blob: c945fba4f54351475ff2efb989f77b23237f60d4 [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
129/**
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700130 * struct sock_common - minimal network layer representation of sockets
Eric Dumazet68835ab2010-11-30 19:04:07 +0000131 * @skc_daddr: Foreign IPv4 addr
132 * @skc_rcv_saddr: Bound local IPv4 addr
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000133 * @skc_hash: hash value used with various protocol lookup tables
Eric Dumazetd4cada42009-11-08 10:17:30 +0000134 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700135 * @skc_family: network address family
136 * @skc_state: Connection state
137 * @skc_reuse: %SO_REUSEADDR setting
138 * @skc_bound_dev_if: bound device index if != 0
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700139 * @skc_bind_node: bind hash linkage for various protocol lookup tables
Eric Dumazet512615b2009-11-08 10:17:58 +0000140 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700141 * @skc_prot: protocol handlers inside a network family
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200142 * @skc_net: reference to the network namespace of this socket
Eric Dumazet68835ab2010-11-30 19:04:07 +0000143 * @skc_node: main hash linkage for various protocol lookup tables
144 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
145 * @skc_tx_queue_mapping: tx queue number for this connection
146 * @skc_refcnt: reference count
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700147 *
148 * This is the minimal network layer representation of sockets, the header
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700149 * for struct sock and struct inet_timewait_sock.
150 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151struct sock_common {
Eric Dumazet68835ab2010-11-30 19:04:07 +0000152 /* skc_daddr and skc_rcv_saddr must be grouped :
153 * cf INET_MATCH() and INET_TW_MATCH()
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000154 */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000155 __be32 skc_daddr;
156 __be32 skc_rcv_saddr;
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000157
Eric Dumazetd4cada42009-11-08 10:17:30 +0000158 union {
159 unsigned int skc_hash;
160 __u16 skc_u16hashes[2];
161 };
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000162 unsigned short skc_family;
163 volatile unsigned char skc_state;
164 unsigned char skc_reuse;
165 int skc_bound_dev_if;
Eric Dumazet512615b2009-11-08 10:17:58 +0000166 union {
167 struct hlist_node skc_bind_node;
168 struct hlist_nulls_node skc_portaddr_node;
169 };
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700170 struct proto *skc_prot;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900171#ifdef CONFIG_NET_NS
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200172 struct net *skc_net;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900173#endif
Eric Dumazet68835ab2010-11-30 19:04:07 +0000174 /*
175 * fields between dontcopy_begin/dontcopy_end
176 * are not copied in sock_copy()
177 */
Randy Dunlap928c41e2011-01-08 17:39:21 +0000178 /* private: */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000179 int skc_dontcopy_begin[0];
Randy Dunlap928c41e2011-01-08 17:39:21 +0000180 /* public: */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000181 union {
182 struct hlist_node skc_node;
183 struct hlist_nulls_node skc_nulls_node;
184 };
185 int skc_tx_queue_mapping;
186 atomic_t skc_refcnt;
Randy Dunlap928c41e2011-01-08 17:39:21 +0000187 /* private: */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000188 int skc_dontcopy_end[0];
Randy Dunlap928c41e2011-01-08 17:39:21 +0000189 /* public: */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190};
191
Glauber Costae1aab162011-12-11 21:47:03 +0000192struct cg_proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193/**
194 * struct sock - network layer representation of sockets
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700195 * @__sk_common: shared layout with inet_timewait_sock
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700196 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
197 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
198 * @sk_lock: synchronizer
199 * @sk_rcvbuf: size of receive buffer in bytes
Eric Dumazet43815482010-04-29 11:01:49 +0000200 * @sk_wq: sock wait queue and async head
Eric Dumazetdeaa5852012-06-24 20:22:49 +0000201 * @sk_rx_dst: receive input route used by early tcp demux
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700202 * @sk_dst_cache: destination cache
203 * @sk_dst_lock: destination cache lock
204 * @sk_policy: flow policy
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700205 * @sk_receive_queue: incoming packets
206 * @sk_wmem_alloc: transmit queue bytes committed
207 * @sk_write_queue: Packet sending queue
Chris Leech97fc2f02006-05-23 17:55:33 -0700208 * @sk_async_wait_queue: DMA copied packets
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700209 * @sk_omem_alloc: "o" is "option" or "other"
210 * @sk_wmem_queued: persistent queue size
211 * @sk_forward_alloc: space allocated forward
212 * @sk_allocation: allocation mode
213 * @sk_sndbuf: size of send buffer in bytes
Wang Chen33c732c2007-11-13 20:30:01 -0800214 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
Patrick Ohly20d49472009-02-12 05:03:38 +0000215 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700216 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
217 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
Eric Dumazeta4654192010-05-16 00:36:33 -0700218 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
Herbert Xubcd76112006-06-30 13:36:35 -0700219 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -0700220 * @sk_gso_max_size: Maximum GSO segment size to build
Ben Hutchings14853482012-07-30 16:11:42 +0000221 * @sk_gso_max_segs: Maximum number of GSO segments
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700222 * @sk_lingertime: %SO_LINGER l_linger setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700223 * @sk_backlog: always used with the per-socket spinlock held
224 * @sk_callback_lock: used with the callbacks in the end of this struct
225 * @sk_error_queue: rarely used
Wang Chen33c732c2007-11-13 20:30:01 -0800226 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
227 * IPV6_ADDRFORM for instance)
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700228 * @sk_err: last error
Wang Chen33c732c2007-11-13 20:30:01 -0800229 * @sk_err_soft: errors that don't cause failure but are the cause of a
230 * persistent failure not just 'timed out'
Eric Dumazetcb61cb92008-06-17 21:04:56 -0700231 * @sk_drops: raw/udp drops counter
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700232 * @sk_ack_backlog: current listen backlog
233 * @sk_max_ack_backlog: listen backlog set in listen()
234 * @sk_priority: %SO_PRIORITY setting
Randy Dunlap1a3bc362012-01-21 09:03:10 +0000235 * @sk_cgrp_prioidx: socket group's priority map index
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700236 * @sk_type: socket type (%SOCK_STREAM, etc)
237 * @sk_protocol: which protocol this socket belongs in this network family
Randy Dunlap53c3fa22010-08-09 13:41:07 +0000238 * @sk_peer_pid: &struct pid for this socket's peer
239 * @sk_peer_cred: %SO_PEERCRED setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700240 * @sk_rcvlowat: %SO_RCVLOWAT setting
241 * @sk_rcvtimeo: %SO_RCVTIMEO setting
242 * @sk_sndtimeo: %SO_SNDTIMEO setting
David S. Millerc58dc012010-04-27 15:05:31 -0700243 * @sk_rxhash: flow hash received from netif layer
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700244 * @sk_filter: socket filtering instructions
245 * @sk_protinfo: private area, net family specific, when not using slab
246 * @sk_timer: sock cleanup timer
247 * @sk_stamp: time stamp of last packet received
248 * @sk_socket: Identd and reporting IO signals
249 * @sk_user_data: RPC layer private data
Eric Dumazet5640f762012-09-23 23:04:42 +0000250 * @sk_frag: cached page frag
Randy Dunlapd3d4f0a2012-04-17 14:03:53 +0000251 * @sk_peek_off: current peek_offset value
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700252 * @sk_send_head: front of stuff to transmit
Martin Waitz67be2dd2005-05-01 08:59:26 -0700253 * @sk_security: used by security modules
Randy Dunlap31729362008-02-18 20:52:13 -0800254 * @sk_mark: generic packet mark
Randy Dunlap53c3fa22010-08-09 13:41:07 +0000255 * @sk_classid: this socket's cgroup classid
Glauber Costae1aab162011-12-11 21:47:03 +0000256 * @sk_cgrp: this socket's cgroup-specific proto data
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700257 * @sk_write_pending: a write to stream socket waits to start
258 * @sk_state_change: callback to indicate change in the state of the sock
259 * @sk_data_ready: callback to indicate there is data to be processed
260 * @sk_write_space: callback to indicate there is bf sending space available
261 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
262 * @sk_backlog_rcv: callback to process the backlog
263 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264 */
265struct sock {
266 /*
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700267 * Now struct inet_timewait_sock also uses sock_common, so please just
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268 * don't add nothing before this first member (__sk_common) --acme
269 */
270 struct sock_common __sk_common;
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000271#define sk_node __sk_common.skc_node
272#define sk_nulls_node __sk_common.skc_nulls_node
273#define sk_refcnt __sk_common.skc_refcnt
Krishna Kumare022f0b2009-10-19 23:46:20 +0000274#define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000275
Eric Dumazet68835ab2010-11-30 19:04:07 +0000276#define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
277#define sk_dontcopy_end __sk_common.skc_dontcopy_end
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000278#define sk_hash __sk_common.skc_hash
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279#define sk_family __sk_common.skc_family
280#define sk_state __sk_common.skc_state
281#define sk_reuse __sk_common.skc_reuse
282#define sk_bound_dev_if __sk_common.skc_bound_dev_if
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283#define sk_bind_node __sk_common.skc_bind_node
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700284#define sk_prot __sk_common.skc_prot
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200285#define sk_net __sk_common.skc_net
Eric Dumazetb178bb32010-11-16 05:56:04 +0000286 socket_lock_t sk_lock;
287 struct sk_buff_head sk_receive_queue;
288 /*
289 * The backlog queue is special, it is always used with
290 * the per-socket spinlock held and requires low latency
291 * access. Therefore we special case it's implementation.
292 * Note : rmem_alloc is in this structure to fill a hole
293 * on 64bit arches, not because its logically part of
294 * backlog.
295 */
296 struct {
297 atomic_t rmem_alloc;
298 int len;
299 struct sk_buff *head;
300 struct sk_buff *tail;
301 } sk_backlog;
302#define sk_rmem_alloc sk_backlog.rmem_alloc
303 int sk_forward_alloc;
304#ifdef CONFIG_RPS
305 __u32 sk_rxhash;
306#endif
307 atomic_t sk_drops;
308 int sk_rcvbuf;
309
310 struct sk_filter __rcu *sk_filter;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000311 struct socket_wq __rcu *sk_wq;
Eric Dumazetb178bb32010-11-16 05:56:04 +0000312
313#ifdef CONFIG_NET_DMA
314 struct sk_buff_head sk_async_wait_queue;
315#endif
316
317#ifdef CONFIG_XFRM
318 struct xfrm_policy *sk_policy[2];
319#endif
320 unsigned long sk_flags;
Eric Dumazetdeaa5852012-06-24 20:22:49 +0000321 struct dst_entry *sk_rx_dst;
Eric Dumazetb178bb32010-11-16 05:56:04 +0000322 struct dst_entry *sk_dst_cache;
323 spinlock_t sk_dst_lock;
324 atomic_t sk_wmem_alloc;
325 atomic_t sk_omem_alloc;
326 int sk_sndbuf;
327 struct sk_buff_head sk_write_queue;
Vegard Nossuma98b65a2009-02-26 14:46:57 +0100328 kmemcheck_bitfield_begin(flags);
Eric Dumazet5fdb9972009-10-08 22:50:25 +0000329 unsigned int sk_shutdown : 2,
330 sk_no_check : 2,
331 sk_userlocks : 4,
332 sk_protocol : 8,
333 sk_type : 16;
Vegard Nossuma98b65a2009-02-26 14:46:57 +0100334 kmemcheck_bitfield_end(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335 int sk_wmem_queued;
Al Viro7d877f32005-10-21 03:20:43 -0400336 gfp_t sk_allocation;
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000337 netdev_features_t sk_route_caps;
338 netdev_features_t sk_route_nocaps;
Herbert Xubcd76112006-06-30 13:36:35 -0700339 int sk_gso_type;
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -0700340 unsigned int sk_gso_max_size;
Ben Hutchings14853482012-07-30 16:11:42 +0000341 u16 sk_gso_max_segs;
David S. Miller9932cf92006-03-24 15:12:37 -0800342 int sk_rcvlowat;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343 unsigned long sk_lingertime;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344 struct sk_buff_head sk_error_queue;
Arnaldo Carvalho de Melo476e19c2005-05-05 13:35:15 -0700345 struct proto *sk_prot_creator;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346 rwlock_t sk_callback_lock;
347 int sk_err,
348 sk_err_soft;
349 unsigned short sk_ack_backlog;
350 unsigned short sk_max_ack_backlog;
351 __u32 sk_priority;
Neil Horman5bc14212011-11-22 05:10:51 +0000352#ifdef CONFIG_CGROUPS
353 __u32 sk_cgrp_prioidx;
354#endif
Eric W. Biederman109f6e32010-06-13 03:30:14 +0000355 struct pid *sk_peer_pid;
356 const struct cred *sk_peer_cred;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357 long sk_rcvtimeo;
358 long sk_sndtimeo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359 void *sk_protinfo;
360 struct timer_list sk_timer;
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700361 ktime_t sk_stamp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362 struct socket *sk_socket;
363 void *sk_user_data;
Eric Dumazet5640f762012-09-23 23:04:42 +0000364 struct page_frag sk_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365 struct sk_buff *sk_send_head;
Pavel Emelyanovef64a542012-02-21 07:31:34 +0000366 __s32 sk_peek_off;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367 int sk_write_pending;
Alexey Dobriyand5f64232008-11-04 14:45:58 -0800368#ifdef CONFIG_SECURITY
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369 void *sk_security;
Alexey Dobriyand5f64232008-11-04 14:45:58 -0800370#endif
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800371 __u32 sk_mark;
Herbert Xuf8451722010-05-24 00:12:34 -0700372 u32 sk_classid;
Glauber Costae1aab162011-12-11 21:47:03 +0000373 struct cg_proto *sk_cgrp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374 void (*sk_state_change)(struct sock *sk);
375 void (*sk_data_ready)(struct sock *sk, int bytes);
376 void (*sk_write_space)(struct sock *sk);
377 void (*sk_error_report)(struct sock *sk);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000378 int (*sk_backlog_rcv)(struct sock *sk,
379 struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380 void (*sk_destruct)(struct sock *sk);
381};
382
Pavel Emelyanov4a17fd52012-04-19 03:39:36 +0000383/*
384 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
385 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
386 * on a socket means that the socket will reuse everybody else's port
387 * without looking at the other's sk_reuse value.
388 */
389
390#define SK_NO_REUSE 0
391#define SK_CAN_REUSE 1
392#define SK_FORCE_REUSE 2
393
Pavel Emelyanovef64a542012-02-21 07:31:34 +0000394static inline int sk_peek_offset(struct sock *sk, int flags)
395{
396 if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0))
397 return sk->sk_peek_off;
398 else
399 return 0;
400}
401
402static inline void sk_peek_offset_bwd(struct sock *sk, int val)
403{
404 if (sk->sk_peek_off >= 0) {
405 if (sk->sk_peek_off >= val)
406 sk->sk_peek_off -= val;
407 else
408 sk->sk_peek_off = 0;
409 }
410}
411
412static inline void sk_peek_offset_fwd(struct sock *sk, int val)
413{
414 if (sk->sk_peek_off >= 0)
415 sk->sk_peek_off += val;
416}
417
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418/*
419 * Hashed lists helper routines
420 */
Li Zefanc4146642010-02-08 23:18:45 +0000421static inline struct sock *sk_entry(const struct hlist_node *node)
422{
423 return hlist_entry(node, struct sock, sk_node);
424}
425
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700426static inline struct sock *__sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427{
428 return hlist_entry(head->first, struct sock, sk_node);
429}
430
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700431static inline struct sock *sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432{
433 return hlist_empty(head) ? NULL : __sk_head(head);
434}
435
Eric Dumazet88ab1932008-11-16 19:39:21 -0800436static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
437{
438 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
439}
440
441static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
442{
443 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
444}
445
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700446static inline struct sock *sk_next(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447{
448 return sk->sk_node.next ?
449 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
450}
451
Eric Dumazet88ab1932008-11-16 19:39:21 -0800452static inline struct sock *sk_nulls_next(const struct sock *sk)
453{
454 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
455 hlist_nulls_entry(sk->sk_nulls_node.next,
456 struct sock, sk_nulls_node) :
457 NULL;
458}
459
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000460static inline bool sk_unhashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461{
462 return hlist_unhashed(&sk->sk_node);
463}
464
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000465static inline bool sk_hashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466{
Akinobu Mitada753be2006-04-28 15:21:23 -0700467 return !sk_unhashed(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468}
469
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000470static inline void sk_node_init(struct hlist_node *node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471{
472 node->pprev = NULL;
473}
474
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000475static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800476{
477 node->pprev = NULL;
478}
479
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000480static inline void __sk_del_node(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481{
482 __hlist_del(&sk->sk_node);
483}
484
stephen hemminger808f5112010-02-22 07:57:18 +0000485/* NB: equivalent to hlist_del_init_rcu */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000486static inline bool __sk_del_node_init(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487{
488 if (sk_hashed(sk)) {
489 __sk_del_node(sk);
490 sk_node_init(&sk->sk_node);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000491 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492 }
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000493 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494}
495
496/* Grab socket reference count. This operation is valid only
497 when sk is ALREADY grabbed f.e. it is found in hash table
498 or a list and the lookup is made under lock preventing hash table
499 modifications.
500 */
501
502static inline void sock_hold(struct sock *sk)
503{
504 atomic_inc(&sk->sk_refcnt);
505}
506
507/* Ungrab socket in the context, which assumes that socket refcnt
508 cannot hit zero, f.e. it is true in context of any socketcall.
509 */
510static inline void __sock_put(struct sock *sk)
511{
512 atomic_dec(&sk->sk_refcnt);
513}
514
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000515static inline bool sk_del_node_init(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516{
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000517 bool rc = __sk_del_node_init(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518
519 if (rc) {
520 /* paranoid for a while -acme */
521 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
522 __sock_put(sk);
523 }
524 return rc;
525}
stephen hemminger808f5112010-02-22 07:57:18 +0000526#define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000528static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700529{
530 if (sk_hashed(sk)) {
Eric Dumazet88ab1932008-11-16 19:39:21 -0800531 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000532 return true;
Eric Dumazet271b72c2008-10-29 02:11:14 -0700533 }
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000534 return false;
Eric Dumazet271b72c2008-10-29 02:11:14 -0700535}
536
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000537static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700538{
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000539 bool rc = __sk_nulls_del_node_init_rcu(sk);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700540
541 if (rc) {
542 /* paranoid for a while -acme */
543 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
544 __sock_put(sk);
545 }
546 return rc;
547}
548
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000549static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550{
551 hlist_add_head(&sk->sk_node, list);
552}
553
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000554static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555{
556 sock_hold(sk);
557 __sk_add_node(sk, list);
558}
559
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000560static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
stephen hemminger808f5112010-02-22 07:57:18 +0000561{
562 sock_hold(sk);
563 hlist_add_head_rcu(&sk->sk_node, list);
564}
565
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000566static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700567{
Eric Dumazet88ab1932008-11-16 19:39:21 -0800568 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700569}
570
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000571static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700572{
573 sock_hold(sk);
Eric Dumazet88ab1932008-11-16 19:39:21 -0800574 __sk_nulls_add_node_rcu(sk, list);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700575}
576
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000577static inline void __sk_del_bind_node(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578{
579 __hlist_del(&sk->sk_bind_node);
580}
581
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000582static inline void sk_add_bind_node(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583 struct hlist_head *list)
584{
585 hlist_add_head(&sk->sk_bind_node, list);
586}
587
588#define sk_for_each(__sk, node, list) \
589 hlist_for_each_entry(__sk, node, list, sk_node)
stephen hemminger808f5112010-02-22 07:57:18 +0000590#define sk_for_each_rcu(__sk, node, list) \
591 hlist_for_each_entry_rcu(__sk, node, list, sk_node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800592#define sk_nulls_for_each(__sk, node, list) \
593 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
594#define sk_nulls_for_each_rcu(__sk, node, list) \
595 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596#define sk_for_each_from(__sk, node) \
597 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
598 hlist_for_each_entry_from(__sk, node, sk_node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800599#define sk_nulls_for_each_from(__sk, node) \
600 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
601 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602#define sk_for_each_safe(__sk, node, tmp, list) \
603 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
604#define sk_for_each_bound(__sk, node, list) \
605 hlist_for_each_entry(__sk, node, list, sk_bind_node)
606
Eric W. Biedermanc336d142012-05-24 17:56:43 -0600607static inline struct user_namespace *sk_user_ns(struct sock *sk)
608{
609 /* Careful only use this in a context where these parameters
610 * can not change and must all be valid, such as recvmsg from
611 * userspace.
612 */
613 return sk->sk_socket->file->f_cred->user_ns;
614}
615
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616/* Sock flags */
617enum sock_flags {
618 SOCK_DEAD,
619 SOCK_DONE,
620 SOCK_URGINLINE,
621 SOCK_KEEPOPEN,
622 SOCK_LINGER,
623 SOCK_DESTROY,
624 SOCK_BROADCAST,
625 SOCK_TIMESTAMP,
626 SOCK_ZAPPED,
627 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
628 SOCK_DBG, /* %SO_DEBUG setting */
629 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700630 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
632 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
Mel Gorman7cb02402012-07-31 16:44:16 -0700633 SOCK_MEMALLOC, /* VM depends on this socket for swapping */
Patrick Ohly20d49472009-02-12 05:03:38 +0000634 SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */
635 SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
636 SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */
637 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
638 SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */
639 SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
640 SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
Eric Dumazetbcdce712009-10-06 17:28:29 -0700641 SOCK_FASYNC, /* fasync() active */
Neil Horman3b885782009-10-12 13:26:31 -0700642 SOCK_RXQ_OVFL,
Shirley Ma1cdebb42011-07-06 12:17:30 +0000643 SOCK_ZEROCOPY, /* buffers from userspace */
Johannes Berg6e3e9392011-11-09 10:15:42 +0100644 SOCK_WIFI_STATUS, /* push wifi status to userspace */
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000645 SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
646 * Will use last 4 bytes of packet sent from
647 * user-space instead.
648 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649};
650
Ralf Baechle53b924b2005-08-23 10:11:30 -0700651static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
652{
653 nsk->sk_flags = osk->sk_flags;
654}
655
Linus Torvalds1da177e2005-04-16 15:20:36 -0700656static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
657{
658 __set_bit(flag, &sk->sk_flags);
659}
660
661static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
662{
663 __clear_bit(flag, &sk->sk_flags);
664}
665
Eric Dumazet1b23a5d2012-05-16 05:57:07 +0000666static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667{
668 return test_bit(flag, &sk->sk_flags);
669}
670
Mel Gormanc93bdd02012-07-31 16:44:19 -0700671#ifdef CONFIG_NET
672extern struct static_key memalloc_socks;
673static inline int sk_memalloc_socks(void)
674{
675 return static_key_false(&memalloc_socks);
676}
677#else
678
679static inline int sk_memalloc_socks(void)
680{
681 return 0;
682}
683
684#endif
685
Mel Gorman99a1dec2012-07-31 16:44:14 -0700686static inline gfp_t sk_gfp_atomic(struct sock *sk, gfp_t gfp_mask)
687{
Mel Gorman7cb02402012-07-31 16:44:16 -0700688 return GFP_ATOMIC | (sk->sk_allocation & __GFP_MEMALLOC);
Mel Gorman99a1dec2012-07-31 16:44:14 -0700689}
690
Linus Torvalds1da177e2005-04-16 15:20:36 -0700691static inline void sk_acceptq_removed(struct sock *sk)
692{
693 sk->sk_ack_backlog--;
694}
695
696static inline void sk_acceptq_added(struct sock *sk)
697{
698 sk->sk_ack_backlog++;
699}
700
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000701static inline bool sk_acceptq_is_full(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702{
David S. Miller64a14652007-03-06 11:21:05 -0800703 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700704}
705
706/*
707 * Compute minimal free write space needed to queue new packets.
708 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000709static inline int sk_stream_min_wspace(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710{
Eric Dumazet8df09ea2007-12-21 03:07:41 -0800711 return sk->sk_wmem_queued >> 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700712}
713
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000714static inline int sk_stream_wspace(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715{
716 return sk->sk_sndbuf - sk->sk_wmem_queued;
717}
718
719extern void sk_stream_write_space(struct sock *sk);
720
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000721static inline bool sk_stream_memory_free(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722{
723 return sk->sk_wmem_queued < sk->sk_sndbuf;
724}
725
Zhu Yi8eae9392010-03-04 18:01:40 +0000726/* OOB backlog add */
Zhu Yia3a858f2010-03-04 18:01:47 +0000727static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800728{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000729 /* dont let skb dst not refcounted, we are going to leave rcu lock */
730 skb_dst_force(skb);
731
732 if (!sk->sk_backlog.tail)
733 sk->sk_backlog.head = skb;
734 else
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800735 sk->sk_backlog.tail->next = skb;
Eric Dumazet7fee2262010-05-11 23:19:48 +0000736
737 sk->sk_backlog.tail = skb;
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800738 skb->next = NULL;
739}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740
Eric Dumazetc3774112010-04-27 15:13:20 -0700741/*
742 * Take into account size of receive queue and backlog queue
Eric Dumazet0fd7bac2011-12-21 07:11:44 +0000743 * Do not take into account this skb truesize,
744 * to allow even a single big packet to come.
Eric Dumazetc3774112010-04-27 15:13:20 -0700745 */
Eric Dumazetf545a382012-04-22 23:34:26 +0000746static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb,
747 unsigned int limit)
Eric Dumazetc3774112010-04-27 15:13:20 -0700748{
749 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
750
Eric Dumazetf545a382012-04-22 23:34:26 +0000751 return qsize > limit;
Eric Dumazetc3774112010-04-27 15:13:20 -0700752}
753
Zhu Yi8eae9392010-03-04 18:01:40 +0000754/* The per-socket spinlock must be held here. */
Eric Dumazetf545a382012-04-22 23:34:26 +0000755static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
756 unsigned int limit)
Zhu Yi8eae9392010-03-04 18:01:40 +0000757{
Eric Dumazetf545a382012-04-22 23:34:26 +0000758 if (sk_rcvqueues_full(sk, skb, limit))
Zhu Yi8eae9392010-03-04 18:01:40 +0000759 return -ENOBUFS;
760
Zhu Yia3a858f2010-03-04 18:01:47 +0000761 __sk_add_backlog(sk, skb);
Zhu Yi8eae9392010-03-04 18:01:40 +0000762 sk->sk_backlog.len += skb->truesize;
763 return 0;
764}
765
Mel Gormanb4b9e352012-07-31 16:44:26 -0700766extern int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
767
Peter Zijlstrac57943a2008-10-07 14:18:42 -0700768static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
769{
Mel Gormanb4b9e352012-07-31 16:44:26 -0700770 if (sk_memalloc_socks() && skb_pfmemalloc(skb))
771 return __sk_backlog_rcv(sk, skb);
772
Peter Zijlstrac57943a2008-10-07 14:18:42 -0700773 return sk->sk_backlog_rcv(sk, skb);
774}
775
David S. Millerc58dc012010-04-27 15:05:31 -0700776static inline void sock_rps_record_flow(const struct sock *sk)
777{
778#ifdef CONFIG_RPS
779 struct rps_sock_flow_table *sock_flow_table;
780
781 rcu_read_lock();
782 sock_flow_table = rcu_dereference(rps_sock_flow_table);
783 rps_record_sock_flow(sock_flow_table, sk->sk_rxhash);
784 rcu_read_unlock();
785#endif
786}
787
788static inline void sock_rps_reset_flow(const struct sock *sk)
789{
790#ifdef CONFIG_RPS
791 struct rps_sock_flow_table *sock_flow_table;
792
793 rcu_read_lock();
794 sock_flow_table = rcu_dereference(rps_sock_flow_table);
795 rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash);
796 rcu_read_unlock();
797#endif
798}
799
Tom Herbertbdeab992011-08-14 19:45:55 +0000800static inline void sock_rps_save_rxhash(struct sock *sk,
801 const struct sk_buff *skb)
David S. Millerc58dc012010-04-27 15:05:31 -0700802{
803#ifdef CONFIG_RPS
Tom Herbertbdeab992011-08-14 19:45:55 +0000804 if (unlikely(sk->sk_rxhash != skb->rxhash)) {
David S. Millerc58dc012010-04-27 15:05:31 -0700805 sock_rps_reset_flow(sk);
Tom Herbertbdeab992011-08-14 19:45:55 +0000806 sk->sk_rxhash = skb->rxhash;
David S. Millerc58dc012010-04-27 15:05:31 -0700807 }
808#endif
809}
810
Tom Herbertbdeab992011-08-14 19:45:55 +0000811static inline void sock_rps_reset_rxhash(struct sock *sk)
812{
813#ifdef CONFIG_RPS
814 sock_rps_reset_flow(sk);
815 sk->sk_rxhash = 0;
816#endif
817}
818
Stephen Hemmingercfcabdc2007-10-09 01:59:42 -0700819#define sk_wait_event(__sk, __timeo, __condition) \
820 ({ int __rc; \
821 release_sock(__sk); \
822 __rc = __condition; \
823 if (!__rc) { \
824 *(__timeo) = schedule_timeout(*(__timeo)); \
825 } \
826 lock_sock(__sk); \
827 __rc = __condition; \
828 __rc; \
829 })
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830
831extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
832extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
833extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
834extern int sk_stream_error(struct sock *sk, int flags, int err);
835extern void sk_stream_kill_queues(struct sock *sk);
Mel Gorman7cb02402012-07-31 16:44:16 -0700836extern void sk_set_memalloc(struct sock *sk);
837extern void sk_clear_memalloc(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838
839extern int sk_wait_data(struct sock *sk, long *timeo);
840
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700841struct request_sock_ops;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800842struct timewait_sock_ops;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -0800843struct inet_hashinfo;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -0700844struct raw_hashinfo;
Paul Gortmakerde477252011-05-26 13:46:22 -0400845struct module;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700846
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847/* Networking protocol blocks we attach to sockets.
848 * socket layer -> transport layer interface
849 * transport -> network interface is defined by struct inet_proto
850 */
851struct proto {
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000852 void (*close)(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853 long timeout);
854 int (*connect)(struct sock *sk,
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000855 struct sockaddr *uaddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856 int addr_len);
857 int (*disconnect)(struct sock *sk, int flags);
858
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000859 struct sock * (*accept)(struct sock *sk, int flags, int *err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860
861 int (*ioctl)(struct sock *sk, int cmd,
862 unsigned long arg);
863 int (*init)(struct sock *sk);
Brian Haley7d06b2e2008-06-14 17:04:49 -0700864 void (*destroy)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865 void (*shutdown)(struct sock *sk, int how);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000866 int (*setsockopt)(struct sock *sk, int level,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700867 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700868 unsigned int optlen);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000869 int (*getsockopt)(struct sock *sk, int level,
870 int optname, char __user *optval,
871 int __user *option);
Alexey Dobriyanaf01d532008-08-28 02:53:51 -0700872#ifdef CONFIG_COMPAT
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800873 int (*compat_setsockopt)(struct sock *sk,
874 int level,
875 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700876 unsigned int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800877 int (*compat_getsockopt)(struct sock *sk,
878 int level,
879 int optname, char __user *optval,
880 int __user *option);
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000881 int (*compat_ioctl)(struct sock *sk,
882 unsigned int cmd, unsigned long arg);
Alexey Dobriyanaf01d532008-08-28 02:53:51 -0700883#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
885 struct msghdr *msg, size_t len);
886 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
887 struct msghdr *msg,
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000888 size_t len, int noblock, int flags,
889 int *addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890 int (*sendpage)(struct sock *sk, struct page *page,
891 int offset, size_t size, int flags);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000892 int (*bind)(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893 struct sockaddr *uaddr, int addr_len);
894
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000895 int (*backlog_rcv) (struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896 struct sk_buff *skb);
897
Eric Dumazet46d3cea2012-07-11 05:50:31 +0000898 void (*release_cb)(struct sock *sk);
Eric Dumazet563d34d2012-07-23 09:48:52 +0200899 void (*mtu_reduced)(struct sock *sk);
Eric Dumazet46d3cea2012-07-11 05:50:31 +0000900
Linus Torvalds1da177e2005-04-16 15:20:36 -0700901 /* Keeping track of sk's, looking them up, and port selection methods. */
902 void (*hash)(struct sock *sk);
903 void (*unhash)(struct sock *sk);
Eric Dumazet719f8352010-09-08 05:08:44 +0000904 void (*rehash)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905 int (*get_port)(struct sock *sk, unsigned short snum);
Octavian Purdilafcbdf092010-12-16 14:26:56 -0800906 void (*clear_sk)(struct sock *sk, int size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907
Eric Dumazet286ab3d2007-11-05 23:38:39 -0800908 /* Keeping track of sockets in use */
Eric Dumazet65f76512008-01-03 20:46:48 -0800909#ifdef CONFIG_PROC_FS
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -0700910 unsigned int inuse_idx;
Eric Dumazet65f76512008-01-03 20:46:48 -0800911#endif
Arnaldo Carvalho de Meloebb53d72007-11-21 22:08:50 +0800912
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913 /* Memory pressure */
Pavel Emelyanov5c52ba12008-07-16 20:28:10 -0700914 void (*enter_memory_pressure)(struct sock *sk);
Eric Dumazet8d987e52010-11-09 23:24:26 +0000915 atomic_long_t *memory_allocated; /* Current allocated memory. */
Eric Dumazet17483762008-11-25 21:16:35 -0800916 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700917 /*
918 * Pressure flag: try to collapse.
919 * Technical note: it is used by multiple contexts non atomically.
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800920 * All the __sk_mem_schedule() is of this nature: accounting
Linus Torvalds1da177e2005-04-16 15:20:36 -0700921 * is strict, actions are advisory and have some latency.
922 */
923 int *memory_pressure;
Eric Dumazet8d987e52010-11-09 23:24:26 +0000924 long *sysctl_mem;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700925 int *sysctl_wmem;
926 int *sysctl_rmem;
927 int max_header;
Changli Gao7ba42912010-07-10 20:41:55 +0000928 bool no_autobind;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929
Eric Dumazet271b72c2008-10-29 02:11:14 -0700930 struct kmem_cache *slab;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931 unsigned int obj_size;
Eric Dumazet271b72c2008-10-29 02:11:14 -0700932 int slab_flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700933
Eric Dumazetdd24c002008-11-25 21:17:14 -0800934 struct percpu_counter *orphan_count;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700935
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700936 struct request_sock_ops *rsk_prot;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800937 struct timewait_sock_ops *twsk_prot;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700938
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -0700939 union {
940 struct inet_hashinfo *hashinfo;
Eric Dumazet645ca702008-10-29 01:41:45 -0700941 struct udp_table *udp_table;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -0700942 struct raw_hashinfo *raw_hash;
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -0700943 } h;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -0800944
Linus Torvalds1da177e2005-04-16 15:20:36 -0700945 struct module *owner;
946
947 char name[32];
948
949 struct list_head node;
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700950#ifdef SOCK_REFCNT_DEBUG
951 atomic_t socks;
952#endif
Andrew Mortonc255a452012-07-31 16:43:02 -0700953#ifdef CONFIG_MEMCG_KMEM
Glauber Costae1aab162011-12-11 21:47:03 +0000954 /*
955 * cgroup specific init/deinit functions. Called once for all
956 * protocols that implement it, from cgroups populate function.
957 * This function has to setup any files the protocol want to
958 * appear in the kmem cgroup filesystem.
959 */
Glauber Costa1d62e432012-04-09 19:36:33 -0300960 int (*init_cgroup)(struct mem_cgroup *memcg,
Glauber Costae1aab162011-12-11 21:47:03 +0000961 struct cgroup_subsys *ss);
Glauber Costa1d62e432012-04-09 19:36:33 -0300962 void (*destroy_cgroup)(struct mem_cgroup *memcg);
Glauber Costae1aab162011-12-11 21:47:03 +0000963 struct cg_proto *(*proto_cgroup)(struct mem_cgroup *memcg);
964#endif
965};
966
Glauber Costa3f134612012-05-29 15:07:11 -0700967/*
968 * Bits in struct cg_proto.flags
969 */
970enum cg_proto_flags {
971 /* Currently active and new sockets should be assigned to cgroups */
972 MEMCG_SOCK_ACTIVE,
973 /* It was ever activated; we must disarm static keys on destruction */
974 MEMCG_SOCK_ACTIVATED,
975};
976
Glauber Costae1aab162011-12-11 21:47:03 +0000977struct cg_proto {
978 void (*enter_memory_pressure)(struct sock *sk);
979 struct res_counter *memory_allocated; /* Current allocated memory. */
980 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
981 int *memory_pressure;
982 long *sysctl_mem;
Glauber Costa3f134612012-05-29 15:07:11 -0700983 unsigned long flags;
Glauber Costae1aab162011-12-11 21:47:03 +0000984 /*
985 * memcg field is used to find which memcg we belong directly
986 * Each memcg struct can hold more than one cg_proto, so container_of
987 * won't really cut.
988 *
989 * The elegant solution would be having an inverse function to
990 * proto_cgroup in struct proto, but that means polluting the structure
991 * for everybody, instead of just for memcg users.
992 */
993 struct mem_cgroup *memcg;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994};
995
996extern int proto_register(struct proto *prot, int alloc_slab);
997extern void proto_unregister(struct proto *prot);
998
Glauber Costa3f134612012-05-29 15:07:11 -0700999static inline bool memcg_proto_active(struct cg_proto *cg_proto)
1000{
1001 return test_bit(MEMCG_SOCK_ACTIVE, &cg_proto->flags);
1002}
1003
1004static inline bool memcg_proto_activated(struct cg_proto *cg_proto)
1005{
1006 return test_bit(MEMCG_SOCK_ACTIVATED, &cg_proto->flags);
1007}
1008
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -07001009#ifdef SOCK_REFCNT_DEBUG
1010static inline void sk_refcnt_debug_inc(struct sock *sk)
1011{
1012 atomic_inc(&sk->sk_prot->socks);
1013}
1014
1015static inline void sk_refcnt_debug_dec(struct sock *sk)
1016{
1017 atomic_dec(&sk->sk_prot->socks);
1018 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
1019 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
1020}
1021
Glauber Costae1aab162011-12-11 21:47:03 +00001022inline void sk_refcnt_debug_release(const struct sock *sk)
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -07001023{
1024 if (atomic_read(&sk->sk_refcnt) != 1)
1025 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
1026 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
1027}
1028#else /* SOCK_REFCNT_DEBUG */
1029#define sk_refcnt_debug_inc(sk) do { } while (0)
1030#define sk_refcnt_debug_dec(sk) do { } while (0)
1031#define sk_refcnt_debug_release(sk) do { } while (0)
1032#endif /* SOCK_REFCNT_DEBUG */
1033
Andrew Mortonc255a452012-07-31 16:43:02 -07001034#if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_NET)
Ingo Molnarc5905af2012-02-24 08:31:31 +01001035extern struct static_key memcg_socket_limit_enabled;
Glauber Costae1aab162011-12-11 21:47:03 +00001036static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1037 struct cg_proto *cg_proto)
1038{
1039 return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg));
1040}
Ingo Molnarc5905af2012-02-24 08:31:31 +01001041#define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
Glauber Costae1aab162011-12-11 21:47:03 +00001042#else
1043#define mem_cgroup_sockets_enabled 0
1044static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1045 struct cg_proto *cg_proto)
1046{
1047 return NULL;
1048}
1049#endif
1050
1051
Glauber Costa180d8cd2011-12-11 21:47:02 +00001052static inline bool sk_has_memory_pressure(const struct sock *sk)
1053{
1054 return sk->sk_prot->memory_pressure != NULL;
1055}
1056
1057static inline bool sk_under_memory_pressure(const struct sock *sk)
1058{
1059 if (!sk->sk_prot->memory_pressure)
1060 return false;
Glauber Costae1aab162011-12-11 21:47:03 +00001061
1062 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1063 return !!*sk->sk_cgrp->memory_pressure;
1064
Glauber Costa180d8cd2011-12-11 21:47:02 +00001065 return !!*sk->sk_prot->memory_pressure;
1066}
1067
1068static inline void sk_leave_memory_pressure(struct sock *sk)
1069{
1070 int *memory_pressure = sk->sk_prot->memory_pressure;
1071
Glauber Costae1aab162011-12-11 21:47:03 +00001072 if (!memory_pressure)
1073 return;
1074
1075 if (*memory_pressure)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001076 *memory_pressure = 0;
Glauber Costae1aab162011-12-11 21:47:03 +00001077
1078 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1079 struct cg_proto *cg_proto = sk->sk_cgrp;
1080 struct proto *prot = sk->sk_prot;
1081
1082 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1083 if (*cg_proto->memory_pressure)
1084 *cg_proto->memory_pressure = 0;
1085 }
1086
Glauber Costa180d8cd2011-12-11 21:47:02 +00001087}
1088
1089static inline void sk_enter_memory_pressure(struct sock *sk)
1090{
Glauber Costae1aab162011-12-11 21:47:03 +00001091 if (!sk->sk_prot->enter_memory_pressure)
1092 return;
1093
1094 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1095 struct cg_proto *cg_proto = sk->sk_cgrp;
1096 struct proto *prot = sk->sk_prot;
1097
1098 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1099 cg_proto->enter_memory_pressure(sk);
1100 }
1101
1102 sk->sk_prot->enter_memory_pressure(sk);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001103}
1104
1105static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1106{
1107 long *prot = sk->sk_prot->sysctl_mem;
Glauber Costae1aab162011-12-11 21:47:03 +00001108 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1109 prot = sk->sk_cgrp->sysctl_mem;
Glauber Costa180d8cd2011-12-11 21:47:02 +00001110 return prot[index];
1111}
1112
Glauber Costae1aab162011-12-11 21:47:03 +00001113static inline void memcg_memory_allocated_add(struct cg_proto *prot,
1114 unsigned long amt,
1115 int *parent_status)
1116{
1117 struct res_counter *fail;
1118 int ret;
1119
Glauber Costa0e90b312012-01-20 04:57:16 +00001120 ret = res_counter_charge_nofail(prot->memory_allocated,
1121 amt << PAGE_SHIFT, &fail);
Glauber Costae1aab162011-12-11 21:47:03 +00001122 if (ret < 0)
1123 *parent_status = OVER_LIMIT;
1124}
1125
1126static inline void memcg_memory_allocated_sub(struct cg_proto *prot,
1127 unsigned long amt)
1128{
1129 res_counter_uncharge(prot->memory_allocated, amt << PAGE_SHIFT);
1130}
1131
1132static inline u64 memcg_memory_allocated_read(struct cg_proto *prot)
1133{
1134 u64 ret;
1135 ret = res_counter_read_u64(prot->memory_allocated, RES_USAGE);
1136 return ret >> PAGE_SHIFT;
1137}
1138
Glauber Costa180d8cd2011-12-11 21:47:02 +00001139static inline long
1140sk_memory_allocated(const struct sock *sk)
1141{
1142 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001143 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1144 return memcg_memory_allocated_read(sk->sk_cgrp);
1145
Glauber Costa180d8cd2011-12-11 21:47:02 +00001146 return atomic_long_read(prot->memory_allocated);
1147}
1148
1149static inline long
Glauber Costae1aab162011-12-11 21:47:03 +00001150sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001151{
1152 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001153
1154 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1155 memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status);
1156 /* update the root cgroup regardless */
1157 atomic_long_add_return(amt, prot->memory_allocated);
1158 return memcg_memory_allocated_read(sk->sk_cgrp);
1159 }
1160
Glauber Costa180d8cd2011-12-11 21:47:02 +00001161 return atomic_long_add_return(amt, prot->memory_allocated);
1162}
1163
1164static inline void
Glauber Costa0e90b312012-01-20 04:57:16 +00001165sk_memory_allocated_sub(struct sock *sk, int amt)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001166{
1167 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001168
Glauber Costa0e90b312012-01-20 04:57:16 +00001169 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
Glauber Costae1aab162011-12-11 21:47:03 +00001170 memcg_memory_allocated_sub(sk->sk_cgrp, amt);
1171
Glauber Costa180d8cd2011-12-11 21:47:02 +00001172 atomic_long_sub(amt, prot->memory_allocated);
1173}
1174
1175static inline void sk_sockets_allocated_dec(struct sock *sk)
1176{
1177 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001178
1179 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1180 struct cg_proto *cg_proto = sk->sk_cgrp;
1181
1182 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1183 percpu_counter_dec(cg_proto->sockets_allocated);
1184 }
1185
Glauber Costa180d8cd2011-12-11 21:47:02 +00001186 percpu_counter_dec(prot->sockets_allocated);
1187}
1188
1189static inline void sk_sockets_allocated_inc(struct sock *sk)
1190{
1191 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001192
1193 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1194 struct cg_proto *cg_proto = sk->sk_cgrp;
1195
1196 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1197 percpu_counter_inc(cg_proto->sockets_allocated);
1198 }
1199
Glauber Costa180d8cd2011-12-11 21:47:02 +00001200 percpu_counter_inc(prot->sockets_allocated);
1201}
1202
1203static inline int
1204sk_sockets_allocated_read_positive(struct sock *sk)
1205{
1206 struct proto *prot = sk->sk_prot;
1207
Glauber Costae1aab162011-12-11 21:47:03 +00001208 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
Eric Dumazet518fbf92012-04-28 23:21:56 +00001209 return percpu_counter_read_positive(sk->sk_cgrp->sockets_allocated);
Glauber Costae1aab162011-12-11 21:47:03 +00001210
Eric Dumazet518fbf92012-04-28 23:21:56 +00001211 return percpu_counter_read_positive(prot->sockets_allocated);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001212}
1213
1214static inline int
1215proto_sockets_allocated_sum_positive(struct proto *prot)
1216{
1217 return percpu_counter_sum_positive(prot->sockets_allocated);
1218}
1219
1220static inline long
1221proto_memory_allocated(struct proto *prot)
1222{
1223 return atomic_long_read(prot->memory_allocated);
1224}
1225
1226static inline bool
1227proto_memory_pressure(struct proto *prot)
1228{
1229 if (!prot->memory_pressure)
1230 return false;
1231 return !!*prot->memory_pressure;
1232}
1233
Eric Dumazet65f76512008-01-03 20:46:48 -08001234
1235#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001236/* Called with local bh disabled */
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -07001237extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1238extern int sock_prot_inuse_get(struct net *net, struct proto *proto);
Eric Dumazet65f76512008-01-03 20:46:48 -08001239#else
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001240static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -07001241 int inc)
Eric Dumazet65f76512008-01-03 20:46:48 -08001242{
1243}
Eric Dumazet65f76512008-01-03 20:46:48 -08001244#endif
1245
Linus Torvalds1da177e2005-04-16 15:20:36 -07001246
Arnaldo Carvalho de Melo614c6cb2005-08-09 19:47:37 -07001247/* With per-bucket locks this operation is not-atomic, so that
1248 * this version is not worse.
1249 */
1250static inline void __sk_prot_rehash(struct sock *sk)
1251{
1252 sk->sk_prot->unhash(sk);
1253 sk->sk_prot->hash(sk);
1254}
1255
Octavian Purdilafcbdf092010-12-16 14:26:56 -08001256void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
1257
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258/* About 10 seconds */
1259#define SOCK_DESTROY_TIME (10*HZ)
1260
1261/* Sockets 0-1023 can't be bound to unless you are superuser */
1262#define PROT_SOCK 1024
1263
1264#define SHUTDOWN_MASK 3
1265#define RCV_SHUTDOWN 1
1266#define SEND_SHUTDOWN 2
1267
1268#define SOCK_SNDBUF_LOCK 1
1269#define SOCK_RCVBUF_LOCK 2
1270#define SOCK_BINDADDR_LOCK 4
1271#define SOCK_BINDPORT_LOCK 8
1272
1273/* sock_iocb: used to kick off async processing of socket ios */
1274struct sock_iocb {
1275 struct list_head list;
1276
1277 int flags;
1278 int size;
1279 struct socket *sock;
1280 struct sock *sk;
1281 struct scm_cookie *scm;
1282 struct msghdr *msg, async_msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283 struct kiocb *kiocb;
1284};
1285
1286static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
1287{
1288 return (struct sock_iocb *)iocb->private;
1289}
1290
1291static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
1292{
1293 return si->kiocb;
1294}
1295
1296struct socket_alloc {
1297 struct socket socket;
1298 struct inode vfs_inode;
1299};
1300
1301static inline struct socket *SOCKET_I(struct inode *inode)
1302{
1303 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1304}
1305
1306static inline struct inode *SOCK_INODE(struct socket *socket)
1307{
1308 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1309}
1310
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001311/*
1312 * Functions for memory accounting
1313 */
1314extern int __sk_mem_schedule(struct sock *sk, int size, int kind);
1315extern void __sk_mem_reclaim(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001317#define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1318#define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1319#define SK_MEM_SEND 0
1320#define SK_MEM_RECV 1
Linus Torvalds1da177e2005-04-16 15:20:36 -07001321
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001322static inline int sk_mem_pages(int amt)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001324 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325}
1326
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001327static inline bool sk_has_account(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001329 /* return true if protocol supports memory accounting */
1330 return !!sk->sk_prot->memory_allocated;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001331}
1332
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001333static inline bool sk_wmem_schedule(struct sock *sk, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001334{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001335 if (!sk_has_account(sk))
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001336 return true;
Herbert Xud80d99d2005-09-01 17:48:23 -07001337 return size <= sk->sk_forward_alloc ||
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001338 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1339}
1340
Mel Gormanc76562b2012-07-31 16:44:41 -07001341static inline bool
Chuck Lever35c448a2012-09-17 14:09:11 -07001342sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size)
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001343{
1344 if (!sk_has_account(sk))
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001345 return true;
Mel Gormanc76562b2012-07-31 16:44:41 -07001346 return size<= sk->sk_forward_alloc ||
1347 __sk_mem_schedule(sk, size, SK_MEM_RECV) ||
1348 skb_pfmemalloc(skb);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001349}
1350
1351static inline void sk_mem_reclaim(struct sock *sk)
1352{
1353 if (!sk_has_account(sk))
1354 return;
1355 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1356 __sk_mem_reclaim(sk);
1357}
1358
David S. Miller9993e7d2008-01-10 21:56:38 -08001359static inline void sk_mem_reclaim_partial(struct sock *sk)
1360{
1361 if (!sk_has_account(sk))
1362 return;
1363 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1364 __sk_mem_reclaim(sk);
1365}
1366
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001367static inline void sk_mem_charge(struct sock *sk, int size)
1368{
1369 if (!sk_has_account(sk))
1370 return;
1371 sk->sk_forward_alloc -= size;
1372}
1373
1374static inline void sk_mem_uncharge(struct sock *sk, int size)
1375{
1376 if (!sk_has_account(sk))
1377 return;
1378 sk->sk_forward_alloc += size;
1379}
1380
1381static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1382{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001383 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1384 sk->sk_wmem_queued -= skb->truesize;
1385 sk_mem_uncharge(sk, skb->truesize);
1386 __kfree_skb(skb);
Herbert Xud80d99d2005-09-01 17:48:23 -07001387}
1388
Linus Torvalds1da177e2005-04-16 15:20:36 -07001389/* Used by processes to "lock" a socket state, so that
1390 * interrupts and bottom half handlers won't change it
1391 * from under us. It essentially blocks any incoming
1392 * packets, so that we won't get any new data or any
1393 * packets that change the state of the socket.
1394 *
1395 * While locked, BH processing will add new packets to
1396 * the backlog queue. This queue is processed by the
1397 * owner of the socket lock right before it is released.
1398 *
1399 * Since ~2.3.5 it is also exclusive sleep lock serializing
1400 * accesses from user process context.
1401 */
John Heffnerd2e91172007-09-12 10:44:19 +02001402#define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001403
Peter Zijlstraed075362006-12-06 20:35:24 -08001404/*
1405 * Macro so as to not evaluate some arguments when
1406 * lockdep is not enabled.
1407 *
1408 * Mark both the sk_lock and the sk_lock.slock as a
1409 * per-address-family lock class.
1410 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001411#define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
Peter Zijlstraed075362006-12-06 20:35:24 -08001412do { \
Ingo Molnare8f6fbf2008-11-12 01:38:36 +00001413 sk->sk_lock.owned = 0; \
Peter Zijlstraed075362006-12-06 20:35:24 -08001414 init_waitqueue_head(&sk->sk_lock.wq); \
1415 spin_lock_init(&(sk)->sk_lock.slock); \
1416 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1417 sizeof((sk)->sk_lock)); \
1418 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001419 (skey), (sname)); \
Peter Zijlstraed075362006-12-06 20:35:24 -08001420 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1421} while (0)
1422
Harvey Harrison41380932007-12-12 10:46:51 -08001423extern void lock_sock_nested(struct sock *sk, int subclass);
Peter Zijlstrafcc70d52006-11-08 22:44:35 -08001424
1425static inline void lock_sock(struct sock *sk)
1426{
1427 lock_sock_nested(sk, 0);
1428}
1429
Harvey Harrison41380932007-12-12 10:46:51 -08001430extern void release_sock(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431
1432/* BH context may only use the following locking interface. */
1433#define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
Ingo Molnarc6366182006-07-03 00:25:13 -07001434#define bh_lock_sock_nested(__sk) \
1435 spin_lock_nested(&((__sk)->sk_lock.slock), \
1436 SINGLE_DEPTH_NESTING)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437#define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1438
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001439extern bool lock_sock_fast(struct sock *sk);
1440/**
1441 * unlock_sock_fast - complement of lock_sock_fast
1442 * @sk: socket
1443 * @slow: slow mode
1444 *
1445 * fast unlock socket for user context.
1446 * If slow mode is on, we call regular release_sock()
1447 */
1448static inline void unlock_sock_fast(struct sock *sk, bool slow)
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001449{
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001450 if (slow)
1451 release_sock(sk);
1452 else
1453 spin_unlock_bh(&sk->sk_lock.slock);
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001454}
1455
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001456
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001457extern struct sock *sk_alloc(struct net *net, int family,
Al Virodd0fc662005-10-07 07:46:04 +01001458 gfp_t priority,
Pavel Emelyanov6257ff22007-11-01 00:39:31 -07001459 struct proto *prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460extern void sk_free(struct sock *sk);
Denis V. Lunevedf02082008-02-29 11:18:32 -08001461extern void sk_release_kernel(struct sock *sk);
Eric Dumazete56c57d2011-11-08 17:07:07 -05001462extern struct sock *sk_clone_lock(const struct sock *sk,
1463 const gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001464
1465extern struct sk_buff *sock_wmalloc(struct sock *sk,
1466 unsigned long size, int force,
Al Virodd0fc662005-10-07 07:46:04 +01001467 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001468extern struct sk_buff *sock_rmalloc(struct sock *sk,
1469 unsigned long size, int force,
Al Virodd0fc662005-10-07 07:46:04 +01001470 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001471extern void sock_wfree(struct sk_buff *skb);
1472extern void sock_rfree(struct sk_buff *skb);
David S. Miller41063e92012-06-19 21:22:05 -07001473extern void sock_edemux(struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474
1475extern int sock_setsockopt(struct socket *sock, int level,
1476 int op, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07001477 unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478
1479extern int sock_getsockopt(struct socket *sock, int level,
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001480 int op, char __user *optval,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481 int __user *optlen);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001482extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483 unsigned long size,
1484 int noblock,
1485 int *errcode);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001486extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
Herbert Xu4cc7f682009-02-04 16:55:54 -08001487 unsigned long header_len,
1488 unsigned long data_len,
1489 int noblock,
1490 int *errcode);
Victor Fusco86a76ca2005-07-08 14:57:47 -07001491extern void *sock_kmalloc(struct sock *sk, int size,
Al Virodd0fc662005-10-07 07:46:04 +01001492 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493extern void sock_kfree_s(struct sock *sk, void *mem, int size);
1494extern void sk_send_sigurg(struct sock *sk);
1495
1496/*
1497 * Functions to fill in entries in struct proto_ops when a protocol
1498 * does not implement a particular function.
1499 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001500extern int sock_no_bind(struct socket *,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001501 struct sockaddr *, int);
1502extern int sock_no_connect(struct socket *,
1503 struct sockaddr *, int, int);
1504extern int sock_no_socketpair(struct socket *,
1505 struct socket *);
1506extern int sock_no_accept(struct socket *,
1507 struct socket *, int);
1508extern int sock_no_getname(struct socket *,
1509 struct sockaddr *, int *, int);
1510extern unsigned int sock_no_poll(struct file *, struct socket *,
1511 struct poll_table_struct *);
1512extern int sock_no_ioctl(struct socket *, unsigned int,
1513 unsigned long);
1514extern int sock_no_listen(struct socket *, int);
1515extern int sock_no_shutdown(struct socket *, int);
1516extern int sock_no_getsockopt(struct socket *, int , int,
1517 char __user *, int __user *);
1518extern int sock_no_setsockopt(struct socket *, int, int,
David S. Millerb7058842009-09-30 16:12:20 -07001519 char __user *, unsigned int);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520extern int sock_no_sendmsg(struct kiocb *, struct socket *,
1521 struct msghdr *, size_t);
1522extern int sock_no_recvmsg(struct kiocb *, struct socket *,
1523 struct msghdr *, size_t, int);
1524extern int sock_no_mmap(struct file *file,
1525 struct socket *sock,
1526 struct vm_area_struct *vma);
1527extern ssize_t sock_no_sendpage(struct socket *sock,
1528 struct page *page,
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001529 int offset, size_t size,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001530 int flags);
1531
1532/*
1533 * Functions to fill in entries in struct proto_ops when a protocol
1534 * uses the inet style.
1535 */
1536extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
1537 char __user *optval, int __user *optlen);
1538extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1539 struct msghdr *msg, size_t size, int flags);
1540extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07001541 char __user *optval, unsigned int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001542extern int compat_sock_common_getsockopt(struct socket *sock, int level,
1543 int optname, char __user *optval, int __user *optlen);
1544extern int compat_sock_common_setsockopt(struct socket *sock, int level,
David S. Millerb7058842009-09-30 16:12:20 -07001545 int optname, char __user *optval, unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001546
1547extern void sk_common_release(struct sock *sk);
1548
1549/*
1550 * Default socket callbacks and setup code
1551 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001552
Linus Torvalds1da177e2005-04-16 15:20:36 -07001553/* Initialise core socket variables */
1554extern void sock_init_data(struct socket *sock, struct sock *sk);
1555
Eric Dumazet46bcf142010-12-06 09:29:43 -08001556extern void sk_filter_release_rcu(struct rcu_head *rcu);
1557
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001559 * sk_filter_release - release a socket filter
Paul Bonserdc9b3342006-11-23 17:56:13 -08001560 * @fp: filter to remove
1561 *
1562 * Remove a filter from a socket and release its resources.
1563 */
1564
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -07001565static inline void sk_filter_release(struct sk_filter *fp)
1566{
1567 if (atomic_dec_and_test(&fp->refcnt))
Eric Dumazet80f8f102011-01-18 07:46:52 +00001568 call_rcu(&fp->rcu, sk_filter_release_rcu);
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -07001569}
1570
1571static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001572{
1573 unsigned int size = sk_filter_len(fp);
1574
1575 atomic_sub(size, &sk->sk_omem_alloc);
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -07001576 sk_filter_release(fp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001577}
1578
1579static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
1580{
1581 atomic_inc(&fp->refcnt);
1582 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
1583}
1584
1585/*
1586 * Socket reference counting postulates.
1587 *
1588 * * Each user of socket SHOULD hold a reference count.
1589 * * Each access point to socket (an hash table bucket, reference from a list,
1590 * running timer, skb in flight MUST hold a reference count.
1591 * * When reference count hits 0, it means it will never increase back.
1592 * * When reference count hits 0, it means that no references from
1593 * outside exist to this socket and current process on current CPU
1594 * is last user and may/should destroy this socket.
1595 * * sk_free is called from any context: process, BH, IRQ. When
1596 * it is called, socket has no references from outside -> sk_free
1597 * may release descendant resources allocated by the socket, but
1598 * to the time when it is called, socket is NOT referenced by any
1599 * hash tables, lists etc.
1600 * * Packets, delivered from outside (from network or from another process)
1601 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1602 * when they sit in queue. Otherwise, packets will leak to hole, when
1603 * socket is looked up by one cpu and unhasing is made by another CPU.
1604 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1605 * (leak to backlog). Packet socket does all the processing inside
1606 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1607 * use separate SMP lock, so that they are prone too.
1608 */
1609
1610/* Ungrab socket and destroy it, if it was the last reference. */
1611static inline void sock_put(struct sock *sk)
1612{
1613 if (atomic_dec_and_test(&sk->sk_refcnt))
1614 sk_free(sk);
1615}
1616
Arnaldo Carvalho de Melo58a5a7b2006-11-16 14:06:06 -02001617extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1618 const int nested);
Arnaldo Carvalho de Melo25995ff2005-12-27 02:42:22 -02001619
Krishna Kumare022f0b2009-10-19 23:46:20 +00001620static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1621{
1622 sk->sk_tx_queue_mapping = tx_queue;
1623}
1624
1625static inline void sk_tx_queue_clear(struct sock *sk)
1626{
1627 sk->sk_tx_queue_mapping = -1;
1628}
1629
1630static inline int sk_tx_queue_get(const struct sock *sk)
1631{
Tom Herbertb0f77d02010-07-14 20:50:29 -07001632 return sk ? sk->sk_tx_queue_mapping : -1;
Krishna Kumare022f0b2009-10-19 23:46:20 +00001633}
1634
David S. Miller972692e2008-06-17 22:41:38 -07001635static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1636{
Krishna Kumare022f0b2009-10-19 23:46:20 +00001637 sk_tx_queue_clear(sk);
David S. Miller972692e2008-06-17 22:41:38 -07001638 sk->sk_socket = sock;
1639}
1640
Eric Dumazetaa395142010-04-20 13:03:51 +00001641static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1642{
Eric Dumazeteaefd112011-02-18 03:26:36 +00001643 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1644 return &rcu_dereference_raw(sk->sk_wq)->wait;
Eric Dumazetaa395142010-04-20 13:03:51 +00001645}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001646/* Detach socket from process context.
1647 * Announce socket dead, detach it from wait queue and inode.
1648 * Note that parent inode held reference count on this struct sock,
1649 * we do not release it in this function, because protocol
1650 * probably wants some additional cleanups or even continuing
1651 * to work with this socket (TCP).
1652 */
1653static inline void sock_orphan(struct sock *sk)
1654{
1655 write_lock_bh(&sk->sk_callback_lock);
1656 sock_set_flag(sk, SOCK_DEAD);
David S. Miller972692e2008-06-17 22:41:38 -07001657 sk_set_socket(sk, NULL);
Eric Dumazet43815482010-04-29 11:01:49 +00001658 sk->sk_wq = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001659 write_unlock_bh(&sk->sk_callback_lock);
1660}
1661
1662static inline void sock_graft(struct sock *sk, struct socket *parent)
1663{
1664 write_lock_bh(&sk->sk_callback_lock);
Eric Dumazeteaefd112011-02-18 03:26:36 +00001665 sk->sk_wq = parent->wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001666 parent->sk = sk;
David S. Miller972692e2008-06-17 22:41:38 -07001667 sk_set_socket(sk, parent);
Venkat Yekkirala4237c752006-07-24 23:32:50 -07001668 security_sock_graft(sk, parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669 write_unlock_bh(&sk->sk_callback_lock);
1670}
1671
Eric W. Biederman976d02012012-05-23 17:16:53 -06001672extern kuid_t sock_i_uid(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001673extern unsigned long sock_i_ino(struct sock *sk);
1674
1675static inline struct dst_entry *
1676__sk_dst_get(struct sock *sk)
1677{
Michal Hockod8bf4ca2011-07-08 14:39:41 +02001678 return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
Eric Dumazetf68c2242010-04-22 16:06:59 -07001679 lockdep_is_held(&sk->sk_lock.slock));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001680}
1681
1682static inline struct dst_entry *
1683sk_dst_get(struct sock *sk)
1684{
1685 struct dst_entry *dst;
1686
Eric Dumazetb6c67122010-04-08 23:03:29 +00001687 rcu_read_lock();
1688 dst = rcu_dereference(sk->sk_dst_cache);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001689 if (dst)
1690 dst_hold(dst);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001691 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001692 return dst;
1693}
1694
Eric Dumazetb6c67122010-04-08 23:03:29 +00001695extern void sk_reset_txq(struct sock *sk);
1696
1697static inline void dst_negative_advice(struct sock *sk)
1698{
1699 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1700
1701 if (dst && dst->ops->negative_advice) {
1702 ndst = dst->ops->negative_advice(dst);
1703
1704 if (ndst != dst) {
1705 rcu_assign_pointer(sk->sk_dst_cache, ndst);
1706 sk_reset_txq(sk);
1707 }
1708 }
1709}
1710
Linus Torvalds1da177e2005-04-16 15:20:36 -07001711static inline void
1712__sk_dst_set(struct sock *sk, struct dst_entry *dst)
1713{
1714 struct dst_entry *old_dst;
1715
Krishna Kumare022f0b2009-10-19 23:46:20 +00001716 sk_tx_queue_clear(sk);
Eric Dumazet0b53ff22010-04-26 20:40:43 +00001717 /*
1718 * This can be called while sk is owned by the caller only,
1719 * with no state that can be checked in a rcu_dereference_check() cond
1720 */
1721 old_dst = rcu_dereference_raw(sk->sk_dst_cache);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001722 rcu_assign_pointer(sk->sk_dst_cache, dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001723 dst_release(old_dst);
1724}
1725
1726static inline void
1727sk_dst_set(struct sock *sk, struct dst_entry *dst)
1728{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001729 spin_lock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001730 __sk_dst_set(sk, dst);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001731 spin_unlock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001732}
1733
1734static inline void
1735__sk_dst_reset(struct sock *sk)
1736{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001737 __sk_dst_set(sk, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001738}
1739
1740static inline void
1741sk_dst_reset(struct sock *sk)
1742{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001743 spin_lock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744 __sk_dst_reset(sk);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001745 spin_unlock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746}
1747
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001748extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001749
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001750extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001752static inline bool sk_can_gso(const struct sock *sk)
Herbert Xubcd76112006-06-30 13:36:35 -07001753{
1754 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1755}
1756
Andi Kleen99580892007-04-20 17:12:43 -07001757extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
Arnaldo Carvalho de Melo6cbb0df2005-08-09 19:49:02 -07001758
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001759static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
Eric Dumazeta4654192010-05-16 00:36:33 -07001760{
1761 sk->sk_route_nocaps |= flags;
1762 sk->sk_route_caps &= ~flags;
1763}
1764
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001765static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1766 char __user *from, char *to,
Wei Yongjun912d3982011-04-06 18:40:12 +00001767 int copy, int offset)
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001768{
1769 if (skb->ip_summed == CHECKSUM_NONE) {
1770 int err = 0;
1771 __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err);
1772 if (err)
1773 return err;
Wei Yongjun912d3982011-04-06 18:40:12 +00001774 skb->csum = csum_block_add(skb->csum, csum, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001775 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1776 if (!access_ok(VERIFY_READ, from, copy) ||
1777 __copy_from_user_nocache(to, from, copy))
1778 return -EFAULT;
1779 } else if (copy_from_user(to, from, copy))
1780 return -EFAULT;
1781
1782 return 0;
1783}
1784
1785static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
1786 char __user *from, int copy)
1787{
Wei Yongjun912d3982011-04-06 18:40:12 +00001788 int err, offset = skb->len;
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001789
Wei Yongjun912d3982011-04-06 18:40:12 +00001790 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1791 copy, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001792 if (err)
Wei Yongjun912d3982011-04-06 18:40:12 +00001793 __skb_trim(skb, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001794
1795 return err;
1796}
1797
1798static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from,
1799 struct sk_buff *skb,
1800 struct page *page,
1801 int off, int copy)
1802{
1803 int err;
1804
Wei Yongjun912d3982011-04-06 18:40:12 +00001805 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1806 copy, skb->len);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001807 if (err)
1808 return err;
1809
1810 skb->len += copy;
1811 skb->data_len += copy;
1812 skb->truesize += copy;
1813 sk->sk_wmem_queued += copy;
1814 sk_mem_charge(sk, copy);
1815 return 0;
1816}
1817
Linus Torvalds1da177e2005-04-16 15:20:36 -07001818static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1819 struct sk_buff *skb, struct page *page,
1820 int off, int copy)
1821{
1822 if (skb->ip_summed == CHECKSUM_NONE) {
1823 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08001824 __wsum csum = csum_and_copy_from_user(from,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001825 page_address(page) + off,
1826 copy, 0, &err);
1827 if (err)
1828 return err;
1829 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1830 } else if (copy_from_user(page_address(page) + off, from, copy))
1831 return -EFAULT;
1832
1833 skb->len += copy;
1834 skb->data_len += copy;
1835 skb->truesize += copy;
1836 sk->sk_wmem_queued += copy;
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001837 sk_mem_charge(sk, copy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001838 return 0;
1839}
1840
Eric Dumazetc5640392009-06-16 10:12:03 +00001841/**
1842 * sk_wmem_alloc_get - returns write allocations
1843 * @sk: socket
1844 *
1845 * Returns sk_wmem_alloc minus initial offset of one
1846 */
1847static inline int sk_wmem_alloc_get(const struct sock *sk)
1848{
1849 return atomic_read(&sk->sk_wmem_alloc) - 1;
1850}
1851
1852/**
1853 * sk_rmem_alloc_get - returns read allocations
1854 * @sk: socket
1855 *
1856 * Returns sk_rmem_alloc
1857 */
1858static inline int sk_rmem_alloc_get(const struct sock *sk)
1859{
1860 return atomic_read(&sk->sk_rmem_alloc);
1861}
1862
1863/**
1864 * sk_has_allocations - check if allocations are outstanding
1865 * @sk: socket
1866 *
1867 * Returns true if socket has write or read allocations
1868 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001869static inline bool sk_has_allocations(const struct sock *sk)
Eric Dumazetc5640392009-06-16 10:12:03 +00001870{
1871 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1872}
1873
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001874/**
Eric Dumazet43815482010-04-29 11:01:49 +00001875 * wq_has_sleeper - check if there are any waiting processes
Randy Dunlapacfbe962010-05-24 23:54:18 -07001876 * @wq: struct socket_wq
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001877 *
Eric Dumazet43815482010-04-29 11:01:49 +00001878 * Returns true if socket_wq has waiting processes
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001879 *
Eric Dumazet43815482010-04-29 11:01:49 +00001880 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001881 * barrier call. They were added due to the race found within the tcp code.
1882 *
1883 * Consider following tcp code paths:
1884 *
1885 * CPU1 CPU2
1886 *
1887 * sys_select receive packet
1888 * ... ...
1889 * __add_wait_queue update tp->rcv_nxt
1890 * ... ...
1891 * tp->rcv_nxt check sock_def_readable
1892 * ... {
Eric Dumazet43815482010-04-29 11:01:49 +00001893 * schedule rcu_read_lock();
1894 * wq = rcu_dereference(sk->sk_wq);
1895 * if (wq && waitqueue_active(&wq->wait))
1896 * wake_up_interruptible(&wq->wait)
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001897 * ...
1898 * }
1899 *
1900 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1901 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1902 * could then endup calling schedule and sleep forever if there are no more
1903 * data on the socket.
Jiri Olsaad462762009-07-08 12:10:31 +00001904 *
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001905 */
Eric Dumazet43815482010-04-29 11:01:49 +00001906static inline bool wq_has_sleeper(struct socket_wq *wq)
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001907{
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001908 /* We need to be sure we are in sync with the
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001909 * add_wait_queue modifications to the wait queue.
1910 *
1911 * This memory barrier is paired in the sock_poll_wait.
1912 */
Eric Dumazet43815482010-04-29 11:01:49 +00001913 smp_mb();
1914 return wq && waitqueue_active(&wq->wait);
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001915}
1916
1917/**
1918 * sock_poll_wait - place memory barrier behind the poll_wait call.
1919 * @filp: file
1920 * @wait_address: socket wait queue
1921 * @p: poll_table
1922 *
Eric Dumazet43815482010-04-29 11:01:49 +00001923 * See the comments in the wq_has_sleeper function.
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001924 */
1925static inline void sock_poll_wait(struct file *filp,
1926 wait_queue_head_t *wait_address, poll_table *p)
1927{
Hans Verkuil626cf232012-03-23 15:02:27 -07001928 if (!poll_does_not_wait(p) && wait_address) {
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001929 poll_wait(filp, wait_address, p);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001930 /* We need to be sure we are in sync with the
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001931 * socket flags modification.
1932 *
Eric Dumazet43815482010-04-29 11:01:49 +00001933 * This memory barrier is paired in the wq_has_sleeper.
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001934 */
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001935 smp_mb();
1936 }
1937}
1938
Linus Torvalds1da177e2005-04-16 15:20:36 -07001939/*
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001940 * Queue a received datagram if it will fit. Stream and sequenced
Linus Torvalds1da177e2005-04-16 15:20:36 -07001941 * protocols can't normally use this as they need to fit buffers in
1942 * and play with them.
1943 *
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001944 * Inlined as it's very short and called for pretty much every
Linus Torvalds1da177e2005-04-16 15:20:36 -07001945 * packet ever received.
1946 */
1947
1948static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1949{
Herbert Xud55d87f2009-06-22 02:25:25 +00001950 skb_orphan(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001951 skb->sk = sk;
1952 skb->destructor = sock_wfree;
Eric Dumazet2b85a342009-06-11 02:55:43 -07001953 /*
1954 * We used to take a refcount on sk, but following operation
1955 * is enough to guarantee sk_free() wont free this sock until
1956 * all in-flight packets are completed
1957 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001958 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1959}
1960
1961static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1962{
Herbert Xud55d87f2009-06-22 02:25:25 +00001963 skb_orphan(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001964 skb->sk = sk;
1965 skb->destructor = sock_rfree;
1966 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001967 sk_mem_charge(sk, skb->truesize);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001968}
1969
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001970extern void sk_reset_timer(struct sock *sk, struct timer_list *timer,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001971 unsigned long expires);
1972
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001973extern void sk_stop_timer(struct sock *sk, struct timer_list *timer);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001974
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001975extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001976
Eric Dumazetb1faf562010-05-31 23:44:05 -07001977extern int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001978
1979/*
1980 * Recover an error report and clear atomically
1981 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001982
Linus Torvalds1da177e2005-04-16 15:20:36 -07001983static inline int sock_error(struct sock *sk)
1984{
Benjamin LaHaisec1cbe4b2005-12-13 23:22:19 -08001985 int err;
1986 if (likely(!sk->sk_err))
1987 return 0;
1988 err = xchg(&sk->sk_err, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001989 return -err;
1990}
1991
1992static inline unsigned long sock_wspace(struct sock *sk)
1993{
1994 int amt = 0;
1995
1996 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1997 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001998 if (amt < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001999 amt = 0;
2000 }
2001 return amt;
2002}
2003
2004static inline void sk_wake_async(struct sock *sk, int how, int band)
2005{
Eric Dumazetbcdce712009-10-06 17:28:29 -07002006 if (sock_flag(sk, SOCK_FASYNC))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002007 sock_wake_async(sk->sk_socket, how, band);
2008}
2009
2010#define SOCK_MIN_SNDBUF 2048
Eric Dumazet7a91b4342010-09-26 18:53:07 -07002011/*
2012 * Since sk_rmem_alloc sums skb->truesize, even a small frame might need
2013 * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak
2014 */
2015#define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002016
2017static inline void sk_stream_moderate_sndbuf(struct sock *sk)
2018{
2019 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
Eric Dumazet8df09ea2007-12-21 03:07:41 -08002020 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
2022 }
2023}
2024
Pavel Emelyanovdf97c702007-11-29 21:22:33 +11002025struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002026
Eric Dumazet5640f762012-09-23 23:04:42 +00002027/**
2028 * sk_page_frag - return an appropriate page_frag
2029 * @sk: socket
2030 *
2031 * If socket allocation mode allows current thread to sleep, it means its
2032 * safe to use the per task page_frag instead of the per socket one.
2033 */
2034static inline struct page_frag *sk_page_frag(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002035{
Eric Dumazet5640f762012-09-23 23:04:42 +00002036 if (sk->sk_allocation & __GFP_WAIT)
2037 return &current->task_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002038
Eric Dumazet5640f762012-09-23 23:04:42 +00002039 return &sk->sk_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002040}
2041
Eric Dumazet5640f762012-09-23 23:04:42 +00002042extern bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag);
2043
Linus Torvalds1da177e2005-04-16 15:20:36 -07002044/*
2045 * Default write policy as shown to user space via poll/select/SIGIO
2046 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002047static inline bool sock_writeable(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002048{
Eric Dumazet8df09ea2007-12-21 03:07:41 -08002049 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002050}
2051
Al Virodd0fc662005-10-07 07:46:04 +01002052static inline gfp_t gfp_any(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002053{
Andrew Morton99709372009-02-12 16:43:17 -08002054 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002055}
2056
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002057static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002058{
2059 return noblock ? 0 : sk->sk_rcvtimeo;
2060}
2061
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002062static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002063{
2064 return noblock ? 0 : sk->sk_sndtimeo;
2065}
2066
2067static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2068{
2069 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
2070}
2071
2072/* Alas, with timeout socket operations are not restartable.
2073 * Compare this to poll().
2074 */
2075static inline int sock_intr_errno(long timeo)
2076{
2077 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2078}
2079
Eric Dumazet92f37fd2007-03-25 22:14:49 -07002080extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2081 struct sk_buff *skb);
Johannes Berg6e3e9392011-11-09 10:15:42 +01002082extern void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2083 struct sk_buff *skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -07002084
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002085static inline void
Linus Torvalds1da177e2005-04-16 15:20:36 -07002086sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2087{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002088 ktime_t kt = skb->tstamp;
Patrick Ohly20d49472009-02-12 05:03:38 +00002089 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002090
Patrick Ohly20d49472009-02-12 05:03:38 +00002091 /*
2092 * generate control messages if
2093 * - receive time stamping in software requested (SOCK_RCVTSTAMP
2094 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
2095 * - software time stamp available and wanted
2096 * (SOCK_TIMESTAMPING_SOFTWARE)
2097 * - hardware time stamps available and wanted
2098 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
2099 * SOCK_TIMESTAMPING_RAW_HARDWARE)
2100 */
2101 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
2102 sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
2103 (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
2104 (hwtstamps->hwtstamp.tv64 &&
2105 sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
2106 (hwtstamps->syststamp.tv64 &&
2107 sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
Eric Dumazet92f37fd2007-03-25 22:14:49 -07002108 __sock_recv_timestamp(msg, sk, skb);
2109 else
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002110 sk->sk_stamp = kt;
Johannes Berg6e3e9392011-11-09 10:15:42 +01002111
2112 if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2113 __sock_recv_wifi_status(msg, sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002114}
2115
Eric Dumazet767dd032010-04-28 19:14:43 +00002116extern void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2117 struct sk_buff *skb);
2118
2119static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2120 struct sk_buff *skb)
2121{
2122#define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2123 (1UL << SOCK_RCVTSTAMP) | \
2124 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
2125 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002126 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
Eric Dumazet767dd032010-04-28 19:14:43 +00002127 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
2128
2129 if (sk->sk_flags & FLAGS_TS_OR_DROPS)
2130 __sock_recv_ts_and_drops(msg, sk, skb);
2131 else
2132 sk->sk_stamp = skb->tstamp;
2133}
Neil Horman3b885782009-10-12 13:26:31 -07002134
Linus Torvalds1da177e2005-04-16 15:20:36 -07002135/**
Patrick Ohly20d49472009-02-12 05:03:38 +00002136 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
Patrick Ohly20d49472009-02-12 05:03:38 +00002137 * @sk: socket sending this packet
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002138 * @tx_flags: filled with instructions for time stamping
Patrick Ohly20d49472009-02-12 05:03:38 +00002139 *
2140 * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if
2141 * parameters are invalid.
2142 */
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002143extern int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags);
Patrick Ohly20d49472009-02-12 05:03:38 +00002144
2145/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002146 * sk_eat_skb - Release a skb if it is no longer needed
Pavel Pisa4dc3b162005-05-01 08:59:25 -07002147 * @sk: socket to eat this skb from
2148 * @skb: socket buffer to eat
Randy Dunlapf4b8ea72006-06-22 16:00:11 -07002149 * @copied_early: flag indicating whether DMA operations copied this data early
Linus Torvalds1da177e2005-04-16 15:20:36 -07002150 *
2151 * This routine must be called with interrupts disabled or with the socket
2152 * locked so that the sk_buff queue operation is ok.
2153*/
Chris Leech624d1162006-05-23 18:01:28 -07002154#ifdef CONFIG_NET_DMA
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002155static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
Chris Leech624d1162006-05-23 18:01:28 -07002156{
2157 __skb_unlink(skb, &sk->sk_receive_queue);
2158 if (!copied_early)
2159 __kfree_skb(skb);
2160 else
2161 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
2162}
2163#else
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002164static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002165{
2166 __skb_unlink(skb, &sk->sk_receive_queue);
2167 __kfree_skb(skb);
2168}
Chris Leech624d1162006-05-23 18:01:28 -07002169#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002170
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002171static inline
2172struct net *sock_net(const struct sock *sk)
2173{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002174 return read_pnet(&sk->sk_net);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002175}
2176
2177static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07002178void sock_net_set(struct sock *sk, struct net *net)
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002179{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002180 write_pnet(&sk->sk_net, net);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002181}
2182
Denis V. Lunevedf02082008-02-29 11:18:32 -08002183/*
2184 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
Lucas De Marchi25985ed2011-03-30 22:57:33 -03002185 * They should not hold a reference to a namespace in order to allow
Denis V. Lunevedf02082008-02-29 11:18:32 -08002186 * to stop it.
2187 * Sockets after sk_change_net should be released using sk_release_kernel
2188 */
2189static inline void sk_change_net(struct sock *sk, struct net *net)
2190{
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002191 put_net(sock_net(sk));
Denis V. Lunev65a18ec2008-04-16 01:59:46 -07002192 sock_net_set(sk, hold_net(net));
Denis V. Lunevedf02082008-02-29 11:18:32 -08002193}
2194
KOVACS Krisztian23542612008-10-07 12:41:01 -07002195static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2196{
Vijay Subramanianefc27f82012-06-24 13:03:07 +00002197 if (skb->sk) {
KOVACS Krisztian23542612008-10-07 12:41:01 -07002198 struct sock *sk = skb->sk;
2199
2200 skb->destructor = NULL;
2201 skb->sk = NULL;
2202 return sk;
2203 }
2204 return NULL;
2205}
2206
Patrick Ohly20d49472009-02-12 05:03:38 +00002207extern void sock_enable_timestamp(struct sock *sk, int flag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002208extern int sock_get_timestamp(struct sock *, struct timeval __user *);
Eric Dumazetae40eb12007-03-18 17:33:16 -07002209extern int sock_get_timestampns(struct sock *, struct timespec __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002210
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002211/*
2212 * Enable debug/info messages
Linus Torvalds1da177e2005-04-16 15:20:36 -07002213 */
Stephen Hemmingera2a316f2007-03-08 20:41:08 -08002214extern int net_msg_warn;
2215#define NETDEBUG(fmt, args...) \
2216 do { if (net_msg_warn) printk(fmt,##args); } while (0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002217
Stephen Hemmingera2a316f2007-03-08 20:41:08 -08002218#define LIMIT_NETDEBUG(fmt, args...) \
2219 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002220
Linus Torvalds1da177e2005-04-16 15:20:36 -07002221extern __u32 sysctl_wmem_max;
2222extern __u32 sysctl_rmem_max;
2223
David S. Miller6baf1f42005-09-05 18:14:11 -07002224extern int sysctl_optmem_max;
2225
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002226extern __u32 sysctl_wmem_default;
2227extern __u32 sysctl_rmem_default;
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002228
Linus Torvalds1da177e2005-04-16 15:20:36 -07002229#endif /* _SOCK_H */