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.
6 * Definitions for the AF_INET socket handler.
8 * Version: @(#)sock.h 1.0.4 05/13/93
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>
16 * Alan Cox : Volatiles in skbuff pointers. See
17 * skbuff comments. May be overdone,
18 * better to prove they can be removed
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
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.
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 */
54 #include <linux/security.h>
55 #include <linux/slab.h>
56 #include <linux/uaccess.h>
57 #include <linux/page_counter.h>
58 #include <linux/memcontrol.h>
59 #include <linux/static_key.h>
60 #include <linux/sched.h>
61 #include <linux/wait.h>
62 #include <linux/cgroup-defs.h>
64 #include <linux/filter.h>
65 #include <linux/rculist_nulls.h>
66 #include <linux/poll.h>
68 #include <linux/atomic.h>
70 #include <net/checksum.h>
71 #include <net/tcp_states.h>
72 #include <linux/net_tstamp.h>
76 * This structure really needs to be cleaned up.
77 * Most of it is for TCP, and not used by any of
78 * the other protocols.
81 /* Define this to get the SOCK_DBG debugging facility. */
82 #define SOCK_DEBUGGING
84 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
85 printk(KERN_DEBUG msg); } while (0)
87 /* Validate arguments and do nothing */
88 static inline __printf(2, 3)
89 void SOCK_DEBUG(const struct sock *sk, const char *msg, ...)
94 /* This is the per-socket lock. The spinlock provides a synchronization
95 * between user contexts and software interrupt processing, whereas the
96 * mini-semaphore synchronizes multiple users amongst themselves.
101 wait_queue_head_t wq;
103 * We express the mutex-alike socket_lock semantics
104 * to the lock validator by explicitly managing
105 * the slock as a lock variant (in addition to
108 #ifdef CONFIG_DEBUG_LOCK_ALLOC
109 struct lockdep_map dep_map;
117 typedef __u32 __bitwise __portpair;
118 typedef __u64 __bitwise __addrpair;
121 * struct sock_common - minimal network layer representation of sockets
122 * @skc_daddr: Foreign IPv4 addr
123 * @skc_rcv_saddr: Bound local IPv4 addr
124 * @skc_hash: hash value used with various protocol lookup tables
125 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
126 * @skc_dport: placeholder for inet_dport/tw_dport
127 * @skc_num: placeholder for inet_num/tw_num
128 * @skc_family: network address family
129 * @skc_state: Connection state
130 * @skc_reuse: %SO_REUSEADDR setting
131 * @skc_reuseport: %SO_REUSEPORT setting
132 * @skc_bound_dev_if: bound device index if != 0
133 * @skc_bind_node: bind hash linkage for various protocol lookup tables
134 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
135 * @skc_prot: protocol handlers inside a network family
136 * @skc_net: reference to the network namespace of this socket
137 * @skc_node: main hash linkage for various protocol lookup tables
138 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
139 * @skc_tx_queue_mapping: tx queue number for this connection
140 * @skc_flags: place holder for sk_flags
141 * %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
142 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
143 * @skc_incoming_cpu: record/match cpu processing incoming packets
144 * @skc_refcnt: reference count
146 * This is the minimal network layer representation of sockets, the header
147 * for struct sock and struct inet_timewait_sock.
150 /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned
151 * address on 64bit arches : cf INET_MATCH()
154 __addrpair skc_addrpair;
157 __be32 skc_rcv_saddr;
161 unsigned int skc_hash;
162 __u16 skc_u16hashes[2];
164 /* skc_dport && skc_num must be grouped as well */
166 __portpair skc_portpair;
173 unsigned short skc_family;
174 volatile unsigned char skc_state;
175 unsigned char skc_reuse:4;
176 unsigned char skc_reuseport:1;
177 unsigned char skc_ipv6only:1;
178 unsigned char skc_net_refcnt:1;
179 int skc_bound_dev_if;
181 struct hlist_node skc_bind_node;
182 struct hlist_node skc_portaddr_node;
184 struct proto *skc_prot;
185 possible_net_t skc_net;
187 #if IS_ENABLED(CONFIG_IPV6)
188 struct in6_addr skc_v6_daddr;
189 struct in6_addr skc_v6_rcv_saddr;
192 atomic64_t skc_cookie;
194 /* following fields are padding to force
195 * offset(struct sock, sk_refcnt) == 128 on 64bit arches
196 * assuming IPV6 is enabled. We use this padding differently
197 * for different kind of 'sockets'
200 unsigned long skc_flags;
201 struct sock *skc_listener; /* request_sock */
202 struct inet_timewait_death_row *skc_tw_dr; /* inet_timewait_sock */
205 * fields between dontcopy_begin/dontcopy_end
206 * are not copied in sock_copy()
209 int skc_dontcopy_begin[0];
212 struct hlist_node skc_node;
213 struct hlist_nulls_node skc_nulls_node;
215 int skc_tx_queue_mapping;
217 int skc_incoming_cpu;
219 u32 skc_tw_rcv_nxt; /* struct tcp_timewait_sock */
224 int skc_dontcopy_end[0];
227 u32 skc_window_clamp;
228 u32 skc_tw_snd_nxt; /* struct tcp_timewait_sock */
234 * struct sock - network layer representation of sockets
235 * @__sk_common: shared layout with inet_timewait_sock
236 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
237 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
238 * @sk_lock: synchronizer
239 * @sk_kern_sock: True if sock is using kernel lock classes
240 * @sk_rcvbuf: size of receive buffer in bytes
241 * @sk_wq: sock wait queue and async head
242 * @sk_rx_dst: receive input route used by early demux
243 * @sk_dst_cache: destination cache
244 * @sk_dst_pending_confirm: need to confirm neighbour
245 * @sk_policy: flow policy
246 * @sk_receive_queue: incoming packets
247 * @sk_wmem_alloc: transmit queue bytes committed
248 * @sk_write_queue: Packet sending queue
249 * @sk_omem_alloc: "o" is "option" or "other"
250 * @sk_wmem_queued: persistent queue size
251 * @sk_forward_alloc: space allocated forward
252 * @sk_napi_id: id of the last napi context to receive data for sk
253 * @sk_ll_usec: usecs to busypoll when there is no data
254 * @sk_allocation: allocation mode
255 * @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
256 * @sk_pacing_status: Pacing status (requested, handled by sch_fq)
257 * @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE)
258 * @sk_sndbuf: size of send buffer in bytes
259 * @sk_padding: unused element for alignment
260 * @sk_no_check_tx: %SO_NO_CHECK setting, set checksum in TX packets
261 * @sk_no_check_rx: allow zero checksum in RX packets
262 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
263 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
264 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
265 * @sk_gso_max_size: Maximum GSO segment size to build
266 * @sk_gso_max_segs: Maximum number of GSO segments
267 * @sk_lingertime: %SO_LINGER l_linger setting
268 * @sk_backlog: always used with the per-socket spinlock held
269 * @sk_callback_lock: used with the callbacks in the end of this struct
270 * @sk_error_queue: rarely used
271 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
272 * IPV6_ADDRFORM for instance)
273 * @sk_err: last error
274 * @sk_err_soft: errors that don't cause failure but are the cause of a
275 * persistent failure not just 'timed out'
276 * @sk_drops: raw/udp drops counter
277 * @sk_ack_backlog: current listen backlog
278 * @sk_max_ack_backlog: listen backlog set in listen()
279 * @sk_priority: %SO_PRIORITY setting
280 * @sk_type: socket type (%SOCK_STREAM, etc)
281 * @sk_protocol: which protocol this socket belongs in this network family
282 * @sk_peer_pid: &struct pid for this socket's peer
283 * @sk_peer_cred: %SO_PEERCRED setting
284 * @sk_rcvlowat: %SO_RCVLOWAT setting
285 * @sk_rcvtimeo: %SO_RCVTIMEO setting
286 * @sk_sndtimeo: %SO_SNDTIMEO setting
287 * @sk_txhash: computed flow hash for use on transmit
288 * @sk_filter: socket filtering instructions
289 * @sk_timer: sock cleanup timer
290 * @sk_stamp: time stamp of last packet received
291 * @sk_tsflags: SO_TIMESTAMPING socket options
292 * @sk_tskey: counter to disambiguate concurrent tstamp requests
293 * @sk_socket: Identd and reporting IO signals
294 * @sk_user_data: RPC layer private data
295 * @sk_frag: cached page frag
296 * @sk_peek_off: current peek_offset value
297 * @sk_send_head: front of stuff to transmit
298 * @sk_security: used by security modules
299 * @sk_mark: generic packet mark
300 * @sk_cgrp_data: cgroup data for this cgroup
301 * @sk_memcg: this socket's memory cgroup association
302 * @sk_write_pending: a write to stream socket waits to start
303 * @sk_state_change: callback to indicate change in the state of the sock
304 * @sk_data_ready: callback to indicate there is data to be processed
305 * @sk_write_space: callback to indicate there is bf sending space available
306 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
307 * @sk_backlog_rcv: callback to process the backlog
308 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
309 * @sk_reuseport_cb: reuseport group container
310 * @sk_rcu: used during RCU grace period
314 * Now struct inet_timewait_sock also uses sock_common, so please just
315 * don't add nothing before this first member (__sk_common) --acme
317 struct sock_common __sk_common;
318 #define sk_node __sk_common.skc_node
319 #define sk_nulls_node __sk_common.skc_nulls_node
320 #define sk_refcnt __sk_common.skc_refcnt
321 #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
323 #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
324 #define sk_dontcopy_end __sk_common.skc_dontcopy_end
325 #define sk_hash __sk_common.skc_hash
326 #define sk_portpair __sk_common.skc_portpair
327 #define sk_num __sk_common.skc_num
328 #define sk_dport __sk_common.skc_dport
329 #define sk_addrpair __sk_common.skc_addrpair
330 #define sk_daddr __sk_common.skc_daddr
331 #define sk_rcv_saddr __sk_common.skc_rcv_saddr
332 #define sk_family __sk_common.skc_family
333 #define sk_state __sk_common.skc_state
334 #define sk_reuse __sk_common.skc_reuse
335 #define sk_reuseport __sk_common.skc_reuseport
336 #define sk_ipv6only __sk_common.skc_ipv6only
337 #define sk_net_refcnt __sk_common.skc_net_refcnt
338 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
339 #define sk_bind_node __sk_common.skc_bind_node
340 #define sk_prot __sk_common.skc_prot
341 #define sk_net __sk_common.skc_net
342 #define sk_v6_daddr __sk_common.skc_v6_daddr
343 #define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr
344 #define sk_cookie __sk_common.skc_cookie
345 #define sk_incoming_cpu __sk_common.skc_incoming_cpu
346 #define sk_flags __sk_common.skc_flags
347 #define sk_rxhash __sk_common.skc_rxhash
349 socket_lock_t sk_lock;
352 struct sk_buff_head sk_error_queue;
353 struct sk_buff_head sk_receive_queue;
355 * The backlog queue is special, it is always used with
356 * the per-socket spinlock held and requires low latency
357 * access. Therefore we special case it's implementation.
358 * Note : rmem_alloc is in this structure to fill a hole
359 * on 64bit arches, not because its logically part of
365 struct sk_buff *head;
366 struct sk_buff *tail;
368 #define sk_rmem_alloc sk_backlog.rmem_alloc
370 int sk_forward_alloc;
371 #ifdef CONFIG_NET_RX_BUSY_POLL
372 unsigned int sk_ll_usec;
373 /* ===== mostly read cache line ===== */
374 unsigned int sk_napi_id;
378 struct sk_filter __rcu *sk_filter;
380 struct socket_wq __rcu *sk_wq;
381 struct socket_wq *sk_wq_raw;
384 struct xfrm_policy __rcu *sk_policy[2];
386 struct dst_entry *sk_rx_dst;
387 struct dst_entry __rcu *sk_dst_cache;
388 atomic_t sk_omem_alloc;
391 /* ===== cache line for TX ===== */
393 atomic_t sk_wmem_alloc;
394 unsigned long sk_tsq_flags;
395 struct sk_buff *sk_send_head;
396 struct sk_buff_head sk_write_queue;
398 int sk_write_pending;
399 __u32 sk_dst_pending_confirm;
400 u32 sk_pacing_status; /* see enum sk_pacing */
402 struct timer_list sk_timer;
405 u32 sk_pacing_rate; /* bytes per second */
406 u32 sk_max_pacing_rate;
407 struct page_frag sk_frag;
408 netdev_features_t sk_route_caps;
409 netdev_features_t sk_route_nocaps;
411 unsigned int sk_gso_max_size;
416 * Because of non atomicity rules, all
417 * changes are protected by socket lock.
419 unsigned int __sk_flags_offset[0];
420 #ifdef __BIG_ENDIAN_BITFIELD
421 #define SK_FL_PROTO_SHIFT 16
422 #define SK_FL_PROTO_MASK 0x00ff0000
424 #define SK_FL_TYPE_SHIFT 0
425 #define SK_FL_TYPE_MASK 0x0000ffff
427 #define SK_FL_PROTO_SHIFT 8
428 #define SK_FL_PROTO_MASK 0x0000ff00
430 #define SK_FL_TYPE_SHIFT 16
431 #define SK_FL_TYPE_MASK 0xffff0000
434 kmemcheck_bitfield_begin(flags);
435 unsigned int sk_padding : 1,
442 #define SK_PROTOCOL_MAX U8_MAX
443 kmemcheck_bitfield_end(flags);
446 unsigned long sk_lingertime;
447 struct proto *sk_prot_creator;
448 rwlock_t sk_callback_lock;
452 u32 sk_max_ack_backlog;
454 struct pid *sk_peer_pid;
455 const struct cred *sk_peer_cred;
461 struct socket *sk_socket;
463 #ifdef CONFIG_SECURITY
466 struct sock_cgroup_data sk_cgrp_data;
467 struct mem_cgroup *sk_memcg;
468 void (*sk_state_change)(struct sock *sk);
469 void (*sk_data_ready)(struct sock *sk);
470 void (*sk_write_space)(struct sock *sk);
471 void (*sk_error_report)(struct sock *sk);
472 int (*sk_backlog_rcv)(struct sock *sk,
473 struct sk_buff *skb);
474 void (*sk_destruct)(struct sock *sk);
475 struct sock_reuseport __rcu *sk_reuseport_cb;
476 struct rcu_head sk_rcu;
481 SK_PACING_NEEDED = 1,
485 #define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
487 #define rcu_dereference_sk_user_data(sk) rcu_dereference(__sk_user_data((sk)))
488 #define rcu_assign_sk_user_data(sk, ptr) rcu_assign_pointer(__sk_user_data((sk)), ptr)
491 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
492 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
493 * on a socket means that the socket will reuse everybody else's port
494 * without looking at the other's sk_reuse value.
497 #define SK_NO_REUSE 0
498 #define SK_CAN_REUSE 1
499 #define SK_FORCE_REUSE 2
501 int sk_set_peek_off(struct sock *sk, int val);
503 static inline int sk_peek_offset(struct sock *sk, int flags)
505 if (unlikely(flags & MSG_PEEK)) {
506 s32 off = READ_ONCE(sk->sk_peek_off);
514 static inline void sk_peek_offset_bwd(struct sock *sk, int val)
516 s32 off = READ_ONCE(sk->sk_peek_off);
518 if (unlikely(off >= 0)) {
519 off = max_t(s32, off - val, 0);
520 WRITE_ONCE(sk->sk_peek_off, off);
524 static inline void sk_peek_offset_fwd(struct sock *sk, int val)
526 sk_peek_offset_bwd(sk, -val);
530 * Hashed lists helper routines
532 static inline struct sock *sk_entry(const struct hlist_node *node)
534 return hlist_entry(node, struct sock, sk_node);
537 static inline struct sock *__sk_head(const struct hlist_head *head)
539 return hlist_entry(head->first, struct sock, sk_node);
542 static inline struct sock *sk_head(const struct hlist_head *head)
544 return hlist_empty(head) ? NULL : __sk_head(head);
547 static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
549 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
552 static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
554 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
557 static inline struct sock *sk_next(const struct sock *sk)
559 return hlist_entry_safe(sk->sk_node.next, struct sock, sk_node);
562 static inline struct sock *sk_nulls_next(const struct sock *sk)
564 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
565 hlist_nulls_entry(sk->sk_nulls_node.next,
566 struct sock, sk_nulls_node) :
570 static inline bool sk_unhashed(const struct sock *sk)
572 return hlist_unhashed(&sk->sk_node);
575 static inline bool sk_hashed(const struct sock *sk)
577 return !sk_unhashed(sk);
580 static inline void sk_node_init(struct hlist_node *node)
585 static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
590 static inline void __sk_del_node(struct sock *sk)
592 __hlist_del(&sk->sk_node);
595 /* NB: equivalent to hlist_del_init_rcu */
596 static inline bool __sk_del_node_init(struct sock *sk)
600 sk_node_init(&sk->sk_node);
606 /* Grab socket reference count. This operation is valid only
607 when sk is ALREADY grabbed f.e. it is found in hash table
608 or a list and the lookup is made under lock preventing hash table
612 static __always_inline void sock_hold(struct sock *sk)
614 atomic_inc(&sk->sk_refcnt);
617 /* Ungrab socket in the context, which assumes that socket refcnt
618 cannot hit zero, f.e. it is true in context of any socketcall.
620 static __always_inline void __sock_put(struct sock *sk)
622 atomic_dec(&sk->sk_refcnt);
625 static inline bool sk_del_node_init(struct sock *sk)
627 bool rc = __sk_del_node_init(sk);
630 /* paranoid for a while -acme */
631 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
636 #define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
638 static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
641 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
647 static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
649 bool rc = __sk_nulls_del_node_init_rcu(sk);
652 /* paranoid for a while -acme */
653 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
659 static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
661 hlist_add_head(&sk->sk_node, list);
664 static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
667 __sk_add_node(sk, list);
670 static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
673 if (IS_ENABLED(CONFIG_IPV6) && sk->sk_reuseport &&
674 sk->sk_family == AF_INET6)
675 hlist_add_tail_rcu(&sk->sk_node, list);
677 hlist_add_head_rcu(&sk->sk_node, list);
680 static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
682 if (IS_ENABLED(CONFIG_IPV6) && sk->sk_reuseport &&
683 sk->sk_family == AF_INET6)
684 hlist_nulls_add_tail_rcu(&sk->sk_nulls_node, list);
686 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
689 static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
692 __sk_nulls_add_node_rcu(sk, list);
695 static inline void __sk_del_bind_node(struct sock *sk)
697 __hlist_del(&sk->sk_bind_node);
700 static inline void sk_add_bind_node(struct sock *sk,
701 struct hlist_head *list)
703 hlist_add_head(&sk->sk_bind_node, list);
706 #define sk_for_each(__sk, list) \
707 hlist_for_each_entry(__sk, list, sk_node)
708 #define sk_for_each_rcu(__sk, list) \
709 hlist_for_each_entry_rcu(__sk, list, sk_node)
710 #define sk_nulls_for_each(__sk, node, list) \
711 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
712 #define sk_nulls_for_each_rcu(__sk, node, list) \
713 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
714 #define sk_for_each_from(__sk) \
715 hlist_for_each_entry_from(__sk, sk_node)
716 #define sk_nulls_for_each_from(__sk, node) \
717 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
718 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
719 #define sk_for_each_safe(__sk, tmp, list) \
720 hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
721 #define sk_for_each_bound(__sk, list) \
722 hlist_for_each_entry(__sk, list, sk_bind_node)
725 * sk_for_each_entry_offset_rcu - iterate over a list at a given struct offset
726 * @tpos: the type * to use as a loop cursor.
727 * @pos: the &struct hlist_node to use as a loop cursor.
728 * @head: the head for your list.
729 * @offset: offset of hlist_node within the struct.
732 #define sk_for_each_entry_offset_rcu(tpos, pos, head, offset) \
733 for (pos = rcu_dereference((head)->first); \
735 ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;}); \
736 pos = rcu_dereference(pos->next))
738 static inline struct user_namespace *sk_user_ns(struct sock *sk)
740 /* Careful only use this in a context where these parameters
741 * can not change and must all be valid, such as recvmsg from
744 return sk->sk_socket->file->f_cred->user_ns;
758 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
759 SOCK_DBG, /* %SO_DEBUG setting */
760 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
761 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
762 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
763 SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
764 SOCK_MEMALLOC, /* VM depends on this socket for swapping */
765 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
766 SOCK_FASYNC, /* fasync() active */
768 SOCK_ZEROCOPY, /* buffers from userspace */
769 SOCK_WIFI_STATUS, /* push wifi status to userspace */
770 SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
771 * Will use last 4 bytes of packet sent from
772 * user-space instead.
774 SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */
775 SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */
776 SOCK_RCU_FREE, /* wait rcu grace period in sk_destruct() */
779 #define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
781 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
783 nsk->sk_flags = osk->sk_flags;
786 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
788 __set_bit(flag, &sk->sk_flags);
791 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
793 __clear_bit(flag, &sk->sk_flags);
796 static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
798 return test_bit(flag, &sk->sk_flags);
802 extern struct static_key memalloc_socks;
803 static inline int sk_memalloc_socks(void)
805 return static_key_false(&memalloc_socks);
809 static inline int sk_memalloc_socks(void)
816 static inline gfp_t sk_gfp_mask(const struct sock *sk, gfp_t gfp_mask)
818 return gfp_mask | (sk->sk_allocation & __GFP_MEMALLOC);
821 static inline void sk_acceptq_removed(struct sock *sk)
823 sk->sk_ack_backlog--;
826 static inline void sk_acceptq_added(struct sock *sk)
828 sk->sk_ack_backlog++;
831 static inline bool sk_acceptq_is_full(const struct sock *sk)
833 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
837 * Compute minimal free write space needed to queue new packets.
839 static inline int sk_stream_min_wspace(const struct sock *sk)
841 return sk->sk_wmem_queued >> 1;
844 static inline int sk_stream_wspace(const struct sock *sk)
846 return sk->sk_sndbuf - sk->sk_wmem_queued;
849 void sk_stream_write_space(struct sock *sk);
851 /* OOB backlog add */
852 static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
854 /* dont let skb dst not refcounted, we are going to leave rcu lock */
855 skb_dst_force_safe(skb);
857 if (!sk->sk_backlog.tail)
858 sk->sk_backlog.head = skb;
860 sk->sk_backlog.tail->next = skb;
862 sk->sk_backlog.tail = skb;
867 * Take into account size of receive queue and backlog queue
868 * Do not take into account this skb truesize,
869 * to allow even a single big packet to come.
871 static inline bool sk_rcvqueues_full(const struct sock *sk, unsigned int limit)
873 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
875 return qsize > limit;
878 /* The per-socket spinlock must be held here. */
879 static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
882 if (sk_rcvqueues_full(sk, limit))
886 * If the skb was allocated from pfmemalloc reserves, only
887 * allow SOCK_MEMALLOC sockets to use it as this socket is
888 * helping free memory
890 if (skb_pfmemalloc(skb) && !sock_flag(sk, SOCK_MEMALLOC))
893 __sk_add_backlog(sk, skb);
894 sk->sk_backlog.len += skb->truesize;
898 int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
900 static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
902 if (sk_memalloc_socks() && skb_pfmemalloc(skb))
903 return __sk_backlog_rcv(sk, skb);
905 return sk->sk_backlog_rcv(sk, skb);
908 static inline void sk_incoming_cpu_update(struct sock *sk)
910 sk->sk_incoming_cpu = raw_smp_processor_id();
913 static inline void sock_rps_record_flow_hash(__u32 hash)
916 struct rps_sock_flow_table *sock_flow_table;
919 sock_flow_table = rcu_dereference(rps_sock_flow_table);
920 rps_record_sock_flow(sock_flow_table, hash);
925 static inline void sock_rps_record_flow(const struct sock *sk)
928 if (static_key_false(&rfs_needed)) {
929 /* Reading sk->sk_rxhash might incur an expensive cache line
932 * TCP_ESTABLISHED does cover almost all states where RFS
933 * might be useful, and is cheaper [1] than testing :
934 * IPv4: inet_sk(sk)->inet_daddr
935 * IPv6: ipv6_addr_any(&sk->sk_v6_daddr)
936 * OR an additional socket flag
937 * [1] : sk_state and sk_prot are in the same cache line.
939 if (sk->sk_state == TCP_ESTABLISHED)
940 sock_rps_record_flow_hash(sk->sk_rxhash);
945 static inline void sock_rps_save_rxhash(struct sock *sk,
946 const struct sk_buff *skb)
949 if (unlikely(sk->sk_rxhash != skb->hash))
950 sk->sk_rxhash = skb->hash;
954 static inline void sock_rps_reset_rxhash(struct sock *sk)
961 #define sk_wait_event(__sk, __timeo, __condition, __wait) \
963 release_sock(__sk); \
964 __rc = __condition; \
966 *(__timeo) = wait_woken(__wait, \
967 TASK_INTERRUPTIBLE, \
970 sched_annotate_sleep(); \
972 __rc = __condition; \
976 int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
977 int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
978 void sk_stream_wait_close(struct sock *sk, long timeo_p);
979 int sk_stream_error(struct sock *sk, int flags, int err);
980 void sk_stream_kill_queues(struct sock *sk);
981 void sk_set_memalloc(struct sock *sk);
982 void sk_clear_memalloc(struct sock *sk);
984 void __sk_flush_backlog(struct sock *sk);
986 static inline bool sk_flush_backlog(struct sock *sk)
988 if (unlikely(READ_ONCE(sk->sk_backlog.tail))) {
989 __sk_flush_backlog(sk);
995 int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb);
997 struct request_sock_ops;
998 struct timewait_sock_ops;
999 struct inet_hashinfo;
1000 struct raw_hashinfo;
1001 struct smc_hashinfo;
1005 * caches using SLAB_TYPESAFE_BY_RCU should let .next pointer from nulls nodes
1006 * un-modified. Special care is taken when initializing object to zero.
1008 static inline void sk_prot_clear_nulls(struct sock *sk, int size)
1010 if (offsetof(struct sock, sk_node.next) != 0)
1011 memset(sk, 0, offsetof(struct sock, sk_node.next));
1012 memset(&sk->sk_node.pprev, 0,
1013 size - offsetof(struct sock, sk_node.pprev));
1016 /* Networking protocol blocks we attach to sockets.
1017 * socket layer -> transport layer interface
1020 void (*close)(struct sock *sk,
1022 int (*connect)(struct sock *sk,
1023 struct sockaddr *uaddr,
1025 int (*disconnect)(struct sock *sk, int flags);
1027 struct sock * (*accept)(struct sock *sk, int flags, int *err,
1030 int (*ioctl)(struct sock *sk, int cmd,
1032 int (*init)(struct sock *sk);
1033 void (*destroy)(struct sock *sk);
1034 void (*shutdown)(struct sock *sk, int how);
1035 int (*setsockopt)(struct sock *sk, int level,
1036 int optname, char __user *optval,
1037 unsigned int optlen);
1038 int (*getsockopt)(struct sock *sk, int level,
1039 int optname, char __user *optval,
1040 int __user *option);
1041 void (*keepalive)(struct sock *sk, int valbool);
1042 #ifdef CONFIG_COMPAT
1043 int (*compat_setsockopt)(struct sock *sk,
1045 int optname, char __user *optval,
1046 unsigned int optlen);
1047 int (*compat_getsockopt)(struct sock *sk,
1049 int optname, char __user *optval,
1050 int __user *option);
1051 int (*compat_ioctl)(struct sock *sk,
1052 unsigned int cmd, unsigned long arg);
1054 int (*sendmsg)(struct sock *sk, struct msghdr *msg,
1056 int (*recvmsg)(struct sock *sk, struct msghdr *msg,
1057 size_t len, int noblock, int flags,
1059 int (*sendpage)(struct sock *sk, struct page *page,
1060 int offset, size_t size, int flags);
1061 int (*bind)(struct sock *sk,
1062 struct sockaddr *uaddr, int addr_len);
1064 int (*backlog_rcv) (struct sock *sk,
1065 struct sk_buff *skb);
1067 void (*release_cb)(struct sock *sk);
1069 /* Keeping track of sk's, looking them up, and port selection methods. */
1070 int (*hash)(struct sock *sk);
1071 void (*unhash)(struct sock *sk);
1072 void (*rehash)(struct sock *sk);
1073 int (*get_port)(struct sock *sk, unsigned short snum);
1075 /* Keeping track of sockets in use */
1076 #ifdef CONFIG_PROC_FS
1077 unsigned int inuse_idx;
1080 bool (*stream_memory_free)(const struct sock *sk);
1081 /* Memory pressure */
1082 void (*enter_memory_pressure)(struct sock *sk);
1083 void (*leave_memory_pressure)(struct sock *sk);
1084 atomic_long_t *memory_allocated; /* Current allocated memory. */
1085 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
1087 * Pressure flag: try to collapse.
1088 * Technical note: it is used by multiple contexts non atomically.
1089 * All the __sk_mem_schedule() is of this nature: accounting
1090 * is strict, actions are advisory and have some latency.
1092 unsigned long *memory_pressure;
1099 struct kmem_cache *slab;
1100 unsigned int obj_size;
1103 struct percpu_counter *orphan_count;
1105 struct request_sock_ops *rsk_prot;
1106 struct timewait_sock_ops *twsk_prot;
1109 struct inet_hashinfo *hashinfo;
1110 struct udp_table *udp_table;
1111 struct raw_hashinfo *raw_hash;
1112 struct smc_hashinfo *smc_hash;
1115 struct module *owner;
1119 struct list_head node;
1120 #ifdef SOCK_REFCNT_DEBUG
1123 int (*diag_destroy)(struct sock *sk, int err);
1126 int proto_register(struct proto *prot, int alloc_slab);
1127 void proto_unregister(struct proto *prot);
1129 #ifdef SOCK_REFCNT_DEBUG
1130 static inline void sk_refcnt_debug_inc(struct sock *sk)
1132 atomic_inc(&sk->sk_prot->socks);
1135 static inline void sk_refcnt_debug_dec(struct sock *sk)
1137 atomic_dec(&sk->sk_prot->socks);
1138 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
1139 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
1142 static inline void sk_refcnt_debug_release(const struct sock *sk)
1144 if (atomic_read(&sk->sk_refcnt) != 1)
1145 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
1146 sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
1148 #else /* SOCK_REFCNT_DEBUG */
1149 #define sk_refcnt_debug_inc(sk) do { } while (0)
1150 #define sk_refcnt_debug_dec(sk) do { } while (0)
1151 #define sk_refcnt_debug_release(sk) do { } while (0)
1152 #endif /* SOCK_REFCNT_DEBUG */
1154 static inline bool sk_stream_memory_free(const struct sock *sk)
1156 if (sk->sk_wmem_queued >= sk->sk_sndbuf)
1159 return sk->sk_prot->stream_memory_free ?
1160 sk->sk_prot->stream_memory_free(sk) : true;
1163 static inline bool sk_stream_is_writeable(const struct sock *sk)
1165 return sk_stream_wspace(sk) >= sk_stream_min_wspace(sk) &&
1166 sk_stream_memory_free(sk);
1169 static inline int sk_under_cgroup_hierarchy(struct sock *sk,
1170 struct cgroup *ancestor)
1172 #ifdef CONFIG_SOCK_CGROUP_DATA
1173 return cgroup_is_descendant(sock_cgroup_ptr(&sk->sk_cgrp_data),
1180 static inline bool sk_has_memory_pressure(const struct sock *sk)
1182 return sk->sk_prot->memory_pressure != NULL;
1185 static inline bool sk_under_memory_pressure(const struct sock *sk)
1187 if (!sk->sk_prot->memory_pressure)
1190 if (mem_cgroup_sockets_enabled && sk->sk_memcg &&
1191 mem_cgroup_under_socket_pressure(sk->sk_memcg))
1194 return !!*sk->sk_prot->memory_pressure;
1198 sk_memory_allocated(const struct sock *sk)
1200 return atomic_long_read(sk->sk_prot->memory_allocated);
1204 sk_memory_allocated_add(struct sock *sk, int amt)
1206 return atomic_long_add_return(amt, sk->sk_prot->memory_allocated);
1210 sk_memory_allocated_sub(struct sock *sk, int amt)
1212 atomic_long_sub(amt, sk->sk_prot->memory_allocated);
1215 static inline void sk_sockets_allocated_dec(struct sock *sk)
1217 percpu_counter_dec(sk->sk_prot->sockets_allocated);
1220 static inline void sk_sockets_allocated_inc(struct sock *sk)
1222 percpu_counter_inc(sk->sk_prot->sockets_allocated);
1226 sk_sockets_allocated_read_positive(struct sock *sk)
1228 return percpu_counter_read_positive(sk->sk_prot->sockets_allocated);
1232 proto_sockets_allocated_sum_positive(struct proto *prot)
1234 return percpu_counter_sum_positive(prot->sockets_allocated);
1238 proto_memory_allocated(struct proto *prot)
1240 return atomic_long_read(prot->memory_allocated);
1244 proto_memory_pressure(struct proto *prot)
1246 if (!prot->memory_pressure)
1248 return !!*prot->memory_pressure;
1252 #ifdef CONFIG_PROC_FS
1253 /* Called with local bh disabled */
1254 void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1255 int sock_prot_inuse_get(struct net *net, struct proto *proto);
1257 static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
1264 /* With per-bucket locks this operation is not-atomic, so that
1265 * this version is not worse.
1267 static inline int __sk_prot_rehash(struct sock *sk)
1269 sk->sk_prot->unhash(sk);
1270 return sk->sk_prot->hash(sk);
1273 /* About 10 seconds */
1274 #define SOCK_DESTROY_TIME (10*HZ)
1276 /* Sockets 0-1023 can't be bound to unless you are superuser */
1277 #define PROT_SOCK 1024
1279 #define SHUTDOWN_MASK 3
1280 #define RCV_SHUTDOWN 1
1281 #define SEND_SHUTDOWN 2
1283 #define SOCK_SNDBUF_LOCK 1
1284 #define SOCK_RCVBUF_LOCK 2
1285 #define SOCK_BINDADDR_LOCK 4
1286 #define SOCK_BINDPORT_LOCK 8
1288 struct socket_alloc {
1289 struct socket socket;
1290 struct inode vfs_inode;
1293 static inline struct socket *SOCKET_I(struct inode *inode)
1295 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1298 static inline struct inode *SOCK_INODE(struct socket *socket)
1300 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1304 * Functions for memory accounting
1306 int __sk_mem_raise_allocated(struct sock *sk, int size, int amt, int kind);
1307 int __sk_mem_schedule(struct sock *sk, int size, int kind);
1308 void __sk_mem_reduce_allocated(struct sock *sk, int amount);
1309 void __sk_mem_reclaim(struct sock *sk, int amount);
1311 /* We used to have PAGE_SIZE here, but systems with 64KB pages
1312 * do not necessarily have 16x time more memory than 4KB ones.
1314 #define SK_MEM_QUANTUM 4096
1315 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1316 #define SK_MEM_SEND 0
1317 #define SK_MEM_RECV 1
1319 /* sysctl_mem values are in pages, we convert them in SK_MEM_QUANTUM units */
1320 static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1322 long val = sk->sk_prot->sysctl_mem[index];
1324 #if PAGE_SIZE > SK_MEM_QUANTUM
1325 val <<= PAGE_SHIFT - SK_MEM_QUANTUM_SHIFT;
1326 #elif PAGE_SIZE < SK_MEM_QUANTUM
1327 val >>= SK_MEM_QUANTUM_SHIFT - PAGE_SHIFT;
1332 static inline int sk_mem_pages(int amt)
1334 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
1337 static inline bool sk_has_account(struct sock *sk)
1339 /* return true if protocol supports memory accounting */
1340 return !!sk->sk_prot->memory_allocated;
1343 static inline bool sk_wmem_schedule(struct sock *sk, int size)
1345 if (!sk_has_account(sk))
1347 return size <= sk->sk_forward_alloc ||
1348 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1352 sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size)
1354 if (!sk_has_account(sk))
1356 return size<= sk->sk_forward_alloc ||
1357 __sk_mem_schedule(sk, size, SK_MEM_RECV) ||
1358 skb_pfmemalloc(skb);
1361 static inline void sk_mem_reclaim(struct sock *sk)
1363 if (!sk_has_account(sk))
1365 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1366 __sk_mem_reclaim(sk, sk->sk_forward_alloc);
1369 static inline void sk_mem_reclaim_partial(struct sock *sk)
1371 if (!sk_has_account(sk))
1373 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1374 __sk_mem_reclaim(sk, sk->sk_forward_alloc - 1);
1377 static inline void sk_mem_charge(struct sock *sk, int size)
1379 if (!sk_has_account(sk))
1381 sk->sk_forward_alloc -= size;
1384 static inline void sk_mem_uncharge(struct sock *sk, int size)
1386 if (!sk_has_account(sk))
1388 sk->sk_forward_alloc += size;
1390 /* Avoid a possible overflow.
1391 * TCP send queues can make this happen, if sk_mem_reclaim()
1392 * is not called and more than 2 GBytes are released at once.
1394 * If we reach 2 MBytes, reclaim 1 MBytes right now, there is
1395 * no need to hold that much forward allocation anyway.
1397 if (unlikely(sk->sk_forward_alloc >= 1 << 21))
1398 __sk_mem_reclaim(sk, 1 << 20);
1401 static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1403 sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1404 sk->sk_wmem_queued -= skb->truesize;
1405 sk_mem_uncharge(sk, skb->truesize);
1409 static inline void sock_release_ownership(struct sock *sk)
1411 if (sk->sk_lock.owned) {
1412 sk->sk_lock.owned = 0;
1414 /* The sk_lock has mutex_unlock() semantics: */
1415 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
1420 * Macro so as to not evaluate some arguments when
1421 * lockdep is not enabled.
1423 * Mark both the sk_lock and the sk_lock.slock as a
1424 * per-address-family lock class.
1426 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1428 sk->sk_lock.owned = 0; \
1429 init_waitqueue_head(&sk->sk_lock.wq); \
1430 spin_lock_init(&(sk)->sk_lock.slock); \
1431 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1432 sizeof((sk)->sk_lock)); \
1433 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1435 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1438 #ifdef CONFIG_LOCKDEP
1439 static inline bool lockdep_sock_is_held(const struct sock *csk)
1441 struct sock *sk = (struct sock *)csk;
1443 return lockdep_is_held(&sk->sk_lock) ||
1444 lockdep_is_held(&sk->sk_lock.slock);
1448 void lock_sock_nested(struct sock *sk, int subclass);
1450 static inline void lock_sock(struct sock *sk)
1452 lock_sock_nested(sk, 0);
1455 void release_sock(struct sock *sk);
1457 /* BH context may only use the following locking interface. */
1458 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
1459 #define bh_lock_sock_nested(__sk) \
1460 spin_lock_nested(&((__sk)->sk_lock.slock), \
1461 SINGLE_DEPTH_NESTING)
1462 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1464 bool lock_sock_fast(struct sock *sk);
1466 * unlock_sock_fast - complement of lock_sock_fast
1470 * fast unlock socket for user context.
1471 * If slow mode is on, we call regular release_sock()
1473 static inline void unlock_sock_fast(struct sock *sk, bool slow)
1478 spin_unlock_bh(&sk->sk_lock.slock);
1481 /* Used by processes to "lock" a socket state, so that
1482 * interrupts and bottom half handlers won't change it
1483 * from under us. It essentially blocks any incoming
1484 * packets, so that we won't get any new data or any
1485 * packets that change the state of the socket.
1487 * While locked, BH processing will add new packets to
1488 * the backlog queue. This queue is processed by the
1489 * owner of the socket lock right before it is released.
1491 * Since ~2.3.5 it is also exclusive sleep lock serializing
1492 * accesses from user process context.
1495 static inline void sock_owned_by_me(const struct sock *sk)
1497 #ifdef CONFIG_LOCKDEP
1498 WARN_ON_ONCE(!lockdep_sock_is_held(sk) && debug_locks);
1502 static inline bool sock_owned_by_user(const struct sock *sk)
1504 sock_owned_by_me(sk);
1505 return sk->sk_lock.owned;
1508 /* no reclassification while locks are held */
1509 static inline bool sock_allow_reclassification(const struct sock *csk)
1511 struct sock *sk = (struct sock *)csk;
1513 return !sk->sk_lock.owned && !spin_is_locked(&sk->sk_lock.slock);
1516 struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
1517 struct proto *prot, int kern);
1518 void sk_free(struct sock *sk);
1519 void sk_destruct(struct sock *sk);
1520 struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority);
1521 void sk_free_unlock_clone(struct sock *sk);
1523 struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
1525 void __sock_wfree(struct sk_buff *skb);
1526 void sock_wfree(struct sk_buff *skb);
1527 void skb_orphan_partial(struct sk_buff *skb);
1528 void sock_rfree(struct sk_buff *skb);
1529 void sock_efree(struct sk_buff *skb);
1531 void sock_edemux(struct sk_buff *skb);
1533 #define sock_edemux sock_efree
1536 int sock_setsockopt(struct socket *sock, int level, int op,
1537 char __user *optval, unsigned int optlen);
1539 int sock_getsockopt(struct socket *sock, int level, int op,
1540 char __user *optval, int __user *optlen);
1541 struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1542 int noblock, int *errcode);
1543 struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1544 unsigned long data_len, int noblock,
1545 int *errcode, int max_page_order);
1546 void *sock_kmalloc(struct sock *sk, int size, gfp_t priority);
1547 void sock_kfree_s(struct sock *sk, void *mem, int size);
1548 void sock_kzfree_s(struct sock *sk, void *mem, int size);
1549 void sk_send_sigurg(struct sock *sk);
1551 struct sockcm_cookie {
1556 int __sock_cmsg_send(struct sock *sk, struct msghdr *msg, struct cmsghdr *cmsg,
1557 struct sockcm_cookie *sockc);
1558 int sock_cmsg_send(struct sock *sk, struct msghdr *msg,
1559 struct sockcm_cookie *sockc);
1562 * Functions to fill in entries in struct proto_ops when a protocol
1563 * does not implement a particular function.
1565 int sock_no_bind(struct socket *, struct sockaddr *, int);
1566 int sock_no_connect(struct socket *, struct sockaddr *, int, int);
1567 int sock_no_socketpair(struct socket *, struct socket *);
1568 int sock_no_accept(struct socket *, struct socket *, int, bool);
1569 int sock_no_getname(struct socket *, struct sockaddr *, int *, int);
1570 unsigned int sock_no_poll(struct file *, struct socket *,
1571 struct poll_table_struct *);
1572 int sock_no_ioctl(struct socket *, unsigned int, unsigned long);
1573 int sock_no_listen(struct socket *, int);
1574 int sock_no_shutdown(struct socket *, int);
1575 int sock_no_getsockopt(struct socket *, int , int, char __user *, int __user *);
1576 int sock_no_setsockopt(struct socket *, int, int, char __user *, unsigned int);
1577 int sock_no_sendmsg(struct socket *, struct msghdr *, size_t);
1578 int sock_no_recvmsg(struct socket *, struct msghdr *, size_t, int);
1579 int sock_no_mmap(struct file *file, struct socket *sock,
1580 struct vm_area_struct *vma);
1581 ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset,
1582 size_t size, int flags);
1585 * Functions to fill in entries in struct proto_ops when a protocol
1586 * uses the inet style.
1588 int sock_common_getsockopt(struct socket *sock, int level, int optname,
1589 char __user *optval, int __user *optlen);
1590 int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
1592 int sock_common_setsockopt(struct socket *sock, int level, int optname,
1593 char __user *optval, unsigned int optlen);
1594 int compat_sock_common_getsockopt(struct socket *sock, int level,
1595 int optname, char __user *optval, int __user *optlen);
1596 int compat_sock_common_setsockopt(struct socket *sock, int level,
1597 int optname, char __user *optval, unsigned int optlen);
1599 void sk_common_release(struct sock *sk);
1602 * Default socket callbacks and setup code
1605 /* Initialise core socket variables */
1606 void sock_init_data(struct socket *sock, struct sock *sk);
1609 * Socket reference counting postulates.
1611 * * Each user of socket SHOULD hold a reference count.
1612 * * Each access point to socket (an hash table bucket, reference from a list,
1613 * running timer, skb in flight MUST hold a reference count.
1614 * * When reference count hits 0, it means it will never increase back.
1615 * * When reference count hits 0, it means that no references from
1616 * outside exist to this socket and current process on current CPU
1617 * is last user and may/should destroy this socket.
1618 * * sk_free is called from any context: process, BH, IRQ. When
1619 * it is called, socket has no references from outside -> sk_free
1620 * may release descendant resources allocated by the socket, but
1621 * to the time when it is called, socket is NOT referenced by any
1622 * hash tables, lists etc.
1623 * * Packets, delivered from outside (from network or from another process)
1624 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1625 * when they sit in queue. Otherwise, packets will leak to hole, when
1626 * socket is looked up by one cpu and unhasing is made by another CPU.
1627 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1628 * (leak to backlog). Packet socket does all the processing inside
1629 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1630 * use separate SMP lock, so that they are prone too.
1633 /* Ungrab socket and destroy it, if it was the last reference. */
1634 static inline void sock_put(struct sock *sk)
1636 if (atomic_dec_and_test(&sk->sk_refcnt))
1639 /* Generic version of sock_put(), dealing with all sockets
1640 * (TCP_TIMEWAIT, TCP_NEW_SYN_RECV, ESTABLISHED...)
1642 void sock_gen_put(struct sock *sk);
1644 int __sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested,
1645 unsigned int trim_cap, bool refcounted);
1646 static inline int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1649 return __sk_receive_skb(sk, skb, nested, 1, true);
1652 static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1654 sk->sk_tx_queue_mapping = tx_queue;
1657 static inline void sk_tx_queue_clear(struct sock *sk)
1659 sk->sk_tx_queue_mapping = -1;
1662 static inline int sk_tx_queue_get(const struct sock *sk)
1664 return sk ? sk->sk_tx_queue_mapping : -1;
1667 static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1669 sk_tx_queue_clear(sk);
1670 sk->sk_socket = sock;
1673 static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1675 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1676 return &rcu_dereference_raw(sk->sk_wq)->wait;
1678 /* Detach socket from process context.
1679 * Announce socket dead, detach it from wait queue and inode.
1680 * Note that parent inode held reference count on this struct sock,
1681 * we do not release it in this function, because protocol
1682 * probably wants some additional cleanups or even continuing
1683 * to work with this socket (TCP).
1685 static inline void sock_orphan(struct sock *sk)
1687 write_lock_bh(&sk->sk_callback_lock);
1688 sock_set_flag(sk, SOCK_DEAD);
1689 sk_set_socket(sk, NULL);
1691 write_unlock_bh(&sk->sk_callback_lock);
1694 static inline void sock_graft(struct sock *sk, struct socket *parent)
1696 write_lock_bh(&sk->sk_callback_lock);
1697 sk->sk_wq = parent->wq;
1699 sk_set_socket(sk, parent);
1700 sk->sk_uid = SOCK_INODE(parent)->i_uid;
1701 security_sock_graft(sk, parent);
1702 write_unlock_bh(&sk->sk_callback_lock);
1705 kuid_t sock_i_uid(struct sock *sk);
1706 unsigned long sock_i_ino(struct sock *sk);
1708 static inline kuid_t sock_net_uid(const struct net *net, const struct sock *sk)
1710 return sk ? sk->sk_uid : make_kuid(net->user_ns, 0);
1713 static inline u32 net_tx_rndhash(void)
1715 u32 v = prandom_u32();
1720 static inline void sk_set_txhash(struct sock *sk)
1722 sk->sk_txhash = net_tx_rndhash();
1725 static inline void sk_rethink_txhash(struct sock *sk)
1731 static inline struct dst_entry *
1732 __sk_dst_get(struct sock *sk)
1734 return rcu_dereference_check(sk->sk_dst_cache,
1735 lockdep_sock_is_held(sk));
1738 static inline struct dst_entry *
1739 sk_dst_get(struct sock *sk)
1741 struct dst_entry *dst;
1744 dst = rcu_dereference(sk->sk_dst_cache);
1745 if (dst && !atomic_inc_not_zero(&dst->__refcnt))
1751 static inline void dst_negative_advice(struct sock *sk)
1753 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1755 sk_rethink_txhash(sk);
1757 if (dst && dst->ops->negative_advice) {
1758 ndst = dst->ops->negative_advice(dst);
1761 rcu_assign_pointer(sk->sk_dst_cache, ndst);
1762 sk_tx_queue_clear(sk);
1763 sk->sk_dst_pending_confirm = 0;
1769 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
1771 struct dst_entry *old_dst;
1773 sk_tx_queue_clear(sk);
1774 sk->sk_dst_pending_confirm = 0;
1775 old_dst = rcu_dereference_protected(sk->sk_dst_cache,
1776 lockdep_sock_is_held(sk));
1777 rcu_assign_pointer(sk->sk_dst_cache, dst);
1778 dst_release(old_dst);
1782 sk_dst_set(struct sock *sk, struct dst_entry *dst)
1784 struct dst_entry *old_dst;
1786 sk_tx_queue_clear(sk);
1787 sk->sk_dst_pending_confirm = 0;
1788 old_dst = xchg((__force struct dst_entry **)&sk->sk_dst_cache, dst);
1789 dst_release(old_dst);
1793 __sk_dst_reset(struct sock *sk)
1795 __sk_dst_set(sk, NULL);
1799 sk_dst_reset(struct sock *sk)
1801 sk_dst_set(sk, NULL);
1804 struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1806 struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1808 static inline void sk_dst_confirm(struct sock *sk)
1810 if (!sk->sk_dst_pending_confirm)
1811 sk->sk_dst_pending_confirm = 1;
1814 static inline void sock_confirm_neigh(struct sk_buff *skb, struct neighbour *n)
1816 if (skb_get_dst_pending_confirm(skb)) {
1817 struct sock *sk = skb->sk;
1818 unsigned long now = jiffies;
1820 /* avoid dirtying neighbour */
1821 if (n->confirmed != now)
1823 if (sk && sk->sk_dst_pending_confirm)
1824 sk->sk_dst_pending_confirm = 0;
1828 bool sk_mc_loop(struct sock *sk);
1830 static inline bool sk_can_gso(const struct sock *sk)
1832 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1835 void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
1837 static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
1839 sk->sk_route_nocaps |= flags;
1840 sk->sk_route_caps &= ~flags;
1843 static inline bool sk_check_csum_caps(struct sock *sk)
1845 return (sk->sk_route_caps & NETIF_F_HW_CSUM) ||
1846 (sk->sk_family == PF_INET &&
1847 (sk->sk_route_caps & NETIF_F_IP_CSUM)) ||
1848 (sk->sk_family == PF_INET6 &&
1849 (sk->sk_route_caps & NETIF_F_IPV6_CSUM));
1852 static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1853 struct iov_iter *from, char *to,
1854 int copy, int offset)
1856 if (skb->ip_summed == CHECKSUM_NONE) {
1858 if (!csum_and_copy_from_iter_full(to, copy, &csum, from))
1860 skb->csum = csum_block_add(skb->csum, csum, offset);
1861 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1862 if (!copy_from_iter_full_nocache(to, copy, from))
1864 } else if (!copy_from_iter_full(to, copy, from))
1870 static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
1871 struct iov_iter *from, int copy)
1873 int err, offset = skb->len;
1875 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1878 __skb_trim(skb, offset);
1883 static inline int skb_copy_to_page_nocache(struct sock *sk, struct iov_iter *from,
1884 struct sk_buff *skb,
1890 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1896 skb->data_len += copy;
1897 skb->truesize += copy;
1898 sk->sk_wmem_queued += copy;
1899 sk_mem_charge(sk, copy);
1904 * sk_wmem_alloc_get - returns write allocations
1907 * Returns sk_wmem_alloc minus initial offset of one
1909 static inline int sk_wmem_alloc_get(const struct sock *sk)
1911 return atomic_read(&sk->sk_wmem_alloc) - 1;
1915 * sk_rmem_alloc_get - returns read allocations
1918 * Returns sk_rmem_alloc
1920 static inline int sk_rmem_alloc_get(const struct sock *sk)
1922 return atomic_read(&sk->sk_rmem_alloc);
1926 * sk_has_allocations - check if allocations are outstanding
1929 * Returns true if socket has write or read allocations
1931 static inline bool sk_has_allocations(const struct sock *sk)
1933 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1937 * skwq_has_sleeper - check if there are any waiting processes
1938 * @wq: struct socket_wq
1940 * Returns true if socket_wq has waiting processes
1942 * The purpose of the skwq_has_sleeper and sock_poll_wait is to wrap the memory
1943 * barrier call. They were added due to the race found within the tcp code.
1945 * Consider following tcp code paths:
1949 * sys_select receive packet
1951 * __add_wait_queue update tp->rcv_nxt
1953 * tp->rcv_nxt check sock_def_readable
1955 * schedule rcu_read_lock();
1956 * wq = rcu_dereference(sk->sk_wq);
1957 * if (wq && waitqueue_active(&wq->wait))
1958 * wake_up_interruptible(&wq->wait)
1962 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1963 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1964 * could then endup calling schedule and sleep forever if there are no more
1965 * data on the socket.
1968 static inline bool skwq_has_sleeper(struct socket_wq *wq)
1970 return wq && wq_has_sleeper(&wq->wait);
1974 * sock_poll_wait - place memory barrier behind the poll_wait call.
1976 * @wait_address: socket wait queue
1979 * See the comments in the wq_has_sleeper function.
1981 static inline void sock_poll_wait(struct file *filp,
1982 wait_queue_head_t *wait_address, poll_table *p)
1984 if (!poll_does_not_wait(p) && wait_address) {
1985 poll_wait(filp, wait_address, p);
1986 /* We need to be sure we are in sync with the
1987 * socket flags modification.
1989 * This memory barrier is paired in the wq_has_sleeper.
1995 static inline void skb_set_hash_from_sk(struct sk_buff *skb, struct sock *sk)
1997 if (sk->sk_txhash) {
1999 skb->hash = sk->sk_txhash;
2003 void skb_set_owner_w(struct sk_buff *skb, struct sock *sk);
2006 * Queue a received datagram if it will fit. Stream and sequenced
2007 * protocols can't normally use this as they need to fit buffers in
2008 * and play with them.
2010 * Inlined as it's very short and called for pretty much every
2011 * packet ever received.
2013 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
2017 skb->destructor = sock_rfree;
2018 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
2019 sk_mem_charge(sk, skb->truesize);
2022 void sk_reset_timer(struct sock *sk, struct timer_list *timer,
2023 unsigned long expires);
2025 void sk_stop_timer(struct sock *sk, struct timer_list *timer);
2027 int __sk_queue_drop_skb(struct sock *sk, struct sk_buff_head *sk_queue,
2028 struct sk_buff *skb, unsigned int flags,
2029 void (*destructor)(struct sock *sk,
2030 struct sk_buff *skb));
2031 int __sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
2032 int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
2034 int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
2035 struct sk_buff *sock_dequeue_err_skb(struct sock *sk);
2038 * Recover an error report and clear atomically
2041 static inline int sock_error(struct sock *sk)
2044 if (likely(!sk->sk_err))
2046 err = xchg(&sk->sk_err, 0);
2050 static inline unsigned long sock_wspace(struct sock *sk)
2054 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
2055 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
2063 * We use sk->sk_wq_raw, from contexts knowing this
2064 * pointer is not NULL and cannot disappear/change.
2066 static inline void sk_set_bit(int nr, struct sock *sk)
2068 if ((nr == SOCKWQ_ASYNC_NOSPACE || nr == SOCKWQ_ASYNC_WAITDATA) &&
2069 !sock_flag(sk, SOCK_FASYNC))
2072 set_bit(nr, &sk->sk_wq_raw->flags);
2075 static inline void sk_clear_bit(int nr, struct sock *sk)
2077 if ((nr == SOCKWQ_ASYNC_NOSPACE || nr == SOCKWQ_ASYNC_WAITDATA) &&
2078 !sock_flag(sk, SOCK_FASYNC))
2081 clear_bit(nr, &sk->sk_wq_raw->flags);
2084 static inline void sk_wake_async(const struct sock *sk, int how, int band)
2086 if (sock_flag(sk, SOCK_FASYNC)) {
2088 sock_wake_async(rcu_dereference(sk->sk_wq), how, band);
2093 /* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
2094 * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
2095 * Note: for send buffers, TCP works better if we can build two skbs at
2098 #define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
2100 #define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2)
2101 #define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE
2103 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
2105 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
2106 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
2107 sk->sk_sndbuf = max_t(u32, sk->sk_sndbuf, SOCK_MIN_SNDBUF);
2111 struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp,
2112 bool force_schedule);
2115 * sk_page_frag - return an appropriate page_frag
2118 * If socket allocation mode allows current thread to sleep, it means its
2119 * safe to use the per task page_frag instead of the per socket one.
2121 static inline struct page_frag *sk_page_frag(struct sock *sk)
2123 if (gfpflags_allow_blocking(sk->sk_allocation))
2124 return ¤t->task_frag;
2126 return &sk->sk_frag;
2129 bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag);
2132 * Default write policy as shown to user space via poll/select/SIGIO
2134 static inline bool sock_writeable(const struct sock *sk)
2136 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
2139 static inline gfp_t gfp_any(void)
2141 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
2144 static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
2146 return noblock ? 0 : sk->sk_rcvtimeo;
2149 static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
2151 return noblock ? 0 : sk->sk_sndtimeo;
2154 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2156 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
2159 /* Alas, with timeout socket operations are not restartable.
2160 * Compare this to poll().
2162 static inline int sock_intr_errno(long timeo)
2164 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2167 struct sock_skb_cb {
2171 /* Store sock_skb_cb at the end of skb->cb[] so protocol families
2172 * using skb->cb[] would keep using it directly and utilize its
2173 * alignement guarantee.
2175 #define SOCK_SKB_CB_OFFSET ((FIELD_SIZEOF(struct sk_buff, cb) - \
2176 sizeof(struct sock_skb_cb)))
2178 #define SOCK_SKB_CB(__skb) ((struct sock_skb_cb *)((__skb)->cb + \
2179 SOCK_SKB_CB_OFFSET))
2181 #define sock_skb_cb_check_size(size) \
2182 BUILD_BUG_ON((size) > SOCK_SKB_CB_OFFSET)
2185 sock_skb_set_dropcount(const struct sock *sk, struct sk_buff *skb)
2187 SOCK_SKB_CB(skb)->dropcount = sock_flag(sk, SOCK_RXQ_OVFL) ?
2188 atomic_read(&sk->sk_drops) : 0;
2191 static inline void sk_drops_add(struct sock *sk, const struct sk_buff *skb)
2193 int segs = max_t(u16, 1, skb_shinfo(skb)->gso_segs);
2195 atomic_add(segs, &sk->sk_drops);
2198 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2199 struct sk_buff *skb);
2200 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2201 struct sk_buff *skb);
2204 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2206 ktime_t kt = skb->tstamp;
2207 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
2210 * generate control messages if
2211 * - receive time stamping in software requested
2212 * - software time stamp available and wanted
2213 * - hardware time stamps available and wanted
2215 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
2216 (sk->sk_tsflags & SOF_TIMESTAMPING_RX_SOFTWARE) ||
2217 (kt && sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) ||
2218 (hwtstamps->hwtstamp &&
2219 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE)))
2220 __sock_recv_timestamp(msg, sk, skb);
2224 if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2225 __sock_recv_wifi_status(msg, sk, skb);
2228 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2229 struct sk_buff *skb);
2231 #define SK_DEFAULT_STAMP (-1L * NSEC_PER_SEC)
2232 static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2233 struct sk_buff *skb)
2235 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
2236 (1UL << SOCK_RCVTSTAMP))
2237 #define TSFLAGS_ANY (SOF_TIMESTAMPING_SOFTWARE | \
2238 SOF_TIMESTAMPING_RAW_HARDWARE)
2240 if (sk->sk_flags & FLAGS_TS_OR_DROPS || sk->sk_tsflags & TSFLAGS_ANY)
2241 __sock_recv_ts_and_drops(msg, sk, skb);
2242 else if (unlikely(sock_flag(sk, SOCK_TIMESTAMP)))
2243 sk->sk_stamp = skb->tstamp;
2244 else if (unlikely(sk->sk_stamp == SK_DEFAULT_STAMP))
2248 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags);
2251 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
2252 * @sk: socket sending this packet
2253 * @tsflags: timestamping flags to use
2254 * @tx_flags: completed with instructions for time stamping
2256 * Note : callers should take care of initial *tx_flags value (usually 0)
2258 static inline void sock_tx_timestamp(const struct sock *sk, __u16 tsflags,
2261 if (unlikely(tsflags))
2262 __sock_tx_timestamp(tsflags, tx_flags);
2263 if (unlikely(sock_flag(sk, SOCK_WIFI_STATUS)))
2264 *tx_flags |= SKBTX_WIFI_STATUS;
2268 * sk_eat_skb - Release a skb if it is no longer needed
2269 * @sk: socket to eat this skb from
2270 * @skb: socket buffer to eat
2272 * This routine must be called with interrupts disabled or with the socket
2273 * locked so that the sk_buff queue operation is ok.
2275 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb)
2277 __skb_unlink(skb, &sk->sk_receive_queue);
2282 struct net *sock_net(const struct sock *sk)
2284 return read_pnet(&sk->sk_net);
2288 void sock_net_set(struct sock *sk, struct net *net)
2290 write_pnet(&sk->sk_net, net);
2293 static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2296 struct sock *sk = skb->sk;
2298 skb->destructor = NULL;
2305 /* This helper checks if a socket is a full socket,
2306 * ie _not_ a timewait or request socket.
2308 static inline bool sk_fullsock(const struct sock *sk)
2310 return (1 << sk->sk_state) & ~(TCPF_TIME_WAIT | TCPF_NEW_SYN_RECV);
2313 /* This helper checks if a socket is a LISTEN or NEW_SYN_RECV
2314 * SYNACK messages can be attached to either ones (depending on SYNCOOKIE)
2316 static inline bool sk_listener(const struct sock *sk)
2318 return (1 << sk->sk_state) & (TCPF_LISTEN | TCPF_NEW_SYN_RECV);
2322 * sk_state_load - read sk->sk_state for lockless contexts
2323 * @sk: socket pointer
2325 * Paired with sk_state_store(). Used in places we do not hold socket lock :
2326 * tcp_diag_get_info(), tcp_get_info(), tcp_poll(), get_tcp4_sock() ...
2328 static inline int sk_state_load(const struct sock *sk)
2330 return smp_load_acquire(&sk->sk_state);
2334 * sk_state_store - update sk->sk_state
2335 * @sk: socket pointer
2336 * @newstate: new state
2338 * Paired with sk_state_load(). Should be used in contexts where
2339 * state change might impact lockless readers.
2341 static inline void sk_state_store(struct sock *sk, int newstate)
2343 smp_store_release(&sk->sk_state, newstate);
2346 void sock_enable_timestamp(struct sock *sk, int flag);
2347 int sock_get_timestamp(struct sock *, struct timeval __user *);
2348 int sock_get_timestampns(struct sock *, struct timespec __user *);
2349 int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len, int level,
2352 bool sk_ns_capable(const struct sock *sk,
2353 struct user_namespace *user_ns, int cap);
2354 bool sk_capable(const struct sock *sk, int cap);
2355 bool sk_net_capable(const struct sock *sk, int cap);
2357 void sk_get_meminfo(const struct sock *sk, u32 *meminfo);
2359 extern __u32 sysctl_wmem_max;
2360 extern __u32 sysctl_rmem_max;
2362 extern int sysctl_tstamp_allow_data;
2363 extern int sysctl_optmem_max;
2365 extern __u32 sysctl_wmem_default;
2366 extern __u32 sysctl_rmem_default;
2368 #endif /* _SOCK_H */