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