blob: efab1200cd0e7833a2d8f1fb24ac1d73be749080 [file] [log] [blame]
lh9ed821d2023-04-07 01:36:19 -07001/*
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 *
10 * Authors: Ross Biro
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
43#include <linux/hardirq.h>
44#include <linux/kernel.h>
45#include <linux/list.h>
46#include <linux/list_nulls.h>
47#include <linux/timer.h>
48#include <linux/cache.h>
49#include <linux/lockdep.h>
50#include <linux/netdevice.h>
51#include <linux/skbuff.h> /* struct sk_buff */
52#include <linux/mm.h>
53#include <linux/security.h>
54#include <linux/slab.h>
55#include <linux/uaccess.h>
56#include <linux/memcontrol.h>
57#include <linux/res_counter.h>
58#include <linux/static_key.h>
59#include <linux/aio.h>
60#include <linux/sched.h>
61
62#include <linux/filter.h>
63#include <linux/rculist_nulls.h>
64#include <linux/poll.h>
65
66#include <linux/atomic.h>
67#include <net/dst.h>
68#include <net/checksum.h>
69
70#include <net/SI/sock_track.h>
71
72struct cgroup;
73struct cgroup_subsys;
74#ifdef CONFIG_NET
75int mem_cgroup_sockets_init(struct cgroup *cgrp, struct cgroup_subsys *ss);
76void mem_cgroup_sockets_destroy(struct cgroup *cgrp);
77#else
78static inline
79int mem_cgroup_sockets_init(struct cgroup *cgrp, struct cgroup_subsys *ss)
80{
81 return 0;
82}
83static inline
84void mem_cgroup_sockets_destroy(struct cgroup *cgrp)
85{
86}
87#endif
88/*
89 * This structure really needs to be cleaned up.
90 * Most of it is for TCP, and not used by any of
91 * the other protocols.
92 */
93
94/* Define this to get the SOCK_DBG debugging facility. */
95#define SOCK_DEBUGGING
96#ifdef SOCK_DEBUGGING
97#define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
98 printk(KERN_DEBUG msg); } while (0)
99#else
100/* Validate arguments and do nothing */
101static inline __printf(2, 3)
102void SOCK_DEBUG(struct sock *sk, const char *msg, ...)
103{
104}
105#endif
106
107/* This is the per-socket lock. The spinlock provides a synchronization
108 * between user contexts and software interrupt processing, whereas the
109 * mini-semaphore synchronizes multiple users amongst themselves.
110 */
111typedef struct {
112 spinlock_t slock;
113 int owned;
114 wait_queue_head_t wq;
115 /*
116 * We express the mutex-alike socket_lock semantics
117 * to the lock validator by explicitly managing
118 * the slock as a lock variant (in addition to
119 * the slock itself):
120 */
121#ifdef CONFIG_DEBUG_LOCK_ALLOC
122 struct lockdep_map dep_map;
123#endif
124} socket_lock_t;
125
126struct sock;
127struct proto;
128struct net;
129
130/**
131 * struct sock_common - minimal network layer representation of sockets
132 * @skc_daddr: Foreign IPv4 addr
133 * @skc_rcv_saddr: Bound local IPv4 addr
134 * @skc_hash: hash value used with various protocol lookup tables
135 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
136 * @skc_family: network address family
137 * @skc_state: Connection state
138 * @skc_reuse: %SO_REUSEADDR setting
139 * @skc_bound_dev_if: bound device index if != 0
140 * @skc_bind_node: bind hash linkage for various protocol lookup tables
141 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
142 * @skc_prot: protocol handlers inside a network family
143 * @skc_net: reference to the network namespace of this socket
144 * @skc_node: main hash linkage for various protocol lookup tables
145 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
146 * @skc_tx_queue_mapping: tx queue number for this connection
147 * @skc_refcnt: reference count
148 *
149 * This is the minimal network layer representation of sockets, the header
150 * for struct sock and struct inet_timewait_sock.
151 */
152struct sock_common {
153 /* skc_daddr and skc_rcv_saddr must be grouped :
154 * cf INET_MATCH() and INET_TW_MATCH()
155 */
156 __be32 skc_daddr;
157 __be32 skc_rcv_saddr;
158
159 union {
160 unsigned int skc_hash;
161 __u16 skc_u16hashes[2];
162 };
163 unsigned short skc_family;
164 volatile unsigned char skc_state;
165 unsigned char skc_reuse;
166 int skc_bound_dev_if;
167 union {
168 struct hlist_node skc_bind_node;
169 struct hlist_nulls_node skc_portaddr_node;
170 };
171 struct proto *skc_prot;
172#ifdef CONFIG_NET_NS
173 struct net *skc_net;
174#endif
175 /*
176 * fields between dontcopy_begin/dontcopy_end
177 * are not copied in sock_copy()
178 */
179 /* private: */
180 int skc_dontcopy_begin[0];
181 /* public: */
182 union {
183 struct hlist_node skc_node;
184 struct hlist_nulls_node skc_nulls_node;
185 };
186 int skc_tx_queue_mapping;
187 atomic_t skc_refcnt;
188 /* private: */
189 int skc_dontcopy_end[0];
190 /* public: */
191};
192
193struct cg_proto;
194/**
195 * struct sock - network layer representation of sockets
196 * @__sk_common: shared layout with inet_timewait_sock
197 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
198 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
199 * @sk_lock: synchronizer
200 * @sk_rcvbuf: size of receive buffer in bytes
201 * @sk_wq: sock wait queue and async head
202 * @sk_dst_cache: destination cache
203 * @sk_dst_lock: destination cache lock
204 * @sk_policy: flow policy
205 * @sk_receive_queue: incoming packets
206 * @sk_wmem_alloc: transmit queue bytes committed
207 * @sk_write_queue: Packet sending queue
208 * @sk_async_wait_queue: DMA copied packets
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
214 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
215 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
216 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
217 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
218 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
219 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
220 * @sk_gso_max_size: Maximum GSO segment size to build
221 * @sk_gso_max_segs: Maximum number of GSO segments
222 * @sk_lingertime: %SO_LINGER l_linger setting
223 * @sk_backlog: always used with the per-socket spinlock held
224 * @sk_callback_lock: used with the callbacks in the end of this struct
225 * @sk_error_queue: rarely used
226 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
227 * IPV6_ADDRFORM for instance)
228 * @sk_err: last error
229 * @sk_err_soft: errors that don't cause failure but are the cause of a
230 * persistent failure not just 'timed out'
231 * @sk_drops: raw/udp drops counter
232 * @sk_ack_backlog: current listen backlog
233 * @sk_max_ack_backlog: listen backlog set in listen()
234 * @sk_priority: %SO_PRIORITY setting
235 * @sk_cgrp_prioidx: socket group's priority map index
236 * @sk_type: socket type (%SOCK_STREAM, etc)
237 * @sk_protocol: which protocol this socket belongs in this network family
238 * @sk_peer_pid: &struct pid for this socket's peer
239 * @sk_peer_cred: %SO_PEERCRED setting
240 * @sk_rcvlowat: %SO_RCVLOWAT setting
241 * @sk_rcvtimeo: %SO_RCVTIMEO setting
242 * @sk_sndtimeo: %SO_SNDTIMEO setting
243 * @sk_rxhash: flow hash received from netif layer
244 * @sk_filter: socket filtering instructions
245 * @sk_protinfo: private area, net family specific, when not using slab
246 * @sk_timer: sock cleanup timer
247 * @sk_stamp: time stamp of last packet received
248 * @sk_socket: Identd and reporting IO signals
249 * @sk_user_data: RPC layer private data
250 * @sk_sndmsg_page: cached page for sendmsg
251 * @sk_sndmsg_off: cached offset for sendmsg
252 * @sk_peek_off: current peek_offset value
253 * @sk_send_head: front of stuff to transmit
254 * @sk_security: used by security modules
255 * @sk_mark: generic packet mark
256 * @sk_classid: this socket's cgroup classid
257 * @sk_cgrp: this socket's cgroup-specific proto data
258 * @sk_write_pending: a write to stream socket waits to start
259 * @sk_state_change: callback to indicate change in the state of the sock
260 * @sk_data_ready: callback to indicate there is data to be processed
261 * @sk_write_space: callback to indicate there is bf sending space available
262 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
263 * @sk_backlog_rcv: callback to process the backlog
264 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
265 */
266struct sock {
267 /*
268 * Now struct inet_timewait_sock also uses sock_common, so please just
269 * don't add nothing before this first member (__sk_common) --acme
270 */
271 struct sock_common __sk_common;
272#define sk_node __sk_common.skc_node
273#define sk_nulls_node __sk_common.skc_nulls_node
274#define sk_refcnt __sk_common.skc_refcnt
275#define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
276
277#define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
278#define sk_dontcopy_end __sk_common.skc_dontcopy_end
279#define sk_hash __sk_common.skc_hash
280#define sk_family __sk_common.skc_family
281#define sk_state __sk_common.skc_state
282#define sk_reuse __sk_common.skc_reuse
283#define sk_bound_dev_if __sk_common.skc_bound_dev_if
284#define sk_bind_node __sk_common.skc_bind_node
285#define sk_prot __sk_common.skc_prot
286#define sk_net __sk_common.skc_net
287 socket_lock_t sk_lock;
288 struct sk_buff_head sk_receive_queue;
289 /*
290 * The backlog queue is special, it is always used with
291 * the per-socket spinlock held and requires low latency
292 * access. Therefore we special case it's implementation.
293 * Note : rmem_alloc is in this structure to fill a hole
294 * on 64bit arches, not because its logically part of
295 * backlog.
296 */
297 struct {
298 atomic_t rmem_alloc;
299 int len;
300 struct sk_buff *head;
301 struct sk_buff *tail;
302 } sk_backlog;
303#define sk_rmem_alloc sk_backlog.rmem_alloc
304 int sk_forward_alloc;
305#ifdef CONFIG_RPS
306 __u32 sk_rxhash;
307#endif
308 atomic_t sk_drops;
309 int sk_rcvbuf;
310 struct sk_filter __rcu *sk_filter;
311 struct socket_wq __rcu *sk_wq;
312
313#ifdef CONFIG_NET_DMA
314 struct sk_buff_head sk_async_wait_queue;
315#endif
316
317#ifdef CONFIG_XFRM
318 struct xfrm_policy *sk_policy[2];
319#endif
320 unsigned long sk_flags;
321 struct dst_entry *sk_dst_cache;
322 spinlock_t sk_dst_lock;
323 atomic_t sk_wmem_alloc;
324 atomic_t sk_omem_alloc;
325 int sk_sndbuf;
326 struct sk_buff_head sk_write_queue;
327 kmemcheck_bitfield_begin(flags);
328 unsigned int sk_shutdown : 2,
329 sk_no_check : 2,
330 sk_userlocks : 4,
331 sk_protocol : 8,
332#define SK_PROTOCOL_MAX U8_MAX //CVE-2015-8543
333 sk_type : 16;
334 kmemcheck_bitfield_end(flags);
335 int sk_wmem_queued;
336 gfp_t sk_allocation;
337 netdev_features_t sk_route_caps;
338 netdev_features_t sk_route_nocaps;
339 int sk_gso_type;
340 unsigned int sk_gso_max_size;
341 u16 sk_gso_max_segs;
342 int sk_rcvlowat;
343 unsigned long sk_lingertime;
344 struct sk_buff_head sk_error_queue;
345 struct proto *sk_prot_creator;
346 rwlock_t sk_callback_lock;
347 int sk_err,
348 sk_err_soft;
349 unsigned short sk_ack_backlog;
350 unsigned short sk_max_ack_backlog;
351 __u32 sk_priority;
352#ifdef CONFIG_CGROUPS
353 __u32 sk_cgrp_prioidx;
354#endif
355 struct pid *sk_peer_pid;
356 const struct cred *sk_peer_cred;
357 long sk_rcvtimeo;
358 long sk_sndtimeo;
359 void *sk_protinfo;
360 struct timer_list sk_timer;
361 ktime_t sk_stamp;
362 struct socket *sk_socket;
363 void *sk_user_data;
364 struct page *sk_sndmsg_page;
365 struct sk_buff *sk_send_head;
366 __u32 sk_sndmsg_off;
367 __s32 sk_peek_off;
368 int sk_write_pending;
369#ifdef CONFIG_SECURITY
370 void *sk_security;
371#endif
372 __u32 sk_mark;
373 u32 sk_classid;
374 struct cg_proto *sk_cgrp;
375 unsigned long sk_send_sum; //µ½Ä¿Ç°ÎªÖ¹·¢ËͳöÈ¥µÄÊý¾Ý°ü×ܸöÊý
376 int conn_list_init; //¼Ç¼µÄconnÁ´±íÊÇ·ñ³õʼ»¯
377 struct list_head conn_head; //connÁ´±í
378 void (*sk_state_change)(struct sock *sk);
379 void (*sk_data_ready)(struct sock *sk, int bytes);
380 void (*sk_write_space)(struct sock *sk);
381 void (*sk_error_report)(struct sock *sk);
382 int (*sk_backlog_rcv)(struct sock *sk,
383 struct sk_buff *skb);
384 void (*sk_destruct)(struct sock *sk);
385};
386
387static inline int sk_peek_offset(struct sock *sk, int flags)
388{
389 if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0))
390 return sk->sk_peek_off;
391 else
392 return 0;
393}
394
395static inline void sk_peek_offset_bwd(struct sock *sk, int val)
396{
397 if (sk->sk_peek_off >= 0) {
398 if (sk->sk_peek_off >= val)
399 sk->sk_peek_off -= val;
400 else
401 sk->sk_peek_off = 0;
402 }
403}
404
405static inline void sk_peek_offset_fwd(struct sock *sk, int val)
406{
407 if (sk->sk_peek_off >= 0)
408 sk->sk_peek_off += val;
409}
410
411/*
412 * Hashed lists helper routines
413 */
414static inline struct sock *sk_entry(const struct hlist_node *node)
415{
416 return hlist_entry(node, struct sock, sk_node);
417}
418
419static inline struct sock *__sk_head(const struct hlist_head *head)
420{
421 return hlist_entry(head->first, struct sock, sk_node);
422}
423
424static inline struct sock *sk_head(const struct hlist_head *head)
425{
426 return hlist_empty(head) ? NULL : __sk_head(head);
427}
428
429static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
430{
431 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
432}
433
434static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
435{
436 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
437}
438
439static inline struct sock *sk_next(const struct sock *sk)
440{
441 return sk->sk_node.next ?
442 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
443}
444
445static inline struct sock *sk_nulls_next(const struct sock *sk)
446{
447 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
448 hlist_nulls_entry(sk->sk_nulls_node.next,
449 struct sock, sk_nulls_node) :
450 NULL;
451}
452
453static inline int sk_unhashed(const struct sock *sk)
454{
455 return hlist_unhashed(&sk->sk_node);
456}
457
458static inline int sk_hashed(const struct sock *sk)
459{
460 return !sk_unhashed(sk);
461}
462
463static __inline__ void sk_node_init(struct hlist_node *node)
464{
465 node->pprev = NULL;
466}
467
468static __inline__ void sk_nulls_node_init(struct hlist_nulls_node *node)
469{
470 node->pprev = NULL;
471}
472
473static __inline__ void __sk_del_node(struct sock *sk)
474{
475 __hlist_del(&sk->sk_node);
476}
477
478/* NB: equivalent to hlist_del_init_rcu */
479static __inline__ int __sk_del_node_init(struct sock *sk)
480{
481 if (sk_hashed(sk)) {
482 __sk_del_node(sk);
483 sk_node_init(&sk->sk_node);
484 return 1;
485 }
486 return 0;
487}
488
489/* Grab socket reference count. This operation is valid only
490 when sk is ALREADY grabbed f.e. it is found in hash table
491 or a list and the lookup is made under lock preventing hash table
492 modifications.
493 */
494
495static inline void sock_hold(struct sock *sk)
496{
497 atomic_inc(&sk->sk_refcnt);
498}
499
500/* Ungrab socket in the context, which assumes that socket refcnt
501 cannot hit zero, f.e. it is true in context of any socketcall.
502 */
503static inline void __sock_put(struct sock *sk)
504{
505 atomic_dec(&sk->sk_refcnt);
506}
507
508static __inline__ int sk_del_node_init(struct sock *sk)
509{
510 int rc = __sk_del_node_init(sk);
511
512 if (rc) {
513 /* paranoid for a while -acme */
514 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
515 __sock_put(sk);
516 }
517 return rc;
518}
519#define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
520
521static __inline__ int __sk_nulls_del_node_init_rcu(struct sock *sk)
522{
523 if (sk_hashed(sk)) {
524 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
525 return 1;
526 }
527 return 0;
528}
529
530static __inline__ int sk_nulls_del_node_init_rcu(struct sock *sk)
531{
532 int rc = __sk_nulls_del_node_init_rcu(sk);
533
534 if (rc) {
535 /* paranoid for a while -acme */
536 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
537 __sock_put(sk);
538 }
539 return rc;
540}
541
542static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
543{
544 hlist_add_head(&sk->sk_node, list);
545}
546
547static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
548{
549 sock_hold(sk);
550 __sk_add_node(sk, list);
551}
552
553static __inline__ void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
554{
555 sock_hold(sk);
556 hlist_add_head_rcu(&sk->sk_node, list);
557}
558
559static __inline__ void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
560{
561 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
562}
563
564static __inline__ void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
565{
566 sock_hold(sk);
567 __sk_nulls_add_node_rcu(sk, list);
568}
569
570static __inline__ void __sk_del_bind_node(struct sock *sk)
571{
572 __hlist_del(&sk->sk_bind_node);
573}
574
575static __inline__ void sk_add_bind_node(struct sock *sk,
576 struct hlist_head *list)
577{
578 hlist_add_head(&sk->sk_bind_node, list);
579}
580
581#define sk_for_each(__sk, node, list) \
582 hlist_for_each_entry(__sk, node, list, sk_node)
583#define sk_for_each_rcu(__sk, node, list) \
584 hlist_for_each_entry_rcu(__sk, node, list, sk_node)
585#define sk_nulls_for_each(__sk, node, list) \
586 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
587#define sk_nulls_for_each_rcu(__sk, node, list) \
588 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
589#define sk_for_each_from(__sk, node) \
590 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
591 hlist_for_each_entry_from(__sk, node, sk_node)
592#define sk_nulls_for_each_from(__sk, node) \
593 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
594 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
595#define sk_for_each_safe(__sk, node, tmp, list) \
596 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
597#define sk_for_each_bound(__sk, node, list) \
598 hlist_for_each_entry(__sk, node, list, sk_bind_node)
599
600/* Sock flags */
601enum sock_flags {
602 SOCK_DEAD,
603 SOCK_DONE,
604 SOCK_URGINLINE,
605 SOCK_KEEPOPEN,
606 SOCK_LINGER,
607 SOCK_DESTROY,
608 SOCK_BROADCAST,
609 SOCK_TIMESTAMP,
610 SOCK_ZAPPED,
611 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
612 SOCK_DBG, /* %SO_DEBUG setting */
613 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
614 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
615 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
616 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
617 SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */
618 SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
619 SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */
620 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
621 SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */
622 SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
623 SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
624 SOCK_FASYNC, /* fasync() active */
625 SOCK_RXQ_OVFL,
626 SOCK_ZEROCOPY, /* buffers from userspace */
627 SOCK_WIFI_STATUS, /* push wifi status to userspace */
628 SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
629 * Will use last 4 bytes of packet sent from
630 * user-space instead.
631 */
632};
633
634static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
635{
636 nsk->sk_flags = osk->sk_flags;
637}
638
639static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
640{
641 __set_bit(flag, &sk->sk_flags);
642}
643
644static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
645{
646 __clear_bit(flag, &sk->sk_flags);
647}
648
649static inline int sock_flag(struct sock *sk, enum sock_flags flag)
650{
651 return test_bit(flag, &sk->sk_flags);
652}
653
654static inline void sk_acceptq_removed(struct sock *sk)
655{
656 sk->sk_ack_backlog--;
657}
658
659static inline void sk_acceptq_added(struct sock *sk)
660{
661 sk->sk_ack_backlog++;
662}
663
664static inline int sk_acceptq_is_full(struct sock *sk)
665{
666 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
667}
668
669/*
670 * Compute minimal free write space needed to queue new packets.
671 */
672static inline int sk_stream_min_wspace(struct sock *sk)
673{
674 return sk->sk_wmem_queued >> 1;
675}
676
677static inline int sk_stream_wspace(struct sock *sk)
678{
679 return sk->sk_sndbuf - sk->sk_wmem_queued;
680}
681
682extern void sk_stream_write_space(struct sock *sk);
683
684static inline int sk_stream_memory_free(struct sock *sk)
685{
686 return sk->sk_wmem_queued < sk->sk_sndbuf;
687}
688
689/* OOB backlog add */
690static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
691{
692 /* dont let skb dst not refcounted, we are going to leave rcu lock */
693 skb_dst_force(skb);
694
695 if (!sk->sk_backlog.tail)
696 sk->sk_backlog.head = skb;
697 else
698 sk->sk_backlog.tail->next = skb;
699
700 sk->sk_backlog.tail = skb;
701 skb->next = NULL;
702}
703
704/*
705 * Take into account size of receive queue and backlog queue
706 * Do not take into account this skb truesize,
707 * to allow even a single big packet to come.
708 */
709static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb)
710{
711 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
712
713 return qsize > sk->sk_rcvbuf;
714}
715
716/* The per-socket spinlock must be held here. */
717static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb)
718{
719 if (sk_rcvqueues_full(sk, skb))
720 {
721 TCP_SOCK_TRACK(sk, TCP_RECV_BUFF_FULL);
722 return -ENOBUFS;
723 }
724
725 __sk_add_backlog(sk, skb);
726 sk->sk_backlog.len += skb->truesize;
727 return 0;
728}
729
730static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
731{
732 return sk->sk_backlog_rcv(sk, skb);
733}
734
735static inline void sock_rps_record_flow(const struct sock *sk)
736{
737#ifdef CONFIG_RPS
738 struct rps_sock_flow_table *sock_flow_table;
739
740 rcu_read_lock();
741 sock_flow_table = rcu_dereference(rps_sock_flow_table);
742 rps_record_sock_flow(sock_flow_table, sk->sk_rxhash);
743 rcu_read_unlock();
744#endif
745}
746
747static inline void sock_rps_reset_flow(const struct sock *sk)
748{
749#ifdef CONFIG_RPS
750 struct rps_sock_flow_table *sock_flow_table;
751
752 rcu_read_lock();
753 sock_flow_table = rcu_dereference(rps_sock_flow_table);
754 rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash);
755 rcu_read_unlock();
756#endif
757}
758
759static inline void sock_rps_save_rxhash(struct sock *sk,
760 const struct sk_buff *skb)
761{
762#ifdef CONFIG_RPS
763 if (unlikely(sk->sk_rxhash != skb->rxhash)) {
764 sock_rps_reset_flow(sk);
765 sk->sk_rxhash = skb->rxhash;
766 }
767#endif
768}
769
770static inline void sock_rps_reset_rxhash(struct sock *sk)
771{
772#ifdef CONFIG_RPS
773 sock_rps_reset_flow(sk);
774 sk->sk_rxhash = 0;
775#endif
776}
777
778#define sk_wait_event(__sk, __timeo, __condition) \
779 ({ int __rc; \
780 release_sock(__sk); \
781 __rc = __condition; \
782 if (!__rc) { \
783 *(__timeo) = schedule_timeout(*(__timeo)); \
784 } \
785 lock_sock(__sk); \
786 __rc = __condition; \
787 __rc; \
788 })
789
790extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
791extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
792extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
793extern int sk_stream_error(struct sock *sk, int flags, int err);
794extern void sk_stream_kill_queues(struct sock *sk);
795
796extern int sk_wait_data(struct sock *sk, long *timeo);
797
798struct request_sock_ops;
799struct timewait_sock_ops;
800struct inet_hashinfo;
801struct raw_hashinfo;
802struct module;
803
804/*
805 * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
806 * un-modified. Special care is taken when initializing object to zero.
807 */
808static inline void sk_prot_clear_nulls(struct sock *sk, int size)
809{
810 if (offsetof(struct sock, sk_node.next) != 0)
811 memset(sk, 0, offsetof(struct sock, sk_node.next));
812 memset(&sk->sk_node.pprev, 0,
813 size - offsetof(struct sock, sk_node.pprev));
814}
815
816/* Networking protocol blocks we attach to sockets.
817 * socket layer -> transport layer interface
818 * transport -> network interface is defined by struct inet_proto
819 */
820struct proto {
821 void (*close)(struct sock *sk,
822 long timeout);
823 int (*connect)(struct sock *sk,
824 struct sockaddr *uaddr,
825 int addr_len);
826 int (*disconnect)(struct sock *sk, int flags);
827
828 struct sock * (*accept) (struct sock *sk, int flags, int *err);
829
830 int (*ioctl)(struct sock *sk, int cmd,
831 unsigned long arg);
832 int (*init)(struct sock *sk);
833 void (*destroy)(struct sock *sk);
834 void (*shutdown)(struct sock *sk, int how);
835 int (*setsockopt)(struct sock *sk, int level,
836 int optname, char __user *optval,
837 unsigned int optlen);
838 int (*getsockopt)(struct sock *sk, int level,
839 int optname, char __user *optval,
840 int __user *option);
841#ifdef CONFIG_COMPAT
842 int (*compat_setsockopt)(struct sock *sk,
843 int level,
844 int optname, char __user *optval,
845 unsigned int optlen);
846 int (*compat_getsockopt)(struct sock *sk,
847 int level,
848 int optname, char __user *optval,
849 int __user *option);
850 int (*compat_ioctl)(struct sock *sk,
851 unsigned int cmd, unsigned long arg);
852#endif
853 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
854 struct msghdr *msg, size_t len);
855 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
856 struct msghdr *msg,
857 size_t len, int noblock, int flags,
858 int *addr_len);
859 int (*sendpage)(struct sock *sk, struct page *page,
860 int offset, size_t size, int flags);
861 int (*bind)(struct sock *sk,
862 struct sockaddr *uaddr, int addr_len);
863
864 int (*backlog_rcv) (struct sock *sk,
865 struct sk_buff *skb);
866
867 /* Keeping track of sk's, looking them up, and port selection methods. */
868 void (*hash)(struct sock *sk);
869 void (*unhash)(struct sock *sk);
870 void (*rehash)(struct sock *sk);
871 int (*get_port)(struct sock *sk, unsigned short snum);
872 void (*clear_sk)(struct sock *sk, int size);
873
874 /* Keeping track of sockets in use */
875#ifdef CONFIG_PROC_FS
876 unsigned int inuse_idx;
877#endif
878
879 /* Memory pressure */
880 void (*enter_memory_pressure)(struct sock *sk);
881 atomic_long_t *memory_allocated; /* Current allocated memory. */
882 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
883 /*
884 * Pressure flag: try to collapse.
885 * Technical note: it is used by multiple contexts non atomically.
886 * All the __sk_mem_schedule() is of this nature: accounting
887 * is strict, actions are advisory and have some latency.
888 */
889 int *memory_pressure;
890 long *sysctl_mem;
891 int *sysctl_wmem;
892 int *sysctl_rmem;
893 int max_header;
894 bool no_autobind;
895
896 struct kmem_cache *slab;
897 unsigned int obj_size;
898 int slab_flags;
899
900 struct percpu_counter *orphan_count;
901
902 struct request_sock_ops *rsk_prot;
903 struct timewait_sock_ops *twsk_prot;
904
905 union {
906 struct inet_hashinfo *hashinfo;
907 struct udp_table *udp_table;
908 struct raw_hashinfo *raw_hash;
909 } h;
910
911 struct module *owner;
912
913 char name[32];
914
915 struct list_head node;
916#ifdef SOCK_REFCNT_DEBUG
917 atomic_t socks;
918#endif
919#ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
920 /*
921 * cgroup specific init/deinit functions. Called once for all
922 * protocols that implement it, from cgroups populate function.
923 * This function has to setup any files the protocol want to
924 * appear in the kmem cgroup filesystem.
925 */
926 int (*init_cgroup)(struct cgroup *cgrp,
927 struct cgroup_subsys *ss);
928 void (*destroy_cgroup)(struct cgroup *cgrp);
929 struct cg_proto *(*proto_cgroup)(struct mem_cgroup *memcg);
930#endif
931};
932
933struct cg_proto {
934 void (*enter_memory_pressure)(struct sock *sk);
935 struct res_counter *memory_allocated; /* Current allocated memory. */
936 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
937 int *memory_pressure;
938 long *sysctl_mem;
939 /*
940 * memcg field is used to find which memcg we belong directly
941 * Each memcg struct can hold more than one cg_proto, so container_of
942 * won't really cut.
943 *
944 * The elegant solution would be having an inverse function to
945 * proto_cgroup in struct proto, but that means polluting the structure
946 * for everybody, instead of just for memcg users.
947 */
948 struct mem_cgroup *memcg;
949};
950
951extern int proto_register(struct proto *prot, int alloc_slab);
952extern void proto_unregister(struct proto *prot);
953
954#ifdef SOCK_REFCNT_DEBUG
955static inline void sk_refcnt_debug_inc(struct sock *sk)
956{
957 atomic_inc(&sk->sk_prot->socks);
958}
959
960static inline void sk_refcnt_debug_dec(struct sock *sk)
961{
962 atomic_dec(&sk->sk_prot->socks);
963 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
964 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
965}
966
967static inline void sk_refcnt_debug_release(const struct sock *sk)
968{
969 if (atomic_read(&sk->sk_refcnt) != 1)
970 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
971 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
972}
973#else /* SOCK_REFCNT_DEBUG */
974#define sk_refcnt_debug_inc(sk) do { } while (0)
975#define sk_refcnt_debug_dec(sk) do { } while (0)
976#define sk_refcnt_debug_release(sk) do { } while (0)
977#endif /* SOCK_REFCNT_DEBUG */
978
979#if defined(CONFIG_CGROUP_MEM_RES_CTLR_KMEM) && defined(CONFIG_NET)
980extern struct static_key memcg_socket_limit_enabled;
981static inline struct cg_proto *parent_cg_proto(struct proto *proto,
982 struct cg_proto *cg_proto)
983{
984 return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg));
985}
986#define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
987#else
988#define mem_cgroup_sockets_enabled 0
989static inline struct cg_proto *parent_cg_proto(struct proto *proto,
990 struct cg_proto *cg_proto)
991{
992 return NULL;
993}
994#endif
995
996
997static inline bool sk_has_memory_pressure(const struct sock *sk)
998{
999 return sk->sk_prot->memory_pressure != NULL;
1000}
1001
1002static inline bool sk_under_memory_pressure(const struct sock *sk)
1003{
1004 if (!sk->sk_prot->memory_pressure)
1005 return false;
1006
1007 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1008 return !!*sk->sk_cgrp->memory_pressure;
1009
1010 return !!*sk->sk_prot->memory_pressure;
1011}
1012
1013static inline void sk_leave_memory_pressure(struct sock *sk)
1014{
1015 int *memory_pressure = sk->sk_prot->memory_pressure;
1016
1017 if (!memory_pressure)
1018 return;
1019
1020 if (*memory_pressure)
1021 *memory_pressure = 0;
1022
1023 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1024 struct cg_proto *cg_proto = sk->sk_cgrp;
1025 struct proto *prot = sk->sk_prot;
1026
1027 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1028 if (*cg_proto->memory_pressure)
1029 *cg_proto->memory_pressure = 0;
1030 }
1031
1032}
1033
1034static inline void sk_enter_memory_pressure(struct sock *sk)
1035{
1036 if (!sk->sk_prot->enter_memory_pressure)
1037 return;
1038
1039 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1040 struct cg_proto *cg_proto = sk->sk_cgrp;
1041 struct proto *prot = sk->sk_prot;
1042
1043 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1044 cg_proto->enter_memory_pressure(sk);
1045 }
1046
1047 sk->sk_prot->enter_memory_pressure(sk);
1048}
1049
1050static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1051{
1052 long *prot = sk->sk_prot->sysctl_mem;
1053 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1054 prot = sk->sk_cgrp->sysctl_mem;
1055 return prot[index];
1056}
1057
1058static inline void memcg_memory_allocated_add(struct cg_proto *prot,
1059 unsigned long amt,
1060 int *parent_status)
1061{
1062 struct res_counter *fail;
1063 int ret;
1064
1065 ret = res_counter_charge_nofail(prot->memory_allocated,
1066 amt << PAGE_SHIFT, &fail);
1067 if (ret < 0)
1068 *parent_status = OVER_LIMIT;
1069}
1070
1071static inline void memcg_memory_allocated_sub(struct cg_proto *prot,
1072 unsigned long amt)
1073{
1074 res_counter_uncharge(prot->memory_allocated, amt << PAGE_SHIFT);
1075}
1076
1077static inline u64 memcg_memory_allocated_read(struct cg_proto *prot)
1078{
1079 u64 ret;
1080 ret = res_counter_read_u64(prot->memory_allocated, RES_USAGE);
1081 return ret >> PAGE_SHIFT;
1082}
1083
1084static inline long
1085sk_memory_allocated(const struct sock *sk)
1086{
1087 struct proto *prot = sk->sk_prot;
1088 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1089 return memcg_memory_allocated_read(sk->sk_cgrp);
1090
1091 return atomic_long_read(prot->memory_allocated);
1092}
1093
1094static inline long
1095sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status)
1096{
1097 struct proto *prot = sk->sk_prot;
1098
1099 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1100 memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status);
1101 /* update the root cgroup regardless */
1102 atomic_long_add_return(amt, prot->memory_allocated);
1103 return memcg_memory_allocated_read(sk->sk_cgrp);
1104 }
1105
1106 return atomic_long_add_return(amt, prot->memory_allocated);
1107}
1108
1109static inline void
1110sk_memory_allocated_sub(struct sock *sk, int amt)
1111{
1112 struct proto *prot = sk->sk_prot;
1113
1114 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1115 memcg_memory_allocated_sub(sk->sk_cgrp, amt);
1116
1117 atomic_long_sub(amt, prot->memory_allocated);
1118}
1119
1120static inline void sk_sockets_allocated_dec(struct sock *sk)
1121{
1122 struct proto *prot = sk->sk_prot;
1123
1124 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1125 struct cg_proto *cg_proto = sk->sk_cgrp;
1126
1127 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1128 percpu_counter_dec(cg_proto->sockets_allocated);
1129 }
1130
1131 percpu_counter_dec(prot->sockets_allocated);
1132}
1133
1134static inline void sk_sockets_allocated_inc(struct sock *sk)
1135{
1136 struct proto *prot = sk->sk_prot;
1137
1138 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1139 struct cg_proto *cg_proto = sk->sk_cgrp;
1140
1141 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1142 percpu_counter_inc(cg_proto->sockets_allocated);
1143 }
1144
1145 percpu_counter_inc(prot->sockets_allocated);
1146}
1147
1148static inline int
1149sk_sockets_allocated_read_positive(struct sock *sk)
1150{
1151 struct proto *prot = sk->sk_prot;
1152
1153 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1154 return percpu_counter_read_positive(sk->sk_cgrp->sockets_allocated);
1155
1156 return percpu_counter_read_positive(prot->sockets_allocated);
1157}
1158
1159static inline int
1160proto_sockets_allocated_sum_positive(struct proto *prot)
1161{
1162 return percpu_counter_sum_positive(prot->sockets_allocated);
1163}
1164
1165static inline long
1166proto_memory_allocated(struct proto *prot)
1167{
1168 return atomic_long_read(prot->memory_allocated);
1169}
1170
1171static inline bool
1172proto_memory_pressure(struct proto *prot)
1173{
1174 if (!prot->memory_pressure)
1175 return false;
1176 return !!*prot->memory_pressure;
1177}
1178
1179
1180#ifdef CONFIG_PROC_FS
1181/* Called with local bh disabled */
1182extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1183extern int sock_prot_inuse_get(struct net *net, struct proto *proto);
1184#else
1185static void inline sock_prot_inuse_add(struct net *net, struct proto *prot,
1186 int inc)
1187{
1188}
1189#endif
1190
1191
1192/* With per-bucket locks this operation is not-atomic, so that
1193 * this version is not worse.
1194 */
1195static inline void __sk_prot_rehash(struct sock *sk)
1196{
1197 sk->sk_prot->unhash(sk);
1198 sk->sk_prot->hash(sk);
1199}
1200
1201void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
1202
1203/* About 10 seconds */
1204#define SOCK_DESTROY_TIME (10*HZ)
1205
1206/* Sockets 0-1023 can't be bound to unless you are superuser */
1207#define PROT_SOCK 1024
1208
1209#define SHUTDOWN_MASK 3
1210#define RCV_SHUTDOWN 1
1211#define SEND_SHUTDOWN 2
1212
1213#define SOCK_SNDBUF_LOCK 1
1214#define SOCK_RCVBUF_LOCK 2
1215#define SOCK_BINDADDR_LOCK 4
1216#define SOCK_BINDPORT_LOCK 8
1217
1218/* sock_iocb: used to kick off async processing of socket ios */
1219struct sock_iocb {
1220 struct list_head list;
1221
1222 int flags;
1223 int size;
1224 struct socket *sock;
1225 struct sock *sk;
1226 struct scm_cookie *scm;
1227 struct msghdr *msg, async_msg;
1228 struct kiocb *kiocb;
1229};
1230
1231static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
1232{
1233 return (struct sock_iocb *)iocb->private;
1234}
1235
1236static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
1237{
1238 return si->kiocb;
1239}
1240
1241struct socket_alloc {
1242 struct socket socket;
1243 struct inode vfs_inode;
1244};
1245
1246static inline struct socket *SOCKET_I(struct inode *inode)
1247{
1248 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1249}
1250
1251static inline struct inode *SOCK_INODE(struct socket *socket)
1252{
1253 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1254}
1255
1256/*
1257 * Functions for memory accounting
1258 */
1259extern int __sk_mem_schedule(struct sock *sk, int size, int kind);
1260extern void __sk_mem_reclaim(struct sock *sk);
1261
1262#define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1263#define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1264#define SK_MEM_SEND 0
1265#define SK_MEM_RECV 1
1266
1267static inline int sk_mem_pages(int amt)
1268{
1269 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
1270}
1271
1272static inline int sk_has_account(struct sock *sk)
1273{
1274 /* return true if protocol supports memory accounting */
1275 return !!sk->sk_prot->memory_allocated;
1276}
1277
1278static inline int sk_wmem_schedule(struct sock *sk, int size)
1279{
1280 if (!sk_has_account(sk))
1281 return 1;
1282 return size <= sk->sk_forward_alloc ||
1283 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1284}
1285
1286static inline int sk_rmem_schedule(struct sock *sk, int size)
1287{
1288 if (!sk_has_account(sk))
1289 return 1;
1290 return size <= sk->sk_forward_alloc ||
1291 __sk_mem_schedule(sk, size, SK_MEM_RECV);
1292}
1293
1294static inline void sk_mem_reclaim(struct sock *sk)
1295{
1296 if (!sk_has_account(sk))
1297 return;
1298 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1299 __sk_mem_reclaim(sk);
1300}
1301
1302static inline void sk_mem_reclaim_partial(struct sock *sk)
1303{
1304 if (!sk_has_account(sk))
1305 return;
1306 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1307 __sk_mem_reclaim(sk);
1308}
1309
1310static inline void sk_mem_charge(struct sock *sk, int size)
1311{
1312 if (!sk_has_account(sk))
1313 return;
1314 sk->sk_forward_alloc -= size;
1315}
1316
1317static inline void sk_mem_uncharge(struct sock *sk, int size)
1318{
1319 if (!sk_has_account(sk))
1320 return;
1321 sk->sk_forward_alloc += size;
1322}
1323
1324static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1325{
1326 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1327 sk->sk_wmem_queued -= skb->truesize;
1328 sk_mem_uncharge(sk, skb->truesize);
1329 __kfree_skb(skb);
1330}
1331
1332/* Used by processes to "lock" a socket state, so that
1333 * interrupts and bottom half handlers won't change it
1334 * from under us. It essentially blocks any incoming
1335 * packets, so that we won't get any new data or any
1336 * packets that change the state of the socket.
1337 *
1338 * While locked, BH processing will add new packets to
1339 * the backlog queue. This queue is processed by the
1340 * owner of the socket lock right before it is released.
1341 *
1342 * Since ~2.3.5 it is also exclusive sleep lock serializing
1343 * accesses from user process context.
1344 */
1345#define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
1346
1347/*
1348 * Macro so as to not evaluate some arguments when
1349 * lockdep is not enabled.
1350 *
1351 * Mark both the sk_lock and the sk_lock.slock as a
1352 * per-address-family lock class.
1353 */
1354#define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1355do { \
1356 sk->sk_lock.owned = 0; \
1357 init_waitqueue_head(&sk->sk_lock.wq); \
1358 spin_lock_init(&(sk)->sk_lock.slock); \
1359 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1360 sizeof((sk)->sk_lock)); \
1361 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1362 (skey), (sname)); \
1363 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1364} while (0)
1365
1366extern void lock_sock_nested(struct sock *sk, int subclass);
1367
1368static inline void lock_sock(struct sock *sk)
1369{
1370 lock_sock_nested(sk, 0);
1371}
1372
1373extern void release_sock(struct sock *sk);
1374
1375/* BH context may only use the following locking interface. */
1376#define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
1377#define bh_lock_sock_nested(__sk) \
1378 spin_lock_nested(&((__sk)->sk_lock.slock), \
1379 SINGLE_DEPTH_NESTING)
1380#define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1381
1382extern bool lock_sock_fast(struct sock *sk);
1383/**
1384 * unlock_sock_fast - complement of lock_sock_fast
1385 * @sk: socket
1386 * @slow: slow mode
1387 *
1388 * fast unlock socket for user context.
1389 * If slow mode is on, we call regular release_sock()
1390 */
1391static inline void unlock_sock_fast(struct sock *sk, bool slow)
1392{
1393 if (slow)
1394 release_sock(sk);
1395 else
1396 spin_unlock_bh(&sk->sk_lock.slock);
1397}
1398
1399
1400extern struct sock *sk_alloc(struct net *net, int family,
1401 gfp_t priority,
1402 struct proto *prot);
1403extern void sk_free(struct sock *sk);
1404extern void sk_release_kernel(struct sock *sk);
1405extern struct sock *sk_clone_lock(const struct sock *sk,
1406 const gfp_t priority);
1407
1408extern struct sk_buff *sock_wmalloc(struct sock *sk,
1409 unsigned long size, int force,
1410 gfp_t priority);
1411extern struct sk_buff *sock_rmalloc(struct sock *sk,
1412 unsigned long size, int force,
1413 gfp_t priority);
1414extern void sock_wfree(struct sk_buff *skb);
1415extern void sock_rfree(struct sk_buff *skb);
1416
1417extern int sock_setsockopt(struct socket *sock, int level,
1418 int op, char __user *optval,
1419 unsigned int optlen);
1420
1421extern int sock_getsockopt(struct socket *sock, int level,
1422 int op, char __user *optval,
1423 int __user *optlen);
1424extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
1425 unsigned long size,
1426 int noblock,
1427 int *errcode);
1428extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
1429 unsigned long header_len,
1430 unsigned long data_len,
1431 int noblock,
1432 int *errcode);
1433extern void *sock_kmalloc(struct sock *sk, int size,
1434 gfp_t priority);
1435extern void sock_kfree_s(struct sock *sk, void *mem, int size);
1436extern void sk_send_sigurg(struct sock *sk);
1437
1438#ifdef CONFIG_CGROUPS
1439extern void sock_update_classid(struct sock *sk);
1440#else
1441static inline void sock_update_classid(struct sock *sk)
1442{
1443}
1444#endif
1445
1446/*
1447 * Functions to fill in entries in struct proto_ops when a protocol
1448 * does not implement a particular function.
1449 */
1450extern int sock_no_bind(struct socket *,
1451 struct sockaddr *, int);
1452extern int sock_no_connect(struct socket *,
1453 struct sockaddr *, int, int);
1454extern int sock_no_socketpair(struct socket *,
1455 struct socket *);
1456extern int sock_no_accept(struct socket *,
1457 struct socket *, int);
1458extern int sock_no_getname(struct socket *,
1459 struct sockaddr *, int *, int);
1460extern unsigned int sock_no_poll(struct file *, struct socket *,
1461 struct poll_table_struct *);
1462extern int sock_no_ioctl(struct socket *, unsigned int,
1463 unsigned long);
1464extern int sock_no_listen(struct socket *, int);
1465extern int sock_no_shutdown(struct socket *, int);
1466extern int sock_no_getsockopt(struct socket *, int , int,
1467 char __user *, int __user *);
1468extern int sock_no_setsockopt(struct socket *, int, int,
1469 char __user *, unsigned int);
1470extern int sock_no_sendmsg(struct kiocb *, struct socket *,
1471 struct msghdr *, size_t);
1472extern int sock_no_recvmsg(struct kiocb *, struct socket *,
1473 struct msghdr *, size_t, int);
1474extern int sock_no_mmap(struct file *file,
1475 struct socket *sock,
1476 struct vm_area_struct *vma);
1477extern ssize_t sock_no_sendpage(struct socket *sock,
1478 struct page *page,
1479 int offset, size_t size,
1480 int flags);
1481
1482/*
1483 * Functions to fill in entries in struct proto_ops when a protocol
1484 * uses the inet style.
1485 */
1486extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
1487 char __user *optval, int __user *optlen);
1488extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1489 struct msghdr *msg, size_t size, int flags);
1490extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
1491 char __user *optval, unsigned int optlen);
1492extern int compat_sock_common_getsockopt(struct socket *sock, int level,
1493 int optname, char __user *optval, int __user *optlen);
1494extern int compat_sock_common_setsockopt(struct socket *sock, int level,
1495 int optname, char __user *optval, unsigned int optlen);
1496
1497extern void sk_common_release(struct sock *sk);
1498
1499/*
1500 * Default socket callbacks and setup code
1501 */
1502
1503/* Initialise core socket variables */
1504extern void sock_init_data(struct socket *sock, struct sock *sk);
1505
1506extern void sk_filter_release_rcu(struct rcu_head *rcu);
1507
1508/**
1509 * sk_filter_release - release a socket filter
1510 * @fp: filter to remove
1511 *
1512 * Remove a filter from a socket and release its resources.
1513 */
1514
1515static inline void sk_filter_release(struct sk_filter *fp)
1516{
1517 if (atomic_dec_and_test(&fp->refcnt))
1518 call_rcu(&fp->rcu, sk_filter_release_rcu);
1519}
1520
1521static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
1522{
1523 unsigned int size = sk_filter_len(fp);
1524
1525 atomic_sub(size, &sk->sk_omem_alloc);
1526 sk_filter_release(fp);
1527}
1528
1529static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
1530{
1531 atomic_inc(&fp->refcnt);
1532 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
1533}
1534
1535/*
1536 * Socket reference counting postulates.
1537 *
1538 * * Each user of socket SHOULD hold a reference count.
1539 * * Each access point to socket (an hash table bucket, reference from a list,
1540 * running timer, skb in flight MUST hold a reference count.
1541 * * When reference count hits 0, it means it will never increase back.
1542 * * When reference count hits 0, it means that no references from
1543 * outside exist to this socket and current process on current CPU
1544 * is last user and may/should destroy this socket.
1545 * * sk_free is called from any context: process, BH, IRQ. When
1546 * it is called, socket has no references from outside -> sk_free
1547 * may release descendant resources allocated by the socket, but
1548 * to the time when it is called, socket is NOT referenced by any
1549 * hash tables, lists etc.
1550 * * Packets, delivered from outside (from network or from another process)
1551 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1552 * when they sit in queue. Otherwise, packets will leak to hole, when
1553 * socket is looked up by one cpu and unhasing is made by another CPU.
1554 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1555 * (leak to backlog). Packet socket does all the processing inside
1556 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1557 * use separate SMP lock, so that they are prone too.
1558 */
1559
1560/* Ungrab socket and destroy it, if it was the last reference. */
1561//ʵÏַŵ½sock.cÖУ¬ÒÔ±ãÌí¼Ó×ÔÑдúÂë
1562void sock_put(struct sock *sk);
1563#if 0
1564static inline void sock_put(struct sock *sk)
1565{
1566 if (atomic_dec_and_test(&sk->sk_refcnt))
1567 sk_free(sk);
1568}
1569#endif
1570extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1571 const int nested);
1572
1573static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1574{
1575 sk->sk_tx_queue_mapping = tx_queue;
1576}
1577
1578static inline void sk_tx_queue_clear(struct sock *sk)
1579{
1580 sk->sk_tx_queue_mapping = -1;
1581}
1582
1583static inline int sk_tx_queue_get(const struct sock *sk)
1584{
1585 return sk ? sk->sk_tx_queue_mapping : -1;
1586}
1587
1588static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1589{
1590 sk_tx_queue_clear(sk);
1591 sk->sk_socket = sock;
1592}
1593
1594static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1595{
1596 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1597 return &rcu_dereference_raw(sk->sk_wq)->wait;
1598}
1599/* Detach socket from process context.
1600 * Announce socket dead, detach it from wait queue and inode.
1601 * Note that parent inode held reference count on this struct sock,
1602 * we do not release it in this function, because protocol
1603 * probably wants some additional cleanups or even continuing
1604 * to work with this socket (TCP).
1605 */
1606static inline void sock_orphan(struct sock *sk)
1607{
1608 write_lock_bh(&sk->sk_callback_lock);
1609 sock_set_flag(sk, SOCK_DEAD);
1610 sk_set_socket(sk, NULL);
1611 sk->sk_wq = NULL;
1612 write_unlock_bh(&sk->sk_callback_lock);
1613}
1614
1615static inline void sock_graft(struct sock *sk, struct socket *parent)
1616{
1617 write_lock_bh(&sk->sk_callback_lock);
1618 sk->sk_wq = parent->wq;
1619 parent->sk = sk;
1620 sk_set_socket(sk, parent);
1621 security_sock_graft(sk, parent);
1622 write_unlock_bh(&sk->sk_callback_lock);
1623}
1624
1625extern int sock_i_uid(struct sock *sk);
1626extern unsigned long sock_i_ino(struct sock *sk);
1627
1628static inline struct dst_entry *
1629__sk_dst_get(struct sock *sk)
1630{
1631 return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
1632 lockdep_is_held(&sk->sk_lock.slock));
1633}
1634
1635static inline struct dst_entry *
1636sk_dst_get(struct sock *sk)
1637{
1638 struct dst_entry *dst;
1639
1640 rcu_read_lock();
1641 dst = rcu_dereference(sk->sk_dst_cache);
1642 if (dst)
1643 dst_hold(dst);
1644 rcu_read_unlock();
1645 return dst;
1646}
1647
1648extern void sk_reset_txq(struct sock *sk);
1649
1650static inline void dst_negative_advice(struct sock *sk)
1651{
1652 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1653
1654 if (dst && dst->ops->negative_advice) {
1655 ndst = dst->ops->negative_advice(dst);
1656
1657 if (ndst != dst) {
1658 rcu_assign_pointer(sk->sk_dst_cache, ndst);
1659 sk_reset_txq(sk);
1660 }
1661 }
1662}
1663
1664static inline void
1665__sk_dst_set(struct sock *sk, struct dst_entry *dst)
1666{
1667 struct dst_entry *old_dst;
1668
1669 sk_tx_queue_clear(sk);
1670 /*
1671 * This can be called while sk is owned by the caller only,
1672 * with no state that can be checked in a rcu_dereference_check() cond
1673 */
1674 old_dst = rcu_dereference_raw(sk->sk_dst_cache);
1675 rcu_assign_pointer(sk->sk_dst_cache, dst);
1676 dst_release(old_dst);
1677}
1678
1679static inline void
1680sk_dst_set(struct sock *sk, struct dst_entry *dst)
1681{
1682 spin_lock(&sk->sk_dst_lock);
1683 __sk_dst_set(sk, dst);
1684 spin_unlock(&sk->sk_dst_lock);
1685}
1686
1687static inline void
1688__sk_dst_reset(struct sock *sk)
1689{
1690 __sk_dst_set(sk, NULL);
1691}
1692
1693static inline void
1694sk_dst_reset(struct sock *sk)
1695{
1696 spin_lock(&sk->sk_dst_lock);
1697 __sk_dst_reset(sk);
1698 spin_unlock(&sk->sk_dst_lock);
1699}
1700
1701extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1702
1703extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1704
1705static inline int sk_can_gso(const struct sock *sk)
1706{
1707 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1708}
1709
1710extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
1711
1712static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
1713{
1714 sk->sk_route_nocaps |= flags;
1715 sk->sk_route_caps &= ~flags;
1716}
1717
1718static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1719 char __user *from, char *to,
1720 int copy, int offset)
1721{
1722 if (skb->ip_summed == CHECKSUM_NONE) {
1723 int err = 0;
1724 __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err);
1725 if (err)
1726 return err;
1727 skb->csum = csum_block_add(skb->csum, csum, offset);
1728 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1729 if (!access_ok(VERIFY_READ, from, copy) ||
1730 __copy_from_user_nocache(to, from, copy))
1731 return -EFAULT;
1732 } else if (copy_from_user(to, from, copy))
1733 return -EFAULT;
1734
1735 return 0;
1736}
1737
1738static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
1739 char __user *from, int copy)
1740{
1741 int err, offset = skb->len;
1742
1743 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1744 copy, offset);
1745 if (err)
1746 __skb_trim(skb, offset);
1747
1748 return err;
1749}
1750
1751static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from,
1752 struct sk_buff *skb,
1753 struct page *page,
1754 int off, int copy)
1755{
1756 int err;
1757
1758 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1759 copy, skb->len);
1760 if (err)
1761 return err;
1762
1763 skb->len += copy;
1764 skb->data_len += copy;
1765 skb->truesize += copy;
1766 sk->sk_wmem_queued += copy;
1767 sk_mem_charge(sk, copy);
1768 return 0;
1769}
1770
1771static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1772 struct sk_buff *skb, struct page *page,
1773 int off, int copy)
1774{
1775 if (skb->ip_summed == CHECKSUM_NONE) {
1776 int err = 0;
1777 __wsum csum = csum_and_copy_from_user(from,
1778 page_address(page) + off,
1779 copy, 0, &err);
1780 if (err)
1781 return err;
1782 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1783 } else if (copy_from_user(page_address(page) + off, from, copy))
1784 return -EFAULT;
1785
1786 skb->len += copy;
1787 skb->data_len += copy;
1788 skb->truesize += copy;
1789 sk->sk_wmem_queued += copy;
1790 sk_mem_charge(sk, copy);
1791 return 0;
1792}
1793
1794/**
1795 * sk_wmem_alloc_get - returns write allocations
1796 * @sk: socket
1797 *
1798 * Returns sk_wmem_alloc minus initial offset of one
1799 */
1800static inline int sk_wmem_alloc_get(const struct sock *sk)
1801{
1802 return atomic_read(&sk->sk_wmem_alloc) - 1;
1803}
1804
1805/**
1806 * sk_rmem_alloc_get - returns read allocations
1807 * @sk: socket
1808 *
1809 * Returns sk_rmem_alloc
1810 */
1811static inline int sk_rmem_alloc_get(const struct sock *sk)
1812{
1813 return atomic_read(&sk->sk_rmem_alloc);
1814}
1815
1816/**
1817 * sk_has_allocations - check if allocations are outstanding
1818 * @sk: socket
1819 *
1820 * Returns true if socket has write or read allocations
1821 */
1822static inline int sk_has_allocations(const struct sock *sk)
1823{
1824 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1825}
1826
1827/**
1828 * wq_has_sleeper - check if there are any waiting processes
1829 * @wq: struct socket_wq
1830 *
1831 * Returns true if socket_wq has waiting processes
1832 *
1833 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
1834 * barrier call. They were added due to the race found within the tcp code.
1835 *
1836 * Consider following tcp code paths:
1837 *
1838 * CPU1 CPU2
1839 *
1840 * sys_select receive packet
1841 * ... ...
1842 * __add_wait_queue update tp->rcv_nxt
1843 * ... ...
1844 * tp->rcv_nxt check sock_def_readable
1845 * ... {
1846 * schedule rcu_read_lock();
1847 * wq = rcu_dereference(sk->sk_wq);
1848 * if (wq && waitqueue_active(&wq->wait))
1849 * wake_up_interruptible(&wq->wait)
1850 * ...
1851 * }
1852 *
1853 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1854 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1855 * could then endup calling schedule and sleep forever if there are no more
1856 * data on the socket.
1857 *
1858 */
1859static inline bool wq_has_sleeper(struct socket_wq *wq)
1860{
1861
1862 /*
1863 * We need to be sure we are in sync with the
1864 * add_wait_queue modifications to the wait queue.
1865 *
1866 * This memory barrier is paired in the sock_poll_wait.
1867 */
1868 smp_mb();
1869 return wq && waitqueue_active(&wq->wait);
1870}
1871
1872/**
1873 * sock_poll_wait - place memory barrier behind the poll_wait call.
1874 * @filp: file
1875 * @wait_address: socket wait queue
1876 * @p: poll_table
1877 *
1878 * See the comments in the wq_has_sleeper function.
1879 */
1880static inline void sock_poll_wait(struct file *filp,
1881 wait_queue_head_t *wait_address, poll_table *p)
1882{
1883 if (!poll_does_not_wait(p) && wait_address) {
1884 poll_wait(filp, wait_address, p);
1885 /*
1886 * We need to be sure we are in sync with the
1887 * socket flags modification.
1888 *
1889 * This memory barrier is paired in the wq_has_sleeper.
1890 */
1891 smp_mb();
1892 }
1893}
1894
1895/*
1896 * Queue a received datagram if it will fit. Stream and sequenced
1897 * protocols can't normally use this as they need to fit buffers in
1898 * and play with them.
1899 *
1900 * Inlined as it's very short and called for pretty much every
1901 * packet ever received.
1902 */
1903
1904static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1905{
1906 skb_orphan(skb);
1907 skb->sk = sk;
1908 skb->destructor = sock_wfree;
1909 /*
1910 * We used to take a refcount on sk, but following operation
1911 * is enough to guarantee sk_free() wont free this sock until
1912 * all in-flight packets are completed
1913 */
1914 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1915}
1916
1917static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1918{
1919 skb_orphan(skb);
1920 skb->sk = sk;
1921 skb->destructor = sock_rfree;
1922 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1923 sk_mem_charge(sk, skb->truesize);
1924}
1925
1926extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1927 unsigned long expires);
1928
1929extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1930
1931extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
1932
1933extern int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
1934
1935/*
1936 * Recover an error report and clear atomically
1937 */
1938
1939static inline int sock_error(struct sock *sk)
1940{
1941 int err;
1942 if (likely(!sk->sk_err))
1943 return 0;
1944 err = xchg(&sk->sk_err, 0);
1945 return -err;
1946}
1947
1948static inline unsigned long sock_wspace(struct sock *sk)
1949{
1950 int amt = 0;
1951
1952 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1953 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1954 if (amt < 0)
1955 amt = 0;
1956 }
1957 return amt;
1958}
1959
1960static inline void sk_wake_async(struct sock *sk, int how, int band)
1961{
1962 if (sock_flag(sk, SOCK_FASYNC))
1963 sock_wake_async(sk->sk_socket, how, band);
1964}
1965
1966#define SOCK_MIN_SNDBUF 2048
1967/*
1968 * Since sk_rmem_alloc sums skb->truesize, even a small frame might need
1969 * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak
1970 */
1971#define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff))
1972
1973static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1974{
1975 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
1976 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
1977 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1978 }
1979}
1980
1981struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
1982
1983static inline struct page *sk_stream_alloc_page(struct sock *sk)
1984{
1985 struct page *page = NULL;
1986
1987 page = alloc_pages(sk->sk_allocation, 0);
1988 if (!page) {
1989 sk_enter_memory_pressure(sk);
1990 sk_stream_moderate_sndbuf(sk);
1991 }
1992 netslab_inc(SOCK_ALLOC_PAGES);
1993 return page;
1994}
1995
1996/*
1997 * Default write policy as shown to user space via poll/select/SIGIO
1998 */
1999static inline int sock_writeable(const struct sock *sk)
2000{
2001 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
2002}
2003
2004static inline gfp_t gfp_any(void)
2005{
2006 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
2007}
2008
2009static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
2010{
2011 return noblock ? 0 : sk->sk_rcvtimeo;
2012}
2013
2014static inline long sock_sndtimeo(const struct sock *sk, int noblock)
2015{
2016 return noblock ? 0 : sk->sk_sndtimeo;
2017}
2018
2019static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2020{
2021 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
2022}
2023
2024/* Alas, with timeout socket operations are not restartable.
2025 * Compare this to poll().
2026 */
2027static inline int sock_intr_errno(long timeo)
2028{
2029 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2030}
2031
2032extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2033 struct sk_buff *skb);
2034extern void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2035 struct sk_buff *skb);
2036
2037static __inline__ void
2038sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2039{
2040 ktime_t kt = skb->tstamp;
2041 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
2042
2043 /*
2044 * generate control messages if
2045 * - receive time stamping in software requested (SOCK_RCVTSTAMP
2046 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
2047 * - software time stamp available and wanted
2048 * (SOCK_TIMESTAMPING_SOFTWARE)
2049 * - hardware time stamps available and wanted
2050 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
2051 * SOCK_TIMESTAMPING_RAW_HARDWARE)
2052 */
2053 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
2054 sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
2055 (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
2056 (hwtstamps->hwtstamp.tv64 &&
2057 sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
2058 (hwtstamps->syststamp.tv64 &&
2059 sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
2060 __sock_recv_timestamp(msg, sk, skb);
2061 else
2062 sk->sk_stamp = kt;
2063
2064 if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2065 __sock_recv_wifi_status(msg, sk, skb);
2066}
2067
2068extern void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2069 struct sk_buff *skb);
2070
2071static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2072 struct sk_buff *skb)
2073{
2074#define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2075 (1UL << SOCK_RCVTSTAMP) | \
2076 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
2077 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
2078 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
2079 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
2080
2081 if (sk->sk_flags & FLAGS_TS_OR_DROPS)
2082 __sock_recv_ts_and_drops(msg, sk, skb);
2083 else
2084 sk->sk_stamp = skb->tstamp;
2085}
2086
2087/**
2088 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
2089 * @sk: socket sending this packet
2090 * @tx_flags: filled with instructions for time stamping
2091 *
2092 * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if
2093 * parameters are invalid.
2094 */
2095extern int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags);
2096
2097/**
2098 * sk_eat_skb - Release a skb if it is no longer needed
2099 * @sk: socket to eat this skb from
2100 * @skb: socket buffer to eat
2101 * @copied_early: flag indicating whether DMA operations copied this data early
2102 *
2103 * This routine must be called with interrupts disabled or with the socket
2104 * locked so that the sk_buff queue operation is ok.
2105*/
2106#ifdef CONFIG_NET_DMA
2107static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
2108{
2109 __skb_unlink(skb, &sk->sk_receive_queue);
2110 if (!copied_early)
2111 __kfree_skb(skb);
2112 else
2113 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
2114}
2115#else
2116static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early)
2117{
2118 __skb_unlink(skb, &sk->sk_receive_queue);
2119 __kfree_skb(skb);
2120}
2121#endif
2122
2123static inline
2124struct net *sock_net(const struct sock *sk)
2125{
2126 return read_pnet(&sk->sk_net);
2127}
2128
2129static inline
2130void sock_net_set(struct sock *sk, struct net *net)
2131{
2132 write_pnet(&sk->sk_net, net);
2133}
2134
2135/*
2136 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
2137 * They should not hold a reference to a namespace in order to allow
2138 * to stop it.
2139 * Sockets after sk_change_net should be released using sk_release_kernel
2140 */
2141static inline void sk_change_net(struct sock *sk, struct net *net)
2142{
2143 put_net(sock_net(sk));
2144 sock_net_set(sk, hold_net(net));
2145}
2146
2147static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2148{
2149 if (unlikely(skb->sk)) {
2150 struct sock *sk = skb->sk;
2151
2152 skb->destructor = NULL;
2153 skb->sk = NULL;
2154 return sk;
2155 }
2156 return NULL;
2157}
2158
2159extern void sock_enable_timestamp(struct sock *sk, int flag);
2160extern int sock_get_timestamp(struct sock *, struct timeval __user *);
2161extern int sock_get_timestampns(struct sock *, struct timespec __user *);
2162
2163/*
2164 * Enable debug/info messages
2165 */
2166extern int net_msg_warn;
2167#define NETDEBUG(fmt, args...) \
2168 do { if (net_msg_warn) printk(fmt,##args); } while (0)
2169
2170#define LIMIT_NETDEBUG(fmt, args...) \
2171 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
2172
2173extern __u32 sysctl_wmem_max;
2174extern __u32 sysctl_rmem_max;
2175
2176extern void sk_init(void);
2177
2178extern int sysctl_optmem_max;
2179
2180extern __u32 sysctl_wmem_default;
2181extern __u32 sysctl_rmem_default;
2182
2183#endif /* _SOCK_H */