blob: ac8e1818380c2c2a669cd9725dd4f018d86092b5 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the AF_INET socket handler.
7 *
8 * Version: @(#)sock.h 1.0.4 05/13/93
9 *
Jesper Juhl02c30a82005-05-05 16:16:16 -070010 * Authors: Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -070011 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Florian La Roche <flla@stud.uni-sb.de>
14 *
15 * Fixes:
16 * Alan Cox : Volatiles in skbuff pointers. See
17 * skbuff comments. May be overdone,
18 * better to prove they can be removed
19 * than the reverse.
20 * Alan Cox : Added a zapped field for tcp to note
21 * a socket is reset and must stay shut up
22 * Alan Cox : New fields for options
23 * Pauline Middelink : identd support
24 * Alan Cox : Eliminate low level recv/recvfrom
25 * David S. Miller : New socket lookup architecture.
26 * Steve Whitehouse: Default routines for sock_ops
27 * Arnaldo C. Melo : removed net_pinfo, tp_pinfo and made
28 * protinfo be just a void pointer, as the
29 * protocol specific parts were moved to
30 * respective headers and ipv4/v6, etc now
31 * use private slabcaches for its socks
32 * Pedro Hortas : New flags field for socket options
33 *
34 *
35 * This program is free software; you can redistribute it and/or
36 * modify it under the terms of the GNU General Public License
37 * as published by the Free Software Foundation; either version
38 * 2 of the License, or (at your option) any later version.
39 */
40#ifndef _SOCK_H
41#define _SOCK_H
42
Alexey Dobriyana6b7a402011-06-06 10:43:46 +000043#include <linux/hardirq.h>
Ilpo Järvinen172589c2007-08-28 15:50:33 -070044#include <linux/kernel.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070045#include <linux/list.h>
Eric Dumazet88ab1932008-11-16 19:39:21 -080046#include <linux/list_nulls.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070047#include <linux/timer.h>
48#include <linux/cache.h>
Glauber Costa3f134612012-05-29 15:07:11 -070049#include <linux/bitops.h>
Ingo Molnara5b5bb92006-07-03 00:25:35 -070050#include <linux/lockdep.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070051#include <linux/netdevice.h>
52#include <linux/skbuff.h> /* struct sk_buff */
Al Virod7fe0f22006-12-03 23:15:30 -050053#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070054#include <linux/security.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090055#include <linux/slab.h>
Tom Herbertc6e1a0d2011-04-04 22:30:30 -070056#include <linux/uaccess.h>
Glauber Costa180d8cd2011-12-11 21:47:02 +000057#include <linux/memcontrol.h>
Glauber Costae1aab162011-12-11 21:47:03 +000058#include <linux/res_counter.h>
Ingo Molnarc5905af2012-02-24 08:31:31 +010059#include <linux/static_key.h>
Al Viro40401532012-02-13 03:58:52 +000060#include <linux/aio.h>
61#include <linux/sched.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070062
63#include <linux/filter.h>
Eric Dumazet88ab1932008-11-16 19:39:21 -080064#include <linux/rculist_nulls.h>
Jiri Olsaa57de0b2009-07-08 12:09:13 +000065#include <linux/poll.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070066
Eric Dumazetc31504d2010-11-15 19:58:26 +000067#include <linux/atomic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070068#include <net/dst.h>
69#include <net/checksum.h>
70
Eric Dumazet9f048bf2011-12-13 03:59:08 +000071struct cgroup;
72struct cgroup_subsys;
Glauber Costac607b2e2011-12-16 00:52:00 +000073#ifdef CONFIG_NET
Glauber Costa1d62e432012-04-09 19:36:33 -030074int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss);
75void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg);
Glauber Costac607b2e2011-12-16 00:52:00 +000076#else
77static inline
Glauber Costa1d62e432012-04-09 19:36:33 -030078int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
Glauber Costac607b2e2011-12-16 00:52:00 +000079{
80 return 0;
81}
82static inline
Glauber Costa1d62e432012-04-09 19:36:33 -030083void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg)
Glauber Costac607b2e2011-12-16 00:52:00 +000084{
85}
86#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -070087/*
88 * This structure really needs to be cleaned up.
89 * Most of it is for TCP, and not used by any of
90 * the other protocols.
91 */
92
93/* Define this to get the SOCK_DBG debugging facility. */
94#define SOCK_DEBUGGING
95#ifdef SOCK_DEBUGGING
96#define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
97 printk(KERN_DEBUG msg); } while (0)
98#else
Stephen Hemminger4cd90292008-03-21 15:54:53 -070099/* Validate arguments and do nothing */
Joe Perchesb9075fa2011-10-31 17:11:33 -0700100static inline __printf(2, 3)
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000101void SOCK_DEBUG(const struct sock *sk, const char *msg, ...)
Stephen Hemminger4cd90292008-03-21 15:54:53 -0700102{
103}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104#endif
105
106/* This is the per-socket lock. The spinlock provides a synchronization
107 * between user contexts and software interrupt processing, whereas the
108 * mini-semaphore synchronizes multiple users amongst themselves.
109 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700110typedef struct {
111 spinlock_t slock;
John Heffnerd2e91172007-09-12 10:44:19 +0200112 int owned;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113 wait_queue_head_t wq;
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700114 /*
115 * We express the mutex-alike socket_lock semantics
116 * to the lock validator by explicitly managing
117 * the slock as a lock variant (in addition to
118 * the slock itself):
119 */
120#ifdef CONFIG_DEBUG_LOCK_ALLOC
121 struct lockdep_map dep_map;
122#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123} socket_lock_t;
124
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125struct sock;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700126struct proto;
Denis V. Lunev0eeb8ff2007-12-04 01:15:45 -0800127struct net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128
Eric Dumazet077b3932012-12-02 07:33:10 +0000129typedef __u32 __bitwise __portpair;
130typedef __u64 __bitwise __addrpair;
131
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132/**
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700133 * struct sock_common - minimal network layer representation of sockets
Eric Dumazet68835ab2010-11-30 19:04:07 +0000134 * @skc_daddr: Foreign IPv4 addr
135 * @skc_rcv_saddr: Bound local IPv4 addr
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000136 * @skc_hash: hash value used with various protocol lookup tables
Eric Dumazetd4cada42009-11-08 10:17:30 +0000137 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
Eric Dumazetce43b032012-11-30 09:49:27 +0000138 * @skc_dport: placeholder for inet_dport/tw_dport
139 * @skc_num: placeholder for inet_num/tw_num
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700140 * @skc_family: network address family
141 * @skc_state: Connection state
142 * @skc_reuse: %SO_REUSEADDR setting
Tom Herbert055dc212013-01-22 09:49:50 +0000143 * @skc_reuseport: %SO_REUSEPORT setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700144 * @skc_bound_dev_if: bound device index if != 0
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700145 * @skc_bind_node: bind hash linkage for various protocol lookup tables
Eric Dumazet512615b2009-11-08 10:17:58 +0000146 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700147 * @skc_prot: protocol handlers inside a network family
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200148 * @skc_net: reference to the network namespace of this socket
Eric Dumazet68835ab2010-11-30 19:04:07 +0000149 * @skc_node: main hash linkage for various protocol lookup tables
150 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
151 * @skc_tx_queue_mapping: tx queue number for this connection
152 * @skc_refcnt: reference count
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700153 *
154 * This is the minimal network layer representation of sockets, the header
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700155 * for struct sock and struct inet_timewait_sock.
156 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157struct sock_common {
Eric Dumazetce43b032012-11-30 09:49:27 +0000158 /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned
159 * address on 64bit arches : cf INET_MATCH() and INET_TW_MATCH()
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000160 */
Eric Dumazetce43b032012-11-30 09:49:27 +0000161 union {
Eric Dumazet077b3932012-12-02 07:33:10 +0000162 __addrpair skc_addrpair;
Eric Dumazetce43b032012-11-30 09:49:27 +0000163 struct {
164 __be32 skc_daddr;
165 __be32 skc_rcv_saddr;
166 };
167 };
Eric Dumazetd4cada42009-11-08 10:17:30 +0000168 union {
169 unsigned int skc_hash;
170 __u16 skc_u16hashes[2];
171 };
Eric Dumazetce43b032012-11-30 09:49:27 +0000172 /* skc_dport && skc_num must be grouped as well */
173 union {
Eric Dumazet077b3932012-12-02 07:33:10 +0000174 __portpair skc_portpair;
Eric Dumazetce43b032012-11-30 09:49:27 +0000175 struct {
176 __be16 skc_dport;
177 __u16 skc_num;
178 };
179 };
180
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000181 unsigned short skc_family;
182 volatile unsigned char skc_state;
Tom Herbert055dc212013-01-22 09:49:50 +0000183 unsigned char skc_reuse:4;
184 unsigned char skc_reuseport:4;
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000185 int skc_bound_dev_if;
Eric Dumazet512615b2009-11-08 10:17:58 +0000186 union {
187 struct hlist_node skc_bind_node;
188 struct hlist_nulls_node skc_portaddr_node;
189 };
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700190 struct proto *skc_prot;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900191#ifdef CONFIG_NET_NS
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200192 struct net *skc_net;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900193#endif
Eric Dumazet68835ab2010-11-30 19:04:07 +0000194 /*
195 * fields between dontcopy_begin/dontcopy_end
196 * are not copied in sock_copy()
197 */
Randy Dunlap928c41e2011-01-08 17:39:21 +0000198 /* private: */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000199 int skc_dontcopy_begin[0];
Randy Dunlap928c41e2011-01-08 17:39:21 +0000200 /* public: */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000201 union {
202 struct hlist_node skc_node;
203 struct hlist_nulls_node skc_nulls_node;
204 };
205 int skc_tx_queue_mapping;
206 atomic_t skc_refcnt;
Randy Dunlap928c41e2011-01-08 17:39:21 +0000207 /* private: */
Eric Dumazet68835ab2010-11-30 19:04:07 +0000208 int skc_dontcopy_end[0];
Randy Dunlap928c41e2011-01-08 17:39:21 +0000209 /* public: */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210};
211
Glauber Costae1aab162011-12-11 21:47:03 +0000212struct cg_proto;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213/**
214 * struct sock - network layer representation of sockets
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700215 * @__sk_common: shared layout with inet_timewait_sock
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700216 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
217 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
218 * @sk_lock: synchronizer
219 * @sk_rcvbuf: size of receive buffer in bytes
Eric Dumazet43815482010-04-29 11:01:49 +0000220 * @sk_wq: sock wait queue and async head
Eric Dumazetdeaa5852012-06-24 20:22:49 +0000221 * @sk_rx_dst: receive input route used by early tcp demux
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700222 * @sk_dst_cache: destination cache
223 * @sk_dst_lock: destination cache lock
224 * @sk_policy: flow policy
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700225 * @sk_receive_queue: incoming packets
226 * @sk_wmem_alloc: transmit queue bytes committed
227 * @sk_write_queue: Packet sending queue
Chris Leech97fc2f02006-05-23 17:55:33 -0700228 * @sk_async_wait_queue: DMA copied packets
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700229 * @sk_omem_alloc: "o" is "option" or "other"
230 * @sk_wmem_queued: persistent queue size
231 * @sk_forward_alloc: space allocated forward
Eliezer Tamir06021292013-06-10 11:39:50 +0300232 * @sk_napi_id: id of the last napi context to receive data for sk
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700233 * @sk_allocation: allocation mode
234 * @sk_sndbuf: size of send buffer in bytes
Wang Chen33c732c2007-11-13 20:30:01 -0800235 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
Patrick Ohly20d49472009-02-12 05:03:38 +0000236 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
Adam Buchbinder48fc7f72012-09-19 21:48:00 -0400237 * @sk_no_check: %SO_NO_CHECK setting, whether or not checkup packets
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700238 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
Eric Dumazeta4654192010-05-16 00:36:33 -0700239 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
Herbert Xubcd76112006-06-30 13:36:35 -0700240 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -0700241 * @sk_gso_max_size: Maximum GSO segment size to build
Ben Hutchings14853482012-07-30 16:11:42 +0000242 * @sk_gso_max_segs: Maximum number of GSO segments
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700243 * @sk_lingertime: %SO_LINGER l_linger setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700244 * @sk_backlog: always used with the per-socket spinlock held
245 * @sk_callback_lock: used with the callbacks in the end of this struct
246 * @sk_error_queue: rarely used
Wang Chen33c732c2007-11-13 20:30:01 -0800247 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
248 * IPV6_ADDRFORM for instance)
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700249 * @sk_err: last error
Wang Chen33c732c2007-11-13 20:30:01 -0800250 * @sk_err_soft: errors that don't cause failure but are the cause of a
251 * persistent failure not just 'timed out'
Eric Dumazetcb61cb92008-06-17 21:04:56 -0700252 * @sk_drops: raw/udp drops counter
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700253 * @sk_ack_backlog: current listen backlog
254 * @sk_max_ack_backlog: listen backlog set in listen()
255 * @sk_priority: %SO_PRIORITY setting
Randy Dunlap1a3bc362012-01-21 09:03:10 +0000256 * @sk_cgrp_prioidx: socket group's priority map index
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700257 * @sk_type: socket type (%SOCK_STREAM, etc)
258 * @sk_protocol: which protocol this socket belongs in this network family
Randy Dunlap53c3fa22010-08-09 13:41:07 +0000259 * @sk_peer_pid: &struct pid for this socket's peer
260 * @sk_peer_cred: %SO_PEERCRED setting
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700261 * @sk_rcvlowat: %SO_RCVLOWAT setting
262 * @sk_rcvtimeo: %SO_RCVTIMEO setting
263 * @sk_sndtimeo: %SO_SNDTIMEO setting
David S. Millerc58dc012010-04-27 15:05:31 -0700264 * @sk_rxhash: flow hash received from netif layer
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700265 * @sk_filter: socket filtering instructions
266 * @sk_protinfo: private area, net family specific, when not using slab
267 * @sk_timer: sock cleanup timer
268 * @sk_stamp: time stamp of last packet received
269 * @sk_socket: Identd and reporting IO signals
270 * @sk_user_data: RPC layer private data
Eric Dumazet5640f762012-09-23 23:04:42 +0000271 * @sk_frag: cached page frag
Randy Dunlapd3d4f0a2012-04-17 14:03:53 +0000272 * @sk_peek_off: current peek_offset value
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700273 * @sk_send_head: front of stuff to transmit
Martin Waitz67be2dd2005-05-01 08:59:26 -0700274 * @sk_security: used by security modules
Randy Dunlap31729362008-02-18 20:52:13 -0800275 * @sk_mark: generic packet mark
Randy Dunlap53c3fa22010-08-09 13:41:07 +0000276 * @sk_classid: this socket's cgroup classid
Glauber Costae1aab162011-12-11 21:47:03 +0000277 * @sk_cgrp: this socket's cgroup-specific proto data
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700278 * @sk_write_pending: a write to stream socket waits to start
279 * @sk_state_change: callback to indicate change in the state of the sock
280 * @sk_data_ready: callback to indicate there is data to be processed
281 * @sk_write_space: callback to indicate there is bf sending space available
282 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
283 * @sk_backlog_rcv: callback to process the backlog
284 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285 */
286struct sock {
287 /*
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700288 * Now struct inet_timewait_sock also uses sock_common, so please just
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289 * don't add nothing before this first member (__sk_common) --acme
290 */
291 struct sock_common __sk_common;
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000292#define sk_node __sk_common.skc_node
293#define sk_nulls_node __sk_common.skc_nulls_node
294#define sk_refcnt __sk_common.skc_refcnt
Krishna Kumare022f0b2009-10-19 23:46:20 +0000295#define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000296
Eric Dumazet68835ab2010-11-30 19:04:07 +0000297#define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
298#define sk_dontcopy_end __sk_common.skc_dontcopy_end
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000299#define sk_hash __sk_common.skc_hash
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300#define sk_family __sk_common.skc_family
301#define sk_state __sk_common.skc_state
302#define sk_reuse __sk_common.skc_reuse
Tom Herbert055dc212013-01-22 09:49:50 +0000303#define sk_reuseport __sk_common.skc_reuseport
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304#define sk_bound_dev_if __sk_common.skc_bound_dev_if
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305#define sk_bind_node __sk_common.skc_bind_node
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700306#define sk_prot __sk_common.skc_prot
Eric W. Biederman07feaeb2007-09-12 11:58:02 +0200307#define sk_net __sk_common.skc_net
Eric Dumazetb178bb32010-11-16 05:56:04 +0000308 socket_lock_t sk_lock;
309 struct sk_buff_head sk_receive_queue;
310 /*
311 * The backlog queue is special, it is always used with
312 * the per-socket spinlock held and requires low latency
313 * access. Therefore we special case it's implementation.
314 * Note : rmem_alloc is in this structure to fill a hole
315 * on 64bit arches, not because its logically part of
316 * backlog.
317 */
318 struct {
319 atomic_t rmem_alloc;
320 int len;
321 struct sk_buff *head;
322 struct sk_buff *tail;
323 } sk_backlog;
324#define sk_rmem_alloc sk_backlog.rmem_alloc
325 int sk_forward_alloc;
326#ifdef CONFIG_RPS
327 __u32 sk_rxhash;
328#endif
Eliezer Tamir06021292013-06-10 11:39:50 +0300329#ifdef CONFIG_NET_LL_RX_POLL
330 unsigned int sk_napi_id;
331#endif
Eric Dumazetb178bb32010-11-16 05:56:04 +0000332 atomic_t sk_drops;
333 int sk_rcvbuf;
334
335 struct sk_filter __rcu *sk_filter;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000336 struct socket_wq __rcu *sk_wq;
Eric Dumazetb178bb32010-11-16 05:56:04 +0000337
338#ifdef CONFIG_NET_DMA
339 struct sk_buff_head sk_async_wait_queue;
340#endif
341
342#ifdef CONFIG_XFRM
343 struct xfrm_policy *sk_policy[2];
344#endif
345 unsigned long sk_flags;
Eric Dumazetdeaa5852012-06-24 20:22:49 +0000346 struct dst_entry *sk_rx_dst;
Cong Wang0e36cbb2013-01-22 21:09:51 +0000347 struct dst_entry __rcu *sk_dst_cache;
Eric Dumazetb178bb32010-11-16 05:56:04 +0000348 spinlock_t sk_dst_lock;
349 atomic_t sk_wmem_alloc;
350 atomic_t sk_omem_alloc;
351 int sk_sndbuf;
352 struct sk_buff_head sk_write_queue;
Vegard Nossuma98b65a2009-02-26 14:46:57 +0100353 kmemcheck_bitfield_begin(flags);
Eric Dumazet5fdb9972009-10-08 22:50:25 +0000354 unsigned int sk_shutdown : 2,
355 sk_no_check : 2,
356 sk_userlocks : 4,
357 sk_protocol : 8,
358 sk_type : 16;
Vegard Nossuma98b65a2009-02-26 14:46:57 +0100359 kmemcheck_bitfield_end(flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360 int sk_wmem_queued;
Al Viro7d877f32005-10-21 03:20:43 -0400361 gfp_t sk_allocation;
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000362 netdev_features_t sk_route_caps;
363 netdev_features_t sk_route_nocaps;
Herbert Xubcd76112006-06-30 13:36:35 -0700364 int sk_gso_type;
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -0700365 unsigned int sk_gso_max_size;
Ben Hutchings14853482012-07-30 16:11:42 +0000366 u16 sk_gso_max_segs;
David S. Miller9932cf92006-03-24 15:12:37 -0800367 int sk_rcvlowat;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368 unsigned long sk_lingertime;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369 struct sk_buff_head sk_error_queue;
Arnaldo Carvalho de Melo476e19c2005-05-05 13:35:15 -0700370 struct proto *sk_prot_creator;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371 rwlock_t sk_callback_lock;
372 int sk_err,
373 sk_err_soft;
374 unsigned short sk_ack_backlog;
375 unsigned short sk_max_ack_backlog;
376 __u32 sk_priority;
Li Zefan3d0dcfb2012-12-25 20:48:24 +0000377#if IS_ENABLED(CONFIG_NETPRIO_CGROUP)
Neil Horman5bc14212011-11-22 05:10:51 +0000378 __u32 sk_cgrp_prioidx;
379#endif
Eric W. Biederman109f6e32010-06-13 03:30:14 +0000380 struct pid *sk_peer_pid;
381 const struct cred *sk_peer_cred;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382 long sk_rcvtimeo;
383 long sk_sndtimeo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384 void *sk_protinfo;
385 struct timer_list sk_timer;
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700386 ktime_t sk_stamp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700387 struct socket *sk_socket;
388 void *sk_user_data;
Eric Dumazet5640f762012-09-23 23:04:42 +0000389 struct page_frag sk_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390 struct sk_buff *sk_send_head;
Pavel Emelyanovef64a542012-02-21 07:31:34 +0000391 __s32 sk_peek_off;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392 int sk_write_pending;
Alexey Dobriyand5f64232008-11-04 14:45:58 -0800393#ifdef CONFIG_SECURITY
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394 void *sk_security;
Alexey Dobriyand5f64232008-11-04 14:45:58 -0800395#endif
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800396 __u32 sk_mark;
Herbert Xuf8451722010-05-24 00:12:34 -0700397 u32 sk_classid;
Glauber Costae1aab162011-12-11 21:47:03 +0000398 struct cg_proto *sk_cgrp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399 void (*sk_state_change)(struct sock *sk);
400 void (*sk_data_ready)(struct sock *sk, int bytes);
401 void (*sk_write_space)(struct sock *sk);
402 void (*sk_error_report)(struct sock *sk);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000403 int (*sk_backlog_rcv)(struct sock *sk,
404 struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 void (*sk_destruct)(struct sock *sk);
406};
407
Pavel Emelyanov4a17fd52012-04-19 03:39:36 +0000408/*
409 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
410 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
411 * on a socket means that the socket will reuse everybody else's port
412 * without looking at the other's sk_reuse value.
413 */
414
415#define SK_NO_REUSE 0
416#define SK_CAN_REUSE 1
417#define SK_FORCE_REUSE 2
418
Pavel Emelyanovef64a542012-02-21 07:31:34 +0000419static inline int sk_peek_offset(struct sock *sk, int flags)
420{
421 if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0))
422 return sk->sk_peek_off;
423 else
424 return 0;
425}
426
427static inline void sk_peek_offset_bwd(struct sock *sk, int val)
428{
429 if (sk->sk_peek_off >= 0) {
430 if (sk->sk_peek_off >= val)
431 sk->sk_peek_off -= val;
432 else
433 sk->sk_peek_off = 0;
434 }
435}
436
437static inline void sk_peek_offset_fwd(struct sock *sk, int val)
438{
439 if (sk->sk_peek_off >= 0)
440 sk->sk_peek_off += val;
441}
442
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443/*
444 * Hashed lists helper routines
445 */
Li Zefanc4146642010-02-08 23:18:45 +0000446static inline struct sock *sk_entry(const struct hlist_node *node)
447{
448 return hlist_entry(node, struct sock, sk_node);
449}
450
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700451static inline struct sock *__sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452{
453 return hlist_entry(head->first, struct sock, sk_node);
454}
455
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700456static inline struct sock *sk_head(const struct hlist_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457{
458 return hlist_empty(head) ? NULL : __sk_head(head);
459}
460
Eric Dumazet88ab1932008-11-16 19:39:21 -0800461static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
462{
463 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
464}
465
466static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
467{
468 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
469}
470
Arnaldo Carvalho de Meloe48c4142005-08-09 20:09:46 -0700471static inline struct sock *sk_next(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472{
473 return sk->sk_node.next ?
474 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
475}
476
Eric Dumazet88ab1932008-11-16 19:39:21 -0800477static inline struct sock *sk_nulls_next(const struct sock *sk)
478{
479 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
480 hlist_nulls_entry(sk->sk_nulls_node.next,
481 struct sock, sk_nulls_node) :
482 NULL;
483}
484
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000485static inline bool sk_unhashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486{
487 return hlist_unhashed(&sk->sk_node);
488}
489
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000490static inline bool sk_hashed(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491{
Akinobu Mitada753be2006-04-28 15:21:23 -0700492 return !sk_unhashed(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493}
494
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000495static inline void sk_node_init(struct hlist_node *node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496{
497 node->pprev = NULL;
498}
499
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000500static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800501{
502 node->pprev = NULL;
503}
504
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000505static inline void __sk_del_node(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506{
507 __hlist_del(&sk->sk_node);
508}
509
stephen hemminger808f5112010-02-22 07:57:18 +0000510/* NB: equivalent to hlist_del_init_rcu */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000511static inline bool __sk_del_node_init(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512{
513 if (sk_hashed(sk)) {
514 __sk_del_node(sk);
515 sk_node_init(&sk->sk_node);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000516 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517 }
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000518 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519}
520
521/* Grab socket reference count. This operation is valid only
522 when sk is ALREADY grabbed f.e. it is found in hash table
523 or a list and the lookup is made under lock preventing hash table
524 modifications.
525 */
526
527static inline void sock_hold(struct sock *sk)
528{
529 atomic_inc(&sk->sk_refcnt);
530}
531
532/* Ungrab socket in the context, which assumes that socket refcnt
533 cannot hit zero, f.e. it is true in context of any socketcall.
534 */
535static inline void __sock_put(struct sock *sk)
536{
537 atomic_dec(&sk->sk_refcnt);
538}
539
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000540static inline bool sk_del_node_init(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541{
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000542 bool rc = __sk_del_node_init(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543
544 if (rc) {
545 /* paranoid for a while -acme */
546 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
547 __sock_put(sk);
548 }
549 return rc;
550}
stephen hemminger808f5112010-02-22 07:57:18 +0000551#define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000553static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700554{
555 if (sk_hashed(sk)) {
Eric Dumazet88ab1932008-11-16 19:39:21 -0800556 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000557 return true;
Eric Dumazet271b72c2008-10-29 02:11:14 -0700558 }
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000559 return false;
Eric Dumazet271b72c2008-10-29 02:11:14 -0700560}
561
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000562static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700563{
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000564 bool rc = __sk_nulls_del_node_init_rcu(sk);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700565
566 if (rc) {
567 /* paranoid for a while -acme */
568 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
569 __sock_put(sk);
570 }
571 return rc;
572}
573
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000574static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575{
576 hlist_add_head(&sk->sk_node, list);
577}
578
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000579static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580{
581 sock_hold(sk);
582 __sk_add_node(sk, list);
583}
584
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000585static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
stephen hemminger808f5112010-02-22 07:57:18 +0000586{
587 sock_hold(sk);
588 hlist_add_head_rcu(&sk->sk_node, list);
589}
590
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000591static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700592{
Eric Dumazet88ab1932008-11-16 19:39:21 -0800593 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700594}
595
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000596static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
Eric Dumazet271b72c2008-10-29 02:11:14 -0700597{
598 sock_hold(sk);
Eric Dumazet88ab1932008-11-16 19:39:21 -0800599 __sk_nulls_add_node_rcu(sk, list);
Eric Dumazet271b72c2008-10-29 02:11:14 -0700600}
601
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000602static inline void __sk_del_bind_node(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603{
604 __hlist_del(&sk->sk_bind_node);
605}
606
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000607static inline void sk_add_bind_node(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608 struct hlist_head *list)
609{
610 hlist_add_head(&sk->sk_bind_node, list);
611}
612
Sasha Levinb67bfe02013-02-27 17:06:00 -0800613#define sk_for_each(__sk, list) \
614 hlist_for_each_entry(__sk, list, sk_node)
615#define sk_for_each_rcu(__sk, list) \
616 hlist_for_each_entry_rcu(__sk, list, sk_node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800617#define sk_nulls_for_each(__sk, node, list) \
618 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
619#define sk_nulls_for_each_rcu(__sk, node, list) \
620 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
Sasha Levinb67bfe02013-02-27 17:06:00 -0800621#define sk_for_each_from(__sk) \
622 hlist_for_each_entry_from(__sk, sk_node)
Eric Dumazet88ab1932008-11-16 19:39:21 -0800623#define sk_nulls_for_each_from(__sk, node) \
624 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
625 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
Sasha Levinb67bfe02013-02-27 17:06:00 -0800626#define sk_for_each_safe(__sk, tmp, list) \
627 hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
628#define sk_for_each_bound(__sk, list) \
629 hlist_for_each_entry(__sk, list, sk_bind_node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630
Eric W. Biedermanc336d142012-05-24 17:56:43 -0600631static inline struct user_namespace *sk_user_ns(struct sock *sk)
632{
633 /* Careful only use this in a context where these parameters
634 * can not change and must all be valid, such as recvmsg from
635 * userspace.
636 */
637 return sk->sk_socket->file->f_cred->user_ns;
638}
639
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640/* Sock flags */
641enum sock_flags {
642 SOCK_DEAD,
643 SOCK_DONE,
644 SOCK_URGINLINE,
645 SOCK_KEEPOPEN,
646 SOCK_LINGER,
647 SOCK_DESTROY,
648 SOCK_BROADCAST,
649 SOCK_TIMESTAMP,
650 SOCK_ZAPPED,
651 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
652 SOCK_DBG, /* %SO_DEBUG setting */
653 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700654 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
656 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
Mel Gorman7cb02402012-07-31 16:44:16 -0700657 SOCK_MEMALLOC, /* VM depends on this socket for swapping */
Patrick Ohly20d49472009-02-12 05:03:38 +0000658 SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */
659 SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
660 SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */
661 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
662 SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */
663 SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
664 SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
Eric Dumazetbcdce712009-10-06 17:28:29 -0700665 SOCK_FASYNC, /* fasync() active */
Neil Horman3b885782009-10-12 13:26:31 -0700666 SOCK_RXQ_OVFL,
Shirley Ma1cdebb42011-07-06 12:17:30 +0000667 SOCK_ZEROCOPY, /* buffers from userspace */
Johannes Berg6e3e9392011-11-09 10:15:42 +0100668 SOCK_WIFI_STATUS, /* push wifi status to userspace */
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000669 SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
670 * Will use last 4 bytes of packet sent from
671 * user-space instead.
672 */
Vincent Bernatd59577b2013-01-16 22:55:49 +0100673 SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */
Keller, Jacob E7d4c04f2013-03-28 11:19:25 +0000674 SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675};
676
Ralf Baechle53b924b2005-08-23 10:11:30 -0700677static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
678{
679 nsk->sk_flags = osk->sk_flags;
680}
681
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
683{
684 __set_bit(flag, &sk->sk_flags);
685}
686
687static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
688{
689 __clear_bit(flag, &sk->sk_flags);
690}
691
Eric Dumazet1b23a5d2012-05-16 05:57:07 +0000692static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693{
694 return test_bit(flag, &sk->sk_flags);
695}
696
Mel Gormanc93bdd02012-07-31 16:44:19 -0700697#ifdef CONFIG_NET
698extern struct static_key memalloc_socks;
699static inline int sk_memalloc_socks(void)
700{
701 return static_key_false(&memalloc_socks);
702}
703#else
704
705static inline int sk_memalloc_socks(void)
706{
707 return 0;
708}
709
710#endif
711
Mel Gorman99a1dec2012-07-31 16:44:14 -0700712static inline gfp_t sk_gfp_atomic(struct sock *sk, gfp_t gfp_mask)
713{
Mel Gorman7cb02402012-07-31 16:44:16 -0700714 return GFP_ATOMIC | (sk->sk_allocation & __GFP_MEMALLOC);
Mel Gorman99a1dec2012-07-31 16:44:14 -0700715}
716
Linus Torvalds1da177e2005-04-16 15:20:36 -0700717static inline void sk_acceptq_removed(struct sock *sk)
718{
719 sk->sk_ack_backlog--;
720}
721
722static inline void sk_acceptq_added(struct sock *sk)
723{
724 sk->sk_ack_backlog++;
725}
726
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000727static inline bool sk_acceptq_is_full(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728{
David S. Miller64a14652007-03-06 11:21:05 -0800729 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730}
731
732/*
733 * Compute minimal free write space needed to queue new packets.
734 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000735static inline int sk_stream_min_wspace(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736{
Eric Dumazet8df09ea2007-12-21 03:07:41 -0800737 return sk->sk_wmem_queued >> 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738}
739
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000740static inline int sk_stream_wspace(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741{
742 return sk->sk_sndbuf - sk->sk_wmem_queued;
743}
744
745extern void sk_stream_write_space(struct sock *sk);
746
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000747static inline bool sk_stream_memory_free(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748{
749 return sk->sk_wmem_queued < sk->sk_sndbuf;
750}
751
Zhu Yi8eae9392010-03-04 18:01:40 +0000752/* OOB backlog add */
Zhu Yia3a858f2010-03-04 18:01:47 +0000753static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800754{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000755 /* dont let skb dst not refcounted, we are going to leave rcu lock */
756 skb_dst_force(skb);
757
758 if (!sk->sk_backlog.tail)
759 sk->sk_backlog.head = skb;
760 else
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800761 sk->sk_backlog.tail->next = skb;
Eric Dumazet7fee2262010-05-11 23:19:48 +0000762
763 sk->sk_backlog.tail = skb;
Stephen Hemminger9ee6b532005-11-08 09:39:42 -0800764 skb->next = NULL;
765}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766
Eric Dumazetc3774112010-04-27 15:13:20 -0700767/*
768 * Take into account size of receive queue and backlog queue
Eric Dumazet0fd7bac2011-12-21 07:11:44 +0000769 * Do not take into account this skb truesize,
770 * to allow even a single big packet to come.
Eric Dumazetc3774112010-04-27 15:13:20 -0700771 */
Eric Dumazetf545a382012-04-22 23:34:26 +0000772static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb,
773 unsigned int limit)
Eric Dumazetc3774112010-04-27 15:13:20 -0700774{
775 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
776
Eric Dumazetf545a382012-04-22 23:34:26 +0000777 return qsize > limit;
Eric Dumazetc3774112010-04-27 15:13:20 -0700778}
779
Zhu Yi8eae9392010-03-04 18:01:40 +0000780/* The per-socket spinlock must be held here. */
Eric Dumazetf545a382012-04-22 23:34:26 +0000781static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
782 unsigned int limit)
Zhu Yi8eae9392010-03-04 18:01:40 +0000783{
Eric Dumazetf545a382012-04-22 23:34:26 +0000784 if (sk_rcvqueues_full(sk, skb, limit))
Zhu Yi8eae9392010-03-04 18:01:40 +0000785 return -ENOBUFS;
786
Zhu Yia3a858f2010-03-04 18:01:47 +0000787 __sk_add_backlog(sk, skb);
Zhu Yi8eae9392010-03-04 18:01:40 +0000788 sk->sk_backlog.len += skb->truesize;
789 return 0;
790}
791
Mel Gormanb4b9e352012-07-31 16:44:26 -0700792extern int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
793
Peter Zijlstrac57943a2008-10-07 14:18:42 -0700794static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
795{
Mel Gormanb4b9e352012-07-31 16:44:26 -0700796 if (sk_memalloc_socks() && skb_pfmemalloc(skb))
797 return __sk_backlog_rcv(sk, skb);
798
Peter Zijlstrac57943a2008-10-07 14:18:42 -0700799 return sk->sk_backlog_rcv(sk, skb);
800}
801
David S. Millerc58dc012010-04-27 15:05:31 -0700802static inline void sock_rps_record_flow(const struct sock *sk)
803{
804#ifdef CONFIG_RPS
805 struct rps_sock_flow_table *sock_flow_table;
806
807 rcu_read_lock();
808 sock_flow_table = rcu_dereference(rps_sock_flow_table);
809 rps_record_sock_flow(sock_flow_table, sk->sk_rxhash);
810 rcu_read_unlock();
811#endif
812}
813
814static inline void sock_rps_reset_flow(const struct sock *sk)
815{
816#ifdef CONFIG_RPS
817 struct rps_sock_flow_table *sock_flow_table;
818
819 rcu_read_lock();
820 sock_flow_table = rcu_dereference(rps_sock_flow_table);
821 rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash);
822 rcu_read_unlock();
823#endif
824}
825
Tom Herbertbdeab992011-08-14 19:45:55 +0000826static inline void sock_rps_save_rxhash(struct sock *sk,
827 const struct sk_buff *skb)
David S. Millerc58dc012010-04-27 15:05:31 -0700828{
829#ifdef CONFIG_RPS
Tom Herbertbdeab992011-08-14 19:45:55 +0000830 if (unlikely(sk->sk_rxhash != skb->rxhash)) {
David S. Millerc58dc012010-04-27 15:05:31 -0700831 sock_rps_reset_flow(sk);
Tom Herbertbdeab992011-08-14 19:45:55 +0000832 sk->sk_rxhash = skb->rxhash;
David S. Millerc58dc012010-04-27 15:05:31 -0700833 }
834#endif
835}
836
Tom Herbertbdeab992011-08-14 19:45:55 +0000837static inline void sock_rps_reset_rxhash(struct sock *sk)
838{
839#ifdef CONFIG_RPS
840 sock_rps_reset_flow(sk);
841 sk->sk_rxhash = 0;
842#endif
843}
844
Stephen Hemmingercfcabdc2007-10-09 01:59:42 -0700845#define sk_wait_event(__sk, __timeo, __condition) \
846 ({ int __rc; \
847 release_sock(__sk); \
848 __rc = __condition; \
849 if (!__rc) { \
850 *(__timeo) = schedule_timeout(*(__timeo)); \
851 } \
852 lock_sock(__sk); \
853 __rc = __condition; \
854 __rc; \
855 })
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856
857extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
858extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
859extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
860extern int sk_stream_error(struct sock *sk, int flags, int err);
861extern void sk_stream_kill_queues(struct sock *sk);
Mel Gorman7cb02402012-07-31 16:44:16 -0700862extern void sk_set_memalloc(struct sock *sk);
863extern void sk_clear_memalloc(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864
865extern int sk_wait_data(struct sock *sk, long *timeo);
866
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700867struct request_sock_ops;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800868struct timewait_sock_ops;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -0800869struct inet_hashinfo;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -0700870struct raw_hashinfo;
Paul Gortmakerde477252011-05-26 13:46:22 -0400871struct module;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700872
Eric Dumazetf77d6022013-05-09 10:28:16 +0000873/*
874 * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
875 * un-modified. Special care is taken when initializing object to zero.
876 */
877static inline void sk_prot_clear_nulls(struct sock *sk, int size)
878{
879 if (offsetof(struct sock, sk_node.next) != 0)
880 memset(sk, 0, offsetof(struct sock, sk_node.next));
881 memset(&sk->sk_node.pprev, 0,
882 size - offsetof(struct sock, sk_node.pprev));
883}
884
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885/* Networking protocol blocks we attach to sockets.
886 * socket layer -> transport layer interface
887 * transport -> network interface is defined by struct inet_proto
888 */
889struct proto {
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000890 void (*close)(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891 long timeout);
892 int (*connect)(struct sock *sk,
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000893 struct sockaddr *uaddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894 int addr_len);
895 int (*disconnect)(struct sock *sk, int flags);
896
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000897 struct sock * (*accept)(struct sock *sk, int flags, int *err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898
899 int (*ioctl)(struct sock *sk, int cmd,
900 unsigned long arg);
901 int (*init)(struct sock *sk);
Brian Haley7d06b2e2008-06-14 17:04:49 -0700902 void (*destroy)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903 void (*shutdown)(struct sock *sk, int how);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000904 int (*setsockopt)(struct sock *sk, int level,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700906 unsigned int optlen);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000907 int (*getsockopt)(struct sock *sk, int level,
908 int optname, char __user *optval,
909 int __user *option);
Alexey Dobriyanaf01d532008-08-28 02:53:51 -0700910#ifdef CONFIG_COMPAT
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800911 int (*compat_setsockopt)(struct sock *sk,
912 int level,
913 int optname, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -0700914 unsigned int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -0800915 int (*compat_getsockopt)(struct sock *sk,
916 int level,
917 int optname, char __user *optval,
918 int __user *option);
Eric W. Biederman709b46e2011-01-29 16:15:56 +0000919 int (*compat_ioctl)(struct sock *sk,
920 unsigned int cmd, unsigned long arg);
Alexey Dobriyanaf01d532008-08-28 02:53:51 -0700921#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
923 struct msghdr *msg, size_t len);
924 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
925 struct msghdr *msg,
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000926 size_t len, int noblock, int flags,
927 int *addr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700928 int (*sendpage)(struct sock *sk, struct page *page,
929 int offset, size_t size, int flags);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000930 int (*bind)(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931 struct sockaddr *uaddr, int addr_len);
932
Eric Dumazetdc6b9b72012-05-16 22:48:15 +0000933 int (*backlog_rcv) (struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700934 struct sk_buff *skb);
935
Eric Dumazet46d3cea2012-07-11 05:50:31 +0000936 void (*release_cb)(struct sock *sk);
Eric Dumazet563d34d2012-07-23 09:48:52 +0200937 void (*mtu_reduced)(struct sock *sk);
Eric Dumazet46d3cea2012-07-11 05:50:31 +0000938
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939 /* Keeping track of sk's, looking them up, and port selection methods. */
940 void (*hash)(struct sock *sk);
941 void (*unhash)(struct sock *sk);
Eric Dumazet719f8352010-09-08 05:08:44 +0000942 void (*rehash)(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700943 int (*get_port)(struct sock *sk, unsigned short snum);
Octavian Purdilafcbdf092010-12-16 14:26:56 -0800944 void (*clear_sk)(struct sock *sk, int size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700945
Eric Dumazet286ab3d2007-11-05 23:38:39 -0800946 /* Keeping track of sockets in use */
Eric Dumazet65f76512008-01-03 20:46:48 -0800947#ifdef CONFIG_PROC_FS
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -0700948 unsigned int inuse_idx;
Eric Dumazet65f76512008-01-03 20:46:48 -0800949#endif
Arnaldo Carvalho de Meloebb53d72007-11-21 22:08:50 +0800950
Linus Torvalds1da177e2005-04-16 15:20:36 -0700951 /* Memory pressure */
Pavel Emelyanov5c52ba12008-07-16 20:28:10 -0700952 void (*enter_memory_pressure)(struct sock *sk);
Eric Dumazet8d987e52010-11-09 23:24:26 +0000953 atomic_long_t *memory_allocated; /* Current allocated memory. */
Eric Dumazet17483762008-11-25 21:16:35 -0800954 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955 /*
956 * Pressure flag: try to collapse.
957 * Technical note: it is used by multiple contexts non atomically.
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800958 * All the __sk_mem_schedule() is of this nature: accounting
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959 * is strict, actions are advisory and have some latency.
960 */
961 int *memory_pressure;
Eric Dumazet8d987e52010-11-09 23:24:26 +0000962 long *sysctl_mem;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963 int *sysctl_wmem;
964 int *sysctl_rmem;
965 int max_header;
Changli Gao7ba42912010-07-10 20:41:55 +0000966 bool no_autobind;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967
Eric Dumazet271b72c2008-10-29 02:11:14 -0700968 struct kmem_cache *slab;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969 unsigned int obj_size;
Eric Dumazet271b72c2008-10-29 02:11:14 -0700970 int slab_flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971
Eric Dumazetdd24c002008-11-25 21:17:14 -0800972 struct percpu_counter *orphan_count;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -0700973
Arnaldo Carvalho de Melo60236fd2005-06-18 22:47:21 -0700974 struct request_sock_ops *rsk_prot;
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -0800975 struct timewait_sock_ops *twsk_prot;
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700976
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -0700977 union {
978 struct inet_hashinfo *hashinfo;
Eric Dumazet645ca702008-10-29 01:41:45 -0700979 struct udp_table *udp_table;
Pavel Emelyanovfc8717b2008-03-22 16:56:51 -0700980 struct raw_hashinfo *raw_hash;
Pavel Emelyanov39d8cda2008-03-22 16:50:58 -0700981 } h;
Arnaldo Carvalho de Meloab1e0a12008-02-03 04:06:04 -0800982
Linus Torvalds1da177e2005-04-16 15:20:36 -0700983 struct module *owner;
984
985 char name[32];
986
987 struct list_head node;
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -0700988#ifdef SOCK_REFCNT_DEBUG
989 atomic_t socks;
990#endif
Andrew Mortonc255a452012-07-31 16:43:02 -0700991#ifdef CONFIG_MEMCG_KMEM
Glauber Costae1aab162011-12-11 21:47:03 +0000992 /*
993 * cgroup specific init/deinit functions. Called once for all
994 * protocols that implement it, from cgroups populate function.
995 * This function has to setup any files the protocol want to
996 * appear in the kmem cgroup filesystem.
997 */
Glauber Costa1d62e432012-04-09 19:36:33 -0300998 int (*init_cgroup)(struct mem_cgroup *memcg,
Glauber Costae1aab162011-12-11 21:47:03 +0000999 struct cgroup_subsys *ss);
Glauber Costa1d62e432012-04-09 19:36:33 -03001000 void (*destroy_cgroup)(struct mem_cgroup *memcg);
Glauber Costae1aab162011-12-11 21:47:03 +00001001 struct cg_proto *(*proto_cgroup)(struct mem_cgroup *memcg);
1002#endif
1003};
1004
Glauber Costa3f134612012-05-29 15:07:11 -07001005/*
1006 * Bits in struct cg_proto.flags
1007 */
1008enum cg_proto_flags {
1009 /* Currently active and new sockets should be assigned to cgroups */
1010 MEMCG_SOCK_ACTIVE,
1011 /* It was ever activated; we must disarm static keys on destruction */
1012 MEMCG_SOCK_ACTIVATED,
1013};
1014
Glauber Costae1aab162011-12-11 21:47:03 +00001015struct cg_proto {
1016 void (*enter_memory_pressure)(struct sock *sk);
1017 struct res_counter *memory_allocated; /* Current allocated memory. */
1018 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
1019 int *memory_pressure;
1020 long *sysctl_mem;
Glauber Costa3f134612012-05-29 15:07:11 -07001021 unsigned long flags;
Glauber Costae1aab162011-12-11 21:47:03 +00001022 /*
1023 * memcg field is used to find which memcg we belong directly
1024 * Each memcg struct can hold more than one cg_proto, so container_of
1025 * won't really cut.
1026 *
1027 * The elegant solution would be having an inverse function to
1028 * proto_cgroup in struct proto, but that means polluting the structure
1029 * for everybody, instead of just for memcg users.
1030 */
1031 struct mem_cgroup *memcg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032};
1033
1034extern int proto_register(struct proto *prot, int alloc_slab);
1035extern void proto_unregister(struct proto *prot);
1036
Glauber Costa3f134612012-05-29 15:07:11 -07001037static inline bool memcg_proto_active(struct cg_proto *cg_proto)
1038{
1039 return test_bit(MEMCG_SOCK_ACTIVE, &cg_proto->flags);
1040}
1041
1042static inline bool memcg_proto_activated(struct cg_proto *cg_proto)
1043{
1044 return test_bit(MEMCG_SOCK_ACTIVATED, &cg_proto->flags);
1045}
1046
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -07001047#ifdef SOCK_REFCNT_DEBUG
1048static inline void sk_refcnt_debug_inc(struct sock *sk)
1049{
1050 atomic_inc(&sk->sk_prot->socks);
1051}
1052
1053static inline void sk_refcnt_debug_dec(struct sock *sk)
1054{
1055 atomic_dec(&sk->sk_prot->socks);
1056 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
1057 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
1058}
1059
Ying Xuedec34fb2013-02-15 22:28:25 +00001060static inline void sk_refcnt_debug_release(const struct sock *sk)
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -07001061{
1062 if (atomic_read(&sk->sk_refcnt) != 1)
1063 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
1064 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
1065}
1066#else /* SOCK_REFCNT_DEBUG */
1067#define sk_refcnt_debug_inc(sk) do { } while (0)
1068#define sk_refcnt_debug_dec(sk) do { } while (0)
1069#define sk_refcnt_debug_release(sk) do { } while (0)
1070#endif /* SOCK_REFCNT_DEBUG */
1071
Andrew Mortonc255a452012-07-31 16:43:02 -07001072#if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_NET)
Ingo Molnarc5905af2012-02-24 08:31:31 +01001073extern struct static_key memcg_socket_limit_enabled;
Glauber Costae1aab162011-12-11 21:47:03 +00001074static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1075 struct cg_proto *cg_proto)
1076{
1077 return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg));
1078}
Ingo Molnarc5905af2012-02-24 08:31:31 +01001079#define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
Glauber Costae1aab162011-12-11 21:47:03 +00001080#else
1081#define mem_cgroup_sockets_enabled 0
1082static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1083 struct cg_proto *cg_proto)
1084{
1085 return NULL;
1086}
1087#endif
1088
1089
Glauber Costa180d8cd2011-12-11 21:47:02 +00001090static inline bool sk_has_memory_pressure(const struct sock *sk)
1091{
1092 return sk->sk_prot->memory_pressure != NULL;
1093}
1094
1095static inline bool sk_under_memory_pressure(const struct sock *sk)
1096{
1097 if (!sk->sk_prot->memory_pressure)
1098 return false;
Glauber Costae1aab162011-12-11 21:47:03 +00001099
1100 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1101 return !!*sk->sk_cgrp->memory_pressure;
1102
Glauber Costa180d8cd2011-12-11 21:47:02 +00001103 return !!*sk->sk_prot->memory_pressure;
1104}
1105
1106static inline void sk_leave_memory_pressure(struct sock *sk)
1107{
1108 int *memory_pressure = sk->sk_prot->memory_pressure;
1109
Glauber Costae1aab162011-12-11 21:47:03 +00001110 if (!memory_pressure)
1111 return;
1112
1113 if (*memory_pressure)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001114 *memory_pressure = 0;
Glauber Costae1aab162011-12-11 21:47:03 +00001115
1116 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1117 struct cg_proto *cg_proto = sk->sk_cgrp;
1118 struct proto *prot = sk->sk_prot;
1119
1120 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1121 if (*cg_proto->memory_pressure)
1122 *cg_proto->memory_pressure = 0;
1123 }
1124
Glauber Costa180d8cd2011-12-11 21:47:02 +00001125}
1126
1127static inline void sk_enter_memory_pressure(struct sock *sk)
1128{
Glauber Costae1aab162011-12-11 21:47:03 +00001129 if (!sk->sk_prot->enter_memory_pressure)
1130 return;
1131
1132 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1133 struct cg_proto *cg_proto = sk->sk_cgrp;
1134 struct proto *prot = sk->sk_prot;
1135
1136 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1137 cg_proto->enter_memory_pressure(sk);
1138 }
1139
1140 sk->sk_prot->enter_memory_pressure(sk);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001141}
1142
1143static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1144{
1145 long *prot = sk->sk_prot->sysctl_mem;
Glauber Costae1aab162011-12-11 21:47:03 +00001146 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1147 prot = sk->sk_cgrp->sysctl_mem;
Glauber Costa180d8cd2011-12-11 21:47:02 +00001148 return prot[index];
1149}
1150
Glauber Costae1aab162011-12-11 21:47:03 +00001151static inline void memcg_memory_allocated_add(struct cg_proto *prot,
1152 unsigned long amt,
1153 int *parent_status)
1154{
1155 struct res_counter *fail;
1156 int ret;
1157
Glauber Costa0e90b312012-01-20 04:57:16 +00001158 ret = res_counter_charge_nofail(prot->memory_allocated,
1159 amt << PAGE_SHIFT, &fail);
Glauber Costae1aab162011-12-11 21:47:03 +00001160 if (ret < 0)
1161 *parent_status = OVER_LIMIT;
1162}
1163
1164static inline void memcg_memory_allocated_sub(struct cg_proto *prot,
1165 unsigned long amt)
1166{
1167 res_counter_uncharge(prot->memory_allocated, amt << PAGE_SHIFT);
1168}
1169
1170static inline u64 memcg_memory_allocated_read(struct cg_proto *prot)
1171{
1172 u64 ret;
1173 ret = res_counter_read_u64(prot->memory_allocated, RES_USAGE);
1174 return ret >> PAGE_SHIFT;
1175}
1176
Glauber Costa180d8cd2011-12-11 21:47:02 +00001177static inline long
1178sk_memory_allocated(const struct sock *sk)
1179{
1180 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001181 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1182 return memcg_memory_allocated_read(sk->sk_cgrp);
1183
Glauber Costa180d8cd2011-12-11 21:47:02 +00001184 return atomic_long_read(prot->memory_allocated);
1185}
1186
1187static inline long
Glauber Costae1aab162011-12-11 21:47:03 +00001188sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001189{
1190 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001191
1192 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1193 memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status);
1194 /* update the root cgroup regardless */
1195 atomic_long_add_return(amt, prot->memory_allocated);
1196 return memcg_memory_allocated_read(sk->sk_cgrp);
1197 }
1198
Glauber Costa180d8cd2011-12-11 21:47:02 +00001199 return atomic_long_add_return(amt, prot->memory_allocated);
1200}
1201
1202static inline void
Glauber Costa0e90b312012-01-20 04:57:16 +00001203sk_memory_allocated_sub(struct sock *sk, int amt)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001204{
1205 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001206
Glauber Costa0e90b312012-01-20 04:57:16 +00001207 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
Glauber Costae1aab162011-12-11 21:47:03 +00001208 memcg_memory_allocated_sub(sk->sk_cgrp, amt);
1209
Glauber Costa180d8cd2011-12-11 21:47:02 +00001210 atomic_long_sub(amt, prot->memory_allocated);
1211}
1212
1213static inline void sk_sockets_allocated_dec(struct sock *sk)
1214{
1215 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001216
1217 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1218 struct cg_proto *cg_proto = sk->sk_cgrp;
1219
1220 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1221 percpu_counter_dec(cg_proto->sockets_allocated);
1222 }
1223
Glauber Costa180d8cd2011-12-11 21:47:02 +00001224 percpu_counter_dec(prot->sockets_allocated);
1225}
1226
1227static inline void sk_sockets_allocated_inc(struct sock *sk)
1228{
1229 struct proto *prot = sk->sk_prot;
Glauber Costae1aab162011-12-11 21:47:03 +00001230
1231 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1232 struct cg_proto *cg_proto = sk->sk_cgrp;
1233
1234 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1235 percpu_counter_inc(cg_proto->sockets_allocated);
1236 }
1237
Glauber Costa180d8cd2011-12-11 21:47:02 +00001238 percpu_counter_inc(prot->sockets_allocated);
1239}
1240
1241static inline int
1242sk_sockets_allocated_read_positive(struct sock *sk)
1243{
1244 struct proto *prot = sk->sk_prot;
1245
Glauber Costae1aab162011-12-11 21:47:03 +00001246 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
Eric Dumazet518fbf92012-04-28 23:21:56 +00001247 return percpu_counter_read_positive(sk->sk_cgrp->sockets_allocated);
Glauber Costae1aab162011-12-11 21:47:03 +00001248
Eric Dumazet518fbf92012-04-28 23:21:56 +00001249 return percpu_counter_read_positive(prot->sockets_allocated);
Glauber Costa180d8cd2011-12-11 21:47:02 +00001250}
1251
1252static inline int
1253proto_sockets_allocated_sum_positive(struct proto *prot)
1254{
1255 return percpu_counter_sum_positive(prot->sockets_allocated);
1256}
1257
1258static inline long
1259proto_memory_allocated(struct proto *prot)
1260{
1261 return atomic_long_read(prot->memory_allocated);
1262}
1263
1264static inline bool
1265proto_memory_pressure(struct proto *prot)
1266{
1267 if (!prot->memory_pressure)
1268 return false;
1269 return !!*prot->memory_pressure;
1270}
1271
Eric Dumazet65f76512008-01-03 20:46:48 -08001272
1273#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274/* Called with local bh disabled */
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -07001275extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1276extern int sock_prot_inuse_get(struct net *net, struct proto *proto);
Eric Dumazet65f76512008-01-03 20:46:48 -08001277#else
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001278static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -07001279 int inc)
Eric Dumazet65f76512008-01-03 20:46:48 -08001280{
1281}
Eric Dumazet65f76512008-01-03 20:46:48 -08001282#endif
1283
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284
Arnaldo Carvalho de Melo614c6cb2005-08-09 19:47:37 -07001285/* With per-bucket locks this operation is not-atomic, so that
1286 * this version is not worse.
1287 */
1288static inline void __sk_prot_rehash(struct sock *sk)
1289{
1290 sk->sk_prot->unhash(sk);
1291 sk->sk_prot->hash(sk);
1292}
1293
Octavian Purdilafcbdf092010-12-16 14:26:56 -08001294void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
1295
Linus Torvalds1da177e2005-04-16 15:20:36 -07001296/* About 10 seconds */
1297#define SOCK_DESTROY_TIME (10*HZ)
1298
1299/* Sockets 0-1023 can't be bound to unless you are superuser */
1300#define PROT_SOCK 1024
1301
1302#define SHUTDOWN_MASK 3
1303#define RCV_SHUTDOWN 1
1304#define SEND_SHUTDOWN 2
1305
1306#define SOCK_SNDBUF_LOCK 1
1307#define SOCK_RCVBUF_LOCK 2
1308#define SOCK_BINDADDR_LOCK 4
1309#define SOCK_BINDPORT_LOCK 8
1310
1311/* sock_iocb: used to kick off async processing of socket ios */
1312struct sock_iocb {
1313 struct list_head list;
1314
1315 int flags;
1316 int size;
1317 struct socket *sock;
1318 struct sock *sk;
1319 struct scm_cookie *scm;
1320 struct msghdr *msg, async_msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001321 struct kiocb *kiocb;
1322};
1323
1324static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
1325{
1326 return (struct sock_iocb *)iocb->private;
1327}
1328
1329static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
1330{
1331 return si->kiocb;
1332}
1333
1334struct socket_alloc {
1335 struct socket socket;
1336 struct inode vfs_inode;
1337};
1338
1339static inline struct socket *SOCKET_I(struct inode *inode)
1340{
1341 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1342}
1343
1344static inline struct inode *SOCK_INODE(struct socket *socket)
1345{
1346 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1347}
1348
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001349/*
1350 * Functions for memory accounting
1351 */
1352extern int __sk_mem_schedule(struct sock *sk, int size, int kind);
1353extern void __sk_mem_reclaim(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001354
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001355#define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1356#define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1357#define SK_MEM_SEND 0
1358#define SK_MEM_RECV 1
Linus Torvalds1da177e2005-04-16 15:20:36 -07001359
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001360static inline int sk_mem_pages(int amt)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001361{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001362 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363}
1364
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001365static inline bool sk_has_account(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001366{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001367 /* return true if protocol supports memory accounting */
1368 return !!sk->sk_prot->memory_allocated;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001369}
1370
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001371static inline bool sk_wmem_schedule(struct sock *sk, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001373 if (!sk_has_account(sk))
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001374 return true;
Herbert Xud80d99d2005-09-01 17:48:23 -07001375 return size <= sk->sk_forward_alloc ||
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001376 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1377}
1378
Mel Gormanc76562b2012-07-31 16:44:41 -07001379static inline bool
Chuck Lever35c448a2012-09-17 14:09:11 -07001380sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size)
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001381{
1382 if (!sk_has_account(sk))
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001383 return true;
Mel Gormanc76562b2012-07-31 16:44:41 -07001384 return size<= sk->sk_forward_alloc ||
1385 __sk_mem_schedule(sk, size, SK_MEM_RECV) ||
1386 skb_pfmemalloc(skb);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001387}
1388
1389static inline void sk_mem_reclaim(struct sock *sk)
1390{
1391 if (!sk_has_account(sk))
1392 return;
1393 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1394 __sk_mem_reclaim(sk);
1395}
1396
David S. Miller9993e7d2008-01-10 21:56:38 -08001397static inline void sk_mem_reclaim_partial(struct sock *sk)
1398{
1399 if (!sk_has_account(sk))
1400 return;
1401 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1402 __sk_mem_reclaim(sk);
1403}
1404
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001405static inline void sk_mem_charge(struct sock *sk, int size)
1406{
1407 if (!sk_has_account(sk))
1408 return;
1409 sk->sk_forward_alloc -= size;
1410}
1411
1412static inline void sk_mem_uncharge(struct sock *sk, int size)
1413{
1414 if (!sk_has_account(sk))
1415 return;
1416 sk->sk_forward_alloc += size;
1417}
1418
1419static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1420{
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001421 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1422 sk->sk_wmem_queued -= skb->truesize;
1423 sk_mem_uncharge(sk, skb->truesize);
1424 __kfree_skb(skb);
Herbert Xud80d99d2005-09-01 17:48:23 -07001425}
1426
Linus Torvalds1da177e2005-04-16 15:20:36 -07001427/* Used by processes to "lock" a socket state, so that
1428 * interrupts and bottom half handlers won't change it
1429 * from under us. It essentially blocks any incoming
1430 * packets, so that we won't get any new data or any
1431 * packets that change the state of the socket.
1432 *
1433 * While locked, BH processing will add new packets to
1434 * the backlog queue. This queue is processed by the
1435 * owner of the socket lock right before it is released.
1436 *
1437 * Since ~2.3.5 it is also exclusive sleep lock serializing
1438 * accesses from user process context.
1439 */
John Heffnerd2e91172007-09-12 10:44:19 +02001440#define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001441
Peter Zijlstraed075362006-12-06 20:35:24 -08001442/*
1443 * Macro so as to not evaluate some arguments when
1444 * lockdep is not enabled.
1445 *
1446 * Mark both the sk_lock and the sk_lock.slock as a
1447 * per-address-family lock class.
1448 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001449#define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
Peter Zijlstraed075362006-12-06 20:35:24 -08001450do { \
Ingo Molnare8f6fbf2008-11-12 01:38:36 +00001451 sk->sk_lock.owned = 0; \
Peter Zijlstraed075362006-12-06 20:35:24 -08001452 init_waitqueue_head(&sk->sk_lock.wq); \
1453 spin_lock_init(&(sk)->sk_lock.slock); \
1454 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1455 sizeof((sk)->sk_lock)); \
1456 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001457 (skey), (sname)); \
Peter Zijlstraed075362006-12-06 20:35:24 -08001458 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1459} while (0)
1460
Harvey Harrison41380932007-12-12 10:46:51 -08001461extern void lock_sock_nested(struct sock *sk, int subclass);
Peter Zijlstrafcc70d52006-11-08 22:44:35 -08001462
1463static inline void lock_sock(struct sock *sk)
1464{
1465 lock_sock_nested(sk, 0);
1466}
1467
Harvey Harrison41380932007-12-12 10:46:51 -08001468extern void release_sock(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001469
1470/* BH context may only use the following locking interface. */
1471#define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
Ingo Molnarc6366182006-07-03 00:25:13 -07001472#define bh_lock_sock_nested(__sk) \
1473 spin_lock_nested(&((__sk)->sk_lock.slock), \
1474 SINGLE_DEPTH_NESTING)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001475#define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1476
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001477extern bool lock_sock_fast(struct sock *sk);
1478/**
1479 * unlock_sock_fast - complement of lock_sock_fast
1480 * @sk: socket
1481 * @slow: slow mode
1482 *
1483 * fast unlock socket for user context.
1484 * If slow mode is on, we call regular release_sock()
1485 */
1486static inline void unlock_sock_fast(struct sock *sk, bool slow)
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001487{
Eric Dumazet8a74ad62010-05-26 19:20:18 +00001488 if (slow)
1489 release_sock(sk);
1490 else
1491 spin_unlock_bh(&sk->sk_lock.slock);
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001492}
1493
Eric Dumazet4b0b72f2010-04-28 14:35:48 -07001494
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001495extern struct sock *sk_alloc(struct net *net, int family,
Al Virodd0fc662005-10-07 07:46:04 +01001496 gfp_t priority,
Pavel Emelyanov6257ff22007-11-01 00:39:31 -07001497 struct proto *prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001498extern void sk_free(struct sock *sk);
Denis V. Lunevedf02082008-02-29 11:18:32 -08001499extern void sk_release_kernel(struct sock *sk);
Eric Dumazete56c57d2011-11-08 17:07:07 -05001500extern struct sock *sk_clone_lock(const struct sock *sk,
1501 const gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001502
1503extern struct sk_buff *sock_wmalloc(struct sock *sk,
1504 unsigned long size, int force,
Al Virodd0fc662005-10-07 07:46:04 +01001505 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001506extern struct sk_buff *sock_rmalloc(struct sock *sk,
1507 unsigned long size, int force,
Al Virodd0fc662005-10-07 07:46:04 +01001508 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001509extern void sock_wfree(struct sk_buff *skb);
1510extern void sock_rfree(struct sk_buff *skb);
David S. Miller41063e92012-06-19 21:22:05 -07001511extern void sock_edemux(struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001512
1513extern int sock_setsockopt(struct socket *sock, int level,
1514 int op, char __user *optval,
David S. Millerb7058842009-09-30 16:12:20 -07001515 unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001516
1517extern int sock_getsockopt(struct socket *sock, int level,
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001518 int op, char __user *optval,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519 int __user *optlen);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001520extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001521 unsigned long size,
1522 int noblock,
1523 int *errcode);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001524extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
Herbert Xu4cc7f682009-02-04 16:55:54 -08001525 unsigned long header_len,
1526 unsigned long data_len,
1527 int noblock,
1528 int *errcode);
Victor Fusco86a76ca2005-07-08 14:57:47 -07001529extern void *sock_kmalloc(struct sock *sk, int size,
Al Virodd0fc662005-10-07 07:46:04 +01001530 gfp_t priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531extern void sock_kfree_s(struct sock *sk, void *mem, int size);
1532extern void sk_send_sigurg(struct sock *sk);
1533
1534/*
1535 * Functions to fill in entries in struct proto_ops when a protocol
1536 * does not implement a particular function.
1537 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001538extern int sock_no_bind(struct socket *,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539 struct sockaddr *, int);
1540extern int sock_no_connect(struct socket *,
1541 struct sockaddr *, int, int);
1542extern int sock_no_socketpair(struct socket *,
1543 struct socket *);
1544extern int sock_no_accept(struct socket *,
1545 struct socket *, int);
1546extern int sock_no_getname(struct socket *,
1547 struct sockaddr *, int *, int);
1548extern unsigned int sock_no_poll(struct file *, struct socket *,
1549 struct poll_table_struct *);
1550extern int sock_no_ioctl(struct socket *, unsigned int,
1551 unsigned long);
1552extern int sock_no_listen(struct socket *, int);
1553extern int sock_no_shutdown(struct socket *, int);
1554extern int sock_no_getsockopt(struct socket *, int , int,
1555 char __user *, int __user *);
1556extern int sock_no_setsockopt(struct socket *, int, int,
David S. Millerb7058842009-09-30 16:12:20 -07001557 char __user *, unsigned int);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558extern int sock_no_sendmsg(struct kiocb *, struct socket *,
1559 struct msghdr *, size_t);
1560extern int sock_no_recvmsg(struct kiocb *, struct socket *,
1561 struct msghdr *, size_t, int);
1562extern int sock_no_mmap(struct file *file,
1563 struct socket *sock,
1564 struct vm_area_struct *vma);
1565extern ssize_t sock_no_sendpage(struct socket *sock,
1566 struct page *page,
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001567 int offset, size_t size,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001568 int flags);
1569
1570/*
1571 * Functions to fill in entries in struct proto_ops when a protocol
1572 * uses the inet style.
1573 */
1574extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
1575 char __user *optval, int __user *optlen);
1576extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1577 struct msghdr *msg, size_t size, int flags);
1578extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07001579 char __user *optval, unsigned int optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001580extern int compat_sock_common_getsockopt(struct socket *sock, int level,
1581 int optname, char __user *optval, int __user *optlen);
1582extern int compat_sock_common_setsockopt(struct socket *sock, int level,
David S. Millerb7058842009-09-30 16:12:20 -07001583 int optname, char __user *optval, unsigned int optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001584
1585extern void sk_common_release(struct sock *sk);
1586
1587/*
1588 * Default socket callbacks and setup code
1589 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001590
Linus Torvalds1da177e2005-04-16 15:20:36 -07001591/* Initialise core socket variables */
1592extern void sock_init_data(struct socket *sock, struct sock *sk);
1593
Eric Dumazet46bcf142010-12-06 09:29:43 -08001594extern void sk_filter_release_rcu(struct rcu_head *rcu);
1595
Linus Torvalds1da177e2005-04-16 15:20:36 -07001596/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001597 * sk_filter_release - release a socket filter
Paul Bonserdc9b3342006-11-23 17:56:13 -08001598 * @fp: filter to remove
1599 *
1600 * Remove a filter from a socket and release its resources.
1601 */
1602
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -07001603static inline void sk_filter_release(struct sk_filter *fp)
1604{
1605 if (atomic_dec_and_test(&fp->refcnt))
Eric Dumazet80f8f102011-01-18 07:46:52 +00001606 call_rcu(&fp->rcu, sk_filter_release_rcu);
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -07001607}
1608
1609static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001610{
1611 unsigned int size = sk_filter_len(fp);
1612
1613 atomic_sub(size, &sk->sk_omem_alloc);
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -07001614 sk_filter_release(fp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001615}
1616
1617static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
1618{
1619 atomic_inc(&fp->refcnt);
1620 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
1621}
1622
1623/*
1624 * Socket reference counting postulates.
1625 *
1626 * * Each user of socket SHOULD hold a reference count.
1627 * * Each access point to socket (an hash table bucket, reference from a list,
1628 * running timer, skb in flight MUST hold a reference count.
1629 * * When reference count hits 0, it means it will never increase back.
1630 * * When reference count hits 0, it means that no references from
1631 * outside exist to this socket and current process on current CPU
1632 * is last user and may/should destroy this socket.
1633 * * sk_free is called from any context: process, BH, IRQ. When
1634 * it is called, socket has no references from outside -> sk_free
1635 * may release descendant resources allocated by the socket, but
1636 * to the time when it is called, socket is NOT referenced by any
1637 * hash tables, lists etc.
1638 * * Packets, delivered from outside (from network or from another process)
1639 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1640 * when they sit in queue. Otherwise, packets will leak to hole, when
1641 * socket is looked up by one cpu and unhasing is made by another CPU.
1642 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1643 * (leak to backlog). Packet socket does all the processing inside
1644 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1645 * use separate SMP lock, so that they are prone too.
1646 */
1647
1648/* Ungrab socket and destroy it, if it was the last reference. */
1649static inline void sock_put(struct sock *sk)
1650{
1651 if (atomic_dec_and_test(&sk->sk_refcnt))
1652 sk_free(sk);
1653}
1654
Arnaldo Carvalho de Melo58a5a7b2006-11-16 14:06:06 -02001655extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1656 const int nested);
Arnaldo Carvalho de Melo25995ff2005-12-27 02:42:22 -02001657
Krishna Kumare022f0b2009-10-19 23:46:20 +00001658static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1659{
1660 sk->sk_tx_queue_mapping = tx_queue;
1661}
1662
1663static inline void sk_tx_queue_clear(struct sock *sk)
1664{
1665 sk->sk_tx_queue_mapping = -1;
1666}
1667
1668static inline int sk_tx_queue_get(const struct sock *sk)
1669{
Tom Herbertb0f77d02010-07-14 20:50:29 -07001670 return sk ? sk->sk_tx_queue_mapping : -1;
Krishna Kumare022f0b2009-10-19 23:46:20 +00001671}
1672
David S. Miller972692e2008-06-17 22:41:38 -07001673static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1674{
Krishna Kumare022f0b2009-10-19 23:46:20 +00001675 sk_tx_queue_clear(sk);
David S. Miller972692e2008-06-17 22:41:38 -07001676 sk->sk_socket = sock;
1677}
1678
Eric Dumazetaa395142010-04-20 13:03:51 +00001679static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1680{
Eric Dumazeteaefd112011-02-18 03:26:36 +00001681 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1682 return &rcu_dereference_raw(sk->sk_wq)->wait;
Eric Dumazetaa395142010-04-20 13:03:51 +00001683}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001684/* Detach socket from process context.
1685 * Announce socket dead, detach it from wait queue and inode.
1686 * Note that parent inode held reference count on this struct sock,
1687 * we do not release it in this function, because protocol
1688 * probably wants some additional cleanups or even continuing
1689 * to work with this socket (TCP).
1690 */
1691static inline void sock_orphan(struct sock *sk)
1692{
1693 write_lock_bh(&sk->sk_callback_lock);
1694 sock_set_flag(sk, SOCK_DEAD);
David S. Miller972692e2008-06-17 22:41:38 -07001695 sk_set_socket(sk, NULL);
Eric Dumazet43815482010-04-29 11:01:49 +00001696 sk->sk_wq = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001697 write_unlock_bh(&sk->sk_callback_lock);
1698}
1699
1700static inline void sock_graft(struct sock *sk, struct socket *parent)
1701{
1702 write_lock_bh(&sk->sk_callback_lock);
Eric Dumazeteaefd112011-02-18 03:26:36 +00001703 sk->sk_wq = parent->wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001704 parent->sk = sk;
David S. Miller972692e2008-06-17 22:41:38 -07001705 sk_set_socket(sk, parent);
Venkat Yekkirala4237c752006-07-24 23:32:50 -07001706 security_sock_graft(sk, parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001707 write_unlock_bh(&sk->sk_callback_lock);
1708}
1709
Eric W. Biederman976d02012012-05-23 17:16:53 -06001710extern kuid_t sock_i_uid(struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001711extern unsigned long sock_i_ino(struct sock *sk);
1712
1713static inline struct dst_entry *
1714__sk_dst_get(struct sock *sk)
1715{
Michal Hockod8bf4ca2011-07-08 14:39:41 +02001716 return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
Eric Dumazetf68c2242010-04-22 16:06:59 -07001717 lockdep_is_held(&sk->sk_lock.slock));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001718}
1719
1720static inline struct dst_entry *
1721sk_dst_get(struct sock *sk)
1722{
1723 struct dst_entry *dst;
1724
Eric Dumazetb6c67122010-04-08 23:03:29 +00001725 rcu_read_lock();
1726 dst = rcu_dereference(sk->sk_dst_cache);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001727 if (dst)
1728 dst_hold(dst);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001729 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001730 return dst;
1731}
1732
Eric Dumazetb6c67122010-04-08 23:03:29 +00001733extern void sk_reset_txq(struct sock *sk);
1734
1735static inline void dst_negative_advice(struct sock *sk)
1736{
1737 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1738
1739 if (dst && dst->ops->negative_advice) {
1740 ndst = dst->ops->negative_advice(dst);
1741
1742 if (ndst != dst) {
1743 rcu_assign_pointer(sk->sk_dst_cache, ndst);
1744 sk_reset_txq(sk);
1745 }
1746 }
1747}
1748
Linus Torvalds1da177e2005-04-16 15:20:36 -07001749static inline void
1750__sk_dst_set(struct sock *sk, struct dst_entry *dst)
1751{
1752 struct dst_entry *old_dst;
1753
Krishna Kumare022f0b2009-10-19 23:46:20 +00001754 sk_tx_queue_clear(sk);
Eric Dumazet0b53ff22010-04-26 20:40:43 +00001755 /*
1756 * This can be called while sk is owned by the caller only,
1757 * with no state that can be checked in a rcu_dereference_check() cond
1758 */
1759 old_dst = rcu_dereference_raw(sk->sk_dst_cache);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001760 rcu_assign_pointer(sk->sk_dst_cache, dst);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001761 dst_release(old_dst);
1762}
1763
1764static inline void
1765sk_dst_set(struct sock *sk, struct dst_entry *dst)
1766{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001767 spin_lock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001768 __sk_dst_set(sk, dst);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001769 spin_unlock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001770}
1771
1772static inline void
1773__sk_dst_reset(struct sock *sk)
1774{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001775 __sk_dst_set(sk, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001776}
1777
1778static inline void
1779sk_dst_reset(struct sock *sk)
1780{
Eric Dumazetb6c67122010-04-08 23:03:29 +00001781 spin_lock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001782 __sk_dst_reset(sk);
Eric Dumazetb6c67122010-04-08 23:03:29 +00001783 spin_unlock(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001784}
1785
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001786extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001787
Denis Vlasenkof0088a52006-03-28 01:08:21 -08001788extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001789
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001790static inline bool sk_can_gso(const struct sock *sk)
Herbert Xubcd76112006-06-30 13:36:35 -07001791{
1792 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1793}
1794
Andi Kleen99580892007-04-20 17:12:43 -07001795extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
Arnaldo Carvalho de Melo6cbb0df2005-08-09 19:49:02 -07001796
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001797static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
Eric Dumazeta4654192010-05-16 00:36:33 -07001798{
1799 sk->sk_route_nocaps |= flags;
1800 sk->sk_route_caps &= ~flags;
1801}
1802
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001803static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1804 char __user *from, char *to,
Wei Yongjun912d3982011-04-06 18:40:12 +00001805 int copy, int offset)
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001806{
1807 if (skb->ip_summed == CHECKSUM_NONE) {
1808 int err = 0;
1809 __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err);
1810 if (err)
1811 return err;
Wei Yongjun912d3982011-04-06 18:40:12 +00001812 skb->csum = csum_block_add(skb->csum, csum, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001813 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1814 if (!access_ok(VERIFY_READ, from, copy) ||
1815 __copy_from_user_nocache(to, from, copy))
1816 return -EFAULT;
1817 } else if (copy_from_user(to, from, copy))
1818 return -EFAULT;
1819
1820 return 0;
1821}
1822
1823static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
1824 char __user *from, int copy)
1825{
Wei Yongjun912d3982011-04-06 18:40:12 +00001826 int err, offset = skb->len;
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001827
Wei Yongjun912d3982011-04-06 18:40:12 +00001828 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1829 copy, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001830 if (err)
Wei Yongjun912d3982011-04-06 18:40:12 +00001831 __skb_trim(skb, offset);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001832
1833 return err;
1834}
1835
1836static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from,
1837 struct sk_buff *skb,
1838 struct page *page,
1839 int off, int copy)
1840{
1841 int err;
1842
Wei Yongjun912d3982011-04-06 18:40:12 +00001843 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1844 copy, skb->len);
Tom Herbertc6e1a0d2011-04-04 22:30:30 -07001845 if (err)
1846 return err;
1847
1848 skb->len += copy;
1849 skb->data_len += copy;
1850 skb->truesize += copy;
1851 sk->sk_wmem_queued += copy;
1852 sk_mem_charge(sk, copy);
1853 return 0;
1854}
1855
Linus Torvalds1da177e2005-04-16 15:20:36 -07001856static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1857 struct sk_buff *skb, struct page *page,
1858 int off, int copy)
1859{
1860 if (skb->ip_summed == CHECKSUM_NONE) {
1861 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08001862 __wsum csum = csum_and_copy_from_user(from,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001863 page_address(page) + off,
1864 copy, 0, &err);
1865 if (err)
1866 return err;
1867 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1868 } else if (copy_from_user(page_address(page) + off, from, copy))
1869 return -EFAULT;
1870
1871 skb->len += copy;
1872 skb->data_len += copy;
1873 skb->truesize += copy;
1874 sk->sk_wmem_queued += copy;
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001875 sk_mem_charge(sk, copy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001876 return 0;
1877}
1878
Eric Dumazetc5640392009-06-16 10:12:03 +00001879/**
1880 * sk_wmem_alloc_get - returns write allocations
1881 * @sk: socket
1882 *
1883 * Returns sk_wmem_alloc minus initial offset of one
1884 */
1885static inline int sk_wmem_alloc_get(const struct sock *sk)
1886{
1887 return atomic_read(&sk->sk_wmem_alloc) - 1;
1888}
1889
1890/**
1891 * sk_rmem_alloc_get - returns read allocations
1892 * @sk: socket
1893 *
1894 * Returns sk_rmem_alloc
1895 */
1896static inline int sk_rmem_alloc_get(const struct sock *sk)
1897{
1898 return atomic_read(&sk->sk_rmem_alloc);
1899}
1900
1901/**
1902 * sk_has_allocations - check if allocations are outstanding
1903 * @sk: socket
1904 *
1905 * Returns true if socket has write or read allocations
1906 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001907static inline bool sk_has_allocations(const struct sock *sk)
Eric Dumazetc5640392009-06-16 10:12:03 +00001908{
1909 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1910}
1911
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001912/**
Eric Dumazet43815482010-04-29 11:01:49 +00001913 * wq_has_sleeper - check if there are any waiting processes
Randy Dunlapacfbe962010-05-24 23:54:18 -07001914 * @wq: struct socket_wq
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001915 *
Eric Dumazet43815482010-04-29 11:01:49 +00001916 * Returns true if socket_wq has waiting processes
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001917 *
Eric Dumazet43815482010-04-29 11:01:49 +00001918 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001919 * barrier call. They were added due to the race found within the tcp code.
1920 *
1921 * Consider following tcp code paths:
1922 *
1923 * CPU1 CPU2
1924 *
1925 * sys_select receive packet
1926 * ... ...
1927 * __add_wait_queue update tp->rcv_nxt
1928 * ... ...
1929 * tp->rcv_nxt check sock_def_readable
1930 * ... {
Eric Dumazet43815482010-04-29 11:01:49 +00001931 * schedule rcu_read_lock();
1932 * wq = rcu_dereference(sk->sk_wq);
1933 * if (wq && waitqueue_active(&wq->wait))
1934 * wake_up_interruptible(&wq->wait)
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001935 * ...
1936 * }
1937 *
1938 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1939 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1940 * could then endup calling schedule and sleep forever if there are no more
1941 * data on the socket.
Jiri Olsaad462762009-07-08 12:10:31 +00001942 *
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001943 */
Eric Dumazet43815482010-04-29 11:01:49 +00001944static inline bool wq_has_sleeper(struct socket_wq *wq)
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001945{
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001946 /* We need to be sure we are in sync with the
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001947 * add_wait_queue modifications to the wait queue.
1948 *
1949 * This memory barrier is paired in the sock_poll_wait.
1950 */
Eric Dumazet43815482010-04-29 11:01:49 +00001951 smp_mb();
1952 return wq && waitqueue_active(&wq->wait);
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001953}
1954
1955/**
1956 * sock_poll_wait - place memory barrier behind the poll_wait call.
1957 * @filp: file
1958 * @wait_address: socket wait queue
1959 * @p: poll_table
1960 *
Eric Dumazet43815482010-04-29 11:01:49 +00001961 * See the comments in the wq_has_sleeper function.
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001962 */
1963static inline void sock_poll_wait(struct file *filp,
1964 wait_queue_head_t *wait_address, poll_table *p)
1965{
Hans Verkuil626cf232012-03-23 15:02:27 -07001966 if (!poll_does_not_wait(p) && wait_address) {
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001967 poll_wait(filp, wait_address, p);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001968 /* We need to be sure we are in sync with the
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001969 * socket flags modification.
1970 *
Eric Dumazet43815482010-04-29 11:01:49 +00001971 * This memory barrier is paired in the wq_has_sleeper.
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001972 */
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001973 smp_mb();
1974 }
1975}
1976
Linus Torvalds1da177e2005-04-16 15:20:36 -07001977/*
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001978 * Queue a received datagram if it will fit. Stream and sequenced
Linus Torvalds1da177e2005-04-16 15:20:36 -07001979 * protocols can't normally use this as they need to fit buffers in
1980 * and play with them.
1981 *
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00001982 * Inlined as it's very short and called for pretty much every
Linus Torvalds1da177e2005-04-16 15:20:36 -07001983 * packet ever received.
1984 */
1985
1986static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1987{
Herbert Xud55d87f2009-06-22 02:25:25 +00001988 skb_orphan(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001989 skb->sk = sk;
1990 skb->destructor = sock_wfree;
Eric Dumazet2b85a342009-06-11 02:55:43 -07001991 /*
1992 * We used to take a refcount on sk, but following operation
1993 * is enough to guarantee sk_free() wont free this sock until
1994 * all in-flight packets are completed
1995 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001996 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1997}
1998
1999static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
2000{
Herbert Xud55d87f2009-06-22 02:25:25 +00002001 skb_orphan(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002002 skb->sk = sk;
2003 skb->destructor = sock_rfree;
2004 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08002005 sk_mem_charge(sk, skb->truesize);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002006}
2007
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002008extern void sk_reset_timer(struct sock *sk, struct timer_list *timer,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009 unsigned long expires);
2010
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002011extern void sk_stop_timer(struct sock *sk, struct timer_list *timer);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002012
Denis Vlasenkof0088a52006-03-28 01:08:21 -08002013extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002014
Eric Dumazetb1faf562010-05-31 23:44:05 -07002015extern int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002016
2017/*
2018 * Recover an error report and clear atomically
2019 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002020
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021static inline int sock_error(struct sock *sk)
2022{
Benjamin LaHaisec1cbe4b2005-12-13 23:22:19 -08002023 int err;
2024 if (likely(!sk->sk_err))
2025 return 0;
2026 err = xchg(&sk->sk_err, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002027 return -err;
2028}
2029
2030static inline unsigned long sock_wspace(struct sock *sk)
2031{
2032 int amt = 0;
2033
2034 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
2035 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002036 if (amt < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002037 amt = 0;
2038 }
2039 return amt;
2040}
2041
2042static inline void sk_wake_async(struct sock *sk, int how, int band)
2043{
Eric Dumazetbcdce712009-10-06 17:28:29 -07002044 if (sock_flag(sk, SOCK_FASYNC))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002045 sock_wake_async(sk->sk_socket, how, band);
2046}
2047
2048#define SOCK_MIN_SNDBUF 2048
Eric Dumazet7a91b4342010-09-26 18:53:07 -07002049/*
2050 * Since sk_rmem_alloc sums skb->truesize, even a small frame might need
2051 * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak
2052 */
2053#define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002054
2055static inline void sk_stream_moderate_sndbuf(struct sock *sk)
2056{
2057 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
Eric Dumazet8df09ea2007-12-21 03:07:41 -08002058 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002059 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
2060 }
2061}
2062
Pavel Emelyanovdf97c702007-11-29 21:22:33 +11002063struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002064
Eric Dumazet5640f762012-09-23 23:04:42 +00002065/**
2066 * sk_page_frag - return an appropriate page_frag
2067 * @sk: socket
2068 *
2069 * If socket allocation mode allows current thread to sleep, it means its
2070 * safe to use the per task page_frag instead of the per socket one.
2071 */
2072static inline struct page_frag *sk_page_frag(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002073{
Eric Dumazet5640f762012-09-23 23:04:42 +00002074 if (sk->sk_allocation & __GFP_WAIT)
2075 return &current->task_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002076
Eric Dumazet5640f762012-09-23 23:04:42 +00002077 return &sk->sk_frag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002078}
2079
Eric Dumazet5640f762012-09-23 23:04:42 +00002080extern bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag);
2081
Linus Torvalds1da177e2005-04-16 15:20:36 -07002082/*
2083 * Default write policy as shown to user space via poll/select/SIGIO
2084 */
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002085static inline bool sock_writeable(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002086{
Eric Dumazet8df09ea2007-12-21 03:07:41 -08002087 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002088}
2089
Al Virodd0fc662005-10-07 07:46:04 +01002090static inline gfp_t gfp_any(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002091{
Andrew Morton99709372009-02-12 16:43:17 -08002092 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002093}
2094
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002095static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002096{
2097 return noblock ? 0 : sk->sk_rcvtimeo;
2098}
2099
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002100static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002101{
2102 return noblock ? 0 : sk->sk_sndtimeo;
2103}
2104
2105static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2106{
2107 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
2108}
2109
2110/* Alas, with timeout socket operations are not restartable.
2111 * Compare this to poll().
2112 */
2113static inline int sock_intr_errno(long timeo)
2114{
2115 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2116}
2117
Eric Dumazet92f37fd2007-03-25 22:14:49 -07002118extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2119 struct sk_buff *skb);
Johannes Berg6e3e9392011-11-09 10:15:42 +01002120extern void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2121 struct sk_buff *skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -07002122
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002123static inline void
Linus Torvalds1da177e2005-04-16 15:20:36 -07002124sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2125{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002126 ktime_t kt = skb->tstamp;
Patrick Ohly20d49472009-02-12 05:03:38 +00002127 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002128
Patrick Ohly20d49472009-02-12 05:03:38 +00002129 /*
2130 * generate control messages if
2131 * - receive time stamping in software requested (SOCK_RCVTSTAMP
2132 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
2133 * - software time stamp available and wanted
2134 * (SOCK_TIMESTAMPING_SOFTWARE)
2135 * - hardware time stamps available and wanted
2136 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
2137 * SOCK_TIMESTAMPING_RAW_HARDWARE)
2138 */
2139 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
2140 sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
2141 (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
2142 (hwtstamps->hwtstamp.tv64 &&
2143 sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
2144 (hwtstamps->syststamp.tv64 &&
2145 sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
Eric Dumazet92f37fd2007-03-25 22:14:49 -07002146 __sock_recv_timestamp(msg, sk, skb);
2147 else
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002148 sk->sk_stamp = kt;
Johannes Berg6e3e9392011-11-09 10:15:42 +01002149
2150 if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2151 __sock_recv_wifi_status(msg, sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002152}
2153
Eric Dumazet767dd032010-04-28 19:14:43 +00002154extern void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2155 struct sk_buff *skb);
2156
2157static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2158 struct sk_buff *skb)
2159{
2160#define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2161 (1UL << SOCK_RCVTSTAMP) | \
2162 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
2163 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002164 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
Eric Dumazet767dd032010-04-28 19:14:43 +00002165 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
2166
2167 if (sk->sk_flags & FLAGS_TS_OR_DROPS)
2168 __sock_recv_ts_and_drops(msg, sk, skb);
2169 else
2170 sk->sk_stamp = skb->tstamp;
2171}
Neil Horman3b885782009-10-12 13:26:31 -07002172
Linus Torvalds1da177e2005-04-16 15:20:36 -07002173/**
Patrick Ohly20d49472009-02-12 05:03:38 +00002174 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
Patrick Ohly20d49472009-02-12 05:03:38 +00002175 * @sk: socket sending this packet
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002176 * @tx_flags: filled with instructions for time stamping
Patrick Ohly20d49472009-02-12 05:03:38 +00002177 *
Daniel Borkmannbf84a012013-04-14 08:08:13 +00002178 * Currently only depends on SOCK_TIMESTAMPING* flags.
Patrick Ohly20d49472009-02-12 05:03:38 +00002179 */
Daniel Borkmannbf84a012013-04-14 08:08:13 +00002180extern void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags);
Patrick Ohly20d49472009-02-12 05:03:38 +00002181
2182/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002183 * sk_eat_skb - Release a skb if it is no longer needed
Pavel Pisa4dc3b162005-05-01 08:59:25 -07002184 * @sk: socket to eat this skb from
2185 * @skb: socket buffer to eat
Randy Dunlapf4b8ea72006-06-22 16:00:11 -07002186 * @copied_early: flag indicating whether DMA operations copied this data early
Linus Torvalds1da177e2005-04-16 15:20:36 -07002187 *
2188 * This routine must be called with interrupts disabled or with the socket
2189 * locked so that the sk_buff queue operation is ok.
2190*/
Chris Leech624d1162006-05-23 18:01:28 -07002191#ifdef CONFIG_NET_DMA
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002192static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
Chris Leech624d1162006-05-23 18:01:28 -07002193{
2194 __skb_unlink(skb, &sk->sk_receive_queue);
2195 if (!copied_early)
2196 __kfree_skb(skb);
2197 else
2198 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
2199}
2200#else
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002201static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002202{
2203 __skb_unlink(skb, &sk->sk_receive_queue);
2204 __kfree_skb(skb);
2205}
Chris Leech624d1162006-05-23 18:01:28 -07002206#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002207
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002208static inline
2209struct net *sock_net(const struct sock *sk)
2210{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002211 return read_pnet(&sk->sk_net);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002212}
2213
2214static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07002215void sock_net_set(struct sock *sk, struct net *net)
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002216{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002217 write_pnet(&sk->sk_net, net);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002218}
2219
Denis V. Lunevedf02082008-02-29 11:18:32 -08002220/*
2221 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
Lucas De Marchi25985ed2011-03-30 22:57:33 -03002222 * They should not hold a reference to a namespace in order to allow
Denis V. Lunevedf02082008-02-29 11:18:32 -08002223 * to stop it.
2224 * Sockets after sk_change_net should be released using sk_release_kernel
2225 */
2226static inline void sk_change_net(struct sock *sk, struct net *net)
2227{
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09002228 put_net(sock_net(sk));
Denis V. Lunev65a18ec2008-04-16 01:59:46 -07002229 sock_net_set(sk, hold_net(net));
Denis V. Lunevedf02082008-02-29 11:18:32 -08002230}
2231
KOVACS Krisztian23542612008-10-07 12:41:01 -07002232static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2233{
Vijay Subramanianefc27f82012-06-24 13:03:07 +00002234 if (skb->sk) {
KOVACS Krisztian23542612008-10-07 12:41:01 -07002235 struct sock *sk = skb->sk;
2236
2237 skb->destructor = NULL;
2238 skb->sk = NULL;
2239 return sk;
2240 }
2241 return NULL;
2242}
2243
Patrick Ohly20d49472009-02-12 05:03:38 +00002244extern void sock_enable_timestamp(struct sock *sk, int flag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002245extern int sock_get_timestamp(struct sock *, struct timeval __user *);
Eric Dumazetae40eb12007-03-18 17:33:16 -07002246extern int sock_get_timestampns(struct sock *, struct timespec __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002247
Eric Dumazetdc6b9b72012-05-16 22:48:15 +00002248/*
2249 * Enable debug/info messages
Linus Torvalds1da177e2005-04-16 15:20:36 -07002250 */
Stephen Hemmingera2a316f2007-03-08 20:41:08 -08002251extern int net_msg_warn;
2252#define NETDEBUG(fmt, args...) \
2253 do { if (net_msg_warn) printk(fmt,##args); } while (0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002254
Stephen Hemmingera2a316f2007-03-08 20:41:08 -08002255#define LIMIT_NETDEBUG(fmt, args...) \
2256 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002257
Linus Torvalds1da177e2005-04-16 15:20:36 -07002258extern __u32 sysctl_wmem_max;
2259extern __u32 sysctl_rmem_max;
2260
David S. Miller6baf1f42005-09-05 18:14:11 -07002261extern int sysctl_optmem_max;
2262
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002263extern __u32 sysctl_wmem_default;
2264extern __u32 sysctl_rmem_default;
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002265
Linus Torvalds1da177e2005-04-16 15:20:36 -07002266#endif /* _SOCK_H */