fq_codel: should use qdisc backlog as threshold
[deliverable/linux.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>
a5b5bb9a 49#include <linux/lockdep.h>
1da177e4
LT
50#include <linux/netdevice.h>
51#include <linux/skbuff.h> /* struct sk_buff */
d7fe0f24 52#include <linux/mm.h>
1da177e4 53#include <linux/security.h>
5a0e3ad6 54#include <linux/slab.h>
c6e1a0d1 55#include <linux/uaccess.h>
180d8cd9 56#include <linux/memcontrol.h>
e1aab161 57#include <linux/res_counter.h>
c5905afb 58#include <linux/static_key.h>
40401530
AV
59#include <linux/aio.h>
60#include <linux/sched.h>
1da177e4
LT
61
62#include <linux/filter.h>
88ab1932 63#include <linux/rculist_nulls.h>
a57de0b4 64#include <linux/poll.h>
1da177e4 65
c31504dc 66#include <linux/atomic.h>
1da177e4
LT
67#include <net/dst.h>
68#include <net/checksum.h>
69
9f048bfb
ED
70struct cgroup;
71struct cgroup_subsys;
c607b2ed 72#ifdef CONFIG_NET
d1a4c0b3 73int mem_cgroup_sockets_init(struct cgroup *cgrp, struct cgroup_subsys *ss);
761b3ef5 74void mem_cgroup_sockets_destroy(struct cgroup *cgrp);
c607b2ed
GC
75#else
76static inline
77int mem_cgroup_sockets_init(struct cgroup *cgrp, struct cgroup_subsys *ss)
78{
79 return 0;
80}
81static inline
761b3ef5 82void mem_cgroup_sockets_destroy(struct cgroup *cgrp)
c607b2ed
GC
83{
84}
85#endif
1da177e4
LT
86/*
87 * This structure really needs to be cleaned up.
88 * Most of it is for TCP, and not used by any of
89 * the other protocols.
90 */
91
92/* Define this to get the SOCK_DBG debugging facility. */
93#define SOCK_DEBUGGING
94#ifdef SOCK_DEBUGGING
95#define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
96 printk(KERN_DEBUG msg); } while (0)
97#else
4cd9029d 98/* Validate arguments and do nothing */
b9075fa9
JP
99static inline __printf(2, 3)
100void SOCK_DEBUG(struct sock *sk, const char *msg, ...)
4cd9029d
SH
101{
102}
1da177e4
LT
103#endif
104
105/* This is the per-socket lock. The spinlock provides a synchronization
106 * between user contexts and software interrupt processing, whereas the
107 * mini-semaphore synchronizes multiple users amongst themselves.
108 */
1da177e4
LT
109typedef struct {
110 spinlock_t slock;
d2e9117c 111 int owned;
1da177e4 112 wait_queue_head_t wq;
a5b5bb9a
IM
113 /*
114 * We express the mutex-alike socket_lock semantics
115 * to the lock validator by explicitly managing
116 * the slock as a lock variant (in addition to
117 * the slock itself):
118 */
119#ifdef CONFIG_DEBUG_LOCK_ALLOC
120 struct lockdep_map dep_map;
121#endif
1da177e4
LT
122} socket_lock_t;
123
1da177e4 124struct sock;
8feaf0c0 125struct proto;
0eeb8ffc 126struct net;
1da177e4
LT
127
128/**
4dc3b16b 129 * struct sock_common - minimal network layer representation of sockets
68835aba
ED
130 * @skc_daddr: Foreign IPv4 addr
131 * @skc_rcv_saddr: Bound local IPv4 addr
4dc6dc71 132 * @skc_hash: hash value used with various protocol lookup tables
d4cada4a 133 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
4dc3b16b
PP
134 * @skc_family: network address family
135 * @skc_state: Connection state
136 * @skc_reuse: %SO_REUSEADDR setting
137 * @skc_bound_dev_if: bound device index if != 0
4dc3b16b 138 * @skc_bind_node: bind hash linkage for various protocol lookup tables
512615b6 139 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
8feaf0c0 140 * @skc_prot: protocol handlers inside a network family
07feaebf 141 * @skc_net: reference to the network namespace of this socket
68835aba
ED
142 * @skc_node: main hash linkage for various protocol lookup tables
143 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
144 * @skc_tx_queue_mapping: tx queue number for this connection
145 * @skc_refcnt: reference count
4dc3b16b
PP
146 *
147 * This is the minimal network layer representation of sockets, the header
8feaf0c0
ACM
148 * for struct sock and struct inet_timewait_sock.
149 */
1da177e4 150struct sock_common {
68835aba
ED
151 /* skc_daddr and skc_rcv_saddr must be grouped :
152 * cf INET_MATCH() and INET_TW_MATCH()
4dc6dc71 153 */
68835aba
ED
154 __be32 skc_daddr;
155 __be32 skc_rcv_saddr;
4dc6dc71 156
d4cada4a
ED
157 union {
158 unsigned int skc_hash;
159 __u16 skc_u16hashes[2];
160 };
4dc6dc71
ED
161 unsigned short skc_family;
162 volatile unsigned char skc_state;
163 unsigned char skc_reuse;
164 int skc_bound_dev_if;
512615b6
ED
165 union {
166 struct hlist_node skc_bind_node;
167 struct hlist_nulls_node skc_portaddr_node;
168 };
8feaf0c0 169 struct proto *skc_prot;
3b1e0a65 170#ifdef CONFIG_NET_NS
07feaebf 171 struct net *skc_net;
3b1e0a65 172#endif
68835aba
ED
173 /*
174 * fields between dontcopy_begin/dontcopy_end
175 * are not copied in sock_copy()
176 */
928c41e7 177 /* private: */
68835aba 178 int skc_dontcopy_begin[0];
928c41e7 179 /* public: */
68835aba
ED
180 union {
181 struct hlist_node skc_node;
182 struct hlist_nulls_node skc_nulls_node;
183 };
184 int skc_tx_queue_mapping;
185 atomic_t skc_refcnt;
928c41e7 186 /* private: */
68835aba 187 int skc_dontcopy_end[0];
928c41e7 188 /* public: */
1da177e4
LT
189};
190
e1aab161 191struct cg_proto;
1da177e4
LT
192/**
193 * struct sock - network layer representation of sockets
8feaf0c0 194 * @__sk_common: shared layout with inet_timewait_sock
4dc3b16b
PP
195 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
196 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
197 * @sk_lock: synchronizer
198 * @sk_rcvbuf: size of receive buffer in bytes
43815482 199 * @sk_wq: sock wait queue and async head
4dc3b16b
PP
200 * @sk_dst_cache: destination cache
201 * @sk_dst_lock: destination cache lock
202 * @sk_policy: flow policy
4dc3b16b
PP
203 * @sk_receive_queue: incoming packets
204 * @sk_wmem_alloc: transmit queue bytes committed
205 * @sk_write_queue: Packet sending queue
97fc2f08 206 * @sk_async_wait_queue: DMA copied packets
4dc3b16b
PP
207 * @sk_omem_alloc: "o" is "option" or "other"
208 * @sk_wmem_queued: persistent queue size
209 * @sk_forward_alloc: space allocated forward
210 * @sk_allocation: allocation mode
211 * @sk_sndbuf: size of send buffer in bytes
33c732c3 212 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
20d49473 213 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
4dc3b16b
PP
214 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
215 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
a465419b 216 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
bcd76111 217 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
82cc1a7a 218 * @sk_gso_max_size: Maximum GSO segment size to build
4dc3b16b 219 * @sk_lingertime: %SO_LINGER l_linger setting
4dc3b16b
PP
220 * @sk_backlog: always used with the per-socket spinlock held
221 * @sk_callback_lock: used with the callbacks in the end of this struct
222 * @sk_error_queue: rarely used
33c732c3
WC
223 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
224 * IPV6_ADDRFORM for instance)
4dc3b16b 225 * @sk_err: last error
33c732c3
WC
226 * @sk_err_soft: errors that don't cause failure but are the cause of a
227 * persistent failure not just 'timed out'
cb61cb9b 228 * @sk_drops: raw/udp drops counter
4dc3b16b
PP
229 * @sk_ack_backlog: current listen backlog
230 * @sk_max_ack_backlog: listen backlog set in listen()
231 * @sk_priority: %SO_PRIORITY setting
1a3bc369 232 * @sk_cgrp_prioidx: socket group's priority map index
4dc3b16b
PP
233 * @sk_type: socket type (%SOCK_STREAM, etc)
234 * @sk_protocol: which protocol this socket belongs in this network family
53c3fa20
RD
235 * @sk_peer_pid: &struct pid for this socket's peer
236 * @sk_peer_cred: %SO_PEERCRED setting
4dc3b16b
PP
237 * @sk_rcvlowat: %SO_RCVLOWAT setting
238 * @sk_rcvtimeo: %SO_RCVTIMEO setting
239 * @sk_sndtimeo: %SO_SNDTIMEO setting
c58dc01b 240 * @sk_rxhash: flow hash received from netif layer
4dc3b16b
PP
241 * @sk_filter: socket filtering instructions
242 * @sk_protinfo: private area, net family specific, when not using slab
243 * @sk_timer: sock cleanup timer
244 * @sk_stamp: time stamp of last packet received
245 * @sk_socket: Identd and reporting IO signals
246 * @sk_user_data: RPC layer private data
247 * @sk_sndmsg_page: cached page for sendmsg
248 * @sk_sndmsg_off: cached offset for sendmsg
d3d4f0a0 249 * @sk_peek_off: current peek_offset value
4dc3b16b 250 * @sk_send_head: front of stuff to transmit
67be2dd1 251 * @sk_security: used by security modules
31729363 252 * @sk_mark: generic packet mark
53c3fa20 253 * @sk_classid: this socket's cgroup classid
e1aab161 254 * @sk_cgrp: this socket's cgroup-specific proto data
4dc3b16b
PP
255 * @sk_write_pending: a write to stream socket waits to start
256 * @sk_state_change: callback to indicate change in the state of the sock
257 * @sk_data_ready: callback to indicate there is data to be processed
258 * @sk_write_space: callback to indicate there is bf sending space available
259 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
260 * @sk_backlog_rcv: callback to process the backlog
261 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
1da177e4
LT
262 */
263struct sock {
264 /*
8feaf0c0 265 * Now struct inet_timewait_sock also uses sock_common, so please just
1da177e4
LT
266 * don't add nothing before this first member (__sk_common) --acme
267 */
268 struct sock_common __sk_common;
4dc6dc71
ED
269#define sk_node __sk_common.skc_node
270#define sk_nulls_node __sk_common.skc_nulls_node
271#define sk_refcnt __sk_common.skc_refcnt
e022f0b4 272#define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
4dc6dc71 273
68835aba
ED
274#define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
275#define sk_dontcopy_end __sk_common.skc_dontcopy_end
4dc6dc71 276#define sk_hash __sk_common.skc_hash
1da177e4
LT
277#define sk_family __sk_common.skc_family
278#define sk_state __sk_common.skc_state
279#define sk_reuse __sk_common.skc_reuse
280#define sk_bound_dev_if __sk_common.skc_bound_dev_if
1da177e4 281#define sk_bind_node __sk_common.skc_bind_node
8feaf0c0 282#define sk_prot __sk_common.skc_prot
07feaebf 283#define sk_net __sk_common.skc_net
1da177e4 284 socket_lock_t sk_lock;
b178bb3d 285 struct sk_buff_head sk_receive_queue;
fa438ccf
ED
286 /*
287 * The backlog queue is special, it is always used with
288 * the per-socket spinlock held and requires low latency
289 * access. Therefore we special case it's implementation.
b178bb3d
ED
290 * Note : rmem_alloc is in this structure to fill a hole
291 * on 64bit arches, not because its logically part of
292 * backlog.
fa438ccf
ED
293 */
294 struct {
b178bb3d
ED
295 atomic_t rmem_alloc;
296 int len;
297 struct sk_buff *head;
298 struct sk_buff *tail;
fa438ccf 299 } sk_backlog;
b178bb3d
ED
300#define sk_rmem_alloc sk_backlog.rmem_alloc
301 int sk_forward_alloc;
302#ifdef CONFIG_RPS
303 __u32 sk_rxhash;
304#endif
305 atomic_t sk_drops;
306 int sk_rcvbuf;
307
308 struct sk_filter __rcu *sk_filter;
eaefd110 309 struct socket_wq __rcu *sk_wq;
b178bb3d
ED
310
311#ifdef CONFIG_NET_DMA
312 struct sk_buff_head sk_async_wait_queue;
313#endif
314
def8b4fa 315#ifdef CONFIG_XFRM
1da177e4 316 struct xfrm_policy *sk_policy[2];
def8b4fa 317#endif
b178bb3d
ED
318 unsigned long sk_flags;
319 struct dst_entry *sk_dst_cache;
b6c6712a 320 spinlock_t sk_dst_lock;
1da177e4
LT
321 atomic_t sk_wmem_alloc;
322 atomic_t sk_omem_alloc;
4e07a91c 323 int sk_sndbuf;
1da177e4 324 struct sk_buff_head sk_write_queue;
b178bb3d
ED
325 kmemcheck_bitfield_begin(flags);
326 unsigned int sk_shutdown : 2,
327 sk_no_check : 2,
328 sk_userlocks : 4,
329 sk_protocol : 8,
330 sk_type : 16;
331 kmemcheck_bitfield_end(flags);
1da177e4 332 int sk_wmem_queued;
7d877f3b 333 gfp_t sk_allocation;
c8f44aff
MM
334 netdev_features_t sk_route_caps;
335 netdev_features_t sk_route_nocaps;
bcd76111 336 int sk_gso_type;
82cc1a7a 337 unsigned int sk_gso_max_size;
9932cf95 338 int sk_rcvlowat;
1da177e4 339 unsigned long sk_lingertime;
1da177e4 340 struct sk_buff_head sk_error_queue;
476e19cf 341 struct proto *sk_prot_creator;
1da177e4
LT
342 rwlock_t sk_callback_lock;
343 int sk_err,
344 sk_err_soft;
345 unsigned short sk_ack_backlog;
346 unsigned short sk_max_ack_backlog;
347 __u32 sk_priority;
5bc1421e
NH
348#ifdef CONFIG_CGROUPS
349 __u32 sk_cgrp_prioidx;
350#endif
109f6e39
EB
351 struct pid *sk_peer_pid;
352 const struct cred *sk_peer_cred;
1da177e4
LT
353 long sk_rcvtimeo;
354 long sk_sndtimeo;
1da177e4
LT
355 void *sk_protinfo;
356 struct timer_list sk_timer;
b7aa0bf7 357 ktime_t sk_stamp;
1da177e4
LT
358 struct socket *sk_socket;
359 void *sk_user_data;
360 struct page *sk_sndmsg_page;
361 struct sk_buff *sk_send_head;
362 __u32 sk_sndmsg_off;
ef64a54f 363 __s32 sk_peek_off;
1da177e4 364 int sk_write_pending;
d5f64238 365#ifdef CONFIG_SECURITY
1da177e4 366 void *sk_security;
d5f64238 367#endif
4a19ec58 368 __u32 sk_mark;
f8451725 369 u32 sk_classid;
e1aab161 370 struct cg_proto *sk_cgrp;
1da177e4
LT
371 void (*sk_state_change)(struct sock *sk);
372 void (*sk_data_ready)(struct sock *sk, int bytes);
373 void (*sk_write_space)(struct sock *sk);
374 void (*sk_error_report)(struct sock *sk);
375 int (*sk_backlog_rcv)(struct sock *sk,
376 struct sk_buff *skb);
377 void (*sk_destruct)(struct sock *sk);
378};
379
4a17fd52
PE
380/*
381 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
382 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
383 * on a socket means that the socket will reuse everybody else's port
384 * without looking at the other's sk_reuse value.
385 */
386
387#define SK_NO_REUSE 0
388#define SK_CAN_REUSE 1
389#define SK_FORCE_REUSE 2
390
ef64a54f
PE
391static inline int sk_peek_offset(struct sock *sk, int flags)
392{
393 if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0))
394 return sk->sk_peek_off;
395 else
396 return 0;
397}
398
399static inline void sk_peek_offset_bwd(struct sock *sk, int val)
400{
401 if (sk->sk_peek_off >= 0) {
402 if (sk->sk_peek_off >= val)
403 sk->sk_peek_off -= val;
404 else
405 sk->sk_peek_off = 0;
406 }
407}
408
409static inline void sk_peek_offset_fwd(struct sock *sk, int val)
410{
411 if (sk->sk_peek_off >= 0)
412 sk->sk_peek_off += val;
413}
414
1da177e4
LT
415/*
416 * Hashed lists helper routines
417 */
c4146644
LZ
418static inline struct sock *sk_entry(const struct hlist_node *node)
419{
420 return hlist_entry(node, struct sock, sk_node);
421}
422
e48c414e 423static inline struct sock *__sk_head(const struct hlist_head *head)
1da177e4
LT
424{
425 return hlist_entry(head->first, struct sock, sk_node);
426}
427
e48c414e 428static inline struct sock *sk_head(const struct hlist_head *head)
1da177e4
LT
429{
430 return hlist_empty(head) ? NULL : __sk_head(head);
431}
432
88ab1932
ED
433static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
434{
435 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
436}
437
438static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
439{
440 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
441}
442
e48c414e 443static inline struct sock *sk_next(const struct sock *sk)
1da177e4
LT
444{
445 return sk->sk_node.next ?
446 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
447}
448
88ab1932
ED
449static inline struct sock *sk_nulls_next(const struct sock *sk)
450{
451 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
452 hlist_nulls_entry(sk->sk_nulls_node.next,
453 struct sock, sk_nulls_node) :
454 NULL;
455}
456
e48c414e 457static inline int sk_unhashed(const struct sock *sk)
1da177e4
LT
458{
459 return hlist_unhashed(&sk->sk_node);
460}
461
e48c414e 462static inline int sk_hashed(const struct sock *sk)
1da177e4 463{
da753bea 464 return !sk_unhashed(sk);
1da177e4
LT
465}
466
467static __inline__ void sk_node_init(struct hlist_node *node)
468{
469 node->pprev = NULL;
470}
471
88ab1932
ED
472static __inline__ void sk_nulls_node_init(struct hlist_nulls_node *node)
473{
474 node->pprev = NULL;
475}
476
1da177e4
LT
477static __inline__ void __sk_del_node(struct sock *sk)
478{
479 __hlist_del(&sk->sk_node);
480}
481
808f5114 482/* NB: equivalent to hlist_del_init_rcu */
1da177e4
LT
483static __inline__ int __sk_del_node_init(struct sock *sk)
484{
485 if (sk_hashed(sk)) {
486 __sk_del_node(sk);
487 sk_node_init(&sk->sk_node);
488 return 1;
489 }
490 return 0;
491}
492
493/* Grab socket reference count. This operation is valid only
494 when sk is ALREADY grabbed f.e. it is found in hash table
495 or a list and the lookup is made under lock preventing hash table
496 modifications.
497 */
498
499static inline void sock_hold(struct sock *sk)
500{
501 atomic_inc(&sk->sk_refcnt);
502}
503
504/* Ungrab socket in the context, which assumes that socket refcnt
505 cannot hit zero, f.e. it is true in context of any socketcall.
506 */
507static inline void __sock_put(struct sock *sk)
508{
509 atomic_dec(&sk->sk_refcnt);
510}
511
512static __inline__ int sk_del_node_init(struct sock *sk)
513{
514 int rc = __sk_del_node_init(sk);
515
516 if (rc) {
517 /* paranoid for a while -acme */
518 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
519 __sock_put(sk);
520 }
521 return rc;
522}
808f5114 523#define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
1da177e4 524
88ab1932 525static __inline__ int __sk_nulls_del_node_init_rcu(struct sock *sk)
271b72c7
ED
526{
527 if (sk_hashed(sk)) {
88ab1932 528 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
271b72c7
ED
529 return 1;
530 }
531 return 0;
532}
533
88ab1932 534static __inline__ int sk_nulls_del_node_init_rcu(struct sock *sk)
271b72c7 535{
88ab1932 536 int rc = __sk_nulls_del_node_init_rcu(sk);
271b72c7
ED
537
538 if (rc) {
539 /* paranoid for a while -acme */
540 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
541 __sock_put(sk);
542 }
543 return rc;
544}
545
1da177e4
LT
546static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
547{
548 hlist_add_head(&sk->sk_node, list);
549}
550
551static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
552{
553 sock_hold(sk);
554 __sk_add_node(sk, list);
555}
556
808f5114 557static __inline__ void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
558{
559 sock_hold(sk);
560 hlist_add_head_rcu(&sk->sk_node, list);
561}
562
88ab1932 563static __inline__ void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
271b72c7 564{
88ab1932 565 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
271b72c7
ED
566}
567
88ab1932 568static __inline__ void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
271b72c7
ED
569{
570 sock_hold(sk);
88ab1932 571 __sk_nulls_add_node_rcu(sk, list);
271b72c7
ED
572}
573
1da177e4
LT
574static __inline__ void __sk_del_bind_node(struct sock *sk)
575{
576 __hlist_del(&sk->sk_bind_node);
577}
578
579static __inline__ void sk_add_bind_node(struct sock *sk,
580 struct hlist_head *list)
581{
582 hlist_add_head(&sk->sk_bind_node, list);
583}
584
585#define sk_for_each(__sk, node, list) \
586 hlist_for_each_entry(__sk, node, list, sk_node)
808f5114 587#define sk_for_each_rcu(__sk, node, list) \
588 hlist_for_each_entry_rcu(__sk, node, list, sk_node)
88ab1932
ED
589#define sk_nulls_for_each(__sk, node, list) \
590 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
591#define sk_nulls_for_each_rcu(__sk, node, list) \
592 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
1da177e4
LT
593#define sk_for_each_from(__sk, node) \
594 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
595 hlist_for_each_entry_from(__sk, node, sk_node)
88ab1932
ED
596#define sk_nulls_for_each_from(__sk, node) \
597 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
598 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
1da177e4
LT
599#define sk_for_each_safe(__sk, node, tmp, list) \
600 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
601#define sk_for_each_bound(__sk, node, list) \
602 hlist_for_each_entry(__sk, node, list, sk_bind_node)
603
604/* Sock flags */
605enum sock_flags {
606 SOCK_DEAD,
607 SOCK_DONE,
608 SOCK_URGINLINE,
609 SOCK_KEEPOPEN,
610 SOCK_LINGER,
611 SOCK_DESTROY,
612 SOCK_BROADCAST,
613 SOCK_TIMESTAMP,
614 SOCK_ZAPPED,
615 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
616 SOCK_DBG, /* %SO_DEBUG setting */
617 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
92f37fd2 618 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
1da177e4
LT
619 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
620 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
20d49473
PO
621 SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */
622 SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
623 SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */
624 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
625 SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */
626 SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
627 SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
bcdce719 628 SOCK_FASYNC, /* fasync() active */
3b885787 629 SOCK_RXQ_OVFL,
1cdebb42 630 SOCK_ZEROCOPY, /* buffers from userspace */
6e3e939f 631 SOCK_WIFI_STATUS, /* push wifi status to userspace */
3bdc0eba
BG
632 SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
633 * Will use last 4 bytes of packet sent from
634 * user-space instead.
635 */
1da177e4
LT
636};
637
53b924b3
RB
638static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
639{
640 nsk->sk_flags = osk->sk_flags;
641}
642
1da177e4
LT
643static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
644{
645 __set_bit(flag, &sk->sk_flags);
646}
647
648static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
649{
650 __clear_bit(flag, &sk->sk_flags);
651}
652
653static inline int sock_flag(struct sock *sk, enum sock_flags flag)
654{
655 return test_bit(flag, &sk->sk_flags);
656}
657
658static inline void sk_acceptq_removed(struct sock *sk)
659{
660 sk->sk_ack_backlog--;
661}
662
663static inline void sk_acceptq_added(struct sock *sk)
664{
665 sk->sk_ack_backlog++;
666}
667
668static inline int sk_acceptq_is_full(struct sock *sk)
669{
64a14651 670 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
1da177e4
LT
671}
672
673/*
674 * Compute minimal free write space needed to queue new packets.
675 */
676static inline int sk_stream_min_wspace(struct sock *sk)
677{
8df09ea3 678 return sk->sk_wmem_queued >> 1;
1da177e4
LT
679}
680
681static inline int sk_stream_wspace(struct sock *sk)
682{
683 return sk->sk_sndbuf - sk->sk_wmem_queued;
684}
685
686extern void sk_stream_write_space(struct sock *sk);
687
688static inline int sk_stream_memory_free(struct sock *sk)
689{
690 return sk->sk_wmem_queued < sk->sk_sndbuf;
691}
692
8eae939f 693/* OOB backlog add */
a3a858ff 694static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
9ee6b535 695{
7fee226a
ED
696 /* dont let skb dst not refcounted, we are going to leave rcu lock */
697 skb_dst_force(skb);
698
699 if (!sk->sk_backlog.tail)
700 sk->sk_backlog.head = skb;
701 else
9ee6b535 702 sk->sk_backlog.tail->next = skb;
7fee226a
ED
703
704 sk->sk_backlog.tail = skb;
9ee6b535
SH
705 skb->next = NULL;
706}
1da177e4 707
c377411f
ED
708/*
709 * Take into account size of receive queue and backlog queue
0fd7bac6
ED
710 * Do not take into account this skb truesize,
711 * to allow even a single big packet to come.
c377411f 712 */
f545a38f
ED
713static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb,
714 unsigned int limit)
c377411f
ED
715{
716 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
717
f545a38f 718 return qsize > limit;
c377411f
ED
719}
720
8eae939f 721/* The per-socket spinlock must be held here. */
f545a38f
ED
722static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
723 unsigned int limit)
8eae939f 724{
f545a38f 725 if (sk_rcvqueues_full(sk, skb, limit))
8eae939f
ZY
726 return -ENOBUFS;
727
a3a858ff 728 __sk_add_backlog(sk, skb);
8eae939f
ZY
729 sk->sk_backlog.len += skb->truesize;
730 return 0;
731}
732
c57943a1
PZ
733static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
734{
735 return sk->sk_backlog_rcv(sk, skb);
736}
737
c58dc01b
DM
738static inline void sock_rps_record_flow(const struct sock *sk)
739{
740#ifdef CONFIG_RPS
741 struct rps_sock_flow_table *sock_flow_table;
742
743 rcu_read_lock();
744 sock_flow_table = rcu_dereference(rps_sock_flow_table);
745 rps_record_sock_flow(sock_flow_table, sk->sk_rxhash);
746 rcu_read_unlock();
747#endif
748}
749
750static inline void sock_rps_reset_flow(const struct sock *sk)
751{
752#ifdef CONFIG_RPS
753 struct rps_sock_flow_table *sock_flow_table;
754
755 rcu_read_lock();
756 sock_flow_table = rcu_dereference(rps_sock_flow_table);
757 rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash);
758 rcu_read_unlock();
759#endif
760}
761
bdeab991
TH
762static inline void sock_rps_save_rxhash(struct sock *sk,
763 const struct sk_buff *skb)
c58dc01b
DM
764{
765#ifdef CONFIG_RPS
bdeab991 766 if (unlikely(sk->sk_rxhash != skb->rxhash)) {
c58dc01b 767 sock_rps_reset_flow(sk);
bdeab991 768 sk->sk_rxhash = skb->rxhash;
c58dc01b
DM
769 }
770#endif
771}
772
bdeab991
TH
773static inline void sock_rps_reset_rxhash(struct sock *sk)
774{
775#ifdef CONFIG_RPS
776 sock_rps_reset_flow(sk);
777 sk->sk_rxhash = 0;
778#endif
779}
780
cfcabdcc
SH
781#define sk_wait_event(__sk, __timeo, __condition) \
782 ({ int __rc; \
783 release_sock(__sk); \
784 __rc = __condition; \
785 if (!__rc) { \
786 *(__timeo) = schedule_timeout(*(__timeo)); \
787 } \
788 lock_sock(__sk); \
789 __rc = __condition; \
790 __rc; \
791 })
1da177e4
LT
792
793extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
794extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
795extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
796extern int sk_stream_error(struct sock *sk, int flags, int err);
797extern void sk_stream_kill_queues(struct sock *sk);
798
799extern int sk_wait_data(struct sock *sk, long *timeo);
800
60236fdd 801struct request_sock_ops;
6d6ee43e 802struct timewait_sock_ops;
ab1e0a13 803struct inet_hashinfo;
fc8717ba 804struct raw_hashinfo;
de477254 805struct module;
2e6599cb 806
1da177e4
LT
807/* Networking protocol blocks we attach to sockets.
808 * socket layer -> transport layer interface
809 * transport -> network interface is defined by struct inet_proto
810 */
811struct proto {
812 void (*close)(struct sock *sk,
813 long timeout);
814 int (*connect)(struct sock *sk,
815 struct sockaddr *uaddr,
816 int addr_len);
817 int (*disconnect)(struct sock *sk, int flags);
818
819 struct sock * (*accept) (struct sock *sk, int flags, int *err);
820
821 int (*ioctl)(struct sock *sk, int cmd,
822 unsigned long arg);
823 int (*init)(struct sock *sk);
7d06b2e0 824 void (*destroy)(struct sock *sk);
1da177e4
LT
825 void (*shutdown)(struct sock *sk, int how);
826 int (*setsockopt)(struct sock *sk, int level,
827 int optname, char __user *optval,
b7058842 828 unsigned int optlen);
1da177e4
LT
829 int (*getsockopt)(struct sock *sk, int level,
830 int optname, char __user *optval,
831 int __user *option);
af01d537 832#ifdef CONFIG_COMPAT
3fdadf7d
DM
833 int (*compat_setsockopt)(struct sock *sk,
834 int level,
835 int optname, char __user *optval,
b7058842 836 unsigned int optlen);
3fdadf7d
DM
837 int (*compat_getsockopt)(struct sock *sk,
838 int level,
839 int optname, char __user *optval,
840 int __user *option);
709b46e8
EB
841 int (*compat_ioctl)(struct sock *sk,
842 unsigned int cmd, unsigned long arg);
af01d537 843#endif
1da177e4
LT
844 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
845 struct msghdr *msg, size_t len);
846 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
847 struct msghdr *msg,
848 size_t len, int noblock, int flags,
849 int *addr_len);
850 int (*sendpage)(struct sock *sk, struct page *page,
851 int offset, size_t size, int flags);
852 int (*bind)(struct sock *sk,
853 struct sockaddr *uaddr, int addr_len);
854
855 int (*backlog_rcv) (struct sock *sk,
856 struct sk_buff *skb);
857
858 /* Keeping track of sk's, looking them up, and port selection methods. */
859 void (*hash)(struct sock *sk);
860 void (*unhash)(struct sock *sk);
719f8358 861 void (*rehash)(struct sock *sk);
1da177e4 862 int (*get_port)(struct sock *sk, unsigned short snum);
fcbdf09d 863 void (*clear_sk)(struct sock *sk, int size);
1da177e4 864
286ab3d4 865 /* Keeping track of sockets in use */
65f76517 866#ifdef CONFIG_PROC_FS
13ff3d6f 867 unsigned int inuse_idx;
65f76517 868#endif
ebb53d75 869
1da177e4 870 /* Memory pressure */
5c52ba17 871 void (*enter_memory_pressure)(struct sock *sk);
8d987e5c 872 atomic_long_t *memory_allocated; /* Current allocated memory. */
1748376b 873 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
1da177e4
LT
874 /*
875 * Pressure flag: try to collapse.
876 * Technical note: it is used by multiple contexts non atomically.
3ab224be 877 * All the __sk_mem_schedule() is of this nature: accounting
1da177e4
LT
878 * is strict, actions are advisory and have some latency.
879 */
880 int *memory_pressure;
8d987e5c 881 long *sysctl_mem;
1da177e4
LT
882 int *sysctl_wmem;
883 int *sysctl_rmem;
884 int max_header;
7ba42910 885 bool no_autobind;
1da177e4 886
271b72c7 887 struct kmem_cache *slab;
1da177e4 888 unsigned int obj_size;
271b72c7 889 int slab_flags;
1da177e4 890
dd24c001 891 struct percpu_counter *orphan_count;
8feaf0c0 892
60236fdd 893 struct request_sock_ops *rsk_prot;
6d6ee43e 894 struct timewait_sock_ops *twsk_prot;
2e6599cb 895
39d8cda7
PE
896 union {
897 struct inet_hashinfo *hashinfo;
645ca708 898 struct udp_table *udp_table;
fc8717ba 899 struct raw_hashinfo *raw_hash;
39d8cda7 900 } h;
ab1e0a13 901
1da177e4
LT
902 struct module *owner;
903
904 char name[32];
905
906 struct list_head node;
e6848976
ACM
907#ifdef SOCK_REFCNT_DEBUG
908 atomic_t socks;
909#endif
e1aab161
GC
910#ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
911 /*
912 * cgroup specific init/deinit functions. Called once for all
913 * protocols that implement it, from cgroups populate function.
914 * This function has to setup any files the protocol want to
915 * appear in the kmem cgroup filesystem.
916 */
917 int (*init_cgroup)(struct cgroup *cgrp,
918 struct cgroup_subsys *ss);
761b3ef5 919 void (*destroy_cgroup)(struct cgroup *cgrp);
e1aab161
GC
920 struct cg_proto *(*proto_cgroup)(struct mem_cgroup *memcg);
921#endif
922};
923
924struct cg_proto {
925 void (*enter_memory_pressure)(struct sock *sk);
926 struct res_counter *memory_allocated; /* Current allocated memory. */
927 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
928 int *memory_pressure;
929 long *sysctl_mem;
930 /*
931 * memcg field is used to find which memcg we belong directly
932 * Each memcg struct can hold more than one cg_proto, so container_of
933 * won't really cut.
934 *
935 * The elegant solution would be having an inverse function to
936 * proto_cgroup in struct proto, but that means polluting the structure
937 * for everybody, instead of just for memcg users.
938 */
939 struct mem_cgroup *memcg;
1da177e4
LT
940};
941
942extern int proto_register(struct proto *prot, int alloc_slab);
943extern void proto_unregister(struct proto *prot);
944
e6848976
ACM
945#ifdef SOCK_REFCNT_DEBUG
946static inline void sk_refcnt_debug_inc(struct sock *sk)
947{
948 atomic_inc(&sk->sk_prot->socks);
949}
950
951static inline void sk_refcnt_debug_dec(struct sock *sk)
952{
953 atomic_dec(&sk->sk_prot->socks);
954 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
955 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
956}
957
e1aab161 958inline void sk_refcnt_debug_release(const struct sock *sk)
e6848976
ACM
959{
960 if (atomic_read(&sk->sk_refcnt) != 1)
961 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
962 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
963}
964#else /* SOCK_REFCNT_DEBUG */
965#define sk_refcnt_debug_inc(sk) do { } while (0)
966#define sk_refcnt_debug_dec(sk) do { } while (0)
967#define sk_refcnt_debug_release(sk) do { } while (0)
968#endif /* SOCK_REFCNT_DEBUG */
969
376be5ff 970#if defined(CONFIG_CGROUP_MEM_RES_CTLR_KMEM) && defined(CONFIG_NET)
c5905afb 971extern struct static_key memcg_socket_limit_enabled;
e1aab161
GC
972static inline struct cg_proto *parent_cg_proto(struct proto *proto,
973 struct cg_proto *cg_proto)
974{
975 return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg));
976}
c5905afb 977#define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
e1aab161
GC
978#else
979#define mem_cgroup_sockets_enabled 0
980static inline struct cg_proto *parent_cg_proto(struct proto *proto,
981 struct cg_proto *cg_proto)
982{
983 return NULL;
984}
985#endif
986
987
180d8cd9
GC
988static inline bool sk_has_memory_pressure(const struct sock *sk)
989{
990 return sk->sk_prot->memory_pressure != NULL;
991}
992
993static inline bool sk_under_memory_pressure(const struct sock *sk)
994{
995 if (!sk->sk_prot->memory_pressure)
996 return false;
e1aab161
GC
997
998 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
999 return !!*sk->sk_cgrp->memory_pressure;
1000
180d8cd9
GC
1001 return !!*sk->sk_prot->memory_pressure;
1002}
1003
1004static inline void sk_leave_memory_pressure(struct sock *sk)
1005{
1006 int *memory_pressure = sk->sk_prot->memory_pressure;
1007
e1aab161
GC
1008 if (!memory_pressure)
1009 return;
1010
1011 if (*memory_pressure)
180d8cd9 1012 *memory_pressure = 0;
e1aab161
GC
1013
1014 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1015 struct cg_proto *cg_proto = sk->sk_cgrp;
1016 struct proto *prot = sk->sk_prot;
1017
1018 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1019 if (*cg_proto->memory_pressure)
1020 *cg_proto->memory_pressure = 0;
1021 }
1022
180d8cd9
GC
1023}
1024
1025static inline void sk_enter_memory_pressure(struct sock *sk)
1026{
e1aab161
GC
1027 if (!sk->sk_prot->enter_memory_pressure)
1028 return;
1029
1030 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1031 struct cg_proto *cg_proto = sk->sk_cgrp;
1032 struct proto *prot = sk->sk_prot;
1033
1034 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1035 cg_proto->enter_memory_pressure(sk);
1036 }
1037
1038 sk->sk_prot->enter_memory_pressure(sk);
180d8cd9
GC
1039}
1040
1041static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1042{
1043 long *prot = sk->sk_prot->sysctl_mem;
e1aab161
GC
1044 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1045 prot = sk->sk_cgrp->sysctl_mem;
180d8cd9
GC
1046 return prot[index];
1047}
1048
e1aab161
GC
1049static inline void memcg_memory_allocated_add(struct cg_proto *prot,
1050 unsigned long amt,
1051 int *parent_status)
1052{
1053 struct res_counter *fail;
1054 int ret;
1055
0e90b31f
GC
1056 ret = res_counter_charge_nofail(prot->memory_allocated,
1057 amt << PAGE_SHIFT, &fail);
e1aab161
GC
1058 if (ret < 0)
1059 *parent_status = OVER_LIMIT;
1060}
1061
1062static inline void memcg_memory_allocated_sub(struct cg_proto *prot,
1063 unsigned long amt)
1064{
1065 res_counter_uncharge(prot->memory_allocated, amt << PAGE_SHIFT);
1066}
1067
1068static inline u64 memcg_memory_allocated_read(struct cg_proto *prot)
1069{
1070 u64 ret;
1071 ret = res_counter_read_u64(prot->memory_allocated, RES_USAGE);
1072 return ret >> PAGE_SHIFT;
1073}
1074
180d8cd9
GC
1075static inline long
1076sk_memory_allocated(const struct sock *sk)
1077{
1078 struct proto *prot = sk->sk_prot;
e1aab161
GC
1079 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1080 return memcg_memory_allocated_read(sk->sk_cgrp);
1081
180d8cd9
GC
1082 return atomic_long_read(prot->memory_allocated);
1083}
1084
1085static inline long
e1aab161 1086sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status)
180d8cd9
GC
1087{
1088 struct proto *prot = sk->sk_prot;
e1aab161
GC
1089
1090 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1091 memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status);
1092 /* update the root cgroup regardless */
1093 atomic_long_add_return(amt, prot->memory_allocated);
1094 return memcg_memory_allocated_read(sk->sk_cgrp);
1095 }
1096
180d8cd9
GC
1097 return atomic_long_add_return(amt, prot->memory_allocated);
1098}
1099
1100static inline void
0e90b31f 1101sk_memory_allocated_sub(struct sock *sk, int amt)
180d8cd9
GC
1102{
1103 struct proto *prot = sk->sk_prot;
e1aab161 1104
0e90b31f 1105 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
e1aab161
GC
1106 memcg_memory_allocated_sub(sk->sk_cgrp, amt);
1107
180d8cd9
GC
1108 atomic_long_sub(amt, prot->memory_allocated);
1109}
1110
1111static inline void sk_sockets_allocated_dec(struct sock *sk)
1112{
1113 struct proto *prot = sk->sk_prot;
e1aab161
GC
1114
1115 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1116 struct cg_proto *cg_proto = sk->sk_cgrp;
1117
1118 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1119 percpu_counter_dec(cg_proto->sockets_allocated);
1120 }
1121
180d8cd9
GC
1122 percpu_counter_dec(prot->sockets_allocated);
1123}
1124
1125static inline void sk_sockets_allocated_inc(struct sock *sk)
1126{
1127 struct proto *prot = sk->sk_prot;
e1aab161
GC
1128
1129 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1130 struct cg_proto *cg_proto = sk->sk_cgrp;
1131
1132 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1133 percpu_counter_inc(cg_proto->sockets_allocated);
1134 }
1135
180d8cd9
GC
1136 percpu_counter_inc(prot->sockets_allocated);
1137}
1138
1139static inline int
1140sk_sockets_allocated_read_positive(struct sock *sk)
1141{
1142 struct proto *prot = sk->sk_prot;
1143
e1aab161 1144 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
518fbf9c 1145 return percpu_counter_read_positive(sk->sk_cgrp->sockets_allocated);
e1aab161 1146
518fbf9c 1147 return percpu_counter_read_positive(prot->sockets_allocated);
180d8cd9
GC
1148}
1149
1150static inline int
1151proto_sockets_allocated_sum_positive(struct proto *prot)
1152{
1153 return percpu_counter_sum_positive(prot->sockets_allocated);
1154}
1155
1156static inline long
1157proto_memory_allocated(struct proto *prot)
1158{
1159 return atomic_long_read(prot->memory_allocated);
1160}
1161
1162static inline bool
1163proto_memory_pressure(struct proto *prot)
1164{
1165 if (!prot->memory_pressure)
1166 return false;
1167 return !!*prot->memory_pressure;
1168}
1169
65f76517
ED
1170
1171#ifdef CONFIG_PROC_FS
1da177e4 1172/* Called with local bh disabled */
c29a0bc4
PE
1173extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1174extern int sock_prot_inuse_get(struct net *net, struct proto *proto);
65f76517 1175#else
c29a0bc4
PE
1176static void inline sock_prot_inuse_add(struct net *net, struct proto *prot,
1177 int inc)
65f76517
ED
1178{
1179}
65f76517
ED
1180#endif
1181
1da177e4 1182
614c6cb4
ACM
1183/* With per-bucket locks this operation is not-atomic, so that
1184 * this version is not worse.
1185 */
1186static inline void __sk_prot_rehash(struct sock *sk)
1187{
1188 sk->sk_prot->unhash(sk);
1189 sk->sk_prot->hash(sk);
1190}
1191
fcbdf09d
OP
1192void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
1193
1da177e4
LT
1194/* About 10 seconds */
1195#define SOCK_DESTROY_TIME (10*HZ)
1196
1197/* Sockets 0-1023 can't be bound to unless you are superuser */
1198#define PROT_SOCK 1024
1199
1200#define SHUTDOWN_MASK 3
1201#define RCV_SHUTDOWN 1
1202#define SEND_SHUTDOWN 2
1203
1204#define SOCK_SNDBUF_LOCK 1
1205#define SOCK_RCVBUF_LOCK 2
1206#define SOCK_BINDADDR_LOCK 4
1207#define SOCK_BINDPORT_LOCK 8
1208
1209/* sock_iocb: used to kick off async processing of socket ios */
1210struct sock_iocb {
1211 struct list_head list;
1212
1213 int flags;
1214 int size;
1215 struct socket *sock;
1216 struct sock *sk;
1217 struct scm_cookie *scm;
1218 struct msghdr *msg, async_msg;
1da177e4
LT
1219 struct kiocb *kiocb;
1220};
1221
1222static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
1223{
1224 return (struct sock_iocb *)iocb->private;
1225}
1226
1227static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
1228{
1229 return si->kiocb;
1230}
1231
1232struct socket_alloc {
1233 struct socket socket;
1234 struct inode vfs_inode;
1235};
1236
1237static inline struct socket *SOCKET_I(struct inode *inode)
1238{
1239 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1240}
1241
1242static inline struct inode *SOCK_INODE(struct socket *socket)
1243{
1244 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1245}
1246
3ab224be
HA
1247/*
1248 * Functions for memory accounting
1249 */
1250extern int __sk_mem_schedule(struct sock *sk, int size, int kind);
1251extern void __sk_mem_reclaim(struct sock *sk);
1da177e4 1252
3ab224be
HA
1253#define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1254#define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1255#define SK_MEM_SEND 0
1256#define SK_MEM_RECV 1
1da177e4 1257
3ab224be 1258static inline int sk_mem_pages(int amt)
1da177e4 1259{
3ab224be 1260 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
1da177e4
LT
1261}
1262
3ab224be 1263static inline int sk_has_account(struct sock *sk)
1da177e4 1264{
3ab224be
HA
1265 /* return true if protocol supports memory accounting */
1266 return !!sk->sk_prot->memory_allocated;
1da177e4
LT
1267}
1268
3ab224be 1269static inline int sk_wmem_schedule(struct sock *sk, int size)
1da177e4 1270{
3ab224be
HA
1271 if (!sk_has_account(sk))
1272 return 1;
1273 return size <= sk->sk_forward_alloc ||
1274 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1da177e4
LT
1275}
1276
3ab224be 1277static inline int sk_rmem_schedule(struct sock *sk, int size)
d80d99d6 1278{
3ab224be
HA
1279 if (!sk_has_account(sk))
1280 return 1;
d80d99d6 1281 return size <= sk->sk_forward_alloc ||
3ab224be
HA
1282 __sk_mem_schedule(sk, size, SK_MEM_RECV);
1283}
1284
1285static inline void sk_mem_reclaim(struct sock *sk)
1286{
1287 if (!sk_has_account(sk))
1288 return;
1289 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1290 __sk_mem_reclaim(sk);
1291}
1292
9993e7d3
DM
1293static inline void sk_mem_reclaim_partial(struct sock *sk)
1294{
1295 if (!sk_has_account(sk))
1296 return;
1297 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1298 __sk_mem_reclaim(sk);
1299}
1300
3ab224be
HA
1301static inline void sk_mem_charge(struct sock *sk, int size)
1302{
1303 if (!sk_has_account(sk))
1304 return;
1305 sk->sk_forward_alloc -= size;
1306}
1307
1308static inline void sk_mem_uncharge(struct sock *sk, int size)
1309{
1310 if (!sk_has_account(sk))
1311 return;
1312 sk->sk_forward_alloc += size;
1313}
1314
1315static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1316{
3ab224be
HA
1317 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1318 sk->sk_wmem_queued -= skb->truesize;
1319 sk_mem_uncharge(sk, skb->truesize);
1320 __kfree_skb(skb);
d80d99d6
HX
1321}
1322
1da177e4
LT
1323/* Used by processes to "lock" a socket state, so that
1324 * interrupts and bottom half handlers won't change it
1325 * from under us. It essentially blocks any incoming
1326 * packets, so that we won't get any new data or any
1327 * packets that change the state of the socket.
1328 *
1329 * While locked, BH processing will add new packets to
1330 * the backlog queue. This queue is processed by the
1331 * owner of the socket lock right before it is released.
1332 *
1333 * Since ~2.3.5 it is also exclusive sleep lock serializing
1334 * accesses from user process context.
1335 */
d2e9117c 1336#define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
1da177e4 1337
ed07536e
PZ
1338/*
1339 * Macro so as to not evaluate some arguments when
1340 * lockdep is not enabled.
1341 *
1342 * Mark both the sk_lock and the sk_lock.slock as a
1343 * per-address-family lock class.
1344 */
1345#define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1346do { \
e8f6fbf6 1347 sk->sk_lock.owned = 0; \
ed07536e
PZ
1348 init_waitqueue_head(&sk->sk_lock.wq); \
1349 spin_lock_init(&(sk)->sk_lock.slock); \
1350 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1351 sizeof((sk)->sk_lock)); \
1352 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1353 (skey), (sname)); \
1354 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1355} while (0)
1356
41380930 1357extern void lock_sock_nested(struct sock *sk, int subclass);
fcc70d5f
PZ
1358
1359static inline void lock_sock(struct sock *sk)
1360{
1361 lock_sock_nested(sk, 0);
1362}
1363
41380930 1364extern void release_sock(struct sock *sk);
1da177e4
LT
1365
1366/* BH context may only use the following locking interface. */
1367#define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
c6366184
IM
1368#define bh_lock_sock_nested(__sk) \
1369 spin_lock_nested(&((__sk)->sk_lock.slock), \
1370 SINGLE_DEPTH_NESTING)
1da177e4
LT
1371#define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1372
8a74ad60
ED
1373extern bool lock_sock_fast(struct sock *sk);
1374/**
1375 * unlock_sock_fast - complement of lock_sock_fast
1376 * @sk: socket
1377 * @slow: slow mode
1378 *
1379 * fast unlock socket for user context.
1380 * If slow mode is on, we call regular release_sock()
1381 */
1382static inline void unlock_sock_fast(struct sock *sk, bool slow)
4b0b72f7 1383{
8a74ad60
ED
1384 if (slow)
1385 release_sock(sk);
1386 else
1387 spin_unlock_bh(&sk->sk_lock.slock);
4b0b72f7
ED
1388}
1389
4b0b72f7 1390
1b8d7ae4 1391extern struct sock *sk_alloc(struct net *net, int family,
dd0fc66f 1392 gfp_t priority,
6257ff21 1393 struct proto *prot);
1da177e4 1394extern void sk_free(struct sock *sk);
edf02087 1395extern void sk_release_kernel(struct sock *sk);
e56c57d0
ED
1396extern struct sock *sk_clone_lock(const struct sock *sk,
1397 const gfp_t priority);
1da177e4
LT
1398
1399extern struct sk_buff *sock_wmalloc(struct sock *sk,
1400 unsigned long size, int force,
dd0fc66f 1401 gfp_t priority);
1da177e4
LT
1402extern struct sk_buff *sock_rmalloc(struct sock *sk,
1403 unsigned long size, int force,
dd0fc66f 1404 gfp_t priority);
1da177e4
LT
1405extern void sock_wfree(struct sk_buff *skb);
1406extern void sock_rfree(struct sk_buff *skb);
1407
1408extern int sock_setsockopt(struct socket *sock, int level,
1409 int op, char __user *optval,
b7058842 1410 unsigned int optlen);
1da177e4
LT
1411
1412extern int sock_getsockopt(struct socket *sock, int level,
1413 int op, char __user *optval,
1414 int __user *optlen);
1415extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
1416 unsigned long size,
1417 int noblock,
1418 int *errcode);
4cc7f68d
HX
1419extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
1420 unsigned long header_len,
1421 unsigned long data_len,
1422 int noblock,
1423 int *errcode);
86a76caf 1424extern void *sock_kmalloc(struct sock *sk, int size,
dd0fc66f 1425 gfp_t priority);
1da177e4
LT
1426extern void sock_kfree_s(struct sock *sk, void *mem, int size);
1427extern void sk_send_sigurg(struct sock *sk);
1428
f8451725
HX
1429#ifdef CONFIG_CGROUPS
1430extern void sock_update_classid(struct sock *sk);
1431#else
1432static inline void sock_update_classid(struct sock *sk)
1433{
1434}
1435#endif
1436
1da177e4
LT
1437/*
1438 * Functions to fill in entries in struct proto_ops when a protocol
1439 * does not implement a particular function.
1440 */
1441extern int sock_no_bind(struct socket *,
1442 struct sockaddr *, int);
1443extern int sock_no_connect(struct socket *,
1444 struct sockaddr *, int, int);
1445extern int sock_no_socketpair(struct socket *,
1446 struct socket *);
1447extern int sock_no_accept(struct socket *,
1448 struct socket *, int);
1449extern int sock_no_getname(struct socket *,
1450 struct sockaddr *, int *, int);
1451extern unsigned int sock_no_poll(struct file *, struct socket *,
1452 struct poll_table_struct *);
1453extern int sock_no_ioctl(struct socket *, unsigned int,
1454 unsigned long);
1455extern int sock_no_listen(struct socket *, int);
1456extern int sock_no_shutdown(struct socket *, int);
1457extern int sock_no_getsockopt(struct socket *, int , int,
1458 char __user *, int __user *);
1459extern int sock_no_setsockopt(struct socket *, int, int,
b7058842 1460 char __user *, unsigned int);
1da177e4
LT
1461extern int sock_no_sendmsg(struct kiocb *, struct socket *,
1462 struct msghdr *, size_t);
1463extern int sock_no_recvmsg(struct kiocb *, struct socket *,
1464 struct msghdr *, size_t, int);
1465extern int sock_no_mmap(struct file *file,
1466 struct socket *sock,
1467 struct vm_area_struct *vma);
1468extern ssize_t sock_no_sendpage(struct socket *sock,
1469 struct page *page,
1470 int offset, size_t size,
1471 int flags);
1472
1473/*
1474 * Functions to fill in entries in struct proto_ops when a protocol
1475 * uses the inet style.
1476 */
1477extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
1478 char __user *optval, int __user *optlen);
1479extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1480 struct msghdr *msg, size_t size, int flags);
1481extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 1482 char __user *optval, unsigned int optlen);
3fdadf7d
DM
1483extern int compat_sock_common_getsockopt(struct socket *sock, int level,
1484 int optname, char __user *optval, int __user *optlen);
1485extern int compat_sock_common_setsockopt(struct socket *sock, int level,
b7058842 1486 int optname, char __user *optval, unsigned int optlen);
1da177e4
LT
1487
1488extern void sk_common_release(struct sock *sk);
1489
1490/*
1491 * Default socket callbacks and setup code
1492 */
1493
1494/* Initialise core socket variables */
1495extern void sock_init_data(struct socket *sock, struct sock *sk);
1496
46bcf14f
ED
1497extern void sk_filter_release_rcu(struct rcu_head *rcu);
1498
dc9b3346 1499/**
1a5778aa 1500 * sk_filter_release - release a socket filter
dc9b3346
PB
1501 * @fp: filter to remove
1502 *
1503 * Remove a filter from a socket and release its resources.
1504 */
1505
309dd5fc
PE
1506static inline void sk_filter_release(struct sk_filter *fp)
1507{
1508 if (atomic_dec_and_test(&fp->refcnt))
80f8f102 1509 call_rcu(&fp->rcu, sk_filter_release_rcu);
309dd5fc
PE
1510}
1511
1512static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
1da177e4
LT
1513{
1514 unsigned int size = sk_filter_len(fp);
1515
1516 atomic_sub(size, &sk->sk_omem_alloc);
309dd5fc 1517 sk_filter_release(fp);
1da177e4
LT
1518}
1519
1520static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
1521{
1522 atomic_inc(&fp->refcnt);
1523 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
1524}
1525
1526/*
1527 * Socket reference counting postulates.
1528 *
1529 * * Each user of socket SHOULD hold a reference count.
1530 * * Each access point to socket (an hash table bucket, reference from a list,
1531 * running timer, skb in flight MUST hold a reference count.
1532 * * When reference count hits 0, it means it will never increase back.
1533 * * When reference count hits 0, it means that no references from
1534 * outside exist to this socket and current process on current CPU
1535 * is last user and may/should destroy this socket.
1536 * * sk_free is called from any context: process, BH, IRQ. When
1537 * it is called, socket has no references from outside -> sk_free
1538 * may release descendant resources allocated by the socket, but
1539 * to the time when it is called, socket is NOT referenced by any
1540 * hash tables, lists etc.
1541 * * Packets, delivered from outside (from network or from another process)
1542 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1543 * when they sit in queue. Otherwise, packets will leak to hole, when
1544 * socket is looked up by one cpu and unhasing is made by another CPU.
1545 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1546 * (leak to backlog). Packet socket does all the processing inside
1547 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1548 * use separate SMP lock, so that they are prone too.
1549 */
1550
1551/* Ungrab socket and destroy it, if it was the last reference. */
1552static inline void sock_put(struct sock *sk)
1553{
1554 if (atomic_dec_and_test(&sk->sk_refcnt))
1555 sk_free(sk);
1556}
1557
58a5a7b9
ACM
1558extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1559 const int nested);
25995ff5 1560
e022f0b4
KK
1561static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1562{
1563 sk->sk_tx_queue_mapping = tx_queue;
1564}
1565
1566static inline void sk_tx_queue_clear(struct sock *sk)
1567{
1568 sk->sk_tx_queue_mapping = -1;
1569}
1570
1571static inline int sk_tx_queue_get(const struct sock *sk)
1572{
b0f77d0e 1573 return sk ? sk->sk_tx_queue_mapping : -1;
e022f0b4
KK
1574}
1575
972692e0
DM
1576static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1577{
e022f0b4 1578 sk_tx_queue_clear(sk);
972692e0
DM
1579 sk->sk_socket = sock;
1580}
1581
aa395145
ED
1582static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1583{
eaefd110
ED
1584 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1585 return &rcu_dereference_raw(sk->sk_wq)->wait;
aa395145 1586}
1da177e4
LT
1587/* Detach socket from process context.
1588 * Announce socket dead, detach it from wait queue and inode.
1589 * Note that parent inode held reference count on this struct sock,
1590 * we do not release it in this function, because protocol
1591 * probably wants some additional cleanups or even continuing
1592 * to work with this socket (TCP).
1593 */
1594static inline void sock_orphan(struct sock *sk)
1595{
1596 write_lock_bh(&sk->sk_callback_lock);
1597 sock_set_flag(sk, SOCK_DEAD);
972692e0 1598 sk_set_socket(sk, NULL);
43815482 1599 sk->sk_wq = NULL;
1da177e4
LT
1600 write_unlock_bh(&sk->sk_callback_lock);
1601}
1602
1603static inline void sock_graft(struct sock *sk, struct socket *parent)
1604{
1605 write_lock_bh(&sk->sk_callback_lock);
eaefd110 1606 sk->sk_wq = parent->wq;
1da177e4 1607 parent->sk = sk;
972692e0 1608 sk_set_socket(sk, parent);
4237c75c 1609 security_sock_graft(sk, parent);
1da177e4
LT
1610 write_unlock_bh(&sk->sk_callback_lock);
1611}
1612
1613extern int sock_i_uid(struct sock *sk);
1614extern unsigned long sock_i_ino(struct sock *sk);
1615
1616static inline struct dst_entry *
1617__sk_dst_get(struct sock *sk)
1618{
d8bf4ca9 1619 return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
f68c224f 1620 lockdep_is_held(&sk->sk_lock.slock));
1da177e4
LT
1621}
1622
1623static inline struct dst_entry *
1624sk_dst_get(struct sock *sk)
1625{
1626 struct dst_entry *dst;
1627
b6c6712a
ED
1628 rcu_read_lock();
1629 dst = rcu_dereference(sk->sk_dst_cache);
1da177e4
LT
1630 if (dst)
1631 dst_hold(dst);
b6c6712a 1632 rcu_read_unlock();
1da177e4
LT
1633 return dst;
1634}
1635
b6c6712a
ED
1636extern void sk_reset_txq(struct sock *sk);
1637
1638static inline void dst_negative_advice(struct sock *sk)
1639{
1640 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1641
1642 if (dst && dst->ops->negative_advice) {
1643 ndst = dst->ops->negative_advice(dst);
1644
1645 if (ndst != dst) {
1646 rcu_assign_pointer(sk->sk_dst_cache, ndst);
1647 sk_reset_txq(sk);
1648 }
1649 }
1650}
1651
1da177e4
LT
1652static inline void
1653__sk_dst_set(struct sock *sk, struct dst_entry *dst)
1654{
1655 struct dst_entry *old_dst;
1656
e022f0b4 1657 sk_tx_queue_clear(sk);
0b53ff2e
ED
1658 /*
1659 * This can be called while sk is owned by the caller only,
1660 * with no state that can be checked in a rcu_dereference_check() cond
1661 */
1662 old_dst = rcu_dereference_raw(sk->sk_dst_cache);
b6c6712a 1663 rcu_assign_pointer(sk->sk_dst_cache, dst);
1da177e4
LT
1664 dst_release(old_dst);
1665}
1666
1667static inline void
1668sk_dst_set(struct sock *sk, struct dst_entry *dst)
1669{
b6c6712a 1670 spin_lock(&sk->sk_dst_lock);
1da177e4 1671 __sk_dst_set(sk, dst);
b6c6712a 1672 spin_unlock(&sk->sk_dst_lock);
1da177e4
LT
1673}
1674
1675static inline void
1676__sk_dst_reset(struct sock *sk)
1677{
b6c6712a 1678 __sk_dst_set(sk, NULL);
1da177e4
LT
1679}
1680
1681static inline void
1682sk_dst_reset(struct sock *sk)
1683{
b6c6712a 1684 spin_lock(&sk->sk_dst_lock);
1da177e4 1685 __sk_dst_reset(sk);
b6c6712a 1686 spin_unlock(&sk->sk_dst_lock);
1da177e4
LT
1687}
1688
f0088a50 1689extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1da177e4 1690
f0088a50 1691extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1da177e4 1692
bcd76111
HX
1693static inline int sk_can_gso(const struct sock *sk)
1694{
1695 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1696}
1697
9958089a 1698extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
6cbb0df7 1699
c8f44aff 1700static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
a465419b
ED
1701{
1702 sk->sk_route_nocaps |= flags;
1703 sk->sk_route_caps &= ~flags;
1704}
1705
c6e1a0d1
TH
1706static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1707 char __user *from, char *to,
912d398d 1708 int copy, int offset)
c6e1a0d1
TH
1709{
1710 if (skb->ip_summed == CHECKSUM_NONE) {
1711 int err = 0;
1712 __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err);
1713 if (err)
1714 return err;
912d398d 1715 skb->csum = csum_block_add(skb->csum, csum, offset);
c6e1a0d1
TH
1716 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1717 if (!access_ok(VERIFY_READ, from, copy) ||
1718 __copy_from_user_nocache(to, from, copy))
1719 return -EFAULT;
1720 } else if (copy_from_user(to, from, copy))
1721 return -EFAULT;
1722
1723 return 0;
1724}
1725
1726static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
1727 char __user *from, int copy)
1728{
912d398d 1729 int err, offset = skb->len;
c6e1a0d1 1730
912d398d
WY
1731 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1732 copy, offset);
c6e1a0d1 1733 if (err)
912d398d 1734 __skb_trim(skb, offset);
c6e1a0d1
TH
1735
1736 return err;
1737}
1738
1739static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from,
1740 struct sk_buff *skb,
1741 struct page *page,
1742 int off, int copy)
1743{
1744 int err;
1745
912d398d
WY
1746 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1747 copy, skb->len);
c6e1a0d1
TH
1748 if (err)
1749 return err;
1750
1751 skb->len += copy;
1752 skb->data_len += copy;
1753 skb->truesize += copy;
1754 sk->sk_wmem_queued += copy;
1755 sk_mem_charge(sk, copy);
1756 return 0;
1757}
1758
1da177e4
LT
1759static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1760 struct sk_buff *skb, struct page *page,
1761 int off, int copy)
1762{
1763 if (skb->ip_summed == CHECKSUM_NONE) {
1764 int err = 0;
5084205f 1765 __wsum csum = csum_and_copy_from_user(from,
1da177e4
LT
1766 page_address(page) + off,
1767 copy, 0, &err);
1768 if (err)
1769 return err;
1770 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1771 } else if (copy_from_user(page_address(page) + off, from, copy))
1772 return -EFAULT;
1773
1774 skb->len += copy;
1775 skb->data_len += copy;
1776 skb->truesize += copy;
1777 sk->sk_wmem_queued += copy;
3ab224be 1778 sk_mem_charge(sk, copy);
1da177e4
LT
1779 return 0;
1780}
1781
c564039f
ED
1782/**
1783 * sk_wmem_alloc_get - returns write allocations
1784 * @sk: socket
1785 *
1786 * Returns sk_wmem_alloc minus initial offset of one
1787 */
1788static inline int sk_wmem_alloc_get(const struct sock *sk)
1789{
1790 return atomic_read(&sk->sk_wmem_alloc) - 1;
1791}
1792
1793/**
1794 * sk_rmem_alloc_get - returns read allocations
1795 * @sk: socket
1796 *
1797 * Returns sk_rmem_alloc
1798 */
1799static inline int sk_rmem_alloc_get(const struct sock *sk)
1800{
1801 return atomic_read(&sk->sk_rmem_alloc);
1802}
1803
1804/**
1805 * sk_has_allocations - check if allocations are outstanding
1806 * @sk: socket
1807 *
1808 * Returns true if socket has write or read allocations
1809 */
1810static inline int sk_has_allocations(const struct sock *sk)
1811{
1812 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1813}
1814
a57de0b4 1815/**
43815482 1816 * wq_has_sleeper - check if there are any waiting processes
acfbe96a 1817 * @wq: struct socket_wq
a57de0b4 1818 *
43815482 1819 * Returns true if socket_wq has waiting processes
a57de0b4 1820 *
43815482 1821 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
a57de0b4
JO
1822 * barrier call. They were added due to the race found within the tcp code.
1823 *
1824 * Consider following tcp code paths:
1825 *
1826 * CPU1 CPU2
1827 *
1828 * sys_select receive packet
1829 * ... ...
1830 * __add_wait_queue update tp->rcv_nxt
1831 * ... ...
1832 * tp->rcv_nxt check sock_def_readable
1833 * ... {
43815482
ED
1834 * schedule rcu_read_lock();
1835 * wq = rcu_dereference(sk->sk_wq);
1836 * if (wq && waitqueue_active(&wq->wait))
1837 * wake_up_interruptible(&wq->wait)
a57de0b4
JO
1838 * ...
1839 * }
1840 *
1841 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1842 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1843 * could then endup calling schedule and sleep forever if there are no more
1844 * data on the socket.
ad462769 1845 *
a57de0b4 1846 */
43815482 1847static inline bool wq_has_sleeper(struct socket_wq *wq)
a57de0b4 1848{
43815482 1849
a57de0b4
JO
1850 /*
1851 * We need to be sure we are in sync with the
1852 * add_wait_queue modifications to the wait queue.
1853 *
1854 * This memory barrier is paired in the sock_poll_wait.
1855 */
43815482
ED
1856 smp_mb();
1857 return wq && waitqueue_active(&wq->wait);
a57de0b4
JO
1858}
1859
1860/**
1861 * sock_poll_wait - place memory barrier behind the poll_wait call.
1862 * @filp: file
1863 * @wait_address: socket wait queue
1864 * @p: poll_table
1865 *
43815482 1866 * See the comments in the wq_has_sleeper function.
a57de0b4
JO
1867 */
1868static inline void sock_poll_wait(struct file *filp,
1869 wait_queue_head_t *wait_address, poll_table *p)
1870{
626cf236 1871 if (!poll_does_not_wait(p) && wait_address) {
a57de0b4
JO
1872 poll_wait(filp, wait_address, p);
1873 /*
1874 * We need to be sure we are in sync with the
1875 * socket flags modification.
1876 *
43815482 1877 * This memory barrier is paired in the wq_has_sleeper.
a57de0b4
JO
1878 */
1879 smp_mb();
1880 }
1881}
1882
1da177e4
LT
1883/*
1884 * Queue a received datagram if it will fit. Stream and sequenced
1885 * protocols can't normally use this as they need to fit buffers in
1886 * and play with them.
1887 *
1888 * Inlined as it's very short and called for pretty much every
1889 * packet ever received.
1890 */
1891
1892static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1893{
d55d87fd 1894 skb_orphan(skb);
1da177e4
LT
1895 skb->sk = sk;
1896 skb->destructor = sock_wfree;
2b85a34e
ED
1897 /*
1898 * We used to take a refcount on sk, but following operation
1899 * is enough to guarantee sk_free() wont free this sock until
1900 * all in-flight packets are completed
1901 */
1da177e4
LT
1902 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1903}
1904
1905static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1906{
d55d87fd 1907 skb_orphan(skb);
1da177e4
LT
1908 skb->sk = sk;
1909 skb->destructor = sock_rfree;
1910 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
3ab224be 1911 sk_mem_charge(sk, skb->truesize);
1da177e4
LT
1912}
1913
1914extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1915 unsigned long expires);
1916
1917extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1918
f0088a50 1919extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
1da177e4 1920
b1faf566 1921extern int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
1da177e4
LT
1922
1923/*
1924 * Recover an error report and clear atomically
1925 */
1926
1927static inline int sock_error(struct sock *sk)
1928{
c1cbe4b7
BL
1929 int err;
1930 if (likely(!sk->sk_err))
1931 return 0;
1932 err = xchg(&sk->sk_err, 0);
1da177e4
LT
1933 return -err;
1934}
1935
1936static inline unsigned long sock_wspace(struct sock *sk)
1937{
1938 int amt = 0;
1939
1940 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1941 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1942 if (amt < 0)
1943 amt = 0;
1944 }
1945 return amt;
1946}
1947
1948static inline void sk_wake_async(struct sock *sk, int how, int band)
1949{
bcdce719 1950 if (sock_flag(sk, SOCK_FASYNC))
1da177e4
LT
1951 sock_wake_async(sk->sk_socket, how, band);
1952}
1953
1954#define SOCK_MIN_SNDBUF 2048
7a91b434
ED
1955/*
1956 * Since sk_rmem_alloc sums skb->truesize, even a small frame might need
1957 * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak
1958 */
1959#define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff))
1da177e4
LT
1960
1961static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1962{
1963 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
8df09ea3 1964 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
1da177e4
LT
1965 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1966 }
1967}
1968
df97c708 1969struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
1da177e4
LT
1970
1971static inline struct page *sk_stream_alloc_page(struct sock *sk)
1972{
1973 struct page *page = NULL;
1974
ef015786
HX
1975 page = alloc_pages(sk->sk_allocation, 0);
1976 if (!page) {
180d8cd9 1977 sk_enter_memory_pressure(sk);
1da177e4
LT
1978 sk_stream_moderate_sndbuf(sk);
1979 }
1980 return page;
1981}
1982
1da177e4
LT
1983/*
1984 * Default write policy as shown to user space via poll/select/SIGIO
1985 */
1986static inline int sock_writeable(const struct sock *sk)
1987{
8df09ea3 1988 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
1da177e4
LT
1989}
1990
dd0fc66f 1991static inline gfp_t gfp_any(void)
1da177e4 1992{
99709372 1993 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1da177e4
LT
1994}
1995
1996static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1997{
1998 return noblock ? 0 : sk->sk_rcvtimeo;
1999}
2000
2001static inline long sock_sndtimeo(const struct sock *sk, int noblock)
2002{
2003 return noblock ? 0 : sk->sk_sndtimeo;
2004}
2005
2006static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2007{
2008 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
2009}
2010
2011/* Alas, with timeout socket operations are not restartable.
2012 * Compare this to poll().
2013 */
2014static inline int sock_intr_errno(long timeo)
2015{
2016 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2017}
2018
92f37fd2
ED
2019extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2020 struct sk_buff *skb);
6e3e939f
JB
2021extern void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2022 struct sk_buff *skb);
92f37fd2 2023
1da177e4
LT
2024static __inline__ void
2025sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2026{
b7aa0bf7 2027 ktime_t kt = skb->tstamp;
20d49473 2028 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
a61bbcf2 2029
20d49473
PO
2030 /*
2031 * generate control messages if
2032 * - receive time stamping in software requested (SOCK_RCVTSTAMP
2033 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
2034 * - software time stamp available and wanted
2035 * (SOCK_TIMESTAMPING_SOFTWARE)
2036 * - hardware time stamps available and wanted
2037 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
2038 * SOCK_TIMESTAMPING_RAW_HARDWARE)
2039 */
2040 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
2041 sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
2042 (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
2043 (hwtstamps->hwtstamp.tv64 &&
2044 sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
2045 (hwtstamps->syststamp.tv64 &&
2046 sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
92f37fd2
ED
2047 __sock_recv_timestamp(msg, sk, skb);
2048 else
b7aa0bf7 2049 sk->sk_stamp = kt;
6e3e939f
JB
2050
2051 if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2052 __sock_recv_wifi_status(msg, sk, skb);
1da177e4
LT
2053}
2054
767dd033
ED
2055extern void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2056 struct sk_buff *skb);
2057
2058static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2059 struct sk_buff *skb)
2060{
2061#define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2062 (1UL << SOCK_RCVTSTAMP) | \
2063 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
2064 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
2065 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
2066 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
2067
2068 if (sk->sk_flags & FLAGS_TS_OR_DROPS)
2069 __sock_recv_ts_and_drops(msg, sk, skb);
2070 else
2071 sk->sk_stamp = skb->tstamp;
2072}
3b885787 2073
20d49473
PO
2074/**
2075 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
20d49473 2076 * @sk: socket sending this packet
2244d07b 2077 * @tx_flags: filled with instructions for time stamping
20d49473
PO
2078 *
2079 * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if
2080 * parameters are invalid.
2081 */
2244d07b 2082extern int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags);
20d49473 2083
1da177e4
LT
2084/**
2085 * sk_eat_skb - Release a skb if it is no longer needed
4dc3b16b
PP
2086 * @sk: socket to eat this skb from
2087 * @skb: socket buffer to eat
f4b8ea78 2088 * @copied_early: flag indicating whether DMA operations copied this data early
1da177e4
LT
2089 *
2090 * This routine must be called with interrupts disabled or with the socket
2091 * locked so that the sk_buff queue operation is ok.
2092*/
624d1164
CL
2093#ifdef CONFIG_NET_DMA
2094static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
2095{
2096 __skb_unlink(skb, &sk->sk_receive_queue);
2097 if (!copied_early)
2098 __kfree_skb(skb);
2099 else
2100 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
2101}
2102#else
2103static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
1da177e4
LT
2104{
2105 __skb_unlink(skb, &sk->sk_receive_queue);
2106 __kfree_skb(skb);
2107}
624d1164 2108#endif
1da177e4 2109
3b1e0a65
YH
2110static inline
2111struct net *sock_net(const struct sock *sk)
2112{
c2d9ba9b 2113 return read_pnet(&sk->sk_net);
3b1e0a65
YH
2114}
2115
2116static inline
f5aa23fd 2117void sock_net_set(struct sock *sk, struct net *net)
3b1e0a65 2118{
c2d9ba9b 2119 write_pnet(&sk->sk_net, net);
3b1e0a65
YH
2120}
2121
edf02087
DL
2122/*
2123 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
25985edc 2124 * They should not hold a reference to a namespace in order to allow
edf02087
DL
2125 * to stop it.
2126 * Sockets after sk_change_net should be released using sk_release_kernel
2127 */
2128static inline void sk_change_net(struct sock *sk, struct net *net)
2129{
3b1e0a65 2130 put_net(sock_net(sk));
65a18ec5 2131 sock_net_set(sk, hold_net(net));
edf02087
DL
2132}
2133
23542618
KK
2134static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2135{
2136 if (unlikely(skb->sk)) {
2137 struct sock *sk = skb->sk;
2138
2139 skb->destructor = NULL;
2140 skb->sk = NULL;
2141 return sk;
2142 }
2143 return NULL;
2144}
2145
20d49473 2146extern void sock_enable_timestamp(struct sock *sk, int flag);
1da177e4 2147extern int sock_get_timestamp(struct sock *, struct timeval __user *);
ae40eb1e 2148extern int sock_get_timestampns(struct sock *, struct timespec __user *);
1da177e4
LT
2149
2150/*
2151 * Enable debug/info messages
2152 */
a2a316fd
SH
2153extern int net_msg_warn;
2154#define NETDEBUG(fmt, args...) \
2155 do { if (net_msg_warn) printk(fmt,##args); } while (0)
1da177e4 2156
a2a316fd
SH
2157#define LIMIT_NETDEBUG(fmt, args...) \
2158 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
1da177e4 2159
1da177e4
LT
2160extern __u32 sysctl_wmem_max;
2161extern __u32 sysctl_rmem_max;
2162
20380731
ACM
2163extern void sk_init(void);
2164
6baf1f41
DM
2165extern int sysctl_optmem_max;
2166
20380731
ACM
2167extern __u32 sysctl_wmem_default;
2168extern __u32 sysctl_rmem_default;
20380731 2169
1da177e4 2170#endif /* _SOCK_H */
This page took 1.203046 seconds and 5 git commands to generate.