]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blame - include/net/sock.h
net: introduce sk_gfp_atomic() to allow addition of GFP flags depending on the indivi...
[mirror_ubuntu-hirsute-kernel.git] / include / net / sock.h
CommitLineData
1da177e4
LT
1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the AF_INET socket handler.
7 *
8 * Version: @(#)sock.h 1.0.4 05/13/93
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Florian La Roche <flla@stud.uni-sb.de>
14 *
15 * Fixes:
16 * Alan Cox : Volatiles in skbuff pointers. See
17 * skbuff comments. May be overdone,
18 * better to prove they can be removed
19 * than the reverse.
20 * Alan Cox : Added a zapped field for tcp to note
21 * a socket is reset and must stay shut up
22 * Alan Cox : New fields for options
23 * Pauline Middelink : identd support
24 * Alan Cox : Eliminate low level recv/recvfrom
25 * David S. Miller : New socket lookup architecture.
26 * Steve Whitehouse: Default routines for sock_ops
27 * Arnaldo C. Melo : removed net_pinfo, tp_pinfo and made
28 * protinfo be just a void pointer, as the
29 * protocol specific parts were moved to
30 * respective headers and ipv4/v6, etc now
31 * use private slabcaches for its socks
32 * Pedro Hortas : New flags field for socket options
33 *
34 *
35 * This program is free software; you can redistribute it and/or
36 * modify it under the terms of the GNU General Public License
37 * as published by the Free Software Foundation; either version
38 * 2 of the License, or (at your option) any later version.
39 */
40#ifndef _SOCK_H
41#define _SOCK_H
42
a6b7a407 43#include <linux/hardirq.h>
172589cc 44#include <linux/kernel.h>
1da177e4 45#include <linux/list.h>
88ab1932 46#include <linux/list_nulls.h>
1da177e4
LT
47#include <linux/timer.h>
48#include <linux/cache.h>
3f134619 49#include <linux/bitops.h>
a5b5bb9a 50#include <linux/lockdep.h>
1da177e4
LT
51#include <linux/netdevice.h>
52#include <linux/skbuff.h> /* struct sk_buff */
d7fe0f24 53#include <linux/mm.h>
1da177e4 54#include <linux/security.h>
5a0e3ad6 55#include <linux/slab.h>
c6e1a0d1 56#include <linux/uaccess.h>
180d8cd9 57#include <linux/memcontrol.h>
e1aab161 58#include <linux/res_counter.h>
c5905afb 59#include <linux/static_key.h>
40401530
AV
60#include <linux/aio.h>
61#include <linux/sched.h>
1da177e4
LT
62
63#include <linux/filter.h>
88ab1932 64#include <linux/rculist_nulls.h>
a57de0b4 65#include <linux/poll.h>
1da177e4 66
c31504dc 67#include <linux/atomic.h>
1da177e4
LT
68#include <net/dst.h>
69#include <net/checksum.h>
70
9f048bfb
ED
71struct cgroup;
72struct cgroup_subsys;
c607b2ed 73#ifdef CONFIG_NET
1d62e436
GC
74int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss);
75void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg);
c607b2ed
GC
76#else
77static inline
1d62e436 78int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
c607b2ed
GC
79{
80 return 0;
81}
82static inline
1d62e436 83void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg)
c607b2ed
GC
84{
85}
86#endif
1da177e4
LT
87/*
88 * This structure really needs to be cleaned up.
89 * Most of it is for TCP, and not used by any of
90 * the other protocols.
91 */
92
93/* Define this to get the SOCK_DBG debugging facility. */
94#define SOCK_DEBUGGING
95#ifdef SOCK_DEBUGGING
96#define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
97 printk(KERN_DEBUG msg); } while (0)
98#else
4cd9029d 99/* Validate arguments and do nothing */
b9075fa9 100static inline __printf(2, 3)
dc6b9b78 101void SOCK_DEBUG(const struct sock *sk, const char *msg, ...)
4cd9029d
SH
102{
103}
1da177e4
LT
104#endif
105
106/* This is the per-socket lock. The spinlock provides a synchronization
107 * between user contexts and software interrupt processing, whereas the
108 * mini-semaphore synchronizes multiple users amongst themselves.
109 */
1da177e4
LT
110typedef struct {
111 spinlock_t slock;
d2e9117c 112 int owned;
1da177e4 113 wait_queue_head_t wq;
a5b5bb9a
IM
114 /*
115 * We express the mutex-alike socket_lock semantics
116 * to the lock validator by explicitly managing
117 * the slock as a lock variant (in addition to
118 * the slock itself):
119 */
120#ifdef CONFIG_DEBUG_LOCK_ALLOC
121 struct lockdep_map dep_map;
122#endif
1da177e4
LT
123} socket_lock_t;
124
1da177e4 125struct sock;
8feaf0c0 126struct proto;
0eeb8ffc 127struct net;
1da177e4
LT
128
129/**
4dc3b16b 130 * struct sock_common - minimal network layer representation of sockets
68835aba
ED
131 * @skc_daddr: Foreign IPv4 addr
132 * @skc_rcv_saddr: Bound local IPv4 addr
4dc6dc71 133 * @skc_hash: hash value used with various protocol lookup tables
d4cada4a 134 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
4dc3b16b
PP
135 * @skc_family: network address family
136 * @skc_state: Connection state
137 * @skc_reuse: %SO_REUSEADDR setting
138 * @skc_bound_dev_if: bound device index if != 0
4dc3b16b 139 * @skc_bind_node: bind hash linkage for various protocol lookup tables
512615b6 140 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
8feaf0c0 141 * @skc_prot: protocol handlers inside a network family
07feaebf 142 * @skc_net: reference to the network namespace of this socket
68835aba
ED
143 * @skc_node: main hash linkage for various protocol lookup tables
144 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
145 * @skc_tx_queue_mapping: tx queue number for this connection
146 * @skc_refcnt: reference count
4dc3b16b
PP
147 *
148 * This is the minimal network layer representation of sockets, the header
8feaf0c0
ACM
149 * for struct sock and struct inet_timewait_sock.
150 */
1da177e4 151struct sock_common {
68835aba
ED
152 /* skc_daddr and skc_rcv_saddr must be grouped :
153 * cf INET_MATCH() and INET_TW_MATCH()
4dc6dc71 154 */
68835aba
ED
155 __be32 skc_daddr;
156 __be32 skc_rcv_saddr;
4dc6dc71 157
d4cada4a
ED
158 union {
159 unsigned int skc_hash;
160 __u16 skc_u16hashes[2];
161 };
4dc6dc71
ED
162 unsigned short skc_family;
163 volatile unsigned char skc_state;
164 unsigned char skc_reuse;
165 int skc_bound_dev_if;
512615b6
ED
166 union {
167 struct hlist_node skc_bind_node;
168 struct hlist_nulls_node skc_portaddr_node;
169 };
8feaf0c0 170 struct proto *skc_prot;
3b1e0a65 171#ifdef CONFIG_NET_NS
07feaebf 172 struct net *skc_net;
3b1e0a65 173#endif
68835aba
ED
174 /*
175 * fields between dontcopy_begin/dontcopy_end
176 * are not copied in sock_copy()
177 */
928c41e7 178 /* private: */
68835aba 179 int skc_dontcopy_begin[0];
928c41e7 180 /* public: */
68835aba
ED
181 union {
182 struct hlist_node skc_node;
183 struct hlist_nulls_node skc_nulls_node;
184 };
185 int skc_tx_queue_mapping;
186 atomic_t skc_refcnt;
928c41e7 187 /* private: */
68835aba 188 int skc_dontcopy_end[0];
928c41e7 189 /* public: */
1da177e4
LT
190};
191
e1aab161 192struct cg_proto;
1da177e4
LT
193/**
194 * struct sock - network layer representation of sockets
8feaf0c0 195 * @__sk_common: shared layout with inet_timewait_sock
4dc3b16b
PP
196 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
197 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
198 * @sk_lock: synchronizer
199 * @sk_rcvbuf: size of receive buffer in bytes
43815482 200 * @sk_wq: sock wait queue and async head
deaa5854 201 * @sk_rx_dst: receive input route used by early tcp demux
4dc3b16b
PP
202 * @sk_dst_cache: destination cache
203 * @sk_dst_lock: destination cache lock
204 * @sk_policy: flow policy
4dc3b16b
PP
205 * @sk_receive_queue: incoming packets
206 * @sk_wmem_alloc: transmit queue bytes committed
207 * @sk_write_queue: Packet sending queue
97fc2f08 208 * @sk_async_wait_queue: DMA copied packets
4dc3b16b
PP
209 * @sk_omem_alloc: "o" is "option" or "other"
210 * @sk_wmem_queued: persistent queue size
211 * @sk_forward_alloc: space allocated forward
212 * @sk_allocation: allocation mode
213 * @sk_sndbuf: size of send buffer in bytes
33c732c3 214 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
20d49473 215 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
4dc3b16b
PP
216 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
217 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
a465419b 218 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
bcd76111 219 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
82cc1a7a 220 * @sk_gso_max_size: Maximum GSO segment size to build
4dc3b16b 221 * @sk_lingertime: %SO_LINGER l_linger setting
4dc3b16b
PP
222 * @sk_backlog: always used with the per-socket spinlock held
223 * @sk_callback_lock: used with the callbacks in the end of this struct
224 * @sk_error_queue: rarely used
33c732c3
WC
225 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
226 * IPV6_ADDRFORM for instance)
4dc3b16b 227 * @sk_err: last error
33c732c3
WC
228 * @sk_err_soft: errors that don't cause failure but are the cause of a
229 * persistent failure not just 'timed out'
cb61cb9b 230 * @sk_drops: raw/udp drops counter
4dc3b16b
PP
231 * @sk_ack_backlog: current listen backlog
232 * @sk_max_ack_backlog: listen backlog set in listen()
233 * @sk_priority: %SO_PRIORITY setting
1a3bc369 234 * @sk_cgrp_prioidx: socket group's priority map index
4dc3b16b
PP
235 * @sk_type: socket type (%SOCK_STREAM, etc)
236 * @sk_protocol: which protocol this socket belongs in this network family
53c3fa20
RD
237 * @sk_peer_pid: &struct pid for this socket's peer
238 * @sk_peer_cred: %SO_PEERCRED setting
4dc3b16b
PP
239 * @sk_rcvlowat: %SO_RCVLOWAT setting
240 * @sk_rcvtimeo: %SO_RCVTIMEO setting
241 * @sk_sndtimeo: %SO_SNDTIMEO setting
c58dc01b 242 * @sk_rxhash: flow hash received from netif layer
4dc3b16b
PP
243 * @sk_filter: socket filtering instructions
244 * @sk_protinfo: private area, net family specific, when not using slab
245 * @sk_timer: sock cleanup timer
246 * @sk_stamp: time stamp of last packet received
247 * @sk_socket: Identd and reporting IO signals
248 * @sk_user_data: RPC layer private data
249 * @sk_sndmsg_page: cached page for sendmsg
250 * @sk_sndmsg_off: cached offset for sendmsg
d3d4f0a0 251 * @sk_peek_off: current peek_offset value
4dc3b16b 252 * @sk_send_head: front of stuff to transmit
67be2dd1 253 * @sk_security: used by security modules
31729363 254 * @sk_mark: generic packet mark
53c3fa20 255 * @sk_classid: this socket's cgroup classid
e1aab161 256 * @sk_cgrp: this socket's cgroup-specific proto data
4dc3b16b
PP
257 * @sk_write_pending: a write to stream socket waits to start
258 * @sk_state_change: callback to indicate change in the state of the sock
259 * @sk_data_ready: callback to indicate there is data to be processed
260 * @sk_write_space: callback to indicate there is bf sending space available
261 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
262 * @sk_backlog_rcv: callback to process the backlog
263 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
1da177e4
LT
264 */
265struct sock {
266 /*
8feaf0c0 267 * Now struct inet_timewait_sock also uses sock_common, so please just
1da177e4
LT
268 * don't add nothing before this first member (__sk_common) --acme
269 */
270 struct sock_common __sk_common;
4dc6dc71
ED
271#define sk_node __sk_common.skc_node
272#define sk_nulls_node __sk_common.skc_nulls_node
273#define sk_refcnt __sk_common.skc_refcnt
e022f0b4 274#define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
4dc6dc71 275
68835aba
ED
276#define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
277#define sk_dontcopy_end __sk_common.skc_dontcopy_end
4dc6dc71 278#define sk_hash __sk_common.skc_hash
1da177e4
LT
279#define sk_family __sk_common.skc_family
280#define sk_state __sk_common.skc_state
281#define sk_reuse __sk_common.skc_reuse
282#define sk_bound_dev_if __sk_common.skc_bound_dev_if
1da177e4 283#define sk_bind_node __sk_common.skc_bind_node
8feaf0c0 284#define sk_prot __sk_common.skc_prot
07feaebf 285#define sk_net __sk_common.skc_net
1da177e4 286 socket_lock_t sk_lock;
b178bb3d 287 struct sk_buff_head sk_receive_queue;
fa438ccf
ED
288 /*
289 * The backlog queue is special, it is always used with
290 * the per-socket spinlock held and requires low latency
291 * access. Therefore we special case it's implementation.
b178bb3d
ED
292 * Note : rmem_alloc is in this structure to fill a hole
293 * on 64bit arches, not because its logically part of
294 * backlog.
fa438ccf
ED
295 */
296 struct {
b178bb3d
ED
297 atomic_t rmem_alloc;
298 int len;
299 struct sk_buff *head;
300 struct sk_buff *tail;
fa438ccf 301 } sk_backlog;
b178bb3d
ED
302#define sk_rmem_alloc sk_backlog.rmem_alloc
303 int sk_forward_alloc;
304#ifdef CONFIG_RPS
305 __u32 sk_rxhash;
306#endif
307 atomic_t sk_drops;
308 int sk_rcvbuf;
309
310 struct sk_filter __rcu *sk_filter;
eaefd110 311 struct socket_wq __rcu *sk_wq;
b178bb3d
ED
312
313#ifdef CONFIG_NET_DMA
314 struct sk_buff_head sk_async_wait_queue;
315#endif
316
def8b4fa 317#ifdef CONFIG_XFRM
1da177e4 318 struct xfrm_policy *sk_policy[2];
def8b4fa 319#endif
b178bb3d 320 unsigned long sk_flags;
deaa5854 321 struct dst_entry *sk_rx_dst;
b178bb3d 322 struct dst_entry *sk_dst_cache;
b6c6712a 323 spinlock_t sk_dst_lock;
1da177e4
LT
324 atomic_t sk_wmem_alloc;
325 atomic_t sk_omem_alloc;
4e07a91c 326 int sk_sndbuf;
1da177e4 327 struct sk_buff_head sk_write_queue;
b178bb3d
ED
328 kmemcheck_bitfield_begin(flags);
329 unsigned int sk_shutdown : 2,
330 sk_no_check : 2,
331 sk_userlocks : 4,
332 sk_protocol : 8,
333 sk_type : 16;
334 kmemcheck_bitfield_end(flags);
1da177e4 335 int sk_wmem_queued;
7d877f3b 336 gfp_t sk_allocation;
c8f44aff
MM
337 netdev_features_t sk_route_caps;
338 netdev_features_t sk_route_nocaps;
bcd76111 339 int sk_gso_type;
82cc1a7a 340 unsigned int sk_gso_max_size;
9932cf95 341 int sk_rcvlowat;
1da177e4 342 unsigned long sk_lingertime;
1da177e4 343 struct sk_buff_head sk_error_queue;
476e19cf 344 struct proto *sk_prot_creator;
1da177e4
LT
345 rwlock_t sk_callback_lock;
346 int sk_err,
347 sk_err_soft;
348 unsigned short sk_ack_backlog;
349 unsigned short sk_max_ack_backlog;
350 __u32 sk_priority;
5bc1421e
NH
351#ifdef CONFIG_CGROUPS
352 __u32 sk_cgrp_prioidx;
353#endif
109f6e39
EB
354 struct pid *sk_peer_pid;
355 const struct cred *sk_peer_cred;
1da177e4
LT
356 long sk_rcvtimeo;
357 long sk_sndtimeo;
1da177e4
LT
358 void *sk_protinfo;
359 struct timer_list sk_timer;
b7aa0bf7 360 ktime_t sk_stamp;
1da177e4
LT
361 struct socket *sk_socket;
362 void *sk_user_data;
363 struct page *sk_sndmsg_page;
364 struct sk_buff *sk_send_head;
365 __u32 sk_sndmsg_off;
ef64a54f 366 __s32 sk_peek_off;
1da177e4 367 int sk_write_pending;
d5f64238 368#ifdef CONFIG_SECURITY
1da177e4 369 void *sk_security;
d5f64238 370#endif
4a19ec58 371 __u32 sk_mark;
f8451725 372 u32 sk_classid;
e1aab161 373 struct cg_proto *sk_cgrp;
1da177e4
LT
374 void (*sk_state_change)(struct sock *sk);
375 void (*sk_data_ready)(struct sock *sk, int bytes);
376 void (*sk_write_space)(struct sock *sk);
377 void (*sk_error_report)(struct sock *sk);
dc6b9b78
ED
378 int (*sk_backlog_rcv)(struct sock *sk,
379 struct sk_buff *skb);
1da177e4
LT
380 void (*sk_destruct)(struct sock *sk);
381};
382
4a17fd52
PE
383/*
384 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
385 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
386 * on a socket means that the socket will reuse everybody else's port
387 * without looking at the other's sk_reuse value.
388 */
389
390#define SK_NO_REUSE 0
391#define SK_CAN_REUSE 1
392#define SK_FORCE_REUSE 2
393
ef64a54f
PE
394static inline int sk_peek_offset(struct sock *sk, int flags)
395{
396 if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0))
397 return sk->sk_peek_off;
398 else
399 return 0;
400}
401
402static inline void sk_peek_offset_bwd(struct sock *sk, int val)
403{
404 if (sk->sk_peek_off >= 0) {
405 if (sk->sk_peek_off >= val)
406 sk->sk_peek_off -= val;
407 else
408 sk->sk_peek_off = 0;
409 }
410}
411
412static inline void sk_peek_offset_fwd(struct sock *sk, int val)
413{
414 if (sk->sk_peek_off >= 0)
415 sk->sk_peek_off += val;
416}
417
1da177e4
LT
418/*
419 * Hashed lists helper routines
420 */
c4146644
LZ
421static inline struct sock *sk_entry(const struct hlist_node *node)
422{
423 return hlist_entry(node, struct sock, sk_node);
424}
425
e48c414e 426static inline struct sock *__sk_head(const struct hlist_head *head)
1da177e4
LT
427{
428 return hlist_entry(head->first, struct sock, sk_node);
429}
430
e48c414e 431static inline struct sock *sk_head(const struct hlist_head *head)
1da177e4
LT
432{
433 return hlist_empty(head) ? NULL : __sk_head(head);
434}
435
88ab1932
ED
436static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
437{
438 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
439}
440
441static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
442{
443 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
444}
445
e48c414e 446static inline struct sock *sk_next(const struct sock *sk)
1da177e4
LT
447{
448 return sk->sk_node.next ?
449 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
450}
451
88ab1932
ED
452static inline struct sock *sk_nulls_next(const struct sock *sk)
453{
454 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
455 hlist_nulls_entry(sk->sk_nulls_node.next,
456 struct sock, sk_nulls_node) :
457 NULL;
458}
459
dc6b9b78 460static inline bool sk_unhashed(const struct sock *sk)
1da177e4
LT
461{
462 return hlist_unhashed(&sk->sk_node);
463}
464
dc6b9b78 465static inline bool sk_hashed(const struct sock *sk)
1da177e4 466{
da753bea 467 return !sk_unhashed(sk);
1da177e4
LT
468}
469
dc6b9b78 470static inline void sk_node_init(struct hlist_node *node)
1da177e4
LT
471{
472 node->pprev = NULL;
473}
474
dc6b9b78 475static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
88ab1932
ED
476{
477 node->pprev = NULL;
478}
479
dc6b9b78 480static inline void __sk_del_node(struct sock *sk)
1da177e4
LT
481{
482 __hlist_del(&sk->sk_node);
483}
484
808f5114 485/* NB: equivalent to hlist_del_init_rcu */
dc6b9b78 486static inline bool __sk_del_node_init(struct sock *sk)
1da177e4
LT
487{
488 if (sk_hashed(sk)) {
489 __sk_del_node(sk);
490 sk_node_init(&sk->sk_node);
dc6b9b78 491 return true;
1da177e4 492 }
dc6b9b78 493 return false;
1da177e4
LT
494}
495
496/* Grab socket reference count. This operation is valid only
497 when sk is ALREADY grabbed f.e. it is found in hash table
498 or a list and the lookup is made under lock preventing hash table
499 modifications.
500 */
501
502static inline void sock_hold(struct sock *sk)
503{
504 atomic_inc(&sk->sk_refcnt);
505}
506
507/* Ungrab socket in the context, which assumes that socket refcnt
508 cannot hit zero, f.e. it is true in context of any socketcall.
509 */
510static inline void __sock_put(struct sock *sk)
511{
512 atomic_dec(&sk->sk_refcnt);
513}
514
dc6b9b78 515static inline bool sk_del_node_init(struct sock *sk)
1da177e4 516{
dc6b9b78 517 bool rc = __sk_del_node_init(sk);
1da177e4
LT
518
519 if (rc) {
520 /* paranoid for a while -acme */
521 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
522 __sock_put(sk);
523 }
524 return rc;
525}
808f5114 526#define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
1da177e4 527
dc6b9b78 528static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
271b72c7
ED
529{
530 if (sk_hashed(sk)) {
88ab1932 531 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
dc6b9b78 532 return true;
271b72c7 533 }
dc6b9b78 534 return false;
271b72c7
ED
535}
536
dc6b9b78 537static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
271b72c7 538{
dc6b9b78 539 bool rc = __sk_nulls_del_node_init_rcu(sk);
271b72c7
ED
540
541 if (rc) {
542 /* paranoid for a while -acme */
543 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
544 __sock_put(sk);
545 }
546 return rc;
547}
548
dc6b9b78 549static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
1da177e4
LT
550{
551 hlist_add_head(&sk->sk_node, list);
552}
553
dc6b9b78 554static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
1da177e4
LT
555{
556 sock_hold(sk);
557 __sk_add_node(sk, list);
558}
559
dc6b9b78 560static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
808f5114 561{
562 sock_hold(sk);
563 hlist_add_head_rcu(&sk->sk_node, list);
564}
565
dc6b9b78 566static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
271b72c7 567{
88ab1932 568 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
271b72c7
ED
569}
570
dc6b9b78 571static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
271b72c7
ED
572{
573 sock_hold(sk);
88ab1932 574 __sk_nulls_add_node_rcu(sk, list);
271b72c7
ED
575}
576
dc6b9b78 577static inline void __sk_del_bind_node(struct sock *sk)
1da177e4
LT
578{
579 __hlist_del(&sk->sk_bind_node);
580}
581
dc6b9b78 582static inline void sk_add_bind_node(struct sock *sk,
1da177e4
LT
583 struct hlist_head *list)
584{
585 hlist_add_head(&sk->sk_bind_node, list);
586}
587
588#define sk_for_each(__sk, node, list) \
589 hlist_for_each_entry(__sk, node, list, sk_node)
808f5114 590#define sk_for_each_rcu(__sk, node, list) \
591 hlist_for_each_entry_rcu(__sk, node, list, sk_node)
88ab1932
ED
592#define sk_nulls_for_each(__sk, node, list) \
593 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
594#define sk_nulls_for_each_rcu(__sk, node, list) \
595 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
1da177e4
LT
596#define sk_for_each_from(__sk, node) \
597 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
598 hlist_for_each_entry_from(__sk, node, sk_node)
88ab1932
ED
599#define sk_nulls_for_each_from(__sk, node) \
600 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
601 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
1da177e4
LT
602#define sk_for_each_safe(__sk, node, tmp, list) \
603 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
604#define sk_for_each_bound(__sk, node, list) \
605 hlist_for_each_entry(__sk, node, list, sk_bind_node)
606
607/* Sock flags */
608enum sock_flags {
609 SOCK_DEAD,
610 SOCK_DONE,
611 SOCK_URGINLINE,
612 SOCK_KEEPOPEN,
613 SOCK_LINGER,
614 SOCK_DESTROY,
615 SOCK_BROADCAST,
616 SOCK_TIMESTAMP,
617 SOCK_ZAPPED,
618 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
619 SOCK_DBG, /* %SO_DEBUG setting */
620 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
92f37fd2 621 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
1da177e4
LT
622 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
623 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
20d49473
PO
624 SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */
625 SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
626 SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */
627 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
628 SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */
629 SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
630 SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
bcdce719 631 SOCK_FASYNC, /* fasync() active */
3b885787 632 SOCK_RXQ_OVFL,
1cdebb42 633 SOCK_ZEROCOPY, /* buffers from userspace */
6e3e939f 634 SOCK_WIFI_STATUS, /* push wifi status to userspace */
3bdc0eba
BG
635 SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
636 * Will use last 4 bytes of packet sent from
637 * user-space instead.
638 */
1da177e4
LT
639};
640
53b924b3
RB
641static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
642{
643 nsk->sk_flags = osk->sk_flags;
644}
645
1da177e4
LT
646static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
647{
648 __set_bit(flag, &sk->sk_flags);
649}
650
651static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
652{
653 __clear_bit(flag, &sk->sk_flags);
654}
655
1b23a5df 656static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
1da177e4
LT
657{
658 return test_bit(flag, &sk->sk_flags);
659}
660
99a1dec7
MG
661static inline gfp_t sk_gfp_atomic(struct sock *sk, gfp_t gfp_mask)
662{
663 return GFP_ATOMIC;
664}
665
1da177e4
LT
666static inline void sk_acceptq_removed(struct sock *sk)
667{
668 sk->sk_ack_backlog--;
669}
670
671static inline void sk_acceptq_added(struct sock *sk)
672{
673 sk->sk_ack_backlog++;
674}
675
dc6b9b78 676static inline bool sk_acceptq_is_full(const struct sock *sk)
1da177e4 677{
64a14651 678 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
1da177e4
LT
679}
680
681/*
682 * Compute minimal free write space needed to queue new packets.
683 */
dc6b9b78 684static inline int sk_stream_min_wspace(const struct sock *sk)
1da177e4 685{
8df09ea3 686 return sk->sk_wmem_queued >> 1;
1da177e4
LT
687}
688
dc6b9b78 689static inline int sk_stream_wspace(const struct sock *sk)
1da177e4
LT
690{
691 return sk->sk_sndbuf - sk->sk_wmem_queued;
692}
693
694extern void sk_stream_write_space(struct sock *sk);
695
dc6b9b78 696static inline bool sk_stream_memory_free(const struct sock *sk)
1da177e4
LT
697{
698 return sk->sk_wmem_queued < sk->sk_sndbuf;
699}
700
8eae939f 701/* OOB backlog add */
a3a858ff 702static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
9ee6b535 703{
7fee226a
ED
704 /* dont let skb dst not refcounted, we are going to leave rcu lock */
705 skb_dst_force(skb);
706
707 if (!sk->sk_backlog.tail)
708 sk->sk_backlog.head = skb;
709 else
9ee6b535 710 sk->sk_backlog.tail->next = skb;
7fee226a
ED
711
712 sk->sk_backlog.tail = skb;
9ee6b535
SH
713 skb->next = NULL;
714}
1da177e4 715
c377411f
ED
716/*
717 * Take into account size of receive queue and backlog queue
0fd7bac6
ED
718 * Do not take into account this skb truesize,
719 * to allow even a single big packet to come.
c377411f 720 */
f545a38f
ED
721static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb,
722 unsigned int limit)
c377411f
ED
723{
724 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
725
f545a38f 726 return qsize > limit;
c377411f
ED
727}
728
8eae939f 729/* The per-socket spinlock must be held here. */
f545a38f
ED
730static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
731 unsigned int limit)
8eae939f 732{
f545a38f 733 if (sk_rcvqueues_full(sk, skb, limit))
8eae939f
ZY
734 return -ENOBUFS;
735
a3a858ff 736 __sk_add_backlog(sk, skb);
8eae939f
ZY
737 sk->sk_backlog.len += skb->truesize;
738 return 0;
739}
740
c57943a1
PZ
741static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
742{
743 return sk->sk_backlog_rcv(sk, skb);
744}
745
c58dc01b
DM
746static inline void sock_rps_record_flow(const struct sock *sk)
747{
748#ifdef CONFIG_RPS
749 struct rps_sock_flow_table *sock_flow_table;
750
751 rcu_read_lock();
752 sock_flow_table = rcu_dereference(rps_sock_flow_table);
753 rps_record_sock_flow(sock_flow_table, sk->sk_rxhash);
754 rcu_read_unlock();
755#endif
756}
757
758static inline void sock_rps_reset_flow(const struct sock *sk)
759{
760#ifdef CONFIG_RPS
761 struct rps_sock_flow_table *sock_flow_table;
762
763 rcu_read_lock();
764 sock_flow_table = rcu_dereference(rps_sock_flow_table);
765 rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash);
766 rcu_read_unlock();
767#endif
768}
769
bdeab991
TH
770static inline void sock_rps_save_rxhash(struct sock *sk,
771 const struct sk_buff *skb)
c58dc01b
DM
772{
773#ifdef CONFIG_RPS
bdeab991 774 if (unlikely(sk->sk_rxhash != skb->rxhash)) {
c58dc01b 775 sock_rps_reset_flow(sk);
bdeab991 776 sk->sk_rxhash = skb->rxhash;
c58dc01b
DM
777 }
778#endif
779}
780
bdeab991
TH
781static inline void sock_rps_reset_rxhash(struct sock *sk)
782{
783#ifdef CONFIG_RPS
784 sock_rps_reset_flow(sk);
785 sk->sk_rxhash = 0;
786#endif
787}
788
cfcabdcc
SH
789#define sk_wait_event(__sk, __timeo, __condition) \
790 ({ int __rc; \
791 release_sock(__sk); \
792 __rc = __condition; \
793 if (!__rc) { \
794 *(__timeo) = schedule_timeout(*(__timeo)); \
795 } \
796 lock_sock(__sk); \
797 __rc = __condition; \
798 __rc; \
799 })
1da177e4
LT
800
801extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
802extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
803extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
804extern int sk_stream_error(struct sock *sk, int flags, int err);
805extern void sk_stream_kill_queues(struct sock *sk);
806
807extern int sk_wait_data(struct sock *sk, long *timeo);
808
60236fdd 809struct request_sock_ops;
6d6ee43e 810struct timewait_sock_ops;
ab1e0a13 811struct inet_hashinfo;
fc8717ba 812struct raw_hashinfo;
de477254 813struct module;
2e6599cb 814
1da177e4
LT
815/* Networking protocol blocks we attach to sockets.
816 * socket layer -> transport layer interface
817 * transport -> network interface is defined by struct inet_proto
818 */
819struct proto {
dc6b9b78 820 void (*close)(struct sock *sk,
1da177e4
LT
821 long timeout);
822 int (*connect)(struct sock *sk,
dc6b9b78 823 struct sockaddr *uaddr,
1da177e4
LT
824 int addr_len);
825 int (*disconnect)(struct sock *sk, int flags);
826
dc6b9b78 827 struct sock * (*accept)(struct sock *sk, int flags, int *err);
1da177e4
LT
828
829 int (*ioctl)(struct sock *sk, int cmd,
830 unsigned long arg);
831 int (*init)(struct sock *sk);
7d06b2e0 832 void (*destroy)(struct sock *sk);
1da177e4 833 void (*shutdown)(struct sock *sk, int how);
dc6b9b78 834 int (*setsockopt)(struct sock *sk, int level,
1da177e4 835 int optname, char __user *optval,
b7058842 836 unsigned int optlen);
dc6b9b78
ED
837 int (*getsockopt)(struct sock *sk, int level,
838 int optname, char __user *optval,
839 int __user *option);
af01d537 840#ifdef CONFIG_COMPAT
3fdadf7d
DM
841 int (*compat_setsockopt)(struct sock *sk,
842 int level,
843 int optname, char __user *optval,
b7058842 844 unsigned int optlen);
3fdadf7d
DM
845 int (*compat_getsockopt)(struct sock *sk,
846 int level,
847 int optname, char __user *optval,
848 int __user *option);
709b46e8
EB
849 int (*compat_ioctl)(struct sock *sk,
850 unsigned int cmd, unsigned long arg);
af01d537 851#endif
1da177e4
LT
852 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
853 struct msghdr *msg, size_t len);
854 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
855 struct msghdr *msg,
dc6b9b78
ED
856 size_t len, int noblock, int flags,
857 int *addr_len);
1da177e4
LT
858 int (*sendpage)(struct sock *sk, struct page *page,
859 int offset, size_t size, int flags);
dc6b9b78 860 int (*bind)(struct sock *sk,
1da177e4
LT
861 struct sockaddr *uaddr, int addr_len);
862
dc6b9b78 863 int (*backlog_rcv) (struct sock *sk,
1da177e4
LT
864 struct sk_buff *skb);
865
46d3ceab 866 void (*release_cb)(struct sock *sk);
563d34d0 867 void (*mtu_reduced)(struct sock *sk);
46d3ceab 868
1da177e4
LT
869 /* Keeping track of sk's, looking them up, and port selection methods. */
870 void (*hash)(struct sock *sk);
871 void (*unhash)(struct sock *sk);
719f8358 872 void (*rehash)(struct sock *sk);
1da177e4 873 int (*get_port)(struct sock *sk, unsigned short snum);
fcbdf09d 874 void (*clear_sk)(struct sock *sk, int size);
1da177e4 875
286ab3d4 876 /* Keeping track of sockets in use */
65f76517 877#ifdef CONFIG_PROC_FS
13ff3d6f 878 unsigned int inuse_idx;
65f76517 879#endif
ebb53d75 880
1da177e4 881 /* Memory pressure */
5c52ba17 882 void (*enter_memory_pressure)(struct sock *sk);
8d987e5c 883 atomic_long_t *memory_allocated; /* Current allocated memory. */
1748376b 884 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
1da177e4
LT
885 /*
886 * Pressure flag: try to collapse.
887 * Technical note: it is used by multiple contexts non atomically.
3ab224be 888 * All the __sk_mem_schedule() is of this nature: accounting
1da177e4
LT
889 * is strict, actions are advisory and have some latency.
890 */
891 int *memory_pressure;
8d987e5c 892 long *sysctl_mem;
1da177e4
LT
893 int *sysctl_wmem;
894 int *sysctl_rmem;
895 int max_header;
7ba42910 896 bool no_autobind;
1da177e4 897
271b72c7 898 struct kmem_cache *slab;
1da177e4 899 unsigned int obj_size;
271b72c7 900 int slab_flags;
1da177e4 901
dd24c001 902 struct percpu_counter *orphan_count;
8feaf0c0 903
60236fdd 904 struct request_sock_ops *rsk_prot;
6d6ee43e 905 struct timewait_sock_ops *twsk_prot;
2e6599cb 906
39d8cda7
PE
907 union {
908 struct inet_hashinfo *hashinfo;
645ca708 909 struct udp_table *udp_table;
fc8717ba 910 struct raw_hashinfo *raw_hash;
39d8cda7 911 } h;
ab1e0a13 912
1da177e4
LT
913 struct module *owner;
914
915 char name[32];
916
917 struct list_head node;
e6848976
ACM
918#ifdef SOCK_REFCNT_DEBUG
919 atomic_t socks;
920#endif
c255a458 921#ifdef CONFIG_MEMCG_KMEM
e1aab161
GC
922 /*
923 * cgroup specific init/deinit functions. Called once for all
924 * protocols that implement it, from cgroups populate function.
925 * This function has to setup any files the protocol want to
926 * appear in the kmem cgroup filesystem.
927 */
1d62e436 928 int (*init_cgroup)(struct mem_cgroup *memcg,
e1aab161 929 struct cgroup_subsys *ss);
1d62e436 930 void (*destroy_cgroup)(struct mem_cgroup *memcg);
e1aab161
GC
931 struct cg_proto *(*proto_cgroup)(struct mem_cgroup *memcg);
932#endif
933};
934
3f134619
GC
935/*
936 * Bits in struct cg_proto.flags
937 */
938enum cg_proto_flags {
939 /* Currently active and new sockets should be assigned to cgroups */
940 MEMCG_SOCK_ACTIVE,
941 /* It was ever activated; we must disarm static keys on destruction */
942 MEMCG_SOCK_ACTIVATED,
943};
944
e1aab161
GC
945struct cg_proto {
946 void (*enter_memory_pressure)(struct sock *sk);
947 struct res_counter *memory_allocated; /* Current allocated memory. */
948 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
949 int *memory_pressure;
950 long *sysctl_mem;
3f134619 951 unsigned long flags;
e1aab161
GC
952 /*
953 * memcg field is used to find which memcg we belong directly
954 * Each memcg struct can hold more than one cg_proto, so container_of
955 * won't really cut.
956 *
957 * The elegant solution would be having an inverse function to
958 * proto_cgroup in struct proto, but that means polluting the structure
959 * for everybody, instead of just for memcg users.
960 */
961 struct mem_cgroup *memcg;
1da177e4
LT
962};
963
964extern int proto_register(struct proto *prot, int alloc_slab);
965extern void proto_unregister(struct proto *prot);
966
3f134619
GC
967static inline bool memcg_proto_active(struct cg_proto *cg_proto)
968{
969 return test_bit(MEMCG_SOCK_ACTIVE, &cg_proto->flags);
970}
971
972static inline bool memcg_proto_activated(struct cg_proto *cg_proto)
973{
974 return test_bit(MEMCG_SOCK_ACTIVATED, &cg_proto->flags);
975}
976
e6848976
ACM
977#ifdef SOCK_REFCNT_DEBUG
978static inline void sk_refcnt_debug_inc(struct sock *sk)
979{
980 atomic_inc(&sk->sk_prot->socks);
981}
982
983static inline void sk_refcnt_debug_dec(struct sock *sk)
984{
985 atomic_dec(&sk->sk_prot->socks);
986 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
987 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
988}
989
e1aab161 990inline void sk_refcnt_debug_release(const struct sock *sk)
e6848976
ACM
991{
992 if (atomic_read(&sk->sk_refcnt) != 1)
993 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
994 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
995}
996#else /* SOCK_REFCNT_DEBUG */
997#define sk_refcnt_debug_inc(sk) do { } while (0)
998#define sk_refcnt_debug_dec(sk) do { } while (0)
999#define sk_refcnt_debug_release(sk) do { } while (0)
1000#endif /* SOCK_REFCNT_DEBUG */
1001
c255a458 1002#if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_NET)
c5905afb 1003extern struct static_key memcg_socket_limit_enabled;
e1aab161
GC
1004static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1005 struct cg_proto *cg_proto)
1006{
1007 return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg));
1008}
c5905afb 1009#define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
e1aab161
GC
1010#else
1011#define mem_cgroup_sockets_enabled 0
1012static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1013 struct cg_proto *cg_proto)
1014{
1015 return NULL;
1016}
1017#endif
1018
1019
180d8cd9
GC
1020static inline bool sk_has_memory_pressure(const struct sock *sk)
1021{
1022 return sk->sk_prot->memory_pressure != NULL;
1023}
1024
1025static inline bool sk_under_memory_pressure(const struct sock *sk)
1026{
1027 if (!sk->sk_prot->memory_pressure)
1028 return false;
e1aab161
GC
1029
1030 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1031 return !!*sk->sk_cgrp->memory_pressure;
1032
180d8cd9
GC
1033 return !!*sk->sk_prot->memory_pressure;
1034}
1035
1036static inline void sk_leave_memory_pressure(struct sock *sk)
1037{
1038 int *memory_pressure = sk->sk_prot->memory_pressure;
1039
e1aab161
GC
1040 if (!memory_pressure)
1041 return;
1042
1043 if (*memory_pressure)
180d8cd9 1044 *memory_pressure = 0;
e1aab161
GC
1045
1046 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1047 struct cg_proto *cg_proto = sk->sk_cgrp;
1048 struct proto *prot = sk->sk_prot;
1049
1050 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1051 if (*cg_proto->memory_pressure)
1052 *cg_proto->memory_pressure = 0;
1053 }
1054
180d8cd9
GC
1055}
1056
1057static inline void sk_enter_memory_pressure(struct sock *sk)
1058{
e1aab161
GC
1059 if (!sk->sk_prot->enter_memory_pressure)
1060 return;
1061
1062 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1063 struct cg_proto *cg_proto = sk->sk_cgrp;
1064 struct proto *prot = sk->sk_prot;
1065
1066 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1067 cg_proto->enter_memory_pressure(sk);
1068 }
1069
1070 sk->sk_prot->enter_memory_pressure(sk);
180d8cd9
GC
1071}
1072
1073static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1074{
1075 long *prot = sk->sk_prot->sysctl_mem;
e1aab161
GC
1076 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1077 prot = sk->sk_cgrp->sysctl_mem;
180d8cd9
GC
1078 return prot[index];
1079}
1080
e1aab161
GC
1081static inline void memcg_memory_allocated_add(struct cg_proto *prot,
1082 unsigned long amt,
1083 int *parent_status)
1084{
1085 struct res_counter *fail;
1086 int ret;
1087
0e90b31f
GC
1088 ret = res_counter_charge_nofail(prot->memory_allocated,
1089 amt << PAGE_SHIFT, &fail);
e1aab161
GC
1090 if (ret < 0)
1091 *parent_status = OVER_LIMIT;
1092}
1093
1094static inline void memcg_memory_allocated_sub(struct cg_proto *prot,
1095 unsigned long amt)
1096{
1097 res_counter_uncharge(prot->memory_allocated, amt << PAGE_SHIFT);
1098}
1099
1100static inline u64 memcg_memory_allocated_read(struct cg_proto *prot)
1101{
1102 u64 ret;
1103 ret = res_counter_read_u64(prot->memory_allocated, RES_USAGE);
1104 return ret >> PAGE_SHIFT;
1105}
1106
180d8cd9
GC
1107static inline long
1108sk_memory_allocated(const struct sock *sk)
1109{
1110 struct proto *prot = sk->sk_prot;
e1aab161
GC
1111 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1112 return memcg_memory_allocated_read(sk->sk_cgrp);
1113
180d8cd9
GC
1114 return atomic_long_read(prot->memory_allocated);
1115}
1116
1117static inline long
e1aab161 1118sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status)
180d8cd9
GC
1119{
1120 struct proto *prot = sk->sk_prot;
e1aab161
GC
1121
1122 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1123 memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status);
1124 /* update the root cgroup regardless */
1125 atomic_long_add_return(amt, prot->memory_allocated);
1126 return memcg_memory_allocated_read(sk->sk_cgrp);
1127 }
1128
180d8cd9
GC
1129 return atomic_long_add_return(amt, prot->memory_allocated);
1130}
1131
1132static inline void
0e90b31f 1133sk_memory_allocated_sub(struct sock *sk, int amt)
180d8cd9
GC
1134{
1135 struct proto *prot = sk->sk_prot;
e1aab161 1136
0e90b31f 1137 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
e1aab161
GC
1138 memcg_memory_allocated_sub(sk->sk_cgrp, amt);
1139
180d8cd9
GC
1140 atomic_long_sub(amt, prot->memory_allocated);
1141}
1142
1143static inline void sk_sockets_allocated_dec(struct sock *sk)
1144{
1145 struct proto *prot = sk->sk_prot;
e1aab161
GC
1146
1147 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1148 struct cg_proto *cg_proto = sk->sk_cgrp;
1149
1150 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1151 percpu_counter_dec(cg_proto->sockets_allocated);
1152 }
1153
180d8cd9
GC
1154 percpu_counter_dec(prot->sockets_allocated);
1155}
1156
1157static inline void sk_sockets_allocated_inc(struct sock *sk)
1158{
1159 struct proto *prot = sk->sk_prot;
e1aab161
GC
1160
1161 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1162 struct cg_proto *cg_proto = sk->sk_cgrp;
1163
1164 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1165 percpu_counter_inc(cg_proto->sockets_allocated);
1166 }
1167
180d8cd9
GC
1168 percpu_counter_inc(prot->sockets_allocated);
1169}
1170
1171static inline int
1172sk_sockets_allocated_read_positive(struct sock *sk)
1173{
1174 struct proto *prot = sk->sk_prot;
1175
e1aab161 1176 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
518fbf9c 1177 return percpu_counter_read_positive(sk->sk_cgrp->sockets_allocated);
e1aab161 1178
518fbf9c 1179 return percpu_counter_read_positive(prot->sockets_allocated);
180d8cd9
GC
1180}
1181
1182static inline int
1183proto_sockets_allocated_sum_positive(struct proto *prot)
1184{
1185 return percpu_counter_sum_positive(prot->sockets_allocated);
1186}
1187
1188static inline long
1189proto_memory_allocated(struct proto *prot)
1190{
1191 return atomic_long_read(prot->memory_allocated);
1192}
1193
1194static inline bool
1195proto_memory_pressure(struct proto *prot)
1196{
1197 if (!prot->memory_pressure)
1198 return false;
1199 return !!*prot->memory_pressure;
1200}
1201
65f76517
ED
1202
1203#ifdef CONFIG_PROC_FS
1da177e4 1204/* Called with local bh disabled */
c29a0bc4
PE
1205extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1206extern int sock_prot_inuse_get(struct net *net, struct proto *proto);
65f76517 1207#else
dc6b9b78 1208static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
c29a0bc4 1209 int inc)
65f76517
ED
1210{
1211}
65f76517
ED
1212#endif
1213
1da177e4 1214
614c6cb4
ACM
1215/* With per-bucket locks this operation is not-atomic, so that
1216 * this version is not worse.
1217 */
1218static inline void __sk_prot_rehash(struct sock *sk)
1219{
1220 sk->sk_prot->unhash(sk);
1221 sk->sk_prot->hash(sk);
1222}
1223
fcbdf09d
OP
1224void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
1225
1da177e4
LT
1226/* About 10 seconds */
1227#define SOCK_DESTROY_TIME (10*HZ)
1228
1229/* Sockets 0-1023 can't be bound to unless you are superuser */
1230#define PROT_SOCK 1024
1231
1232#define SHUTDOWN_MASK 3
1233#define RCV_SHUTDOWN 1
1234#define SEND_SHUTDOWN 2
1235
1236#define SOCK_SNDBUF_LOCK 1
1237#define SOCK_RCVBUF_LOCK 2
1238#define SOCK_BINDADDR_LOCK 4
1239#define SOCK_BINDPORT_LOCK 8
1240
1241/* sock_iocb: used to kick off async processing of socket ios */
1242struct sock_iocb {
1243 struct list_head list;
1244
1245 int flags;
1246 int size;
1247 struct socket *sock;
1248 struct sock *sk;
1249 struct scm_cookie *scm;
1250 struct msghdr *msg, async_msg;
1da177e4
LT
1251 struct kiocb *kiocb;
1252};
1253
1254static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
1255{
1256 return (struct sock_iocb *)iocb->private;
1257}
1258
1259static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
1260{
1261 return si->kiocb;
1262}
1263
1264struct socket_alloc {
1265 struct socket socket;
1266 struct inode vfs_inode;
1267};
1268
1269static inline struct socket *SOCKET_I(struct inode *inode)
1270{
1271 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1272}
1273
1274static inline struct inode *SOCK_INODE(struct socket *socket)
1275{
1276 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1277}
1278
3ab224be
HA
1279/*
1280 * Functions for memory accounting
1281 */
1282extern int __sk_mem_schedule(struct sock *sk, int size, int kind);
1283extern void __sk_mem_reclaim(struct sock *sk);
1da177e4 1284
3ab224be
HA
1285#define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1286#define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1287#define SK_MEM_SEND 0
1288#define SK_MEM_RECV 1
1da177e4 1289
3ab224be 1290static inline int sk_mem_pages(int amt)
1da177e4 1291{
3ab224be 1292 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
1da177e4
LT
1293}
1294
dc6b9b78 1295static inline bool sk_has_account(struct sock *sk)
1da177e4 1296{
3ab224be
HA
1297 /* return true if protocol supports memory accounting */
1298 return !!sk->sk_prot->memory_allocated;
1da177e4
LT
1299}
1300
dc6b9b78 1301static inline bool sk_wmem_schedule(struct sock *sk, int size)
1da177e4 1302{
3ab224be 1303 if (!sk_has_account(sk))
dc6b9b78 1304 return true;
3ab224be
HA
1305 return size <= sk->sk_forward_alloc ||
1306 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1da177e4
LT
1307}
1308
dc6b9b78 1309static inline bool sk_rmem_schedule(struct sock *sk, int size)
d80d99d6 1310{
3ab224be 1311 if (!sk_has_account(sk))
dc6b9b78 1312 return true;
d80d99d6 1313 return size <= sk->sk_forward_alloc ||
3ab224be
HA
1314 __sk_mem_schedule(sk, size, SK_MEM_RECV);
1315}
1316
1317static inline void sk_mem_reclaim(struct sock *sk)
1318{
1319 if (!sk_has_account(sk))
1320 return;
1321 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1322 __sk_mem_reclaim(sk);
1323}
1324
9993e7d3
DM
1325static inline void sk_mem_reclaim_partial(struct sock *sk)
1326{
1327 if (!sk_has_account(sk))
1328 return;
1329 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1330 __sk_mem_reclaim(sk);
1331}
1332
3ab224be
HA
1333static inline void sk_mem_charge(struct sock *sk, int size)
1334{
1335 if (!sk_has_account(sk))
1336 return;
1337 sk->sk_forward_alloc -= size;
1338}
1339
1340static inline void sk_mem_uncharge(struct sock *sk, int size)
1341{
1342 if (!sk_has_account(sk))
1343 return;
1344 sk->sk_forward_alloc += size;
1345}
1346
1347static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1348{
3ab224be
HA
1349 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1350 sk->sk_wmem_queued -= skb->truesize;
1351 sk_mem_uncharge(sk, skb->truesize);
1352 __kfree_skb(skb);
d80d99d6
HX
1353}
1354
1da177e4
LT
1355/* Used by processes to "lock" a socket state, so that
1356 * interrupts and bottom half handlers won't change it
1357 * from under us. It essentially blocks any incoming
1358 * packets, so that we won't get any new data or any
1359 * packets that change the state of the socket.
1360 *
1361 * While locked, BH processing will add new packets to
1362 * the backlog queue. This queue is processed by the
1363 * owner of the socket lock right before it is released.
1364 *
1365 * Since ~2.3.5 it is also exclusive sleep lock serializing
1366 * accesses from user process context.
1367 */
d2e9117c 1368#define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
1da177e4 1369
ed07536e
PZ
1370/*
1371 * Macro so as to not evaluate some arguments when
1372 * lockdep is not enabled.
1373 *
1374 * Mark both the sk_lock and the sk_lock.slock as a
1375 * per-address-family lock class.
1376 */
dc6b9b78 1377#define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
ed07536e 1378do { \
e8f6fbf6 1379 sk->sk_lock.owned = 0; \
ed07536e
PZ
1380 init_waitqueue_head(&sk->sk_lock.wq); \
1381 spin_lock_init(&(sk)->sk_lock.slock); \
1382 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1383 sizeof((sk)->sk_lock)); \
1384 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
dc6b9b78 1385 (skey), (sname)); \
ed07536e
PZ
1386 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1387} while (0)
1388
41380930 1389extern void lock_sock_nested(struct sock *sk, int subclass);
fcc70d5f
PZ
1390
1391static inline void lock_sock(struct sock *sk)
1392{
1393 lock_sock_nested(sk, 0);
1394}
1395
41380930 1396extern void release_sock(struct sock *sk);
1da177e4
LT
1397
1398/* BH context may only use the following locking interface. */
1399#define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
c6366184
IM
1400#define bh_lock_sock_nested(__sk) \
1401 spin_lock_nested(&((__sk)->sk_lock.slock), \
1402 SINGLE_DEPTH_NESTING)
1da177e4
LT
1403#define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1404
8a74ad60
ED
1405extern bool lock_sock_fast(struct sock *sk);
1406/**
1407 * unlock_sock_fast - complement of lock_sock_fast
1408 * @sk: socket
1409 * @slow: slow mode
1410 *
1411 * fast unlock socket for user context.
1412 * If slow mode is on, we call regular release_sock()
1413 */
1414static inline void unlock_sock_fast(struct sock *sk, bool slow)
4b0b72f7 1415{
8a74ad60
ED
1416 if (slow)
1417 release_sock(sk);
1418 else
1419 spin_unlock_bh(&sk->sk_lock.slock);
4b0b72f7
ED
1420}
1421
4b0b72f7 1422
1b8d7ae4 1423extern struct sock *sk_alloc(struct net *net, int family,
dd0fc66f 1424 gfp_t priority,
6257ff21 1425 struct proto *prot);
1da177e4 1426extern void sk_free(struct sock *sk);
edf02087 1427extern void sk_release_kernel(struct sock *sk);
e56c57d0
ED
1428extern struct sock *sk_clone_lock(const struct sock *sk,
1429 const gfp_t priority);
1da177e4
LT
1430
1431extern struct sk_buff *sock_wmalloc(struct sock *sk,
1432 unsigned long size, int force,
dd0fc66f 1433 gfp_t priority);
1da177e4
LT
1434extern struct sk_buff *sock_rmalloc(struct sock *sk,
1435 unsigned long size, int force,
dd0fc66f 1436 gfp_t priority);
1da177e4
LT
1437extern void sock_wfree(struct sk_buff *skb);
1438extern void sock_rfree(struct sk_buff *skb);
41063e9d 1439extern void sock_edemux(struct sk_buff *skb);
1da177e4
LT
1440
1441extern int sock_setsockopt(struct socket *sock, int level,
1442 int op, char __user *optval,
b7058842 1443 unsigned int optlen);
1da177e4
LT
1444
1445extern int sock_getsockopt(struct socket *sock, int level,
dc6b9b78 1446 int op, char __user *optval,
1da177e4 1447 int __user *optlen);
dc6b9b78 1448extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
1da177e4
LT
1449 unsigned long size,
1450 int noblock,
1451 int *errcode);
dc6b9b78 1452extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
4cc7f68d
HX
1453 unsigned long header_len,
1454 unsigned long data_len,
1455 int noblock,
1456 int *errcode);
86a76caf 1457extern void *sock_kmalloc(struct sock *sk, int size,
dd0fc66f 1458 gfp_t priority);
1da177e4
LT
1459extern void sock_kfree_s(struct sock *sk, void *mem, int size);
1460extern void sk_send_sigurg(struct sock *sk);
1461
f8451725
HX
1462#ifdef CONFIG_CGROUPS
1463extern void sock_update_classid(struct sock *sk);
1464#else
1465static inline void sock_update_classid(struct sock *sk)
1466{
1467}
1468#endif
1469
1da177e4
LT
1470/*
1471 * Functions to fill in entries in struct proto_ops when a protocol
1472 * does not implement a particular function.
1473 */
dc6b9b78 1474extern int sock_no_bind(struct socket *,
1da177e4
LT
1475 struct sockaddr *, int);
1476extern int sock_no_connect(struct socket *,
1477 struct sockaddr *, int, int);
1478extern int sock_no_socketpair(struct socket *,
1479 struct socket *);
1480extern int sock_no_accept(struct socket *,
1481 struct socket *, int);
1482extern int sock_no_getname(struct socket *,
1483 struct sockaddr *, int *, int);
1484extern unsigned int sock_no_poll(struct file *, struct socket *,
1485 struct poll_table_struct *);
1486extern int sock_no_ioctl(struct socket *, unsigned int,
1487 unsigned long);
1488extern int sock_no_listen(struct socket *, int);
1489extern int sock_no_shutdown(struct socket *, int);
1490extern int sock_no_getsockopt(struct socket *, int , int,
1491 char __user *, int __user *);
1492extern int sock_no_setsockopt(struct socket *, int, int,
b7058842 1493 char __user *, unsigned int);
1da177e4
LT
1494extern int sock_no_sendmsg(struct kiocb *, struct socket *,
1495 struct msghdr *, size_t);
1496extern int sock_no_recvmsg(struct kiocb *, struct socket *,
1497 struct msghdr *, size_t, int);
1498extern int sock_no_mmap(struct file *file,
1499 struct socket *sock,
1500 struct vm_area_struct *vma);
1501extern ssize_t sock_no_sendpage(struct socket *sock,
1502 struct page *page,
dc6b9b78 1503 int offset, size_t size,
1da177e4
LT
1504 int flags);
1505
1506/*
1507 * Functions to fill in entries in struct proto_ops when a protocol
1508 * uses the inet style.
1509 */
1510extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
1511 char __user *optval, int __user *optlen);
1512extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1513 struct msghdr *msg, size_t size, int flags);
1514extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 1515 char __user *optval, unsigned int optlen);
3fdadf7d
DM
1516extern int compat_sock_common_getsockopt(struct socket *sock, int level,
1517 int optname, char __user *optval, int __user *optlen);
1518extern int compat_sock_common_setsockopt(struct socket *sock, int level,
b7058842 1519 int optname, char __user *optval, unsigned int optlen);
1da177e4
LT
1520
1521extern void sk_common_release(struct sock *sk);
1522
1523/*
1524 * Default socket callbacks and setup code
1525 */
dc6b9b78 1526
1da177e4
LT
1527/* Initialise core socket variables */
1528extern void sock_init_data(struct socket *sock, struct sock *sk);
1529
46bcf14f
ED
1530extern void sk_filter_release_rcu(struct rcu_head *rcu);
1531
dc9b3346 1532/**
1a5778aa 1533 * sk_filter_release - release a socket filter
dc9b3346
PB
1534 * @fp: filter to remove
1535 *
1536 * Remove a filter from a socket and release its resources.
1537 */
1538
309dd5fc
PE
1539static inline void sk_filter_release(struct sk_filter *fp)
1540{
1541 if (atomic_dec_and_test(&fp->refcnt))
80f8f102 1542 call_rcu(&fp->rcu, sk_filter_release_rcu);
309dd5fc
PE
1543}
1544
1545static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
1da177e4
LT
1546{
1547 unsigned int size = sk_filter_len(fp);
1548
1549 atomic_sub(size, &sk->sk_omem_alloc);
309dd5fc 1550 sk_filter_release(fp);
1da177e4
LT
1551}
1552
1553static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
1554{
1555 atomic_inc(&fp->refcnt);
1556 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
1557}
1558
1559/*
1560 * Socket reference counting postulates.
1561 *
1562 * * Each user of socket SHOULD hold a reference count.
1563 * * Each access point to socket (an hash table bucket, reference from a list,
1564 * running timer, skb in flight MUST hold a reference count.
1565 * * When reference count hits 0, it means it will never increase back.
1566 * * When reference count hits 0, it means that no references from
1567 * outside exist to this socket and current process on current CPU
1568 * is last user and may/should destroy this socket.
1569 * * sk_free is called from any context: process, BH, IRQ. When
1570 * it is called, socket has no references from outside -> sk_free
1571 * may release descendant resources allocated by the socket, but
1572 * to the time when it is called, socket is NOT referenced by any
1573 * hash tables, lists etc.
1574 * * Packets, delivered from outside (from network or from another process)
1575 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1576 * when they sit in queue. Otherwise, packets will leak to hole, when
1577 * socket is looked up by one cpu and unhasing is made by another CPU.
1578 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1579 * (leak to backlog). Packet socket does all the processing inside
1580 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1581 * use separate SMP lock, so that they are prone too.
1582 */
1583
1584/* Ungrab socket and destroy it, if it was the last reference. */
1585static inline void sock_put(struct sock *sk)
1586{
1587 if (atomic_dec_and_test(&sk->sk_refcnt))
1588 sk_free(sk);
1589}
1590
58a5a7b9
ACM
1591extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1592 const int nested);
25995ff5 1593
e022f0b4
KK
1594static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1595{
1596 sk->sk_tx_queue_mapping = tx_queue;
1597}
1598
1599static inline void sk_tx_queue_clear(struct sock *sk)
1600{
1601 sk->sk_tx_queue_mapping = -1;
1602}
1603
1604static inline int sk_tx_queue_get(const struct sock *sk)
1605{
b0f77d0e 1606 return sk ? sk->sk_tx_queue_mapping : -1;
e022f0b4
KK
1607}
1608
972692e0
DM
1609static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1610{
e022f0b4 1611 sk_tx_queue_clear(sk);
972692e0
DM
1612 sk->sk_socket = sock;
1613}
1614
aa395145
ED
1615static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1616{
eaefd110
ED
1617 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1618 return &rcu_dereference_raw(sk->sk_wq)->wait;
aa395145 1619}
1da177e4
LT
1620/* Detach socket from process context.
1621 * Announce socket dead, detach it from wait queue and inode.
1622 * Note that parent inode held reference count on this struct sock,
1623 * we do not release it in this function, because protocol
1624 * probably wants some additional cleanups or even continuing
1625 * to work with this socket (TCP).
1626 */
1627static inline void sock_orphan(struct sock *sk)
1628{
1629 write_lock_bh(&sk->sk_callback_lock);
1630 sock_set_flag(sk, SOCK_DEAD);
972692e0 1631 sk_set_socket(sk, NULL);
43815482 1632 sk->sk_wq = NULL;
1da177e4
LT
1633 write_unlock_bh(&sk->sk_callback_lock);
1634}
1635
1636static inline void sock_graft(struct sock *sk, struct socket *parent)
1637{
1638 write_lock_bh(&sk->sk_callback_lock);
eaefd110 1639 sk->sk_wq = parent->wq;
1da177e4 1640 parent->sk = sk;
972692e0 1641 sk_set_socket(sk, parent);
4237c75c 1642 security_sock_graft(sk, parent);
1da177e4
LT
1643 write_unlock_bh(&sk->sk_callback_lock);
1644}
1645
1646extern int sock_i_uid(struct sock *sk);
1647extern unsigned long sock_i_ino(struct sock *sk);
1648
1649static inline struct dst_entry *
1650__sk_dst_get(struct sock *sk)
1651{
d8bf4ca9 1652 return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
f68c224f 1653 lockdep_is_held(&sk->sk_lock.slock));
1da177e4
LT
1654}
1655
1656static inline struct dst_entry *
1657sk_dst_get(struct sock *sk)
1658{
1659 struct dst_entry *dst;
1660
b6c6712a
ED
1661 rcu_read_lock();
1662 dst = rcu_dereference(sk->sk_dst_cache);
1da177e4
LT
1663 if (dst)
1664 dst_hold(dst);
b6c6712a 1665 rcu_read_unlock();
1da177e4
LT
1666 return dst;
1667}
1668
b6c6712a
ED
1669extern void sk_reset_txq(struct sock *sk);
1670
1671static inline void dst_negative_advice(struct sock *sk)
1672{
1673 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1674
1675 if (dst && dst->ops->negative_advice) {
1676 ndst = dst->ops->negative_advice(dst);
1677
1678 if (ndst != dst) {
1679 rcu_assign_pointer(sk->sk_dst_cache, ndst);
1680 sk_reset_txq(sk);
1681 }
1682 }
1683}
1684
1da177e4
LT
1685static inline void
1686__sk_dst_set(struct sock *sk, struct dst_entry *dst)
1687{
1688 struct dst_entry *old_dst;
1689
e022f0b4 1690 sk_tx_queue_clear(sk);
0b53ff2e
ED
1691 /*
1692 * This can be called while sk is owned by the caller only,
1693 * with no state that can be checked in a rcu_dereference_check() cond
1694 */
1695 old_dst = rcu_dereference_raw(sk->sk_dst_cache);
b6c6712a 1696 rcu_assign_pointer(sk->sk_dst_cache, dst);
1da177e4
LT
1697 dst_release(old_dst);
1698}
1699
1700static inline void
1701sk_dst_set(struct sock *sk, struct dst_entry *dst)
1702{
b6c6712a 1703 spin_lock(&sk->sk_dst_lock);
1da177e4 1704 __sk_dst_set(sk, dst);
b6c6712a 1705 spin_unlock(&sk->sk_dst_lock);
1da177e4
LT
1706}
1707
1708static inline void
1709__sk_dst_reset(struct sock *sk)
1710{
b6c6712a 1711 __sk_dst_set(sk, NULL);
1da177e4
LT
1712}
1713
1714static inline void
1715sk_dst_reset(struct sock *sk)
1716{
b6c6712a 1717 spin_lock(&sk->sk_dst_lock);
1da177e4 1718 __sk_dst_reset(sk);
b6c6712a 1719 spin_unlock(&sk->sk_dst_lock);
1da177e4
LT
1720}
1721
f0088a50 1722extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1da177e4 1723
f0088a50 1724extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1da177e4 1725
dc6b9b78 1726static inline bool sk_can_gso(const struct sock *sk)
bcd76111
HX
1727{
1728 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1729}
1730
9958089a 1731extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
6cbb0df7 1732
c8f44aff 1733static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
a465419b
ED
1734{
1735 sk->sk_route_nocaps |= flags;
1736 sk->sk_route_caps &= ~flags;
1737}
1738
c6e1a0d1
TH
1739static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1740 char __user *from, char *to,
912d398d 1741 int copy, int offset)
c6e1a0d1
TH
1742{
1743 if (skb->ip_summed == CHECKSUM_NONE) {
1744 int err = 0;
1745 __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err);
1746 if (err)
1747 return err;
912d398d 1748 skb->csum = csum_block_add(skb->csum, csum, offset);
c6e1a0d1
TH
1749 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1750 if (!access_ok(VERIFY_READ, from, copy) ||
1751 __copy_from_user_nocache(to, from, copy))
1752 return -EFAULT;
1753 } else if (copy_from_user(to, from, copy))
1754 return -EFAULT;
1755
1756 return 0;
1757}
1758
1759static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
1760 char __user *from, int copy)
1761{
912d398d 1762 int err, offset = skb->len;
c6e1a0d1 1763
912d398d
WY
1764 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1765 copy, offset);
c6e1a0d1 1766 if (err)
912d398d 1767 __skb_trim(skb, offset);
c6e1a0d1
TH
1768
1769 return err;
1770}
1771
1772static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from,
1773 struct sk_buff *skb,
1774 struct page *page,
1775 int off, int copy)
1776{
1777 int err;
1778
912d398d
WY
1779 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1780 copy, skb->len);
c6e1a0d1
TH
1781 if (err)
1782 return err;
1783
1784 skb->len += copy;
1785 skb->data_len += copy;
1786 skb->truesize += copy;
1787 sk->sk_wmem_queued += copy;
1788 sk_mem_charge(sk, copy);
1789 return 0;
1790}
1791
1da177e4
LT
1792static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1793 struct sk_buff *skb, struct page *page,
1794 int off, int copy)
1795{
1796 if (skb->ip_summed == CHECKSUM_NONE) {
1797 int err = 0;
5084205f 1798 __wsum csum = csum_and_copy_from_user(from,
1da177e4
LT
1799 page_address(page) + off,
1800 copy, 0, &err);
1801 if (err)
1802 return err;
1803 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1804 } else if (copy_from_user(page_address(page) + off, from, copy))
1805 return -EFAULT;
1806
1807 skb->len += copy;
1808 skb->data_len += copy;
1809 skb->truesize += copy;
1810 sk->sk_wmem_queued += copy;
3ab224be 1811 sk_mem_charge(sk, copy);
1da177e4
LT
1812 return 0;
1813}
1814
c564039f
ED
1815/**
1816 * sk_wmem_alloc_get - returns write allocations
1817 * @sk: socket
1818 *
1819 * Returns sk_wmem_alloc minus initial offset of one
1820 */
1821static inline int sk_wmem_alloc_get(const struct sock *sk)
1822{
1823 return atomic_read(&sk->sk_wmem_alloc) - 1;
1824}
1825
1826/**
1827 * sk_rmem_alloc_get - returns read allocations
1828 * @sk: socket
1829 *
1830 * Returns sk_rmem_alloc
1831 */
1832static inline int sk_rmem_alloc_get(const struct sock *sk)
1833{
1834 return atomic_read(&sk->sk_rmem_alloc);
1835}
1836
1837/**
1838 * sk_has_allocations - check if allocations are outstanding
1839 * @sk: socket
1840 *
1841 * Returns true if socket has write or read allocations
1842 */
dc6b9b78 1843static inline bool sk_has_allocations(const struct sock *sk)
c564039f
ED
1844{
1845 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1846}
1847
a57de0b4 1848/**
43815482 1849 * wq_has_sleeper - check if there are any waiting processes
acfbe96a 1850 * @wq: struct socket_wq
a57de0b4 1851 *
43815482 1852 * Returns true if socket_wq has waiting processes
a57de0b4 1853 *
43815482 1854 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
a57de0b4
JO
1855 * barrier call. They were added due to the race found within the tcp code.
1856 *
1857 * Consider following tcp code paths:
1858 *
1859 * CPU1 CPU2
1860 *
1861 * sys_select receive packet
1862 * ... ...
1863 * __add_wait_queue update tp->rcv_nxt
1864 * ... ...
1865 * tp->rcv_nxt check sock_def_readable
1866 * ... {
43815482
ED
1867 * schedule rcu_read_lock();
1868 * wq = rcu_dereference(sk->sk_wq);
1869 * if (wq && waitqueue_active(&wq->wait))
1870 * wake_up_interruptible(&wq->wait)
a57de0b4
JO
1871 * ...
1872 * }
1873 *
1874 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1875 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1876 * could then endup calling schedule and sleep forever if there are no more
1877 * data on the socket.
ad462769 1878 *
a57de0b4 1879 */
43815482 1880static inline bool wq_has_sleeper(struct socket_wq *wq)
a57de0b4 1881{
dc6b9b78 1882 /* We need to be sure we are in sync with the
a57de0b4
JO
1883 * add_wait_queue modifications to the wait queue.
1884 *
1885 * This memory barrier is paired in the sock_poll_wait.
1886 */
43815482
ED
1887 smp_mb();
1888 return wq && waitqueue_active(&wq->wait);
a57de0b4
JO
1889}
1890
1891/**
1892 * sock_poll_wait - place memory barrier behind the poll_wait call.
1893 * @filp: file
1894 * @wait_address: socket wait queue
1895 * @p: poll_table
1896 *
43815482 1897 * See the comments in the wq_has_sleeper function.
a57de0b4
JO
1898 */
1899static inline void sock_poll_wait(struct file *filp,
1900 wait_queue_head_t *wait_address, poll_table *p)
1901{
626cf236 1902 if (!poll_does_not_wait(p) && wait_address) {
a57de0b4 1903 poll_wait(filp, wait_address, p);
dc6b9b78 1904 /* We need to be sure we are in sync with the
a57de0b4
JO
1905 * socket flags modification.
1906 *
43815482 1907 * This memory barrier is paired in the wq_has_sleeper.
dc6b9b78 1908 */
a57de0b4
JO
1909 smp_mb();
1910 }
1911}
1912
1da177e4 1913/*
dc6b9b78 1914 * Queue a received datagram if it will fit. Stream and sequenced
1da177e4
LT
1915 * protocols can't normally use this as they need to fit buffers in
1916 * and play with them.
1917 *
dc6b9b78 1918 * Inlined as it's very short and called for pretty much every
1da177e4
LT
1919 * packet ever received.
1920 */
1921
1922static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1923{
d55d87fd 1924 skb_orphan(skb);
1da177e4
LT
1925 skb->sk = sk;
1926 skb->destructor = sock_wfree;
2b85a34e
ED
1927 /*
1928 * We used to take a refcount on sk, but following operation
1929 * is enough to guarantee sk_free() wont free this sock until
1930 * all in-flight packets are completed
1931 */
1da177e4
LT
1932 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1933}
1934
1935static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1936{
d55d87fd 1937 skb_orphan(skb);
1da177e4
LT
1938 skb->sk = sk;
1939 skb->destructor = sock_rfree;
1940 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
3ab224be 1941 sk_mem_charge(sk, skb->truesize);
1da177e4
LT
1942}
1943
dc6b9b78 1944extern void sk_reset_timer(struct sock *sk, struct timer_list *timer,
1da177e4
LT
1945 unsigned long expires);
1946
dc6b9b78 1947extern void sk_stop_timer(struct sock *sk, struct timer_list *timer);
1da177e4 1948
f0088a50 1949extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
1da177e4 1950
b1faf566 1951extern int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
1da177e4
LT
1952
1953/*
1954 * Recover an error report and clear atomically
1955 */
dc6b9b78 1956
1da177e4
LT
1957static inline int sock_error(struct sock *sk)
1958{
c1cbe4b7
BL
1959 int err;
1960 if (likely(!sk->sk_err))
1961 return 0;
1962 err = xchg(&sk->sk_err, 0);
1da177e4
LT
1963 return -err;
1964}
1965
1966static inline unsigned long sock_wspace(struct sock *sk)
1967{
1968 int amt = 0;
1969
1970 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1971 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
dc6b9b78 1972 if (amt < 0)
1da177e4
LT
1973 amt = 0;
1974 }
1975 return amt;
1976}
1977
1978static inline void sk_wake_async(struct sock *sk, int how, int band)
1979{
bcdce719 1980 if (sock_flag(sk, SOCK_FASYNC))
1da177e4
LT
1981 sock_wake_async(sk->sk_socket, how, band);
1982}
1983
1984#define SOCK_MIN_SNDBUF 2048
7a91b434
ED
1985/*
1986 * Since sk_rmem_alloc sums skb->truesize, even a small frame might need
1987 * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak
1988 */
1989#define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff))
1da177e4
LT
1990
1991static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1992{
1993 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
8df09ea3 1994 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
1da177e4
LT
1995 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1996 }
1997}
1998
df97c708 1999struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
1da177e4
LT
2000
2001static inline struct page *sk_stream_alloc_page(struct sock *sk)
2002{
2003 struct page *page = NULL;
2004
ef015786
HX
2005 page = alloc_pages(sk->sk_allocation, 0);
2006 if (!page) {
180d8cd9 2007 sk_enter_memory_pressure(sk);
1da177e4
LT
2008 sk_stream_moderate_sndbuf(sk);
2009 }
2010 return page;
2011}
2012
1da177e4
LT
2013/*
2014 * Default write policy as shown to user space via poll/select/SIGIO
2015 */
dc6b9b78 2016static inline bool sock_writeable(const struct sock *sk)
1da177e4 2017{
8df09ea3 2018 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
1da177e4
LT
2019}
2020
dd0fc66f 2021static inline gfp_t gfp_any(void)
1da177e4 2022{
99709372 2023 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1da177e4
LT
2024}
2025
dc6b9b78 2026static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
1da177e4
LT
2027{
2028 return noblock ? 0 : sk->sk_rcvtimeo;
2029}
2030
dc6b9b78 2031static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
1da177e4
LT
2032{
2033 return noblock ? 0 : sk->sk_sndtimeo;
2034}
2035
2036static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2037{
2038 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
2039}
2040
2041/* Alas, with timeout socket operations are not restartable.
2042 * Compare this to poll().
2043 */
2044static inline int sock_intr_errno(long timeo)
2045{
2046 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2047}
2048
92f37fd2
ED
2049extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2050 struct sk_buff *skb);
6e3e939f
JB
2051extern void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2052 struct sk_buff *skb);
92f37fd2 2053
dc6b9b78 2054static inline void
1da177e4
LT
2055sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2056{
b7aa0bf7 2057 ktime_t kt = skb->tstamp;
20d49473 2058 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
a61bbcf2 2059
20d49473
PO
2060 /*
2061 * generate control messages if
2062 * - receive time stamping in software requested (SOCK_RCVTSTAMP
2063 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
2064 * - software time stamp available and wanted
2065 * (SOCK_TIMESTAMPING_SOFTWARE)
2066 * - hardware time stamps available and wanted
2067 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
2068 * SOCK_TIMESTAMPING_RAW_HARDWARE)
2069 */
2070 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
2071 sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
2072 (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
2073 (hwtstamps->hwtstamp.tv64 &&
2074 sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
2075 (hwtstamps->syststamp.tv64 &&
2076 sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
92f37fd2
ED
2077 __sock_recv_timestamp(msg, sk, skb);
2078 else
b7aa0bf7 2079 sk->sk_stamp = kt;
6e3e939f
JB
2080
2081 if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2082 __sock_recv_wifi_status(msg, sk, skb);
1da177e4
LT
2083}
2084
767dd033
ED
2085extern void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2086 struct sk_buff *skb);
2087
2088static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2089 struct sk_buff *skb)
2090{
2091#define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2092 (1UL << SOCK_RCVTSTAMP) | \
2093 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
2094 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
dc6b9b78 2095 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
767dd033
ED
2096 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
2097
2098 if (sk->sk_flags & FLAGS_TS_OR_DROPS)
2099 __sock_recv_ts_and_drops(msg, sk, skb);
2100 else
2101 sk->sk_stamp = skb->tstamp;
2102}
3b885787 2103
20d49473
PO
2104/**
2105 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
20d49473 2106 * @sk: socket sending this packet
2244d07b 2107 * @tx_flags: filled with instructions for time stamping
20d49473
PO
2108 *
2109 * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if
2110 * parameters are invalid.
2111 */
2244d07b 2112extern int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags);
20d49473 2113
1da177e4
LT
2114/**
2115 * sk_eat_skb - Release a skb if it is no longer needed
4dc3b16b
PP
2116 * @sk: socket to eat this skb from
2117 * @skb: socket buffer to eat
f4b8ea78 2118 * @copied_early: flag indicating whether DMA operations copied this data early
1da177e4
LT
2119 *
2120 * This routine must be called with interrupts disabled or with the socket
2121 * locked so that the sk_buff queue operation is ok.
2122*/
624d1164 2123#ifdef CONFIG_NET_DMA
dc6b9b78 2124static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
624d1164
CL
2125{
2126 __skb_unlink(skb, &sk->sk_receive_queue);
2127 if (!copied_early)
2128 __kfree_skb(skb);
2129 else
2130 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
2131}
2132#else
dc6b9b78 2133static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
1da177e4
LT
2134{
2135 __skb_unlink(skb, &sk->sk_receive_queue);
2136 __kfree_skb(skb);
2137}
624d1164 2138#endif
1da177e4 2139
3b1e0a65
YH
2140static inline
2141struct net *sock_net(const struct sock *sk)
2142{
c2d9ba9b 2143 return read_pnet(&sk->sk_net);
3b1e0a65
YH
2144}
2145
2146static inline
f5aa23fd 2147void sock_net_set(struct sock *sk, struct net *net)
3b1e0a65 2148{
c2d9ba9b 2149 write_pnet(&sk->sk_net, net);
3b1e0a65
YH
2150}
2151
edf02087
DL
2152/*
2153 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
25985edc 2154 * They should not hold a reference to a namespace in order to allow
edf02087
DL
2155 * to stop it.
2156 * Sockets after sk_change_net should be released using sk_release_kernel
2157 */
2158static inline void sk_change_net(struct sock *sk, struct net *net)
2159{
3b1e0a65 2160 put_net(sock_net(sk));
65a18ec5 2161 sock_net_set(sk, hold_net(net));
edf02087
DL
2162}
2163
23542618
KK
2164static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2165{
efc27f8c 2166 if (skb->sk) {
23542618
KK
2167 struct sock *sk = skb->sk;
2168
2169 skb->destructor = NULL;
2170 skb->sk = NULL;
2171 return sk;
2172 }
2173 return NULL;
2174}
2175
20d49473 2176extern void sock_enable_timestamp(struct sock *sk, int flag);
1da177e4 2177extern int sock_get_timestamp(struct sock *, struct timeval __user *);
ae40eb1e 2178extern int sock_get_timestampns(struct sock *, struct timespec __user *);
1da177e4 2179
dc6b9b78
ED
2180/*
2181 * Enable debug/info messages
1da177e4 2182 */
a2a316fd
SH
2183extern int net_msg_warn;
2184#define NETDEBUG(fmt, args...) \
2185 do { if (net_msg_warn) printk(fmt,##args); } while (0)
1da177e4 2186
a2a316fd
SH
2187#define LIMIT_NETDEBUG(fmt, args...) \
2188 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
1da177e4 2189
1da177e4
LT
2190extern __u32 sysctl_wmem_max;
2191extern __u32 sysctl_rmem_max;
2192
20380731
ACM
2193extern void sk_init(void);
2194
6baf1f41
DM
2195extern int sysctl_optmem_max;
2196
20380731
ACM
2197extern __u32 sysctl_wmem_default;
2198extern __u32 sysctl_rmem_default;
20380731 2199
1da177e4 2200#endif /* _SOCK_H */