Merge tag 'v3.10.61' into update
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / include / net / sock.h
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 *
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/bitops.h>
50 #include <linux/lockdep.h>
51 #include <linux/netdevice.h>
52 #include <linux/skbuff.h> /* struct sk_buff */
53 #include <linux/mm.h>
54 #include <linux/security.h>
55 #include <linux/slab.h>
56 #include <linux/uaccess.h>
57 #include <linux/memcontrol.h>
58 #include <linux/res_counter.h>
59 #include <linux/static_key.h>
60 #include <linux/aio.h>
61 #include <linux/sched.h>
62
63 #include <linux/filter.h>
64 #include <linux/rculist_nulls.h>
65 #include <linux/poll.h>
66
67 #include <linux/atomic.h>
68 #include <net/dst.h>
69 #include <net/checksum.h>
70
71 struct cgroup;
72 struct cgroup_subsys;
73 #ifdef CONFIG_NET
74 int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss);
75 void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg);
76 #else
77 static inline
78 int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
79 {
80 return 0;
81 }
82 static inline
83 void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg)
84 {
85 }
86 #endif
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
99 /* Validate arguments and do nothing */
100 static inline __printf(2, 3)
101 void SOCK_DEBUG(const struct sock *sk, const char *msg, ...)
102 {
103 }
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 */
110 typedef struct {
111 spinlock_t slock;
112 int owned;
113 wait_queue_head_t wq;
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
123 } socket_lock_t;
124
125 struct sock;
126 struct proto;
127 struct net;
128
129 typedef __u32 __bitwise __portpair;
130 typedef __u64 __bitwise __addrpair;
131
132 /**
133 * struct sock_common - minimal network layer representation of sockets
134 * @skc_daddr: Foreign IPv4 addr
135 * @skc_rcv_saddr: Bound local IPv4 addr
136 * @skc_hash: hash value used with various protocol lookup tables
137 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
138 * @skc_dport: placeholder for inet_dport/tw_dport
139 * @skc_num: placeholder for inet_num/tw_num
140 * @skc_family: network address family
141 * @skc_state: Connection state
142 * @skc_reuse: %SO_REUSEADDR setting
143 * @skc_reuseport: %SO_REUSEPORT setting
144 * @skc_bound_dev_if: bound device index if != 0
145 * @skc_bind_node: bind hash linkage for various protocol lookup tables
146 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
147 * @skc_prot: protocol handlers inside a network family
148 * @skc_net: reference to the network namespace of this socket
149 * @skc_node: main hash linkage for various protocol lookup tables
150 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
151 * @skc_tx_queue_mapping: tx queue number for this connection
152 * @skc_refcnt: reference count
153 *
154 * This is the minimal network layer representation of sockets, the header
155 * for struct sock and struct inet_timewait_sock.
156 */
157 struct sock_common {
158 /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned
159 * address on 64bit arches : cf INET_MATCH() and INET_TW_MATCH()
160 */
161 union {
162 __addrpair skc_addrpair;
163 struct {
164 __be32 skc_daddr;
165 __be32 skc_rcv_saddr;
166 };
167 };
168 union {
169 unsigned int skc_hash;
170 __u16 skc_u16hashes[2];
171 };
172 /* skc_dport && skc_num must be grouped as well */
173 union {
174 __portpair skc_portpair;
175 struct {
176 __be16 skc_dport;
177 __u16 skc_num;
178 };
179 };
180
181 unsigned short skc_family;
182 volatile unsigned char skc_state;
183 unsigned char skc_reuse:4;
184 unsigned char skc_reuseport:4;
185 int skc_bound_dev_if;
186 union {
187 struct hlist_node skc_bind_node;
188 struct hlist_nulls_node skc_portaddr_node;
189 };
190 struct proto *skc_prot;
191 #ifdef CONFIG_NET_NS
192 struct net *skc_net;
193 #endif
194 /*
195 * fields between dontcopy_begin/dontcopy_end
196 * are not copied in sock_copy()
197 */
198 /* private: */
199 int skc_dontcopy_begin[0];
200 /* public: */
201 union {
202 struct hlist_node skc_node;
203 struct hlist_nulls_node skc_nulls_node;
204 };
205 int skc_tx_queue_mapping;
206 atomic_t skc_refcnt;
207 /* private: */
208 int skc_dontcopy_end[0];
209 /* public: */
210 };
211
212 struct cg_proto;
213 /**
214 * struct sock - network layer representation of sockets
215 * @__sk_common: shared layout with inet_timewait_sock
216 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
217 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
218 * @sk_lock: synchronizer
219 * @sk_rcvbuf: size of receive buffer in bytes
220 * @sk_wq: sock wait queue and async head
221 * @sk_rx_dst: receive input route used by early tcp demux
222 * @sk_dst_cache: destination cache
223 * @sk_dst_lock: destination cache lock
224 * @sk_policy: flow policy
225 * @sk_receive_queue: incoming packets
226 * @sk_wmem_alloc: transmit queue bytes committed
227 * @sk_write_queue: Packet sending queue
228 * @sk_async_wait_queue: DMA copied packets
229 * @sk_omem_alloc: "o" is "option" or "other"
230 * @sk_wmem_queued: persistent queue size
231 * @sk_forward_alloc: space allocated forward
232 * @sk_allocation: allocation mode
233 * @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
234 * @sk_sndbuf: size of send buffer in bytes
235 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
236 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
237 * @sk_no_check: %SO_NO_CHECK setting, whether or not checkup packets
238 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
239 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
240 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
241 * @sk_gso_max_size: Maximum GSO segment size to build
242 * @sk_gso_max_segs: Maximum number of GSO segments
243 * @sk_lingertime: %SO_LINGER l_linger setting
244 * @sk_backlog: always used with the per-socket spinlock held
245 * @sk_callback_lock: used with the callbacks in the end of this struct
246 * @sk_error_queue: rarely used
247 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
248 * IPV6_ADDRFORM for instance)
249 * @sk_err: last error
250 * @sk_err_soft: errors that don't cause failure but are the cause of a
251 * persistent failure not just 'timed out'
252 * @sk_drops: raw/udp drops counter
253 * @sk_ack_backlog: current listen backlog
254 * @sk_max_ack_backlog: listen backlog set in listen()
255 * @sk_priority: %SO_PRIORITY setting
256 * @sk_cgrp_prioidx: socket group's priority map index
257 * @sk_type: socket type (%SOCK_STREAM, etc)
258 * @sk_protocol: which protocol this socket belongs in this network family
259 * @sk_peer_pid: &struct pid for this socket's peer
260 * @sk_peer_cred: %SO_PEERCRED setting
261 * @sk_rcvlowat: %SO_RCVLOWAT setting
262 * @sk_rcvtimeo: %SO_RCVTIMEO setting
263 * @sk_sndtimeo: %SO_SNDTIMEO setting
264 * @sk_rxhash: flow hash received from netif layer
265 * @sk_filter: socket filtering instructions
266 * @sk_protinfo: private area, net family specific, when not using slab
267 * @sk_timer: sock cleanup timer
268 * @sk_stamp: time stamp of last packet received
269 * @sk_socket: Identd and reporting IO signals
270 * @sk_user_data: RPC layer private data
271 * @sk_frag: cached page frag
272 * @sk_peek_off: current peek_offset value
273 * @sk_send_head: front of stuff to transmit
274 * @sk_security: used by security modules
275 * @sk_mark: generic packet mark
276 * @sk_classid: this socket's cgroup classid
277 * @sk_cgrp: this socket's cgroup-specific proto data
278 * @sk_write_pending: a write to stream socket waits to start
279 * @sk_state_change: callback to indicate change in the state of the sock
280 * @sk_data_ready: callback to indicate there is data to be processed
281 * @sk_write_space: callback to indicate there is bf sending space available
282 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
283 * @sk_backlog_rcv: callback to process the backlog
284 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
285 */
286 struct sock {
287 /*
288 * Now struct inet_timewait_sock also uses sock_common, so please just
289 * don't add nothing before this first member (__sk_common) --acme
290 */
291 struct sock_common __sk_common;
292 #define sk_node __sk_common.skc_node
293 #define sk_nulls_node __sk_common.skc_nulls_node
294 #define sk_refcnt __sk_common.skc_refcnt
295 #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
296
297 #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
298 #define sk_dontcopy_end __sk_common.skc_dontcopy_end
299 #define sk_hash __sk_common.skc_hash
300 #define sk_family __sk_common.skc_family
301 #define sk_state __sk_common.skc_state
302 #define sk_reuse __sk_common.skc_reuse
303 #define sk_reuseport __sk_common.skc_reuseport
304 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
305 #define sk_bind_node __sk_common.skc_bind_node
306 #define sk_prot __sk_common.skc_prot
307 #define sk_net __sk_common.skc_net
308 socket_lock_t sk_lock;
309 struct sk_buff_head sk_receive_queue;
310 /*
311 * The backlog queue is special, it is always used with
312 * the per-socket spinlock held and requires low latency
313 * access. Therefore we special case it's implementation.
314 * Note : rmem_alloc is in this structure to fill a hole
315 * on 64bit arches, not because its logically part of
316 * backlog.
317 */
318 struct {
319 atomic_t rmem_alloc;
320 int len;
321 struct sk_buff *head;
322 struct sk_buff *tail;
323 } sk_backlog;
324 #define sk_rmem_alloc sk_backlog.rmem_alloc
325 int sk_forward_alloc;
326 #ifdef CONFIG_RPS
327 __u32 sk_rxhash;
328 #endif
329 atomic_t sk_drops;
330 int sk_rcvbuf;
331
332 struct sk_filter __rcu *sk_filter;
333 struct socket_wq __rcu *sk_wq;
334
335 #ifdef CONFIG_NET_DMA
336 struct sk_buff_head sk_async_wait_queue;
337 #endif
338
339 #ifdef CONFIG_XFRM
340 struct xfrm_policy *sk_policy[2];
341 #endif
342 unsigned long sk_flags;
343 struct dst_entry *sk_rx_dst;
344 struct dst_entry __rcu *sk_dst_cache;
345 spinlock_t sk_dst_lock;
346 atomic_t sk_wmem_alloc;
347 atomic_t sk_omem_alloc;
348 int sk_sndbuf;
349 struct sk_buff_head sk_write_queue;
350 kmemcheck_bitfield_begin(flags);
351 unsigned int sk_shutdown : 2,
352 sk_no_check : 2,
353 sk_userlocks : 4,
354 sk_protocol : 8,
355 sk_type : 16;
356 kmemcheck_bitfield_end(flags);
357 int sk_wmem_queued;
358 gfp_t sk_allocation;
359 u32 sk_pacing_rate; /* bytes per second */
360 netdev_features_t sk_route_caps;
361 netdev_features_t sk_route_nocaps;
362 int sk_gso_type;
363 unsigned int sk_gso_max_size;
364 u16 sk_gso_max_segs;
365 int sk_rcvlowat;
366 unsigned long sk_lingertime;
367 struct sk_buff_head sk_error_queue;
368 struct proto *sk_prot_creator;
369 rwlock_t sk_callback_lock;
370 int sk_err,
371 sk_err_soft;
372 unsigned short sk_ack_backlog;
373 unsigned short sk_max_ack_backlog;
374 __u32 sk_priority;
375 #if IS_ENABLED(CONFIG_NETPRIO_CGROUP)
376 __u32 sk_cgrp_prioidx;
377 #endif
378 struct pid *sk_peer_pid;
379 const struct cred *sk_peer_cred;
380 long sk_rcvtimeo;
381 long sk_sndtimeo;
382 void *sk_protinfo;
383 struct timer_list sk_timer;
384 ktime_t sk_stamp;
385 struct socket *sk_socket;
386 void *sk_user_data;
387 struct page_frag sk_frag;
388 struct sk_buff *sk_send_head;
389 __s32 sk_peek_off;
390 int sk_write_pending;
391 #ifdef CONFIG_SECURITY
392 void *sk_security;
393 #endif
394 __u32 sk_mark;
395 u32 sk_classid;
396 struct cg_proto *sk_cgrp;
397 void (*sk_state_change)(struct sock *sk);
398 void (*sk_data_ready)(struct sock *sk, int bytes);
399 void (*sk_write_space)(struct sock *sk);
400 void (*sk_error_report)(struct sock *sk);
401 int (*sk_backlog_rcv)(struct sock *sk,
402 struct sk_buff *skb);
403 void (*sk_destruct)(struct sock *sk);
404 };
405
406 /*
407 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
408 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
409 * on a socket means that the socket will reuse everybody else's port
410 * without looking at the other's sk_reuse value.
411 */
412
413 #define SK_NO_REUSE 0
414 #define SK_CAN_REUSE 1
415 #define SK_FORCE_REUSE 2
416
417 static inline int sk_peek_offset(struct sock *sk, int flags)
418 {
419 if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0))
420 return sk->sk_peek_off;
421 else
422 return 0;
423 }
424
425 static inline void sk_peek_offset_bwd(struct sock *sk, int val)
426 {
427 if (sk->sk_peek_off >= 0) {
428 if (sk->sk_peek_off >= val)
429 sk->sk_peek_off -= val;
430 else
431 sk->sk_peek_off = 0;
432 }
433 }
434
435 static inline void sk_peek_offset_fwd(struct sock *sk, int val)
436 {
437 if (sk->sk_peek_off >= 0)
438 sk->sk_peek_off += val;
439 }
440
441 /*
442 * Hashed lists helper routines
443 */
444 static inline struct sock *sk_entry(const struct hlist_node *node)
445 {
446 return hlist_entry(node, struct sock, sk_node);
447 }
448
449 static inline struct sock *__sk_head(const struct hlist_head *head)
450 {
451 return hlist_entry(head->first, struct sock, sk_node);
452 }
453
454 static inline struct sock *sk_head(const struct hlist_head *head)
455 {
456 return hlist_empty(head) ? NULL : __sk_head(head);
457 }
458
459 static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
460 {
461 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
462 }
463
464 static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
465 {
466 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
467 }
468
469 static inline struct sock *sk_next(const struct sock *sk)
470 {
471 return sk->sk_node.next ?
472 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
473 }
474
475 static inline struct sock *sk_nulls_next(const struct sock *sk)
476 {
477 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
478 hlist_nulls_entry(sk->sk_nulls_node.next,
479 struct sock, sk_nulls_node) :
480 NULL;
481 }
482
483 static inline bool sk_unhashed(const struct sock *sk)
484 {
485 return hlist_unhashed(&sk->sk_node);
486 }
487
488 static inline bool sk_hashed(const struct sock *sk)
489 {
490 return !sk_unhashed(sk);
491 }
492
493 static inline void sk_node_init(struct hlist_node *node)
494 {
495 node->pprev = NULL;
496 }
497
498 static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
499 {
500 node->pprev = NULL;
501 }
502
503 static inline void __sk_del_node(struct sock *sk)
504 {
505 __hlist_del(&sk->sk_node);
506 }
507
508 /* NB: equivalent to hlist_del_init_rcu */
509 static inline bool __sk_del_node_init(struct sock *sk)
510 {
511 if (sk_hashed(sk)) {
512 __sk_del_node(sk);
513 sk_node_init(&sk->sk_node);
514 return true;
515 }
516 return false;
517 }
518
519 /* Grab socket reference count. This operation is valid only
520 when sk is ALREADY grabbed f.e. it is found in hash table
521 or a list and the lookup is made under lock preventing hash table
522 modifications.
523 */
524
525 static inline void sock_hold(struct sock *sk)
526 {
527 atomic_inc(&sk->sk_refcnt);
528 }
529
530 /* Ungrab socket in the context, which assumes that socket refcnt
531 cannot hit zero, f.e. it is true in context of any socketcall.
532 */
533 static inline void __sock_put(struct sock *sk)
534 {
535 atomic_dec(&sk->sk_refcnt);
536 }
537
538 static inline bool sk_del_node_init(struct sock *sk)
539 {
540 bool rc = __sk_del_node_init(sk);
541
542 if (rc) {
543 /* paranoid for a while -acme */
544 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
545 __sock_put(sk);
546 }
547 return rc;
548 }
549 #define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
550
551 static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
552 {
553 if (sk_hashed(sk)) {
554 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
555 return true;
556 }
557 return false;
558 }
559
560 static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
561 {
562 bool rc = __sk_nulls_del_node_init_rcu(sk);
563
564 if (rc) {
565 /* paranoid for a while -acme */
566 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
567 __sock_put(sk);
568 }
569 return rc;
570 }
571
572 static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
573 {
574 hlist_add_head(&sk->sk_node, list);
575 }
576
577 static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
578 {
579 sock_hold(sk);
580 __sk_add_node(sk, list);
581 }
582
583 static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
584 {
585 sock_hold(sk);
586 hlist_add_head_rcu(&sk->sk_node, list);
587 }
588
589 static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
590 {
591 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
592 }
593
594 static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
595 {
596 sock_hold(sk);
597 __sk_nulls_add_node_rcu(sk, list);
598 }
599
600 static inline void __sk_del_bind_node(struct sock *sk)
601 {
602 __hlist_del(&sk->sk_bind_node);
603 }
604
605 static inline void sk_add_bind_node(struct sock *sk,
606 struct hlist_head *list)
607 {
608 hlist_add_head(&sk->sk_bind_node, list);
609 }
610
611 #define sk_for_each(__sk, list) \
612 hlist_for_each_entry(__sk, list, sk_node)
613 #define sk_for_each_rcu(__sk, list) \
614 hlist_for_each_entry_rcu(__sk, list, sk_node)
615 #define sk_nulls_for_each(__sk, node, list) \
616 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
617 #define sk_nulls_for_each_rcu(__sk, node, list) \
618 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
619 #define sk_for_each_from(__sk) \
620 hlist_for_each_entry_from(__sk, sk_node)
621 #define sk_nulls_for_each_from(__sk, node) \
622 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
623 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
624 #define sk_for_each_safe(__sk, tmp, list) \
625 hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
626 #define sk_for_each_bound(__sk, list) \
627 hlist_for_each_entry(__sk, list, sk_bind_node)
628
629 static inline struct user_namespace *sk_user_ns(struct sock *sk)
630 {
631 /* Careful only use this in a context where these parameters
632 * can not change and must all be valid, such as recvmsg from
633 * userspace.
634 */
635 return sk->sk_socket->file->f_cred->user_ns;
636 }
637
638 /* Sock flags */
639 enum sock_flags {
640 SOCK_DEAD,
641 SOCK_DONE,
642 SOCK_URGINLINE,
643 SOCK_KEEPOPEN,
644 SOCK_LINGER,
645 SOCK_DESTROY,
646 SOCK_BROADCAST,
647 SOCK_TIMESTAMP,
648 SOCK_ZAPPED,
649 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
650 SOCK_DBG, /* %SO_DEBUG setting */
651 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
652 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
653 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
654 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
655 SOCK_MEMALLOC, /* VM depends on this socket for swapping */
656 SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */
657 SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
658 SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */
659 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
660 SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */
661 SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
662 SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
663 SOCK_FASYNC, /* fasync() active */
664 SOCK_RXQ_OVFL,
665 SOCK_ZEROCOPY, /* buffers from userspace */
666 SOCK_WIFI_STATUS, /* push wifi status to userspace */
667 SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
668 * Will use last 4 bytes of packet sent from
669 * user-space instead.
670 */
671 SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */
672 SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */
673 };
674
675 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
676 {
677 nsk->sk_flags = osk->sk_flags;
678 }
679
680 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
681 {
682 __set_bit(flag, &sk->sk_flags);
683 }
684
685 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
686 {
687 __clear_bit(flag, &sk->sk_flags);
688 }
689
690 static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
691 {
692 return test_bit(flag, &sk->sk_flags);
693 }
694
695 #ifdef CONFIG_NET
696 extern struct static_key memalloc_socks;
697 static inline int sk_memalloc_socks(void)
698 {
699 return static_key_false(&memalloc_socks);
700 }
701 #else
702
703 static inline int sk_memalloc_socks(void)
704 {
705 return 0;
706 }
707
708 #endif
709
710 static inline gfp_t sk_gfp_atomic(struct sock *sk, gfp_t gfp_mask)
711 {
712 return GFP_ATOMIC | (sk->sk_allocation & __GFP_MEMALLOC);
713 }
714
715 static inline void sk_acceptq_removed(struct sock *sk)
716 {
717 sk->sk_ack_backlog--;
718 }
719
720 static inline void sk_acceptq_added(struct sock *sk)
721 {
722 sk->sk_ack_backlog++;
723 }
724
725 static inline bool sk_acceptq_is_full(const struct sock *sk)
726 {
727 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
728 }
729
730 /*
731 * Compute minimal free write space needed to queue new packets.
732 */
733 static inline int sk_stream_min_wspace(const struct sock *sk)
734 {
735 return sk->sk_wmem_queued >> 1;
736 }
737
738 static inline int sk_stream_wspace(const struct sock *sk)
739 {
740 return sk->sk_sndbuf - sk->sk_wmem_queued;
741 }
742
743 extern void sk_stream_write_space(struct sock *sk);
744
745 static inline bool sk_stream_memory_free(const struct sock *sk)
746 {
747 return sk->sk_wmem_queued < sk->sk_sndbuf;
748 }
749
750 /* OOB backlog add */
751 static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
752 {
753 /* dont let skb dst not refcounted, we are going to leave rcu lock */
754 skb_dst_force(skb);
755
756 if (!sk->sk_backlog.tail)
757 sk->sk_backlog.head = skb;
758 else
759 sk->sk_backlog.tail->next = skb;
760
761 sk->sk_backlog.tail = skb;
762 skb->next = NULL;
763 }
764
765 /*
766 * Take into account size of receive queue and backlog queue
767 * Do not take into account this skb truesize,
768 * to allow even a single big packet to come.
769 */
770 static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb,
771 unsigned int limit)
772 {
773 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
774
775 return qsize > limit;
776 }
777
778 /* The per-socket spinlock must be held here. */
779 static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
780 unsigned int limit)
781 {
782 if (sk_rcvqueues_full(sk, skb, limit))
783 {
784 #ifdef CONFIG_MTK_NET_LOGGING
785 printk(KERN_ERR "[mtk_net][sock]sk_add_backlog->sk_rcvqueues_full sk->sk_rcvbuf:%d,sk->sk_sndbuf:%d ",sk->sk_rcvbuf,sk->sk_sndbuf);
786 #endif
787 return -ENOBUFS;
788 }
789 __sk_add_backlog(sk, skb);
790 sk->sk_backlog.len += skb->truesize;
791 return 0;
792 }
793
794 extern int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
795
796 static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
797 {
798 if (sk_memalloc_socks() && skb_pfmemalloc(skb))
799 return __sk_backlog_rcv(sk, skb);
800
801 return sk->sk_backlog_rcv(sk, skb);
802 }
803
804 static inline void sock_rps_record_flow(const struct sock *sk)
805 {
806 #ifdef CONFIG_RPS
807 struct rps_sock_flow_table *sock_flow_table;
808
809 rcu_read_lock();
810 sock_flow_table = rcu_dereference(rps_sock_flow_table);
811 rps_record_sock_flow(sock_flow_table, sk->sk_rxhash);
812 rcu_read_unlock();
813 #endif
814 }
815
816 static inline void sock_rps_reset_flow(const struct sock *sk)
817 {
818 #ifdef CONFIG_RPS
819 struct rps_sock_flow_table *sock_flow_table;
820
821 rcu_read_lock();
822 sock_flow_table = rcu_dereference(rps_sock_flow_table);
823 rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash);
824 rcu_read_unlock();
825 #endif
826 }
827
828 static inline void sock_rps_save_rxhash(struct sock *sk,
829 const struct sk_buff *skb)
830 {
831 #ifdef CONFIG_RPS
832 if (unlikely(sk->sk_rxhash != skb->rxhash)) {
833 sock_rps_reset_flow(sk);
834 sk->sk_rxhash = skb->rxhash;
835 }
836 #endif
837 }
838
839 static inline void sock_rps_reset_rxhash(struct sock *sk)
840 {
841 #ifdef CONFIG_RPS
842 sock_rps_reset_flow(sk);
843 sk->sk_rxhash = 0;
844 #endif
845 }
846
847 #define sk_wait_event(__sk, __timeo, __condition) \
848 ({ int __rc; \
849 release_sock(__sk); \
850 __rc = __condition; \
851 if (!__rc) { \
852 *(__timeo) = schedule_timeout(*(__timeo)); \
853 } \
854 lock_sock(__sk); \
855 __rc = __condition; \
856 __rc; \
857 })
858
859 extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
860 extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
861 extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
862 extern int sk_stream_error(struct sock *sk, int flags, int err);
863 extern void sk_stream_kill_queues(struct sock *sk);
864 extern void sk_set_memalloc(struct sock *sk);
865 extern void sk_clear_memalloc(struct sock *sk);
866
867 extern int sk_wait_data(struct sock *sk, long *timeo);
868
869 struct request_sock_ops;
870 struct timewait_sock_ops;
871 struct inet_hashinfo;
872 struct raw_hashinfo;
873 struct module;
874
875 /*
876 * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
877 * un-modified. Special care is taken when initializing object to zero.
878 */
879 static inline void sk_prot_clear_nulls(struct sock *sk, int size)
880 {
881 if (offsetof(struct sock, sk_node.next) != 0)
882 memset(sk, 0, offsetof(struct sock, sk_node.next));
883 memset(&sk->sk_node.pprev, 0,
884 size - offsetof(struct sock, sk_node.pprev));
885 }
886
887 /* Networking protocol blocks we attach to sockets.
888 * socket layer -> transport layer interface
889 * transport -> network interface is defined by struct inet_proto
890 */
891 struct proto {
892 void (*close)(struct sock *sk,
893 long timeout);
894 int (*connect)(struct sock *sk,
895 struct sockaddr *uaddr,
896 int addr_len);
897 int (*disconnect)(struct sock *sk, int flags);
898
899 struct sock * (*accept)(struct sock *sk, int flags, int *err);
900
901 int (*ioctl)(struct sock *sk, int cmd,
902 unsigned long arg);
903 int (*init)(struct sock *sk);
904 void (*destroy)(struct sock *sk);
905 void (*shutdown)(struct sock *sk, int how);
906 int (*setsockopt)(struct sock *sk, int level,
907 int optname, char __user *optval,
908 unsigned int optlen);
909 int (*getsockopt)(struct sock *sk, int level,
910 int optname, char __user *optval,
911 int __user *option);
912 #ifdef CONFIG_COMPAT
913 int (*compat_setsockopt)(struct sock *sk,
914 int level,
915 int optname, char __user *optval,
916 unsigned int optlen);
917 int (*compat_getsockopt)(struct sock *sk,
918 int level,
919 int optname, char __user *optval,
920 int __user *option);
921 int (*compat_ioctl)(struct sock *sk,
922 unsigned int cmd, unsigned long arg);
923 #endif
924 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
925 struct msghdr *msg, size_t len);
926 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
927 struct msghdr *msg,
928 size_t len, int noblock, int flags,
929 int *addr_len);
930 int (*sendpage)(struct sock *sk, struct page *page,
931 int offset, size_t size, int flags);
932 int (*bind)(struct sock *sk,
933 struct sockaddr *uaddr, int addr_len);
934
935 int (*backlog_rcv) (struct sock *sk,
936 struct sk_buff *skb);
937
938 void (*release_cb)(struct sock *sk);
939
940 /* Keeping track of sk's, looking them up, and port selection methods. */
941 void (*hash)(struct sock *sk);
942 void (*unhash)(struct sock *sk);
943 void (*rehash)(struct sock *sk);
944 int (*get_port)(struct sock *sk, unsigned short snum);
945 void (*clear_sk)(struct sock *sk, int size);
946
947 /* Keeping track of sockets in use */
948 #ifdef CONFIG_PROC_FS
949 unsigned int inuse_idx;
950 #endif
951
952 /* Memory pressure */
953 void (*enter_memory_pressure)(struct sock *sk);
954 atomic_long_t *memory_allocated; /* Current allocated memory. */
955 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
956 /*
957 * Pressure flag: try to collapse.
958 * Technical note: it is used by multiple contexts non atomically.
959 * All the __sk_mem_schedule() is of this nature: accounting
960 * is strict, actions are advisory and have some latency.
961 */
962 int *memory_pressure;
963 long *sysctl_mem;
964 int *sysctl_wmem;
965 int *sysctl_rmem;
966 int max_header;
967 bool no_autobind;
968
969 struct kmem_cache *slab;
970 unsigned int obj_size;
971 int slab_flags;
972
973 struct percpu_counter *orphan_count;
974
975 struct request_sock_ops *rsk_prot;
976 struct timewait_sock_ops *twsk_prot;
977
978 union {
979 struct inet_hashinfo *hashinfo;
980 struct udp_table *udp_table;
981 struct raw_hashinfo *raw_hash;
982 } h;
983
984 struct module *owner;
985
986 char name[32];
987
988 struct list_head node;
989 #ifdef SOCK_REFCNT_DEBUG
990 atomic_t socks;
991 #endif
992 #ifdef CONFIG_MEMCG_KMEM
993 /*
994 * cgroup specific init/deinit functions. Called once for all
995 * protocols that implement it, from cgroups populate function.
996 * This function has to setup any files the protocol want to
997 * appear in the kmem cgroup filesystem.
998 */
999 int (*init_cgroup)(struct mem_cgroup *memcg,
1000 struct cgroup_subsys *ss);
1001 void (*destroy_cgroup)(struct mem_cgroup *memcg);
1002 struct cg_proto *(*proto_cgroup)(struct mem_cgroup *memcg);
1003 #endif
1004 };
1005
1006 /*
1007 * Bits in struct cg_proto.flags
1008 */
1009 enum cg_proto_flags {
1010 /* Currently active and new sockets should be assigned to cgroups */
1011 MEMCG_SOCK_ACTIVE,
1012 /* It was ever activated; we must disarm static keys on destruction */
1013 MEMCG_SOCK_ACTIVATED,
1014 };
1015
1016 struct cg_proto {
1017 void (*enter_memory_pressure)(struct sock *sk);
1018 struct res_counter *memory_allocated; /* Current allocated memory. */
1019 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
1020 int *memory_pressure;
1021 long *sysctl_mem;
1022 unsigned long flags;
1023 /*
1024 * memcg field is used to find which memcg we belong directly
1025 * Each memcg struct can hold more than one cg_proto, so container_of
1026 * won't really cut.
1027 *
1028 * The elegant solution would be having an inverse function to
1029 * proto_cgroup in struct proto, but that means polluting the structure
1030 * for everybody, instead of just for memcg users.
1031 */
1032 struct mem_cgroup *memcg;
1033 };
1034
1035 extern int proto_register(struct proto *prot, int alloc_slab);
1036 extern void proto_unregister(struct proto *prot);
1037
1038 static inline bool memcg_proto_active(struct cg_proto *cg_proto)
1039 {
1040 return test_bit(MEMCG_SOCK_ACTIVE, &cg_proto->flags);
1041 }
1042
1043 static inline bool memcg_proto_activated(struct cg_proto *cg_proto)
1044 {
1045 return test_bit(MEMCG_SOCK_ACTIVATED, &cg_proto->flags);
1046 }
1047
1048 #ifdef SOCK_REFCNT_DEBUG
1049 static inline void sk_refcnt_debug_inc(struct sock *sk)
1050 {
1051 atomic_inc(&sk->sk_prot->socks);
1052 }
1053
1054 static inline void sk_refcnt_debug_dec(struct sock *sk)
1055 {
1056 atomic_dec(&sk->sk_prot->socks);
1057 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
1058 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
1059 }
1060
1061 static inline void sk_refcnt_debug_release(const struct sock *sk)
1062 {
1063 if (atomic_read(&sk->sk_refcnt) != 1)
1064 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
1065 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
1066 }
1067 #else /* SOCK_REFCNT_DEBUG */
1068 #define sk_refcnt_debug_inc(sk) do { } while (0)
1069 #define sk_refcnt_debug_dec(sk) do { } while (0)
1070 #define sk_refcnt_debug_release(sk) do { } while (0)
1071 #endif /* SOCK_REFCNT_DEBUG */
1072
1073 #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_NET)
1074 extern struct static_key memcg_socket_limit_enabled;
1075 static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1076 struct cg_proto *cg_proto)
1077 {
1078 return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg));
1079 }
1080 #define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
1081 #else
1082 #define mem_cgroup_sockets_enabled 0
1083 static inline struct cg_proto *parent_cg_proto(struct proto *proto,
1084 struct cg_proto *cg_proto)
1085 {
1086 return NULL;
1087 }
1088 #endif
1089
1090
1091 static inline bool sk_has_memory_pressure(const struct sock *sk)
1092 {
1093 return sk->sk_prot->memory_pressure != NULL;
1094 }
1095
1096 static inline bool sk_under_memory_pressure(const struct sock *sk)
1097 {
1098 if (!sk->sk_prot->memory_pressure)
1099 return false;
1100
1101 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1102 return !!*sk->sk_cgrp->memory_pressure;
1103
1104 return !!*sk->sk_prot->memory_pressure;
1105 }
1106
1107 static inline void sk_leave_memory_pressure(struct sock *sk)
1108 {
1109 int *memory_pressure = sk->sk_prot->memory_pressure;
1110
1111 if (!memory_pressure)
1112 return;
1113
1114 if (*memory_pressure)
1115 *memory_pressure = 0;
1116
1117 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1118 struct cg_proto *cg_proto = sk->sk_cgrp;
1119 struct proto *prot = sk->sk_prot;
1120
1121 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1122 if (*cg_proto->memory_pressure)
1123 *cg_proto->memory_pressure = 0;
1124 }
1125
1126 }
1127
1128 static inline void sk_enter_memory_pressure(struct sock *sk)
1129 {
1130 if (!sk->sk_prot->enter_memory_pressure)
1131 return;
1132
1133 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1134 struct cg_proto *cg_proto = sk->sk_cgrp;
1135 struct proto *prot = sk->sk_prot;
1136
1137 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1138 cg_proto->enter_memory_pressure(sk);
1139 }
1140
1141 sk->sk_prot->enter_memory_pressure(sk);
1142 }
1143
1144 static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1145 {
1146 long *prot = sk->sk_prot->sysctl_mem;
1147 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1148 prot = sk->sk_cgrp->sysctl_mem;
1149 return prot[index];
1150 }
1151
1152 static inline void memcg_memory_allocated_add(struct cg_proto *prot,
1153 unsigned long amt,
1154 int *parent_status)
1155 {
1156 struct res_counter *fail;
1157 int ret;
1158
1159 ret = res_counter_charge_nofail(prot->memory_allocated,
1160 amt << PAGE_SHIFT, &fail);
1161 if (ret < 0)
1162 *parent_status = OVER_LIMIT;
1163 }
1164
1165 static inline void memcg_memory_allocated_sub(struct cg_proto *prot,
1166 unsigned long amt)
1167 {
1168 res_counter_uncharge(prot->memory_allocated, amt << PAGE_SHIFT);
1169 }
1170
1171 static inline u64 memcg_memory_allocated_read(struct cg_proto *prot)
1172 {
1173 u64 ret;
1174 ret = res_counter_read_u64(prot->memory_allocated, RES_USAGE);
1175 return ret >> PAGE_SHIFT;
1176 }
1177
1178 static inline long
1179 sk_memory_allocated(const struct sock *sk)
1180 {
1181 struct proto *prot = sk->sk_prot;
1182 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1183 return memcg_memory_allocated_read(sk->sk_cgrp);
1184
1185 return atomic_long_read(prot->memory_allocated);
1186 }
1187
1188 static inline long
1189 sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status)
1190 {
1191 struct proto *prot = sk->sk_prot;
1192
1193 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1194 memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status);
1195 /* update the root cgroup regardless */
1196 atomic_long_add_return(amt, prot->memory_allocated);
1197 return memcg_memory_allocated_read(sk->sk_cgrp);
1198 }
1199
1200 return atomic_long_add_return(amt, prot->memory_allocated);
1201 }
1202
1203 static inline void
1204 sk_memory_allocated_sub(struct sock *sk, int amt)
1205 {
1206 struct proto *prot = sk->sk_prot;
1207
1208 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1209 memcg_memory_allocated_sub(sk->sk_cgrp, amt);
1210
1211 atomic_long_sub(amt, prot->memory_allocated);
1212 }
1213
1214 static inline void sk_sockets_allocated_dec(struct sock *sk)
1215 {
1216 struct proto *prot = sk->sk_prot;
1217
1218 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1219 struct cg_proto *cg_proto = sk->sk_cgrp;
1220
1221 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1222 percpu_counter_dec(cg_proto->sockets_allocated);
1223 }
1224
1225 percpu_counter_dec(prot->sockets_allocated);
1226 }
1227
1228 static inline void sk_sockets_allocated_inc(struct sock *sk)
1229 {
1230 struct proto *prot = sk->sk_prot;
1231
1232 if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
1233 struct cg_proto *cg_proto = sk->sk_cgrp;
1234
1235 for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
1236 percpu_counter_inc(cg_proto->sockets_allocated);
1237 }
1238
1239 percpu_counter_inc(prot->sockets_allocated);
1240 }
1241
1242 static inline int
1243 sk_sockets_allocated_read_positive(struct sock *sk)
1244 {
1245 struct proto *prot = sk->sk_prot;
1246
1247 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1248 return percpu_counter_read_positive(sk->sk_cgrp->sockets_allocated);
1249
1250 return percpu_counter_read_positive(prot->sockets_allocated);
1251 }
1252
1253 static inline int
1254 proto_sockets_allocated_sum_positive(struct proto *prot)
1255 {
1256 return percpu_counter_sum_positive(prot->sockets_allocated);
1257 }
1258
1259 static inline long
1260 proto_memory_allocated(struct proto *prot)
1261 {
1262 return atomic_long_read(prot->memory_allocated);
1263 }
1264
1265 static inline bool
1266 proto_memory_pressure(struct proto *prot)
1267 {
1268 if (!prot->memory_pressure)
1269 return false;
1270 return !!*prot->memory_pressure;
1271 }
1272
1273
1274 #ifdef CONFIG_PROC_FS
1275 /* Called with local bh disabled */
1276 extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1277 extern int sock_prot_inuse_get(struct net *net, struct proto *proto);
1278 #else
1279 static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
1280 int inc)
1281 {
1282 }
1283 #endif
1284
1285
1286 /* With per-bucket locks this operation is not-atomic, so that
1287 * this version is not worse.
1288 */
1289 static inline void __sk_prot_rehash(struct sock *sk)
1290 {
1291 sk->sk_prot->unhash(sk);
1292 sk->sk_prot->hash(sk);
1293 }
1294
1295 void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
1296
1297 /* About 10 seconds */
1298 #define SOCK_DESTROY_TIME (10*HZ)
1299
1300 /* Sockets 0-1023 can't be bound to unless you are superuser */
1301 #define PROT_SOCK 1024
1302
1303 #define SHUTDOWN_MASK 3
1304 #define RCV_SHUTDOWN 1
1305 #define SEND_SHUTDOWN 2
1306
1307 #define SOCK_SNDBUF_LOCK 1
1308 #define SOCK_RCVBUF_LOCK 2
1309 #define SOCK_BINDADDR_LOCK 4
1310 #define SOCK_BINDPORT_LOCK 8
1311
1312 /* sock_iocb: used to kick off async processing of socket ios */
1313 struct sock_iocb {
1314 struct list_head list;
1315
1316 int flags;
1317 int size;
1318 struct socket *sock;
1319 struct sock *sk;
1320 struct scm_cookie *scm;
1321 struct msghdr *msg, async_msg;
1322 struct kiocb *kiocb;
1323 };
1324
1325 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
1326 {
1327 return (struct sock_iocb *)iocb->private;
1328 }
1329
1330 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
1331 {
1332 return si->kiocb;
1333 }
1334
1335 struct socket_alloc {
1336 struct socket socket;
1337 struct inode vfs_inode;
1338 };
1339
1340 static inline struct socket *SOCKET_I(struct inode *inode)
1341 {
1342 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1343 }
1344
1345 static inline struct inode *SOCK_INODE(struct socket *socket)
1346 {
1347 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1348 }
1349
1350 /*
1351 * Functions for memory accounting
1352 */
1353 extern int __sk_mem_schedule(struct sock *sk, int size, int kind);
1354 extern void __sk_mem_reclaim(struct sock *sk);
1355
1356 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1357 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1358 #define SK_MEM_SEND 0
1359 #define SK_MEM_RECV 1
1360
1361 static inline int sk_mem_pages(int amt)
1362 {
1363 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
1364 }
1365
1366 static inline bool sk_has_account(struct sock *sk)
1367 {
1368 /* return true if protocol supports memory accounting */
1369 return !!sk->sk_prot->memory_allocated;
1370 }
1371
1372 static inline bool sk_wmem_schedule(struct sock *sk, int size)
1373 {
1374 if (!sk_has_account(sk))
1375 return true;
1376 return size <= sk->sk_forward_alloc ||
1377 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1378 }
1379
1380 static inline bool
1381 sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size)
1382 {
1383 if (!sk_has_account(sk))
1384 return true;
1385 return size<= sk->sk_forward_alloc ||
1386 __sk_mem_schedule(sk, size, SK_MEM_RECV) ||
1387 skb_pfmemalloc(skb);
1388 }
1389
1390 static inline void sk_mem_reclaim(struct sock *sk)
1391 {
1392 if (!sk_has_account(sk))
1393 return;
1394 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1395 __sk_mem_reclaim(sk);
1396 }
1397
1398 static inline void sk_mem_reclaim_partial(struct sock *sk)
1399 {
1400 if (!sk_has_account(sk))
1401 return;
1402 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1403 __sk_mem_reclaim(sk);
1404 }
1405
1406 static inline void sk_mem_charge(struct sock *sk, int size)
1407 {
1408 if (!sk_has_account(sk))
1409 return;
1410 sk->sk_forward_alloc -= size;
1411 }
1412
1413 static inline void sk_mem_uncharge(struct sock *sk, int size)
1414 {
1415 if (!sk_has_account(sk))
1416 return;
1417 sk->sk_forward_alloc += size;
1418 }
1419
1420 static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1421 {
1422 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1423 sk->sk_wmem_queued -= skb->truesize;
1424 sk_mem_uncharge(sk, skb->truesize);
1425 __kfree_skb(skb);
1426 }
1427
1428 /* Used by processes to "lock" a socket state, so that
1429 * interrupts and bottom half handlers won't change it
1430 * from under us. It essentially blocks any incoming
1431 * packets, so that we won't get any new data or any
1432 * packets that change the state of the socket.
1433 *
1434 * While locked, BH processing will add new packets to
1435 * the backlog queue. This queue is processed by the
1436 * owner of the socket lock right before it is released.
1437 *
1438 * Since ~2.3.5 it is also exclusive sleep lock serializing
1439 * accesses from user process context.
1440 */
1441 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
1442
1443 static inline void sock_release_ownership(struct sock *sk)
1444 {
1445 sk->sk_lock.owned = 0;
1446 }
1447
1448 /*
1449 * Macro so as to not evaluate some arguments when
1450 * lockdep is not enabled.
1451 *
1452 * Mark both the sk_lock and the sk_lock.slock as a
1453 * per-address-family lock class.
1454 */
1455 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1456 do { \
1457 sk->sk_lock.owned = 0; \
1458 init_waitqueue_head(&sk->sk_lock.wq); \
1459 spin_lock_init(&(sk)->sk_lock.slock); \
1460 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1461 sizeof((sk)->sk_lock)); \
1462 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1463 (skey), (sname)); \
1464 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1465 } while (0)
1466
1467 extern void lock_sock_nested(struct sock *sk, int subclass);
1468
1469 static inline void lock_sock(struct sock *sk)
1470 {
1471 lock_sock_nested(sk, 0);
1472 }
1473
1474 extern void release_sock(struct sock *sk);
1475
1476 /* BH context may only use the following locking interface. */
1477 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
1478 #define bh_lock_sock_nested(__sk) \
1479 spin_lock_nested(&((__sk)->sk_lock.slock), \
1480 SINGLE_DEPTH_NESTING)
1481 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1482
1483 extern bool lock_sock_fast(struct sock *sk);
1484 /**
1485 * unlock_sock_fast - complement of lock_sock_fast
1486 * @sk: socket
1487 * @slow: slow mode
1488 *
1489 * fast unlock socket for user context.
1490 * If slow mode is on, we call regular release_sock()
1491 */
1492 static inline void unlock_sock_fast(struct sock *sk, bool slow)
1493 {
1494 if (slow)
1495 release_sock(sk);
1496 else
1497 spin_unlock_bh(&sk->sk_lock.slock);
1498 }
1499
1500
1501 extern struct sock *sk_alloc(struct net *net, int family,
1502 gfp_t priority,
1503 struct proto *prot);
1504 extern void sk_free(struct sock *sk);
1505 extern void sk_release_kernel(struct sock *sk);
1506 extern struct sock *sk_clone_lock(const struct sock *sk,
1507 const gfp_t priority);
1508
1509 extern struct sk_buff *sock_wmalloc(struct sock *sk,
1510 unsigned long size, int force,
1511 gfp_t priority);
1512 extern struct sk_buff *sock_rmalloc(struct sock *sk,
1513 unsigned long size, int force,
1514 gfp_t priority);
1515 extern void sock_wfree(struct sk_buff *skb);
1516 extern void sock_rfree(struct sk_buff *skb);
1517 extern void sock_edemux(struct sk_buff *skb);
1518
1519 extern int sock_setsockopt(struct socket *sock, int level,
1520 int op, char __user *optval,
1521 unsigned int optlen);
1522
1523 extern int sock_getsockopt(struct socket *sock, int level,
1524 int op, char __user *optval,
1525 int __user *optlen);
1526 extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
1527 unsigned long size,
1528 int noblock,
1529 int *errcode);
1530 extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
1531 unsigned long header_len,
1532 unsigned long data_len,
1533 int noblock,
1534 int *errcode);
1535 extern void *sock_kmalloc(struct sock *sk, int size,
1536 gfp_t priority);
1537 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
1538 extern void sk_send_sigurg(struct sock *sk);
1539
1540 /*
1541 * Functions to fill in entries in struct proto_ops when a protocol
1542 * does not implement a particular function.
1543 */
1544 extern int sock_no_bind(struct socket *,
1545 struct sockaddr *, int);
1546 extern int sock_no_connect(struct socket *,
1547 struct sockaddr *, int, int);
1548 extern int sock_no_socketpair(struct socket *,
1549 struct socket *);
1550 extern int sock_no_accept(struct socket *,
1551 struct socket *, int);
1552 extern int sock_no_getname(struct socket *,
1553 struct sockaddr *, int *, int);
1554 extern unsigned int sock_no_poll(struct file *, struct socket *,
1555 struct poll_table_struct *);
1556 extern int sock_no_ioctl(struct socket *, unsigned int,
1557 unsigned long);
1558 extern int sock_no_listen(struct socket *, int);
1559 extern int sock_no_shutdown(struct socket *, int);
1560 extern int sock_no_getsockopt(struct socket *, int , int,
1561 char __user *, int __user *);
1562 extern int sock_no_setsockopt(struct socket *, int, int,
1563 char __user *, unsigned int);
1564 extern int sock_no_sendmsg(struct kiocb *, struct socket *,
1565 struct msghdr *, size_t);
1566 extern int sock_no_recvmsg(struct kiocb *, struct socket *,
1567 struct msghdr *, size_t, int);
1568 extern int sock_no_mmap(struct file *file,
1569 struct socket *sock,
1570 struct vm_area_struct *vma);
1571 extern ssize_t sock_no_sendpage(struct socket *sock,
1572 struct page *page,
1573 int offset, size_t size,
1574 int flags);
1575
1576 /*
1577 * Functions to fill in entries in struct proto_ops when a protocol
1578 * uses the inet style.
1579 */
1580 extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
1581 char __user *optval, int __user *optlen);
1582 extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1583 struct msghdr *msg, size_t size, int flags);
1584 extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
1585 char __user *optval, unsigned int optlen);
1586 extern int compat_sock_common_getsockopt(struct socket *sock, int level,
1587 int optname, char __user *optval, int __user *optlen);
1588 extern int compat_sock_common_setsockopt(struct socket *sock, int level,
1589 int optname, char __user *optval, unsigned int optlen);
1590
1591 extern void sk_common_release(struct sock *sk);
1592
1593 /*
1594 * Default socket callbacks and setup code
1595 */
1596
1597 /* Initialise core socket variables */
1598 extern void sock_init_data(struct socket *sock, struct sock *sk);
1599
1600 extern void sk_filter_release_rcu(struct rcu_head *rcu);
1601
1602 /**
1603 * sk_filter_release - release a socket filter
1604 * @fp: filter to remove
1605 *
1606 * Remove a filter from a socket and release its resources.
1607 */
1608
1609 static inline void sk_filter_release(struct sk_filter *fp)
1610 {
1611 if (atomic_dec_and_test(&fp->refcnt))
1612 call_rcu(&fp->rcu, sk_filter_release_rcu);
1613 }
1614
1615 static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
1616 {
1617 unsigned int size = sk_filter_len(fp);
1618
1619 atomic_sub(size, &sk->sk_omem_alloc);
1620 sk_filter_release(fp);
1621 }
1622
1623 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
1624 {
1625 atomic_inc(&fp->refcnt);
1626 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
1627 }
1628
1629 /*
1630 * Socket reference counting postulates.
1631 *
1632 * * Each user of socket SHOULD hold a reference count.
1633 * * Each access point to socket (an hash table bucket, reference from a list,
1634 * running timer, skb in flight MUST hold a reference count.
1635 * * When reference count hits 0, it means it will never increase back.
1636 * * When reference count hits 0, it means that no references from
1637 * outside exist to this socket and current process on current CPU
1638 * is last user and may/should destroy this socket.
1639 * * sk_free is called from any context: process, BH, IRQ. When
1640 * it is called, socket has no references from outside -> sk_free
1641 * may release descendant resources allocated by the socket, but
1642 * to the time when it is called, socket is NOT referenced by any
1643 * hash tables, lists etc.
1644 * * Packets, delivered from outside (from network or from another process)
1645 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1646 * when they sit in queue. Otherwise, packets will leak to hole, when
1647 * socket is looked up by one cpu and unhasing is made by another CPU.
1648 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1649 * (leak to backlog). Packet socket does all the processing inside
1650 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1651 * use separate SMP lock, so that they are prone too.
1652 */
1653
1654 /* Ungrab socket and destroy it, if it was the last reference. */
1655 static inline void sock_put(struct sock *sk)
1656 {
1657 if (atomic_dec_and_test(&sk->sk_refcnt))
1658 sk_free(sk);
1659 }
1660
1661 extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1662 const int nested);
1663
1664 static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1665 {
1666 sk->sk_tx_queue_mapping = tx_queue;
1667 }
1668
1669 static inline void sk_tx_queue_clear(struct sock *sk)
1670 {
1671 sk->sk_tx_queue_mapping = -1;
1672 }
1673
1674 static inline int sk_tx_queue_get(const struct sock *sk)
1675 {
1676 return sk ? sk->sk_tx_queue_mapping : -1;
1677 }
1678
1679 static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1680 {
1681 sk_tx_queue_clear(sk);
1682 sk->sk_socket = sock;
1683 }
1684
1685 static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1686 {
1687 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1688 return &rcu_dereference_raw(sk->sk_wq)->wait;
1689 }
1690 /* Detach socket from process context.
1691 * Announce socket dead, detach it from wait queue and inode.
1692 * Note that parent inode held reference count on this struct sock,
1693 * we do not release it in this function, because protocol
1694 * probably wants some additional cleanups or even continuing
1695 * to work with this socket (TCP).
1696 */
1697 static inline void sock_orphan(struct sock *sk)
1698 {
1699 write_lock_bh(&sk->sk_callback_lock);
1700 sock_set_flag(sk, SOCK_DEAD);
1701 sk_set_socket(sk, NULL);
1702 sk->sk_wq = NULL;
1703 write_unlock_bh(&sk->sk_callback_lock);
1704 }
1705
1706 static inline void sock_graft(struct sock *sk, struct socket *parent)
1707 {
1708 write_lock_bh(&sk->sk_callback_lock);
1709 sk->sk_wq = parent->wq;
1710 parent->sk = sk;
1711 sk_set_socket(sk, parent);
1712 security_sock_graft(sk, parent);
1713 write_unlock_bh(&sk->sk_callback_lock);
1714 }
1715
1716 extern kuid_t sock_i_uid(struct sock *sk);
1717 extern unsigned long sock_i_ino(struct sock *sk);
1718
1719 static inline struct dst_entry *
1720 __sk_dst_get(struct sock *sk)
1721 {
1722 return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
1723 lockdep_is_held(&sk->sk_lock.slock));
1724 }
1725
1726 static inline struct dst_entry *
1727 sk_dst_get(struct sock *sk)
1728 {
1729 struct dst_entry *dst;
1730
1731 rcu_read_lock();
1732 dst = rcu_dereference(sk->sk_dst_cache);
1733 if (dst && !atomic_inc_not_zero(&dst->__refcnt))
1734 dst = NULL;
1735 rcu_read_unlock();
1736 return dst;
1737 }
1738
1739 extern void sk_reset_txq(struct sock *sk);
1740
1741 static inline void dst_negative_advice(struct sock *sk)
1742 {
1743 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1744
1745 if (dst && dst->ops->negative_advice) {
1746 ndst = dst->ops->negative_advice(dst);
1747
1748 if (ndst != dst) {
1749 rcu_assign_pointer(sk->sk_dst_cache, ndst);
1750 sk_reset_txq(sk);
1751 }
1752 }
1753 }
1754
1755 static inline void
1756 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
1757 {
1758 struct dst_entry *old_dst;
1759
1760 sk_tx_queue_clear(sk);
1761 /*
1762 * This can be called while sk is owned by the caller only,
1763 * with no state that can be checked in a rcu_dereference_check() cond
1764 */
1765 old_dst = rcu_dereference_raw(sk->sk_dst_cache);
1766 rcu_assign_pointer(sk->sk_dst_cache, dst);
1767 dst_release(old_dst);
1768 }
1769
1770 static inline void
1771 sk_dst_set(struct sock *sk, struct dst_entry *dst)
1772 {
1773 struct dst_entry *old_dst;
1774
1775 sk_tx_queue_clear(sk);
1776 old_dst = xchg((__force struct dst_entry **)&sk->sk_dst_cache, dst);
1777 dst_release(old_dst);
1778 }
1779
1780 static inline void
1781 __sk_dst_reset(struct sock *sk)
1782 {
1783 __sk_dst_set(sk, NULL);
1784 }
1785
1786 static inline void
1787 sk_dst_reset(struct sock *sk)
1788 {
1789 sk_dst_set(sk, NULL);
1790 }
1791
1792 extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1793
1794 extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1795
1796 static inline bool sk_can_gso(const struct sock *sk)
1797 {
1798 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1799 }
1800
1801 extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
1802
1803 static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
1804 {
1805 sk->sk_route_nocaps |= flags;
1806 sk->sk_route_caps &= ~flags;
1807 }
1808
1809 static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1810 char __user *from, char *to,
1811 int copy, int offset)
1812 {
1813 if (skb->ip_summed == CHECKSUM_NONE) {
1814 int err = 0;
1815 __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err);
1816 if (err)
1817 return err;
1818 skb->csum = csum_block_add(skb->csum, csum, offset);
1819 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1820 if (!access_ok(VERIFY_READ, from, copy) ||
1821 __copy_from_user_nocache(to, from, copy))
1822 return -EFAULT;
1823 } else if (copy_from_user(to, from, copy))
1824 return -EFAULT;
1825
1826 return 0;
1827 }
1828
1829 static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
1830 char __user *from, int copy)
1831 {
1832 int err, offset = skb->len;
1833
1834 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1835 copy, offset);
1836 if (err)
1837 __skb_trim(skb, offset);
1838
1839 return err;
1840 }
1841
1842 static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from,
1843 struct sk_buff *skb,
1844 struct page *page,
1845 int off, int copy)
1846 {
1847 int err;
1848
1849 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1850 copy, skb->len);
1851 if (err)
1852 return err;
1853
1854 skb->len += copy;
1855 skb->data_len += copy;
1856 skb->truesize += copy;
1857 sk->sk_wmem_queued += copy;
1858 sk_mem_charge(sk, copy);
1859 return 0;
1860 }
1861
1862 static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1863 struct sk_buff *skb, struct page *page,
1864 int off, int copy)
1865 {
1866 if (skb->ip_summed == CHECKSUM_NONE) {
1867 int err = 0;
1868 __wsum csum = csum_and_copy_from_user(from,
1869 page_address(page) + off,
1870 copy, 0, &err);
1871 if (err)
1872 return err;
1873 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1874 } else if (copy_from_user(page_address(page) + off, from, copy))
1875 return -EFAULT;
1876
1877 skb->len += copy;
1878 skb->data_len += copy;
1879 skb->truesize += copy;
1880 sk->sk_wmem_queued += copy;
1881 sk_mem_charge(sk, copy);
1882 return 0;
1883 }
1884
1885 /**
1886 * sk_wmem_alloc_get - returns write allocations
1887 * @sk: socket
1888 *
1889 * Returns sk_wmem_alloc minus initial offset of one
1890 */
1891 static inline int sk_wmem_alloc_get(const struct sock *sk)
1892 {
1893 return atomic_read(&sk->sk_wmem_alloc) - 1;
1894 }
1895
1896 /**
1897 * sk_rmem_alloc_get - returns read allocations
1898 * @sk: socket
1899 *
1900 * Returns sk_rmem_alloc
1901 */
1902 static inline int sk_rmem_alloc_get(const struct sock *sk)
1903 {
1904 return atomic_read(&sk->sk_rmem_alloc);
1905 }
1906
1907 /**
1908 * sk_has_allocations - check if allocations are outstanding
1909 * @sk: socket
1910 *
1911 * Returns true if socket has write or read allocations
1912 */
1913 static inline bool sk_has_allocations(const struct sock *sk)
1914 {
1915 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1916 }
1917
1918 /**
1919 * wq_has_sleeper - check if there are any waiting processes
1920 * @wq: struct socket_wq
1921 *
1922 * Returns true if socket_wq has waiting processes
1923 *
1924 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
1925 * barrier call. They were added due to the race found within the tcp code.
1926 *
1927 * Consider following tcp code paths:
1928 *
1929 * CPU1 CPU2
1930 *
1931 * sys_select receive packet
1932 * ... ...
1933 * __add_wait_queue update tp->rcv_nxt
1934 * ... ...
1935 * tp->rcv_nxt check sock_def_readable
1936 * ... {
1937 * schedule rcu_read_lock();
1938 * wq = rcu_dereference(sk->sk_wq);
1939 * if (wq && waitqueue_active(&wq->wait))
1940 * wake_up_interruptible(&wq->wait)
1941 * ...
1942 * }
1943 *
1944 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1945 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1946 * could then endup calling schedule and sleep forever if there are no more
1947 * data on the socket.
1948 *
1949 */
1950 static inline bool wq_has_sleeper(struct socket_wq *wq)
1951 {
1952 /* We need to be sure we are in sync with the
1953 * add_wait_queue modifications to the wait queue.
1954 *
1955 * This memory barrier is paired in the sock_poll_wait.
1956 */
1957 smp_mb();
1958 return wq && waitqueue_active(&wq->wait);
1959 }
1960
1961 /**
1962 * sock_poll_wait - place memory barrier behind the poll_wait call.
1963 * @filp: file
1964 * @wait_address: socket wait queue
1965 * @p: poll_table
1966 *
1967 * See the comments in the wq_has_sleeper function.
1968 */
1969 static inline void sock_poll_wait(struct file *filp,
1970 wait_queue_head_t *wait_address, poll_table *p)
1971 {
1972 if (!poll_does_not_wait(p) && wait_address) {
1973 poll_wait(filp, wait_address, p);
1974 /* We need to be sure we are in sync with the
1975 * socket flags modification.
1976 *
1977 * This memory barrier is paired in the wq_has_sleeper.
1978 */
1979 smp_mb();
1980 }
1981 }
1982
1983 /*
1984 * Queue a received datagram if it will fit. Stream and sequenced
1985 * protocols can't normally use this as they need to fit buffers in
1986 * and play with them.
1987 *
1988 * Inlined as it's very short and called for pretty much every
1989 * packet ever received.
1990 */
1991
1992 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1993 {
1994 skb_orphan(skb);
1995 skb->sk = sk;
1996 skb->destructor = sock_wfree;
1997 /*
1998 * We used to take a refcount on sk, but following operation
1999 * is enough to guarantee sk_free() wont free this sock until
2000 * all in-flight packets are completed
2001 */
2002 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
2003 }
2004
2005 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
2006 {
2007 skb_orphan(skb);
2008 skb->sk = sk;
2009 skb->destructor = sock_rfree;
2010 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
2011 sk_mem_charge(sk, skb->truesize);
2012 }
2013
2014 extern void sk_reset_timer(struct sock *sk, struct timer_list *timer,
2015 unsigned long expires);
2016
2017 extern void sk_stop_timer(struct sock *sk, struct timer_list *timer);
2018
2019 extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
2020
2021 extern int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
2022
2023 /*
2024 * Recover an error report and clear atomically
2025 */
2026
2027 static inline int sock_error(struct sock *sk)
2028 {
2029 int err;
2030 if (likely(!sk->sk_err))
2031 return 0;
2032 err = xchg(&sk->sk_err, 0);
2033 return -err;
2034 }
2035
2036 static inline unsigned long sock_wspace(struct sock *sk)
2037 {
2038 int amt = 0;
2039
2040 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
2041 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
2042 if (amt < 0)
2043 amt = 0;
2044 }
2045 return amt;
2046 }
2047
2048 static inline void sk_wake_async(struct sock *sk, int how, int band)
2049 {
2050 if (sock_flag(sk, SOCK_FASYNC))
2051 sock_wake_async(sk->sk_socket, how, band);
2052 }
2053
2054 #define SOCK_MIN_SNDBUF 2048
2055 /*
2056 * Since sk_rmem_alloc sums skb->truesize, even a small frame might need
2057 * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak
2058 */
2059 #define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff))
2060
2061 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
2062 {
2063 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
2064 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
2065 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
2066 }
2067 }
2068
2069 struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
2070
2071 /**
2072 * sk_page_frag - return an appropriate page_frag
2073 * @sk: socket
2074 *
2075 * If socket allocation mode allows current thread to sleep, it means its
2076 * safe to use the per task page_frag instead of the per socket one.
2077 */
2078 static inline struct page_frag *sk_page_frag(struct sock *sk)
2079 {
2080 if (sk->sk_allocation & __GFP_WAIT)
2081 return &current->task_frag;
2082
2083 return &sk->sk_frag;
2084 }
2085
2086 extern bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag);
2087
2088 /*
2089 * Default write policy as shown to user space via poll/select/SIGIO
2090 */
2091 static inline bool sock_writeable(const struct sock *sk)
2092 {
2093 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
2094 }
2095
2096 static inline gfp_t gfp_any(void)
2097 {
2098 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
2099 }
2100
2101 static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
2102 {
2103 return noblock ? 0 : sk->sk_rcvtimeo;
2104 }
2105
2106 static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
2107 {
2108 return noblock ? 0 : sk->sk_sndtimeo;
2109 }
2110
2111 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2112 {
2113 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
2114 }
2115
2116 /* Alas, with timeout socket operations are not restartable.
2117 * Compare this to poll().
2118 */
2119 static inline int sock_intr_errno(long timeo)
2120 {
2121 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2122 }
2123
2124 extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2125 struct sk_buff *skb);
2126 extern void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2127 struct sk_buff *skb);
2128
2129 static inline void
2130 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2131 {
2132 ktime_t kt = skb->tstamp;
2133 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
2134
2135 /*
2136 * generate control messages if
2137 * - receive time stamping in software requested (SOCK_RCVTSTAMP
2138 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
2139 * - software time stamp available and wanted
2140 * (SOCK_TIMESTAMPING_SOFTWARE)
2141 * - hardware time stamps available and wanted
2142 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
2143 * SOCK_TIMESTAMPING_RAW_HARDWARE)
2144 */
2145 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
2146 sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
2147 (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
2148 (hwtstamps->hwtstamp.tv64 &&
2149 sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
2150 (hwtstamps->syststamp.tv64 &&
2151 sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
2152 __sock_recv_timestamp(msg, sk, skb);
2153 else
2154 sk->sk_stamp = kt;
2155
2156 if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2157 __sock_recv_wifi_status(msg, sk, skb);
2158 }
2159
2160 extern void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2161 struct sk_buff *skb);
2162
2163 static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2164 struct sk_buff *skb)
2165 {
2166 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2167 (1UL << SOCK_RCVTSTAMP) | \
2168 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
2169 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
2170 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
2171 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
2172
2173 if (sk->sk_flags & FLAGS_TS_OR_DROPS)
2174 __sock_recv_ts_and_drops(msg, sk, skb);
2175 else
2176 sk->sk_stamp = skb->tstamp;
2177 }
2178
2179 /**
2180 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
2181 * @sk: socket sending this packet
2182 * @tx_flags: filled with instructions for time stamping
2183 *
2184 * Currently only depends on SOCK_TIMESTAMPING* flags.
2185 */
2186 extern void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags);
2187
2188 /**
2189 * sk_eat_skb - Release a skb if it is no longer needed
2190 * @sk: socket to eat this skb from
2191 * @skb: socket buffer to eat
2192 * @copied_early: flag indicating whether DMA operations copied this data early
2193 *
2194 * This routine must be called with interrupts disabled or with the socket
2195 * locked so that the sk_buff queue operation is ok.
2196 */
2197 #ifdef CONFIG_NET_DMA
2198 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
2199 {
2200 __skb_unlink(skb, &sk->sk_receive_queue);
2201 if (!copied_early)
2202 __kfree_skb(skb);
2203 else
2204 __skb_queue_tail(&sk->sk_async_wait_queue, skb);
2205 }
2206 #else
2207 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
2208 {
2209 __skb_unlink(skb, &sk->sk_receive_queue);
2210 __kfree_skb(skb);
2211 }
2212 #endif
2213
2214 static inline
2215 struct net *sock_net(const struct sock *sk)
2216 {
2217 return read_pnet(&sk->sk_net);
2218 }
2219
2220 static inline
2221 void sock_net_set(struct sock *sk, struct net *net)
2222 {
2223 write_pnet(&sk->sk_net, net);
2224 }
2225
2226 /*
2227 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
2228 * They should not hold a reference to a namespace in order to allow
2229 * to stop it.
2230 * Sockets after sk_change_net should be released using sk_release_kernel
2231 */
2232 static inline void sk_change_net(struct sock *sk, struct net *net)
2233 {
2234 put_net(sock_net(sk));
2235 sock_net_set(sk, hold_net(net));
2236 }
2237
2238 static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2239 {
2240 if (skb->sk) {
2241 struct sock *sk = skb->sk;
2242
2243 skb->destructor = NULL;
2244 skb->sk = NULL;
2245 return sk;
2246 }
2247 return NULL;
2248 }
2249
2250 extern void sock_enable_timestamp(struct sock *sk, int flag);
2251 extern int sock_get_timestamp(struct sock *, struct timeval __user *);
2252 extern int sock_get_timestampns(struct sock *, struct timespec __user *);
2253
2254 bool sk_ns_capable(const struct sock *sk,
2255 struct user_namespace *user_ns, int cap);
2256 bool sk_capable(const struct sock *sk, int cap);
2257 bool sk_net_capable(const struct sock *sk, int cap);
2258
2259 /*
2260 * Enable debug/info messages
2261 */
2262 extern int net_msg_warn;
2263 #define NETDEBUG(fmt, args...) \
2264 do { if (net_msg_warn) printk(fmt,##args); } while (0)
2265
2266 #define LIMIT_NETDEBUG(fmt, args...) \
2267 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
2268
2269 extern __u32 sysctl_wmem_max;
2270 extern __u32 sysctl_rmem_max;
2271
2272 extern int sysctl_optmem_max;
2273
2274 extern __u32 sysctl_wmem_default;
2275 extern __u32 sysctl_rmem_default;
2276
2277 #endif /* _SOCK_H */