]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - include/net/sock.h
Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
[mirror_ubuntu-artful-kernel.git] / include / net / sock.h
CommitLineData
1da177e4
LT
1/*
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 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
11 * 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
a6b7a407 43#include <linux/hardirq.h>
172589cc 44#include <linux/kernel.h>
1da177e4 45#include <linux/list.h>
88ab1932 46#include <linux/list_nulls.h>
1da177e4
LT
47#include <linux/timer.h>
48#include <linux/cache.h>
3f134619 49#include <linux/bitops.h>
a5b5bb9a 50#include <linux/lockdep.h>
1da177e4
LT
51#include <linux/netdevice.h>
52#include <linux/skbuff.h> /* struct sk_buff */
d7fe0f24 53#include <linux/mm.h>
1da177e4 54#include <linux/security.h>
5a0e3ad6 55#include <linux/slab.h>
c6e1a0d1 56#include <linux/uaccess.h>
180d8cd9 57#include <linux/memcontrol.h>
e1aab161 58#include <linux/res_counter.h>
c5905afb 59#include <linux/static_key.h>
40401530
AV
60#include <linux/aio.h>
61#include <linux/sched.h>
1da177e4
LT
62
63#include <linux/filter.h>
88ab1932 64#include <linux/rculist_nulls.h>
a57de0b4 65#include <linux/poll.h>
1da177e4 66
c31504dc 67#include <linux/atomic.h>
1da177e4
LT
68#include <net/dst.h>
69#include <net/checksum.h>
70
9f048bfb
ED
71struct cgroup;
72struct cgroup_subsys;
c607b2ed 73#ifdef CONFIG_NET
1d62e436
GC
74int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss);
75void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg);
c607b2ed
GC
76#else
77static inline
1d62e436 78int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
c607b2ed
GC
79{
80 return 0;
81}
82static inline
1d62e436 83void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg)
c607b2ed
GC
84{
85}
86#endif
1da177e4
LT
87/*
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
4cd9029d 99/* Validate arguments and do nothing */
b9075fa9 100static inline __printf(2, 3)
dc6b9b78 101void SOCK_DEBUG(const struct sock *sk, const char *msg, ...)
4cd9029d
SH
102{
103}
1da177e4
LT
104#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 */
1da177e4
LT
110typedef struct {
111 spinlock_t slock;
d2e9117c 112 int owned;
1da177e4 113 wait_queue_head_t wq;
a5b5bb9a
IM
114 /*
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
1da177e4
LT
123} socket_lock_t;
124
1da177e4 125struct sock;
8feaf0c0 126struct proto;
0eeb8ffc 127struct net;
1da177e4
LT
128
129/**
4dc3b16b 130 * struct sock_common - minimal network layer representation of sockets
68835aba
ED
131 * @skc_daddr: Foreign IPv4 addr
132 * @skc_rcv_saddr: Bound local IPv4 addr
4dc6dc71 133 * @skc_hash: hash value used with various protocol lookup tables
d4cada4a 134 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
4dc3b16b
PP
135 * @skc_family: network address family
136 * @skc_state: Connection state
137 * @skc_reuse: %SO_REUSEADDR setting
138 * @skc_bound_dev_if: bound device index if != 0
4dc3b16b 139 * @skc_bind_node: bind hash linkage for various protocol lookup tables
512615b6 140 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
8feaf0c0 141 * @skc_prot: protocol handlers inside a network family
07feaebf 142 * @skc_net: reference to the network namespace of this socket
68835aba
ED
143 * @skc_node: main hash linkage for various protocol lookup tables
144 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
145 * @skc_tx_queue_mapping: tx queue number for this connection
146 * @skc_refcnt: reference count
4dc3b16b
PP
147 *
148 * This is the minimal network layer representation of sockets, the header
8feaf0c0
ACM
149 * for struct sock and struct inet_timewait_sock.
150 */
1da177e4 151struct sock_common {
68835aba
ED
152 /* skc_daddr and skc_rcv_saddr must be grouped :
153 * cf INET_MATCH() and INET_TW_MATCH()
4dc6dc71 154 */
68835aba
ED
155 __be32 skc_daddr;
156 __be32 skc_rcv_saddr;
4dc6dc71 157
d4cada4a
ED
158 union {
159 unsigned int skc_hash;
160 __u16 skc_u16hashes[2];
161 };
4dc6dc71
ED
162 unsigned short skc_family;
163 volatile unsigned char skc_state;
164 unsigned char skc_reuse;
165 int skc_bound_dev_if;
512615b6
ED
166 union {
167 struct hlist_node skc_bind_node;
168 struct hlist_nulls_node skc_portaddr_node;
169 };
8feaf0c0 170 struct proto *skc_prot;
3b1e0a65 171#ifdef CONFIG_NET_NS
07feaebf 172 struct net *skc_net;
3b1e0a65 173#endif
68835aba
ED
174 /*
175 * fields between dontcopy_begin/dontcopy_end
176 * are not copied in sock_copy()
177 */
928c41e7 178 /* private: */
68835aba 179 int skc_dontcopy_begin[0];
928c41e7 180 /* public: */
68835aba
ED
181 union {
182 struct hlist_node skc_node;
183 struct hlist_nulls_node skc_nulls_node;
184 };
185 int skc_tx_queue_mapping;
186 atomic_t skc_refcnt;
928c41e7 187 /* private: */
68835aba 188 int skc_dontcopy_end[0];
928c41e7 189 /* public: */
1da177e4
LT
190};
191
e1aab161 192struct cg_proto;
1da177e4
LT
193/**
194 * struct sock - network layer representation of sockets
8feaf0c0 195 * @__sk_common: shared layout with inet_timewait_sock
4dc3b16b
PP
196 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
197 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
198 * @sk_lock: synchronizer
199 * @sk_rcvbuf: size of receive buffer in bytes
43815482 200 * @sk_wq: sock wait queue and async head
4dc3b16b
PP
201 * @sk_dst_cache: destination cache
202 * @sk_dst_lock: destination cache lock
203 * @sk_policy: flow policy
4dc3b16b
PP
204 * @sk_receive_queue: incoming packets
205 * @sk_wmem_alloc: transmit queue bytes committed
206 * @sk_write_queue: Packet sending queue
97fc2f08 207 * @sk_async_wait_queue: DMA copied packets
4dc3b16b
PP
208 * @sk_omem_alloc: "o" is "option" or "other"
209 * @sk_wmem_queued: persistent queue size
210 * @sk_forward_alloc: space allocated forward
211 * @sk_allocation: allocation mode
212 * @sk_sndbuf: size of send buffer in bytes
33c732c3 213 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
20d49473 214 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
4dc3b16b
PP
215 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
216 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
a465419b 217 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
bcd76111 218 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
82cc1a7a 219 * @sk_gso_max_size: Maximum GSO segment size to build
4dc3b16b 220 * @sk_lingertime: %SO_LINGER l_linger setting
4dc3b16b
PP
221 * @sk_backlog: always used with the per-socket spinlock held
222 * @sk_callback_lock: used with the callbacks in the end of this struct
223 * @sk_error_queue: rarely used
33c732c3
WC
224 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
225 * IPV6_ADDRFORM for instance)
4dc3b16b 226 * @sk_err: last error
33c732c3
WC
227 * @sk_err_soft: errors that don't cause failure but are the cause of a
228 * persistent failure not just 'timed out'
cb61cb9b 229 * @sk_drops: raw/udp drops counter
4dc3b16b
PP
230 * @sk_ack_backlog: current listen backlog
231 * @sk_max_ack_backlog: listen backlog set in listen()
232 * @sk_priority: %SO_PRIORITY setting
1a3bc369 233 * @sk_cgrp_prioidx: socket group's priority map index
4dc3b16b
PP
234 * @sk_type: socket type (%SOCK_STREAM, etc)
235 * @sk_protocol: which protocol this socket belongs in this network family
53c3fa20
RD
236 * @sk_peer_pid: &struct pid for this socket's peer
237 * @sk_peer_cred: %SO_PEERCRED setting
4dc3b16b
PP
238 * @sk_rcvlowat: %SO_RCVLOWAT setting
239 * @sk_rcvtimeo: %SO_RCVTIMEO setting
240 * @sk_sndtimeo: %SO_SNDTIMEO setting
c58dc01b 241 * @sk_rxhash: flow hash received from netif layer
4dc3b16b
PP
242 * @sk_filter: socket filtering instructions
243 * @sk_protinfo: private area, net family specific, when not using slab
244 * @sk_timer: sock cleanup timer
245 * @sk_stamp: time stamp of last packet received
246 * @sk_socket: Identd and reporting IO signals
247 * @sk_user_data: RPC layer private data
248 * @sk_sndmsg_page: cached page for sendmsg
249 * @sk_sndmsg_off: cached offset for sendmsg
d3d4f0a0 250 * @sk_peek_off: current peek_offset value
4dc3b16b 251 * @sk_send_head: front of stuff to transmit
67be2dd1 252 * @sk_security: used by security modules
31729363 253 * @sk_mark: generic packet mark
53c3fa20 254 * @sk_classid: this socket's cgroup classid
e1aab161 255 * @sk_cgrp: this socket's cgroup-specific proto data
4dc3b16b
PP
256 * @sk_write_pending: a write to stream socket waits to start
257 * @sk_state_change: callback to indicate change in the state of the sock
258 * @sk_data_ready: callback to indicate there is data to be processed
259 * @sk_write_space: callback to indicate there is bf sending space available
260 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
261 * @sk_backlog_rcv: callback to process the backlog
262 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
1da177e4
LT
263 */
264struct sock {
265 /*
8feaf0c0 266 * Now struct inet_timewait_sock also uses sock_common, so please just
1da177e4
LT
267 * don't add nothing before this first member (__sk_common) --acme
268 */
269 struct sock_common __sk_common;
4dc6dc71
ED
270#define sk_node __sk_common.skc_node
271#define sk_nulls_node __sk_common.skc_nulls_node
272#define sk_refcnt __sk_common.skc_refcnt
e022f0b4 273#define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
4dc6dc71 274
68835aba
ED
275#define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
276#define sk_dontcopy_end __sk_common.skc_dontcopy_end
4dc6dc71 277#define sk_hash __sk_common.skc_hash
1da177e4
LT
278#define sk_family __sk_common.skc_family
279#define sk_state __sk_common.skc_state
280#define sk_reuse __sk_common.skc_reuse
281#define sk_bound_dev_if __sk_common.skc_bound_dev_if
1da177e4 282#define sk_bind_node __sk_common.skc_bind_node
8feaf0c0 283#define sk_prot __sk_common.skc_prot
07feaebf 284#define sk_net __sk_common.skc_net
1da177e4 285 socket_lock_t sk_lock;
b178bb3d 286 struct sk_buff_head sk_receive_queue;
fa438ccf
ED
287 /*
288 * The backlog queue is special, it is always used with
289 * the per-socket spinlock held and requires low latency
290 * access. Therefore we special case it's implementation.
b178bb3d
ED
291 * Note : rmem_alloc is in this structure to fill a hole
292 * on 64bit arches, not because its logically part of
293 * backlog.
fa438ccf
ED
294 */
295 struct {
b178bb3d
ED
296 atomic_t rmem_alloc;
297 int len;
298 struct sk_buff *head;
299 struct sk_buff *tail;
fa438ccf 300 } sk_backlog;
b178bb3d
ED
301#define sk_rmem_alloc sk_backlog.rmem_alloc
302 int sk_forward_alloc;
303#ifdef CONFIG_RPS
304 __u32 sk_rxhash;
305#endif
306 atomic_t sk_drops;
307 int sk_rcvbuf;
308
309 struct sk_filter __rcu *sk_filter;
eaefd110 310 struct socket_wq __rcu *sk_wq;
b178bb3d
ED
311
312#ifdef CONFIG_NET_DMA
313 struct sk_buff_head sk_async_wait_queue;
314#endif
315
def8b4fa 316#ifdef CONFIG_XFRM
1da177e4 317 struct xfrm_policy *sk_policy[2];
def8b4fa 318#endif
b178bb3d
ED
319 unsigned long sk_flags;
320 struct dst_entry *sk_dst_cache;
b6c6712a 321 spinlock_t sk_dst_lock;
41063e9d 322 struct dst_entry *sk_rx_dst;
1da177e4
LT
323 atomic_t sk_wmem_alloc;
324 atomic_t sk_omem_alloc;
4e07a91c 325 int sk_sndbuf;
1da177e4 326 struct sk_buff_head sk_write_queue;
b178bb3d
ED
327 kmemcheck_bitfield_begin(flags);
328 unsigned int sk_shutdown : 2,
329 sk_no_check : 2,
330 sk_userlocks : 4,
331 sk_protocol : 8,
332 sk_type : 16;
333 kmemcheck_bitfield_end(flags);
1da177e4 334 int sk_wmem_queued;
7d877f3b 335 gfp_t sk_allocation;
c8f44aff
MM
336 netdev_features_t sk_route_caps;
337 netdev_features_t sk_route_nocaps;
bcd76111 338 int sk_gso_type;
82cc1a7a 339 unsigned int sk_gso_max_size;
9932cf95 340 int sk_rcvlowat;
1da177e4 341 unsigned long sk_lingertime;
1da177e4 342 struct sk_buff_head sk_error_queue;
476e19cf 343 struct proto *sk_prot_creator;
1da177e4
LT
344 rwlock_t sk_callback_lock;
345 int sk_err,
346 sk_err_soft;
347 unsigned short sk_ack_backlog;
348 unsigned short sk_max_ack_backlog;
349 __u32 sk_priority;
5bc1421e
NH
350#ifdef CONFIG_CGROUPS
351 __u32 sk_cgrp_prioidx;
352#endif
109f6e39
EB
353 struct pid *sk_peer_pid;
354 const struct cred *sk_peer_cred;
1da177e4
LT
355 long sk_rcvtimeo;
356 long sk_sndtimeo;
1da177e4
LT
357 void *sk_protinfo;
358 struct timer_list sk_timer;
b7aa0bf7 359 ktime_t sk_stamp;
1da177e4
LT
360 struct socket *sk_socket;
361 void *sk_user_data;
362 struct page *sk_sndmsg_page;
363 struct sk_buff *sk_send_head;
364 __u32 sk_sndmsg_off;
ef64a54f 365 __s32 sk_peek_off;
1da177e4 366 int sk_write_pending;
d5f64238 367#ifdef CONFIG_SECURITY
1da177e4 368 void *sk_security;
d5f64238 369#endif
4a19ec58 370 __u32 sk_mark;
f8451725 371 u32 sk_classid;
e1aab161 372 struct cg_proto *sk_cgrp;
1da177e4
LT
373 void (*sk_state_change)(struct sock *sk);
374 void (*sk_data_ready)(struct sock *sk, int bytes);
375 void (*sk_write_space)(struct sock *sk);
376 void (*sk_error_report)(struct sock *sk);
dc6b9b78
ED
377 int (*sk_backlog_rcv)(struct sock *sk,
378 struct sk_buff *skb);
1da177e4
LT
379 void (*sk_destruct)(struct sock *sk);
380};
381
4a17fd52
PE
382/*
383 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
384 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
385 * on a socket means that the socket will reuse everybody else's port
386 * without looking at the other's sk_reuse value.
387 */
388
389#define SK_NO_REUSE 0
390#define SK_CAN_REUSE 1
391#define SK_FORCE_REUSE 2
392
ef64a54f
PE
393static inline int sk_peek_offset(struct sock *sk, int flags)
394{
395 if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0))
396 return sk->sk_peek_off;
397 else
398 return 0;
399}
400
401static inline void sk_peek_offset_bwd(struct sock *sk, int val)
402{
403 if (sk->sk_peek_off >= 0) {
404 if (sk->sk_peek_off >= val)
405 sk->sk_peek_off -= val;
406 else
407 sk->sk_peek_off = 0;
408 }
409}
410
411static inline void sk_peek_offset_fwd(struct sock *sk, int val)
412{
413 if (sk->sk_peek_off >= 0)
414 sk->sk_peek_off += val;
415}
416
1da177e4
LT
417/*
418 * Hashed lists helper routines
419 */
c4146644
LZ
420static inline struct sock *sk_entry(const struct hlist_node *node)
421{
422 return hlist_entry(node, struct sock, sk_node);
423}
424
e48c414e 425static inline struct sock *__sk_head(const struct hlist_head *head)
1da177e4
LT
426{
427 return hlist_entry(head->first, struct sock, sk_node);
428}
429
e48c414e 430static inline struct sock *sk_head(const struct hlist_head *head)
1da177e4
LT
431{
432 return hlist_empty(head) ? NULL : __sk_head(head);
433}
434
88ab1932
ED
435static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
436{
437 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
438}
439
440static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
441{
442 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
443}
444
e48c414e 445static inline struct sock *sk_next(const struct sock *sk)
1da177e4
LT
446{
447 return sk->sk_node.next ?
448 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
449}
450
88ab1932
ED
451static inline struct sock *sk_nulls_next(const struct sock *sk)
452{
453 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
454 hlist_nulls_entry(sk->sk_nulls_node.next,
455 struct sock, sk_nulls_node) :
456 NULL;
457}
458
dc6b9b78 459static inline bool sk_unhashed(const struct sock *sk)
1da177e4
LT
460{
461 return hlist_unhashed(&sk->sk_node);
462}
463
dc6b9b78 464static inline bool sk_hashed(const struct sock *sk)
1da177e4 465{
da753bea 466 return !sk_unhashed(sk);
1da177e4
LT
467}
468
dc6b9b78 469static inline void sk_node_init(struct hlist_node *node)
1da177e4
LT
470{
471 node->pprev = NULL;
472}
473
dc6b9b78 474static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
88ab1932
ED
475{
476 node->pprev = NULL;
477}
478
dc6b9b78 479static inline void __sk_del_node(struct sock *sk)
1da177e4
LT
480{
481 __hlist_del(&sk->sk_node);
482}
483
808f5114 484/* NB: equivalent to hlist_del_init_rcu */
dc6b9b78 485static inline bool __sk_del_node_init(struct sock *sk)
1da177e4
LT
486{
487 if (sk_hashed(sk)) {
488 __sk_del_node(sk);
489 sk_node_init(&sk->sk_node);
dc6b9b78 490 return true;
1da177e4 491 }
dc6b9b78 492 return false;
1da177e4
LT
493}
494
495/* Grab socket reference count. This operation is valid only
496 when sk is ALREADY grabbed f.e. it is found in hash table
497 or a list and the lookup is made under lock preventing hash table
498 modifications.
499 */
500
501static inline void sock_hold(struct sock *sk)
502{
503 atomic_inc(&sk->sk_refcnt);
504}
505
506/* Ungrab socket in the context, which assumes that socket refcnt
507 cannot hit zero, f.e. it is true in context of any socketcall.
508 */
509static inline void __sock_put(struct sock *sk)
510{
511 atomic_dec(&sk->sk_refcnt);
512}
513
dc6b9b78 514static inline bool sk_del_node_init(struct sock *sk)
1da177e4 515{
dc6b9b78 516 bool rc = __sk_del_node_init(sk);
1da177e4
LT
517
518 if (rc) {
519 /* paranoid for a while -acme */
520 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
521 __sock_put(sk);
522 }
523 return rc;
524}
808f5114 525#define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
1da177e4 526
dc6b9b78 527static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
271b72c7
ED
528{
529 if (sk_hashed(sk)) {
88ab1932 530 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
dc6b9b78 531 return true;
271b72c7 532 }
dc6b9b78 533 return false;
271b72c7
ED
534}
535
dc6b9b78 536static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
271b72c7 537{
dc6b9b78 538 bool rc = __sk_nulls_del_node_init_rcu(sk);
271b72c7
ED
539
540 if (rc) {
541 /* paranoid for a while -acme */
542 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
543 __sock_put(sk);
544 }
545 return rc;
546}
547
dc6b9b78 548static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
1da177e4
LT
549{
550 hlist_add_head(&sk->sk_node, list);
551}
552
dc6b9b78 553static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
1da177e4
LT
554{
555 sock_hold(sk);
556 __sk_add_node(sk, list);
557}
558
dc6b9b78 559static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
808f5114 560{
561 sock_hold(sk);
562 hlist_add_head_rcu(&sk->sk_node, list);
563}
564
dc6b9b78 565static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
271b72c7 566{
88ab1932 567 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
271b72c7
ED
568}
569
dc6b9b78 570static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
271b72c7
ED
571{
572 sock_hold(sk);
88ab1932 573 __sk_nulls_add_node_rcu(sk, list);
271b72c7
ED
574}
575
dc6b9b78 576static inline void __sk_del_bind_node(struct sock *sk)
1da177e4
LT
577{
578 __hlist_del(&sk->sk_bind_node);
579}
580
dc6b9b78 581static inline void sk_add_bind_node(struct sock *sk,
1da177e4
LT
582 struct hlist_head *list)
583{
584 hlist_add_head(&sk->sk_bind_node, list);
585}
586
587#define sk_for_each(__sk, node, list) \
588 hlist_for_each_entry(__sk, node, list, sk_node)
808f5114 589#define sk_for_each_rcu(__sk, node, list) \
590 hlist_for_each_entry_rcu(__sk, node, list, sk_node)
88ab1932
ED
591#define sk_nulls_for_each(__sk, node, list) \
592 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
593#define sk_nulls_for_each_rcu(__sk, node, list) \
594 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
1da177e4
LT
595#define sk_for_each_from(__sk, node) \
596 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
597 hlist_for_each_entry_from(__sk, node, sk_node)
88ab1932
ED
598#define sk_nulls_for_each_from(__sk, node) \
599 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
600 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
1da177e4
LT
601#define sk_for_each_safe(__sk, node, tmp, list) \
602 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
603#define sk_for_each_bound(__sk, node, list) \
604 hlist_for_each_entry(__sk, node, list, sk_bind_node)
605
606/* Sock flags */
607enum sock_flags {
608 SOCK_DEAD,
609 SOCK_DONE,
610 SOCK_URGINLINE,
611 SOCK_KEEPOPEN,
612 SOCK_LINGER,
613 SOCK_DESTROY,
614 SOCK_BROADCAST,
615 SOCK_TIMESTAMP,
616 SOCK_ZAPPED,
617 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
618 SOCK_DBG, /* %SO_DEBUG setting */
619 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
92f37fd2 620 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
1da177e4
LT
621 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
622 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
20d49473
PO
623 SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */
624 SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
625 SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */
626 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
627 SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */
628 SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
629 SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
bcdce719 630 SOCK_FASYNC, /* fasync() active */
3b885787 631 SOCK_RXQ_OVFL,
1cdebb42 632 SOCK_ZEROCOPY, /* buffers from userspace */
6e3e939f 633 SOCK_WIFI_STATUS, /* push wifi status to userspace */
3bdc0eba
BG
634 SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
635 * Will use last 4 bytes of packet sent from
636 * user-space instead.
637 */
1da177e4
LT
638};
639
53b924b3
RB
640static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
641{
642 nsk->sk_flags = osk->sk_flags;
643}
644
1da177e4
LT
645static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
646{
647 __set_bit(flag, &sk->sk_flags);
648}
649
650static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
651{
652 __clear_bit(flag, &sk->sk_flags);
653}
654
1b23a5df 655static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
1da177e4
LT
656{
657 return test_bit(flag, &sk->sk_flags);
658}
659
660static inline void sk_acceptq_removed(struct sock *sk)
661{
662 sk->sk_ack_backlog--;
663}
664
665static inline void sk_acceptq_added(struct sock *sk)
666{
667 sk->sk_ack_backlog++;
668}
669
dc6b9b78 670static inline bool sk_acceptq_is_full(const struct sock *sk)
1da177e4 671{
64a14651 672 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
1da177e4
LT
673}
674
675/*
676 * Compute minimal free write space needed to queue new packets.
677 */
dc6b9b78 678static inline int sk_stream_min_wspace(const struct sock *sk)
1da177e4 679{
8df09ea3 680 return sk->sk_wmem_queued >> 1;
1da177e4
LT
681}
682
dc6b9b78 683static inline int sk_stream_wspace(const struct sock *sk)
1da177e4
LT
684{
685 return sk->sk_sndbuf - sk->sk_wmem_queued;
686}
687
688extern void sk_stream_write_space(struct sock *sk);
689
dc6b9b78 690static inline bool sk_stream_memory_free(const struct sock *sk)
1da177e4
LT
691{
692 return sk->sk_wmem_queued < sk->sk_sndbuf;
693}
694
8eae939f 695/* OOB backlog add */
a3a858ff 696static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
9ee6b535 697{
7fee226a
ED
698 /* dont let skb dst not refcounted, we are going to leave rcu lock */
699 skb_dst_force(skb);
700
701 if (!sk->sk_backlog.tail)
702 sk->sk_backlog.head = skb;
703 else
9ee6b535 704 sk->sk_backlog.tail->next = skb;
7fee226a
ED
705
706 sk->sk_backlog.tail = skb;
9ee6b535
SH
707 skb->next = NULL;
708}
1da177e4 709
c377411f
ED
710/*
711 * Take into account size of receive queue and backlog queue
0fd7bac6
ED
712 * Do not take into account this skb truesize,
713 * to allow even a single big packet to come.
c377411f 714 */
f545a38f
ED
715static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb,
716 unsigned int limit)
c377411f
ED
717{
718 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
719
f545a38f 720 return qsize > limit;
c377411f
ED
721}
722
8eae939f 723/* The per-socket spinlock must be held here. */
f545a38f
ED
724static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
725 unsigned int limit)
8eae939f 726{
f545a38f 727 if (sk_rcvqueues_full(sk, skb, limit))
8eae939f
ZY
728 return -ENOBUFS;
729
a3a858ff 730 __sk_add_backlog(sk, skb);
8eae939f
ZY
731 sk->sk_backlog.len += skb->truesize;
732 return 0;
733}
734
c57943a1
PZ
735static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
736{
737 return sk->sk_backlog_rcv(sk, skb);
738}
739
c58dc01b
DM
740static inline void sock_rps_record_flow(const struct sock *sk)
741{
742#ifdef CONFIG_RPS
743 struct rps_sock_flow_table *sock_flow_table;
744
745 rcu_read_lock();
746 sock_flow_table = rcu_dereference(rps_sock_flow_table);
747 rps_record_sock_flow(sock_flow_table, sk->sk_rxhash);
748 rcu_read_unlock();
749#endif
750}
751
752static inline void sock_rps_reset_flow(const struct sock *sk)
753{
754#ifdef CONFIG_RPS
755 struct rps_sock_flow_table *sock_flow_table;
756
757 rcu_read_lock();
758 sock_flow_table = rcu_dereference(rps_sock_flow_table);
759 rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash);
760 rcu_read_unlock();
761#endif
762}
763
bdeab991
TH
764static inline void sock_rps_save_rxhash(struct sock *sk,
765 const struct sk_buff *skb)
c58dc01b
DM
766{
767#ifdef CONFIG_RPS
bdeab991 768 if (unlikely(sk->sk_rxhash != skb->rxhash)) {
c58dc01b 769 sock_rps_reset_flow(sk);
bdeab991 770 sk->sk_rxhash = skb->rxhash;
c58dc01b
DM
771 }
772#endif
773}
774
bdeab991
TH
775static inline void sock_rps_reset_rxhash(struct sock *sk)
776{
777#ifdef CONFIG_RPS
778 sock_rps_reset_flow(sk);
779 sk->sk_rxhash = 0;
780#endif
781}
782
cfcabdcc
SH
783#define sk_wait_event(__sk, __timeo, __condition) \
784 ({ int __rc; \
785 release_sock(__sk); \
786 __rc = __condition; \
787 if (!__rc) { \
788 *(__timeo) = schedule_timeout(*(__timeo)); \
789 } \
790 lock_sock(__sk); \
791 __rc = __condition; \
792 __rc; \
793 })
1da177e4
LT
794
795extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
796extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
797extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
798extern int sk_stream_error(struct sock *sk, int flags, int err);
799extern void sk_stream_kill_queues(struct sock *sk);
800
801extern int sk_wait_data(struct sock *sk, long *timeo);
802
60236fdd 803struct request_sock_ops;
6d6ee43e 804struct timewait_sock_ops;
ab1e0a13 805struct inet_hashinfo;
fc8717ba 806struct raw_hashinfo;
de477254 807struct module;
2e6599cb 808
1da177e4
LT
809/* Networking protocol blocks we attach to sockets.
810 * socket layer -> transport layer interface
811 * transport -> network interface is defined by struct inet_proto
812 */
813struct proto {
dc6b9b78 814 void (*close)(struct sock *sk,
1da177e4
LT
815 long timeout);
816 int (*connect)(struct sock *sk,
dc6b9b78 817 struct sockaddr *uaddr,
1da177e4
LT
818 int addr_len);
819 int (*disconnect)(struct sock *sk, int flags);
820
dc6b9b78 821 struct sock * (*accept)(struct sock *sk, int flags, int *err);
1da177e4
LT
822
823 int (*ioctl)(struct sock *sk, int cmd,
824 unsigned long arg);
825 int (*init)(struct sock *sk);
7d06b2e0 826 void (*destroy)(struct sock *sk);
1da177e4 827 void (*shutdown)(struct sock *sk, int how);
dc6b9b78 828 int (*setsockopt)(struct sock *sk, int level,
1da177e4 829 int optname, char __user *optval,
b7058842 830 unsigned int optlen);
dc6b9b78
ED
831 int (*getsockopt)(struct sock *sk, int level,
832 int optname, char __user *optval,
833 int __user *option);
af01d537 834#ifdef CONFIG_COMPAT
3fdadf7d
DM
835 int (*compat_setsockopt)(struct sock *sk,
836 int level,
837 int optname, char __user *optval,
b7058842 838 unsigned int optlen);
3fdadf7d
DM
839 int (*compat_getsockopt)(struct sock *sk,
840 int level,
841 int optname, char __user *optval,
842 int __user *option);
709b46e8
EB
843 int (*compat_ioctl)(struct sock *sk,
844 unsigned int cmd, unsigned long arg);
af01d537 845#endif
1da177e4
LT
846 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
847 struct msghdr *msg, size_t len);
848 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
849 struct msghdr *msg,
dc6b9b78
ED
850 size_t len, int noblock, int flags,
851 int *addr_len);
1da177e4
LT
852 int (*sendpage)(struct sock *sk, struct page *page,
853 int offset, size_t size, int flags);
dc6b9b78 854 int (*bind)(struct sock *sk,
1da177e4
LT
855 struct sockaddr *uaddr, int addr_len);
856
dc6b9b78 857 int (*backlog_rcv) (struct sock *sk,
1da177e4
LT
858 struct sk_buff *skb);
859
860 /* Keeping track of sk's, looking them up, and port selection methods. */
861 void (*hash)(struct sock *sk);
862 void (*unhash)(struct sock *sk);
719f8358 863 void (*rehash)(struct sock *sk);
1da177e4 864 int (*get_port)(struct sock *sk, unsigned short snum);
fcbdf09d 865 void (*clear_sk)(struct sock *sk, int size);
1da177e4 866
286ab3d4 867 /* Keeping track of sockets in use */
65f76517 868#ifdef CONFIG_PROC_FS
13ff3d6f 869 unsigned int inuse_idx;
65f76517 870#endif
ebb53d75 871
1da177e4 872 /* Memory pressure */
5c52ba17 873 void (*enter_memory_pressure)(struct sock *sk);
8d987e5c 874 atomic_long_t *memory_allocated; /* Current allocated memory. */
1748376b 875 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
1da177e4
LT
876 /*
877 * Pressure flag: try to collapse.
878 * Technical note: it is used by multiple contexts non atomically.
3ab224be 879 * All the __sk_mem_schedule() is of this nature: accounting
1da177e4
LT
880 * is strict, actions are advisory and have some latency.
881 */
882 int *memory_pressure;
8d987e5c 883 long *sysctl_mem;
1da177e4
LT
884 int *sysctl_wmem;
885 int *sysctl_rmem;
886 int max_header;
7ba42910 887 bool no_autobind;
1da177e4 888
271b72c7 889 struct kmem_cache *slab;
1da177e4 890 unsigned int obj_size;
271b72c7 891 int slab_flags;
1da177e4 892
dd24c001 893 struct percpu_counter *orphan_count;
8feaf0c0 894
60236fdd 895 struct request_sock_ops *rsk_prot;
6d6ee43e 896 struct timewait_sock_ops *twsk_prot;
2e6599cb 897
39d8cda7
PE
898 union {
899 struct inet_hashinfo *hashinfo;
645ca708 900 struct udp_table *udp_table;
fc8717ba 901 struct raw_hashinfo *raw_hash;
39d8cda7 902 } h;
ab1e0a13 903
1da177e4
LT
904 struct module *owner;
905
906 char name[32];
907
908 struct list_head node;
e6848976
ACM
909#ifdef SOCK_REFCNT_DEBUG
910 atomic_t socks;
911#endif
e1aab161
GC
912#ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
913 /*
914 * cgroup specific init/deinit functions. Called once for all
915 * protocols that implement it, from cgroups populate function.
916 * This function has to setup any files the protocol want to
917 * appear in the kmem cgroup filesystem.
918 */
1d62e436 919 int (*init_cgroup)(struct mem_cgroup *memcg,
e1aab161 920 struct cgroup_subsys *ss);
1d62e436 921 void (*destroy_cgroup)(struct mem_cgroup *memcg);
e1aab161
GC
922 struct cg_proto *(*proto_cgroup)(struct mem_cgroup *memcg);
923#endif
924};
925
3f134619
GC
926/*
927 * Bits in struct cg_proto.flags
928 */
929enum cg_proto_flags {
930 /* Currently active and new sockets should be assigned to cgroups */
931 MEMCG_SOCK_ACTIVE,
932 /* It was ever activated; we must disarm static keys on destruction */
933 MEMCG_SOCK_ACTIVATED,
934};
935
e1aab161
GC
936struct cg_proto {
937 void (*enter_memory_pressure)(struct sock *sk);
938 struct res_counter *memory_allocated; /* Current allocated memory. */
939 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
940 int *memory_pressure;
941 long *sysctl_mem;
3f134619 942 unsigned long flags;
e1aab161
GC
943 /*
944 * memcg field is used to find which memcg we belong directly
945 * Each memcg struct can hold more than one cg_proto, so container_of
946 * won't really cut.
947 *
948 * The elegant solution would be having an inverse function to
949 * proto_cgroup in struct proto, but that means polluting the structure
950 * for everybody, instead of just for memcg users.
951 */
952 struct mem_cgroup *memcg;
1da177e4
LT
953};
954
955extern int proto_register(struct proto *prot, int alloc_slab);
956extern void proto_unregister(struct proto *prot);
957
3f134619
GC
958static inline bool memcg_proto_active(struct cg_proto *cg_proto)
959{
960 return test_bit(MEMCG_SOCK_ACTIVE, &cg_proto->flags);
961}
962
963static inline bool memcg_proto_activated(struct cg_proto *cg_proto)
964{
965 return test_bit(MEMCG_SOCK_ACTIVATED, &cg_proto->flags);
966}
967
e6848976
ACM
968#ifdef SOCK_REFCNT_DEBUG
969static inline void sk_refcnt_debug_inc(struct sock *sk)
970{
971 atomic_inc(&sk->sk_prot->socks);
972}
973
974static inline void sk_refcnt_debug_dec(struct sock *sk)
975{
976 atomic_dec(&sk->sk_prot->socks);
977 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
978 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
979}
980
e1aab161 981inline void sk_refcnt_debug_release(const struct sock *sk)
e6848976
ACM
982{
983 if (atomic_read(&sk->sk_refcnt) != 1)
984 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
985 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
986}
987#else /* SOCK_REFCNT_DEBUG */
988#define sk_refcnt_debug_inc(sk) do { } while (0)
989#define sk_refcnt_debug_dec(sk) do { } while (0)
990#define sk_refcnt_debug_release(sk) do { } while (0)
991#endif /* SOCK_REFCNT_DEBUG */
992
376be5ff 993#if defined(CONFIG_CGROUP_MEM_RES_CTLR_KMEM) && defined(CONFIG_NET)
c5905afb 994extern struct static_key memcg_socket_limit_enabled;
e1aab161
GC
995static inline struct cg_proto *parent_cg_proto(struct proto *proto,
996 struct cg_proto *cg_proto)
997{
998 return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg));
999}
c5905afb 1000#define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
e1aab161
GC
1001#else
1002#define mem_cgroup_sockets_enabled 0
1003static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1004 struct cg_proto *cg_proto)
1005{
1006 return NULL;
1007}
1008#endif
1009
1010
180d8cd9
GC
1011static inline bool sk_has_memory_pressure(const struct sock *sk)
1012{
1013 return sk->sk_prot->memory_pressure != NULL;
1014}
1015
1016static inline bool sk_under_memory_pressure(const struct sock *sk)
1017{
1018 if (!sk->sk_prot->memory_pressure)
1019 return false;
e1aab161
GC
1020
1021 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1022 return !!*sk->sk_cgrp->memory_pressure;
1023
180d8cd9
GC
1024 return !!*sk->sk_prot->memory_pressure;
1025}
1026
1027static inline void sk_leave_memory_pressure(struct sock *sk)
1028{
1029 int *memory_pressure = sk->sk_prot->memory_pressure;
1030
e1aab161
GC
1031 if (!memory_pressure)
1032 return;
1033
1034 if (*memory_pressure)
180d8cd9 1035 *memory_pressure = 0;
e1aab161
GC
1036
1037 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1038 struct cg_proto *cg_proto = sk->sk_cgrp;
1039 struct proto *prot = sk->sk_prot;
1040
1041 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1042 if (*cg_proto->memory_pressure)
1043 *cg_proto->memory_pressure = 0;
1044 }
1045
180d8cd9
GC
1046}
1047
1048static inline void sk_enter_memory_pressure(struct sock *sk)
1049{
e1aab161
GC
1050 if (!sk->sk_prot->enter_memory_pressure)
1051 return;
1052
1053 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1054 struct cg_proto *cg_proto = sk->sk_cgrp;
1055 struct proto *prot = sk->sk_prot;
1056
1057 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1058 cg_proto->enter_memory_pressure(sk);
1059 }
1060
1061 sk->sk_prot->enter_memory_pressure(sk);
180d8cd9
GC
1062}
1063
1064static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1065{
1066 long *prot = sk->sk_prot->sysctl_mem;
e1aab161
GC
1067 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1068 prot = sk->sk_cgrp->sysctl_mem;
180d8cd9
GC
1069 return prot[index];
1070}
1071
e1aab161
GC
1072static inline void memcg_memory_allocated_add(struct cg_proto *prot,
1073 unsigned long amt,
1074 int *parent_status)
1075{
1076 struct res_counter *fail;
1077 int ret;
1078
0e90b31f
GC
1079 ret = res_counter_charge_nofail(prot->memory_allocated,
1080 amt << PAGE_SHIFT, &fail);
e1aab161
GC
1081 if (ret < 0)
1082 *parent_status = OVER_LIMIT;
1083}
1084
1085static inline void memcg_memory_allocated_sub(struct cg_proto *prot,
1086 unsigned long amt)
1087{
1088 res_counter_uncharge(prot->memory_allocated, amt << PAGE_SHIFT);
1089}
1090
1091static inline u64 memcg_memory_allocated_read(struct cg_proto *prot)
1092{
1093 u64 ret;
1094 ret = res_counter_read_u64(prot->memory_allocated, RES_USAGE);
1095 return ret >> PAGE_SHIFT;
1096}
1097
180d8cd9
GC
1098static inline long
1099sk_memory_allocated(const struct sock *sk)
1100{
1101 struct proto *prot = sk->sk_prot;
e1aab161
GC
1102 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1103 return memcg_memory_allocated_read(sk->sk_cgrp);
1104
180d8cd9
GC
1105 return atomic_long_read(prot->memory_allocated);
1106}
1107
1108static inline long
e1aab161 1109sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status)
180d8cd9
GC
1110{
1111 struct proto *prot = sk->sk_prot;
e1aab161
GC
1112
1113 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1114 memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status);
1115 /* update the root cgroup regardless */
1116 atomic_long_add_return(amt, prot->memory_allocated);
1117 return memcg_memory_allocated_read(sk->sk_cgrp);
1118 }
1119
180d8cd9
GC
1120 return atomic_long_add_return(amt, prot->memory_allocated);
1121}
1122
1123static inline void
0e90b31f 1124sk_memory_allocated_sub(struct sock *sk, int amt)
180d8cd9
GC
1125{
1126 struct proto *prot = sk->sk_prot;
e1aab161 1127
0e90b31f 1128 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
e1aab161
GC
1129 memcg_memory_allocated_sub(sk->sk_cgrp, amt);
1130
180d8cd9
GC
1131 atomic_long_sub(amt, prot->memory_allocated);
1132}
1133
1134static inline void sk_sockets_allocated_dec(struct sock *sk)
1135{
1136 struct proto *prot = sk->sk_prot;
e1aab161
GC
1137
1138 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1139 struct cg_proto *cg_proto = sk->sk_cgrp;
1140
1141 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1142 percpu_counter_dec(cg_proto->sockets_allocated);
1143 }
1144
180d8cd9
GC
1145 percpu_counter_dec(prot->sockets_allocated);
1146}
1147
1148static inline void sk_sockets_allocated_inc(struct sock *sk)
1149{
1150 struct proto *prot = sk->sk_prot;
e1aab161
GC
1151
1152 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1153 struct cg_proto *cg_proto = sk->sk_cgrp;
1154
1155 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1156 percpu_counter_inc(cg_proto->sockets_allocated);
1157 }
1158
180d8cd9
GC
1159 percpu_counter_inc(prot->sockets_allocated);
1160}
1161
1162static inline int
1163sk_sockets_allocated_read_positive(struct sock *sk)
1164{
1165 struct proto *prot = sk->sk_prot;
1166
e1aab161 1167 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
518fbf9c 1168 return percpu_counter_read_positive(sk->sk_cgrp->sockets_allocated);
e1aab161 1169
518fbf9c 1170 return percpu_counter_read_positive(prot->sockets_allocated);
180d8cd9
GC
1171}
1172
1173static inline int
1174proto_sockets_allocated_sum_positive(struct proto *prot)
1175{
1176 return percpu_counter_sum_positive(prot->sockets_allocated);
1177}
1178
1179static inline long
1180proto_memory_allocated(struct proto *prot)
1181{
1182 return atomic_long_read(prot->memory_allocated);
1183}
1184
1185static inline bool
1186proto_memory_pressure(struct proto *prot)
1187{
1188 if (!prot->memory_pressure)
1189 return false;
1190 return !!*prot->memory_pressure;
1191}
1192
65f76517
ED
1193
1194#ifdef CONFIG_PROC_FS
1da177e4 1195/* Called with local bh disabled */
c29a0bc4
PE
1196extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1197extern int sock_prot_inuse_get(struct net *net, struct proto *proto);
65f76517 1198#else
dc6b9b78 1199static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
c29a0bc4 1200 int inc)
65f76517
ED
1201{
1202}
65f76517
ED
1203#endif
1204
1da177e4 1205
614c6cb4
ACM
1206/* With per-bucket locks this operation is not-atomic, so that
1207 * this version is not worse.
1208 */
1209static inline void __sk_prot_rehash(struct sock *sk)
1210{
1211 sk->sk_prot->unhash(sk);
1212 sk->sk_prot->hash(sk);
1213}
1214
fcbdf09d
OP
1215void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
1216
1da177e4
LT
1217/* About 10 seconds */
1218#define SOCK_DESTROY_TIME (10*HZ)
1219
1220/* Sockets 0-1023 can't be bound to unless you are superuser */
1221#define PROT_SOCK 1024
1222
1223#define SHUTDOWN_MASK 3
1224#define RCV_SHUTDOWN 1
1225#define SEND_SHUTDOWN 2
1226
1227#define SOCK_SNDBUF_LOCK 1
1228#define SOCK_RCVBUF_LOCK 2
1229#define SOCK_BINDADDR_LOCK 4
1230#define SOCK_BINDPORT_LOCK 8
1231
1232/* sock_iocb: used to kick off async processing of socket ios */
1233struct sock_iocb {
1234 struct list_head list;
1235
1236 int flags;
1237 int size;
1238 struct socket *sock;
1239 struct sock *sk;
1240 struct scm_cookie *scm;
1241 struct msghdr *msg, async_msg;
1da177e4
LT
1242 struct kiocb *kiocb;
1243};
1244
1245static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
1246{
1247 return (struct sock_iocb *)iocb->private;
1248}
1249
1250static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
1251{
1252 return si->kiocb;
1253}
1254
1255struct socket_alloc {
1256 struct socket socket;
1257 struct inode vfs_inode;
1258};
1259
1260static inline struct socket *SOCKET_I(struct inode *inode)
1261{
1262 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1263}
1264
1265static inline struct inode *SOCK_INODE(struct socket *socket)
1266{
1267 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1268}
1269
3ab224be
HA
1270/*
1271 * Functions for memory accounting
1272 */
1273extern int __sk_mem_schedule(struct sock *sk, int size, int kind);
1274extern void __sk_mem_reclaim(struct sock *sk);
1da177e4 1275
3ab224be
HA
1276#define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1277#define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1278#define SK_MEM_SEND 0
1279#define SK_MEM_RECV 1
1da177e4 1280
3ab224be 1281static inline int sk_mem_pages(int amt)
1da177e4 1282{
3ab224be 1283 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
1da177e4
LT
1284}
1285
dc6b9b78 1286static inline bool sk_has_account(struct sock *sk)
1da177e4 1287{
3ab224be
HA
1288 /* return true if protocol supports memory accounting */
1289 return !!sk->sk_prot->memory_allocated;
1da177e4
LT
1290}
1291
dc6b9b78 1292static inline bool sk_wmem_schedule(struct sock *sk, int size)
1da177e4 1293{
3ab224be 1294 if (!sk_has_account(sk))
dc6b9b78 1295 return true;
3ab224be
HA
1296 return size <= sk->sk_forward_alloc ||
1297 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1da177e4
LT
1298}
1299
dc6b9b78 1300static inline bool sk_rmem_schedule(struct sock *sk, int size)
d80d99d6 1301{
3ab224be 1302 if (!sk_has_account(sk))
dc6b9b78 1303 return true;
d80d99d6 1304 return size <= sk->sk_forward_alloc ||
3ab224be
HA
1305 __sk_mem_schedule(sk, size, SK_MEM_RECV);
1306}
1307
1308static inline void sk_mem_reclaim(struct sock *sk)
1309{
1310 if (!sk_has_account(sk))
1311 return;
1312 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1313 __sk_mem_reclaim(sk);
1314}
1315
9993e7d3
DM
1316static inline void sk_mem_reclaim_partial(struct sock *sk)
1317{
1318 if (!sk_has_account(sk))
1319 return;
1320 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1321 __sk_mem_reclaim(sk);
1322}
1323
3ab224be
HA
1324static inline void sk_mem_charge(struct sock *sk, int size)
1325{
1326 if (!sk_has_account(sk))
1327 return;
1328 sk->sk_forward_alloc -= size;
1329}
1330
1331static inline void sk_mem_uncharge(struct sock *sk, int size)
1332{
1333 if (!sk_has_account(sk))
1334 return;
1335 sk->sk_forward_alloc += size;
1336}
1337
1338static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1339{
3ab224be
HA
1340 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1341 sk->sk_wmem_queued -= skb->truesize;
1342 sk_mem_uncharge(sk, skb->truesize);
1343 __kfree_skb(skb);
d80d99d6
HX
1344}
1345
1da177e4
LT
1346/* Used by processes to "lock" a socket state, so that
1347 * interrupts and bottom half handlers won't change it
1348 * from under us. It essentially blocks any incoming
1349 * packets, so that we won't get any new data or any
1350 * packets that change the state of the socket.
1351 *
1352 * While locked, BH processing will add new packets to
1353 * the backlog queue. This queue is processed by the
1354 * owner of the socket lock right before it is released.
1355 *
1356 * Since ~2.3.5 it is also exclusive sleep lock serializing
1357 * accesses from user process context.
1358 */
d2e9117c 1359#define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
1da177e4 1360
ed07536e
PZ
1361/*
1362 * Macro so as to not evaluate some arguments when
1363 * lockdep is not enabled.
1364 *
1365 * Mark both the sk_lock and the sk_lock.slock as a
1366 * per-address-family lock class.
1367 */
dc6b9b78 1368#define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
ed07536e 1369do { \
e8f6fbf6 1370 sk->sk_lock.owned = 0; \
ed07536e
PZ
1371 init_waitqueue_head(&sk->sk_lock.wq); \
1372 spin_lock_init(&(sk)->sk_lock.slock); \
1373 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1374 sizeof((sk)->sk_lock)); \
1375 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
dc6b9b78 1376 (skey), (sname)); \
ed07536e
PZ
1377 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1378} while (0)
1379
41380930 1380extern void lock_sock_nested(struct sock *sk, int subclass);
fcc70d5f
PZ
1381
1382static inline void lock_sock(struct sock *sk)
1383{
1384 lock_sock_nested(sk, 0);
1385}
1386
41380930 1387extern void release_sock(struct sock *sk);
1da177e4
LT
1388
1389/* BH context may only use the following locking interface. */
1390#define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
c6366184
IM
1391#define bh_lock_sock_nested(__sk) \
1392 spin_lock_nested(&((__sk)->sk_lock.slock), \
1393 SINGLE_DEPTH_NESTING)
1da177e4
LT
1394#define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1395
8a74ad60
ED
1396extern bool lock_sock_fast(struct sock *sk);
1397/**
1398 * unlock_sock_fast - complement of lock_sock_fast
1399 * @sk: socket
1400 * @slow: slow mode
1401 *
1402 * fast unlock socket for user context.
1403 * If slow mode is on, we call regular release_sock()
1404 */
1405static inline void unlock_sock_fast(struct sock *sk, bool slow)
4b0b72f7 1406{
8a74ad60
ED
1407 if (slow)
1408 release_sock(sk);
1409 else
1410 spin_unlock_bh(&sk->sk_lock.slock);
4b0b72f7
ED
1411}
1412
4b0b72f7 1413
1b8d7ae4 1414extern struct sock *sk_alloc(struct net *net, int family,
dd0fc66f 1415 gfp_t priority,
6257ff21 1416 struct proto *prot);
1da177e4 1417extern void sk_free(struct sock *sk);
edf02087 1418extern void sk_release_kernel(struct sock *sk);
e56c57d0
ED
1419extern struct sock *sk_clone_lock(const struct sock *sk,
1420 const gfp_t priority);
1da177e4
LT
1421
1422extern struct sk_buff *sock_wmalloc(struct sock *sk,
1423 unsigned long size, int force,
dd0fc66f 1424 gfp_t priority);
1da177e4
LT
1425extern struct sk_buff *sock_rmalloc(struct sock *sk,
1426 unsigned long size, int force,
dd0fc66f 1427 gfp_t priority);
1da177e4
LT
1428extern void sock_wfree(struct sk_buff *skb);
1429extern void sock_rfree(struct sk_buff *skb);
41063e9d 1430extern void sock_edemux(struct sk_buff *skb);
1da177e4
LT
1431
1432extern int sock_setsockopt(struct socket *sock, int level,
1433 int op, char __user *optval,
b7058842 1434 unsigned int optlen);
1da177e4
LT
1435
1436extern int sock_getsockopt(struct socket *sock, int level,
dc6b9b78 1437 int op, char __user *optval,
1da177e4 1438 int __user *optlen);
dc6b9b78 1439extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
1da177e4
LT
1440 unsigned long size,
1441 int noblock,
1442 int *errcode);
dc6b9b78 1443extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
4cc7f68d
HX
1444 unsigned long header_len,
1445 unsigned long data_len,
1446 int noblock,
1447 int *errcode);
86a76caf 1448extern void *sock_kmalloc(struct sock *sk, int size,
dd0fc66f 1449 gfp_t priority);
1da177e4
LT
1450extern void sock_kfree_s(struct sock *sk, void *mem, int size);
1451extern void sk_send_sigurg(struct sock *sk);
1452
f8451725
HX
1453#ifdef CONFIG_CGROUPS
1454extern void sock_update_classid(struct sock *sk);
1455#else
1456static inline void sock_update_classid(struct sock *sk)
1457{
1458}
1459#endif
1460
1da177e4
LT
1461/*
1462 * Functions to fill in entries in struct proto_ops when a protocol
1463 * does not implement a particular function.
1464 */
dc6b9b78 1465extern int sock_no_bind(struct socket *,
1da177e4
LT
1466 struct sockaddr *, int);
1467extern int sock_no_connect(struct socket *,
1468 struct sockaddr *, int, int);
1469extern int sock_no_socketpair(struct socket *,
1470 struct socket *);
1471extern int sock_no_accept(struct socket *,
1472 struct socket *, int);
1473extern int sock_no_getname(struct socket *,
1474 struct sockaddr *, int *, int);
1475extern unsigned int sock_no_poll(struct file *, struct socket *,
1476 struct poll_table_struct *);
1477extern int sock_no_ioctl(struct socket *, unsigned int,
1478 unsigned long);
1479extern int sock_no_listen(struct socket *, int);
1480extern int sock_no_shutdown(struct socket *, int);
1481extern int sock_no_getsockopt(struct socket *, int , int,
1482 char __user *, int __user *);
1483extern int sock_no_setsockopt(struct socket *, int, int,
b7058842 1484 char __user *, unsigned int);
1da177e4
LT
1485extern int sock_no_sendmsg(struct kiocb *, struct socket *,
1486 struct msghdr *, size_t);
1487extern int sock_no_recvmsg(struct kiocb *, struct socket *,
1488 struct msghdr *, size_t, int);
1489extern int sock_no_mmap(struct file *file,
1490 struct socket *sock,
1491 struct vm_area_struct *vma);
1492extern ssize_t sock_no_sendpage(struct socket *sock,
1493 struct page *page,
dc6b9b78 1494 int offset, size_t size,
1da177e4
LT
1495 int flags);
1496
1497/*
1498 * Functions to fill in entries in struct proto_ops when a protocol
1499 * uses the inet style.
1500 */
1501extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
1502 char __user *optval, int __user *optlen);
1503extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1504 struct msghdr *msg, size_t size, int flags);
1505extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 1506 char __user *optval, unsigned int optlen);
3fdadf7d
DM
1507extern int compat_sock_common_getsockopt(struct socket *sock, int level,
1508 int optname, char __user *optval, int __user *optlen);
1509extern int compat_sock_common_setsockopt(struct socket *sock, int level,
b7058842 1510 int optname, char __user *optval, unsigned int optlen);
1da177e4
LT
1511
1512extern void sk_common_release(struct sock *sk);
1513
1514/*
1515 * Default socket callbacks and setup code
1516 */
dc6b9b78 1517
1da177e4
LT
1518/* Initialise core socket variables */
1519extern void sock_init_data(struct socket *sock, struct sock *sk);
1520
46bcf14f
ED
1521extern void sk_filter_release_rcu(struct rcu_head *rcu);
1522
dc9b3346 1523/**
1a5778aa 1524 * sk_filter_release - release a socket filter
dc9b3346
PB
1525 * @fp: filter to remove
1526 *
1527 * Remove a filter from a socket and release its resources.
1528 */
1529
309dd5fc
PE
1530static inline void sk_filter_release(struct sk_filter *fp)
1531{
1532 if (atomic_dec_and_test(&fp->refcnt))
80f8f102 1533 call_rcu(&fp->rcu, sk_filter_release_rcu);
309dd5fc
PE
1534}
1535
1536static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
1da177e4
LT
1537{
1538 unsigned int size = sk_filter_len(fp);
1539
1540 atomic_sub(size, &sk->sk_omem_alloc);
309dd5fc 1541 sk_filter_release(fp);
1da177e4
LT
1542}
1543
1544static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
1545{
1546 atomic_inc(&fp->refcnt);
1547 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
1548}
1549
1550/*
1551 * Socket reference counting postulates.
1552 *
1553 * * Each user of socket SHOULD hold a reference count.
1554 * * Each access point to socket (an hash table bucket, reference from a list,
1555 * running timer, skb in flight MUST hold a reference count.
1556 * * When reference count hits 0, it means it will never increase back.
1557 * * When reference count hits 0, it means that no references from
1558 * outside exist to this socket and current process on current CPU
1559 * is last user and may/should destroy this socket.
1560 * * sk_free is called from any context: process, BH, IRQ. When
1561 * it is called, socket has no references from outside -> sk_free
1562 * may release descendant resources allocated by the socket, but
1563 * to the time when it is called, socket is NOT referenced by any
1564 * hash tables, lists etc.
1565 * * Packets, delivered from outside (from network or from another process)
1566 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1567 * when they sit in queue. Otherwise, packets will leak to hole, when
1568 * socket is looked up by one cpu and unhasing is made by another CPU.
1569 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1570 * (leak to backlog). Packet socket does all the processing inside
1571 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1572 * use separate SMP lock, so that they are prone too.
1573 */
1574
1575/* Ungrab socket and destroy it, if it was the last reference. */
1576static inline void sock_put(struct sock *sk)
1577{
1578 if (atomic_dec_and_test(&sk->sk_refcnt))
1579 sk_free(sk);
1580}
1581
58a5a7b9
ACM
1582extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1583 const int nested);
25995ff5 1584
e022f0b4
KK
1585static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1586{
1587 sk->sk_tx_queue_mapping = tx_queue;
1588}
1589
1590static inline void sk_tx_queue_clear(struct sock *sk)
1591{
1592 sk->sk_tx_queue_mapping = -1;
1593}
1594
1595static inline int sk_tx_queue_get(const struct sock *sk)
1596{
b0f77d0e 1597 return sk ? sk->sk_tx_queue_mapping : -1;
e022f0b4
KK
1598}
1599
972692e0
DM
1600static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1601{
e022f0b4 1602 sk_tx_queue_clear(sk);
972692e0
DM
1603 sk->sk_socket = sock;
1604}
1605
aa395145
ED
1606static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1607{
eaefd110
ED
1608 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1609 return &rcu_dereference_raw(sk->sk_wq)->wait;
aa395145 1610}
1da177e4
LT
1611/* Detach socket from process context.
1612 * Announce socket dead, detach it from wait queue and inode.
1613 * Note that parent inode held reference count on this struct sock,
1614 * we do not release it in this function, because protocol
1615 * probably wants some additional cleanups or even continuing
1616 * to work with this socket (TCP).
1617 */
1618static inline void sock_orphan(struct sock *sk)
1619{
1620 write_lock_bh(&sk->sk_callback_lock);
1621 sock_set_flag(sk, SOCK_DEAD);
972692e0 1622 sk_set_socket(sk, NULL);
43815482 1623 sk->sk_wq = NULL;
1da177e4
LT
1624 write_unlock_bh(&sk->sk_callback_lock);
1625}
1626
1627static inline void sock_graft(struct sock *sk, struct socket *parent)
1628{
1629 write_lock_bh(&sk->sk_callback_lock);
eaefd110 1630 sk->sk_wq = parent->wq;
1da177e4 1631 parent->sk = sk;
972692e0 1632 sk_set_socket(sk, parent);
4237c75c 1633 security_sock_graft(sk, parent);
1da177e4
LT
1634 write_unlock_bh(&sk->sk_callback_lock);
1635}
1636
1637extern int sock_i_uid(struct sock *sk);
1638extern unsigned long sock_i_ino(struct sock *sk);
1639
1640static inline struct dst_entry *
1641__sk_dst_get(struct sock *sk)
1642{
d8bf4ca9 1643 return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
f68c224f 1644 lockdep_is_held(&sk->sk_lock.slock));
1da177e4
LT
1645}
1646
1647static inline struct dst_entry *
1648sk_dst_get(struct sock *sk)
1649{
1650 struct dst_entry *dst;
1651
b6c6712a
ED
1652 rcu_read_lock();
1653 dst = rcu_dereference(sk->sk_dst_cache);
1da177e4
LT
1654 if (dst)
1655 dst_hold(dst);
b6c6712a 1656 rcu_read_unlock();
1da177e4
LT
1657 return dst;
1658}
1659
b6c6712a
ED
1660extern void sk_reset_txq(struct sock *sk);
1661
1662static inline void dst_negative_advice(struct sock *sk)
1663{
1664 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1665
1666 if (dst && dst->ops->negative_advice) {
1667 ndst = dst->ops->negative_advice(dst);
1668
1669 if (ndst != dst) {
1670 rcu_assign_pointer(sk->sk_dst_cache, ndst);
1671 sk_reset_txq(sk);
1672 }
1673 }
1674}
1675
1da177e4
LT
1676static inline void
1677__sk_dst_set(struct sock *sk, struct dst_entry *dst)
1678{
1679 struct dst_entry *old_dst;
1680
e022f0b4 1681 sk_tx_queue_clear(sk);
0b53ff2e
ED
1682 /*
1683 * This can be called while sk is owned by the caller only,
1684 * with no state that can be checked in a rcu_dereference_check() cond
1685 */
1686 old_dst = rcu_dereference_raw(sk->sk_dst_cache);
b6c6712a 1687 rcu_assign_pointer(sk->sk_dst_cache, dst);
1da177e4
LT
1688 dst_release(old_dst);
1689}
1690
1691static inline void
1692sk_dst_set(struct sock *sk, struct dst_entry *dst)
1693{
b6c6712a 1694 spin_lock(&sk->sk_dst_lock);
1da177e4 1695 __sk_dst_set(sk, dst);
b6c6712a 1696 spin_unlock(&sk->sk_dst_lock);
1da177e4
LT
1697}
1698
1699static inline void
1700__sk_dst_reset(struct sock *sk)
1701{
b6c6712a 1702 __sk_dst_set(sk, NULL);
1da177e4
LT
1703}
1704
1705static inline void
1706sk_dst_reset(struct sock *sk)
1707{
b6c6712a 1708 spin_lock(&sk->sk_dst_lock);
1da177e4 1709 __sk_dst_reset(sk);
b6c6712a 1710 spin_unlock(&sk->sk_dst_lock);
1da177e4
LT
1711}
1712
f0088a50 1713extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1da177e4 1714
f0088a50 1715extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1da177e4 1716
dc6b9b78 1717static inline bool sk_can_gso(const struct sock *sk)
bcd76111
HX
1718{
1719 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1720}
1721
9958089a 1722extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
6cbb0df7 1723
c8f44aff 1724static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
a465419b
ED
1725{
1726 sk->sk_route_nocaps |= flags;
1727 sk->sk_route_caps &= ~flags;
1728}
1729
c6e1a0d1
TH
1730static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1731 char __user *from, char *to,
912d398d 1732 int copy, int offset)
c6e1a0d1
TH
1733{
1734 if (skb->ip_summed == CHECKSUM_NONE) {
1735 int err = 0;
1736 __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err);
1737 if (err)
1738 return err;
912d398d 1739 skb->csum = csum_block_add(skb->csum, csum, offset);
c6e1a0d1
TH
1740 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1741 if (!access_ok(VERIFY_READ, from, copy) ||
1742 __copy_from_user_nocache(to, from, copy))
1743 return -EFAULT;
1744 } else if (copy_from_user(to, from, copy))
1745 return -EFAULT;
1746
1747 return 0;
1748}
1749
1750static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
1751 char __user *from, int copy)
1752{
912d398d 1753 int err, offset = skb->len;
c6e1a0d1 1754
912d398d
WY
1755 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1756 copy, offset);
c6e1a0d1 1757 if (err)
912d398d 1758 __skb_trim(skb, offset);
c6e1a0d1
TH
1759
1760 return err;
1761}
1762
1763static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from,
1764 struct sk_buff *skb,
1765 struct page *page,
1766 int off, int copy)
1767{
1768 int err;
1769
912d398d
WY
1770 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1771 copy, skb->len);
c6e1a0d1
TH
1772 if (err)
1773 return err;
1774
1775 skb->len += copy;
1776 skb->data_len += copy;
1777 skb->truesize += copy;
1778 sk->sk_wmem_queued += copy;
1779 sk_mem_charge(sk, copy);
1780 return 0;
1781}
1782
1da177e4
LT
1783static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1784 struct sk_buff *skb, struct page *page,
1785 int off, int copy)
1786{
1787 if (skb->ip_summed == CHECKSUM_NONE) {
1788 int err = 0;
5084205f 1789 __wsum csum = csum_and_copy_from_user(from,
1da177e4
LT
1790 page_address(page) + off,
1791 copy, 0, &err);
1792 if (err)
1793 return err;
1794 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1795 } else if (copy_from_user(page_address(page) + off, from, copy))
1796 return -EFAULT;
1797
1798 skb->len += copy;
1799 skb->data_len += copy;
1800 skb->truesize += copy;
1801 sk->sk_wmem_queued += copy;
3ab224be 1802 sk_mem_charge(sk, copy);
1da177e4
LT
1803 return 0;
1804}
1805
c564039f
ED
1806/**
1807 * sk_wmem_alloc_get - returns write allocations
1808 * @sk: socket
1809 *
1810 * Returns sk_wmem_alloc minus initial offset of one
1811 */
1812static inline int sk_wmem_alloc_get(const struct sock *sk)
1813{
1814 return atomic_read(&sk->sk_wmem_alloc) - 1;
1815}
1816
1817/**
1818 * sk_rmem_alloc_get - returns read allocations
1819 * @sk: socket
1820 *
1821 * Returns sk_rmem_alloc
1822 */
1823static inline int sk_rmem_alloc_get(const struct sock *sk)
1824{
1825 return atomic_read(&sk->sk_rmem_alloc);
1826}
1827
1828/**
1829 * sk_has_allocations - check if allocations are outstanding
1830 * @sk: socket
1831 *
1832 * Returns true if socket has write or read allocations
1833 */
dc6b9b78 1834static inline bool sk_has_allocations(const struct sock *sk)
c564039f
ED
1835{
1836 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1837}
1838
a57de0b4 1839/**
43815482 1840 * wq_has_sleeper - check if there are any waiting processes
acfbe96a 1841 * @wq: struct socket_wq
a57de0b4 1842 *
43815482 1843 * Returns true if socket_wq has waiting processes
a57de0b4 1844 *
43815482 1845 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
a57de0b4
JO
1846 * barrier call. They were added due to the race found within the tcp code.
1847 *
1848 * Consider following tcp code paths:
1849 *
1850 * CPU1 CPU2
1851 *
1852 * sys_select receive packet
1853 * ... ...
1854 * __add_wait_queue update tp->rcv_nxt
1855 * ... ...
1856 * tp->rcv_nxt check sock_def_readable
1857 * ... {
43815482
ED
1858 * schedule rcu_read_lock();
1859 * wq = rcu_dereference(sk->sk_wq);
1860 * if (wq && waitqueue_active(&wq->wait))
1861 * wake_up_interruptible(&wq->wait)
a57de0b4
JO
1862 * ...
1863 * }
1864 *
1865 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1866 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1867 * could then endup calling schedule and sleep forever if there are no more
1868 * data on the socket.
ad462769 1869 *
a57de0b4 1870 */
43815482 1871static inline bool wq_has_sleeper(struct socket_wq *wq)
a57de0b4 1872{
dc6b9b78 1873 /* We need to be sure we are in sync with the
a57de0b4
JO
1874 * add_wait_queue modifications to the wait queue.
1875 *
1876 * This memory barrier is paired in the sock_poll_wait.
1877 */
43815482
ED
1878 smp_mb();
1879 return wq && waitqueue_active(&wq->wait);
a57de0b4
JO
1880}
1881
1882/**
1883 * sock_poll_wait - place memory barrier behind the poll_wait call.
1884 * @filp: file
1885 * @wait_address: socket wait queue
1886 * @p: poll_table
1887 *
43815482 1888 * See the comments in the wq_has_sleeper function.
a57de0b4
JO
1889 */
1890static inline void sock_poll_wait(struct file *filp,
1891 wait_queue_head_t *wait_address, poll_table *p)
1892{
626cf236 1893 if (!poll_does_not_wait(p) && wait_address) {
a57de0b4 1894 poll_wait(filp, wait_address, p);
dc6b9b78 1895 /* We need to be sure we are in sync with the
a57de0b4
JO
1896 * socket flags modification.
1897 *
43815482 1898 * This memory barrier is paired in the wq_has_sleeper.
dc6b9b78 1899 */
a57de0b4
JO
1900 smp_mb();
1901 }
1902}
1903
1da177e4 1904/*
dc6b9b78 1905 * Queue a received datagram if it will fit. Stream and sequenced
1da177e4
LT
1906 * protocols can't normally use this as they need to fit buffers in
1907 * and play with them.
1908 *
dc6b9b78 1909 * Inlined as it's very short and called for pretty much every
1da177e4
LT
1910 * packet ever received.
1911 */
1912
1913static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1914{
d55d87fd 1915 skb_orphan(skb);
1da177e4
LT
1916 skb->sk = sk;
1917 skb->destructor = sock_wfree;
2b85a34e
ED
1918 /*
1919 * We used to take a refcount on sk, but following operation
1920 * is enough to guarantee sk_free() wont free this sock until
1921 * all in-flight packets are completed
1922 */
1da177e4
LT
1923 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1924}
1925
1926static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1927{
d55d87fd 1928 skb_orphan(skb);
1da177e4
LT
1929 skb->sk = sk;
1930 skb->destructor = sock_rfree;
1931 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
3ab224be 1932 sk_mem_charge(sk, skb->truesize);
1da177e4
LT
1933}
1934
dc6b9b78 1935extern void sk_reset_timer(struct sock *sk, struct timer_list *timer,
1da177e4
LT
1936 unsigned long expires);
1937
dc6b9b78 1938extern void sk_stop_timer(struct sock *sk, struct timer_list *timer);
1da177e4 1939
f0088a50 1940extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
1da177e4 1941
b1faf566 1942extern int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
1da177e4
LT
1943
1944/*
1945 * Recover an error report and clear atomically
1946 */
dc6b9b78 1947
1da177e4
LT
1948static inline int sock_error(struct sock *sk)
1949{
c1cbe4b7
BL
1950 int err;
1951 if (likely(!sk->sk_err))
1952 return 0;
1953 err = xchg(&sk->sk_err, 0);
1da177e4
LT
1954 return -err;
1955}
1956
1957static inline unsigned long sock_wspace(struct sock *sk)
1958{
1959 int amt = 0;
1960
1961 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1962 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
dc6b9b78 1963 if (amt < 0)
1da177e4
LT
1964 amt = 0;
1965 }
1966 return amt;
1967}
1968
1969static inline void sk_wake_async(struct sock *sk, int how, int band)
1970{
bcdce719 1971 if (sock_flag(sk, SOCK_FASYNC))
1da177e4
LT
1972 sock_wake_async(sk->sk_socket, how, band);
1973}
1974
1975#define SOCK_MIN_SNDBUF 2048
7a91b434
ED
1976/*
1977 * Since sk_rmem_alloc sums skb->truesize, even a small frame might need
1978 * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak
1979 */
1980#define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff))
1da177e4
LT
1981
1982static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1983{
1984 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
8df09ea3 1985 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
1da177e4
LT
1986 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1987 }
1988}
1989
df97c708 1990struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
1da177e4
LT
1991
1992static inline struct page *sk_stream_alloc_page(struct sock *sk)
1993{
1994 struct page *page = NULL;
1995
ef015786
HX
1996 page = alloc_pages(sk->sk_allocation, 0);
1997 if (!page) {
180d8cd9 1998 sk_enter_memory_pressure(sk);
1da177e4
LT
1999 sk_stream_moderate_sndbuf(sk);
2000 }
2001 return page;
2002}
2003
1da177e4
LT
2004/*
2005 * Default write policy as shown to user space via poll/select/SIGIO
2006 */
dc6b9b78 2007static inline bool sock_writeable(const struct sock *sk)
1da177e4 2008{
8df09ea3 2009 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
1da177e4
LT
2010}
2011
dd0fc66f 2012static inline gfp_t gfp_any(void)
1da177e4 2013{
99709372 2014 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1da177e4
LT
2015}
2016
dc6b9b78 2017static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
1da177e4
LT
2018{
2019 return noblock ? 0 : sk->sk_rcvtimeo;
2020}
2021
dc6b9b78 2022static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
1da177e4
LT
2023{
2024 return noblock ? 0 : sk->sk_sndtimeo;
2025}
2026
2027static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2028{
2029 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
2030}
2031
2032/* Alas, with timeout socket operations are not restartable.
2033 * Compare this to poll().
2034 */
2035static inline int sock_intr_errno(long timeo)
2036{
2037 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2038}
2039
92f37fd2
ED
2040extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2041 struct sk_buff *skb);
6e3e939f
JB
2042extern void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2043 struct sk_buff *skb);
92f37fd2 2044
dc6b9b78 2045static inline void
1da177e4
LT
2046sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2047{
b7aa0bf7 2048 ktime_t kt = skb->tstamp;
20d49473 2049 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
a61bbcf2 2050
20d49473
PO
2051 /*
2052 * generate control messages if
2053 * - receive time stamping in software requested (SOCK_RCVTSTAMP
2054 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
2055 * - software time stamp available and wanted
2056 * (SOCK_TIMESTAMPING_SOFTWARE)
2057 * - hardware time stamps available and wanted
2058 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
2059 * SOCK_TIMESTAMPING_RAW_HARDWARE)
2060 */
2061 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
2062 sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
2063 (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
2064 (hwtstamps->hwtstamp.tv64 &&
2065 sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
2066 (hwtstamps->syststamp.tv64 &&
2067 sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
92f37fd2
ED
2068 __sock_recv_timestamp(msg, sk, skb);
2069 else
b7aa0bf7 2070 sk->sk_stamp = kt;
6e3e939f
JB
2071
2072 if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2073 __sock_recv_wifi_status(msg, sk, skb);
1da177e4
LT
2074}
2075
767dd033
ED
2076extern void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2077 struct sk_buff *skb);
2078
2079static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2080 struct sk_buff *skb)
2081{
2082#define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2083 (1UL << SOCK_RCVTSTAMP) | \
2084 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
2085 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
dc6b9b78 2086 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
767dd033
ED
2087 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
2088
2089 if (sk->sk_flags & FLAGS_TS_OR_DROPS)
2090 __sock_recv_ts_and_drops(msg, sk, skb);
2091 else
2092 sk->sk_stamp = skb->tstamp;
2093}
3b885787 2094
20d49473
PO
2095/**
2096 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
20d49473 2097 * @sk: socket sending this packet
2244d07b 2098 * @tx_flags: filled with instructions for time stamping
20d49473
PO
2099 *
2100 * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if
2101 * parameters are invalid.
2102 */
2244d07b 2103extern int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags);
20d49473 2104
1da177e4
LT
2105/**
2106 * sk_eat_skb - Release a skb if it is no longer needed
4dc3b16b
PP
2107 * @sk: socket to eat this skb from
2108 * @skb: socket buffer to eat
f4b8ea78 2109 * @copied_early: flag indicating whether DMA operations copied this data early
1da177e4
LT
2110 *
2111 * This routine must be called with interrupts disabled or with the socket
2112 * locked so that the sk_buff queue operation is ok.
2113*/
624d1164 2114#ifdef CONFIG_NET_DMA
dc6b9b78 2115static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
624d1164
CL
2116{
2117 __skb_unlink(skb, &sk->sk_receive_queue);
2118 if (!copied_early)
2119 __kfree_skb(skb);
2120 else
2121 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
2122}
2123#else
dc6b9b78 2124static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
1da177e4
LT
2125{
2126 __skb_unlink(skb, &sk->sk_receive_queue);
2127 __kfree_skb(skb);
2128}
624d1164 2129#endif
1da177e4 2130
3b1e0a65
YH
2131static inline
2132struct net *sock_net(const struct sock *sk)
2133{
c2d9ba9b 2134 return read_pnet(&sk->sk_net);
3b1e0a65
YH
2135}
2136
2137static inline
f5aa23fd 2138void sock_net_set(struct sock *sk, struct net *net)
3b1e0a65 2139{
c2d9ba9b 2140 write_pnet(&sk->sk_net, net);
3b1e0a65
YH
2141}
2142
edf02087
DL
2143/*
2144 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
25985edc 2145 * They should not hold a reference to a namespace in order to allow
edf02087
DL
2146 * to stop it.
2147 * Sockets after sk_change_net should be released using sk_release_kernel
2148 */
2149static inline void sk_change_net(struct sock *sk, struct net *net)
2150{
3b1e0a65 2151 put_net(sock_net(sk));
65a18ec5 2152 sock_net_set(sk, hold_net(net));
edf02087
DL
2153}
2154
23542618
KK
2155static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2156{
2157 if (unlikely(skb->sk)) {
2158 struct sock *sk = skb->sk;
2159
2160 skb->destructor = NULL;
2161 skb->sk = NULL;
2162 return sk;
2163 }
2164 return NULL;
2165}
2166
20d49473 2167extern void sock_enable_timestamp(struct sock *sk, int flag);
1da177e4 2168extern int sock_get_timestamp(struct sock *, struct timeval __user *);
ae40eb1e 2169extern int sock_get_timestampns(struct sock *, struct timespec __user *);
1da177e4 2170
dc6b9b78
ED
2171/*
2172 * Enable debug/info messages
1da177e4 2173 */
a2a316fd
SH
2174extern int net_msg_warn;
2175#define NETDEBUG(fmt, args...) \
2176 do { if (net_msg_warn) printk(fmt,##args); } while (0)
1da177e4 2177
a2a316fd
SH
2178#define LIMIT_NETDEBUG(fmt, args...) \
2179 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
1da177e4 2180
1da177e4
LT
2181extern __u32 sysctl_wmem_max;
2182extern __u32 sysctl_rmem_max;
2183
20380731
ACM
2184extern void sk_init(void);
2185
6baf1f41
DM
2186extern int sysctl_optmem_max;
2187
20380731
ACM
2188extern __u32 sysctl_wmem_default;
2189extern __u32 sysctl_rmem_default;
20380731 2190
1da177e4 2191#endif /* _SOCK_H */