Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
[GitHub/exynos8895/android_kernel_samsung_universal8895.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 128
077b393d
ED
129typedef __u32 __bitwise __portpair;
130typedef __u64 __bitwise __addrpair;
131
1da177e4 132/**
4dc3b16b 133 * struct sock_common - minimal network layer representation of sockets
68835aba
ED
134 * @skc_daddr: Foreign IPv4 addr
135 * @skc_rcv_saddr: Bound local IPv4 addr
4dc6dc71 136 * @skc_hash: hash value used with various protocol lookup tables
d4cada4a 137 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
ce43b03e
ED
138 * @skc_dport: placeholder for inet_dport/tw_dport
139 * @skc_num: placeholder for inet_num/tw_num
4dc3b16b
PP
140 * @skc_family: network address family
141 * @skc_state: Connection state
142 * @skc_reuse: %SO_REUSEADDR setting
055dc21a 143 * @skc_reuseport: %SO_REUSEPORT setting
4dc3b16b 144 * @skc_bound_dev_if: bound device index if != 0
4dc3b16b 145 * @skc_bind_node: bind hash linkage for various protocol lookup tables
512615b6 146 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
8feaf0c0 147 * @skc_prot: protocol handlers inside a network family
07feaebf 148 * @skc_net: reference to the network namespace of this socket
68835aba
ED
149 * @skc_node: main hash linkage for various protocol lookup tables
150 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
151 * @skc_tx_queue_mapping: tx queue number for this connection
152 * @skc_refcnt: reference count
4dc3b16b
PP
153 *
154 * This is the minimal network layer representation of sockets, the header
8feaf0c0
ACM
155 * for struct sock and struct inet_timewait_sock.
156 */
1da177e4 157struct sock_common {
ce43b03e
ED
158 /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned
159 * address on 64bit arches : cf INET_MATCH() and INET_TW_MATCH()
4dc6dc71 160 */
ce43b03e 161 union {
077b393d 162 __addrpair skc_addrpair;
ce43b03e
ED
163 struct {
164 __be32 skc_daddr;
165 __be32 skc_rcv_saddr;
166 };
167 };
d4cada4a
ED
168 union {
169 unsigned int skc_hash;
170 __u16 skc_u16hashes[2];
171 };
ce43b03e
ED
172 /* skc_dport && skc_num must be grouped as well */
173 union {
077b393d 174 __portpair skc_portpair;
ce43b03e
ED
175 struct {
176 __be16 skc_dport;
177 __u16 skc_num;
178 };
179 };
180
4dc6dc71
ED
181 unsigned short skc_family;
182 volatile unsigned char skc_state;
055dc21a
TH
183 unsigned char skc_reuse:4;
184 unsigned char skc_reuseport:4;
4dc6dc71 185 int skc_bound_dev_if;
512615b6
ED
186 union {
187 struct hlist_node skc_bind_node;
188 struct hlist_nulls_node skc_portaddr_node;
189 };
8feaf0c0 190 struct proto *skc_prot;
3b1e0a65 191#ifdef CONFIG_NET_NS
07feaebf 192 struct net *skc_net;
3b1e0a65 193#endif
68835aba
ED
194 /*
195 * fields between dontcopy_begin/dontcopy_end
196 * are not copied in sock_copy()
197 */
928c41e7 198 /* private: */
68835aba 199 int skc_dontcopy_begin[0];
928c41e7 200 /* public: */
68835aba
ED
201 union {
202 struct hlist_node skc_node;
203 struct hlist_nulls_node skc_nulls_node;
204 };
205 int skc_tx_queue_mapping;
206 atomic_t skc_refcnt;
928c41e7 207 /* private: */
68835aba 208 int skc_dontcopy_end[0];
928c41e7 209 /* public: */
1da177e4
LT
210};
211
e1aab161 212struct cg_proto;
1da177e4
LT
213/**
214 * struct sock - network layer representation of sockets
8feaf0c0 215 * @__sk_common: shared layout with inet_timewait_sock
4dc3b16b
PP
216 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
217 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
218 * @sk_lock: synchronizer
219 * @sk_rcvbuf: size of receive buffer in bytes
43815482 220 * @sk_wq: sock wait queue and async head
421b3885 221 * @sk_rx_dst: receive input route used by early demux
4dc3b16b
PP
222 * @sk_dst_cache: destination cache
223 * @sk_dst_lock: destination cache lock
224 * @sk_policy: flow policy
4dc3b16b
PP
225 * @sk_receive_queue: incoming packets
226 * @sk_wmem_alloc: transmit queue bytes committed
227 * @sk_write_queue: Packet sending queue
97fc2f08 228 * @sk_async_wait_queue: DMA copied packets
4dc3b16b
PP
229 * @sk_omem_alloc: "o" is "option" or "other"
230 * @sk_wmem_queued: persistent queue size
231 * @sk_forward_alloc: space allocated forward
06021292 232 * @sk_napi_id: id of the last napi context to receive data for sk
dafcc438 233 * @sk_ll_usec: usecs to busypoll when there is no data
4dc3b16b 234 * @sk_allocation: allocation mode
95bd09eb 235 * @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
c3f40d7c 236 * @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE)
4dc3b16b 237 * @sk_sndbuf: size of send buffer in bytes
33c732c3 238 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
20d49473 239 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
48fc7f7e 240 * @sk_no_check: %SO_NO_CHECK setting, whether or not checkup packets
4dc3b16b 241 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
a465419b 242 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
bcd76111 243 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
82cc1a7a 244 * @sk_gso_max_size: Maximum GSO segment size to build
1485348d 245 * @sk_gso_max_segs: Maximum number of GSO segments
4dc3b16b 246 * @sk_lingertime: %SO_LINGER l_linger setting
4dc3b16b
PP
247 * @sk_backlog: always used with the per-socket spinlock held
248 * @sk_callback_lock: used with the callbacks in the end of this struct
249 * @sk_error_queue: rarely used
33c732c3
WC
250 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
251 * IPV6_ADDRFORM for instance)
4dc3b16b 252 * @sk_err: last error
33c732c3
WC
253 * @sk_err_soft: errors that don't cause failure but are the cause of a
254 * persistent failure not just 'timed out'
cb61cb9b 255 * @sk_drops: raw/udp drops counter
4dc3b16b
PP
256 * @sk_ack_backlog: current listen backlog
257 * @sk_max_ack_backlog: listen backlog set in listen()
258 * @sk_priority: %SO_PRIORITY setting
1a3bc369 259 * @sk_cgrp_prioidx: socket group's priority map index
4dc3b16b
PP
260 * @sk_type: socket type (%SOCK_STREAM, etc)
261 * @sk_protocol: which protocol this socket belongs in this network family
53c3fa20
RD
262 * @sk_peer_pid: &struct pid for this socket's peer
263 * @sk_peer_cred: %SO_PEERCRED setting
4dc3b16b
PP
264 * @sk_rcvlowat: %SO_RCVLOWAT setting
265 * @sk_rcvtimeo: %SO_RCVTIMEO setting
266 * @sk_sndtimeo: %SO_SNDTIMEO setting
c58dc01b 267 * @sk_rxhash: flow hash received from netif layer
4dc3b16b
PP
268 * @sk_filter: socket filtering instructions
269 * @sk_protinfo: private area, net family specific, when not using slab
270 * @sk_timer: sock cleanup timer
271 * @sk_stamp: time stamp of last packet received
272 * @sk_socket: Identd and reporting IO signals
273 * @sk_user_data: RPC layer private data
5640f768 274 * @sk_frag: cached page frag
d3d4f0a0 275 * @sk_peek_off: current peek_offset value
4dc3b16b 276 * @sk_send_head: front of stuff to transmit
67be2dd1 277 * @sk_security: used by security modules
31729363 278 * @sk_mark: generic packet mark
53c3fa20 279 * @sk_classid: this socket's cgroup classid
e1aab161 280 * @sk_cgrp: this socket's cgroup-specific proto data
4dc3b16b
PP
281 * @sk_write_pending: a write to stream socket waits to start
282 * @sk_state_change: callback to indicate change in the state of the sock
283 * @sk_data_ready: callback to indicate there is data to be processed
284 * @sk_write_space: callback to indicate there is bf sending space available
285 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
286 * @sk_backlog_rcv: callback to process the backlog
287 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
1da177e4
LT
288 */
289struct sock {
290 /*
8feaf0c0 291 * Now struct inet_timewait_sock also uses sock_common, so please just
1da177e4
LT
292 * don't add nothing before this first member (__sk_common) --acme
293 */
294 struct sock_common __sk_common;
4dc6dc71
ED
295#define sk_node __sk_common.skc_node
296#define sk_nulls_node __sk_common.skc_nulls_node
297#define sk_refcnt __sk_common.skc_refcnt
e022f0b4 298#define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
4dc6dc71 299
68835aba
ED
300#define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
301#define sk_dontcopy_end __sk_common.skc_dontcopy_end
4dc6dc71 302#define sk_hash __sk_common.skc_hash
50805466
ED
303#define sk_portpair __sk_common.skc_portpair
304#define sk_addrpair __sk_common.skc_addrpair
305#define sk_daddr __sk_common.skc_daddr
306#define sk_rcv_saddr __sk_common.skc_rcv_saddr
1da177e4
LT
307#define sk_family __sk_common.skc_family
308#define sk_state __sk_common.skc_state
309#define sk_reuse __sk_common.skc_reuse
055dc21a 310#define sk_reuseport __sk_common.skc_reuseport
1da177e4 311#define sk_bound_dev_if __sk_common.skc_bound_dev_if
1da177e4 312#define sk_bind_node __sk_common.skc_bind_node
8feaf0c0 313#define sk_prot __sk_common.skc_prot
07feaebf 314#define sk_net __sk_common.skc_net
1da177e4 315 socket_lock_t sk_lock;
b178bb3d 316 struct sk_buff_head sk_receive_queue;
fa438ccf
ED
317 /*
318 * The backlog queue is special, it is always used with
319 * the per-socket spinlock held and requires low latency
320 * access. Therefore we special case it's implementation.
b178bb3d
ED
321 * Note : rmem_alloc is in this structure to fill a hole
322 * on 64bit arches, not because its logically part of
323 * backlog.
fa438ccf
ED
324 */
325 struct {
b178bb3d
ED
326 atomic_t rmem_alloc;
327 int len;
328 struct sk_buff *head;
329 struct sk_buff *tail;
fa438ccf 330 } sk_backlog;
b178bb3d
ED
331#define sk_rmem_alloc sk_backlog.rmem_alloc
332 int sk_forward_alloc;
333#ifdef CONFIG_RPS
334 __u32 sk_rxhash;
06021292 335#endif
e0d1095a 336#ifdef CONFIG_NET_RX_BUSY_POLL
06021292 337 unsigned int sk_napi_id;
dafcc438 338 unsigned int sk_ll_usec;
b178bb3d
ED
339#endif
340 atomic_t sk_drops;
341 int sk_rcvbuf;
342
343 struct sk_filter __rcu *sk_filter;
eaefd110 344 struct socket_wq __rcu *sk_wq;
b178bb3d
ED
345
346#ifdef CONFIG_NET_DMA
347 struct sk_buff_head sk_async_wait_queue;
348#endif
349
def8b4fa 350#ifdef CONFIG_XFRM
1da177e4 351 struct xfrm_policy *sk_policy[2];
def8b4fa 352#endif
b178bb3d 353 unsigned long sk_flags;
deaa5854 354 struct dst_entry *sk_rx_dst;
0e36cbb3 355 struct dst_entry __rcu *sk_dst_cache;
b6c6712a 356 spinlock_t sk_dst_lock;
1da177e4
LT
357 atomic_t sk_wmem_alloc;
358 atomic_t sk_omem_alloc;
4e07a91c 359 int sk_sndbuf;
1da177e4 360 struct sk_buff_head sk_write_queue;
b178bb3d
ED
361 kmemcheck_bitfield_begin(flags);
362 unsigned int sk_shutdown : 2,
363 sk_no_check : 2,
364 sk_userlocks : 4,
365 sk_protocol : 8,
366 sk_type : 16;
367 kmemcheck_bitfield_end(flags);
1da177e4 368 int sk_wmem_queued;
7d877f3b 369 gfp_t sk_allocation;
95bd09eb 370 u32 sk_pacing_rate; /* bytes per second */
62748f32 371 u32 sk_max_pacing_rate;
c8f44aff
MM
372 netdev_features_t sk_route_caps;
373 netdev_features_t sk_route_nocaps;
bcd76111 374 int sk_gso_type;
82cc1a7a 375 unsigned int sk_gso_max_size;
1485348d 376 u16 sk_gso_max_segs;
9932cf95 377 int sk_rcvlowat;
1da177e4 378 unsigned long sk_lingertime;
1da177e4 379 struct sk_buff_head sk_error_queue;
476e19cf 380 struct proto *sk_prot_creator;
1da177e4
LT
381 rwlock_t sk_callback_lock;
382 int sk_err,
383 sk_err_soft;
384 unsigned short sk_ack_backlog;
385 unsigned short sk_max_ack_backlog;
386 __u32 sk_priority;
3d0dcfbd 387#if IS_ENABLED(CONFIG_NETPRIO_CGROUP)
5bc1421e
NH
388 __u32 sk_cgrp_prioidx;
389#endif
109f6e39
EB
390 struct pid *sk_peer_pid;
391 const struct cred *sk_peer_cred;
1da177e4
LT
392 long sk_rcvtimeo;
393 long sk_sndtimeo;
1da177e4
LT
394 void *sk_protinfo;
395 struct timer_list sk_timer;
b7aa0bf7 396 ktime_t sk_stamp;
1da177e4
LT
397 struct socket *sk_socket;
398 void *sk_user_data;
5640f768 399 struct page_frag sk_frag;
1da177e4 400 struct sk_buff *sk_send_head;
ef64a54f 401 __s32 sk_peek_off;
1da177e4 402 int sk_write_pending;
d5f64238 403#ifdef CONFIG_SECURITY
1da177e4 404 void *sk_security;
d5f64238 405#endif
4a19ec58 406 __u32 sk_mark;
f8451725 407 u32 sk_classid;
e1aab161 408 struct cg_proto *sk_cgrp;
1da177e4
LT
409 void (*sk_state_change)(struct sock *sk);
410 void (*sk_data_ready)(struct sock *sk, int bytes);
411 void (*sk_write_space)(struct sock *sk);
412 void (*sk_error_report)(struct sock *sk);
dc6b9b78
ED
413 int (*sk_backlog_rcv)(struct sock *sk,
414 struct sk_buff *skb);
1da177e4
LT
415 void (*sk_destruct)(struct sock *sk);
416};
417
559835ea
PS
418#define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
419
420#define rcu_dereference_sk_user_data(sk) rcu_dereference(__sk_user_data((sk)))
421#define rcu_assign_sk_user_data(sk, ptr) rcu_assign_pointer(__sk_user_data((sk)), ptr)
422
4a17fd52
PE
423/*
424 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
425 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
426 * on a socket means that the socket will reuse everybody else's port
427 * without looking at the other's sk_reuse value.
428 */
429
430#define SK_NO_REUSE 0
431#define SK_CAN_REUSE 1
432#define SK_FORCE_REUSE 2
433
ef64a54f
PE
434static inline int sk_peek_offset(struct sock *sk, int flags)
435{
436 if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0))
437 return sk->sk_peek_off;
438 else
439 return 0;
440}
441
442static inline void sk_peek_offset_bwd(struct sock *sk, int val)
443{
444 if (sk->sk_peek_off >= 0) {
445 if (sk->sk_peek_off >= val)
446 sk->sk_peek_off -= val;
447 else
448 sk->sk_peek_off = 0;
449 }
450}
451
452static inline void sk_peek_offset_fwd(struct sock *sk, int val)
453{
454 if (sk->sk_peek_off >= 0)
455 sk->sk_peek_off += val;
456}
457
1da177e4
LT
458/*
459 * Hashed lists helper routines
460 */
c4146644
LZ
461static inline struct sock *sk_entry(const struct hlist_node *node)
462{
463 return hlist_entry(node, struct sock, sk_node);
464}
465
e48c414e 466static inline struct sock *__sk_head(const struct hlist_head *head)
1da177e4
LT
467{
468 return hlist_entry(head->first, struct sock, sk_node);
469}
470
e48c414e 471static inline struct sock *sk_head(const struct hlist_head *head)
1da177e4
LT
472{
473 return hlist_empty(head) ? NULL : __sk_head(head);
474}
475
88ab1932
ED
476static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
477{
478 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
479}
480
481static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
482{
483 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
484}
485
e48c414e 486static inline struct sock *sk_next(const struct sock *sk)
1da177e4
LT
487{
488 return sk->sk_node.next ?
489 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
490}
491
88ab1932
ED
492static inline struct sock *sk_nulls_next(const struct sock *sk)
493{
494 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
495 hlist_nulls_entry(sk->sk_nulls_node.next,
496 struct sock, sk_nulls_node) :
497 NULL;
498}
499
dc6b9b78 500static inline bool sk_unhashed(const struct sock *sk)
1da177e4
LT
501{
502 return hlist_unhashed(&sk->sk_node);
503}
504
dc6b9b78 505static inline bool sk_hashed(const struct sock *sk)
1da177e4 506{
da753bea 507 return !sk_unhashed(sk);
1da177e4
LT
508}
509
dc6b9b78 510static inline void sk_node_init(struct hlist_node *node)
1da177e4
LT
511{
512 node->pprev = NULL;
513}
514
dc6b9b78 515static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
88ab1932
ED
516{
517 node->pprev = NULL;
518}
519
dc6b9b78 520static inline void __sk_del_node(struct sock *sk)
1da177e4
LT
521{
522 __hlist_del(&sk->sk_node);
523}
524
808f5114 525/* NB: equivalent to hlist_del_init_rcu */
dc6b9b78 526static inline bool __sk_del_node_init(struct sock *sk)
1da177e4
LT
527{
528 if (sk_hashed(sk)) {
529 __sk_del_node(sk);
530 sk_node_init(&sk->sk_node);
dc6b9b78 531 return true;
1da177e4 532 }
dc6b9b78 533 return false;
1da177e4
LT
534}
535
536/* Grab socket reference count. This operation is valid only
537 when sk is ALREADY grabbed f.e. it is found in hash table
538 or a list and the lookup is made under lock preventing hash table
539 modifications.
540 */
541
542static inline void sock_hold(struct sock *sk)
543{
544 atomic_inc(&sk->sk_refcnt);
545}
546
547/* Ungrab socket in the context, which assumes that socket refcnt
548 cannot hit zero, f.e. it is true in context of any socketcall.
549 */
550static inline void __sock_put(struct sock *sk)
551{
552 atomic_dec(&sk->sk_refcnt);
553}
554
dc6b9b78 555static inline bool sk_del_node_init(struct sock *sk)
1da177e4 556{
dc6b9b78 557 bool rc = __sk_del_node_init(sk);
1da177e4
LT
558
559 if (rc) {
560 /* paranoid for a while -acme */
561 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
562 __sock_put(sk);
563 }
564 return rc;
565}
808f5114 566#define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
1da177e4 567
dc6b9b78 568static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
271b72c7
ED
569{
570 if (sk_hashed(sk)) {
88ab1932 571 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
dc6b9b78 572 return true;
271b72c7 573 }
dc6b9b78 574 return false;
271b72c7
ED
575}
576
dc6b9b78 577static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
271b72c7 578{
dc6b9b78 579 bool rc = __sk_nulls_del_node_init_rcu(sk);
271b72c7
ED
580
581 if (rc) {
582 /* paranoid for a while -acme */
583 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
584 __sock_put(sk);
585 }
586 return rc;
587}
588
dc6b9b78 589static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
1da177e4
LT
590{
591 hlist_add_head(&sk->sk_node, list);
592}
593
dc6b9b78 594static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
1da177e4
LT
595{
596 sock_hold(sk);
597 __sk_add_node(sk, list);
598}
599
dc6b9b78 600static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
808f5114 601{
602 sock_hold(sk);
603 hlist_add_head_rcu(&sk->sk_node, list);
604}
605
dc6b9b78 606static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
271b72c7 607{
88ab1932 608 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
271b72c7
ED
609}
610
dc6b9b78 611static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
271b72c7
ED
612{
613 sock_hold(sk);
88ab1932 614 __sk_nulls_add_node_rcu(sk, list);
271b72c7
ED
615}
616
dc6b9b78 617static inline void __sk_del_bind_node(struct sock *sk)
1da177e4
LT
618{
619 __hlist_del(&sk->sk_bind_node);
620}
621
dc6b9b78 622static inline void sk_add_bind_node(struct sock *sk,
1da177e4
LT
623 struct hlist_head *list)
624{
625 hlist_add_head(&sk->sk_bind_node, list);
626}
627
b67bfe0d
SL
628#define sk_for_each(__sk, list) \
629 hlist_for_each_entry(__sk, list, sk_node)
630#define sk_for_each_rcu(__sk, list) \
631 hlist_for_each_entry_rcu(__sk, list, sk_node)
88ab1932
ED
632#define sk_nulls_for_each(__sk, node, list) \
633 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
634#define sk_nulls_for_each_rcu(__sk, node, list) \
635 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
b67bfe0d
SL
636#define sk_for_each_from(__sk) \
637 hlist_for_each_entry_from(__sk, sk_node)
88ab1932
ED
638#define sk_nulls_for_each_from(__sk, node) \
639 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
640 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
b67bfe0d
SL
641#define sk_for_each_safe(__sk, tmp, list) \
642 hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
643#define sk_for_each_bound(__sk, list) \
644 hlist_for_each_entry(__sk, list, sk_bind_node)
1da177e4 645
c336d148
EB
646static inline struct user_namespace *sk_user_ns(struct sock *sk)
647{
648 /* Careful only use this in a context where these parameters
649 * can not change and must all be valid, such as recvmsg from
650 * userspace.
651 */
652 return sk->sk_socket->file->f_cred->user_ns;
653}
654
1da177e4
LT
655/* Sock flags */
656enum sock_flags {
657 SOCK_DEAD,
658 SOCK_DONE,
659 SOCK_URGINLINE,
660 SOCK_KEEPOPEN,
661 SOCK_LINGER,
662 SOCK_DESTROY,
663 SOCK_BROADCAST,
664 SOCK_TIMESTAMP,
665 SOCK_ZAPPED,
666 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
667 SOCK_DBG, /* %SO_DEBUG setting */
668 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
92f37fd2 669 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
1da177e4
LT
670 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
671 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
7cb02404 672 SOCK_MEMALLOC, /* VM depends on this socket for swapping */
20d49473
PO
673 SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */
674 SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
675 SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */
676 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
677 SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */
678 SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
679 SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
bcdce719 680 SOCK_FASYNC, /* fasync() active */
3b885787 681 SOCK_RXQ_OVFL,
1cdebb42 682 SOCK_ZEROCOPY, /* buffers from userspace */
6e3e939f 683 SOCK_WIFI_STATUS, /* push wifi status to userspace */
3bdc0eba
BG
684 SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
685 * Will use last 4 bytes of packet sent from
686 * user-space instead.
687 */
d59577b6 688 SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */
7d4c04fc 689 SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */
1da177e4
LT
690};
691
53b924b3
RB
692static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
693{
694 nsk->sk_flags = osk->sk_flags;
695}
696
1da177e4
LT
697static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
698{
699 __set_bit(flag, &sk->sk_flags);
700}
701
702static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
703{
704 __clear_bit(flag, &sk->sk_flags);
705}
706
1b23a5df 707static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
1da177e4
LT
708{
709 return test_bit(flag, &sk->sk_flags);
710}
711
c93bdd0e
MG
712#ifdef CONFIG_NET
713extern struct static_key memalloc_socks;
714static inline int sk_memalloc_socks(void)
715{
716 return static_key_false(&memalloc_socks);
717}
718#else
719
720static inline int sk_memalloc_socks(void)
721{
722 return 0;
723}
724
725#endif
726
99a1dec7
MG
727static inline gfp_t sk_gfp_atomic(struct sock *sk, gfp_t gfp_mask)
728{
7cb02404 729 return GFP_ATOMIC | (sk->sk_allocation & __GFP_MEMALLOC);
99a1dec7
MG
730}
731
1da177e4
LT
732static inline void sk_acceptq_removed(struct sock *sk)
733{
734 sk->sk_ack_backlog--;
735}
736
737static inline void sk_acceptq_added(struct sock *sk)
738{
739 sk->sk_ack_backlog++;
740}
741
dc6b9b78 742static inline bool sk_acceptq_is_full(const struct sock *sk)
1da177e4 743{
64a14651 744 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
1da177e4
LT
745}
746
747/*
748 * Compute minimal free write space needed to queue new packets.
749 */
dc6b9b78 750static inline int sk_stream_min_wspace(const struct sock *sk)
1da177e4 751{
8df09ea3 752 return sk->sk_wmem_queued >> 1;
1da177e4
LT
753}
754
dc6b9b78 755static inline int sk_stream_wspace(const struct sock *sk)
1da177e4
LT
756{
757 return sk->sk_sndbuf - sk->sk_wmem_queued;
758}
759
69336bd2 760void sk_stream_write_space(struct sock *sk);
1da177e4 761
8eae939f 762/* OOB backlog add */
a3a858ff 763static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
9ee6b535 764{
7fee226a
ED
765 /* dont let skb dst not refcounted, we are going to leave rcu lock */
766 skb_dst_force(skb);
767
768 if (!sk->sk_backlog.tail)
769 sk->sk_backlog.head = skb;
770 else
9ee6b535 771 sk->sk_backlog.tail->next = skb;
7fee226a
ED
772
773 sk->sk_backlog.tail = skb;
9ee6b535
SH
774 skb->next = NULL;
775}
1da177e4 776
c377411f
ED
777/*
778 * Take into account size of receive queue and backlog queue
0fd7bac6
ED
779 * Do not take into account this skb truesize,
780 * to allow even a single big packet to come.
c377411f 781 */
f545a38f
ED
782static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb,
783 unsigned int limit)
c377411f
ED
784{
785 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
786
f545a38f 787 return qsize > limit;
c377411f
ED
788}
789
8eae939f 790/* The per-socket spinlock must be held here. */
f545a38f
ED
791static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
792 unsigned int limit)
8eae939f 793{
f545a38f 794 if (sk_rcvqueues_full(sk, skb, limit))
8eae939f
ZY
795 return -ENOBUFS;
796
a3a858ff 797 __sk_add_backlog(sk, skb);
8eae939f
ZY
798 sk->sk_backlog.len += skb->truesize;
799 return 0;
800}
801
69336bd2 802int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
b4b9e355 803
c57943a1
PZ
804static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
805{
b4b9e355
MG
806 if (sk_memalloc_socks() && skb_pfmemalloc(skb))
807 return __sk_backlog_rcv(sk, skb);
808
c57943a1
PZ
809 return sk->sk_backlog_rcv(sk, skb);
810}
811
c58dc01b
DM
812static inline void sock_rps_record_flow(const struct sock *sk)
813{
814#ifdef CONFIG_RPS
815 struct rps_sock_flow_table *sock_flow_table;
816
817 rcu_read_lock();
818 sock_flow_table = rcu_dereference(rps_sock_flow_table);
819 rps_record_sock_flow(sock_flow_table, sk->sk_rxhash);
820 rcu_read_unlock();
821#endif
822}
823
824static inline void sock_rps_reset_flow(const struct sock *sk)
825{
826#ifdef CONFIG_RPS
827 struct rps_sock_flow_table *sock_flow_table;
828
829 rcu_read_lock();
830 sock_flow_table = rcu_dereference(rps_sock_flow_table);
831 rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash);
832 rcu_read_unlock();
833#endif
834}
835
bdeab991
TH
836static inline void sock_rps_save_rxhash(struct sock *sk,
837 const struct sk_buff *skb)
c58dc01b
DM
838{
839#ifdef CONFIG_RPS
bdeab991 840 if (unlikely(sk->sk_rxhash != skb->rxhash)) {
c58dc01b 841 sock_rps_reset_flow(sk);
bdeab991 842 sk->sk_rxhash = skb->rxhash;
c58dc01b
DM
843 }
844#endif
845}
846
bdeab991
TH
847static inline void sock_rps_reset_rxhash(struct sock *sk)
848{
849#ifdef CONFIG_RPS
850 sock_rps_reset_flow(sk);
851 sk->sk_rxhash = 0;
852#endif
853}
854
cfcabdcc
SH
855#define sk_wait_event(__sk, __timeo, __condition) \
856 ({ int __rc; \
857 release_sock(__sk); \
858 __rc = __condition; \
859 if (!__rc) { \
860 *(__timeo) = schedule_timeout(*(__timeo)); \
861 } \
862 lock_sock(__sk); \
863 __rc = __condition; \
864 __rc; \
865 })
1da177e4 866
69336bd2
JP
867int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
868int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
869void sk_stream_wait_close(struct sock *sk, long timeo_p);
870int sk_stream_error(struct sock *sk, int flags, int err);
871void sk_stream_kill_queues(struct sock *sk);
872void sk_set_memalloc(struct sock *sk);
873void sk_clear_memalloc(struct sock *sk);
1da177e4 874
69336bd2 875int sk_wait_data(struct sock *sk, long *timeo);
1da177e4 876
60236fdd 877struct request_sock_ops;
6d6ee43e 878struct timewait_sock_ops;
ab1e0a13 879struct inet_hashinfo;
fc8717ba 880struct raw_hashinfo;
de477254 881struct module;
2e6599cb 882
f77d6021
ED
883/*
884 * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
885 * un-modified. Special care is taken when initializing object to zero.
886 */
887static inline void sk_prot_clear_nulls(struct sock *sk, int size)
888{
889 if (offsetof(struct sock, sk_node.next) != 0)
890 memset(sk, 0, offsetof(struct sock, sk_node.next));
891 memset(&sk->sk_node.pprev, 0,
892 size - offsetof(struct sock, sk_node.pprev));
893}
894
1da177e4
LT
895/* Networking protocol blocks we attach to sockets.
896 * socket layer -> transport layer interface
897 * transport -> network interface is defined by struct inet_proto
898 */
899struct proto {
dc6b9b78 900 void (*close)(struct sock *sk,
1da177e4
LT
901 long timeout);
902 int (*connect)(struct sock *sk,
dc6b9b78 903 struct sockaddr *uaddr,
1da177e4
LT
904 int addr_len);
905 int (*disconnect)(struct sock *sk, int flags);
906
dc6b9b78 907 struct sock * (*accept)(struct sock *sk, int flags, int *err);
1da177e4
LT
908
909 int (*ioctl)(struct sock *sk, int cmd,
910 unsigned long arg);
911 int (*init)(struct sock *sk);
7d06b2e0 912 void (*destroy)(struct sock *sk);
1da177e4 913 void (*shutdown)(struct sock *sk, int how);
dc6b9b78 914 int (*setsockopt)(struct sock *sk, int level,
1da177e4 915 int optname, char __user *optval,
b7058842 916 unsigned int optlen);
dc6b9b78
ED
917 int (*getsockopt)(struct sock *sk, int level,
918 int optname, char __user *optval,
919 int __user *option);
af01d537 920#ifdef CONFIG_COMPAT
3fdadf7d
DM
921 int (*compat_setsockopt)(struct sock *sk,
922 int level,
923 int optname, char __user *optval,
b7058842 924 unsigned int optlen);
3fdadf7d
DM
925 int (*compat_getsockopt)(struct sock *sk,
926 int level,
927 int optname, char __user *optval,
928 int __user *option);
709b46e8
EB
929 int (*compat_ioctl)(struct sock *sk,
930 unsigned int cmd, unsigned long arg);
af01d537 931#endif
1da177e4
LT
932 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
933 struct msghdr *msg, size_t len);
934 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
935 struct msghdr *msg,
dc6b9b78
ED
936 size_t len, int noblock, int flags,
937 int *addr_len);
1da177e4
LT
938 int (*sendpage)(struct sock *sk, struct page *page,
939 int offset, size_t size, int flags);
dc6b9b78 940 int (*bind)(struct sock *sk,
1da177e4
LT
941 struct sockaddr *uaddr, int addr_len);
942
dc6b9b78 943 int (*backlog_rcv) (struct sock *sk,
1da177e4
LT
944 struct sk_buff *skb);
945
46d3ceab 946 void (*release_cb)(struct sock *sk);
563d34d0 947 void (*mtu_reduced)(struct sock *sk);
46d3ceab 948
1da177e4
LT
949 /* Keeping track of sk's, looking them up, and port selection methods. */
950 void (*hash)(struct sock *sk);
951 void (*unhash)(struct sock *sk);
719f8358 952 void (*rehash)(struct sock *sk);
1da177e4 953 int (*get_port)(struct sock *sk, unsigned short snum);
fcbdf09d 954 void (*clear_sk)(struct sock *sk, int size);
1da177e4 955
286ab3d4 956 /* Keeping track of sockets in use */
65f76517 957#ifdef CONFIG_PROC_FS
13ff3d6f 958 unsigned int inuse_idx;
65f76517 959#endif
ebb53d75 960
c9bee3b7 961 bool (*stream_memory_free)(const struct sock *sk);
1da177e4 962 /* Memory pressure */
5c52ba17 963 void (*enter_memory_pressure)(struct sock *sk);
8d987e5c 964 atomic_long_t *memory_allocated; /* Current allocated memory. */
1748376b 965 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
1da177e4
LT
966 /*
967 * Pressure flag: try to collapse.
968 * Technical note: it is used by multiple contexts non atomically.
3ab224be 969 * All the __sk_mem_schedule() is of this nature: accounting
1da177e4
LT
970 * is strict, actions are advisory and have some latency.
971 */
972 int *memory_pressure;
8d987e5c 973 long *sysctl_mem;
1da177e4
LT
974 int *sysctl_wmem;
975 int *sysctl_rmem;
976 int max_header;
7ba42910 977 bool no_autobind;
1da177e4 978
271b72c7 979 struct kmem_cache *slab;
1da177e4 980 unsigned int obj_size;
271b72c7 981 int slab_flags;
1da177e4 982
dd24c001 983 struct percpu_counter *orphan_count;
8feaf0c0 984
60236fdd 985 struct request_sock_ops *rsk_prot;
6d6ee43e 986 struct timewait_sock_ops *twsk_prot;
2e6599cb 987
39d8cda7
PE
988 union {
989 struct inet_hashinfo *hashinfo;
645ca708 990 struct udp_table *udp_table;
fc8717ba 991 struct raw_hashinfo *raw_hash;
39d8cda7 992 } h;
ab1e0a13 993
1da177e4
LT
994 struct module *owner;
995
996 char name[32];
997
998 struct list_head node;
e6848976
ACM
999#ifdef SOCK_REFCNT_DEBUG
1000 atomic_t socks;
1001#endif
c255a458 1002#ifdef CONFIG_MEMCG_KMEM
e1aab161
GC
1003 /*
1004 * cgroup specific init/deinit functions. Called once for all
1005 * protocols that implement it, from cgroups populate function.
1006 * This function has to setup any files the protocol want to
1007 * appear in the kmem cgroup filesystem.
1008 */
1d62e436 1009 int (*init_cgroup)(struct mem_cgroup *memcg,
e1aab161 1010 struct cgroup_subsys *ss);
1d62e436 1011 void (*destroy_cgroup)(struct mem_cgroup *memcg);
e1aab161
GC
1012 struct cg_proto *(*proto_cgroup)(struct mem_cgroup *memcg);
1013#endif
1014};
1015
3f134619
GC
1016/*
1017 * Bits in struct cg_proto.flags
1018 */
1019enum cg_proto_flags {
1020 /* Currently active and new sockets should be assigned to cgroups */
1021 MEMCG_SOCK_ACTIVE,
1022 /* It was ever activated; we must disarm static keys on destruction */
1023 MEMCG_SOCK_ACTIVATED,
1024};
1025
e1aab161
GC
1026struct cg_proto {
1027 void (*enter_memory_pressure)(struct sock *sk);
1028 struct res_counter *memory_allocated; /* Current allocated memory. */
1029 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
1030 int *memory_pressure;
1031 long *sysctl_mem;
3f134619 1032 unsigned long flags;
e1aab161
GC
1033 /*
1034 * memcg field is used to find which memcg we belong directly
1035 * Each memcg struct can hold more than one cg_proto, so container_of
1036 * won't really cut.
1037 *
1038 * The elegant solution would be having an inverse function to
1039 * proto_cgroup in struct proto, but that means polluting the structure
1040 * for everybody, instead of just for memcg users.
1041 */
1042 struct mem_cgroup *memcg;
1da177e4
LT
1043};
1044
69336bd2
JP
1045int proto_register(struct proto *prot, int alloc_slab);
1046void proto_unregister(struct proto *prot);
1da177e4 1047
3f134619
GC
1048static inline bool memcg_proto_active(struct cg_proto *cg_proto)
1049{
1050 return test_bit(MEMCG_SOCK_ACTIVE, &cg_proto->flags);
1051}
1052
1053static inline bool memcg_proto_activated(struct cg_proto *cg_proto)
1054{
1055 return test_bit(MEMCG_SOCK_ACTIVATED, &cg_proto->flags);
1056}
1057
e6848976
ACM
1058#ifdef SOCK_REFCNT_DEBUG
1059static inline void sk_refcnt_debug_inc(struct sock *sk)
1060{
1061 atomic_inc(&sk->sk_prot->socks);
1062}
1063
1064static inline void sk_refcnt_debug_dec(struct sock *sk)
1065{
1066 atomic_dec(&sk->sk_prot->socks);
1067 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
1068 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
1069}
1070
dec34fb0 1071static inline void sk_refcnt_debug_release(const struct sock *sk)
e6848976
ACM
1072{
1073 if (atomic_read(&sk->sk_refcnt) != 1)
1074 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
1075 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
1076}
1077#else /* SOCK_REFCNT_DEBUG */
1078#define sk_refcnt_debug_inc(sk) do { } while (0)
1079#define sk_refcnt_debug_dec(sk) do { } while (0)
1080#define sk_refcnt_debug_release(sk) do { } while (0)
1081#endif /* SOCK_REFCNT_DEBUG */
1082
c255a458 1083#if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_NET)
c5905afb 1084extern struct static_key memcg_socket_limit_enabled;
e1aab161
GC
1085static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1086 struct cg_proto *cg_proto)
1087{
1088 return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg));
1089}
c5905afb 1090#define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
e1aab161
GC
1091#else
1092#define mem_cgroup_sockets_enabled 0
1093static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1094 struct cg_proto *cg_proto)
1095{
1096 return NULL;
1097}
1098#endif
1099
c9bee3b7
ED
1100static inline bool sk_stream_memory_free(const struct sock *sk)
1101{
1102 if (sk->sk_wmem_queued >= sk->sk_sndbuf)
1103 return false;
1104
1105 return sk->sk_prot->stream_memory_free ?
1106 sk->sk_prot->stream_memory_free(sk) : true;
1107}
1108
64dc6130
ED
1109static inline bool sk_stream_is_writeable(const struct sock *sk)
1110{
c9bee3b7
ED
1111 return sk_stream_wspace(sk) >= sk_stream_min_wspace(sk) &&
1112 sk_stream_memory_free(sk);
64dc6130 1113}
e1aab161 1114
c9bee3b7 1115
180d8cd9
GC
1116static inline bool sk_has_memory_pressure(const struct sock *sk)
1117{
1118 return sk->sk_prot->memory_pressure != NULL;
1119}
1120
1121static inline bool sk_under_memory_pressure(const struct sock *sk)
1122{
1123 if (!sk->sk_prot->memory_pressure)
1124 return false;
e1aab161
GC
1125
1126 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1127 return !!*sk->sk_cgrp->memory_pressure;
1128
180d8cd9
GC
1129 return !!*sk->sk_prot->memory_pressure;
1130}
1131
1132static inline void sk_leave_memory_pressure(struct sock *sk)
1133{
1134 int *memory_pressure = sk->sk_prot->memory_pressure;
1135
e1aab161
GC
1136 if (!memory_pressure)
1137 return;
1138
1139 if (*memory_pressure)
180d8cd9 1140 *memory_pressure = 0;
e1aab161
GC
1141
1142 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1143 struct cg_proto *cg_proto = sk->sk_cgrp;
1144 struct proto *prot = sk->sk_prot;
1145
1146 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1147 if (*cg_proto->memory_pressure)
1148 *cg_proto->memory_pressure = 0;
1149 }
1150
180d8cd9
GC
1151}
1152
1153static inline void sk_enter_memory_pressure(struct sock *sk)
1154{
e1aab161
GC
1155 if (!sk->sk_prot->enter_memory_pressure)
1156 return;
1157
1158 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1159 struct cg_proto *cg_proto = sk->sk_cgrp;
1160 struct proto *prot = sk->sk_prot;
1161
1162 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1163 cg_proto->enter_memory_pressure(sk);
1164 }
1165
1166 sk->sk_prot->enter_memory_pressure(sk);
180d8cd9
GC
1167}
1168
1169static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1170{
1171 long *prot = sk->sk_prot->sysctl_mem;
e1aab161
GC
1172 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1173 prot = sk->sk_cgrp->sysctl_mem;
180d8cd9
GC
1174 return prot[index];
1175}
1176
e1aab161
GC
1177static inline void memcg_memory_allocated_add(struct cg_proto *prot,
1178 unsigned long amt,
1179 int *parent_status)
1180{
1181 struct res_counter *fail;
1182 int ret;
1183
0e90b31f
GC
1184 ret = res_counter_charge_nofail(prot->memory_allocated,
1185 amt << PAGE_SHIFT, &fail);
e1aab161
GC
1186 if (ret < 0)
1187 *parent_status = OVER_LIMIT;
1188}
1189
1190static inline void memcg_memory_allocated_sub(struct cg_proto *prot,
1191 unsigned long amt)
1192{
1193 res_counter_uncharge(prot->memory_allocated, amt << PAGE_SHIFT);
1194}
1195
1196static inline u64 memcg_memory_allocated_read(struct cg_proto *prot)
1197{
1198 u64 ret;
1199 ret = res_counter_read_u64(prot->memory_allocated, RES_USAGE);
1200 return ret >> PAGE_SHIFT;
1201}
1202
180d8cd9
GC
1203static inline long
1204sk_memory_allocated(const struct sock *sk)
1205{
1206 struct proto *prot = sk->sk_prot;
e1aab161
GC
1207 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1208 return memcg_memory_allocated_read(sk->sk_cgrp);
1209
180d8cd9
GC
1210 return atomic_long_read(prot->memory_allocated);
1211}
1212
1213static inline long
e1aab161 1214sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status)
180d8cd9
GC
1215{
1216 struct proto *prot = sk->sk_prot;
e1aab161
GC
1217
1218 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1219 memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status);
1220 /* update the root cgroup regardless */
1221 atomic_long_add_return(amt, prot->memory_allocated);
1222 return memcg_memory_allocated_read(sk->sk_cgrp);
1223 }
1224
180d8cd9
GC
1225 return atomic_long_add_return(amt, prot->memory_allocated);
1226}
1227
1228static inline void
0e90b31f 1229sk_memory_allocated_sub(struct sock *sk, int amt)
180d8cd9
GC
1230{
1231 struct proto *prot = sk->sk_prot;
e1aab161 1232
0e90b31f 1233 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
e1aab161
GC
1234 memcg_memory_allocated_sub(sk->sk_cgrp, amt);
1235
180d8cd9
GC
1236 atomic_long_sub(amt, prot->memory_allocated);
1237}
1238
1239static inline void sk_sockets_allocated_dec(struct sock *sk)
1240{
1241 struct proto *prot = sk->sk_prot;
e1aab161
GC
1242
1243 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1244 struct cg_proto *cg_proto = sk->sk_cgrp;
1245
1246 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1247 percpu_counter_dec(cg_proto->sockets_allocated);
1248 }
1249
180d8cd9
GC
1250 percpu_counter_dec(prot->sockets_allocated);
1251}
1252
1253static inline void sk_sockets_allocated_inc(struct sock *sk)
1254{
1255 struct proto *prot = sk->sk_prot;
e1aab161
GC
1256
1257 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1258 struct cg_proto *cg_proto = sk->sk_cgrp;
1259
1260 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1261 percpu_counter_inc(cg_proto->sockets_allocated);
1262 }
1263
180d8cd9
GC
1264 percpu_counter_inc(prot->sockets_allocated);
1265}
1266
1267static inline int
1268sk_sockets_allocated_read_positive(struct sock *sk)
1269{
1270 struct proto *prot = sk->sk_prot;
1271
e1aab161 1272 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
518fbf9c 1273 return percpu_counter_read_positive(sk->sk_cgrp->sockets_allocated);
e1aab161 1274
518fbf9c 1275 return percpu_counter_read_positive(prot->sockets_allocated);
180d8cd9
GC
1276}
1277
1278static inline int
1279proto_sockets_allocated_sum_positive(struct proto *prot)
1280{
1281 return percpu_counter_sum_positive(prot->sockets_allocated);
1282}
1283
1284static inline long
1285proto_memory_allocated(struct proto *prot)
1286{
1287 return atomic_long_read(prot->memory_allocated);
1288}
1289
1290static inline bool
1291proto_memory_pressure(struct proto *prot)
1292{
1293 if (!prot->memory_pressure)
1294 return false;
1295 return !!*prot->memory_pressure;
1296}
1297
65f76517
ED
1298
1299#ifdef CONFIG_PROC_FS
1da177e4 1300/* Called with local bh disabled */
69336bd2
JP
1301void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1302int sock_prot_inuse_get(struct net *net, struct proto *proto);
65f76517 1303#else
dc6b9b78 1304static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
c29a0bc4 1305 int inc)
65f76517
ED
1306{
1307}
65f76517
ED
1308#endif
1309
1da177e4 1310
614c6cb4
ACM
1311/* With per-bucket locks this operation is not-atomic, so that
1312 * this version is not worse.
1313 */
1314static inline void __sk_prot_rehash(struct sock *sk)
1315{
1316 sk->sk_prot->unhash(sk);
1317 sk->sk_prot->hash(sk);
1318}
1319
fcbdf09d
OP
1320void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
1321
1da177e4
LT
1322/* About 10 seconds */
1323#define SOCK_DESTROY_TIME (10*HZ)
1324
1325/* Sockets 0-1023 can't be bound to unless you are superuser */
1326#define PROT_SOCK 1024
1327
1328#define SHUTDOWN_MASK 3
1329#define RCV_SHUTDOWN 1
1330#define SEND_SHUTDOWN 2
1331
1332#define SOCK_SNDBUF_LOCK 1
1333#define SOCK_RCVBUF_LOCK 2
1334#define SOCK_BINDADDR_LOCK 4
1335#define SOCK_BINDPORT_LOCK 8
1336
1337/* sock_iocb: used to kick off async processing of socket ios */
1338struct sock_iocb {
1339 struct list_head list;
1340
1341 int flags;
1342 int size;
1343 struct socket *sock;
1344 struct sock *sk;
1345 struct scm_cookie *scm;
1346 struct msghdr *msg, async_msg;
1da177e4
LT
1347 struct kiocb *kiocb;
1348};
1349
1350static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
1351{
1352 return (struct sock_iocb *)iocb->private;
1353}
1354
1355static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
1356{
1357 return si->kiocb;
1358}
1359
1360struct socket_alloc {
1361 struct socket socket;
1362 struct inode vfs_inode;
1363};
1364
1365static inline struct socket *SOCKET_I(struct inode *inode)
1366{
1367 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1368}
1369
1370static inline struct inode *SOCK_INODE(struct socket *socket)
1371{
1372 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1373}
1374
3ab224be
HA
1375/*
1376 * Functions for memory accounting
1377 */
69336bd2
JP
1378int __sk_mem_schedule(struct sock *sk, int size, int kind);
1379void __sk_mem_reclaim(struct sock *sk);
1da177e4 1380
3ab224be
HA
1381#define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1382#define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1383#define SK_MEM_SEND 0
1384#define SK_MEM_RECV 1
1da177e4 1385
3ab224be 1386static inline int sk_mem_pages(int amt)
1da177e4 1387{
3ab224be 1388 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
1da177e4
LT
1389}
1390
dc6b9b78 1391static inline bool sk_has_account(struct sock *sk)
1da177e4 1392{
3ab224be
HA
1393 /* return true if protocol supports memory accounting */
1394 return !!sk->sk_prot->memory_allocated;
1da177e4
LT
1395}
1396
dc6b9b78 1397static inline bool sk_wmem_schedule(struct sock *sk, int size)
1da177e4 1398{
3ab224be 1399 if (!sk_has_account(sk))
dc6b9b78 1400 return true;
3ab224be
HA
1401 return size <= sk->sk_forward_alloc ||
1402 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1da177e4
LT
1403}
1404
c76562b6 1405static inline bool
35c448a8 1406sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size)
d80d99d6 1407{
3ab224be 1408 if (!sk_has_account(sk))
dc6b9b78 1409 return true;
c76562b6
MG
1410 return size<= sk->sk_forward_alloc ||
1411 __sk_mem_schedule(sk, size, SK_MEM_RECV) ||
1412 skb_pfmemalloc(skb);
3ab224be
HA
1413}
1414
1415static inline void sk_mem_reclaim(struct sock *sk)
1416{
1417 if (!sk_has_account(sk))
1418 return;
1419 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1420 __sk_mem_reclaim(sk);
1421}
1422
9993e7d3
DM
1423static inline void sk_mem_reclaim_partial(struct sock *sk)
1424{
1425 if (!sk_has_account(sk))
1426 return;
1427 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1428 __sk_mem_reclaim(sk);
1429}
1430
3ab224be
HA
1431static inline void sk_mem_charge(struct sock *sk, int size)
1432{
1433 if (!sk_has_account(sk))
1434 return;
1435 sk->sk_forward_alloc -= size;
1436}
1437
1438static inline void sk_mem_uncharge(struct sock *sk, int size)
1439{
1440 if (!sk_has_account(sk))
1441 return;
1442 sk->sk_forward_alloc += size;
1443}
1444
1445static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1446{
3ab224be
HA
1447 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1448 sk->sk_wmem_queued -= skb->truesize;
1449 sk_mem_uncharge(sk, skb->truesize);
1450 __kfree_skb(skb);
d80d99d6
HX
1451}
1452
1da177e4
LT
1453/* Used by processes to "lock" a socket state, so that
1454 * interrupts and bottom half handlers won't change it
1455 * from under us. It essentially blocks any incoming
1456 * packets, so that we won't get any new data or any
1457 * packets that change the state of the socket.
1458 *
1459 * While locked, BH processing will add new packets to
1460 * the backlog queue. This queue is processed by the
1461 * owner of the socket lock right before it is released.
1462 *
1463 * Since ~2.3.5 it is also exclusive sleep lock serializing
1464 * accesses from user process context.
1465 */
d2e9117c 1466#define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
1da177e4 1467
ed07536e
PZ
1468/*
1469 * Macro so as to not evaluate some arguments when
1470 * lockdep is not enabled.
1471 *
1472 * Mark both the sk_lock and the sk_lock.slock as a
1473 * per-address-family lock class.
1474 */
dc6b9b78 1475#define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
ed07536e 1476do { \
e8f6fbf6 1477 sk->sk_lock.owned = 0; \
ed07536e
PZ
1478 init_waitqueue_head(&sk->sk_lock.wq); \
1479 spin_lock_init(&(sk)->sk_lock.slock); \
1480 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1481 sizeof((sk)->sk_lock)); \
1482 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
dc6b9b78 1483 (skey), (sname)); \
ed07536e
PZ
1484 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1485} while (0)
1486
69336bd2 1487void lock_sock_nested(struct sock *sk, int subclass);
fcc70d5f
PZ
1488
1489static inline void lock_sock(struct sock *sk)
1490{
1491 lock_sock_nested(sk, 0);
1492}
1493
69336bd2 1494void release_sock(struct sock *sk);
1da177e4
LT
1495
1496/* BH context may only use the following locking interface. */
1497#define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
c6366184
IM
1498#define bh_lock_sock_nested(__sk) \
1499 spin_lock_nested(&((__sk)->sk_lock.slock), \
1500 SINGLE_DEPTH_NESTING)
1da177e4
LT
1501#define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1502
69336bd2 1503bool lock_sock_fast(struct sock *sk);
8a74ad60
ED
1504/**
1505 * unlock_sock_fast - complement of lock_sock_fast
1506 * @sk: socket
1507 * @slow: slow mode
1508 *
1509 * fast unlock socket for user context.
1510 * If slow mode is on, we call regular release_sock()
1511 */
1512static inline void unlock_sock_fast(struct sock *sk, bool slow)
4b0b72f7 1513{
8a74ad60
ED
1514 if (slow)
1515 release_sock(sk);
1516 else
1517 spin_unlock_bh(&sk->sk_lock.slock);
4b0b72f7
ED
1518}
1519
4b0b72f7 1520
69336bd2
JP
1521struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
1522 struct proto *prot);
1523void sk_free(struct sock *sk);
1524void sk_release_kernel(struct sock *sk);
1525struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority);
1526
1527struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
1528 gfp_t priority);
1529struct sk_buff *sock_rmalloc(struct sock *sk, unsigned long size, int force,
1530 gfp_t priority);
1531void sock_wfree(struct sk_buff *skb);
1532void skb_orphan_partial(struct sk_buff *skb);
1533void sock_rfree(struct sk_buff *skb);
1534void sock_edemux(struct sk_buff *skb);
1535
1536int sock_setsockopt(struct socket *sock, int level, int op,
1537 char __user *optval, unsigned int optlen);
1538
1539int sock_getsockopt(struct socket *sock, int level, int op,
1540 char __user *optval, int __user *optlen);
1541struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1542 int noblock, int *errcode);
1543struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1544 unsigned long data_len, int noblock,
1545 int *errcode, int max_page_order);
1546void *sock_kmalloc(struct sock *sk, int size, gfp_t priority);
1547void sock_kfree_s(struct sock *sk, void *mem, int size);
1548void sk_send_sigurg(struct sock *sk);
1da177e4
LT
1549
1550/*
1551 * Functions to fill in entries in struct proto_ops when a protocol
1552 * does not implement a particular function.
1553 */
69336bd2
JP
1554int sock_no_bind(struct socket *, struct sockaddr *, int);
1555int sock_no_connect(struct socket *, struct sockaddr *, int, int);
1556int sock_no_socketpair(struct socket *, struct socket *);
1557int sock_no_accept(struct socket *, struct socket *, int);
1558int sock_no_getname(struct socket *, struct sockaddr *, int *, int);
1559unsigned int sock_no_poll(struct file *, struct socket *,
1560 struct poll_table_struct *);
1561int sock_no_ioctl(struct socket *, unsigned int, unsigned long);
1562int sock_no_listen(struct socket *, int);
1563int sock_no_shutdown(struct socket *, int);
1564int sock_no_getsockopt(struct socket *, int , int, char __user *, int __user *);
1565int sock_no_setsockopt(struct socket *, int, int, char __user *, unsigned int);
1566int sock_no_sendmsg(struct kiocb *, struct socket *, struct msghdr *, size_t);
1567int sock_no_recvmsg(struct kiocb *, struct socket *, struct msghdr *, size_t,
1568 int);
1569int sock_no_mmap(struct file *file, struct socket *sock,
1570 struct vm_area_struct *vma);
1571ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset,
1572 size_t size, int flags);
1da177e4
LT
1573
1574/*
1575 * Functions to fill in entries in struct proto_ops when a protocol
1576 * uses the inet style.
1577 */
69336bd2 1578int sock_common_getsockopt(struct socket *sock, int level, int optname,
1da177e4 1579 char __user *optval, int __user *optlen);
69336bd2 1580int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1da177e4 1581 struct msghdr *msg, size_t size, int flags);
69336bd2 1582int sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 1583 char __user *optval, unsigned int optlen);
69336bd2 1584int compat_sock_common_getsockopt(struct socket *sock, int level,
3fdadf7d 1585 int optname, char __user *optval, int __user *optlen);
69336bd2 1586int compat_sock_common_setsockopt(struct socket *sock, int level,
b7058842 1587 int optname, char __user *optval, unsigned int optlen);
1da177e4 1588
69336bd2 1589void sk_common_release(struct sock *sk);
1da177e4
LT
1590
1591/*
1592 * Default socket callbacks and setup code
1593 */
dc6b9b78 1594
1da177e4 1595/* Initialise core socket variables */
69336bd2 1596void sock_init_data(struct socket *sock, struct sock *sk);
1da177e4 1597
69336bd2 1598void sk_filter_release_rcu(struct rcu_head *rcu);
46bcf14f 1599
dc9b3346 1600/**
1a5778aa 1601 * sk_filter_release - release a socket filter
dc9b3346
PB
1602 * @fp: filter to remove
1603 *
1604 * Remove a filter from a socket and release its resources.
1605 */
1606
309dd5fc
PE
1607static inline void sk_filter_release(struct sk_filter *fp)
1608{
1609 if (atomic_dec_and_test(&fp->refcnt))
80f8f102 1610 call_rcu(&fp->rcu, sk_filter_release_rcu);
309dd5fc
PE
1611}
1612
1613static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
1da177e4 1614{
d45ed4a4 1615 atomic_sub(sk_filter_size(fp->len), &sk->sk_omem_alloc);
309dd5fc 1616 sk_filter_release(fp);
1da177e4
LT
1617}
1618
1619static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
1620{
1621 atomic_inc(&fp->refcnt);
d45ed4a4 1622 atomic_add(sk_filter_size(fp->len), &sk->sk_omem_alloc);
1da177e4
LT
1623}
1624
1625/*
1626 * Socket reference counting postulates.
1627 *
1628 * * Each user of socket SHOULD hold a reference count.
1629 * * Each access point to socket (an hash table bucket, reference from a list,
1630 * running timer, skb in flight MUST hold a reference count.
1631 * * When reference count hits 0, it means it will never increase back.
1632 * * When reference count hits 0, it means that no references from
1633 * outside exist to this socket and current process on current CPU
1634 * is last user and may/should destroy this socket.
1635 * * sk_free is called from any context: process, BH, IRQ. When
1636 * it is called, socket has no references from outside -> sk_free
1637 * may release descendant resources allocated by the socket, but
1638 * to the time when it is called, socket is NOT referenced by any
1639 * hash tables, lists etc.
1640 * * Packets, delivered from outside (from network or from another process)
1641 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1642 * when they sit in queue. Otherwise, packets will leak to hole, when
1643 * socket is looked up by one cpu and unhasing is made by another CPU.
1644 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1645 * (leak to backlog). Packet socket does all the processing inside
1646 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1647 * use separate SMP lock, so that they are prone too.
1648 */
1649
1650/* Ungrab socket and destroy it, if it was the last reference. */
1651static inline void sock_put(struct sock *sk)
1652{
1653 if (atomic_dec_and_test(&sk->sk_refcnt))
1654 sk_free(sk);
1655}
1656
69336bd2 1657int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested);
25995ff5 1658
e022f0b4
KK
1659static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1660{
1661 sk->sk_tx_queue_mapping = tx_queue;
1662}
1663
1664static inline void sk_tx_queue_clear(struct sock *sk)
1665{
1666 sk->sk_tx_queue_mapping = -1;
1667}
1668
1669static inline int sk_tx_queue_get(const struct sock *sk)
1670{
b0f77d0e 1671 return sk ? sk->sk_tx_queue_mapping : -1;
e022f0b4
KK
1672}
1673
972692e0
DM
1674static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1675{
e022f0b4 1676 sk_tx_queue_clear(sk);
972692e0
DM
1677 sk->sk_socket = sock;
1678}
1679
aa395145
ED
1680static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1681{
eaefd110
ED
1682 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1683 return &rcu_dereference_raw(sk->sk_wq)->wait;
aa395145 1684}
1da177e4
LT
1685/* Detach socket from process context.
1686 * Announce socket dead, detach it from wait queue and inode.
1687 * Note that parent inode held reference count on this struct sock,
1688 * we do not release it in this function, because protocol
1689 * probably wants some additional cleanups or even continuing
1690 * to work with this socket (TCP).
1691 */
1692static inline void sock_orphan(struct sock *sk)
1693{
1694 write_lock_bh(&sk->sk_callback_lock);
1695 sock_set_flag(sk, SOCK_DEAD);
972692e0 1696 sk_set_socket(sk, NULL);
43815482 1697 sk->sk_wq = NULL;
1da177e4
LT
1698 write_unlock_bh(&sk->sk_callback_lock);
1699}
1700
1701static inline void sock_graft(struct sock *sk, struct socket *parent)
1702{
1703 write_lock_bh(&sk->sk_callback_lock);
eaefd110 1704 sk->sk_wq = parent->wq;
1da177e4 1705 parent->sk = sk;
972692e0 1706 sk_set_socket(sk, parent);
4237c75c 1707 security_sock_graft(sk, parent);
1da177e4
LT
1708 write_unlock_bh(&sk->sk_callback_lock);
1709}
1710
69336bd2
JP
1711kuid_t sock_i_uid(struct sock *sk);
1712unsigned long sock_i_ino(struct sock *sk);
1da177e4
LT
1713
1714static inline struct dst_entry *
1715__sk_dst_get(struct sock *sk)
1716{
d8bf4ca9 1717 return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
f68c224f 1718 lockdep_is_held(&sk->sk_lock.slock));
1da177e4
LT
1719}
1720
1721static inline struct dst_entry *
1722sk_dst_get(struct sock *sk)
1723{
1724 struct dst_entry *dst;
1725
b6c6712a
ED
1726 rcu_read_lock();
1727 dst = rcu_dereference(sk->sk_dst_cache);
1da177e4
LT
1728 if (dst)
1729 dst_hold(dst);
b6c6712a 1730 rcu_read_unlock();
1da177e4
LT
1731 return dst;
1732}
1733
69336bd2 1734void sk_reset_txq(struct sock *sk);
b6c6712a
ED
1735
1736static inline void dst_negative_advice(struct sock *sk)
1737{
1738 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1739
1740 if (dst && dst->ops->negative_advice) {
1741 ndst = dst->ops->negative_advice(dst);
1742
1743 if (ndst != dst) {
1744 rcu_assign_pointer(sk->sk_dst_cache, ndst);
1745 sk_reset_txq(sk);
1746 }
1747 }
1748}
1749
1da177e4
LT
1750static inline void
1751__sk_dst_set(struct sock *sk, struct dst_entry *dst)
1752{
1753 struct dst_entry *old_dst;
1754
e022f0b4 1755 sk_tx_queue_clear(sk);
0b53ff2e
ED
1756 /*
1757 * This can be called while sk is owned by the caller only,
1758 * with no state that can be checked in a rcu_dereference_check() cond
1759 */
1760 old_dst = rcu_dereference_raw(sk->sk_dst_cache);
b6c6712a 1761 rcu_assign_pointer(sk->sk_dst_cache, dst);
1da177e4
LT
1762 dst_release(old_dst);
1763}
1764
1765static inline void
1766sk_dst_set(struct sock *sk, struct dst_entry *dst)
1767{
b6c6712a 1768 spin_lock(&sk->sk_dst_lock);
1da177e4 1769 __sk_dst_set(sk, dst);
b6c6712a 1770 spin_unlock(&sk->sk_dst_lock);
1da177e4
LT
1771}
1772
1773static inline void
1774__sk_dst_reset(struct sock *sk)
1775{
b6c6712a 1776 __sk_dst_set(sk, NULL);
1da177e4
LT
1777}
1778
1779static inline void
1780sk_dst_reset(struct sock *sk)
1781{
b6c6712a 1782 spin_lock(&sk->sk_dst_lock);
1da177e4 1783 __sk_dst_reset(sk);
b6c6712a 1784 spin_unlock(&sk->sk_dst_lock);
1da177e4
LT
1785}
1786
69336bd2 1787struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1da177e4 1788
69336bd2 1789struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1da177e4 1790
dc6b9b78 1791static inline bool sk_can_gso(const struct sock *sk)
bcd76111
HX
1792{
1793 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1794}
1795
69336bd2 1796void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
6cbb0df7 1797
c8f44aff 1798static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
a465419b
ED
1799{
1800 sk->sk_route_nocaps |= flags;
1801 sk->sk_route_caps &= ~flags;
1802}
1803
c6e1a0d1
TH
1804static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1805 char __user *from, char *to,
912d398d 1806 int copy, int offset)
c6e1a0d1
TH
1807{
1808 if (skb->ip_summed == CHECKSUM_NONE) {
1809 int err = 0;
1810 __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err);
1811 if (err)
1812 return err;
912d398d 1813 skb->csum = csum_block_add(skb->csum, csum, offset);
c6e1a0d1
TH
1814 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1815 if (!access_ok(VERIFY_READ, from, copy) ||
1816 __copy_from_user_nocache(to, from, copy))
1817 return -EFAULT;
1818 } else if (copy_from_user(to, from, copy))
1819 return -EFAULT;
1820
1821 return 0;
1822}
1823
1824static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
1825 char __user *from, int copy)
1826{
912d398d 1827 int err, offset = skb->len;
c6e1a0d1 1828
912d398d
WY
1829 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1830 copy, offset);
c6e1a0d1 1831 if (err)
912d398d 1832 __skb_trim(skb, offset);
c6e1a0d1
TH
1833
1834 return err;
1835}
1836
1837static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from,
1838 struct sk_buff *skb,
1839 struct page *page,
1840 int off, int copy)
1841{
1842 int err;
1843
912d398d
WY
1844 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1845 copy, skb->len);
c6e1a0d1
TH
1846 if (err)
1847 return err;
1848
1849 skb->len += copy;
1850 skb->data_len += copy;
1851 skb->truesize += copy;
1852 sk->sk_wmem_queued += copy;
1853 sk_mem_charge(sk, copy);
1854 return 0;
1855}
1856
1da177e4
LT
1857static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1858 struct sk_buff *skb, struct page *page,
1859 int off, int copy)
1860{
1861 if (skb->ip_summed == CHECKSUM_NONE) {
1862 int err = 0;
5084205f 1863 __wsum csum = csum_and_copy_from_user(from,
1da177e4
LT
1864 page_address(page) + off,
1865 copy, 0, &err);
1866 if (err)
1867 return err;
1868 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1869 } else if (copy_from_user(page_address(page) + off, from, copy))
1870 return -EFAULT;
1871
1872 skb->len += copy;
1873 skb->data_len += copy;
1874 skb->truesize += copy;
1875 sk->sk_wmem_queued += copy;
3ab224be 1876 sk_mem_charge(sk, copy);
1da177e4
LT
1877 return 0;
1878}
1879
c564039f
ED
1880/**
1881 * sk_wmem_alloc_get - returns write allocations
1882 * @sk: socket
1883 *
1884 * Returns sk_wmem_alloc minus initial offset of one
1885 */
1886static inline int sk_wmem_alloc_get(const struct sock *sk)
1887{
1888 return atomic_read(&sk->sk_wmem_alloc) - 1;
1889}
1890
1891/**
1892 * sk_rmem_alloc_get - returns read allocations
1893 * @sk: socket
1894 *
1895 * Returns sk_rmem_alloc
1896 */
1897static inline int sk_rmem_alloc_get(const struct sock *sk)
1898{
1899 return atomic_read(&sk->sk_rmem_alloc);
1900}
1901
1902/**
1903 * sk_has_allocations - check if allocations are outstanding
1904 * @sk: socket
1905 *
1906 * Returns true if socket has write or read allocations
1907 */
dc6b9b78 1908static inline bool sk_has_allocations(const struct sock *sk)
c564039f
ED
1909{
1910 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1911}
1912
a57de0b4 1913/**
43815482 1914 * wq_has_sleeper - check if there are any waiting processes
acfbe96a 1915 * @wq: struct socket_wq
a57de0b4 1916 *
43815482 1917 * Returns true if socket_wq has waiting processes
a57de0b4 1918 *
43815482 1919 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
a57de0b4
JO
1920 * barrier call. They were added due to the race found within the tcp code.
1921 *
1922 * Consider following tcp code paths:
1923 *
1924 * CPU1 CPU2
1925 *
1926 * sys_select receive packet
1927 * ... ...
1928 * __add_wait_queue update tp->rcv_nxt
1929 * ... ...
1930 * tp->rcv_nxt check sock_def_readable
1931 * ... {
43815482
ED
1932 * schedule rcu_read_lock();
1933 * wq = rcu_dereference(sk->sk_wq);
1934 * if (wq && waitqueue_active(&wq->wait))
1935 * wake_up_interruptible(&wq->wait)
a57de0b4
JO
1936 * ...
1937 * }
1938 *
1939 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1940 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1941 * could then endup calling schedule and sleep forever if there are no more
1942 * data on the socket.
ad462769 1943 *
a57de0b4 1944 */
43815482 1945static inline bool wq_has_sleeper(struct socket_wq *wq)
a57de0b4 1946{
dc6b9b78 1947 /* We need to be sure we are in sync with the
a57de0b4
JO
1948 * add_wait_queue modifications to the wait queue.
1949 *
1950 * This memory barrier is paired in the sock_poll_wait.
1951 */
43815482
ED
1952 smp_mb();
1953 return wq && waitqueue_active(&wq->wait);
a57de0b4
JO
1954}
1955
1956/**
1957 * sock_poll_wait - place memory barrier behind the poll_wait call.
1958 * @filp: file
1959 * @wait_address: socket wait queue
1960 * @p: poll_table
1961 *
43815482 1962 * See the comments in the wq_has_sleeper function.
a57de0b4
JO
1963 */
1964static inline void sock_poll_wait(struct file *filp,
1965 wait_queue_head_t *wait_address, poll_table *p)
1966{
626cf236 1967 if (!poll_does_not_wait(p) && wait_address) {
a57de0b4 1968 poll_wait(filp, wait_address, p);
dc6b9b78 1969 /* We need to be sure we are in sync with the
a57de0b4
JO
1970 * socket flags modification.
1971 *
43815482 1972 * This memory barrier is paired in the wq_has_sleeper.
dc6b9b78 1973 */
a57de0b4
JO
1974 smp_mb();
1975 }
1976}
1977
1da177e4 1978/*
dc6b9b78 1979 * Queue a received datagram if it will fit. Stream and sequenced
1da177e4
LT
1980 * protocols can't normally use this as they need to fit buffers in
1981 * and play with them.
1982 *
dc6b9b78 1983 * Inlined as it's very short and called for pretty much every
1da177e4
LT
1984 * packet ever received.
1985 */
1986
1987static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1988{
d55d87fd 1989 skb_orphan(skb);
1da177e4
LT
1990 skb->sk = sk;
1991 skb->destructor = sock_wfree;
2b85a34e
ED
1992 /*
1993 * We used to take a refcount on sk, but following operation
1994 * is enough to guarantee sk_free() wont free this sock until
1995 * all in-flight packets are completed
1996 */
1da177e4
LT
1997 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1998}
1999
2000static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
2001{
d55d87fd 2002 skb_orphan(skb);
1da177e4
LT
2003 skb->sk = sk;
2004 skb->destructor = sock_rfree;
2005 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
3ab224be 2006 sk_mem_charge(sk, skb->truesize);
1da177e4
LT
2007}
2008
69336bd2
JP
2009void sk_reset_timer(struct sock *sk, struct timer_list *timer,
2010 unsigned long expires);
1da177e4 2011
69336bd2 2012void sk_stop_timer(struct sock *sk, struct timer_list *timer);
1da177e4 2013
69336bd2 2014int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
1da177e4 2015
69336bd2 2016int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
1da177e4
LT
2017
2018/*
2019 * Recover an error report and clear atomically
2020 */
dc6b9b78 2021
1da177e4
LT
2022static inline int sock_error(struct sock *sk)
2023{
c1cbe4b7
BL
2024 int err;
2025 if (likely(!sk->sk_err))
2026 return 0;
2027 err = xchg(&sk->sk_err, 0);
1da177e4
LT
2028 return -err;
2029}
2030
2031static inline unsigned long sock_wspace(struct sock *sk)
2032{
2033 int amt = 0;
2034
2035 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
2036 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
dc6b9b78 2037 if (amt < 0)
1da177e4
LT
2038 amt = 0;
2039 }
2040 return amt;
2041}
2042
2043static inline void sk_wake_async(struct sock *sk, int how, int band)
2044{
bcdce719 2045 if (sock_flag(sk, SOCK_FASYNC))
1da177e4
LT
2046 sock_wake_async(sk->sk_socket, how, band);
2047}
2048
eea86af6
DB
2049/* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
2050 * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
2051 * Note: for send buffers, TCP works better if we can build two skbs at
2052 * minimum.
7a91b434 2053 */
9eb5bf83 2054#define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
eea86af6
DB
2055
2056#define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2)
2057#define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE
1da177e4
LT
2058
2059static inline void sk_stream_moderate_sndbuf(struct sock *sk)
2060{
2061 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
8df09ea3 2062 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
eea86af6 2063 sk->sk_sndbuf = max_t(u32, sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1da177e4
LT
2064 }
2065}
2066
df97c708 2067struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
1da177e4 2068
5640f768
ED
2069/**
2070 * sk_page_frag - return an appropriate page_frag
2071 * @sk: socket
2072 *
2073 * If socket allocation mode allows current thread to sleep, it means its
2074 * safe to use the per task page_frag instead of the per socket one.
2075 */
2076static inline struct page_frag *sk_page_frag(struct sock *sk)
1da177e4 2077{
5640f768
ED
2078 if (sk->sk_allocation & __GFP_WAIT)
2079 return &current->task_frag;
1da177e4 2080
5640f768 2081 return &sk->sk_frag;
1da177e4
LT
2082}
2083
69336bd2 2084bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag);
5640f768 2085
1da177e4
LT
2086/*
2087 * Default write policy as shown to user space via poll/select/SIGIO
2088 */
dc6b9b78 2089static inline bool sock_writeable(const struct sock *sk)
1da177e4 2090{
8df09ea3 2091 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
1da177e4
LT
2092}
2093
dd0fc66f 2094static inline gfp_t gfp_any(void)
1da177e4 2095{
99709372 2096 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1da177e4
LT
2097}
2098
dc6b9b78 2099static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
1da177e4
LT
2100{
2101 return noblock ? 0 : sk->sk_rcvtimeo;
2102}
2103
dc6b9b78 2104static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
1da177e4
LT
2105{
2106 return noblock ? 0 : sk->sk_sndtimeo;
2107}
2108
2109static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2110{
2111 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
2112}
2113
2114/* Alas, with timeout socket operations are not restartable.
2115 * Compare this to poll().
2116 */
2117static inline int sock_intr_errno(long timeo)
2118{
2119 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2120}
2121
69336bd2
JP
2122void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2123 struct sk_buff *skb);
2124void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2125 struct sk_buff *skb);
92f37fd2 2126
dc6b9b78 2127static inline void
1da177e4
LT
2128sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2129{
b7aa0bf7 2130 ktime_t kt = skb->tstamp;
20d49473 2131 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
a61bbcf2 2132
20d49473
PO
2133 /*
2134 * generate control messages if
2135 * - receive time stamping in software requested (SOCK_RCVTSTAMP
2136 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
2137 * - software time stamp available and wanted
2138 * (SOCK_TIMESTAMPING_SOFTWARE)
2139 * - hardware time stamps available and wanted
2140 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
2141 * SOCK_TIMESTAMPING_RAW_HARDWARE)
2142 */
2143 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
2144 sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
2145 (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
2146 (hwtstamps->hwtstamp.tv64 &&
2147 sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
2148 (hwtstamps->syststamp.tv64 &&
2149 sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
92f37fd2
ED
2150 __sock_recv_timestamp(msg, sk, skb);
2151 else
b7aa0bf7 2152 sk->sk_stamp = kt;
6e3e939f
JB
2153
2154 if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2155 __sock_recv_wifi_status(msg, sk, skb);
1da177e4
LT
2156}
2157
69336bd2
JP
2158void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2159 struct sk_buff *skb);
767dd033
ED
2160
2161static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2162 struct sk_buff *skb)
2163{
2164#define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2165 (1UL << SOCK_RCVTSTAMP) | \
2166 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
2167 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
dc6b9b78 2168 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
767dd033
ED
2169 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
2170
2171 if (sk->sk_flags & FLAGS_TS_OR_DROPS)
2172 __sock_recv_ts_and_drops(msg, sk, skb);
2173 else
2174 sk->sk_stamp = skb->tstamp;
2175}
3b885787 2176
20d49473
PO
2177/**
2178 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
20d49473 2179 * @sk: socket sending this packet
2244d07b 2180 * @tx_flags: filled with instructions for time stamping
20d49473 2181 *
bf84a010 2182 * Currently only depends on SOCK_TIMESTAMPING* flags.
20d49473 2183 */
69336bd2 2184void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags);
20d49473 2185
1da177e4
LT
2186/**
2187 * sk_eat_skb - Release a skb if it is no longer needed
4dc3b16b
PP
2188 * @sk: socket to eat this skb from
2189 * @skb: socket buffer to eat
f4b8ea78 2190 * @copied_early: flag indicating whether DMA operations copied this data early
1da177e4
LT
2191 *
2192 * This routine must be called with interrupts disabled or with the socket
2193 * locked so that the sk_buff queue operation is ok.
2194*/
624d1164 2195#ifdef CONFIG_NET_DMA
dc6b9b78 2196static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
624d1164
CL
2197{
2198 __skb_unlink(skb, &sk->sk_receive_queue);
2199 if (!copied_early)
2200 __kfree_skb(skb);
2201 else
2202 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
2203}
2204#else
dc6b9b78 2205static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
1da177e4
LT
2206{
2207 __skb_unlink(skb, &sk->sk_receive_queue);
2208 __kfree_skb(skb);
2209}
624d1164 2210#endif
1da177e4 2211
3b1e0a65
YH
2212static inline
2213struct net *sock_net(const struct sock *sk)
2214{
c2d9ba9b 2215 return read_pnet(&sk->sk_net);
3b1e0a65
YH
2216}
2217
2218static inline
f5aa23fd 2219void sock_net_set(struct sock *sk, struct net *net)
3b1e0a65 2220{
c2d9ba9b 2221 write_pnet(&sk->sk_net, net);
3b1e0a65
YH
2222}
2223
edf02087
DL
2224/*
2225 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
25985edc 2226 * They should not hold a reference to a namespace in order to allow
edf02087
DL
2227 * to stop it.
2228 * Sockets after sk_change_net should be released using sk_release_kernel
2229 */
2230static inline void sk_change_net(struct sock *sk, struct net *net)
2231{
3b1e0a65 2232 put_net(sock_net(sk));
65a18ec5 2233 sock_net_set(sk, hold_net(net));
edf02087
DL
2234}
2235
23542618
KK
2236static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2237{
efc27f8c 2238 if (skb->sk) {
23542618
KK
2239 struct sock *sk = skb->sk;
2240
2241 skb->destructor = NULL;
2242 skb->sk = NULL;
2243 return sk;
2244 }
2245 return NULL;
2246}
2247
69336bd2
JP
2248void sock_enable_timestamp(struct sock *sk, int flag);
2249int sock_get_timestamp(struct sock *, struct timeval __user *);
2250int sock_get_timestampns(struct sock *, struct timespec __user *);
2251int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len, int level,
2252 int type);
1da177e4 2253
dc6b9b78
ED
2254/*
2255 * Enable debug/info messages
1da177e4 2256 */
a2a316fd
SH
2257extern int net_msg_warn;
2258#define NETDEBUG(fmt, args...) \
2259 do { if (net_msg_warn) printk(fmt,##args); } while (0)
1da177e4 2260
a2a316fd
SH
2261#define LIMIT_NETDEBUG(fmt, args...) \
2262 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
1da177e4 2263
1da177e4
LT
2264extern __u32 sysctl_wmem_max;
2265extern __u32 sysctl_rmem_max;
2266
6baf1f41
DM
2267extern int sysctl_optmem_max;
2268
20380731
ACM
2269extern __u32 sysctl_wmem_default;
2270extern __u32 sysctl_rmem_default;
20380731 2271
1da177e4 2272#endif /* _SOCK_H */