]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - include/net/sock.h
net: fix sk_page_frag() recursion from memory reclaim
[mirror_ubuntu-bionic-kernel.git] / include / net / sock.h
CommitLineData
1da177e4
LT
1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the AF_INET socket handler.
7 *
8 * Version: @(#)sock.h 1.0.4 05/13/93
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Florian La Roche <flla@stud.uni-sb.de>
14 *
15 * Fixes:
16 * Alan Cox : Volatiles in skbuff pointers. See
17 * skbuff comments. May be overdone,
18 * better to prove they can be removed
19 * than the reverse.
20 * Alan Cox : Added a zapped field for tcp to note
21 * a socket is reset and must stay shut up
22 * Alan Cox : New fields for options
23 * Pauline Middelink : identd support
24 * Alan Cox : Eliminate low level recv/recvfrom
25 * David S. Miller : New socket lookup architecture.
26 * Steve Whitehouse: Default routines for sock_ops
27 * Arnaldo C. Melo : removed net_pinfo, tp_pinfo and made
28 * protinfo be just a void pointer, as the
29 * protocol specific parts were moved to
30 * respective headers and ipv4/v6, etc now
31 * use private slabcaches for its socks
32 * Pedro Hortas : New flags field for socket options
33 *
34 *
35 * This program is free software; you can redistribute it and/or
36 * modify it under the terms of the GNU General Public License
37 * as published by the Free Software Foundation; either version
38 * 2 of the License, or (at your option) any later version.
39 */
40#ifndef _SOCK_H
41#define _SOCK_H
42
a6b7a407 43#include <linux/hardirq.h>
172589cc 44#include <linux/kernel.h>
1da177e4 45#include <linux/list.h>
88ab1932 46#include <linux/list_nulls.h>
1da177e4
LT
47#include <linux/timer.h>
48#include <linux/cache.h>
3f134619 49#include <linux/bitops.h>
a5b5bb9a 50#include <linux/lockdep.h>
1da177e4
LT
51#include <linux/netdevice.h>
52#include <linux/skbuff.h> /* struct sk_buff */
d7fe0f24 53#include <linux/mm.h>
1da177e4 54#include <linux/security.h>
5a0e3ad6 55#include <linux/slab.h>
c6e1a0d1 56#include <linux/uaccess.h>
3e32cb2e 57#include <linux/page_counter.h>
180d8cd9 58#include <linux/memcontrol.h>
c5905afb 59#include <linux/static_key.h>
40401530 60#include <linux/sched.h>
1ce0bf50 61#include <linux/wait.h>
2a56a1fe 62#include <linux/cgroup-defs.h>
75c119af 63#include <linux/rbtree.h>
1da177e4 64#include <linux/filter.h>
88ab1932 65#include <linux/rculist_nulls.h>
a57de0b4 66#include <linux/poll.h>
1da177e4 67
c31504dc 68#include <linux/atomic.h>
41c6d650 69#include <linux/refcount.h>
1da177e4
LT
70#include <net/dst.h>
71#include <net/checksum.h>
1d0ab253 72#include <net/tcp_states.h>
b9f40e21 73#include <linux/net_tstamp.h>
f16a7dd5 74#include <net/smc.h>
1da177e4
LT
75
76/*
77 * This structure really needs to be cleaned up.
78 * Most of it is for TCP, and not used by any of
79 * the other protocols.
80 */
81
82/* Define this to get the SOCK_DBG debugging facility. */
83#define SOCK_DEBUGGING
84#ifdef SOCK_DEBUGGING
85#define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
86 printk(KERN_DEBUG msg); } while (0)
87#else
4cd9029d 88/* Validate arguments and do nothing */
b9075fa9 89static inline __printf(2, 3)
dc6b9b78 90void SOCK_DEBUG(const struct sock *sk, const char *msg, ...)
4cd9029d
SH
91{
92}
1da177e4
LT
93#endif
94
95/* This is the per-socket lock. The spinlock provides a synchronization
96 * between user contexts and software interrupt processing, whereas the
97 * mini-semaphore synchronizes multiple users amongst themselves.
98 */
1da177e4
LT
99typedef struct {
100 spinlock_t slock;
d2e9117c 101 int owned;
1da177e4 102 wait_queue_head_t wq;
a5b5bb9a
IM
103 /*
104 * We express the mutex-alike socket_lock semantics
105 * to the lock validator by explicitly managing
106 * the slock as a lock variant (in addition to
107 * the slock itself):
108 */
109#ifdef CONFIG_DEBUG_LOCK_ALLOC
110 struct lockdep_map dep_map;
111#endif
1da177e4
LT
112} socket_lock_t;
113
1da177e4 114struct sock;
8feaf0c0 115struct proto;
0eeb8ffc 116struct net;
1da177e4 117
077b393d
ED
118typedef __u32 __bitwise __portpair;
119typedef __u64 __bitwise __addrpair;
120
1da177e4 121/**
4dc3b16b 122 * struct sock_common - minimal network layer representation of sockets
68835aba
ED
123 * @skc_daddr: Foreign IPv4 addr
124 * @skc_rcv_saddr: Bound local IPv4 addr
4dc6dc71 125 * @skc_hash: hash value used with various protocol lookup tables
d4cada4a 126 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
ce43b03e
ED
127 * @skc_dport: placeholder for inet_dport/tw_dport
128 * @skc_num: placeholder for inet_num/tw_num
4dc3b16b
PP
129 * @skc_family: network address family
130 * @skc_state: Connection state
131 * @skc_reuse: %SO_REUSEADDR setting
055dc21a 132 * @skc_reuseport: %SO_REUSEPORT setting
4dc3b16b 133 * @skc_bound_dev_if: bound device index if != 0
4dc3b16b 134 * @skc_bind_node: bind hash linkage for various protocol lookup tables
512615b6 135 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
8feaf0c0 136 * @skc_prot: protocol handlers inside a network family
07feaebf 137 * @skc_net: reference to the network namespace of this socket
68835aba
ED
138 * @skc_node: main hash linkage for various protocol lookup tables
139 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
140 * @skc_tx_queue_mapping: tx queue number for this connection
8e5eb54d
ED
141 * @skc_flags: place holder for sk_flags
142 * %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
143 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
70da268b 144 * @skc_incoming_cpu: record/match cpu processing incoming packets
68835aba 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 {
ce43b03e 151 /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned
05dbc7b5 152 * address on 64bit arches : cf INET_MATCH()
4dc6dc71 153 */
ce43b03e 154 union {
077b393d 155 __addrpair skc_addrpair;
ce43b03e
ED
156 struct {
157 __be32 skc_daddr;
158 __be32 skc_rcv_saddr;
159 };
160 };
d4cada4a
ED
161 union {
162 unsigned int skc_hash;
163 __u16 skc_u16hashes[2];
164 };
ce43b03e
ED
165 /* skc_dport && skc_num must be grouped as well */
166 union {
077b393d 167 __portpair skc_portpair;
ce43b03e
ED
168 struct {
169 __be16 skc_dport;
170 __u16 skc_num;
171 };
172 };
173
4dc6dc71
ED
174 unsigned short skc_family;
175 volatile unsigned char skc_state;
055dc21a 176 unsigned char skc_reuse:4;
9fe516ba
ED
177 unsigned char skc_reuseport:1;
178 unsigned char skc_ipv6only:1;
26abe143 179 unsigned char skc_net_refcnt:1;
4dc6dc71 180 int skc_bound_dev_if;
512615b6
ED
181 union {
182 struct hlist_node skc_bind_node;
ca065d0c 183 struct hlist_node skc_portaddr_node;
512615b6 184 };
8feaf0c0 185 struct proto *skc_prot;
0c5c9fb5 186 possible_net_t skc_net;
efe4208f
ED
187
188#if IS_ENABLED(CONFIG_IPV6)
189 struct in6_addr skc_v6_daddr;
190 struct in6_addr skc_v6_rcv_saddr;
191#endif
192
33cf7c90
ED
193 atomic64_t skc_cookie;
194
8e5eb54d
ED
195 /* following fields are padding to force
196 * offset(struct sock, sk_refcnt) == 128 on 64bit arches
197 * assuming IPV6 is enabled. We use this padding differently
198 * for different kind of 'sockets'
199 */
200 union {
201 unsigned long skc_flags;
202 struct sock *skc_listener; /* request_sock */
203 struct inet_timewait_death_row *skc_tw_dr; /* inet_timewait_sock */
204 };
68835aba
ED
205 /*
206 * fields between dontcopy_begin/dontcopy_end
207 * are not copied in sock_copy()
208 */
928c41e7 209 /* private: */
68835aba 210 int skc_dontcopy_begin[0];
928c41e7 211 /* public: */
68835aba
ED
212 union {
213 struct hlist_node skc_node;
214 struct hlist_nulls_node skc_nulls_node;
215 };
216 int skc_tx_queue_mapping;
ed53d0ab
ED
217 union {
218 int skc_incoming_cpu;
219 u32 skc_rcv_wnd;
d475f090 220 u32 skc_tw_rcv_nxt; /* struct tcp_timewait_sock */
ed53d0ab 221 };
70da268b 222
41c6d650 223 refcount_t skc_refcnt;
928c41e7 224 /* private: */
68835aba 225 int skc_dontcopy_end[0];
ed53d0ab
ED
226 union {
227 u32 skc_rxhash;
228 u32 skc_window_clamp;
d475f090 229 u32 skc_tw_snd_nxt; /* struct tcp_timewait_sock */
ed53d0ab 230 };
928c41e7 231 /* public: */
1da177e4
LT
232};
233
234/**
235 * struct sock - network layer representation of sockets
8feaf0c0 236 * @__sk_common: shared layout with inet_timewait_sock
4dc3b16b
PP
237 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
238 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
239 * @sk_lock: synchronizer
cdfbabfb 240 * @sk_kern_sock: True if sock is using kernel lock classes
4dc3b16b 241 * @sk_rcvbuf: size of receive buffer in bytes
43815482 242 * @sk_wq: sock wait queue and async head
421b3885 243 * @sk_rx_dst: receive input route used by early demux
4dc3b16b 244 * @sk_dst_cache: destination cache
9b8805a3 245 * @sk_dst_pending_confirm: need to confirm neighbour
4dc3b16b 246 * @sk_policy: flow policy
4dc3b16b
PP
247 * @sk_receive_queue: incoming packets
248 * @sk_wmem_alloc: transmit queue bytes committed
771edcaf 249 * @sk_tsq_flags: TCP Small Queues flags
4dc3b16b
PP
250 * @sk_write_queue: Packet sending queue
251 * @sk_omem_alloc: "o" is "option" or "other"
252 * @sk_wmem_queued: persistent queue size
253 * @sk_forward_alloc: space allocated forward
06021292 254 * @sk_napi_id: id of the last napi context to receive data for sk
dafcc438 255 * @sk_ll_usec: usecs to busypoll when there is no data
4dc3b16b 256 * @sk_allocation: allocation mode
95bd09eb 257 * @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
218af599 258 * @sk_pacing_status: Pacing status (requested, handled by sch_fq)
c3f40d7c 259 * @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE)
4dc3b16b 260 * @sk_sndbuf: size of send buffer in bytes
771edcaf 261 * @__sk_flags_offset: empty field used to determine location of bitfield
293de7de 262 * @sk_padding: unused element for alignment
28448b80
TH
263 * @sk_no_check_tx: %SO_NO_CHECK setting, set checksum in TX packets
264 * @sk_no_check_rx: allow zero checksum in RX packets
4dc3b16b 265 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
a465419b 266 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
bcd76111 267 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
82cc1a7a 268 * @sk_gso_max_size: Maximum GSO segment size to build
1485348d 269 * @sk_gso_max_segs: Maximum number of GSO segments
3a9b76fd 270 * @sk_pacing_shift: scaling factor for TCP Small Queues
4dc3b16b 271 * @sk_lingertime: %SO_LINGER l_linger setting
4dc3b16b
PP
272 * @sk_backlog: always used with the per-socket spinlock held
273 * @sk_callback_lock: used with the callbacks in the end of this struct
274 * @sk_error_queue: rarely used
33c732c3
WC
275 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
276 * IPV6_ADDRFORM for instance)
4dc3b16b 277 * @sk_err: last error
33c732c3
WC
278 * @sk_err_soft: errors that don't cause failure but are the cause of a
279 * persistent failure not just 'timed out'
cb61cb9b 280 * @sk_drops: raw/udp drops counter
4dc3b16b
PP
281 * @sk_ack_backlog: current listen backlog
282 * @sk_max_ack_backlog: listen backlog set in listen()
771edcaf 283 * @sk_uid: user id of owner
4dc3b16b
PP
284 * @sk_priority: %SO_PRIORITY setting
285 * @sk_type: socket type (%SOCK_STREAM, etc)
286 * @sk_protocol: which protocol this socket belongs in this network family
53c3fa20
RD
287 * @sk_peer_pid: &struct pid for this socket's peer
288 * @sk_peer_cred: %SO_PEERCRED setting
4dc3b16b
PP
289 * @sk_rcvlowat: %SO_RCVLOWAT setting
290 * @sk_rcvtimeo: %SO_RCVTIMEO setting
291 * @sk_sndtimeo: %SO_SNDTIMEO setting
b73c3d0e 292 * @sk_txhash: computed flow hash for use on transmit
4dc3b16b 293 * @sk_filter: socket filtering instructions
4dc3b16b
PP
294 * @sk_timer: sock cleanup timer
295 * @sk_stamp: time stamp of last packet received
2bdc9211 296 * @sk_stamp_seq: lock for accessing sk_stamp on 32 bit architectures only
b9f40e21 297 * @sk_tsflags: SO_TIMESTAMPING socket options
09c2d251 298 * @sk_tskey: counter to disambiguate concurrent tstamp requests
52267790 299 * @sk_zckey: counter to order MSG_ZEROCOPY notifications
4dc3b16b
PP
300 * @sk_socket: Identd and reporting IO signals
301 * @sk_user_data: RPC layer private data
5640f768 302 * @sk_frag: cached page frag
d3d4f0a0 303 * @sk_peek_off: current peek_offset value
4dc3b16b 304 * @sk_send_head: front of stuff to transmit
67be2dd1 305 * @sk_security: used by security modules
31729363 306 * @sk_mark: generic packet mark
2a56a1fe 307 * @sk_cgrp_data: cgroup data for this cgroup
baac50bb 308 * @sk_memcg: this socket's memory cgroup association
4dc3b16b
PP
309 * @sk_write_pending: a write to stream socket waits to start
310 * @sk_state_change: callback to indicate change in the state of the sock
311 * @sk_data_ready: callback to indicate there is data to be processed
312 * @sk_write_space: callback to indicate there is bf sending space available
313 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
314 * @sk_backlog_rcv: callback to process the backlog
315 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
ef456144 316 * @sk_reuseport_cb: reuseport group container
293de7de
SH
317 * @sk_rcu: used during RCU grace period
318 */
1da177e4
LT
319struct sock {
320 /*
8feaf0c0 321 * Now struct inet_timewait_sock also uses sock_common, so please just
1da177e4
LT
322 * don't add nothing before this first member (__sk_common) --acme
323 */
324 struct sock_common __sk_common;
4dc6dc71
ED
325#define sk_node __sk_common.skc_node
326#define sk_nulls_node __sk_common.skc_nulls_node
327#define sk_refcnt __sk_common.skc_refcnt
e022f0b4 328#define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
4dc6dc71 329
68835aba
ED
330#define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
331#define sk_dontcopy_end __sk_common.skc_dontcopy_end
4dc6dc71 332#define sk_hash __sk_common.skc_hash
50805466 333#define sk_portpair __sk_common.skc_portpair
05dbc7b5
ED
334#define sk_num __sk_common.skc_num
335#define sk_dport __sk_common.skc_dport
50805466
ED
336#define sk_addrpair __sk_common.skc_addrpair
337#define sk_daddr __sk_common.skc_daddr
338#define sk_rcv_saddr __sk_common.skc_rcv_saddr
1da177e4
LT
339#define sk_family __sk_common.skc_family
340#define sk_state __sk_common.skc_state
341#define sk_reuse __sk_common.skc_reuse
055dc21a 342#define sk_reuseport __sk_common.skc_reuseport
9fe516ba 343#define sk_ipv6only __sk_common.skc_ipv6only
26abe143 344#define sk_net_refcnt __sk_common.skc_net_refcnt
1da177e4 345#define sk_bound_dev_if __sk_common.skc_bound_dev_if
1da177e4 346#define sk_bind_node __sk_common.skc_bind_node
8feaf0c0 347#define sk_prot __sk_common.skc_prot
07feaebf 348#define sk_net __sk_common.skc_net
efe4208f
ED
349#define sk_v6_daddr __sk_common.skc_v6_daddr
350#define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr
33cf7c90 351#define sk_cookie __sk_common.skc_cookie
70da268b 352#define sk_incoming_cpu __sk_common.skc_incoming_cpu
8e5eb54d 353#define sk_flags __sk_common.skc_flags
ed53d0ab 354#define sk_rxhash __sk_common.skc_rxhash
efe4208f 355
1da177e4 356 socket_lock_t sk_lock;
9115e8cd
ED
357 atomic_t sk_drops;
358 int sk_rcvlowat;
359 struct sk_buff_head sk_error_queue;
b178bb3d 360 struct sk_buff_head sk_receive_queue;
fa438ccf
ED
361 /*
362 * The backlog queue is special, it is always used with
363 * the per-socket spinlock held and requires low latency
364 * access. Therefore we special case it's implementation.
b178bb3d
ED
365 * Note : rmem_alloc is in this structure to fill a hole
366 * on 64bit arches, not because its logically part of
367 * backlog.
fa438ccf
ED
368 */
369 struct {
b178bb3d
ED
370 atomic_t rmem_alloc;
371 int len;
372 struct sk_buff *head;
373 struct sk_buff *tail;
fa438ccf 374 } sk_backlog;
b178bb3d 375#define sk_rmem_alloc sk_backlog.rmem_alloc
2c8c56e1 376
9115e8cd 377 int sk_forward_alloc;
e0d1095a 378#ifdef CONFIG_NET_RX_BUSY_POLL
dafcc438 379 unsigned int sk_ll_usec;
9115e8cd
ED
380 /* ===== mostly read cache line ===== */
381 unsigned int sk_napi_id;
b178bb3d 382#endif
b178bb3d
ED
383 int sk_rcvbuf;
384
385 struct sk_filter __rcu *sk_filter;
ceb5d58b
ED
386 union {
387 struct socket_wq __rcu *sk_wq;
388 struct socket_wq *sk_wq_raw;
389 };
def8b4fa 390#ifdef CONFIG_XFRM
d188ba86 391 struct xfrm_policy __rcu *sk_policy[2];
def8b4fa 392#endif
deaa5854 393 struct dst_entry *sk_rx_dst;
0e36cbb3 394 struct dst_entry __rcu *sk_dst_cache;
1da177e4 395 atomic_t sk_omem_alloc;
4e07a91c 396 int sk_sndbuf;
9115e8cd
ED
397
398 /* ===== cache line for TX ===== */
399 int sk_wmem_queued;
14afee4b 400 refcount_t sk_wmem_alloc;
9115e8cd 401 unsigned long sk_tsq_flags;
75c119af
ED
402 union {
403 struct sk_buff *sk_send_head;
404 struct rb_root tcp_rtx_queue;
405 };
1da177e4 406 struct sk_buff_head sk_write_queue;
9115e8cd
ED
407 __s32 sk_peek_off;
408 int sk_write_pending;
9b8805a3 409 __u32 sk_dst_pending_confirm;
218af599 410 u32 sk_pacing_status; /* see enum sk_pacing */
9115e8cd
ED
411 long sk_sndtimeo;
412 struct timer_list sk_timer;
413 __u32 sk_priority;
414 __u32 sk_mark;
415 u32 sk_pacing_rate; /* bytes per second */
416 u32 sk_max_pacing_rate;
417 struct page_frag sk_frag;
418 netdev_features_t sk_route_caps;
419 netdev_features_t sk_route_nocaps;
420 int sk_gso_type;
421 unsigned int sk_gso_max_size;
422 gfp_t sk_allocation;
423 __u32 sk_txhash;
fc64869c
AR
424
425 /*
426 * Because of non atomicity rules, all
427 * changes are protected by socket lock.
428 */
aa4c1037
DA
429 unsigned int __sk_flags_offset[0];
430#ifdef __BIG_ENDIAN_BITFIELD
431#define SK_FL_PROTO_SHIFT 16
432#define SK_FL_PROTO_MASK 0x00ff0000
433
434#define SK_FL_TYPE_SHIFT 0
435#define SK_FL_TYPE_MASK 0x0000ffff
436#else
437#define SK_FL_PROTO_SHIFT 8
438#define SK_FL_PROTO_MASK 0x0000ff00
439
440#define SK_FL_TYPE_SHIFT 16
441#define SK_FL_TYPE_MASK 0xffff0000
442#endif
443
cdfbabfb
DH
444 unsigned int sk_padding : 1,
445 sk_kern_sock : 1,
28448b80
TH
446 sk_no_check_tx : 1,
447 sk_no_check_rx : 1,
b178bb3d
ED
448 sk_userlocks : 4,
449 sk_protocol : 8,
450 sk_type : 16;
7bbadd2d 451#define SK_PROTOCOL_MAX U8_MAX
1485348d 452 u16 sk_gso_max_segs;
3a9b76fd 453 u8 sk_pacing_shift;
1da177e4 454 unsigned long sk_lingertime;
476e19cf 455 struct proto *sk_prot_creator;
1da177e4
LT
456 rwlock_t sk_callback_lock;
457 int sk_err,
458 sk_err_soft;
becb74f0
ED
459 u32 sk_ack_backlog;
460 u32 sk_max_ack_backlog;
86741ec2 461 kuid_t sk_uid;
109f6e39
EB
462 struct pid *sk_peer_pid;
463 const struct cred *sk_peer_cred;
1da177e4 464 long sk_rcvtimeo;
b7aa0bf7 465 ktime_t sk_stamp;
2bdc9211
DD
466#if BITS_PER_LONG==32
467 seqlock_t sk_stamp_seq;
468#endif
b9f40e21 469 u16 sk_tsflags;
fc64869c 470 u8 sk_shutdown;
09c2d251 471 u32 sk_tskey;
52267790 472 atomic_t sk_zckey;
1da177e4
LT
473 struct socket *sk_socket;
474 void *sk_user_data;
d5f64238 475#ifdef CONFIG_SECURITY
1da177e4 476 void *sk_security;
d5f64238 477#endif
2a56a1fe 478 struct sock_cgroup_data sk_cgrp_data;
baac50bb 479 struct mem_cgroup *sk_memcg;
1da177e4 480 void (*sk_state_change)(struct sock *sk);
676d2369 481 void (*sk_data_ready)(struct sock *sk);
1da177e4
LT
482 void (*sk_write_space)(struct sock *sk);
483 void (*sk_error_report)(struct sock *sk);
dc6b9b78
ED
484 int (*sk_backlog_rcv)(struct sock *sk,
485 struct sk_buff *skb);
1da177e4 486 void (*sk_destruct)(struct sock *sk);
ef456144 487 struct sock_reuseport __rcu *sk_reuseport_cb;
a4298e45 488 struct rcu_head sk_rcu;
1da177e4
LT
489};
490
218af599
ED
491enum sk_pacing {
492 SK_PACING_NONE = 0,
493 SK_PACING_NEEDED = 1,
494 SK_PACING_FQ = 2,
495};
496
559835ea
PS
497#define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
498
499#define rcu_dereference_sk_user_data(sk) rcu_dereference(__sk_user_data((sk)))
500#define rcu_assign_sk_user_data(sk, ptr) rcu_assign_pointer(__sk_user_data((sk)), ptr)
501
4a17fd52
PE
502/*
503 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
504 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
505 * on a socket means that the socket will reuse everybody else's port
506 * without looking at the other's sk_reuse value.
507 */
508
509#define SK_NO_REUSE 0
510#define SK_CAN_REUSE 1
511#define SK_FORCE_REUSE 2
512
627d2d6b 513int sk_set_peek_off(struct sock *sk, int val);
514
ef64a54f
PE
515static inline int sk_peek_offset(struct sock *sk, int flags)
516{
b9bb53f3 517 if (unlikely(flags & MSG_PEEK)) {
a0917e0b 518 return READ_ONCE(sk->sk_peek_off);
b9bb53f3
WB
519 }
520
521 return 0;
ef64a54f
PE
522}
523
524static inline void sk_peek_offset_bwd(struct sock *sk, int val)
525{
b9bb53f3
WB
526 s32 off = READ_ONCE(sk->sk_peek_off);
527
528 if (unlikely(off >= 0)) {
529 off = max_t(s32, off - val, 0);
530 WRITE_ONCE(sk->sk_peek_off, off);
ef64a54f
PE
531 }
532}
533
534static inline void sk_peek_offset_fwd(struct sock *sk, int val)
535{
b9bb53f3 536 sk_peek_offset_bwd(sk, -val);
ef64a54f
PE
537}
538
1da177e4
LT
539/*
540 * Hashed lists helper routines
541 */
c4146644
LZ
542static inline struct sock *sk_entry(const struct hlist_node *node)
543{
544 return hlist_entry(node, struct sock, sk_node);
545}
546
e48c414e 547static inline struct sock *__sk_head(const struct hlist_head *head)
1da177e4
LT
548{
549 return hlist_entry(head->first, struct sock, sk_node);
550}
551
e48c414e 552static inline struct sock *sk_head(const struct hlist_head *head)
1da177e4
LT
553{
554 return hlist_empty(head) ? NULL : __sk_head(head);
555}
556
88ab1932
ED
557static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
558{
559 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
560}
561
562static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
563{
564 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
565}
566
e48c414e 567static inline struct sock *sk_next(const struct sock *sk)
1da177e4 568{
6c59ebd3 569 return hlist_entry_safe(sk->sk_node.next, struct sock, sk_node);
1da177e4
LT
570}
571
88ab1932
ED
572static inline struct sock *sk_nulls_next(const struct sock *sk)
573{
574 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
575 hlist_nulls_entry(sk->sk_nulls_node.next,
576 struct sock, sk_nulls_node) :
577 NULL;
578}
579
dc6b9b78 580static inline bool sk_unhashed(const struct sock *sk)
1da177e4
LT
581{
582 return hlist_unhashed(&sk->sk_node);
583}
584
dc6b9b78 585static inline bool sk_hashed(const struct sock *sk)
1da177e4 586{
da753bea 587 return !sk_unhashed(sk);
1da177e4
LT
588}
589
dc6b9b78 590static inline void sk_node_init(struct hlist_node *node)
1da177e4
LT
591{
592 node->pprev = NULL;
593}
594
dc6b9b78 595static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
88ab1932
ED
596{
597 node->pprev = NULL;
598}
599
dc6b9b78 600static inline void __sk_del_node(struct sock *sk)
1da177e4
LT
601{
602 __hlist_del(&sk->sk_node);
603}
604
808f5114 605/* NB: equivalent to hlist_del_init_rcu */
dc6b9b78 606static inline bool __sk_del_node_init(struct sock *sk)
1da177e4
LT
607{
608 if (sk_hashed(sk)) {
609 __sk_del_node(sk);
610 sk_node_init(&sk->sk_node);
dc6b9b78 611 return true;
1da177e4 612 }
dc6b9b78 613 return false;
1da177e4
LT
614}
615
616/* Grab socket reference count. This operation is valid only
617 when sk is ALREADY grabbed f.e. it is found in hash table
618 or a list and the lookup is made under lock preventing hash table
619 modifications.
620 */
621
f9a7cbbf 622static __always_inline void sock_hold(struct sock *sk)
1da177e4 623{
41c6d650 624 refcount_inc(&sk->sk_refcnt);
1da177e4
LT
625}
626
627/* Ungrab socket in the context, which assumes that socket refcnt
628 cannot hit zero, f.e. it is true in context of any socketcall.
629 */
f9a7cbbf 630static __always_inline void __sock_put(struct sock *sk)
1da177e4 631{
41c6d650 632 refcount_dec(&sk->sk_refcnt);
1da177e4
LT
633}
634
dc6b9b78 635static inline bool sk_del_node_init(struct sock *sk)
1da177e4 636{
dc6b9b78 637 bool rc = __sk_del_node_init(sk);
1da177e4
LT
638
639 if (rc) {
640 /* paranoid for a while -acme */
41c6d650 641 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
1da177e4
LT
642 __sock_put(sk);
643 }
644 return rc;
645}
808f5114 646#define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
1da177e4 647
dc6b9b78 648static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
271b72c7
ED
649{
650 if (sk_hashed(sk)) {
88ab1932 651 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
dc6b9b78 652 return true;
271b72c7 653 }
dc6b9b78 654 return false;
271b72c7
ED
655}
656
dc6b9b78 657static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
271b72c7 658{
dc6b9b78 659 bool rc = __sk_nulls_del_node_init_rcu(sk);
271b72c7
ED
660
661 if (rc) {
662 /* paranoid for a while -acme */
41c6d650 663 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
271b72c7
ED
664 __sock_put(sk);
665 }
666 return rc;
667}
668
dc6b9b78 669static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
1da177e4
LT
670{
671 hlist_add_head(&sk->sk_node, list);
672}
673
dc6b9b78 674static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
1da177e4
LT
675{
676 sock_hold(sk);
677 __sk_add_node(sk, list);
678}
679
dc6b9b78 680static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
808f5114 681{
682 sock_hold(sk);
d296ba60
CG
683 if (IS_ENABLED(CONFIG_IPV6) && sk->sk_reuseport &&
684 sk->sk_family == AF_INET6)
685 hlist_add_tail_rcu(&sk->sk_node, list);
686 else
687 hlist_add_head_rcu(&sk->sk_node, list);
808f5114 688}
689
eb7f2cb5
MC
690static inline void sk_add_node_tail_rcu(struct sock *sk, struct hlist_head *list)
691{
692 sock_hold(sk);
693 hlist_add_tail_rcu(&sk->sk_node, list);
694}
695
dc6b9b78 696static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
271b72c7 697{
d7efc6c1 698 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
271b72c7
ED
699}
700
dc6b9b78 701static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
271b72c7
ED
702{
703 sock_hold(sk);
88ab1932 704 __sk_nulls_add_node_rcu(sk, list);
271b72c7
ED
705}
706
dc6b9b78 707static inline void __sk_del_bind_node(struct sock *sk)
1da177e4
LT
708{
709 __hlist_del(&sk->sk_bind_node);
710}
711
dc6b9b78 712static inline void sk_add_bind_node(struct sock *sk,
1da177e4
LT
713 struct hlist_head *list)
714{
715 hlist_add_head(&sk->sk_bind_node, list);
716}
717
b67bfe0d
SL
718#define sk_for_each(__sk, list) \
719 hlist_for_each_entry(__sk, list, sk_node)
720#define sk_for_each_rcu(__sk, list) \
721 hlist_for_each_entry_rcu(__sk, list, sk_node)
88ab1932
ED
722#define sk_nulls_for_each(__sk, node, list) \
723 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
724#define sk_nulls_for_each_rcu(__sk, node, list) \
725 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
b67bfe0d
SL
726#define sk_for_each_from(__sk) \
727 hlist_for_each_entry_from(__sk, sk_node)
88ab1932
ED
728#define sk_nulls_for_each_from(__sk, node) \
729 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
730 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
b67bfe0d
SL
731#define sk_for_each_safe(__sk, tmp, list) \
732 hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
733#define sk_for_each_bound(__sk, list) \
734 hlist_for_each_entry(__sk, list, sk_bind_node)
1da177e4 735
2dc41cff 736/**
ca065d0c 737 * sk_for_each_entry_offset_rcu - iterate over a list at a given struct offset
2dc41cff
DH
738 * @tpos: the type * to use as a loop cursor.
739 * @pos: the &struct hlist_node to use as a loop cursor.
740 * @head: the head for your list.
741 * @offset: offset of hlist_node within the struct.
742 *
743 */
ca065d0c 744#define sk_for_each_entry_offset_rcu(tpos, pos, head, offset) \
b6f4f848 745 for (pos = rcu_dereference(hlist_first_rcu(head)); \
ca065d0c 746 pos != NULL && \
2dc41cff 747 ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;}); \
b6f4f848 748 pos = rcu_dereference(hlist_next_rcu(pos)))
2dc41cff 749
c336d148
EB
750static inline struct user_namespace *sk_user_ns(struct sock *sk)
751{
752 /* Careful only use this in a context where these parameters
753 * can not change and must all be valid, such as recvmsg from
754 * userspace.
755 */
756 return sk->sk_socket->file->f_cred->user_ns;
757}
758
1da177e4
LT
759/* Sock flags */
760enum sock_flags {
761 SOCK_DEAD,
762 SOCK_DONE,
763 SOCK_URGINLINE,
764 SOCK_KEEPOPEN,
765 SOCK_LINGER,
766 SOCK_DESTROY,
767 SOCK_BROADCAST,
768 SOCK_TIMESTAMP,
769 SOCK_ZAPPED,
770 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
771 SOCK_DBG, /* %SO_DEBUG setting */
772 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
92f37fd2 773 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
1da177e4
LT
774 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
775 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
7cb02404 776 SOCK_MEMALLOC, /* VM depends on this socket for swapping */
20d49473 777 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
bcdce719 778 SOCK_FASYNC, /* fasync() active */
3b885787 779 SOCK_RXQ_OVFL,
1cdebb42 780 SOCK_ZEROCOPY, /* buffers from userspace */
6e3e939f 781 SOCK_WIFI_STATUS, /* push wifi status to userspace */
3bdc0eba
BG
782 SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
783 * Will use last 4 bytes of packet sent from
784 * user-space instead.
785 */
d59577b6 786 SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */
7d4c04fc 787 SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */
a4298e45 788 SOCK_RCU_FREE, /* wait rcu grace period in sk_destruct() */
1da177e4
LT
789};
790
01ce63c9
MRL
791#define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
792
53b924b3
RB
793static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
794{
795 nsk->sk_flags = osk->sk_flags;
796}
797
1da177e4
LT
798static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
799{
800 __set_bit(flag, &sk->sk_flags);
801}
802
803static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
804{
805 __clear_bit(flag, &sk->sk_flags);
806}
807
1b23a5df 808static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
1da177e4
LT
809{
810 return test_bit(flag, &sk->sk_flags);
811}
812
c93bdd0e
MG
813#ifdef CONFIG_NET
814extern struct static_key memalloc_socks;
815static inline int sk_memalloc_socks(void)
816{
817 return static_key_false(&memalloc_socks);
818}
819#else
820
821static inline int sk_memalloc_socks(void)
822{
823 return 0;
824}
825
826#endif
827
7450aaf6 828static inline gfp_t sk_gfp_mask(const struct sock *sk, gfp_t gfp_mask)
99a1dec7 829{
7450aaf6 830 return gfp_mask | (sk->sk_allocation & __GFP_MEMALLOC);
99a1dec7
MG
831}
832
1da177e4
LT
833static inline void sk_acceptq_removed(struct sock *sk)
834{
835 sk->sk_ack_backlog--;
836}
837
838static inline void sk_acceptq_added(struct sock *sk)
839{
840 sk->sk_ack_backlog++;
841}
842
dc6b9b78 843static inline bool sk_acceptq_is_full(const struct sock *sk)
1da177e4 844{
64a14651 845 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
1da177e4
LT
846}
847
848/*
849 * Compute minimal free write space needed to queue new packets.
850 */
dc6b9b78 851static inline int sk_stream_min_wspace(const struct sock *sk)
1da177e4 852{
8df09ea3 853 return sk->sk_wmem_queued >> 1;
1da177e4
LT
854}
855
dc6b9b78 856static inline int sk_stream_wspace(const struct sock *sk)
1da177e4
LT
857{
858 return sk->sk_sndbuf - sk->sk_wmem_queued;
859}
860
69336bd2 861void sk_stream_write_space(struct sock *sk);
1da177e4 862
8eae939f 863/* OOB backlog add */
a3a858ff 864static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
9ee6b535 865{
7fee226a 866 /* dont let skb dst not refcounted, we are going to leave rcu lock */
222d7dbd 867 skb_dst_force(skb);
7fee226a
ED
868
869 if (!sk->sk_backlog.tail)
870 sk->sk_backlog.head = skb;
871 else
9ee6b535 872 sk->sk_backlog.tail->next = skb;
7fee226a
ED
873
874 sk->sk_backlog.tail = skb;
9ee6b535
SH
875 skb->next = NULL;
876}
1da177e4 877
c377411f
ED
878/*
879 * Take into account size of receive queue and backlog queue
0fd7bac6
ED
880 * Do not take into account this skb truesize,
881 * to allow even a single big packet to come.
c377411f 882 */
274f482d 883static inline bool sk_rcvqueues_full(const struct sock *sk, unsigned int limit)
c377411f
ED
884{
885 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
886
f545a38f 887 return qsize > limit;
c377411f
ED
888}
889
8eae939f 890/* The per-socket spinlock must be held here. */
f545a38f
ED
891static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
892 unsigned int limit)
8eae939f 893{
274f482d 894 if (sk_rcvqueues_full(sk, limit))
8eae939f
ZY
895 return -ENOBUFS;
896
c7c49b8f
ED
897 /*
898 * If the skb was allocated from pfmemalloc reserves, only
899 * allow SOCK_MEMALLOC sockets to use it as this socket is
900 * helping free memory
901 */
902 if (skb_pfmemalloc(skb) && !sock_flag(sk, SOCK_MEMALLOC))
903 return -ENOMEM;
904
a3a858ff 905 __sk_add_backlog(sk, skb);
8eae939f
ZY
906 sk->sk_backlog.len += skb->truesize;
907 return 0;
908}
909
69336bd2 910int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
b4b9e355 911
c57943a1
PZ
912static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
913{
b4b9e355
MG
914 if (sk_memalloc_socks() && skb_pfmemalloc(skb))
915 return __sk_backlog_rcv(sk, skb);
916
c57943a1
PZ
917 return sk->sk_backlog_rcv(sk, skb);
918}
919
2c8c56e1
ED
920static inline void sk_incoming_cpu_update(struct sock *sk)
921{
34cfb542
PA
922 int cpu = raw_smp_processor_id();
923
f5f96dc2
ED
924 if (unlikely(READ_ONCE(sk->sk_incoming_cpu) != cpu))
925 WRITE_ONCE(sk->sk_incoming_cpu, cpu);
2c8c56e1
ED
926}
927
fe477558 928static inline void sock_rps_record_flow_hash(__u32 hash)
c58dc01b
DM
929{
930#ifdef CONFIG_RPS
931 struct rps_sock_flow_table *sock_flow_table;
932
933 rcu_read_lock();
934 sock_flow_table = rcu_dereference(rps_sock_flow_table);
fe477558 935 rps_record_sock_flow(sock_flow_table, hash);
c58dc01b
DM
936 rcu_read_unlock();
937#endif
938}
939
fe477558
TH
940static inline void sock_rps_record_flow(const struct sock *sk)
941{
c9d8ca04 942#ifdef CONFIG_RPS
13bfff25
ED
943 if (static_key_false(&rfs_needed)) {
944 /* Reading sk->sk_rxhash might incur an expensive cache line
945 * miss.
946 *
947 * TCP_ESTABLISHED does cover almost all states where RFS
948 * might be useful, and is cheaper [1] than testing :
949 * IPv4: inet_sk(sk)->inet_daddr
950 * IPv6: ipv6_addr_any(&sk->sk_v6_daddr)
951 * OR an additional socket flag
952 * [1] : sk_state and sk_prot are in the same cache line.
953 */
954 if (sk->sk_state == TCP_ESTABLISHED)
955 sock_rps_record_flow_hash(sk->sk_rxhash);
956 }
c9d8ca04 957#endif
fe477558
TH
958}
959
bdeab991
TH
960static inline void sock_rps_save_rxhash(struct sock *sk,
961 const struct sk_buff *skb)
c58dc01b
DM
962{
963#ifdef CONFIG_RPS
567e4b79 964 if (unlikely(sk->sk_rxhash != skb->hash))
61b905da 965 sk->sk_rxhash = skb->hash;
c58dc01b
DM
966#endif
967}
968
bdeab991
TH
969static inline void sock_rps_reset_rxhash(struct sock *sk)
970{
971#ifdef CONFIG_RPS
bdeab991
TH
972 sk->sk_rxhash = 0;
973#endif
974}
975
d9dc8b0f 976#define sk_wait_event(__sk, __timeo, __condition, __wait) \
cfcabdcc
SH
977 ({ int __rc; \
978 release_sock(__sk); \
979 __rc = __condition; \
980 if (!__rc) { \
d9dc8b0f
WC
981 *(__timeo) = wait_woken(__wait, \
982 TASK_INTERRUPTIBLE, \
983 *(__timeo)); \
cfcabdcc 984 } \
d9dc8b0f 985 sched_annotate_sleep(); \
cfcabdcc
SH
986 lock_sock(__sk); \
987 __rc = __condition; \
988 __rc; \
989 })
1da177e4 990
69336bd2
JP
991int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
992int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
993void sk_stream_wait_close(struct sock *sk, long timeo_p);
994int sk_stream_error(struct sock *sk, int flags, int err);
995void sk_stream_kill_queues(struct sock *sk);
996void sk_set_memalloc(struct sock *sk);
997void sk_clear_memalloc(struct sock *sk);
1da177e4 998
d41a69f1
ED
999void __sk_flush_backlog(struct sock *sk);
1000
1001static inline bool sk_flush_backlog(struct sock *sk)
1002{
1003 if (unlikely(READ_ONCE(sk->sk_backlog.tail))) {
1004 __sk_flush_backlog(sk);
1005 return true;
1006 }
1007 return false;
1008}
1009
dfbafc99 1010int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb);
1da177e4 1011
60236fdd 1012struct request_sock_ops;
6d6ee43e 1013struct timewait_sock_ops;
ab1e0a13 1014struct inet_hashinfo;
fc8717ba 1015struct raw_hashinfo;
f16a7dd5 1016struct smc_hashinfo;
de477254 1017struct module;
2e6599cb 1018
f77d6021 1019/*
5f0d5a3a 1020 * caches using SLAB_TYPESAFE_BY_RCU should let .next pointer from nulls nodes
f77d6021
ED
1021 * un-modified. Special care is taken when initializing object to zero.
1022 */
1023static inline void sk_prot_clear_nulls(struct sock *sk, int size)
1024{
1025 if (offsetof(struct sock, sk_node.next) != 0)
1026 memset(sk, 0, offsetof(struct sock, sk_node.next));
1027 memset(&sk->sk_node.pprev, 0,
1028 size - offsetof(struct sock, sk_node.pprev));
1029}
1030
1da177e4
LT
1031/* Networking protocol blocks we attach to sockets.
1032 * socket layer -> transport layer interface
1da177e4
LT
1033 */
1034struct proto {
dc6b9b78 1035 void (*close)(struct sock *sk,
1da177e4
LT
1036 long timeout);
1037 int (*connect)(struct sock *sk,
dc6b9b78 1038 struct sockaddr *uaddr,
1da177e4
LT
1039 int addr_len);
1040 int (*disconnect)(struct sock *sk, int flags);
1041
cdfbabfb
DH
1042 struct sock * (*accept)(struct sock *sk, int flags, int *err,
1043 bool kern);
1da177e4
LT
1044
1045 int (*ioctl)(struct sock *sk, int cmd,
1046 unsigned long arg);
1047 int (*init)(struct sock *sk);
7d06b2e0 1048 void (*destroy)(struct sock *sk);
1da177e4 1049 void (*shutdown)(struct sock *sk, int how);
dc6b9b78 1050 int (*setsockopt)(struct sock *sk, int level,
1da177e4 1051 int optname, char __user *optval,
b7058842 1052 unsigned int optlen);
dc6b9b78
ED
1053 int (*getsockopt)(struct sock *sk, int level,
1054 int optname, char __user *optval,
1055 int __user *option);
4b9d07a4 1056 void (*keepalive)(struct sock *sk, int valbool);
af01d537 1057#ifdef CONFIG_COMPAT
3fdadf7d
DM
1058 int (*compat_setsockopt)(struct sock *sk,
1059 int level,
1060 int optname, char __user *optval,
b7058842 1061 unsigned int optlen);
3fdadf7d
DM
1062 int (*compat_getsockopt)(struct sock *sk,
1063 int level,
1064 int optname, char __user *optval,
1065 int __user *option);
709b46e8
EB
1066 int (*compat_ioctl)(struct sock *sk,
1067 unsigned int cmd, unsigned long arg);
af01d537 1068#endif
1b784140
YX
1069 int (*sendmsg)(struct sock *sk, struct msghdr *msg,
1070 size_t len);
1071 int (*recvmsg)(struct sock *sk, struct msghdr *msg,
dc6b9b78
ED
1072 size_t len, int noblock, int flags,
1073 int *addr_len);
1da177e4
LT
1074 int (*sendpage)(struct sock *sk, struct page *page,
1075 int offset, size_t size, int flags);
dc6b9b78 1076 int (*bind)(struct sock *sk,
1da177e4
LT
1077 struct sockaddr *uaddr, int addr_len);
1078
dc6b9b78 1079 int (*backlog_rcv) (struct sock *sk,
1da177e4
LT
1080 struct sk_buff *skb);
1081
46d3ceab
ED
1082 void (*release_cb)(struct sock *sk);
1083
1da177e4 1084 /* Keeping track of sk's, looking them up, and port selection methods. */
086c653f 1085 int (*hash)(struct sock *sk);
1da177e4 1086 void (*unhash)(struct sock *sk);
719f8358 1087 void (*rehash)(struct sock *sk);
1da177e4
LT
1088 int (*get_port)(struct sock *sk, unsigned short snum);
1089
286ab3d4 1090 /* Keeping track of sockets in use */
65f76517 1091#ifdef CONFIG_PROC_FS
13ff3d6f 1092 unsigned int inuse_idx;
65f76517 1093#endif
ebb53d75 1094
c9bee3b7 1095 bool (*stream_memory_free)(const struct sock *sk);
1da177e4 1096 /* Memory pressure */
5c52ba17 1097 void (*enter_memory_pressure)(struct sock *sk);
06044751 1098 void (*leave_memory_pressure)(struct sock *sk);
8d987e5c 1099 atomic_long_t *memory_allocated; /* Current allocated memory. */
1748376b 1100 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
1da177e4
LT
1101 /*
1102 * Pressure flag: try to collapse.
1103 * Technical note: it is used by multiple contexts non atomically.
3ab224be 1104 * All the __sk_mem_schedule() is of this nature: accounting
1da177e4
LT
1105 * is strict, actions are advisory and have some latency.
1106 */
06044751 1107 unsigned long *memory_pressure;
8d987e5c 1108 long *sysctl_mem;
a3dcaf17 1109
1da177e4
LT
1110 int *sysctl_wmem;
1111 int *sysctl_rmem;
a3dcaf17
ED
1112 u32 sysctl_wmem_offset;
1113 u32 sysctl_rmem_offset;
1114
1da177e4 1115 int max_header;
7ba42910 1116 bool no_autobind;
1da177e4 1117
271b72c7 1118 struct kmem_cache *slab;
1da177e4 1119 unsigned int obj_size;
d50112ed 1120 slab_flags_t slab_flags;
1da177e4 1121
dd24c001 1122 struct percpu_counter *orphan_count;
8feaf0c0 1123
60236fdd 1124 struct request_sock_ops *rsk_prot;
6d6ee43e 1125 struct timewait_sock_ops *twsk_prot;
2e6599cb 1126
39d8cda7
PE
1127 union {
1128 struct inet_hashinfo *hashinfo;
645ca708 1129 struct udp_table *udp_table;
fc8717ba 1130 struct raw_hashinfo *raw_hash;
f16a7dd5 1131 struct smc_hashinfo *smc_hash;
39d8cda7 1132 } h;
ab1e0a13 1133
1da177e4
LT
1134 struct module *owner;
1135
1136 char name[32];
1137
1138 struct list_head node;
e6848976
ACM
1139#ifdef SOCK_REFCNT_DEBUG
1140 atomic_t socks;
e1aab161 1141#endif
64be0aed 1142 int (*diag_destroy)(struct sock *sk, int err);
3859a271 1143} __randomize_layout;
e1aab161 1144
69336bd2
JP
1145int proto_register(struct proto *prot, int alloc_slab);
1146void proto_unregister(struct proto *prot);
1da177e4 1147
e6848976
ACM
1148#ifdef SOCK_REFCNT_DEBUG
1149static inline void sk_refcnt_debug_inc(struct sock *sk)
1150{
1151 atomic_inc(&sk->sk_prot->socks);
1152}
1153
1154static inline void sk_refcnt_debug_dec(struct sock *sk)
1155{
1156 atomic_dec(&sk->sk_prot->socks);
1157 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
1158 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
1159}
1160
dec34fb0 1161static inline void sk_refcnt_debug_release(const struct sock *sk)
e6848976 1162{
41c6d650 1163 if (refcount_read(&sk->sk_refcnt) != 1)
e6848976 1164 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
41c6d650 1165 sk->sk_prot->name, sk, refcount_read(&sk->sk_refcnt));
e6848976
ACM
1166}
1167#else /* SOCK_REFCNT_DEBUG */
1168#define sk_refcnt_debug_inc(sk) do { } while (0)
1169#define sk_refcnt_debug_dec(sk) do { } while (0)
1170#define sk_refcnt_debug_release(sk) do { } while (0)
1171#endif /* SOCK_REFCNT_DEBUG */
1172
c9bee3b7
ED
1173static inline bool sk_stream_memory_free(const struct sock *sk)
1174{
1175 if (sk->sk_wmem_queued >= sk->sk_sndbuf)
1176 return false;
1177
1178 return sk->sk_prot->stream_memory_free ?
1179 sk->sk_prot->stream_memory_free(sk) : true;
1180}
1181
64dc6130
ED
1182static inline bool sk_stream_is_writeable(const struct sock *sk)
1183{
c9bee3b7
ED
1184 return sk_stream_wspace(sk) >= sk_stream_min_wspace(sk) &&
1185 sk_stream_memory_free(sk);
64dc6130 1186}
e1aab161 1187
54fd9c2d
DB
1188static inline int sk_under_cgroup_hierarchy(struct sock *sk,
1189 struct cgroup *ancestor)
1190{
1191#ifdef CONFIG_SOCK_CGROUP_DATA
1192 return cgroup_is_descendant(sock_cgroup_ptr(&sk->sk_cgrp_data),
1193 ancestor);
1194#else
1195 return -ENOTSUPP;
1196#endif
1197}
c9bee3b7 1198
180d8cd9
GC
1199static inline bool sk_has_memory_pressure(const struct sock *sk)
1200{
1201 return sk->sk_prot->memory_pressure != NULL;
1202}
1203
1204static inline bool sk_under_memory_pressure(const struct sock *sk)
1205{
1206 if (!sk->sk_prot->memory_pressure)
1207 return false;
e1aab161 1208
baac50bb
JW
1209 if (mem_cgroup_sockets_enabled && sk->sk_memcg &&
1210 mem_cgroup_under_socket_pressure(sk->sk_memcg))
e805605c 1211 return true;
e1aab161 1212
35b87f6c 1213 return !!*sk->sk_prot->memory_pressure;
180d8cd9
GC
1214}
1215
180d8cd9
GC
1216static inline long
1217sk_memory_allocated(const struct sock *sk)
1218{
e805605c 1219 return atomic_long_read(sk->sk_prot->memory_allocated);
180d8cd9
GC
1220}
1221
1222static inline long
e805605c 1223sk_memory_allocated_add(struct sock *sk, int amt)
180d8cd9 1224{
e805605c 1225 return atomic_long_add_return(amt, sk->sk_prot->memory_allocated);
180d8cd9
GC
1226}
1227
1228static inline void
0e90b31f 1229sk_memory_allocated_sub(struct sock *sk, int amt)
180d8cd9 1230{
e805605c 1231 atomic_long_sub(amt, sk->sk_prot->memory_allocated);
180d8cd9
GC
1232}
1233
1234static inline void sk_sockets_allocated_dec(struct sock *sk)
1235{
af95d7df 1236 percpu_counter_dec(sk->sk_prot->sockets_allocated);
180d8cd9
GC
1237}
1238
1239static inline void sk_sockets_allocated_inc(struct sock *sk)
1240{
af95d7df 1241 percpu_counter_inc(sk->sk_prot->sockets_allocated);
180d8cd9
GC
1242}
1243
1244static inline int
1245sk_sockets_allocated_read_positive(struct sock *sk)
1246{
af95d7df 1247 return percpu_counter_read_positive(sk->sk_prot->sockets_allocated);
180d8cd9
GC
1248}
1249
1250static inline int
1251proto_sockets_allocated_sum_positive(struct proto *prot)
1252{
1253 return percpu_counter_sum_positive(prot->sockets_allocated);
1254}
1255
1256static inline long
1257proto_memory_allocated(struct proto *prot)
1258{
1259 return atomic_long_read(prot->memory_allocated);
1260}
1261
1262static inline bool
1263proto_memory_pressure(struct proto *prot)
1264{
1265 if (!prot->memory_pressure)
1266 return false;
1267 return !!*prot->memory_pressure;
1268}
1269
65f76517
ED
1270
1271#ifdef CONFIG_PROC_FS
1da177e4 1272/* Called with local bh disabled */
69336bd2
JP
1273void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1274int sock_prot_inuse_get(struct net *net, struct proto *proto);
65f76517 1275#else
dc6b9b78 1276static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
c29a0bc4 1277 int inc)
65f76517
ED
1278{
1279}
65f76517
ED
1280#endif
1281
1da177e4 1282
614c6cb4
ACM
1283/* With per-bucket locks this operation is not-atomic, so that
1284 * this version is not worse.
1285 */
086c653f 1286static inline int __sk_prot_rehash(struct sock *sk)
614c6cb4
ACM
1287{
1288 sk->sk_prot->unhash(sk);
086c653f 1289 return sk->sk_prot->hash(sk);
614c6cb4
ACM
1290}
1291
1da177e4
LT
1292/* About 10 seconds */
1293#define SOCK_DESTROY_TIME (10*HZ)
1294
1295/* Sockets 0-1023 can't be bound to unless you are superuser */
1296#define PROT_SOCK 1024
1297
1298#define SHUTDOWN_MASK 3
1299#define RCV_SHUTDOWN 1
1300#define SEND_SHUTDOWN 2
1301
1302#define SOCK_SNDBUF_LOCK 1
1303#define SOCK_RCVBUF_LOCK 2
1304#define SOCK_BINDADDR_LOCK 4
1305#define SOCK_BINDPORT_LOCK 8
1306
1da177e4
LT
1307struct socket_alloc {
1308 struct socket socket;
1309 struct inode vfs_inode;
1310};
1311
1312static inline struct socket *SOCKET_I(struct inode *inode)
1313{
1314 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1315}
1316
1317static inline struct inode *SOCK_INODE(struct socket *socket)
1318{
1319 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1320}
1321
3ab224be
HA
1322/*
1323 * Functions for memory accounting
1324 */
f8c3bf00 1325int __sk_mem_raise_allocated(struct sock *sk, int size, int amt, int kind);
69336bd2 1326int __sk_mem_schedule(struct sock *sk, int size, int kind);
f8c3bf00 1327void __sk_mem_reduce_allocated(struct sock *sk, int amount);
1a24e04e 1328void __sk_mem_reclaim(struct sock *sk, int amount);
1da177e4 1329
bd68a2a8
ED
1330/* We used to have PAGE_SIZE here, but systems with 64KB pages
1331 * do not necessarily have 16x time more memory than 4KB ones.
1332 */
1333#define SK_MEM_QUANTUM 4096
3ab224be
HA
1334#define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1335#define SK_MEM_SEND 0
1336#define SK_MEM_RECV 1
1da177e4 1337
bd68a2a8
ED
1338/* sysctl_mem values are in pages, we convert them in SK_MEM_QUANTUM units */
1339static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1340{
1341 long val = sk->sk_prot->sysctl_mem[index];
1342
1343#if PAGE_SIZE > SK_MEM_QUANTUM
1344 val <<= PAGE_SHIFT - SK_MEM_QUANTUM_SHIFT;
1345#elif PAGE_SIZE < SK_MEM_QUANTUM
1346 val >>= SK_MEM_QUANTUM_SHIFT - PAGE_SHIFT;
1347#endif
1348 return val;
1349}
1350
3ab224be 1351static inline int sk_mem_pages(int amt)
1da177e4 1352{
3ab224be 1353 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
1da177e4
LT
1354}
1355
dc6b9b78 1356static inline bool sk_has_account(struct sock *sk)
1da177e4 1357{
3ab224be
HA
1358 /* return true if protocol supports memory accounting */
1359 return !!sk->sk_prot->memory_allocated;
1da177e4
LT
1360}
1361
dc6b9b78 1362static inline bool sk_wmem_schedule(struct sock *sk, int size)
1da177e4 1363{
3ab224be 1364 if (!sk_has_account(sk))
dc6b9b78 1365 return true;
3ab224be
HA
1366 return size <= sk->sk_forward_alloc ||
1367 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1da177e4
LT
1368}
1369
c76562b6 1370static inline bool
35c448a8 1371sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size)
d80d99d6 1372{
3ab224be 1373 if (!sk_has_account(sk))
dc6b9b78 1374 return true;
c76562b6
MG
1375 return size<= sk->sk_forward_alloc ||
1376 __sk_mem_schedule(sk, size, SK_MEM_RECV) ||
1377 skb_pfmemalloc(skb);
3ab224be
HA
1378}
1379
1380static inline void sk_mem_reclaim(struct sock *sk)
1381{
1382 if (!sk_has_account(sk))
1383 return;
1384 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1a24e04e 1385 __sk_mem_reclaim(sk, sk->sk_forward_alloc);
3ab224be
HA
1386}
1387
9993e7d3
DM
1388static inline void sk_mem_reclaim_partial(struct sock *sk)
1389{
1390 if (!sk_has_account(sk))
1391 return;
1392 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1a24e04e 1393 __sk_mem_reclaim(sk, sk->sk_forward_alloc - 1);
9993e7d3
DM
1394}
1395
3ab224be
HA
1396static inline void sk_mem_charge(struct sock *sk, int size)
1397{
1398 if (!sk_has_account(sk))
1399 return;
1400 sk->sk_forward_alloc -= size;
1401}
1402
1403static inline void sk_mem_uncharge(struct sock *sk, int size)
1404{
1405 if (!sk_has_account(sk))
1406 return;
1407 sk->sk_forward_alloc += size;
20c64d5c
ED
1408
1409 /* Avoid a possible overflow.
1410 * TCP send queues can make this happen, if sk_mem_reclaim()
1411 * is not called and more than 2 GBytes are released at once.
1412 *
1413 * If we reach 2 MBytes, reclaim 1 MBytes right now, there is
1414 * no need to hold that much forward allocation anyway.
1415 */
1416 if (unlikely(sk->sk_forward_alloc >= 1 << 21))
1417 __sk_mem_reclaim(sk, 1 << 20);
3ab224be
HA
1418}
1419
1420static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1421{
3ab224be
HA
1422 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1423 sk->sk_wmem_queued -= skb->truesize;
1424 sk_mem_uncharge(sk, skb->truesize);
1425 __kfree_skb(skb);
d80d99d6
HX
1426}
1427
c3f9b018
ED
1428static inline void sock_release_ownership(struct sock *sk)
1429{
61881cfb
HFS
1430 if (sk->sk_lock.owned) {
1431 sk->sk_lock.owned = 0;
1432
1433 /* The sk_lock has mutex_unlock() semantics: */
1434 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
1435 }
c3f9b018
ED
1436}
1437
ed07536e
PZ
1438/*
1439 * Macro so as to not evaluate some arguments when
1440 * lockdep is not enabled.
1441 *
1442 * Mark both the sk_lock and the sk_lock.slock as a
1443 * per-address-family lock class.
1444 */
dc6b9b78 1445#define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
ed07536e 1446do { \
e8f6fbf6 1447 sk->sk_lock.owned = 0; \
ed07536e
PZ
1448 init_waitqueue_head(&sk->sk_lock.wq); \
1449 spin_lock_init(&(sk)->sk_lock.slock); \
1450 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1451 sizeof((sk)->sk_lock)); \
1452 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
dc6b9b78 1453 (skey), (sname)); \
ed07536e
PZ
1454 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1455} while (0)
1456
b33b0a1b 1457#ifdef CONFIG_LOCKDEP
03be9822 1458static inline bool lockdep_sock_is_held(const struct sock *csk)
1e1d04e6
HFS
1459{
1460 struct sock *sk = (struct sock *)csk;
1461
1462 return lockdep_is_held(&sk->sk_lock) ||
1463 lockdep_is_held(&sk->sk_lock.slock);
1464}
b33b0a1b 1465#endif
1e1d04e6 1466
69336bd2 1467void lock_sock_nested(struct sock *sk, int subclass);
fcc70d5f
PZ
1468
1469static inline void lock_sock(struct sock *sk)
1470{
1471 lock_sock_nested(sk, 0);
1472}
1473
c4f66379 1474void __release_sock(struct sock *sk);
69336bd2 1475void release_sock(struct sock *sk);
1da177e4
LT
1476
1477/* BH context may only use the following locking interface. */
1478#define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
c6366184
IM
1479#define bh_lock_sock_nested(__sk) \
1480 spin_lock_nested(&((__sk)->sk_lock.slock), \
1481 SINGLE_DEPTH_NESTING)
1da177e4
LT
1482#define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1483
69336bd2 1484bool lock_sock_fast(struct sock *sk);
8a74ad60
ED
1485/**
1486 * unlock_sock_fast - complement of lock_sock_fast
1487 * @sk: socket
1488 * @slow: slow mode
1489 *
1490 * fast unlock socket for user context.
1491 * If slow mode is on, we call regular release_sock()
1492 */
1493static inline void unlock_sock_fast(struct sock *sk, bool slow)
4b0b72f7 1494{
8a74ad60
ED
1495 if (slow)
1496 release_sock(sk);
1497 else
1498 spin_unlock_bh(&sk->sk_lock.slock);
4b0b72f7
ED
1499}
1500
fafc4e1e
HFS
1501/* Used by processes to "lock" a socket state, so that
1502 * interrupts and bottom half handlers won't change it
1503 * from under us. It essentially blocks any incoming
1504 * packets, so that we won't get any new data or any
1505 * packets that change the state of the socket.
1506 *
1507 * While locked, BH processing will add new packets to
1508 * the backlog queue. This queue is processed by the
1509 * owner of the socket lock right before it is released.
1510 *
1511 * Since ~2.3.5 it is also exclusive sleep lock serializing
1512 * accesses from user process context.
1513 */
1514
46cc6e49 1515static inline void sock_owned_by_me(const struct sock *sk)
fafc4e1e
HFS
1516{
1517#ifdef CONFIG_LOCKDEP
5e91f6ce 1518 WARN_ON_ONCE(!lockdep_sock_is_held(sk) && debug_locks);
fafc4e1e 1519#endif
46cc6e49
ED
1520}
1521
1522static inline bool sock_owned_by_user(const struct sock *sk)
1523{
1524 sock_owned_by_me(sk);
fafc4e1e
HFS
1525 return sk->sk_lock.owned;
1526}
1527
602f7a27
TH
1528static inline bool sock_owned_by_user_nocheck(const struct sock *sk)
1529{
1530 return sk->sk_lock.owned;
1531}
1532
fafc4e1e
HFS
1533/* no reclassification while locks are held */
1534static inline bool sock_allow_reclassification(const struct sock *csk)
1535{
1536 struct sock *sk = (struct sock *)csk;
1537
1538 return !sk->sk_lock.owned && !spin_is_locked(&sk->sk_lock.slock);
1539}
4b0b72f7 1540
69336bd2 1541struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
11aa9c28 1542 struct proto *prot, int kern);
69336bd2 1543void sk_free(struct sock *sk);
eb4cb008 1544void sk_destruct(struct sock *sk);
69336bd2 1545struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority);
94352d45 1546void sk_free_unlock_clone(struct sock *sk);
69336bd2
JP
1547
1548struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
1549 gfp_t priority);
1d2077ac 1550void __sock_wfree(struct sk_buff *skb);
69336bd2 1551void sock_wfree(struct sk_buff *skb);
98ba0bd5
WB
1552struct sk_buff *sock_omalloc(struct sock *sk, unsigned long size,
1553 gfp_t priority);
69336bd2
JP
1554void skb_orphan_partial(struct sk_buff *skb);
1555void sock_rfree(struct sk_buff *skb);
62bccb8c 1556void sock_efree(struct sk_buff *skb);
82eabd9e 1557#ifdef CONFIG_INET
69336bd2 1558void sock_edemux(struct sk_buff *skb);
82eabd9e 1559#else
158f323b 1560#define sock_edemux sock_efree
82eabd9e 1561#endif
69336bd2
JP
1562
1563int sock_setsockopt(struct socket *sock, int level, int op,
1564 char __user *optval, unsigned int optlen);
1565
1566int sock_getsockopt(struct socket *sock, int level, int op,
1567 char __user *optval, int __user *optlen);
1568struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1569 int noblock, int *errcode);
1570struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1571 unsigned long data_len, int noblock,
1572 int *errcode, int max_page_order);
1573void *sock_kmalloc(struct sock *sk, int size, gfp_t priority);
1574void sock_kfree_s(struct sock *sk, void *mem, int size);
79e88659 1575void sock_kzfree_s(struct sock *sk, void *mem, int size);
69336bd2 1576void sk_send_sigurg(struct sock *sk);
1da177e4 1577
f28ea365
EJ
1578struct sockcm_cookie {
1579 u32 mark;
3dd17e63 1580 u16 tsflags;
f28ea365
EJ
1581};
1582
39771b12
WB
1583int __sock_cmsg_send(struct sock *sk, struct msghdr *msg, struct cmsghdr *cmsg,
1584 struct sockcm_cookie *sockc);
f28ea365
EJ
1585int sock_cmsg_send(struct sock *sk, struct msghdr *msg,
1586 struct sockcm_cookie *sockc);
1587
1da177e4
LT
1588/*
1589 * Functions to fill in entries in struct proto_ops when a protocol
1590 * does not implement a particular function.
1591 */
69336bd2
JP
1592int sock_no_bind(struct socket *, struct sockaddr *, int);
1593int sock_no_connect(struct socket *, struct sockaddr *, int, int);
1594int sock_no_socketpair(struct socket *, struct socket *);
cdfbabfb 1595int sock_no_accept(struct socket *, struct socket *, int, bool);
69336bd2
JP
1596int sock_no_getname(struct socket *, struct sockaddr *, int *, int);
1597unsigned int sock_no_poll(struct file *, struct socket *,
1598 struct poll_table_struct *);
1599int sock_no_ioctl(struct socket *, unsigned int, unsigned long);
1600int sock_no_listen(struct socket *, int);
1601int sock_no_shutdown(struct socket *, int);
1602int sock_no_getsockopt(struct socket *, int , int, char __user *, int __user *);
1603int sock_no_setsockopt(struct socket *, int, int, char __user *, unsigned int);
1b784140 1604int sock_no_sendmsg(struct socket *, struct msghdr *, size_t);
306b13eb 1605int sock_no_sendmsg_locked(struct sock *sk, struct msghdr *msg, size_t len);
1b784140 1606int sock_no_recvmsg(struct socket *, struct msghdr *, size_t, int);
69336bd2
JP
1607int sock_no_mmap(struct file *file, struct socket *sock,
1608 struct vm_area_struct *vma);
1609ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset,
1610 size_t size, int flags);
306b13eb
TH
1611ssize_t sock_no_sendpage_locked(struct sock *sk, struct page *page,
1612 int offset, size_t size, int flags);
1da177e4
LT
1613
1614/*
1615 * Functions to fill in entries in struct proto_ops when a protocol
1616 * uses the inet style.
1617 */
69336bd2 1618int sock_common_getsockopt(struct socket *sock, int level, int optname,
1da177e4 1619 char __user *optval, int __user *optlen);
1b784140
YX
1620int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
1621 int flags);
69336bd2 1622int sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 1623 char __user *optval, unsigned int optlen);
69336bd2 1624int compat_sock_common_getsockopt(struct socket *sock, int level,
3fdadf7d 1625 int optname, char __user *optval, int __user *optlen);
69336bd2 1626int compat_sock_common_setsockopt(struct socket *sock, int level,
b7058842 1627 int optname, char __user *optval, unsigned int optlen);
1da177e4 1628
69336bd2 1629void sk_common_release(struct sock *sk);
1da177e4
LT
1630
1631/*
1632 * Default socket callbacks and setup code
1633 */
dc6b9b78 1634
1da177e4 1635/* Initialise core socket variables */
69336bd2 1636void sock_init_data(struct socket *sock, struct sock *sk);
1da177e4 1637
1da177e4
LT
1638/*
1639 * Socket reference counting postulates.
1640 *
1641 * * Each user of socket SHOULD hold a reference count.
1642 * * Each access point to socket (an hash table bucket, reference from a list,
1643 * running timer, skb in flight MUST hold a reference count.
1644 * * When reference count hits 0, it means it will never increase back.
1645 * * When reference count hits 0, it means that no references from
1646 * outside exist to this socket and current process on current CPU
1647 * is last user and may/should destroy this socket.
1648 * * sk_free is called from any context: process, BH, IRQ. When
1649 * it is called, socket has no references from outside -> sk_free
1650 * may release descendant resources allocated by the socket, but
1651 * to the time when it is called, socket is NOT referenced by any
1652 * hash tables, lists etc.
1653 * * Packets, delivered from outside (from network or from another process)
1654 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1655 * when they sit in queue. Otherwise, packets will leak to hole, when
1656 * socket is looked up by one cpu and unhasing is made by another CPU.
1657 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1658 * (leak to backlog). Packet socket does all the processing inside
1659 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1660 * use separate SMP lock, so that they are prone too.
1661 */
1662
1663/* Ungrab socket and destroy it, if it was the last reference. */
1664static inline void sock_put(struct sock *sk)
1665{
41c6d650 1666 if (refcount_dec_and_test(&sk->sk_refcnt))
1da177e4
LT
1667 sk_free(sk);
1668}
05dbc7b5 1669/* Generic version of sock_put(), dealing with all sockets
41b822c5 1670 * (TCP_TIMEWAIT, TCP_NEW_SYN_RECV, ESTABLISHED...)
05dbc7b5
ED
1671 */
1672void sock_gen_put(struct sock *sk);
1da177e4 1673
4f0c40d9 1674int __sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested,
c3f24cfb 1675 unsigned int trim_cap, bool refcounted);
4f0c40d9
WB
1676static inline int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1677 const int nested)
1678{
c3f24cfb 1679 return __sk_receive_skb(sk, skb, nested, 1, true);
4f0c40d9 1680}
25995ff5 1681
e022f0b4
KK
1682static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1683{
1684 sk->sk_tx_queue_mapping = tx_queue;
1685}
1686
1687static inline void sk_tx_queue_clear(struct sock *sk)
1688{
1689 sk->sk_tx_queue_mapping = -1;
1690}
1691
1692static inline int sk_tx_queue_get(const struct sock *sk)
1693{
b0f77d0e 1694 return sk ? sk->sk_tx_queue_mapping : -1;
e022f0b4
KK
1695}
1696
972692e0
DM
1697static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1698{
e022f0b4 1699 sk_tx_queue_clear(sk);
972692e0
DM
1700 sk->sk_socket = sock;
1701}
1702
aa395145
ED
1703static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1704{
eaefd110
ED
1705 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1706 return &rcu_dereference_raw(sk->sk_wq)->wait;
aa395145 1707}
1da177e4
LT
1708/* Detach socket from process context.
1709 * Announce socket dead, detach it from wait queue and inode.
1710 * Note that parent inode held reference count on this struct sock,
1711 * we do not release it in this function, because protocol
1712 * probably wants some additional cleanups or even continuing
1713 * to work with this socket (TCP).
1714 */
1715static inline void sock_orphan(struct sock *sk)
1716{
1717 write_lock_bh(&sk->sk_callback_lock);
1718 sock_set_flag(sk, SOCK_DEAD);
972692e0 1719 sk_set_socket(sk, NULL);
43815482 1720 sk->sk_wq = NULL;
1da177e4
LT
1721 write_unlock_bh(&sk->sk_callback_lock);
1722}
1723
1724static inline void sock_graft(struct sock *sk, struct socket *parent)
1725{
0ffdaf5b 1726 WARN_ON(parent->sk);
1da177e4 1727 write_lock_bh(&sk->sk_callback_lock);
eaefd110 1728 sk->sk_wq = parent->wq;
1da177e4 1729 parent->sk = sk;
972692e0 1730 sk_set_socket(sk, parent);
86741ec2 1731 sk->sk_uid = SOCK_INODE(parent)->i_uid;
4237c75c 1732 security_sock_graft(sk, parent);
1da177e4
LT
1733 write_unlock_bh(&sk->sk_callback_lock);
1734}
1735
69336bd2
JP
1736kuid_t sock_i_uid(struct sock *sk);
1737unsigned long sock_i_ino(struct sock *sk);
1da177e4 1738
86741ec2
LC
1739static inline kuid_t sock_net_uid(const struct net *net, const struct sock *sk)
1740{
1741 return sk ? sk->sk_uid : make_kuid(net->user_ns, 0);
1742}
1743
58d607d3 1744static inline u32 net_tx_rndhash(void)
877d1f62 1745{
58d607d3
ED
1746 u32 v = prandom_u32();
1747
1748 return v ?: 1;
1749}
877d1f62 1750
58d607d3
ED
1751static inline void sk_set_txhash(struct sock *sk)
1752{
1753 sk->sk_txhash = net_tx_rndhash();
877d1f62
TH
1754}
1755
265f94ff
TH
1756static inline void sk_rethink_txhash(struct sock *sk)
1757{
1758 if (sk->sk_txhash)
1759 sk_set_txhash(sk);
1760}
1761
1da177e4
LT
1762static inline struct dst_entry *
1763__sk_dst_get(struct sock *sk)
1764{
1e1d04e6
HFS
1765 return rcu_dereference_check(sk->sk_dst_cache,
1766 lockdep_sock_is_held(sk));
1da177e4
LT
1767}
1768
1769static inline struct dst_entry *
1770sk_dst_get(struct sock *sk)
1771{
1772 struct dst_entry *dst;
1773
b6c6712a
ED
1774 rcu_read_lock();
1775 dst = rcu_dereference(sk->sk_dst_cache);
f8864972
ED
1776 if (dst && !atomic_inc_not_zero(&dst->__refcnt))
1777 dst = NULL;
b6c6712a 1778 rcu_read_unlock();
1da177e4
LT
1779 return dst;
1780}
1781
b6c6712a
ED
1782static inline void dst_negative_advice(struct sock *sk)
1783{
1784 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1785
265f94ff
TH
1786 sk_rethink_txhash(sk);
1787
b6c6712a
ED
1788 if (dst && dst->ops->negative_advice) {
1789 ndst = dst->ops->negative_advice(dst);
1790
1791 if (ndst != dst) {
1792 rcu_assign_pointer(sk->sk_dst_cache, ndst);
0a6957e7 1793 sk_tx_queue_clear(sk);
9b8805a3 1794 sk->sk_dst_pending_confirm = 0;
b6c6712a
ED
1795 }
1796 }
1797}
1798
1da177e4
LT
1799static inline void
1800__sk_dst_set(struct sock *sk, struct dst_entry *dst)
1801{
1802 struct dst_entry *old_dst;
1803
e022f0b4 1804 sk_tx_queue_clear(sk);
9b8805a3 1805 sk->sk_dst_pending_confirm = 0;
95964c6d
ED
1806 old_dst = rcu_dereference_protected(sk->sk_dst_cache,
1807 lockdep_sock_is_held(sk));
b6c6712a 1808 rcu_assign_pointer(sk->sk_dst_cache, dst);
1da177e4
LT
1809 dst_release(old_dst);
1810}
1811
1812static inline void
1813sk_dst_set(struct sock *sk, struct dst_entry *dst)
1814{
7f502361
ED
1815 struct dst_entry *old_dst;
1816
1817 sk_tx_queue_clear(sk);
9b8805a3 1818 sk->sk_dst_pending_confirm = 0;
5925a055 1819 old_dst = xchg((__force struct dst_entry **)&sk->sk_dst_cache, dst);
7f502361 1820 dst_release(old_dst);
1da177e4
LT
1821}
1822
1823static inline void
1824__sk_dst_reset(struct sock *sk)
1825{
b6c6712a 1826 __sk_dst_set(sk, NULL);
1da177e4
LT
1827}
1828
1829static inline void
1830sk_dst_reset(struct sock *sk)
1831{
7f502361 1832 sk_dst_set(sk, NULL);
1da177e4
LT
1833}
1834
69336bd2 1835struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1da177e4 1836
69336bd2 1837struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1da177e4 1838
9b8805a3
JA
1839static inline void sk_dst_confirm(struct sock *sk)
1840{
1841 if (!sk->sk_dst_pending_confirm)
1842 sk->sk_dst_pending_confirm = 1;
1843}
1844
4ff06203
JA
1845static inline void sock_confirm_neigh(struct sk_buff *skb, struct neighbour *n)
1846{
1847 if (skb_get_dst_pending_confirm(skb)) {
1848 struct sock *sk = skb->sk;
1849 unsigned long now = jiffies;
1850
1851 /* avoid dirtying neighbour */
1852 if (n->confirmed != now)
1853 n->confirmed = now;
1854 if (sk && sk->sk_dst_pending_confirm)
1855 sk->sk_dst_pending_confirm = 0;
1856 }
1857}
1858
f60e5990 1859bool sk_mc_loop(struct sock *sk);
1860
dc6b9b78 1861static inline bool sk_can_gso(const struct sock *sk)
bcd76111
HX
1862{
1863 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1864}
1865
69336bd2 1866void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
6cbb0df7 1867
c8f44aff 1868static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
a465419b
ED
1869{
1870 sk->sk_route_nocaps |= flags;
1871 sk->sk_route_caps &= ~flags;
1872}
1873
9a49850d
TH
1874static inline bool sk_check_csum_caps(struct sock *sk)
1875{
1876 return (sk->sk_route_caps & NETIF_F_HW_CSUM) ||
1877 (sk->sk_family == PF_INET &&
1878 (sk->sk_route_caps & NETIF_F_IP_CSUM)) ||
1879 (sk->sk_family == PF_INET6 &&
1880 (sk->sk_route_caps & NETIF_F_IPV6_CSUM));
1881}
1882
c6e1a0d1 1883static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
57be5bda 1884 struct iov_iter *from, char *to,
912d398d 1885 int copy, int offset)
c6e1a0d1
TH
1886{
1887 if (skb->ip_summed == CHECKSUM_NONE) {
57be5bda 1888 __wsum csum = 0;
15e6cb46 1889 if (!csum_and_copy_from_iter_full(to, copy, &csum, from))
57be5bda 1890 return -EFAULT;
912d398d 1891 skb->csum = csum_block_add(skb->csum, csum, offset);
c6e1a0d1 1892 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
15e6cb46 1893 if (!copy_from_iter_full_nocache(to, copy, from))
c6e1a0d1 1894 return -EFAULT;
15e6cb46 1895 } else if (!copy_from_iter_full(to, copy, from))
c6e1a0d1
TH
1896 return -EFAULT;
1897
1898 return 0;
1899}
1900
1901static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
57be5bda 1902 struct iov_iter *from, int copy)
c6e1a0d1 1903{
912d398d 1904 int err, offset = skb->len;
c6e1a0d1 1905
912d398d
WY
1906 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1907 copy, offset);
c6e1a0d1 1908 if (err)
912d398d 1909 __skb_trim(skb, offset);
c6e1a0d1
TH
1910
1911 return err;
1912}
1913
57be5bda 1914static inline int skb_copy_to_page_nocache(struct sock *sk, struct iov_iter *from,
c6e1a0d1
TH
1915 struct sk_buff *skb,
1916 struct page *page,
1917 int off, int copy)
1918{
1919 int err;
1920
912d398d
WY
1921 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1922 copy, skb->len);
c6e1a0d1
TH
1923 if (err)
1924 return err;
1925
1926 skb->len += copy;
1927 skb->data_len += copy;
1928 skb->truesize += copy;
1929 sk->sk_wmem_queued += copy;
1930 sk_mem_charge(sk, copy);
1931 return 0;
1932}
1933
c564039f
ED
1934/**
1935 * sk_wmem_alloc_get - returns write allocations
1936 * @sk: socket
1937 *
1938 * Returns sk_wmem_alloc minus initial offset of one
1939 */
1940static inline int sk_wmem_alloc_get(const struct sock *sk)
1941{
14afee4b 1942 return refcount_read(&sk->sk_wmem_alloc) - 1;
c564039f
ED
1943}
1944
1945/**
1946 * sk_rmem_alloc_get - returns read allocations
1947 * @sk: socket
1948 *
1949 * Returns sk_rmem_alloc
1950 */
1951static inline int sk_rmem_alloc_get(const struct sock *sk)
1952{
1953 return atomic_read(&sk->sk_rmem_alloc);
1954}
1955
1956/**
1957 * sk_has_allocations - check if allocations are outstanding
1958 * @sk: socket
1959 *
1960 * Returns true if socket has write or read allocations
1961 */
dc6b9b78 1962static inline bool sk_has_allocations(const struct sock *sk)
c564039f
ED
1963{
1964 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1965}
1966
a57de0b4 1967/**
1ce0bf50 1968 * skwq_has_sleeper - check if there are any waiting processes
acfbe96a 1969 * @wq: struct socket_wq
a57de0b4 1970 *
43815482 1971 * Returns true if socket_wq has waiting processes
a57de0b4 1972 *
1ce0bf50 1973 * The purpose of the skwq_has_sleeper and sock_poll_wait is to wrap the memory
a57de0b4
JO
1974 * barrier call. They were added due to the race found within the tcp code.
1975 *
d651983d 1976 * Consider following tcp code paths::
a57de0b4 1977 *
d651983d
MCC
1978 * CPU1 CPU2
1979 * sys_select receive packet
a57de0b4
JO
1980 * ... ...
1981 * __add_wait_queue update tp->rcv_nxt
1982 * ... ...
1983 * tp->rcv_nxt check sock_def_readable
1984 * ... {
43815482
ED
1985 * schedule rcu_read_lock();
1986 * wq = rcu_dereference(sk->sk_wq);
1987 * if (wq && waitqueue_active(&wq->wait))
1988 * wake_up_interruptible(&wq->wait)
a57de0b4
JO
1989 * ...
1990 * }
1991 *
1992 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1993 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1994 * could then endup calling schedule and sleep forever if there are no more
1995 * data on the socket.
ad462769 1996 *
a57de0b4 1997 */
1ce0bf50 1998static inline bool skwq_has_sleeper(struct socket_wq *wq)
a57de0b4 1999{
1ce0bf50 2000 return wq && wq_has_sleeper(&wq->wait);
a57de0b4
JO
2001}
2002
2003/**
2004 * sock_poll_wait - place memory barrier behind the poll_wait call.
2005 * @filp: file
2006 * @wait_address: socket wait queue
2007 * @p: poll_table
2008 *
43815482 2009 * See the comments in the wq_has_sleeper function.
a57de0b4
JO
2010 */
2011static inline void sock_poll_wait(struct file *filp,
2012 wait_queue_head_t *wait_address, poll_table *p)
2013{
626cf236 2014 if (!poll_does_not_wait(p) && wait_address) {
a57de0b4 2015 poll_wait(filp, wait_address, p);
dc6b9b78 2016 /* We need to be sure we are in sync with the
a57de0b4
JO
2017 * socket flags modification.
2018 *
43815482 2019 * This memory barrier is paired in the wq_has_sleeper.
dc6b9b78 2020 */
a57de0b4
JO
2021 smp_mb();
2022 }
2023}
2024
b73c3d0e
TH
2025static inline void skb_set_hash_from_sk(struct sk_buff *skb, struct sock *sk)
2026{
2027 if (sk->sk_txhash) {
2028 skb->l4_hash = 1;
2029 skb->hash = sk->sk_txhash;
2030 }
2031}
2032
9e17f8a4
ED
2033void skb_set_owner_w(struct sk_buff *skb, struct sock *sk);
2034
1da177e4 2035/*
dc6b9b78 2036 * Queue a received datagram if it will fit. Stream and sequenced
1da177e4
LT
2037 * protocols can't normally use this as they need to fit buffers in
2038 * and play with them.
2039 *
dc6b9b78 2040 * Inlined as it's very short and called for pretty much every
1da177e4
LT
2041 * packet ever received.
2042 */
1da177e4
LT
2043static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
2044{
d55d87fd 2045 skb_orphan(skb);
1da177e4
LT
2046 skb->sk = sk;
2047 skb->destructor = sock_rfree;
2048 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
3ab224be 2049 sk_mem_charge(sk, skb->truesize);
1da177e4
LT
2050}
2051
69336bd2
JP
2052void sk_reset_timer(struct sock *sk, struct timer_list *timer,
2053 unsigned long expires);
1da177e4 2054
69336bd2 2055void sk_stop_timer(struct sock *sk, struct timer_list *timer);
1da177e4 2056
65101aec
PA
2057int __sk_queue_drop_skb(struct sock *sk, struct sk_buff_head *sk_queue,
2058 struct sk_buff *skb, unsigned int flags,
69629464
ED
2059 void (*destructor)(struct sock *sk,
2060 struct sk_buff *skb));
e6afc8ac 2061int __sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
69336bd2 2062int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
1da177e4 2063
69336bd2 2064int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
364a9e93 2065struct sk_buff *sock_dequeue_err_skb(struct sock *sk);
1da177e4
LT
2066
2067/*
2068 * Recover an error report and clear atomically
2069 */
dc6b9b78 2070
1da177e4
LT
2071static inline int sock_error(struct sock *sk)
2072{
c1cbe4b7
BL
2073 int err;
2074 if (likely(!sk->sk_err))
2075 return 0;
2076 err = xchg(&sk->sk_err, 0);
1da177e4
LT
2077 return -err;
2078}
2079
2080static inline unsigned long sock_wspace(struct sock *sk)
2081{
2082 int amt = 0;
2083
2084 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
14afee4b 2085 amt = sk->sk_sndbuf - refcount_read(&sk->sk_wmem_alloc);
dc6b9b78 2086 if (amt < 0)
1da177e4
LT
2087 amt = 0;
2088 }
2089 return amt;
2090}
2091
ceb5d58b
ED
2092/* Note:
2093 * We use sk->sk_wq_raw, from contexts knowing this
2094 * pointer is not NULL and cannot disappear/change.
2095 */
9cd3e072 2096static inline void sk_set_bit(int nr, struct sock *sk)
1da177e4 2097{
4be73522
ED
2098 if ((nr == SOCKWQ_ASYNC_NOSPACE || nr == SOCKWQ_ASYNC_WAITDATA) &&
2099 !sock_flag(sk, SOCK_FASYNC))
9317bb69
ED
2100 return;
2101
ceb5d58b 2102 set_bit(nr, &sk->sk_wq_raw->flags);
9cd3e072
ED
2103}
2104
2105static inline void sk_clear_bit(int nr, struct sock *sk)
2106{
4be73522
ED
2107 if ((nr == SOCKWQ_ASYNC_NOSPACE || nr == SOCKWQ_ASYNC_WAITDATA) &&
2108 !sock_flag(sk, SOCK_FASYNC))
9317bb69
ED
2109 return;
2110
ceb5d58b 2111 clear_bit(nr, &sk->sk_wq_raw->flags);
9cd3e072
ED
2112}
2113
ceb5d58b 2114static inline void sk_wake_async(const struct sock *sk, int how, int band)
1da177e4 2115{
ceb5d58b
ED
2116 if (sock_flag(sk, SOCK_FASYNC)) {
2117 rcu_read_lock();
2118 sock_wake_async(rcu_dereference(sk->sk_wq), how, band);
2119 rcu_read_unlock();
2120 }
1da177e4
LT
2121}
2122
eea86af6
DB
2123/* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
2124 * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
2125 * Note: for send buffers, TCP works better if we can build two skbs at
2126 * minimum.
7a91b434 2127 */
9eb5bf83 2128#define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
eea86af6
DB
2129
2130#define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2)
2131#define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE
1da177e4
LT
2132
2133static inline void sk_stream_moderate_sndbuf(struct sock *sk)
2134{
2135 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
8df09ea3 2136 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
eea86af6 2137 sk->sk_sndbuf = max_t(u32, sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1da177e4
LT
2138 }
2139}
2140
eb934478
ED
2141struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp,
2142 bool force_schedule);
1da177e4 2143
5640f768
ED
2144/**
2145 * sk_page_frag - return an appropriate page_frag
2146 * @sk: socket
2147 *
f92bd815
TH
2148 * Use the per task page_frag instead of the per socket one for
2149 * optimization when we know that we're in the normal context and owns
2150 * everything that's associated with %current.
2151 *
2152 * gfpflags_allow_blocking() isn't enough here as direct reclaim may nest
2153 * inside other socket operations and end up recursing into sk_page_frag()
2154 * while it's already in use.
5640f768
ED
2155 */
2156static inline struct page_frag *sk_page_frag(struct sock *sk)
1da177e4 2157{
f92bd815 2158 if (gfpflags_normal_context(sk->sk_allocation))
5640f768 2159 return &current->task_frag;
1da177e4 2160
5640f768 2161 return &sk->sk_frag;
1da177e4
LT
2162}
2163
69336bd2 2164bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag);
5640f768 2165
1da177e4
LT
2166/*
2167 * Default write policy as shown to user space via poll/select/SIGIO
2168 */
dc6b9b78 2169static inline bool sock_writeable(const struct sock *sk)
1da177e4 2170{
14afee4b 2171 return refcount_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
1da177e4
LT
2172}
2173
dd0fc66f 2174static inline gfp_t gfp_any(void)
1da177e4 2175{
99709372 2176 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1da177e4
LT
2177}
2178
dc6b9b78 2179static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
1da177e4
LT
2180{
2181 return noblock ? 0 : sk->sk_rcvtimeo;
2182}
2183
dc6b9b78 2184static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
1da177e4
LT
2185{
2186 return noblock ? 0 : sk->sk_sndtimeo;
2187}
2188
2189static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2190{
2191 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
2192}
2193
2194/* Alas, with timeout socket operations are not restartable.
2195 * Compare this to poll().
2196 */
2197static inline int sock_intr_errno(long timeo)
2198{
2199 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2200}
2201
744d5a3e
EB
2202struct sock_skb_cb {
2203 u32 dropcount;
2204};
2205
2206/* Store sock_skb_cb at the end of skb->cb[] so protocol families
2207 * using skb->cb[] would keep using it directly and utilize its
2208 * alignement guarantee.
2209 */
2210#define SOCK_SKB_CB_OFFSET ((FIELD_SIZEOF(struct sk_buff, cb) - \
2211 sizeof(struct sock_skb_cb)))
2212
2213#define SOCK_SKB_CB(__skb) ((struct sock_skb_cb *)((__skb)->cb + \
2214 SOCK_SKB_CB_OFFSET))
2215
b4772ef8 2216#define sock_skb_cb_check_size(size) \
744d5a3e 2217 BUILD_BUG_ON((size) > SOCK_SKB_CB_OFFSET)
b4772ef8 2218
3bc3b96f
EB
2219static inline void
2220sock_skb_set_dropcount(const struct sock *sk, struct sk_buff *skb)
2221{
3665f381
ED
2222 SOCK_SKB_CB(skb)->dropcount = sock_flag(sk, SOCK_RXQ_OVFL) ?
2223 atomic_read(&sk->sk_drops) : 0;
3bc3b96f
EB
2224}
2225
532182cd
ED
2226static inline void sk_drops_add(struct sock *sk, const struct sk_buff *skb)
2227{
2228 int segs = max_t(u16, 1, skb_shinfo(skb)->gso_segs);
2229
2230 atomic_add(segs, &sk->sk_drops);
2231}
2232
2bdc9211
DD
2233static inline ktime_t sock_read_timestamp(struct sock *sk)
2234{
2235#if BITS_PER_LONG==32
2236 unsigned int seq;
2237 ktime_t kt;
2238
2239 do {
2240 seq = read_seqbegin(&sk->sk_stamp_seq);
2241 kt = sk->sk_stamp;
2242 } while (read_seqretry(&sk->sk_stamp_seq, seq));
2243
2244 return kt;
2245#else
2246 return sk->sk_stamp;
2247#endif
2248}
2249
2250static inline void sock_write_timestamp(struct sock *sk, ktime_t kt)
2251{
2252#if BITS_PER_LONG==32
2253 write_seqlock(&sk->sk_stamp_seq);
2254 sk->sk_stamp = kt;
2255 write_sequnlock(&sk->sk_stamp_seq);
2256#else
2257 sk->sk_stamp = kt;
2258#endif
2259}
2260
69336bd2
JP
2261void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2262 struct sk_buff *skb);
2263void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2264 struct sk_buff *skb);
92f37fd2 2265
dc6b9b78 2266static inline void
1da177e4
LT
2267sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2268{
b7aa0bf7 2269 ktime_t kt = skb->tstamp;
20d49473 2270 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
a61bbcf2 2271
20d49473
PO
2272 /*
2273 * generate control messages if
b9f40e21 2274 * - receive time stamping in software requested
20d49473 2275 * - software time stamp available and wanted
20d49473 2276 * - hardware time stamps available and wanted
20d49473
PO
2277 */
2278 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
b9f40e21 2279 (sk->sk_tsflags & SOF_TIMESTAMPING_RX_SOFTWARE) ||
2456e855
TG
2280 (kt && sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) ||
2281 (hwtstamps->hwtstamp &&
b9f40e21 2282 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE)))
92f37fd2
ED
2283 __sock_recv_timestamp(msg, sk, skb);
2284 else
2bdc9211 2285 sock_write_timestamp(sk, kt);
6e3e939f
JB
2286
2287 if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2288 __sock_recv_wifi_status(msg, sk, skb);
1da177e4
LT
2289}
2290
69336bd2
JP
2291void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2292 struct sk_buff *skb);
767dd033 2293
6c7c98ba 2294#define SK_DEFAULT_STAMP (-1L * NSEC_PER_SEC)
767dd033
ED
2295static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2296 struct sk_buff *skb)
2297{
2298#define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
b9f40e21
WB
2299 (1UL << SOCK_RCVTSTAMP))
2300#define TSFLAGS_ANY (SOF_TIMESTAMPING_SOFTWARE | \
2301 SOF_TIMESTAMPING_RAW_HARDWARE)
767dd033 2302
b9f40e21 2303 if (sk->sk_flags & FLAGS_TS_OR_DROPS || sk->sk_tsflags & TSFLAGS_ANY)
767dd033 2304 __sock_recv_ts_and_drops(msg, sk, skb);
d3fbff30 2305 else if (unlikely(sock_flag(sk, SOCK_TIMESTAMP)))
2bdc9211 2306 sock_write_timestamp(sk, skb->tstamp);
6c7c98ba 2307 else if (unlikely(sk->sk_stamp == SK_DEFAULT_STAMP))
2bdc9211 2308 sock_write_timestamp(sk, 0);
767dd033 2309}
3b885787 2310
c14ac945 2311void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags);
67cc0d40 2312
20d49473
PO
2313/**
2314 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
20d49473 2315 * @sk: socket sending this packet
c14ac945 2316 * @tsflags: timestamping flags to use
140c55d4
ED
2317 * @tx_flags: completed with instructions for time stamping
2318 *
d651983d 2319 * Note: callers should take care of initial ``*tx_flags`` value (usually 0)
20d49473 2320 */
c14ac945
SHY
2321static inline void sock_tx_timestamp(const struct sock *sk, __u16 tsflags,
2322 __u8 *tx_flags)
67cc0d40 2323{
c14ac945
SHY
2324 if (unlikely(tsflags))
2325 __sock_tx_timestamp(tsflags, tx_flags);
67cc0d40
WB
2326 if (unlikely(sock_flag(sk, SOCK_WIFI_STATUS)))
2327 *tx_flags |= SKBTX_WIFI_STATUS;
2328}
20d49473 2329
1da177e4
LT
2330/**
2331 * sk_eat_skb - Release a skb if it is no longer needed
4dc3b16b
PP
2332 * @sk: socket to eat this skb from
2333 * @skb: socket buffer to eat
1da177e4
LT
2334 *
2335 * This routine must be called with interrupts disabled or with the socket
2336 * locked so that the sk_buff queue operation is ok.
2337*/
7bced397 2338static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb)
1da177e4
LT
2339{
2340 __skb_unlink(skb, &sk->sk_receive_queue);
2341 __kfree_skb(skb);
2342}
2343
3b1e0a65
YH
2344static inline
2345struct net *sock_net(const struct sock *sk)
2346{
c2d9ba9b 2347 return read_pnet(&sk->sk_net);
3b1e0a65
YH
2348}
2349
2350static inline
f5aa23fd 2351void sock_net_set(struct sock *sk, struct net *net)
3b1e0a65 2352{
c2d9ba9b 2353 write_pnet(&sk->sk_net, net);
3b1e0a65
YH
2354}
2355
23542618
KK
2356static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2357{
efc27f8c 2358 if (skb->sk) {
23542618
KK
2359 struct sock *sk = skb->sk;
2360
2361 skb->destructor = NULL;
2362 skb->sk = NULL;
2363 return sk;
2364 }
2365 return NULL;
2366}
2367
1d0ab253
ED
2368/* This helper checks if a socket is a full socket,
2369 * ie _not_ a timewait or request socket.
2370 */
2371static inline bool sk_fullsock(const struct sock *sk)
2372{
2373 return (1 << sk->sk_state) & ~(TCPF_TIME_WAIT | TCPF_NEW_SYN_RECV);
2374}
2375
e446f9df
ED
2376/* This helper checks if a socket is a LISTEN or NEW_SYN_RECV
2377 * SYNACK messages can be attached to either ones (depending on SYNCOOKIE)
2378 */
2379static inline bool sk_listener(const struct sock *sk)
2380{
2381 return (1 << sk->sk_state) & (TCPF_LISTEN | TCPF_NEW_SYN_RECV);
2382}
2383
00fd38d9
ED
2384/**
2385 * sk_state_load - read sk->sk_state for lockless contexts
2386 * @sk: socket pointer
2387 *
2388 * Paired with sk_state_store(). Used in places we do not hold socket lock :
2389 * tcp_diag_get_info(), tcp_get_info(), tcp_poll(), get_tcp4_sock() ...
2390 */
2391static inline int sk_state_load(const struct sock *sk)
2392{
2393 return smp_load_acquire(&sk->sk_state);
2394}
2395
2396/**
2397 * sk_state_store - update sk->sk_state
2398 * @sk: socket pointer
2399 * @newstate: new state
2400 *
2401 * Paired with sk_state_load(). Should be used in contexts where
2402 * state change might impact lockless readers.
2403 */
2404static inline void sk_state_store(struct sock *sk, int newstate)
2405{
2406 smp_store_release(&sk->sk_state, newstate);
2407}
2408
69336bd2
JP
2409void sock_enable_timestamp(struct sock *sk, int flag);
2410int sock_get_timestamp(struct sock *, struct timeval __user *);
2411int sock_get_timestampns(struct sock *, struct timespec __user *);
2412int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len, int level,
2413 int type);
1da177e4 2414
a3b299da
EB
2415bool sk_ns_capable(const struct sock *sk,
2416 struct user_namespace *user_ns, int cap);
2417bool sk_capable(const struct sock *sk, int cap);
2418bool sk_net_capable(const struct sock *sk, int cap);
2419
a2d133b1
JH
2420void sk_get_meminfo(const struct sock *sk, u32 *meminfo);
2421
eaa72dc4
ED
2422/* Take into consideration the size of the struct sk_buff overhead in the
2423 * determination of these values, since that is non-constant across
2424 * platforms. This makes socket queueing behavior and performance
2425 * not depend upon such differences.
2426 */
2427#define _SK_MEM_PACKETS 256
2428#define _SK_MEM_OVERHEAD SKB_TRUESIZE(256)
2429#define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
2430#define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
2431
1da177e4
LT
2432extern __u32 sysctl_wmem_max;
2433extern __u32 sysctl_rmem_max;
2434
b245be1f 2435extern int sysctl_tstamp_allow_data;
6baf1f41
DM
2436extern int sysctl_optmem_max;
2437
20380731
ACM
2438extern __u32 sysctl_wmem_default;
2439extern __u32 sysctl_rmem_default;
20380731 2440
a3dcaf17
ED
2441static inline int sk_get_wmem0(const struct sock *sk, const struct proto *proto)
2442{
2443 /* Does this proto have per netns sysctl_wmem ? */
2444 if (proto->sysctl_wmem_offset)
2445 return *(int *)((void *)sock_net(sk) + proto->sysctl_wmem_offset);
2446
2447 return *proto->sysctl_wmem;
2448}
2449
2450static inline int sk_get_rmem0(const struct sock *sk, const struct proto *proto)
2451{
2452 /* Does this proto have per netns sysctl_rmem ? */
2453 if (proto->sysctl_rmem_offset)
2454 return *(int *)((void *)sock_net(sk) + proto->sysctl_rmem_offset);
2455
2456 return *proto->sysctl_rmem;
2457}
2458
1da177e4 2459#endif /* _SOCK_H */