2
1907 *
1908 * sys_select receive packet
1909 * ... ...
1910 * __add_wait_queue update tp->rcv_nxt
1911 * ... ...
1912 * tp->rcv_nxt check sock_def_readable
1913 * ... {
Eric Dumazet43815482010-04-29 11:01:49 +00001914 * schedule rcu_read_lock();
1915 * wq = rcu_dereference(sk->sk_wq);
1916 * if (wq && waitqueue_active(&wq->wait))
1917 * wake_up_interruptible(&wq->wait)
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001918 * ...
1919 * }
1920 *
1921 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1922 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1923 * could then endup calling schedule and sleep forever if there are no more
1924 * data on the socket.
Jiri Olsaad462762009-07-08 12:10:31 +00001925 *
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001926 */
Eric Dumazet43815482010-04-29 11:01:49 +00001927static inline bool wq_has_sleeper(struct socket_wq *wq)
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001928{
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001929 /* We need to be sure we are in sync with the
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001930 * add_wait_queue modifications to the wait queue.
1931 *
1932 * This memory barrier is paired in the sock_poll_wait.
1933 */
Eric Dumazet43815482010-04-29 11:01:49 +00001934 smp_mb();
1935 return wq && waitqueue_active(&wq->wait);
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001936}
1937
1938/**
1939 * sock_poll_wait - place memory barrier behind the poll_wait call.
1940 * @filp: file
1941 * @wait_address: socket wait queue
1942 * @p: poll_table
1943 *
Eric Dumazet43815482010-04-29 11:01:49 +00001944 * See the comments in the wq_has_sleeper function.
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001945 */
1946static inline void sock_poll_wait(struct file *filp,
1947 wait_queue_head_t *wait_address, poll_table *p)
1948{
Hans Verkuil626cf232012-03-23 15:02:27 -07001949 if (!poll_does_not_wait(p) && wait_address) {
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001950 poll_wait(filp, wait_address, p);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001951 /* We need to be sure we are in sync with the
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001952 * socket flags modification.
1953 *
Eric Dumazet43815482010-04-29 11:01:49 +00001954 * This memory barrier is paired in the wq_has_sleeper.
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001955 */
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001956 smp_mb();
1957 }
1958}
1959
Linus Torvalds1da177e2005-04-16 15:20:36 -07001960/*
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001961 * Queue a received datagram if it will fit. Stream and sequenced
Linus Torvalds1da177e2005-04-16 15:20:36 -07001962 * protocols can't normally use this as they need to fit buffers in
1963 * and play with them.
1964 *
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001965 * Inlined as it's very short and called for pretty much every
Linus Torvalds1da177e2005-04-16 15:20:36 -07001966 * packet ever received.
1967 */
1968
1969static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1970{
Herbert Xud55d87f2009-06-22 02:25:25 +00001971 skb_orphan(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001972 skb->sk = sk;
1973 skb->destructor = sock_wfree;
Eric Dumazet2b85a342009-06-11 02:55:43 -07001974 /*
1975 * We used to take a refcount on sk, but following operation
1976 * is enough to guarantee sk_free() wont free this sock until
1977 * all in-flight packets are completed
1978 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001979 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1980}
1981
1982static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1983{
Herbert Xud55d87f2009-06-22 02:25:25 +00001984 skb_orphan(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001985 skb->sk = sk;
1986 skb->destructor = sock_rfree;
1987 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001988 sk_mem_charge(sk, skb->truesize);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001989}
1990
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001991extern void sk_reset_timer(struct sock *sk, struct timer_list *timer,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001992 unsigned long expires);
1993
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001994extern void sk_stop_timer(struct sock *sk, struct timer_list *timer);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001995
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001996extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001997
Eric Dumazetb1faf562010-05-31 23:44:05 -07001998extern int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001999
2000/*
2001 * Recover an error report and clear atomically
2002 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002003
Linus Torvalds1da177e2005-04-16 15:20:36 -07002004static inline int sock_error(struct sock *sk)
2005{
Benjamin LaHaisec1cbe4b2005-12-13 23:22:19 -08002006 int err;
2007 if (likely(!sk->sk_err))
2008 return 0;
2009 err = xchg(&sk->sk_err, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002010 return -err;
2011}
2012
2013static inline unsigned long sock_wspace(struct sock *sk)
2014{
2015 int amt = 0;
2016
2017 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
2018 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002019 if (amt < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002020 amt = 0;
2021 }
2022 return amt;
2023}
2024
2025static inline void sk_wake_async(struct sock *sk, int how, int band)
2026{
Eric Dumazetbcdce712009-10-06 17:28:29 -07002027 if (sock_flag(sk, SOCK_FASYNC))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002028 sock_wake_async(sk->sk_socket, how, band);
2029}
2030
2031#define SOCK_MIN_SNDBUF 2048
Eric Dumazet7a91b4342010-09-26 18:53:07 -07002032/*
2033 * Since sk_rmem_alloc sums skb->truesize, even a small frame might need
2034 * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak
2035 */
2036#define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002037
2038static inline void sk_stream_moderate_sndbuf(struct sock *sk)
2039{
2040 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
Eric Dumazet8df09ea2007-12-21 03:07:41 -08002041 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002042 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
2043 }
2044}
2045
Pavel Emelyanovdf97c702007-11-29 21:22:33 +11002046struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002047
Eric Dumazet5640f762012-09-23 23:04:42 +00002048/**
2049 * sk_page_frag - return an appropriate page_frag
2050 * @sk: socket
2051 *
2052 * If socket allocation mode allows current thread to sleep, it means its
2053 * safe to use the per task page_frag instead of the per socket one.
2054 */
2055static inline struct page_frag *sk_page_frag(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002056{
Eric Dumazet5640f762012-09-23 23:04:42 +00002057 if (sk->sk_allocation & __GFP_WAIT)
2058 return &current->task_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002059
Eric Dumazet5640f762012-09-23 23:04:42 +00002060 return &sk->sk_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002061}
2062
Eric Dumazet5640f762012-09-23 23:04:42 +00002063extern bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag);
2064
Linus Torvalds1da177e2005-04-16 15:20:36 -07002065/*
2066 * Default write policy as shown to user space via poll/select/SIGIO
2067 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002068static inline bool sock_writeable(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002069{
Eric Dumazet8df09ea2007-12-21 03:07:41 -08002070 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002071}
2072
Al Virodd0fc662005-10-07 07:46:04 +01002073static inline gfp_t gfp_any(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002074{
Andrew Morton99709372009-02-12 16:43:17 -08002075 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002076}
2077
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002078static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002079{
2080 return noblock ? 0 : sk->sk_rcvtimeo;
2081}
2082
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002083static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002084{
2085 return noblock ? 0 : sk->sk_sndtimeo;
2086}
2087
2088static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2089{
2090 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
2091}
2092
2093/* Alas, with timeout socket operations are not restartable.
2094 * Compare this to poll().
2095 */
2096static inline int sock_intr_errno(long timeo)
2097{
2098 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2099}
2100
Eric Dumazet92f37fd2007-03-25 22:14:49 -07002101extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2102 struct sk_buff *skb);
Johannes Berg6e3e9392011-11-09 10:15:42 +01002103extern void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2104 struct sk_buff *skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -07002105
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002106static inline void
Linus Torvalds1da177e2005-04-16 15:20:36 -07002107sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2108{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002109 ktime_t kt = skb->tstamp;
Patrick Ohly20d49472009-02-12 05:03:38 +00002110 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002111
Patrick Ohly20d49472009-02-12 05:03:38 +00002112 /*
2113 * generate control messages if
2114 * - receive time stamping in software requested (SOCK_RCVTSTAMP
2115 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
2116 * - software time stamp available and wanted
2117 * (SOCK_TIMESTAMPING_SOFTWARE)
2118 * - hardware time stamps available and wanted
2119 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
2120 * SOCK_TIMESTAMPING_RAW_HARDWARE)
2121 */
2122 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
2123 sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
2124 (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
2125 (hwtstamps->hwtstamp.tv64 &&
2126 sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
2127 (hwtstamps->syststamp.tv64 &&
2128 sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
Eric Dumazet92f37fd2007-03-25 22:14:49 -07002129 __sock_recv_timestamp(msg, sk, skb);
2130 else
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002131 sk->sk_stamp = kt;
Johannes Berg6e3e9392011-11-09 10:15:42 +01002132
2133 if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2134 __sock_recv_wifi_status(msg, sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002135}
2136
Eric Dumazet767dd032010-04-28 19:14:43 +00002137extern void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2138 struct sk_buff *skb);
2139
2140static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2141 struct sk_buff *skb)
2142{
2143#define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2144 (1UL << SOCK_RCVTSTAMP) | \
2145 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
2146 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002147 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
Eric Dumazet767dd032010-04-28 19:14:43 +00002148 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
2149
2150 if (sk->sk_flags & FLAGS_TS_OR_DROPS)
2151 __sock_recv_ts_and_drops(msg, sk, skb);
2152 else
2153 sk->sk_stamp = skb->tstamp;
2154}
Neil Horman3b885782009-10-12 13:26:31 -07002155
Linus Torvalds1da177e2005-04-16 15:20:36 -07002156/**
Patrick Ohly20d49472009-02-12 05:03:38 +00002157 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
Patrick Ohly20d49472009-02-12 05:03:38 +00002158 * @sk: socket sending this packet
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002159 * @tx_flags: filled with instructions for time stamping
Patrick Ohly20d49472009-02-12 05:03:38 +00002160 *
2161 * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if
2162 * parameters are invalid.
2163 */
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002164extern int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags);
Patrick Ohly20d49472009-02-12 05:03:38 +00002165
2166/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002167 * sk_eat_skb - Release a skb if it is no longer needed
Pavel Pisa4dc3b162005-05-01 08:59:25 -07002168 * @sk: socket to eat this skb from
2169 * @skb: socket buffer to eat
Randy Dunlapf4b8ea72006-06-22 16:00:11 -07002170 * @copied_early: flag indicating whether DMA operations copied this data early
Linus Torvalds1da177e2005-04-16 15:20:36 -07002171 *
2172 * This routine must be called with interrupts disabled or with the socket
2173 * locked so that the sk_buff queue operation is ok.
2174*/
Chris Leech624d1162006-05-23 18:01:28 -07002175#ifdef CONFIG_NET_DMA
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002176static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
Chris Leech624d1162006-05-23 18:01:28 -07002177{
2178 __skb_unlink(skb, &sk->sk_receive_queue);
2179 if (!copied_early)
2180 __kfree_skb(skb);
2181 else
2182 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
2183}
2184#else
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002185static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002186{
2187 __skb_unlink(skb, &sk->sk_receive_queue);
2188 __kfree_skb(skb);
2189}
Chris Leech624d1162006-05-23 18:01:28 -07002190#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002191
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002192static inline
2193struct net *sock_net(const struct sock *sk)
2194{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002195 return read_pnet(&sk->sk_net);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002196}
2197
2198static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07002199void sock_net_set(struct sock *sk, struct net *net)
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002200{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002201 write_pnet(&sk->sk_net, net);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002202}
2203
Denis V. Lunevedf02082008-02-29 11:18:32 -08002204/*
2205 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
Lucas De Marchi25985ed2011-03-30 22:57:33 -03002206 * They should not hold a reference to a namespace in order to allow
Denis V. Lunevedf02082008-02-29 11:18:32 -08002207 * to stop it.
2208 * Sockets after sk_change_net should be released using sk_release_kernel
2209 */
2210static inline void sk_change_net(struct sock *sk, struct net *net)
2211{
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002212 put_net(sock_net(sk));
Denis V. Lunev65a18ec2008-04-16 01:59:46 -07002213 sock_net_set(sk, hold_net(net));
Denis V. Lunevedf02082008-02-29 11:18:32 -08002214}
2215
KOVACS Krisztian23542612008-10-07 12:41:01 -07002216static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2217{
Vijay Subramanianefc27f82012-06-24 13:03:07 +00002218 if (skb->sk) {
KOVACS Krisztian23542612008-10-07 12:41:01 -07002219 struct sock *sk = skb->sk;
2220
2221 skb->destructor = NULL;
2222 skb->sk = NULL;
2223 return sk;
2224 }
2225 return NULL;
2226}
2227
Patrick Ohly20d49472009-02-12 05:03:38 +00002228extern void sock_enable_timestamp(struct sock *sk, int flag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002229extern int sock_get_timestamp(struct sock *, struct timeval __user *);
Eric Dumazetae40eb12007-03-18 17:33:16 -07002230extern int sock_get_timestampns(struct sock *, struct timespec __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002231
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002232/*
2233 * Enable debug/info messages
Linus Torvalds1da177e2005-04-16 15:20:36 -07002234 */
Stephen Hemmingera2a316f2007-03-08 20:41:08 -08002235extern int net_msg_warn;
2236#define NETDEBUG(fmt, args...) \
2237 do { if (net_msg_warn) printk(fmt,##args); } while (0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002238
Stephen Hemmingera2a316f2007-03-08 20:41:08 -08002239#define LIMIT_NETDEBUG(fmt, args...) \
2240 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002241
Linus Torvalds1da177e2005-04-16 15:20:36 -07002242extern __u32 sysctl_wmem_max;
2243extern __u32 sysctl_rmem_max;
2244
David S. Miller6baf1f42005-09-05 18:14:11 -07002245extern int sysctl_optmem_max;
2246
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002247extern __u32 sysctl_wmem_default;
2248extern __u32 sysctl_rmem_default;
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002249
Linus Torvalds1da177e2005-04-16 15:20:36 -07002250#endif /* _SOCK_H */